Researching Criminology

Researching Criminology
R e s e a rc h i n g C r i m i n o l o g y
“research
This book provides an essential tool for undergraduate students embarking upon their own
projects in Criminology. It provides clear and informative guidance on a range of
research methods and designs to assist students in their own criminological endeavours.”
D J
R T
ACKI ,
APLEY SENIOR LECTURER IN CRIMINOLOGY, INSTITUTE OF CRIMINAL JUSTICE STUDIES, UNIVERSITY OF PORTSMOUTH
“institutions
…what makes the book stand out is the inclusion of real research into various criminal justice
that have actually been undertaken by the authors. In doing so, what is produced is a
book that offers research 'know how', stimulates interest and interjects research passion into
what can often be a difficult, and sometimes dry area of research. ”
DR TINA PATEL, LECTURER IN CRIMINOLOGY AND SOCIOLOGY, LIVERPOOL JOHN MOORES UNIVERSITY
❙ How do criminologists go about studying crime and its consequences?

❙ How are programmes for offenders and communities evaluated?
❙ How can you collect and analyse criminological material?
Research on crime and criminality is often referred to by the media, policy makers and
practitioners, but where does this research come from and how reliable is it?
Designed especially for students on criminology and criminal justice courses, and professionals
working in the field, Researching Criminology emphasises the importance of research as an
integrated process. It looks at the ways in which a mixture of investigative methods can be used
to analyze a criminological question.
Written by two experienced researchers and lecturers, Researching Criminology is a

comprehensive introduction to the aims, principles and methods of doing criminological
research. The book covers all the key topics that you will encounter when researching crime.
Individual chapters include:
❙ The research process ❙ Evaluation research

❙ Principles of researching criminology
❙ How to design criminological research
❙ Researching ethically
❙ A glossary of essential key concepts
Researching
Crow and Semmens
Structured in three parts, addressing the principles of criminological research, how to collect and
analyse material and providing detailed examples of real world research, Researching
Criminology
Criminology will be of benefit to all students of criminology and criminal justice, for practitioners
interested in criminological research, and for those undertaking criminological research for the
first time.
Iain Crow is Reader in Research Methods at the University of Sheffield, UK. His teaching covers
research methods and statistics, and research ethics. He has undertaken research on a wide
range of topics concerned with the treatment and rehabilitation of offenders and crime reduction.
Natasha Semmens is a lecturer in Criminology at the University of Sheffield, UK. Her research
interests include the fear of crime, crime survey methodology, white collar crime and cybercrime.
Her teaching includes research methods, quantitative
analysis, and white collar crime.
Cover design Hybert Design • www.hybertdesign.com

www.openup.co.uk
Iain Crow and Natasha Semmens
JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 7 OUTPUT: Wed Aug 29 11:17:58 2007 SUM: 0C70A77B
/production/mcgraw−hill/booksxml/crowsemmens−11pt/aplm_01−halftitle
RESEARCHING
CRIMINOLOGY
Kerrypress Ltd – Typeset in XML A Division: aplm_01-halftitle F Sequential 1

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 6 OUTPUT: Wed Aug 29 11:17:58 2007 SUM: 0123905A
/production/mcgraw−hill/booksxml/crowsemmens−11pt/aplm_01−halftitle
Kerrypress Ltd – Typeset in XML A Division: aplm_01-halftitle F Sequential 2

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 14 OUTPUT: Wed Aug 29 11:18:12 2007 SUM: 10FB06BB
/production/mcgraw−hill/booksxml/crowsemmens−11pt/aplm_02−title
RESEARCHING
CRIMINOLOGY
Iain Crow and

Natasha Semmens
Open University Press
Kerrypress Ltd – Typeset in XML A Division: aplm_02-title F Sequential 1

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 20 OUTPUT: Wed Aug 29 11:28:00 2007 SUM: 2591887F
/production/mcgraw−hill/booksxml/crowsemmens−11pt/aplm_03−imprint
Open University Press

McGraw-Hill Education
McGraw-Hill House
Shoppenhangers Road
Maidenhead
Berkshire
England
SL6 2QL
email: [email protected]
world wide web: www.openup.co.uk
and Two Penn Plaza, New York, NY 10121—2289, USA
First published 2006
Copyright © Author’s name 2006
All rights reserved. Except for the quotation of short passages for the purpose of
criticism and review, no part of this publication may reproduced, stored in a
retrieval system, or transmitted, in any form or by any means, electronic,
mechanical, photocopying, recording or otherwise, without the prior written
permission of the publisher or a licence from the Copyright Licensing Agency
Limited. Details of such licences (for reprographic reproduction) may be obtained
from the Copyright Licensing Agency Ltd of Saffron House,
6–10 Kirby Street, London, EC1N 8TS.
A catalogue record of this book is available from the British Library
ISBN-13: 978-0-33-521546-1 (pb) 978-0-33-521547-8 (hb)

ISBN-10: 0-33-522117-3 (pb) 0-33-522118-1 (hb)
Library of Congress Cataloging-in-Publication Data

CIP data applied for
Typeset by Kerrypress, Luton, Bedfordshire

Printed in XXXXXXXXXXXXXXXXXXXXXXXXXXXX
Kerrypress Ltd – Typeset in XML A Division: aplm_03-imprint F Sequential 1
www.kerrypress.co.uk - 01582 451331 - www.xpp-web-services.co.uk

McGraw Hill - 170mm x 240mm - Fonts: Caslon, Gill Sans
JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 12 OUTPUT: Wed Aug 29 11:22:41 2007 SUM: 2F75B9C8
/production/mcgraw−hill/booksxml/crowsemmens−11pt/aplm_04−contents
Contents
About the authors
Acknowledgements
Introduction
Part I: The principles of criminological research

1 The research process
What is research?
Researching criminology
Types of inquiry and types of data
Research as a process
Further reading
2 The principles of researching criminology
The research question: identifying the research question and
exploring possible theoretical explanations
Choosing a research strategy
Method selection and data collection
Data analysis and inference
Conclusion
Further reading
3 Designing criminological research
What is research design and why is it important?
The main types of empirical inquiry
Some common kinds of research design
Choosing a research design: an example
Sampling
Researching criminology ethically
Conclusion
Further reading
Kerrypress Ltd – Typeset in XML A Division: aplm_03-contents F Sequential 1

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 12 OUTPUT: Wed Aug 29 11:22:41 2007 SUM: 39200E0E
vi Contents
4 Criminological evaluation
Defining terms
The evaluation paradigm
Evaluation and theory
Doing criminological evaluation
Further reading
Part II: Collecting and analysing material

5 Researching by reading
Introduction
Reading and reviewing the literature critically
Analysing other people’s data
Analysing documents
Conclusion
Further reading
Appendix
6 Researching by looking
Introduction: watch and learn
The science of observation? Epistemology and research design
Methodological decisions
Data collection
Data analysis
The disadvantages of observation
Conclusion
Further Reading
7 Researching by asking and listening
Introduction
Interviewing
Group interviews and focus groups
Questionnaires
Further reading
8 Analysing criminological research
About analysis
Analysing quantitative material: an introduction
Analysing qualitative material: an introduction
Further reading
Appendix 1
Appendix 2
Part III: Real-world research

9 Researching offenders and employment
Background to the project
Theoretical context

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 12 OUTPUT: Wed Aug 29 11:22:41 2007 SUM: 211AC417
Contents vii
The research
Results
Conclusion
10 Researching the Youth Court
Theoretical context
The research
Analysis
Results
Conclusion
Appendix
11 Researching a Community Safety Programme
Theoretical context
The research
Results
Conclusion
12 Researching the fear of crime
Theoretical context
The research
Results
Conclusion
13 Concluding comments
Taking it further
Glossary
Notes
References
Index


JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 13 OUTPUT: Wed Aug 29 11:22:54 2007 SUM: 340DC8EA
/production/mcgraw−hill/booksxml/crowsemmens−11pt/aplm_05−aboutauthors
About the authors
Iain Crow is a reader in Research Methods at the University of

Sheffield. He has previously worked at the Addiction Research Unit at
the Institute of Psychiatry and was Head of Research at NACRO, the
national organization for offender rehabilitation and crime prevention,
until 1989. His teaching covers research methods and statistics,
research ethics, and the treatment and rehabilitation of offenders. He
has done research on drug use, race and criminal justice, offenders and
unemployment, the Youth Court, community safety. He has also been
involved with community-based programmes for offenders, and done
research on such programmes. His other books include The Treatment
and Rehabilitation of Offenders, Criminal Justice 2000: Strategies for a
New Century, and Unemployment, Crime and Offenders.
Natasha Semmens has been a Lecturer in Criminology at the Univer-
sity of Sheffield since 2001. Her teaching covers research methods,
white-collar crime and foundations in law and legal systems. Her
research interests include the fear of crime, crime survey methodology,
white-collar crime and cybercrime. She has also been involved in
projects with the Home Office, NACRO and the UK Information
Commissioner.
Kerrypress Ltd – Typeset in XML A Division: aplm_04-aboutauthors F Sequential 1

/production/mcgraw−hill/booksxml/crowsemmens−11pt/aplm_05−aboutauthors
Kerrypress Ltd – Typeset in XML A Division: aplm_04-aboutauthors F Sequential 2

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 10 OUTPUT: Wed Aug 29 11:23:07 2007 SUM: 40F8377A
/production/mcgraw−hill/booksxml/crowsemmens−11pt/aplm_06−acknowls
Acknowledgements
Inevitably such a book reflects our own approach to criminological

research, but it also brings together material that reflects the teaching
and research done by fellow criminologists in the School of Law at the
University of Sheffield. It is therefore right and proper to acknowledge
the fact that what is written here owes much to the work we have done
with others. We should also say that teaching and learning go together,
and that we have learned a lot about teaching research methods in
criminology from the generations of students who have taken the
modules we have taught, so we also wish to thank them.
We have used actual criminological examples to illustrate the points
we are trying to convey as much as possible, and this means that we
have drawn on the work of criminologists from many quarters.
However, Part III of this book describes projects that we ourselves
have done over the years. Most criminological research is a collabora-
tive effort involving a team of people. We therefore owe a great debt of
gratitude to those we have worked with on these projects, both for
their contributions to the research we did together, but also for what
we have learned by working with them.
Chapter 9, on the relationship between employment status and
sentencing, is based on a study funded by the Economic and Social
Research Council which was done in collaboration with Frances
Simon. Frances was a statistician who had previously worked with the
Home Office and her contribution to this particular study and to
criminology generally has been immense, and deserving of the widest
possible recognition.
Chapter 10 describes a project on the Youth Court funded by the
Home Office, undertaken in collaboration with Charlotte Allen and
Michael Cavadino. Charlotte subsequently went on to work in the
Home Office Research, Development and Statistics Directorate, and
Mick has since gone to the University of Central Lancashire.
Kerrypress Ltd – Typeset in XML A Division: aplm_05-acknowls F Sequential 1

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 10 OUTPUT: Wed Aug 29 11:23:07 2007 SUM: 35D36070
/production/mcgraw−hill/booksxml/crowsemmens−11pt/aplm_06−acknowls
xii Acknowledgements
Chapter 11 explains how a Joseph Rowntree Foundation-funded

evaluation of three community safety projects was done. In the early
days of this project, Paul Wiles played a key role before he went on to
head the Home Office Research, Development and Statistics Directo-
rate. Alan France was an important formative influence, and after Paul
left became co-director of the project. He has since become Director of
the Centre for Research and Social Policy. Sue Hacking and Mary
Hart were other major contributors to the research, Sue on the
quantitative side, and Mary working mainly on the qualitative mate-
rial.
Chapter 12 describes a project which involved a collaboration with
the British Crime Survey team at the Home Office. Pat Mayhew and
Catriona Mirlees-Black were instrumental in setting up the collabora-
tion back in 1999. Tracy Budd and Joanna Taylor (then Mattinson)
provided valuable support and advice during the analysis stages. In the
later stages of the project, Chris Kershaw was a source of continuing
support. All have now moved on to work in other sections of the
Home Office. This research was conducted for a PhD thesis and many
academic colleagues offered priceless guidance and inspiration, includ-
ing Jason Ditton, Steve Farrall and Ken Pease.
Finally, we would like to thank Chris Cudmore of McGraw-Hill for
his help and encouragement, and for saving us from embarassment by
looking over an early draft of the text.
Although, as we have explained, this book draws on the work of
others, and brings together what we have learned from working with
others, we lay no blame at their doors for what is written here. We take
full responsibility for what is in the book, and they have every right to
criticize us for anything we have got wrong.

University of Sheffield
Kerrypress Ltd – Typeset in XML A Division: aplm_05-acknowls F Sequential 2

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 24 OUTPUT: Wed Aug 29 11:23:20 2007 SUM: 475E334A
/production/mcgraw−hill/booksxml/crowsemmens−11pt/aplm_07−introduction
Introduction
Almost the first question we asked ourselves before embarking on

writing this book was why we wanted to do it. Surely there are
plenty of research methods texts around already? Well, one answer is
that although there are indeed a number of texts covering social
research methods in general, there are not that many specifically
oriented towards criminology and the needs of criminology students.
Those that are around tend to be American, quite old, or collections
of chapters by different authors rather than integrated texts.
The book has come about to a large extent as a result of our own
teaching, and reflects our research experiences. In teaching under-
graduate criminology students, and postgraduate students who had
not encountered social research techniques previously or done any
criminology, we felt there was a need for a textbook that would
introduce them to the principles involved, the various methods used,
and illustrate how these work in real criminological studies. In our
work we have also encountered many practitioners, policy-makers,
journalists and other commentators who have an interest in crime
and criminal justice. We hope that the book will also be of value to
them.
The book is in three parts, covering the main principles of
criminological research, the collection and analysis of material, and
examples of real criminological research projects that we have
undertaken.
The principles that underpin criminological research are similar in
many respects to those that form the basis for other forms of social
inquiry. However, criminological investigation is often shaped by the
fact that it deals with matters that are by their nature illegal and also
often immoral, so criminological inquiry is often situated in a
context that requires this to be taken into account. In Part I we start
by introducing the research process. It would be misleading to think
of research as simply an agglomeration of particular techniques and
methods, and we emphasize the importance of seeing research as an
integrated process. The book focuses mainly on the later stages of
Kerrypress Ltd – Typeset in XML A Division: aplm_06-introduction F Sequential 1

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 24 OUTPUT: Wed Aug 29 11:23:20 2007 SUM: 4C3B0AFC
2 Researching criminology
the process, and many of the topics mentioned in the opening

chapter are considered in more detail subsequently. In Chapter 2, we
consider the relationship between epistemology, theory and methods
of inquiry. In Chapter 3, we look at how research is designed, and at
the important matter of researching criminology in a way that is
ethical. Chapter 4 focuses on a particular form of research design
that has considerable importance in criminology: that which is
directed towards evaluating the impact of various interventions
which involve individuals, communities, or indeed national changes,
such as new legislation.
Part II of the book, Chapters 5–8, covers the collection of
research material. Most research comes down to looking at what
happens (observing), reading, asking questions, and listening to what
is said, and the chapters that comprise Part II reflect this. We also
consider the analysis of research material here. All books have their
limitations, and in this one we do not go into detail about how to
analyse research data. An adequate consideration of how to analyse
quantitative and qualitative data would require at least another one,
if not two, books, and reference can be made to other sources for
this purpose. However, having said already that research has to be
seen as an integrated process, and that consideration must be given
to how research material is to be analysed, we would be remiss if we
did not give some attention to the kind of things that analysis
involves, so that you can prepare adequately.
Reading books on research methods is all very well; they explain
what should happen in theory, in ideal circumstances. But, as we
ourselves have found, when you set out to do a criminological
project, you often have to work it out for yourself. While it is
necessary to understand what criminological research is about, how
to make use of interviews and secondary sources, and what to do
with the results, it is important to have a sense of how these things
need to come together. Furthermore, it is one thing to know about
such matters in the abstract; it is quite another to know how they
work in reality. Throughout the earlier parts of the book we have
tried to use examples wherever possible to illustrate various points,
but this is not the same as understanding how a whole criminology
project works. So Part III of the book, Chapters 9–12, describes
some of the research that we have done. In doing this, we explain
what the projects demonstrate, why we did things in a particular
way, and what alternatives might have been used. The purpose of
these chapters is to show how the principles described earlier are put
into practice. They are four very different projects undertaken at
different times. We explain the basis for each study, its design, the
methods used, how the analysis was conducted, and how the results
were interpreted and used.

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 24 OUTPUT: Wed Aug 29 11:23:20 2007 SUM: 3BFDEDBF
Introduction 3
The first study was conducted at a time of high unemployment,

and examines the relationship between offenders’ employment status
and the sentence they received. Although carried out some years
ago, the methodology is still relevant and could be applied to other
similar inquiries. The second project is a study of the way that
changes in the Youth Court were implemented. Although it illus-
trates how certain methods were used, it is also of interest because
of the sensitive local and national political issues involved. The third
study is one of those projects very common in criminology: an
evaluation. However, it is an evaluation that was both problematic,
and was at the centre of some far-reaching issues. The fourth
project, covering fear of crime, describes how the understanding of
that topic was extended by working alongside a major national
survey to look at credit card fraud. In none of these examples are we
saying that this is the only way of doing the research described.
Criminological research is often a choice between options, and we
explain why the studies happened the way they did.
Criminological research is usually a team effort, and that was very
much the case with the projects described here. We did these
projects in conjunction with fellow researchers, so we cannot take
full credit for them, and duly acknowledge the part played by others.
The full results of the studies can be found elsewhere and it is not
the purpose of the final part of the book to reproduce the substance
of the studies in all particulars. The aim is to illustrate the various
methodological points that they raise, and in order to do this we
make reference to some of the substantive material where necessary
in order to make the methodology intelligible.
A Glossary of key terms is found at the back of the book.
Our hope is that this book will serve the dual purpose of helping
readers to critically appraise criminological research that they
encounter, and also perhaps to go on to experience the satisfaction
that can be derived from researching criminology.

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 24 OUTPUT: Wed Aug 29 11:23:20 2007 SUM: 03F26973

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 32 OUTPUT: Tue Aug 28 15:02:37 2007 SUM: 127BF3C0
/production/mcgraw−hill/booksxml/crowsemmens−11pt/chap01
Part I
The principles of criminological
research
Kerrypress Ltd – Typeset in XML A Division: chap01 F Sequential 1

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 32 OUTPUT: Tue Aug 28 15:02:37 2007 SUM: 0397B57F

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 32 OUTPUT: Tue Aug 28 15:02:37 2007 SUM: 48C81ECC

In this opening chapter we will be looking at what researching
criminology involves as an integrated process. It is an overview of
what is to come in the rest of the book and we will be raising a
series of issues that will be covered in greater detail in subsequent
chapters. The point of this first chapter is to explain how different
elements of the research process hang together. Doing criminological
research is not just a matter of using certain methods of inquiry; it
is about the way such methods are used as part of a more
wide-ranging enterprise that constitutes the scientific method.
What is research?
The term ‘research’ is used to refer to a wide range of activities.

Sometimes it means nothing more than looking up information in
established sources. So is there anything special about the kind of
activity that takes place in academic institutions and certain research
centres? How does academic research differ from, say, good journal-
ism? Is it different in kind, or only in degree? Should we distinguish
between academic researchers and researchers who work for non-
academic organizations, such as market researchers?
There are no recognized standards or regulations that determine
that one person is a bona fide researcher while another is not. We
are not therefore going to attempt a prescriptive or proscriptive
definition of research. But we are going to suggest that the kind of
research engaged in by those who aspire to a genuine pursuit of
knowledge has certain features. First, research is likely to be in some
degree original; it is not merely looking up some information that is
already known and reproducing that information. This does not
necessarily mean that research requires the collection of new data at
first hand. It may be a new analysis of existing data, or the
re-interpretation of extant material; it may be a new synthesis.
Second, the kind of research that we are referring to will usually
be theoretically informed: it attempts to advance a corpus of ideas and
existing knowledge. It is not coming from nowhere and going to
nowhere. From this it follows that research will also be systematic in
nature. This does not mean that it has to follow a set of rules
rigidly, but it will approximate to something called the scientific

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 32 OUTPUT: Tue Aug 28 15:02:37 2007 SUM: 53D1FB66
method. This requires a brief explanation because there is a tendency

among some social researchers to equate the scientific method solely
with a positivist notion of research based on natural science.1 The
scientific method is broader than this and applies in many disci-
plines, not just the natural sciences. It is more a way of approaching
the process of inquiry. In particular, it is characterized by an
interaction between theory and method. Another central concern of
research is the extent to which any conclusions are generalizable.
Here again we encounter the importance of a link between the
methods used and the theoretical context. The kind of research that
we are concerned with should also follow certain ethical principles.
This involves having regard to such matters as informed consent,
and ensuring that the rights of parties to the research are not
violated. These are all matters to which we will return during the
course of this book.
Research is about addressing questions, usually phrased in terms
of hypotheses. We refer to theory for why we asked the questions,
and to interpret the results. Research may take different forms. A
distinction is commonly made, for example, between pure and
applied research, and there are different types of applied research,
such as evaluation research and action research. This book focuses
on aspects of empirical investigation, based on the observation of
social phenomena, rather than on purely abstract or theoretical
inquiry, but this does not mean that such research is atheoretical.
Researching criminology
This book is about researching criminology, so the question arises as

to whether there is anything particularly distinctive about crimino-
logical research. Criminological research has much in common with
social research in general and the principles that apply in other areas
of social science are also relevant to researching criminology.
Consequently, much of this book could be read as being about social
research methodology. However, researching criminology does have
certain distinctive features. For a start, doing criminological research
usually involves the study of things that are illegal. The criminologi-
cal researcher is likely to learn about illegal acts and meet people
who have committed such acts. Doing research on drug misuse, for
example, may well mean being in the company of people in
possession of illegal drugs. This can place researchers in difficult
situations. Apart from the illegality of much of the subject matter, a
criminological researcher is also quite possibly going to be dealing
with sensitive issues, and with people who have had traumatic
experiences. In some situations the research can have an emotional
impact on the researcher, for example, undertaking research on

JOBNAME: Mcgraw−crowsemmens PAGE: 5 SESS: 32 OUTPUT: Tue Aug 28 15:02:37 2007 SUM: 4E05F18A
The research process 9
sexual abuse is likely to be a harrowing experience. So the subject

matter of criminology is distinctive and different from many other
disciplines. This makes the observance of ethical principles especially
important in criminology. It also means that the criminological
researcher needs to develop a range of interpersonal skills. This is
something often neglected in research methods textbooks. The
emphasis is usually, and understandably, on theoretical and methodo-
logical matters; that is on what is colloquially regarded as ‘academic’
skills. But successful criminological research also means that the
researcher has to be able to relate to people and organizations, to
establish ‘rapport’, and sometimes empathy, while also retaining
integrity as a researcher, and that is an important theme in this
book.
The use of the term ‘discipline’ above is also important. Many
would say that criminology is not a discipline in its own right. It is
certainly the case that criminology draws on a number of different
academic disciplines, including sociology, psychology, economics,
geography, and law, to name just some. This disciplinary diversity
means that the methods used in criminological research are likely to
reflect the kinds of techniques used in the parent disciplines. Several
of these parent disciplines use social science research methods, and
will therefore have similar features, but it can lead to a rich diversity
of approaches. Imagine, for example that you were undertaking a
study of sentencing in the courts. A psychologist might use
experimental techniques to present magistrates and judges with a
series of cases in which the features of the cases are varied in such
a way as to explore what is likely to influence sentencers most. A
sociologist might spend time observing what happens in court, or
looking at the backgrounds of sentencers and offenders. An econo-
mist might come up with models of sentencing that take account of
various ‘costs’, a term which goes wider than just financial costs.
Other kinds of study might look at how crime is distributed
geographically. A feature of criminological research is that not only
does it cover an important topic, but also has this richly diverse
character.
Types of inquiry and types of data
Although much is written about research and ways of doing it, most
empirical social inquiry involves three main kinds of data:
1 Observation – looking at things, people and events.

2 Words – verbal or textual communication in its various forms.
3 Numbers – counting things, people and events.

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 32 OUTPUT: Tue Aug 28 15:02:37 2007 SUM: 49C6F534
This forms the basis of Part II of the book, and we will therefore be
considering all three of these activities in more detail there. First,
however, it is worth referring to some important distinctions that are
often found in criminology and in social research generally.
A distinction is often made between quantitative data, involving
the use of numbers, and qualitative data, which usually means the
use of words or images. The former is sometimes referred to as
‘hard’ data, and the latter as ‘soft’ data. Much social research is
quantitative in nature, involving statistical analysis. Examples of such
research include study of the criminal statistics and opinion polls,
and major social surveys, such as the British Household Panel
Survey, and the British Crime Survey. Quantitative research is
sometimes, wrongly, seen as synonymous with a positivist approach
to inquiry. Positivism, as we explain in a later chapter, is a more
broadly based philosophy of science which involves using methods
derived from the natural sciences, which does include numerical
measurement, but this is far from being the whole story. On the
other hand, methods which involve qualitative inquiry, such as
ethnography, participant observation and discourse analysis, are often
associated with a different perspective, most commonly referred to as
interpretive inquiry, but encompassing such things as hermeneutics,
ethnomethodology, phenomenology and structuralism. Such
approaches tend to concentrate on understanding the meaning of
social interaction.
The relationship between these different research perspectives
might be viewed diagrammatically (Figure 1.1). The main point we
want to make is that this simplistic distinction between positivist
and interpretive inquiry is a crude, and even misleading one. Even
those who are experts in statistical analysis are aware of its
limitations, and would probably reject the label positivist, and those
concerned with subcultural theory and ethnography also recognize
the need to sample and count.
Returning to our identification of the three main components of
research as observations, words and numbers, what we want to
emphasize is that there is considerable interchange between them.
Different research perspectives often have more in common with
each other than may at first be apparent, each needing to have
regard to such matters as the unit of observation, sampling and
Figure 1.1 Relationship between quantitative and

qualitative data

JOBNAME: Mcgraw−crowsemmens PAGE: 7 SESS: 32 OUTPUT: Tue Aug 28 15:02:37 2007 SUM: 5346D154
problems of reliability and validity. For example, what people say or

write may be converted into numbers by counting the number of
times they say or do something. Conversely, numbers need to be
examined for the meanings they contain. Although it is possible to
associate quantitative and qualitative data with different epistemo-
logical positions, few modern researchers dogmatically assert that
one type of data is inherently superior to the other; each is a way of
trying to enhance our understanding, with advantages and disadvan-
tages depending on the context in which they are used.
Most modern researchers in criminology and related social sci-
ences are fairly eclectic about their research methods, and recognize
the limitations of their data, however obtained. Indeed, many
researchers would advocate the use of combined methods to improve
the validity of their research. This has been the case for some time.
Almost 20 years ago, Denzin (1988) described the use of ‘triangu-
lation’ in research. He identified four basic kinds of triangulation.
First was the triangulation of different types of data, which could
involve time, space and people. The second was investigator trian-
gulation, involving multiple observers of the same object. Next came
theoretical triangulation, involving different perspectives in relation
to the same set of objects. Finally, methodological triangulation was
defined as a distinction between triangulation within a specific
method, and triangulation combining dissimilar methods to examine
the same thing. An example of within-methods triangulation might
be to use more than one attitude scale within a study to look at
alienation. A between-methods triangulation, on the other hand,
might combine an attitude scale to study alienation with an
observational study of behaviour.2 To give a criminological example,
imagine you want to know how magistrates sentence different types
of offender. You could ask them how they would deal with various
cases, while at the same time asking them questions about each case
(within-methods interviewing). You could also interview magistrates,
observe how cases are dealt with in court, and analyse a sample of
cases using the court records (between-methods).
In recent years there has been a resurgence of interest in what is
now called ‘mixed methods’ research, with recent journal articles and
texts embracing a multi-method approach.3 The studies described in
the final part of this book all employ what Green and Preston
(2005) refer to as ‘methodological hybridity’. The use of numbers
and measurement, as well as words, has always been an important
component of both natural and social scientific research, and they
are often used in combination. More problematic is what happens
when sophisticated research techniques get into the hands of those
who prefer research to provide simplistic, and preferably cheap,
solutions.

JOBNAME: Mcgraw−crowsemmens PAGE: 8 SESS: 32 OUTPUT: Tue Aug 28 15:02:37 2007 SUM: 3402FDAC
Figure 1.2 The research process

Research as a process
It follows from what has been said already that research is not
simply the application of a particular method, such as a question-
naire. It is an integrated process involving theories, hypotheses,
designs, methods and analyses. Only by having regard to the whole
process is it possible to determine when it is appropriate to use
particular methods, and to draw appropriate inferences from them.
This process may be represented diagrammatically as shown in
Figure 1.2.
The first thing to say about this model is that it is a simplified
one, and it is something of an ideal. Research does not always move
definably from one stage in the process to the next. It is not
uncommon for the process to work in the other direction as well,
moving from observation, through the elaboration of hypotheses to
theoretical propositions. This is the approach taken by those who
develop grounded theory (Glaser and Strauss 1967; Strauss and
Corbin 1998). Nonetheless, the model is one that is broadly
recognized, even by researchers who go about inquiry in very
different ways.4 For example, at least one researcher adopts the
model in writing about how to undertake literature reviews (Cooper
1989). Let us now look at the various elements of this process.
Theory
A theory is a set of logically inter-related propositions. The Concise
Oxford Dictionary of Sociology defines theory as ‘an account of the

JOBNAME: Mcgraw−crowsemmens PAGE: 9 SESS: 32 OUTPUT: Tue Aug 28 15:02:37 2007 SUM: 4EFDDC0C
world which goes beyond what we can see and measure’ (Marshall
1994). The nature of social theory, and the role of theory in
research, will be considered further in the next chapter. However,
what needs to be noted here is that empirical criminological inquiry
is seldom if ever theory-free, and it is one of the basic principles of
this book that the theory underlying a criminological research
project, and the methods used in undertaking the research are
closely related.
Like all social research, criminological research usually starts with
the posing of a question or the addressing of a problem. Research
may come about in a number of ways. It may be a response to a
specific problem, such as an apparent rise in a particular type of
crime, or as a result of new criminal justice legislation. It may be
that after reading an article, or reviewing the literature on a topic, a
researcher believes there is a gap in the knowledge, or that previous
research is wrong, or out of date. Not infrequently research is
prompted by a perceived wrong or injustice, such as the belief that
a particular group is being discriminated against. In all these
instances it can seem that the theoretical basis for a particular
investigation is self-evident, and it is all too easy for the theory that
underlies a research project to be unexamined. Whatever prompted
the research, the theoretical basis of inquiry should be critically
analysed. A competent researcher considers the theoretical context of
an inquiry, even if s/he is not undertaking research directly associ-
ated with the investigation of a particular theory.
Having said this, not all research is about the examination of
‘grand’ theory. Much research is usefully conducted at lower levels of
theoretical analysis. Research may be exploratory in nature, con-
cerned with the development and formulation of concepts and
propositions, rather than putting a theory to the test. But even
exploratory research should have a clearly defined focus (Glaser and
Strauss 1967). The point is that some theoretical element is
invariably present in empirical investigation, and the more clearly
this is set out, the more valid and valuable the research will be.
Hypotheses
Theoretical propositions are often of a rather general, abstract,
nature. For example, the proposition may be that certain groups in
society are more alienated than others and therefore more predis-
posed towards committing crimes. This may be related to certain
forms of social theory concerned with class and conflict, and clearly
there is much here that needs examination, definition and clarifica-
tion before the proposition and the theory on which it is based can
be tested empirically. Hypotheses are empirically testable statements,
usually involving relationships between concepts which form part of

JOBNAME: Mcgraw−crowsemmens PAGE: 10 SESS: 32 OUTPUT: Tue Aug 28 15:02:37 2007 SUM: 4E45C568
a theory. Examining a hypothesis involves applying clearly defined

criteria which enable the hypothesis to be refuted or not. Tradition-
ally hypotheses are not accepted, since there is usually the possibility
that some other basis for rejecting the hypothesis could be found.
Hence hypotheses, and by extension our empirical understanding,
are frequently of a provisional nature. There are few laws in the
social sciences in the same way as in the natural sciences.
To give an example, it has been theorized that because of the
nature of our society and the role of the criminal justice system
within it, the unemployed are more likely to be punished by
imprisonment than those with jobs (Box 1987). This theoretical
proposition has been tested (Crow and Simon 1987). A fuller
description of how this was done is given in Chapter 9, but for the
time being what needs to be noted is that in the course of this
inquiry the investigators framed a number of hypotheses about what
the empirical study would find. As often happens in an empirical
inquiry, the investigators found that the relationships were more
complex than theories sometimes suggest. Hence what you tend to
get is a modification and enhancement of the hypotheses, laying the
basis for further inquiry. This is why research so often ends with the
conclusion that ‘more research is needed’. It is not necessarily
because researchers are touting for their next research grant; it
should be because research is a continuous process of refinement and
discovery, reflecting a constant exchange between theory and data.
For this reason hypotheses are pivotal in research, making it
possible to relate data and observations to theoretical propositions.
This is true of both quantitative and qualitative research. Kelle
describes the process in qualitative research:
Qualitative hypotheses, when they first come into a researcher’s
mind, are usually not highly specified and definite propositions
about certain facts, but tentative and imprecise, sometimes very
vague conjectures about possible relationships. (1997: 13)
Kelle goes on to spell out the methodological requirements for

hypothesis testing, which apply as much to quantitative as to
qualitative data:
1 Hypotheses must be empirically testable; that means they must be
precise enough and have empirical content.
2 The codes which are used for hypothesis testing must denote
clearly defined phenomena in a reliable and stable way.
3 The prerequisite of independent testing requires that a hypothesis
is not tested with the same empirical material from which it is
developed.
In statistics, the term hypothesis has a very specific meaning and the
criteria for rejecting or not rejecting a hypothesis are mathematically

JOBNAME: Mcgraw−crowsemmens PAGE: 11 SESS: 32 OUTPUT: Tue Aug 28 15:02:37 2007 SUM: 4C3C6D86
determined. Usually what happens is that, in looking at relationships

between variables, one starts with a ‘null hypothesis’ that there is no
relationship, or no difference, and a calculation is then performed to
see whether this null hypothesis can be rejected on the basis that it
is unlikely to occur by chance at a stated level of probability. In
order to test a hypothesis, either quantitatively or qualitatively, it is
necessary to state precisely what is meant by what initially may be
some rather abstract concepts (e.g. alienation), and rough ideas about
the relationships between them; this is a process rather inelegantly
known as operationalization.
Operationalization
This is a crucial part of the research process. It includes the
following:
1 The definition of any concepts, and the basis for examining them
have to be spelled out. In other words, it is necessary to say just
what is meant by a term such as ‘alienation’, and how you are able
to tell whether someone is more or less alienated.5 If the
hypothesis is that ‘People without family commitments are
inclined to be more radical than those with such commitments’,
then what is meant by ‘family commitments’ has to be specified,
and we have to say what we mean by ‘more radical’. Only when
this is done will it be possible to construct a questionnaire that
can be used on a sample to see whether (at a certain level of
statistical probability) those without family commitments satisfy
the criterion of ‘radicalness’ more than those who have family
commitments.
2 Operationalizing a project involves developing a research design. In
the example above, this may involve obtaining two groups of
people who are comparable in certain respects, except that one
group has family commitments and the other does not. More will
be said about this in Chapter 3.
3 The advantages and disadvantages of alternative methods of
inquiry have to be considered – whether to use a questionnaire,
in-depth interviews, observation, or some other method. The
methods judged to be most appropriate for testing a hypothesis
are chosen and justified. In doing this, an important consideration
will be the resources available to carry out the research; research is
often a compromise between what would be ideal and what is
practical.
4 Any scales, scores or other measures need to be developed and
validated. For example, one may need to develop a scale of
‘alienation’ or ‘radicalness’.

JOBNAME: Mcgraw−crowsemmens PAGE: 12 SESS: 32 OUTPUT: Tue Aug 28 15:02:37 2007 SUM: 43DA2FED
5 The basis for selecting a sample or samples needs to be consid-

ered.
6 Consideration also needs to be given to ethical concerns.
Data collection
Data collection involves a range of skills, and consideration of a
number of issues, including:
1 The need to negotiate access to the data, either with organiza-
tions or the individuals concerned.
2 The constraints and opportunities inherent in the fieldwork
environment. The choice of which methods to use is affected by
the environment in which you will be working. For example, a
study of homeless alcoholics is unlikely to be best pursued by
administering a 30-page questionnaire in a day centre. If the
homeless alcoholics are receiving in-patient treatment in hospital,
such a method might be conceivable, but even then there may be
better alternatives.
3 Exactly how the data are going to be obtained and recorded. This
depends on such things as whether collection is to be done by
extraction from records or documents, by interview, by question-
naire, or by observation. You may need to develop abstraction
proformas, interview schedules, consider whether or not to use a
tape recorder or other form of recording, and how such recordings
are to be processed and analysed.
4 This in turn involves consideration of how the data is structured.
If using a questionnaire, recording may be highly structured, but
qualitative methods allow for structure to emerge during the
course of study. For example, Schatzman and Strauss (1973:
99–100) refer to observational notes, theoretical notes, and meth-
odological notes.6
Planning fieldwork and data collection in advance is a critical part of
any study, because you seldom get a second chance to go back and
do it again!
Data analysis
Because the theoretical issues and data collection are interesting and
demanding, it is easy to put off consideration of how information is
to be analysed until later. It is tempting to wait and see what you
get, and then decide what to do with it. This is a dangerous line of
thought that can result in you ending up with a pile of material that
is not going to be half as useful as it could have been. If the data is
to be analysed statistically, consideration needs to be given to how it

JOBNAME: Mcgraw−crowsemmens PAGE: 13 SESS: 32 OUTPUT: Tue Aug 28 15:02:37 2007 SUM: 51F0DBD0
will be coded, whether you have the right computer program, and,
not least, whether you have sufficient statistical knowledge for the
task that faces you. It is no good, for example, finding out that the
technique you need to use is multi-level modelling if you have
neither the software nor the knowledge to use it. If you need help
from a statistician, it is better to contact someone in advance than
put them on the spot at the last minute. Conversely, there is no
point in spending time planning to use sophisticated multi-variate
techniques if your data will not meet the statistical assumptions
required for, say, a multiple regression, when simpler techniques will
suffice. Qualitative analysis involves a constant checking of emerging
concepts and their interpretation against the fieldwork material (see
Rose 1982: 123–5), and you need to decide in advance such things
as whether you are going to use a computer-based package to
analyse qualitative material, and whether the data needs to be
transcribed verbatim.
Inference and presentation

This is where you draw appropriate conclusions and relate findings
back to the hypotheses and theoretical context considered earlier.
The skill is to make the best use of the data without ‘over-reading’
it or trying to get your material to tell you things that it was never
intended for. It may also include drawing out implications for policy
and practice. The presentation and dissemination of findings are also
important at this stage. This may involve providing feedback to
agencies and individuals who have co-operated with you during the
investigation. This can be done in the form of a presentation, or in
a short and accessible summary of the research.7 Producing feedback
to participants fulfils several functions. First, it is a way of
discharging your ethical obligations, handing back some of the
information to where it came from. Second, it may provide the
opportunity to have your conclusions implemented by practitioners
and policy-makers. Finally, such presentations can be a useful
rehearsal for presenting results to the wider world. This wider
presentation will usually involve publication in some form. Present-
ing a paper at a conference is also a useful way of running through
your research in preparation for publication.
Further reading
There are several books on methods of social research in general.

These include Alan Bryman’s Social Research Methods (2004), which
covers both quantitative and qualitative methods. Research Training
for Social Scientists (2000), edited by Dawn Burton, is reasonably

JOBNAME: Mcgraw−crowsemmens PAGE: 14 SESS: 32 OUTPUT: Tue Aug 28 15:02:37 2007 SUM: 2027EBF2
comprehensive, starting with the philosophy of the social sciences,

moving on through ethical issues, qualitative and quantitative meth-
ods, to writing and presenting research.
One of our main reasons for writing this book is the fact that
there are not many texts concerned specifically with research
methods in criminology. However, there are edited books containing
chapters by various writers. Doing Criminological Research (2000),
edited by Victor Jupp, Pamela Davies and Peter Francis, has chapters
on a variety of issues, including formulating research problems, and
understanding the politics of criminological research, as well as
drawing on what can be learned from research undertaken by the
various contributors. Doing Research on Crime and Justice, by Roy
King and Emma Wincup, also has chapters drawing on the expertise
and experience of well-known criminologists such as Anthony
Bottoms, Sandra Walklate, Mike Maguire and George Mair.

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 31 OUTPUT: Tue Aug 28 15:03:53 2007 SUM: 52F88736
2 The principles of researching

criminology
In Chapter 1, we emphasized that criminological research usually
starts with the posing of a research question. This question may be
developed in response to an emergent crime problem or policy issue,
or it may be the case that a researcher sees the need to fill a gap in
the literature, or update it. It is important to recognize that from
this earliest stage of research, theory underpins the research process.
As the researcher goes on to develop hypotheses, operationalize
concepts, collect and analyse data and then draw conclusions, theory
continues to drive the research process and fundamentally influences
the final outcomes. It is extremely important, therefore, for a
competent researcher to consider the theoretical context of an
inquiry, even if s/he is not undertaking research directly associated
with the investigation of a particular theory.
In this chapter, we will explore in some depth the relationship
between research and theory. We will discover that there are two
different ways of understanding the term ‘theory’ in the context of
research. Most commonly, we understand theories to consist of
explanations which are developed to account for phenomena which
we have observed or experienced. So, for example, subcultural theory
emerged from investigations into gang behaviour and delinquency
(Cohen 1973; Thrasher 1947). It is important to realize that such
theories are developed over time, by different people, using different
methodologies and are in constant need of testing. However, we also
need to consider theory in terms of the theoretical approach we take
to the research process itself. In so doing we must consider one of
the most debated questions in social research: is it possible to
investigate the social world in a scientific way?
This chapter does not contain a discussion of all the possible
theoretical explanations available to the criminological researcher.
Indeed, there are many useful books and resources which already do
this and we refer to some of them in the Further Reading section at
the end of the chapter. The aim of the chapter is to explain how the
relationship between theory and research is essential in criminologi-
cal research as a foundation for ‘quality control’. As we saw in the
previous chapter, research is judged on its validity and reliability and
in the discussion that follows, we hope to illustrate that theory plays
an essential role in the establishment of validity and reliability in

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 31 OUTPUT: Tue Aug 28 15:03:53 2007 SUM: 3C53B065
Table 2.1 Validity and reliability in research
research. Table 2.1 provides a summary of the key concerns in

validity and reliability which we hope will serve as a useful point of
reference as you progress through the chapter. It is helpful to reflect
on these ‘standards’ when thinking about the theoretical underpin-
nings of research.
The research question: identifying the research question and exploring

possible theoretical explanations
A characteristic of criminological research is that the research

question is theoretically informed, even if the research itself is not
directly associated with the investigation of a particular theory. This
sets criminological research apart from other types of research, for
example market research, which is often more concerned with
fact-finding and polling than it is to explain and understand a social
phenomenon. The process of designing a theoretically informed
research question, or set of questions, does not always happen in
exactly the same way but there are a number of important principles
which, if followed, should facilitate the development of clear,
specific, useful questions.
Take some time to think, for a moment, about your own research
interests. If you had to design a research study, what would you
choose as a topic? What are you interested in finding out about and
why? The chances are your initial idea will be expressed in very
broad terms. You might be interested, for example, in the relation-
ship between poverty and crime or the experiences of witnesses in
the courtroom. Whatever your decision, you will have made it in
accordance with your own knowledge, experience and values. How-
ever, before you can start thinking about designing your research
strategy, it is necessary to refine your broad interests into specific
questions for investigation. In order to do this, you are going to have

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 31 OUTPUT: Tue Aug 28 15:03:53 2007 SUM: 58EAF288
The principles of researching criminology 21
to become well acquainted with the literature in the topic area you
have selected. Contained within the literature may be empirical and
theoretical perspectives on any given area and it is the job of the
researcher to decide which items are relevant to the project being
developed. By reading about a subject, you will be able to assess the
current level of knowledge in an area and identify potential gaps in
understanding (i.e. gaps in the existing theory).
All research, then, should have a theoretical element which may
be derived from a well-established theory, such as opportunity
theory or subcultural theory, or it may even be an explanatory
perspective which does not have a technical label. The process of
researching and reviewing literature bases may take place in both
library and electronic environments. Invariably, you will begin by
searching the appropriate bibliographic databases and citation indices
using key words or terms (we discuss this in more depth in
Chapter 5). Once you have drawn up a list of potentially useful
sources, the process of locating, reading and following up additional
sources begins. Often, this is a time-consuming process and it
requires good organizational skills on the part of the researcher.
Robson (2002) also recommends that techniques of ‘networking’ are
employed to ensure that the researcher knows his/her area exten-
sively. He notes that, in many cases, other researchers who have
completed research in the area are able to provide useful resources or
citations. So he encourages dissemination of ‘work in progress’
through presenting seminars, writing reviews for publication and
participation in interdisciplinary collaborations/investigations, with a
view to getting feedback and ideas.
Once the literature review has been completed, you should be in
a position to refine your original broad research idea into a
defensible research question. The ultimate research question should
primarily be focused and clear, but it also needs to be both useful
and answerable. In order to ensure this, the research question should
be accompanied by a clear set of aims and objectives.
The aims of research will usually fall into one of three categories:
to describe something, to explore/understand a phenomenon, or to
explain/measure it (Robson 2002). Importantly, though, research
studies can incorporate combinations of elements of description,
explanation and exploration and therefore it is important for the
researcher to be clear about his/her aims from the outset. A
descriptive research study does not seek to go beyond the simple
reporting of an observation. It is most likely to report a situation or
trend (so, for example, the publication of a police force’s clear-up
rates in the past five years) or it may be the kind of study which
observes a newly emerging phenomenon (or a new perspective on an
existing one). Research which has the aim of explanation and
measurement will seek to test specific hypotheses with a view to

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 31 OUTPUT: Tue Aug 28 15:03:53 2007 SUM: 52228E05
drawing conclusions which may be generalized (a good example is

the study described in Chapter 9). Research which sets out to
explore or understand something will be more open in nature and
will not always be restricted to the testing of hypotheses (see, for
example, the study described in Chapter 12). The objectives of
research may range from the development of theory or knowledge in
a particular area, to the development of methodology or technical
tools, through to a contribution to policy or practice. Or, indeed,
there may be several inter-linked objectives.
Theory (be it criminological, sociological or whatever) not only
feeds into or underpins research, but also can emerge from research,
if, indeed, that is the aim of the enterprise. It is the job of the
researcher to establish at the outset what role theory will take in the
study. In a descriptive study, for example, the role of theory may be
minimal but it may have a more important part to play in
exploratory or explanatory studies. Wherever theory does play a role,
the researcher must finally decide at which level of explanation to
operate. This simply refers to the possibility that one explanation
may be the result of a number of distinct causes, which can be
expressed as different explanations which work at different ‘levels’
(individual, group, community, society, etc.). One can look at the
motivations of individuals, one can look at the different pressures or
opportunities which are specific to certain groups or one can consider
population characteristics. These different accounts are not mutually
exclusive and researchers can validly draw hypotheses about crime at
each different level. However, what a researcher cannot assume is
that a causal factor operating at one level also operates at another.
So, for example, populations with a high proportion of unemployed
people may be more prone to crime than populations where the vast
majority of people have paid employment, but this does not
necessarily mean that unemployment necessarily motivates an indi-
vidual person to steal. To deduce conclusions about individuals on
the basis of population-level data is often referred to as the
‘ecological fallacy’ (Robinson 1950) and it is something which the
researcher needs to avoid.
To recap, then, the initial stages of research design involve some
important decisions which are inevitably underpinned by theoretical
considerations. Up until now, our discussion has centred on the role
of theories in relation to the topic of the research itself. The next
stage of research design is to select a research strategy and that
involves some quite different theoretical questions.
Choosing a research strategy
Crucially, the choice of research strategy, and later the methodology,

should follow the process of research question development. Any

JOBNAME: Mcgraw−crowsemmens PAGE: 5 SESS: 31 OUTPUT: Tue Aug 28 15:03:53 2007 SUM: 55F991CE
research study which emerges from a backward process (deciding for

example that you want to do a survey and then looking for a topic)
is likely to be badly conceived and full of practical and theoretical
problems. But what do we mean by a research strategy? Put simply, it
is a term which we are using to describe the overall design of the
research process in any given project; thus it is a kind of ‘strategic
plan’ which serves as a framework for all of the methodological
decisions which need to be made and the conclusions which are
eventually drawn. There are several different strategies available to
the criminological researcher and we will be considering them in
depth in Chapter 3, but first it is important to understand the role
of theory in the strategic planning process. We must come to
understand how strategies are firmly rooted in epistemology and how
the epistemological basis of all research projects inform the meth-
odological decisions which are later made.
What is epistemology?
Epistemology is a branch of philosophy which examines the concept
of ‘knowledge’ – what it is, where it comes from and whether
absolute, ‘true’ knowledge can be achieved. It is important in the
context of research because a researcher’s epistemological position
will influence the methodological choices s/he makes. The focus of
epistemology, in very simple terms, is that where a belief can be
justified or verified, it may be accepted as knowledge. The question
is, then, how can one justify a belief? There are a range of different
philosophies, each with its own approach, but three main schools of
thought are of interest to us here: empiricism, rationalism and
pragmatism. All were developed most significantly during a period of
history called the Period of Enlightenment (in the eighteenth
century) in which philosophers from different disciplines argued
passionately about the nature of knowledge. While it is not
necessary for us to explore each of these schools in great detail, it is
important to have a basic understanding of the key principles
involved because they each form the basis of the various different
strategies available to a criminological researcher. It is also important
to recognize that although there are clear differences between them,
they are not necessarily mutually exclusive. We offer a very brief
summary of the key points here, but the interested reader is referred
to the Further Reading section at the end of the chapter.
Empiricism has a long history in philosophy and its roots can be
traced back to Aristotle, but it was first explicitly formulated by John
Locke in the seventeenth century. Empiricists argue that the only
way to acquire knowledge is through objective sensory perception.
Thus, we acquire knowledge about the world through experience;
that is by seeing, hearing, smelling or feeling phenomena. Through

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 31 OUTPUT: Tue Aug 28 15:03:53 2007 SUM: 5615CC8B
such experience, we are able to make a posteriori statements about

the world, for example ‘water boils at 100 degrees centigrade’ or, to
use a crime example, ‘crime rates are highest in inner city areas’.
Anything which cannot be verified through experience cannot be
accepted as a truth. For the empiricist, then, the aim of research is
to formulate ‘laws’ which are based on observations of patterns or
recurring relationships, using inductive reasoning. This is the process
of reasoning which moves in a logical manner from an observation
to the drawing of more general conclusions. So, for example, we
might find that all the inner cities in England have high crime rates
and conclude that all inner cities have high crime rates. The point is
that the process of reasoning does not start with a theory – the
theory is derived from the data. The problem is that, although there
may very well be a high probability of the conclusion being true, we
cannot be completely sure.
The school of rationalism also has a long history (Plato, for
example, was a rationalist) but it was developed most notably in the
seventeenth and eighteenth centuries. The two most commonly cited
rationalists are René Descartes (who led the way in introducing
mathematical methods into philosophy) and Immanuel Kant (who is
well known for his work on moral reasoning). Rationalists take a
different approach to knowledge. While they would accept that
knowledge may be acquired through experience, as the empiricists
suggest, they argue that this is not the only way in which knowledge
can be derived. It is proposed that human reason can also be the
source of knowledge; something can be held to be true at a purely
conceptual level. Rationalists believe, then, that a priori knowledge is
possible, i.e. knowledge that is gained and verified by reason alone
and does not need to be justified through actual experience (as does
a posteriori knowledge). An example might be the statement ‘nothing
is red and green all over’. We can arrive at this statement through
logical deduction, it is not necessary to prove it empirically (Kant
1871). Rationalists argue, then, that we can begin with an intuitive
general principle (a theory) and derive complete knowledge by
building on that basic principle using the process of deductive
reasoning. To achieve this, we undergo a logical process of designing
hypotheses, collecting data to test those hypotheses and, depending
on the results, refining our original theories and developing new
hypotheses to test.
The third school we wish to raise here is pragmatism. The
pragmatic approach has been overshadowed slightly by the conflict
between empiricism and rationalism, but several pragmatists have
made an interesting contribution to the epistemological debates
(well-known names include Charles Sanders Peirce and John
Dewey). Pragmatists do not believe that one true reality can be
conceived by the human mind, instead truth is ‘what works’. The

JOBNAME: Mcgraw−crowsemmens PAGE: 7 SESS: 31 OUTPUT: Tue Aug 28 15:03:53 2007 SUM: 584F40F1
key idea in pragmatism is that ‘truth’ changes over time and across
different circumstances and, therefore, all principles should be
viewed as ‘working hypotheses’ rather than absolute truths. As a
result, it is accepted that any set of data is explicable by more than
one theory. Importantly, priority is always given to the practical, over
the theoretical, in inquiry and it is accepted that both inquiry and
facts are value-laden. The pragmatic approach has suffered signifi-
cant criticism in the past because it amounts to little more than
empirical relativism and short-sighted practicalism (Bertrand Russell
was a famous critic).
Epistemology in criminology
Let us now bring the discussion back to criminological research. It is
true to say that a great deal of criminological work is empirical in
nature, and by that we mean that research is conducted using a
working hypothesis that is tested through experiment or observation,
rather than work which is purely theoretical. The empirical approach
is most commonly associated with the positivist school of criminol-
ogy. Positivism developed in the second half of the nineteenth
century, at a time when the natural sciences and engineering were
thriving and scholars from a range of disciplines were inspired by
the advocates of empiricism. The positivist criminologists rejected
the methodology of Classicism1 and developed instead scientific
methods of measuring and quantifying criminal behaviour, producing
biological (Lombroso), psychological (Freud) and social (Durkheim)
theories of crime. However, these early positivistic approaches were
heavily criticized on the grounds that, unlike in the natural sciences,
it is not easy to create universal laws about the social world, nor is
it desirable to exclude from study those things which empiricists
would call ‘soft’ and unreliable such as people’s motivations or
reasoning. In response to such criticism, positivists developed their
methods and adjusted their principles over time in order to adapt to
the peculiarities of the social world, but at the same time retaining
scientific principles.2 Thus, you will see references to neo-positivist
and post-positivist theories which seek to make it possible to
measure ‘soft’ concepts as long as heavy emphasis is placed on the
standardization of measurement procedures. In modern forms of
positivism, the empiricist’s requirement for the establishment of
universal laws is replaced with probabilities that relationships
between concepts will hold true across all circumstances (Fischer
1998 and see also Popper’s work on falsification (1959)).
However, although positivism has had a significant role in
criminological thinking, it is not the only approach to research.
Indeed, a number of anti-positivistic schools have emerged that
criticize the overly ‘fixed’ approach to researching the social world

and advocate a different epistemological and methodological frame-

work. These include ethnography, phenomenology, hermeneutics and
critical research, all of which are commonly referred to as interpre-
tivistic approaches and involve a wide range of methods including
observation, textual analysis, life histories and interviews. These
approaches are not obsessed with fixed methods or fidelity to the
phenomenon, but instead are focused on studying a phenomenon as
it occurs in its natural state, undisturbed by the researcher. The idea
of a standardized approach to data collection is rejected and it is
argued that the method of data collection should be determined by
the nature and setting of the phenomenon under study. The focus
for research is on extracting the meaning of action, rather than
simply calculating the frequencies of actions or characteristics. This,
of course, means that this type of research is particularly vulnerable
to accusations of subjectivity and invalidity.
Additionally, an approach to research which has recently enjoyed
increasing popularity, particularly in the development of evaluation
studies in criminology, is realism. This is an approach that attempts
to combine the strengths of the positivistic and interpretivistic
schools. It places a heavy emphasis on the importance of context,
following the ‘paradigmatic’ approach of the pragmatist school in
epistemology. Thus, it has the respectability which scientific study
carries with it, but at the same time allows for exploratory methods
to be used and, importantly, context to be explored. We will discuss
realism in more detail in Chapter 4 when we look at evaluation
research, but it is important to recognize it now as a potential
research strategy.
Using epistemology in a research strategy

It is possible, then, to see how the epistemological positions play a
significant role in the development of criminological thinking.
Think, for a moment, what the empiricist, rationalist and pragmatist
might say about approaches to learning about crime. The empiricist
would argue that in order to learn about crime we must conduct
scientific, empirical studies to measure the phenomenon. In contrast,
the rationalist would argue that we can legitimately draw conclu-
sions about crime through reasoning alone. The pragmatist would
argue that we can never achieve a universal understanding of crime
– instead we must focus on studying crime-related phenomena in
the contexts of specific cultures or within social groups. By under-
standing the tension which exists between these epistemological
views, we can better understand the development of knowledge in
our discipline and also determine how best to design our own
research, taking into account the potential strengths and weaknesses
of different approaches. Remember, a good researcher places a high

JOBNAME: Mcgraw−crowsemmens PAGE: 9 SESS: 31 OUTPUT: Tue Aug 28 15:03:53 2007 SUM: 4E9B6A1F
priority on the production of research that is valid and reliable and

the epistemological basis of a research strategy is key in achieving
this. Each of the different epistemological approaches has its own
idea of what is and what is not valid, and this has to be
acknowledged in the design of the research strategy.
As we stated at the start of the chapter, it is quite rare for a
researcher to state clearly his/her epistemological position in simple
terms (for example ‘this work is positivistic’) and often it is left for
the reader to work out the theoretical underpinnings of a piece of
published research for him/herself. This is usually a relatively easy
task if you can clearly identify the aims (description, exploration,
explanation), the nature of the hypotheses (narrow or broad), the
methods used (fixed or flexible), the type of data collected (quanti-
tative or qualitative) and the way in which conclusions were drawn
(inductive or deductive reasoning). Even if it is not possible to be
100 per cent certain, usually enough detail can be drawn to allow a
sufficient assessment of the epistemological approach adopted. As a
result, we can interpret other people’s research more effectively
because we are able to question the relationship between theory and
method in the approach they adopt and ultimately assess the validity
and reliability of their work. Similarly, our own work is open to
similar analysis by others and we must be careful to ensure
consistency in our strategy and to anticipate potential weaknesses in
our approach and, wherever possible, incorporate safeguards into our
methods. It is worth noting that, although there is always a link
between epistemology and research strategy, it is not always a perfect
relationship. Sometimes, researchers employ a multi-strategy
approach to research, combining different epistemological
approaches.
Let us pause a moment to look back on the discussion so far and
review our position. We have said that all research begins with the
development of a research question and the setting of aims and
objectives. At this stage of the research, any number of explanatory
theories may be relevant and it is important for the researcher to
clearly identify the area of theory which s/he hopes to explore or
clarify and the level of explanation at which s/he intends to operate.
The next stage is for the researcher to adopt a research strategy and
to build a theoretical framework upon which the whole project can
be built. This framework will itself be built on a series of
epistemological principles and sets some important limits within
which the researcher must make the decisions about methods and
data analysis. It is these decisions which we now move on to
consider in more detail.

JOBNAME: Mcgraw−crowsemmens PAGE: 10 SESS: 31 OUTPUT: Tue Aug 28 15:03:53 2007 SUM: 50ECD6E7
Method selection and data collection
Once a framework for the research has been provided through the
selection of a strategy, the process of method design can begin. This
process must be conducted with continual reference to the aims and
objectives of the research and to the research strategy adopted. The
important theoretical implications that have influenced the research
process thus far will continue to have an important role to play. A
detailed discussion of the various methods available to the crimino-
logical researcher is contained in the chapters that follow, but it is
important for us to be aware of the theoretical underpinnings of the
methodological choices which need to be made. The choices we
discuss in this section relate to the type of data, the hypotheses, the
sampling strategy and the role of the researcher.
Type of data
One of the first decisions to make is whether to take a quantitative
and/or qualitative approach to the data collection. We have already
referred to the difference between quantitative and qualitative
research in Chapter 1. Quantitative research epitomizes a positivistic
approach, promoting the objectivity and remoteness of the
researcher, the testing of hypotheses in a formally structured meth-
odology and the collection of hard, numerical data. In contrast, in
qualitative research, the researcher is encouraged to become involved
with the subjects, looking for meaning and developing emergent
theories through the analysis of soft, non-numerical data (Hammers-
ley 1992).
A multi-strategy approach is actually common in criminological
research and we are strong advocates of such an approach. It is
important, however, to recognize that this position is subject to some
controversy. Bryman (2004) notes two different objections to multi-
strategy approaches. On the one hand, proponents of the ‘embedded
methods argument’ say that specific methods carry with them
irrefutable epistemological commitments. So, for example, if you
choose to carry out participant observation you must adopt an
epistemology consistent with interpretivism (Smith 1983; Smith and
Heshusius 1986). Since different epistemological principles are
inconsistent, it makes no sense to attempt to combine methods. On
the other hand, advocates of the ‘paradigm argument’ say that
quantitative and qualitative research are paradigms which are simply
inconsistent with each other because they have opposing assump-
tions and methods.
However, Bryman illustrates very effectively how the distinction
between quantitative and qualitative research is less apparent (and
indeed problematic) in practice. He notes the many examples of

JOBNAME: Mcgraw−crowsemmens PAGE: 11 SESS: 31 OUTPUT: Tue Aug 28 15:03:53 2007 SUM: 53CA76DE
common ground between the two approaches, illustrating that

research often includes both an inductive and deductive approach to
analysis and a combination of quantitative and qualitative data. He
concludes that the relationship between epistemology and method is
a tendency, rather than fixed and that, although there are some
differences between quantitative and qualitative approaches, it is
important not to exaggerate them.
Nowadays, particularly in criminology, researchers combine meth-
ods, theories and data in a single study to improve validity (Denzin
1988). Such an approach is referred to as triangulation and the idea
behind it is that it improves the quality and validity of research by
taking into account the possible weaknesses found in any single
method. There are four different forms of triangulation: theory,
methodological, researcher and data (see also Denzin’s kinds of
triangulation on p.000). Theory triangulation involves the combina-
tion of two or more theoretical approaches. Methodological triangu-
lation requires the combination of quantitative and qualitative
approaches. Researcher (or observer) triangulation occurs where
more than one researcher is used in the fieldwork. Data triangula-
tion is the combination of different methods of data collection. It is
possible to use any combination of these approaches in a research
study and increasingly this is being recognized as a desirable feature
in criminological research.
Hypotheses
Whether you adopt a quantitative and/or a qualitative approach, it is
likely to be the case that hypotheses are developed. As we saw in
Chapter 1, a hypothesis usually takes the form of a short statement
which speculates a relationship between two variables and is most
commonly used in positivistic research. However, it may be incorrect
to assume that the development of hypotheses is not appropriate to
a non-positivistic study, especially if the aim is to explore or describe
something. Indeed, in interpretive studies it is often the aim to look
for emergent hypotheses, as well as testing them. An example is
Becker’s (1958) participant observation in which he sought to
discover hypotheses as well as to test them. If a positivistic approach
is being adopted, it will be necessary for you to clearly define and
operationalize the key concepts which build the hypotheses. The
danger here is that you are imposing your own concepts and it is
important to ensure as far as possible that your concepts are valid
and reliable. If an interpretivistic approach is adopted, the key
concepts will become defined as the research progresses. There are
also problems of interpretation here and it is again important to
remain as objective as possible. We will consider these issues in more
detail in later chapters.

JOBNAME: Mcgraw−crowsemmens PAGE: 12 SESS: 31 OUTPUT: Tue Aug 28 15:03:53 2007 SUM: 4A514143
Sampling
We will go into the methods of sampling, which can be quite
complicated, in Chapter 3. However, it is important to realize that
the choice of sampling procedure will have theoretical underpinnings
and must be justified in accordance with the aims of the project and
the epistemological approach adopted. There are two main
approaches to sampling – probability and non-probability sampling.
Probability sampling is based on a major assumption – the assump-
tion of homogeneity. The principle is that a sample drawn is
representative of a whole and each element has an equal chance of
being included in the sample. So, when a doctor takes a sample of
your blood, the assumption is that that sample accurately represents
the condition of your blood generally. The doctor could take another
ten, or even one hundred samples and they would all be identical.
Similarly, if I make a cup of coffee, I expect the spoon of coffee in
my mug to represent the whole jar – each sample should result in a
cup of coffee which tastes the same. The methods for conducting
probability samples are strictly controlled and methodical, using
procedures to ensure randomness. The strength of this kind of
sampling is that claims of representativeness can be made by the
researcher. This kind of sampling is most commonly used for survey
research which tends to be positivistic and quantitative but is not
exclusively so.
Non-probability sampling is different. There is no way of predict-
ing the probability that different subjects will be drawn and the
researcher is often prevented from generalizing or making claims of
representativeness. The methods used are tailored very much to the
boundaries of the research. These methods might be used for
reasons of convenience or where access to a closed group of people
is very difficult. Say, for example, I wish to conduct research into
teenagers who use drugs – it would be very hard to locate the
population as the subjects might not want to be open about their
addictions, but if I were able to locate a small group, they might be
able to facilitate meetings with their friends or acquaintances. This
kind of sampling is most commonly used in non-positivistic,
qualitative research but, again, may be used in different kinds of
approach.
Role of the researcher

Once a sampling strategy has been devised, there is one final
decision to be made by the researcher and that is what role will s/he
play in the research. In some cases, the researcher will play an active
role, for example, by acting as an interviewer or participant observer.
In other cases, the researcher will adopt a more objective, scientific

JOBNAME: Mcgraw−crowsemmens PAGE: 13 SESS: 31 OUTPUT: Tue Aug 28 15:03:53 2007 SUM: 547C42D3
role, for example by administering a postal survey or by being a

non-participant observer. Whichever role is selected, thought must
be given to the epistemological framework of the study itself. The
more positivistic a study is, the more detached and procedural the
research must become.
Data analysis and inference
Recall that in Chapter 1 we warned against succumbing to the

temptation to postpone the development of a data analysis strategy
until after the data has been collected. It is, indeed, tempting to wait
and see what you get, and then decide what to do with it but this
approach is likely to result in disappointment and frustration for the
researcher. If the data is quantitative, it may need to be analysed
statistically and therefore consideration needs to be given to how it
will be coded and analysed. Qualitative analysis involves continual
codification and interpretation of emerging concepts and you need
to decide in advance such things as whether you are going to use a
computer-based package to analyse qualitative material, and whether
the data needs to be transcribed verbatim. We will go into the
principles and methods of data analysis in more detail in Chapter 8,
but the point we wish to emphasize here is that it is important to
know from the outset exactly how you wish to use your data to
address your research question, test your hypotheses and/or develop
your theory. Again, the decisions made here will relate closely to the
original aims of the research. Whether you are seeking to describe,
explain or explore a phenomenon will influence the type of analysis
you conduct.
A distinction must be made between two different relationships
between theory and data analysis: deductive and inductive reasoning
(we mentioned these earlier in the discussion about epistemology).
Deductive reasoning involves the testing of theory through the
selection of key concepts and the testing of hypotheses. Inductive
reasoning involves the generation of theory through analysis of data.
Although the two approaches can be, and usually are, distinguished,
it is important to note that most research contains elements of both
approaches, to some extent (Bryman 2004).
If you are adopting a deductive approach (most likely if you are
using a positivistic approach), hypotheses are deduced from existing
theory and subjected to empirical testing. The process of hypothesis
building starts with the identification of key concepts. These key
concepts, then, can be seen as the building blocks of hypotheses
which, in turn, come together to build theories. In order to measure
these concepts, it is necessary to transform them into variables. This
process is called operationalization and will be explained in more

detail in the following section. Once the data has been collected, it
will be possible to either confirm or reject the hypotheses. Finally,
the researcher reflects on the implications of the results for the
existing theory and feeds back into it (Bryman (2004), notes that
this is actually a process of induction tacked onto the end of the
deductive process).
The development of theory through induction is different and is
most often found in non-positivistic studies (i.e. those adopting an
exploratory approach). Once the research question has been posed, it
is not necessary to develop hypotheses for rigid testing. Instead, data
are collected and analysed and the theory is developed from the
findings. It may then be necessary to go back and explore the theory
more. Grounded theory is an approach to research which moves
from the stage of research question design to sampling and data
collection. Data is coded and categorized and the researcher begins
at that stage to generate hypotheses and build theories. By continu-
ally repeating the process until the ‘point of theoretical saturation’ is
reached (i.e. the point at which new data does not add anything new
to the concepts which have been developed), the researcher eventu-
ally develops theory out of the data (Strauss and Corbin 1998).
Conclusion
It is hoped that this chapter has illustrated the various ways in

which theory can (and must) play a part in the research process.
From the early stages of research question design, through to the
design of methodology and analysis of data, theory underpins the
research process. It is extremely important, therefore, for a compe-
tent researcher to consider the theoretical context of an inquiry, even
if s/he is not undertaking research directly associated with the
investigation of a particular theory. Most commonly, we understand
theories to consist of explanations which are developed to account
for phenomena which we have observed or experienced. But we have
seen that theory is also relevant in terms of the theoretical approach
we take to the research strategy. Crucially, a good research study has
a clear research strategy which can be justified by the researcher. As
we move on now to consider the technicalities and finer details of
research design and specific methodologies, it is hoped that the
importance of these issues will be borne in mind and, indeed, that
their relevance will be enforced.
Further reading
We have only been able to scratch the surface of the role of theory
in research in this chapter and there are a number of texts which are

JOBNAME: Mcgraw−crowsemmens PAGE: 15 SESS: 31 OUTPUT: Tue Aug 28 15:03:53 2007 SUM: 291DCFAA
useful in developing a more in-depth understanding of explanatory

and epistemological theories. Bryman (2004) gives a good textbook-
style overview of the epistemological positions in social science and
Delanty and Strydom’s (2003) collection of over sixty extracts from
classic works on the philosophy of social science is an excellent
resource for anyone interested in reading epistemological works in
their original form. For the explanatory theories, there are many
theory texts on the market but one of the most comprehensive (and,
indeed, most popular among our students) is the book by Vold and
Snipes, now in its 5th edition (2001). Bottoms (2000) also provides
a relatively short and very interesting account of the relationship
between theory and research in criminology, focusing on different
categories of criminological theory and providing a useful worked-
through example to illustrate his points. Finally, Garland’s (2002)
chapter on the development of criminology in Britain in The Oxford
Handbook of Criminology gives an interesting historical overview of
the role of theory in criminology.

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 4C3243F1
3 Designing criminological
research
In this chapter we look at what is involved in designing a research
project. In terms of the research process outlined in Chapter 1, this
is the stage between formulating a theoretically informed research
question and collecting the data needed to investigate the question.
It includes deciding on an appropriate research design, choosing an
appropriate sampling technique, and considering the important issue
of researching ethically. During the course of the chapter we will
look at some of the more widely used kinds of research design.
As ever, we emphasize that although we are setting out the steps
involved in researching criminology in a certain sequence, this is not
intended to be a rigid rule. For those who have not done much
research before, this will be a useful framework. The more experi-
enced researcher may well adopt a more flexible approach which
reflects the exigencies of a particular project.
What is research design and why is it important?
Social research is sometimes thought of in terms of particular

methods such as questionnaires and interview schedules, participant
observation, documentary analysis. Such methods are the techniques
used in collecting and handling data. Research design concerns the
overall research plan, within which such methods are to be
employed. To use a military analogy, research design is about
strategy, whereas selecting specific methods is about tactics.
For example, if you are studying how local authorities implement
legislation concerning child care, first you have to decide how you
are going to approach this. Are you going to do a survey of all local
authorities, or look in some detail at just a few, or perhaps a
combination of the two? You then need to decide what methods to
use within this design: a questionnaire if you are doing a survey,
interviews with key officials, an analysis of an authority’s records,
minutes of meetings, and so on. As we mentioned in Chapter 1, it
is quite common to have research designs that use mixed methods,
perhaps combining a survey with case studies.

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 396E18FD
Designing criminological research 35
Causal analysis
If you are hoping to find out what caused something, the research
design is an important means of ensuring that you have the basis for
determining causality.1 Usually this means being able to ‘control’ for,
or take account of, certain factors. If you are looking at the effects of
new legislation, or the introduction of a new policy then it is
important to have information about the situation before the new
policy or legislation is introduced, in order to compare it with the
situation subsequently. You therefore have to decide which measures
need to be compared. You may also need to compare instances
where changes have been made, with instances where they have not.
For example, if you were looking at whether a crime reduction
programme has had any effect, you would need to consider whether
changes in recorded crime have been affected by changes in policing
practices in some areas. In some studies you may need to look at
what is happening in different parts of the country, or among
different groups of people in order to control for such things as
socio-economic differences. A simple way of envisaging causal
relationships is shown in Figure 3.1, where the question being
addressed is whether changes in unemployment rates have an impact
on crime rates:
A variable is simply something that is measurable; that varies (see
Glossary). A dependent variable is the outcome you are interested in
(in the example below, the crime rate). An independent variable (or
variables) is what you hypothesize might be responsible for varia-
tions in the dependent variable. Intervening or extraneous variables
are those things that might influence the relationship between the
independent and dependent variables.
Figure 3.1 Causal analysis

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 556D64D7
In order to determine whether there is a causal relationship it is

necessary to ascertain:
+ whether there is a relationship between two factors (variables);
+ whether the independent variable is logically and temporally prior
to (i.e. comes before) the dependent variable;
+ whether the relationship is true or spurious; can it be attributed
to the presence of other factors (intervening variables)?
See also Bryman and Cramer (1990: 7 ff.).
To give an example, by choosing different geographical areas to
study, we can ‘control’ for socio-economic variation to some extent
by selecting a well-off area, and a not so well-off area.2 The research
design adopted is also critical for determining the basis for any
sampling. In the example concerning local authority child-care
practice, we may take a sample of local authorities and then select a
sample of people within each authority to interview. We will come
back to this shortly.
The main types of empirical inquiry
To enable us to work out what is causing a particular phenomenon

(such as a change in the crime rate) we can employ several research
strategies. In doing so, however, we need to be aware that there are
certain ways of identifying how one variable might be affecting
another. There are important distinctions between experimental
research, quasi-experimental research, and non-experimental
research. All three try to take account of the relevant variables in an
effort to make a causal inference, but do so in different ways.
Experimental studies attempt to arrange a situation in which
variables are controlled prior to the data being collected, often by
allocating people, places (or whatever units are under investigation)
to experimental and control groups. The best way to exercise control
over extraneous variables is to allocate units randomly. Random
assignment to ‘experimental’ and ‘control’ groups is crucial to avoid
selecting groups that vary in some other way, such as friends
choosing to be with each other. This is not easily achieved in
criminological research in real life, but can be simulated in some
situations. For example, magistrates could be randomly allocated to
one of two groups. Those in Group A are given details of a case,
and asked how they would deal with it. Those in Group B are given
the same case, except that one piece of information is different, say,
the defendant’s race, or employment history, or the age of the
victim, to see whether this has any effect.
As another example, is children’s behaviour affected by watching
violent films or television, or playing certain kinds of computer

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 5489A06C
games? You could expose some children to such experiences (‘experi-

mentals’) but not other children (‘controls’), perhaps by showing
them films which are either violent or non-violent in class, and
measure or observe their behaviour during play time before and after
watching the films. You would have to ensure that the controls are
not accidentally or covertly exposed to such experiences and you will
need to take account of the fact that other factors may also change
over the time period concerned. You would then measure the
dependent variable (such as the number of aggressive acts that each
child displays) and determine whether any differences between the
two groups are statistically significant. We have referred elsewhere to
the importance of doing research ethically, and such a study would,
of course raise ethical issues, including the need for parental consent.
While the example above illustrates a methodological point, because
of the ethical and practical problems involved, you are more likely to
encounter other types of study in most criminological research.
Quasi-experimental studies, sometimes called correlational studies,
include surveys. Here the variables frequently have to be manipu-
lated statistically after the data has been collected in order to try to
make inferences about the relationships between variables. It is more
difficult to determine causality because there is less control over
variables, and it may be difficult to determine the time order of
variables. Taking the violent films and behaviour example used
above, in a quasi-experimental design you might take a sample of
school-children, ask them about their viewing, and their behaviour
by means of a questionnaire, or see if they have been in trouble for
behaving badly, and compare them, taking account of their age,
gender, background and other factors.
Non-experimental studies make little or no attempt to control or
manipulate variables. Their purpose might be to look in more depth
at the mechanisms underlying a particular social process, and to gain
insight into the interpretations and perceptions of actors, or other
objects of study such as organizations. Ethnographic studies come
under this heading. Applying this approach to the films/violence
issue, you might spend time with children, watching how they spend
their time, and talking to them in order to understand how their
viewing habits and behaviour could be mediated by their social
habits.
As a further example of the three types of study mentioned above,
imagine that you are looking at whether the police are biased against
certain groups of people. You could investigate this in the following
ways:
1 by giving a random sample of police officers’ descriptions of cases
and asking them how they would proceed: would they caution or
charge, and if so, with what offence? The details of the cases are
exactly the same for all officers with one exception: for half the

JOBNAME: Mcgraw−crowsemmens PAGE: 5 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 50580FD7
group the race of the accused is white, and for the other half
(determined randomly) the accused is black. This would be an
experimental study.
2 Alternatively, you could go through a random selection of cases
the police have dealt with over the last year and abstract
information on variables such as what charges were brought and
the race of the suspect, and any other variables that might be
relevant. You then analyse this data to see whether black
defendants are dealt with differently, using statistical tests to take
account of such factors as the nature of the offence, previous
criminal history, and so on. This would be a quasi-experimental
study.
3 You could spend some time in a particular police force (or forces)
learning about police culture and the way the police think, talk
and act in relation to suspects from different ethnic backgrounds.
This would be a non-experimental study.
You could also, if time and money allow, use a combination of the
three approaches. Note that in all three instances you would need to
make decisions about whether you are going to investigate one
police force or many, and if several, should they be randomly
selected, or selected on purpose to illustrate different parts of the
country (e.g. the Metropolitan Police, one in the rest of the South of
England, one in the North of England, and one in Wales).
Furthermore, all three approaches have ethical implications. These
may be more apparent in the case of experimental designs because of
the random allocation involved, but in all three instances there is the
question of telling those involved what you are doing, and whether
this is likely to affect the way they behave. We return to ethical
issues shortly.
Some common kinds of research design
A thorough review of different kinds of research design would

require a much longer discussion, but it is worth describing a few
examples of different research strategies that are commonly used.
Prospective and retrospective studies

Studying what happens to people coming before the courts may be
done either by sitting in court waiting for cases to come up
(prospective), or going through court records to see what happened
to people who have been through the court (retrospective).
In prospective studies, it is easier to ensure that the information
you want is recorded, but they are very time-consuming. For

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 483334C5
example, there has been a lot of concern over the years about the
way in which mentally disordered offenders are dealt with by the
criminal justice system (e.g. Palmer and Hart 1996). But if you
wanted to look at how mentally disordered people are dealt with in
court, such cases are not that common compared to the total
number of cases going through a court overall (for example, Browne
et al. 1993). You might get much interesting information from such
a sample, but it would involve spending a lot of time in court to get
a viable sample, compared with doing a retrospective study of cases
dealt with in the past by using court records.
Cross-sectional studies
These involve collecting information relating to a phenomenon at a
particular point in time, usually in different locations. For example,
does the rate of school exclusions vary in different parts of the
country, and, if so, what are the differences and why? While such a
design may be a useful basis for inquiries, it does mean that all
kinds of other differences between areas have to be taken into
consideration, and of course by excluding the time variable, it means
you cannot consider how rates are changing.
Longitudinal studies
These are studies that involve collecting information over a period of
time, on people, places, or events. They are usually a major
undertaking, and there are some well-known national longitudinal
studies which have collected information on a cohort of people over
a number of years. For example, the National Child Development
Study is a longitudinal birth cohort study of everyone born in a
single week in 1958. It consists of data collected at intervals from
birth to the present for some 17,000 individuals. One of the
best-known examples of a longitudinal study in criminology is the
Cambridge Study of Delinquent Development, in which a sample of
about 500 boys, aged 8 to 9, was drawn from six primary schools in
a working-class urban area in the UK. Longitudinal cohorts like
these have been used to study the development of criminal careers,
and the kinds of factors that affect whether and when people offend,
and when they cease to offend.
Evaluation studies
There is one particular and very important kind of research design
in criminology that we have not mentioned so far. This is the kind
that involves evaluating the impact of various interventions. We

JOBNAME: Mcgraw−crowsemmens PAGE: 7 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 51C25C6F
mention it now because logically a chapter on research design is

where it rightfully belongs. However, it has played such an impor-
tant role in criminological research that it needs a chapter to itself,
so it is discussed in more detail in Chapter 4.
Choosing a research design: an example
To illustrate the possibilities for choosing and developing an appro-

priate research design, take the following question as an example:
‘Does an increase in unemployment lead to an increase in crime?’
Studies on the relationship between unemployment and crime
tend to occur when unemployment is high. When the rate declines,
such research goes out of fashion – until the next downturn in the
economy. Perhaps in itself this says something about the relation-
ship? Notice that the way the problem is defined is important. The
question could have been phrased differently, such as ‘does unem-
ployment cause crime?’. This would involve approaching the matter
in a different way, requiring a much more fundamental investigation
of the relationship. The question here is more amenable to investi-
gation, but more limited in scope – which for present purposes will
suffice.
In order to investigate such a question, there are several prelimi-
nary considerations:
+ Measuring unemployment: This has always been difficult, and the
way such matters are measured is liable to change (see Chapter 5
on using other people’s data).
+ Measuring crime: Crime figures have several shortcomings, and are
far from being a measure of true crime. Which crimes? All
offences OR should we restrict ourselves to looking at those that
one might theoretically expect to be susceptible to changes in
levels of employment, such as theft?
+ Inference: What is causing what?
+ It is important to compare like with like. If you are comparing
employed with unemployed people, are they similar in other
respects?
+ Rises in unemployment must be prior to rises in crime rate.
+ It is no good pointing to the fact that a lot of offenders are
unemployed. Many alcoholics drink tea, but how many tea
drinkers become alcoholics?
+ Generalization: Be wary of drawing global conclusions based on
limited studies. Researchers may have done a study of the
unemployment–crime relationship in a particular area. For exam-
ple, Gormally et al. (1981) studied the relationship in Northern
Ireland, but how far can their results be used to draw inferences
about the relationship in general?

JOBNAME: Mcgraw−crowsemmens PAGE: 8 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 30DF71AD
Figure 3.2 Crimes per 1000 population for police force areas by
unemployment rate by travel-to-work area
In investigating this question, there are two main kinds of study that
can be considered: studies of aggregate data and studies of individu-
als. Aggregate data are collections of statistics, such as the overall
crime rate or the unemployment rate.
Aggregate studies
Cross-sectional
These compare areas with varying rates of unemployment to see
whether crime is higher in areas with high unemployment rates.
There may be problems matching the boundaries for police force
areas for crime statistics with the travel-to-work areas used for
compiling employment statistics. There may also be other differences
between high and low unemployment areas which cause any differ-
ences. Look at the scatter diagram in Figure 3.2, for instance.
Figure 3.2 does show a correlation between crime rates and
unemployment rates, but this does not take us very far. Even
supposing that police force areas coincide with travel-to-work areas,
it leaves us wanting to know much more about the areas concerned.
Take the case at the bottom with a crime rate of less than 0.5 per
1000 population. It has a fairly average level of unemployment, but
the lowest crime rate. Why is it so exceptional? As it happens, this
represents an area in rural Wales, so one might hypothesize that it
has a strong community identity which helps to protect it from the
worst ravages of unemployment. One might want to investigate this

JOBNAME: Mcgraw−crowsemmens PAGE: 9 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 46F6C1F5
area in more detail. In passing, it is worth noting that this also

demonstrates the value of not just looking at overall trends, but also
examining individual instances within them.
Time series analysis

In contrast to the cross-sectional design, time series designs look at
the same variables over a period of time. To address our question,
we might therefore look at trends to see whether crime goes up
when unemployment goes up and vice versa. This has been studied
by economists (e.g. Brenner 1976), but the period of time chosen
may be important because there may be a relationship over some
periods but not others (Tarling 1982), and this warrants further
examination. It is also necessary to take account of underlying trends
in economic indicators and lag effects – the impact of rising or
decreasing employment levels may not become apparent for a while
after they occur.
In both types of aggregate study it is necessary to control
statistically for other factors which may explain any co-variation.
Such aggregate studies are often based on econometric models, and
tell us little about the people involved, the impact unemployment
has on them and the way in which it affects their options and
behaviour as individuals, hence a second kind of study taking
individuals rather than aggregate measures is useful.
Individual studies
Matching groups
This design compares groups of employed and unemployed people
to see whether the latter have committed more crimes than the
former. The main problem here is comparing like with like (match-
ing). We could follow a cohort of school leavers, comparing those
who get jobs and those who didn’t. What problems might be
encountered here? Perhaps the kinds of people who don’t get jobs
are also the kind of people who get into trouble, either because they
have a particular kind of personality or because they come from a
particular background. If the latter, they may simply be more likely
to be apprehended by the police but not necessarily be more
delinquent.
Longitudinal studies
This design looks at a single group of people, comparing the extent
of their offending when they are out of work with when they are in
work. In this instance they ‘act as their own controls’. One such
study was undertaken using the Cambridge Study of Delinquent

JOBNAME: Mcgraw−crowsemmens PAGE: 10 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 4B497B6D
Development cohort referred to earlier for a period during the 1970s

(Farrington et al. 1986), and another in the early 1980s looked at
young people in Northern Ireland (Gormally et al. 1981).
Studies based on individuals are preferable to aggregate studies in
many ways, but have some disadvantages:
1 It is more time-consuming and expensive to study groups of
people than to examine the statistics on unemployment and crime
which are already available.
2 They are more limited in coverage, and therefore less generaliz-
able.
As you can see, there is no easy, simple solution to the question of
which research design to adopt, and this is often the case in
criminological research. Researching such questions often involves
finding the best option given the circumstances and resources
available, while acknowledging its limitations. This does not mean
that we cannot answer such questions, but it may mean that it is
necessary to look at what emerges from a number of studies using
different approaches in order to get the full picture, and this is the
case with the unemployment and crime question (for a further
consideration of such studies see Crow et al. 1989).
Sampling
Whatever kind of design is adopted, it is likely to involve looking at

only a proportion of the cases, people, places, events, etc. that you
are studying. Consequently, sampling is an integral part of any
research design.
What is sampling?
A sample is part of a population, which is studied in order to obtain
estimates about the population (e.g. the proportion of a population
who are the victims of theft). A population is a set of all the units
about which an inference is to be made. Consequently a sample has
to be selected in such a way that the conclusions drawn are valid, i.e.
it must be representative of the population. Sampling is done
because it is often not possible to study the whole population.
It is important to bear in mind that beyond the population from
which the sample is drawn there may be a wider universe about
which generalizations are being explicitly or implicitly made. For
example, in selecting a sample of schoolchildren from a school in
Nitherington, are you seeking to draw conclusions about that
particular school, about schools in Nitherington, or about the
educational system? If anything other than the first (that particular

school), then clearly a sample from one school, while it may be

suggestive and illuminating, is not sufficient in itself.
Sampling what?
We referred above to ‘units’. This is because a sample can consist of
almost anything: people, organizations, places, objects or events (e.g.
a study of fights in a town centre), or a combination of these. It is
necessary to say what your sampling unit is.
Sampling frame
It is also necessary to identify a sampling frame. This is the list of
all the units in the population from which the sample is to be
selected (e.g. an electoral roll, a list of firms of solicitors, a list of all
the magistrates courts in the country). However, be warned that
sampling frames can be less than perfect. For example, if we are to
take a sample of students from those registered at a particular
university, despite everyone’s best endeavours, the university’s register
may be less than 100 per cent accurate. Similarly, if we are hoping
to sample from all the arrests made by a police force in a particular
year, the force’s records may contain errors.
Sampling fraction
The sampling fraction (f) indicates what proportion the sample (n)
is of the population (N). Thus, f = n ÷ N. For example, if there are
400 students in the first year of a criminology course and we sample
40 of them, the sampling fraction is 40 ÷ 400, or one in ten.
Size of sample
There is no hard and fast rule stating what size a sample should be.
The best rule is the bigger the better, because the bigger a sample is,
the closer it gets to the size of the population it represents. But
clearly if you are trying to make inferences on a large scale about
the population of whole towns, or countries, the proportion you can
realistically sample is not going to be anywhere near the size of the
population. This makes it all the more important to ensure that your
sample isn’t biased.
In practice, the size of a sample depends on several things. One of
them is simply the resources you have at your disposal. With lots of
money it is possible to acquire larger samples, but most research
depends on limited resources. So the next thing is to consider the
nature of your population (is it very diverse, so that you have to

JOBNAME: Mcgraw−crowsemmens PAGE: 12 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 3E1913EA
Table 3.1 Cell frequencies in a cross-tabulation
cover lots of different groups, or relatively homogeneous?), and what

do you hope to do with your data? If you are using quantitative
analysis, then you are likely to want to break your sample down into
sub-groups (e.g. men and women, young offenders and adult
offenders, type of offence committed), and examine those sub-
groups in relation to various dependent variables (such as what kind
of sentence they get). Let’s suppose you want to see whether people
from different ethnic backgrounds receive different sentences, you
would have a cross-tabulation3 of ethnicity with sentence (keeping
the categories as simple as possible) as shown in Table 3.1.
As a general rule, in order to carry out the necessary statistical tests,
you need to have at least five cases in each cell (no pun intended) of
the grid, and there is a convention in statistics that the ideal should
be 30 per cell or more. As you can see, even at the lowest and
simplest level you need a sample of 80 for this analysis. If you are
analysing your data in various ways, perhaps differentiating between
age groups, gender, and so on, then even with a sample of several
hundred, you can quickly end up with quite small numbers in each
cell of the cross-tabulation. This is particularly likely to happen in
criminology, where certain groups in the population, such as women
and those from ethnic minorities, tend to be less frequently
represented than in the population as a whole. Consequently it is
necessary to think backwards from the kind of results that you
anticipate needing.
Sampling, error and inference

In itself size is no guarantee that the sample is representative. An
important aspect of sampling concerns the extent to which your
sample can be said to reflect the population being studied. If you are

JOBNAME: Mcgraw−crowsemmens PAGE: 13 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 4EAF8E39
studying victimization, can the characteristics of the sample inter-

viewed be said to hold for the population from which the sample is
drawn?
Two concepts are relevant when considering representativeness.
One is bias. All samples are liable to various sources of error, for
example, as a result of interviewers not selecting people properly.
Another source of bias is non-response, which occurs as a result of
some people or organizations not taking part in the study for various
reasons, such as choosing not to, or by an oversight get missed out.
Those who should form part of a sample according to the selection
criteria, but do not for whatever reason, may well be different in
some systematic way from those who are included. For example, if
you were trying to study a sample of ex-prisoners to see how well
they managed to settle back into society, the ones most difficult to
include may well be the ones with most problems who are most
likely to re-offend. The sample is biased, and therefore not truly
representative of the population you are interested in. The important
thing to note about this is that bias is not reduced by having a
larger sample. Indeed, if the bias is systematic, a larger sample will
only make things worse, and increasing the size of the sample may
be misleading because it makes it appear as if the sample is
convincing when it isn’t.
The other concept is accuracy. Accuracy can be increased by
increasing the sample size, because increasing sample size causes the
sampling distribution of the mean of a variable to approach a
normal distribution.4 However carefully done, a sample will rarely be
like the population from which it was drawn in every particular. No
one will expect it to be exactly the same. The important thing is to
be clear and open when reporting a study about how it was done,
and how any sources of bias or inaccuracy may have arisen.
Wherever possible (i.e. where the parameters of the population
are known), then you should check to see whether your sample is
representative of the population. Inferential statistics are used to
determine whether a sample came from a particular population. For
example, if you were doing a study of some aspect of youth
offending and took a random sample of young offenders from one
or more Youth Offending Teams, then we would hope that the
principle of random selection would lead to your sample being
typical of all the young people being dealt with by the YOT.
However, it is prudent to check to see whether this is in fact the
case. In such an example, the YOT is likely to have certain
information about the characteristics of those on its caseload, such
as their ages, the proportion of males and females. If we just take
mean age as an example, then by calculating the standard error of
the mean of the sample it is possible to say whether the mean age

JOBNAME: Mcgraw−crowsemmens PAGE: 14 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 4C25BED4
of the population (in this instance, the YOT caseload) falls within
the confidence limits of your sample. More is said about analysis in
Chapter 8.
Types of sample
There are two main kinds of sample: random or non-random. These
are also referred to in some references as probability and non-
probability samples. The distinction is an important one because true
random samples are by nature representative of a population, and
enable you to make statements of probability about the populations
from which they are drawn, whereas non-random ones, however
carefully selected, are not, and this affects the extent to which one
can generalize from the sample to a population.
Random samples
These are also known as probability samples because the essence of a
random sample is that every unit in the population has an equal
probability of being selected, i.e. there is no bias in the selection
process. This is what makes it representative of the population. Even
so, statements made about the population based on a random sample
are statements of probability, not certainties, because error can creep
in. So we can say that there is a 95 per cent probability that the
sample mean for a particular variable such as age is within a certain
range of the true (population) mean, but because there is always the
possibility of error, we can never be 100 per cent certain.
Random samples can take various forms:
+ Simple random sample – This consists of picking a sample at
random by, for example, choosing every tenth person on a list
(the starting point being selected randomly), or by using a table
of random numbers.
+ Stratified random sample – This consists of dividing the population
into identifiable groups, or strata, and then sampling randomly
within strata in order to ensure adequate representation of the
strata. For example, if choosing a sample of students, you may
want to divide them into 1st, 2nd and 3rd year students and
sample each year. Other examples of strata are race, gender or
social class. Women are much less likely to be charged with
offences than men, so in criminology if you want to ensure
adequate representation of women in your study, it is usually
necessary to stratify the sample in order to make sure that you
have an adequate sample of women from which to make
generalizations, as shown below.
A stratified random sample with a constant sampling fraction is one
where the same proportion of each stratum is selected, irrespective

JOBNAME: Mcgraw−crowsemmens PAGE: 15 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 4AB3A315
of the size of the stratum. For example, there may be 200 first year
criminology students, 400 second years, and 300 third years in the
population of a university. If we take one in ten of each (a constant
sampling fraction), we get 20 first years, 40 second years and 30
third years in a sample of 90 students out of a total population of
900
90 1
[f = 900 = 10].
A stratified random sample with a variable sampling fraction is

one where the proportion selected from each stratum is different,
usually in order to obtain roughly equal numbers of each stratum in
the sample regardless of the number of each stratum in the
population. We may, for example, want to compare the way in which
men and women are dealt with by the courts, but because females
only comprise a small proportion of those coming before the courts,
a simple random sample might yield insufficient numbers of females
to make any statistical inference possible. Thus, if there are 6000
male defendants and 2000 female defendants, and we want 50 of
each, we might take 1 in 120 males
50 1 50 1
[f = 6000 = 120], and 1 in 40 females [f = 2000 = 40]
+ Multistage, or cluster, sample – This is where there is first a sample

of large units, chosen at random, then in each of these a random
sample of smaller units is taken. So rather than try to sample all
students in the country, we might first take a sample of
universities, then take a sample of students within these univer-
sities. Several stages may be used, such as constituency, ward,
street, and household. The British Crime Survey uses this kind of
sampling technique. The approach is more economical than trying
to sample the whole population as one group, but is less accurate
for a given size of sample because error is liable to occur at each
stage of the sampling process.
In addition, a sample may be weighted to take account of the fact

that certain categories are likely to be more or less numerous in a
population. Again, the British Crime Survey has done this in order
to increase the proportion of black people in the sample to ascertain
the level of victimization among ethnic minorities. To take another
example, when carrying out a study of sentenced offenders, the
researcher may sample one out of every two people fined, because a
large majority of those sentenced at magistrates courts are dealt with
by means of a fine; it is more economical to allow for this. When
analysing the results it is necessary to take account of this weighting.

JOBNAME: Mcgraw−crowsemmens PAGE: 16 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 5396FF6B
Non-random samples
These are not selected at random, with each unit having an equal
chance of inclusion. This means that probability statements cannot
be made, and it cannot be assumed that they are representative of
the population. Nonetheless, they may be the best that is possible in
certain circumstances. This is true of many studies of problematic
social issues, where a sampling frame may be impossible. Consider
trying to do a study of homeless people, or problem drug users, or
people with debt problems. Criminology is particularly likely to
encounter such situations because it is by nature concerned with
deviant groups. When offenders come into contact with the law and
end up on police files, in the courts, on probation caseloads, or in
prison, then they can be sampled randomly as a sample of court
cases appearing before a court, or as part of a prison population.
Even then, as every criminologist knows, the offenders who end up
in these situations are the ones who get caught, so they are not
representative of offenders generally. This does not mean that
non-random samples are not worth considering, but it does mean
that any generalizations based on them need to be qualified. There
are various kinds of non-random sample:
+ Quota sample – is one chosen in order to obtain a certain balance
of the population in the sample: so many men, so many women,
a certain number of people from each social class, or each racial
group. It is representative with respect to these determining
variables, but not others, and thus the sample may not be typical
of the population as a whole. Researchers debate the merits of
quota sampling5, but it is widely used in political polls and
market research.
+ Snowballing – this is sometimes done when you can gain access to
certain members of a population but not others. For example, you
might interview all the drug misusers coming to a treatment
centre for help, then ask if they will introduce you to friends of
theirs who have not come to the centre. In this way you can find
out why some people seek treatment and others don’t.6
+ Other kinds of non-random sample sometimes referred to are
availability, judgemental and accidental samples. As their names
suggest, these are variations on the theme of using whatever you
can get.
Most considerations of sampling tend to centre on quantitative
studies. However, it is worth noting that qualitative methods of
inquiry also usually involve sampling in some way, in that they ‘dip
into’ the phenomena they are investigating. Here also some consid-
eration needs to be given to how the units of study are chosen, and
what they are intended to tell us. Such is the predominance of the
survey based on probability sampling that it is easy to overlook the

JOBNAME: Mcgraw−crowsemmens PAGE: 17 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 5361CCFC
significance of the atypical and the different. A survey may reflect

the ‘average’ family with its 2.4 children, but an extreme example
can also tell us a lot. Take, for instance, the correlation between
recorded crime, based on a sample of police force areas, and the level
of unemployment in an area referred to earlier. This may show that
the trend is for those areas with higher levels of unemployment to
have higher crime rates, but as mentioned previously, the interesting
case may be the area which has high unemployment, but low crime
rates. Why is this? Can it be explained in terms of the special
characteristics of such an area, which may have important lessons for
how other places can protect themselves from the effects of
unemployment?
A valuable contribution has been made to social research by the
case study, whether it is of individuals, organizations or places, where
one or a very few cases are selected for study. They may be selected
on the basis of their special characteristics, or because they typify the
kind of issues being investigated. Whether the aim is to provide an
accurate picture of the whole population, or to study particular
examples, whatever technique is employed, it is important that the
researcher describes how the sample was acquired (even if it is a
sample of one) so that readers know exactly what its strengths and
limitations are, and the extent of its applicability.
Researching criminology ethically
Ethical principles
Awareness of ethical issues has been increasing in recent years. This
has taken a variety of forms, including among other things, the ways
in which countries and organizations do business with each other,
and an acute awareness of medical ethics in the aftermath of organ
retention scandals. The principles that should govern ethical behav-
iour generally also apply to criminological research, but when
applied in this context they have particular characteristics, and raise
specific issues. We cannot do justice to the full depth of these issues
here, but aim to discuss some of the main considerations.
Guidance on ethical research is set out in several places. Most
academic disciplines and organizations have ethical guidance, and
the first place to go for any criminologist is the British Society of
Criminology (BSC), whose guidance can be found on their website.7
Universities and other bodies engaged in research will also (or
should have) ethical procedures in place, as will funding bodies, such
as the Economic and Social Research Council (ESRC), which has
produced a Research Ethics Framework.8
Important though they are, just pointing to such guidelines is not
the whole answer. One problem is the question of enforceability.

JOBNAME: Mcgraw−crowsemmens PAGE: 18 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 57809678
Membership of the BSC is entirely voluntary, and while a member

who flouted the society’s ethical guidelines could be expelled, this
does not have the same effect that it does if someone is struck off
as a doctor. Nor does the BSC vet research proposals. While there is
provision at UK universities and other bodies employing researchers
to vet proposals, it is nonetheless a good idea when planning your
research to consider the ethical aspects, rather than wait for an
ethics approvals committee to return your application. In order to do
this it is necessary to have an appreciation of the issues involved.
First, it is important to understand why researching ethically is
important. The reasons may appear to be self-evident, but it is
necessary to appreciate the basis on which ethical research is
founded. Research involves a transfer of information, usually to the
researcher and whoever is paying for the research. Thus, research is
a form of power. Information is taken from someone (perhaps in an
interview or questionnaire), and may then be used by the researcher
and others in ways over which the owner of the information has no
control. Research can also empower people. For example, action
research, a methodology in which the research is done in conjunc-
tion with certain groups, can give people more power over their lives
and their options. This is because they can exert a certain amount of
control over the research process and its results. So the key issue is
control, and since research depends on resources, those who have
most resources will be able to better control research and its results.
This means that when research is undertaken, it should normally
only be done with the express permission of those involved.
However, this can lead to problems.
Research ethics are informed by various philosophical and theo-
retical positions. For example, some theorists give individual rights
pre-eminence (e.g. Gewirth 1996), while others stress the need for
community values to take priority (e.g. Sandel 1982). We cannot do
justice to the range and depth of argument here, and more
comprehensive reviews can be found elsewhere (Beauchamp and
Childress 2001). However, (and to oversimplify) one of the main
distinctions in ethical theory is that between consequentialist and
deontological theories of morality. The consequentialist view, derived
from the work of Jeremy Bentham and John Stuart Mill, has a
utilitarian focus. This position would argue that even though
research may sometimes have adverse features, such as invading
someone’s privacy, these should be outweighed by the beneficial
consequences for certain groups or society as a whole. The problem
with this view is who judges what is in people’s best interests, and
this raises the issue of the accountability of research. The deonto-
logical view, on the other hand, has its origins in the writings of
Immanuel Kant. This tells us that certain things must be done or
not done for moral reasons, irrespective of what we wish to be the

JOBNAME: Mcgraw−crowsemmens PAGE: 19 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 4A4A9C1F
case (what Kant referred to as the categorical imperative), and that

no one should be treated as a means to an end. Thus, certain actions
are not permissible regardless of the consequences. So while conse-
quentialists seek to maximize the best possible outcomes, deontolo-
gists argue that an ethical person should act independently of the
consequences. In recent years a rights-based position based on liberal
individualism has been the more persuasive, and measures to protect
individual human rights have tended to prevail. Although theoretical
positions may differ and have different emphases, Beauchamp and
Childress suggest that, when it comes to framing practical ethics, the
distinctions are not as great as they may appear (ibid.: 376–7).
As a result, ethical research needs to be based on clearly defined
principles:9
1 The right to self-determination: people have the right to decide
whether or not to take part in research. This is usually referred to
as informed consent.
2 We should not knowingly cause harm, or expose people to
unacceptable risks.
3 We should not put people at a disadvantage: our research should
be governed by fairness and equality.
4 We should be honest and truthful: we should not deceive people.
We should keep our promises. We should avoid fraud and
plagiarism.
5 We should respect people’s privacy.
6 We should preserve confidentiality.
Ethical dilemmas and criminological research

These principles are high ideals, and not surprisingly they are not as
simple to observe as they may appear. Thus, for example, not
exposing people to unacceptable risks can be problematic because
risk is a relative concept, based on the probability of something
occurring. However, the main problems arise from ethical dilemmas.
One kind of dilemma is whether research should proceed if it
compromises any of the above principles. For example, there is a
particular problem regarding covert research, which by definition
means that informed consent is unlikely to be obtained. The ESRC’s
guidelines recognize that:
Covert research may be undertaken when it may provide
unique forms of evidence or where overt observation might
alter the phenomenon being studied. The broad principle
should be that covert research must not be undertaken lightly
or routinely. It is only justified if important issues are being

JOBNAME: Mcgraw−crowsemmens PAGE: 20 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 4A8A30C5
addressed and if matters of social significance which cannot be

uncovered in other ways are likely to be discovered. (para.
2.1.4)
It is also noted that:

As is recognized elsewhere (see Tri-Council of Canada, 2002.
http://www.pre.ethics.gc.ca/English/policystatement/
introduction.cfm) Research may be ‘deliberately and legiti-
mately opposed to the interests of the research subjects’ in
cases where the objectives of the research are to reveal and
critique fundamental economic, political or cultural disadvan-
tage or exploitation …. Much social science research has a
critical role to play in exploring and questioning social, cultural
and economic structures and processes. (para. 2.1.7)
The ESRC guidelines also recognize that:

Informed consent may be impracticable or meaningless in some
research, such as research on crowd behaviour, or may be
contrary to the research design, as is often the case in
psychological experiments where consent would compromise
the objective of the research. In some circumstances – such as
users of illegal drugs – written consent might also create
unnecessary risks for research subjects. Even in this last case,
however, a researcher should seek informed consent where
possible to secure the trust and confidence of those involved. In
some contexts consent may need to be managed at a point
beyond the completion of research fieldwork, for example,
where covert observation is necessary and warranted. This
might apply to research in the field of deviance especially
where it involves illegal or immoral behaviour. (para. 2.1.4)
Another kind of dilemma arises when one of the principles comes

into conflict with another. This can happen when the requirement of
confidentiality conflicts with the avoidance of harm: what should be
done when, having given an undertaking of confidentiality to
respondents, it is discovered that one of them is likely to cause harm
to other people?
As may be deduced from the quotations from the ESRC
guidelines above, criminology is particularly susceptible to ethical
dilemmas because of the nature of its subject matter. The crimino-
logical researcher is likely to be dealing with illegal activities, with
people who have committed and have been the victims of crime, and
inquiring about difficult topics, such as substance misuse, physical
and sexual abuse, mental disorder, and many other things requiring

JOBNAME: Mcgraw−crowsemmens PAGE: 21 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 441B2C19
care and sensitivity. Perhaps some indication is best conveyed by

referring to a few examples, which are taken from actual situations.
Box 3.1 Example 1

You are undertaking a study of people addicted to drugs, which
involves in-depth interviews with addicts. You have given assurances
of confidentiality. During the course of one interview a female
respondent discloses that her partner sometimes hits her young
child. She begs you not to tell anyone, because being a known addict
she already has problems with the Social Services Department, and
is afraid that her child will be taken away from her.
Box 3.2 Example 2

You are undertaking an ethnographic study of attempts to rehabili-
tate offenders who are in a hostel for ex-prisoners. The success of
your study depends on gaining the trust of the hostel residents, and
on them being able to confide in you. During the course of your
studies you learn that one of the residents has a firearm hidden in
his room. What should you do?
Comment: Both these examples illustrate a conflict between confi-

dentiality and preventing harm. In the first example, in Box 3.1
most people would agree that the interests of the child come first,
but there are various things that could be done. The first is that
situations like these might have been anticipated when planning the
research, and prepared for. It could be that the undertaking of
confidentiality should be modified from the outset by making it
clear to respondents that there are circumstances in which you
would have to do something. It is also worth noting that the young
woman addict has put the researcher in a difficult position, and may
well be expressing her anxieties in a ‘cry for help’. This raises the
issue of the limits of a researcher’s competence: researchers are not
counsellors, and should not try to be so. However, it is possible that
the researcher could refer the respondent to a suitable agency or
individual for help.

JOBNAME: Mcgraw−crowsemmens PAGE: 22 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 54920BAE
Box 3.3 Example 3

Another situation might arise if you were looking at the relationship
between race and sentencing. The most direct way of doing this is to
look at the sentences given by magistrates and judges, but if you tell
the courts that this is what you are doing, then you alert sentencers,
and this will influence their behaviour, at least for the duration of
the study. Should you therefore effectively deceive the courts,
perhaps by a less than truthful claim that you are looking at the
‘factors that play a role in sentencing’?
Comment: Some might argue that in this instance deception is

justified because your research is investigating a potential injustice (a
consequentialist argument), especially since sentencers are in posi-
tions of power. But are the more powerful subjects of research less
entitled to be treated in an ethical way than the less powerful?
Earlier in this chapter, and in the next, we make reference to the
ethical issues that arise in criminological research when considering
random allocation. This can be illustrated by an experiment that
took place some years ago in a juvenile court in the north of
England. This adopted a system whereby delinquents who had been
truanting from school were not dealt with by the court immediately,
but had their cases adjourned for a number of months, during which
time their school attendance was monitored. What happened to
them subsequently depended on their behaviour during the period of
adjournment. This was nicknamed the Damocles experiment, and in
order to test its efficacy, juveniles were randomly allocated either to
be given the Damocles treatment, or to be dealt with in the normal
way (Hullin 1985). The procedure was controversial and was
eventually discontinued.
Conclusion
There are undoubtedly further aspects of these scenarios that might

be considered, and this is only a brief review of research ethics and
the moral theory on which it is based. The aim here is to ensure
that those researching criminology are aware of the duty that they
have towards those who take part in their endeavours. It is also
worth noting that, in accordance with one of the themes that
underlies this book, ethical practice should be regarded as an integral
part of the research process, and the kinds of ethical issue that arise
will reflect the kind of research being undertaken. Thus, surveys
most often raise issues of confidentiality; qualitative methods, such
as participant observation, raise more acutely such matters as

JOBNAME: Mcgraw−crowsemmens PAGE: 23 SESS: 31 OUTPUT: Tue Aug 28 15:05:05 2007 SUM: 496C40AF
informed consent, and random controlled trials may mean that some
of those taking part are put at a disadvantage. Evaluative studies can
have significant implications for the futures not only of those who
are the subject of an intervention, but of the staff involved in an
intervention.
Having indicated some of the problems that can arise, what can
be done about them? Can criminologists not do any research that
involves observing other people’s behaviour without telling them, or
never get involved in any research where they might encounter
illegal behaviour because they cannot give undertakings of confiden-
tiality? Some might think so, but an alternative is to ensure that
there are adequate safeguards.
There are two main ways of doing this. One is careful planning
and preparation prior to the start of any research. While you can
never anticipate all the situations that might arise, as the examples
above illustrate, it is not too difficult to see how problems might
occur. The planning and preparation should involve consultation
with colleagues or supervisors, and with any agencies that might be
involved. Careful scrutiny of ethical guidelines is also important, and
of course any reputable project will need to be reviewed by a
research ethics committee. But this should not be the end of any
ethical concerns. The other main way of responding to ethical
problems is continued monitoring and review of the research as it
progresses. This can be done through formal advisory committees, or
it may be that arrangements can be made with others to see where
problems might arise. This is particularly important in the case of
ethnographic research, where the researcher can become caught up
in events, and not realize that they are stepping over the lines of
what is acceptable professional behaviour.
Further reading
Bryman (2004) has chapters on research design and sampling, and

qualitative, quantitative and mixed method approaches to research
design are considered by Creswell (2002), who also has a section on
ethical issues in Chapter 3.
Regarding research ethics, Beauchamp and Childress’s book is
largely concerned with medical ethics, but Chapters 1 and 8 contain
useful discussions of ethical theories and the general principles. The
guidelines produced by the British Society of Criminology and the
Economic and Social Research Council should also be consulted.

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 4382989F
Defining terms
One particular, and very important kind of research design in

criminology is that which involves evaluating the impact of various
initiatives and interventions. The term ‘intervention’ can refer to a
variety of things. It could mean:
+ the way that individual offenders are dealt with, either by
sentence or by participation in a treatment programme;
+ an initiative designed to reduce crime, such as putting on more
police officers, or more surveillance technology, in an area;
+ new legislation;
+ social policies intended to affect crime, such as more drug
treatment or education programmes.
Evaluation designs are significant in two main respects. The first is
the way in which they fit into criminological and methodological
theory. Evaluation should not just be seen as a technique, but as a
way of theorizing about what is happening. The second is their
political and policy implications; most evaluations are likely to have
come about as a result of some practical or policy initiative, and
their results are therefore likely to have very real consequences.1 In
this section we will start by looking at the nature of evaluation, and
then go on to consider the practical aspects of doing evaluation.
Before doing so, however, it is useful to start by defining some
terms.
+ Evaluation is the process of examining whether interventions
achieve their desired outcomes. However, the simplicity of this
definition conceals a world of complexity. It is also important to
define some other terms commonly used in association with
evaluation (and not infrequently confused with it).
+ Aims are the goals which an initiative or organization is striving
to achieve.

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 310A82DF
+ Monitoring is the process of checking what is happening in

relation to what an initiative or organization planned to do.
+ Inputs are what are put into an initiative, such as time, money,
and activities.
+ Processes are what happen in order to make an initiative occur
(also referred to as implementation).
+ Outputs are the direct results of inputs, such as the number of
people who have been involved in the intervention.
+ Outcomes are the desired and specific results, such as a reduction
in offending, a reduction in the amount of burglary, reduced fear
of crime, reduced incarceration, reduced costs, and so on. Out-
comes should be directly related to aims, and should be based on
measurable criteria.
We will give an example later of how these terms might apply to a
particular project.
The evaluation paradigm
When considering an evaluation design we come back to the model

of causality that was referred to in the previous chapter (see Figure
3.1, p. 41). Here, however, the dependent variable(s) are the
intended outcome(s) and the independent variable is the interven-
tion itself (Figure 4.1).
Figure 4.1 Causal analysis in evaluation

Criminological evaluation 59
Figure 4.2 The OXO Model

The critical bit is the intervening variables. In evaluation it is
important to compare like with like, so that the only thing that
varies is the intervention. The reasoning is that it is only when all
other factors have been controlled that the effect of the intervention
can be seen. The traditional way of doing this is by applying the
intervention to one situation or group of people, but not to another
situation or group, which is like the first in all other respects. This
is sometimes referred to as the OXO model (Figure 4.2).
This model is similar to that employed in clinical trials in
medicine when a new drug or other form of treatment is being
developed. The ideal situation is one where those who get the
treatment are randomly allocated to the intervention or non-
intervention situation – the random controlled trial (RCT). Those
allocated to the non-intervention situation usually get something,
such as an inert substance, known as a placebo, which mimics a
situation where nothing happens.
However, when it comes to applying the model to criminology
problems arise. First, there are difficulties in applying the model in
its pure form in real life. It is difficult enough if one is trying to see
whether a particular sentence or programme for offenders is more
effective than another, but even more difficult if you are looking at
whether, for example, initiatives that affect a whole community are
effective. To some extent these can be overcome by what is referred
to as quasi-experimental designs (see Chapter 3; Cook and Camp-
bell 1979), where statistics are used to take account of possible
intervening variables when comparing intervention with non-
intervention situations, but this is not as effective as RCT. For a
start you have less control over what has happening in the
non-intervention situation. Rarely will nothing be happening; indi-
viduals and communities that are not receiving an intervention are
quite likely to be experiencing something, and researchers need to
document what this is. There are also ethical concerns about the use
of random allocation, which involves subjecting one group of people
to something while denying it to another group.2

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 345B088E
In recent years American researchers have produced a more highly

developed version of the OXO model. As a result of reviewing
evaluations of crime prevention projects Sherman et al. (1998)
developed the Maryland Scale of Scientific Methods as a way of
determining how strong the internal validity of each study was.3 The
scale ranked each evaluation study on internal validity, from 1
(weakest) to 5 (strongest). Rating was based primarily on three
factors:
1 Control of other variables in the analysis that might have been the
true causes of any connection between a programme and crime
prevention.
2 Measurement error from such things as subjects lost over time, or
low response rates.
3 Statistical capability to detect programme effects (as a result of
things like sample size, and the base rate of crime at the outset).
Table 4.1 summarizes the key elements in the scoring of various
evaluation research designs (where X = present and O = not present).
‘Before-After’ refers to whether measures of outcome variables were
taken before and after implementation of an initiative. ‘Control’
refers to whether there were any control groups, or control situa-
tions, where intervention did not occur. ‘Multiple units’ refers to
whether there were a number of units involved (such as people,
communities, or agencies). ‘Randomization’ refers to whether there
was randomization in the allocation of units to intervention or
non-intervention situations. The levels can be further explained as
follows (with our comments in italics):
Table 4.1 The Maryland Scale of Scientific Methods (a)

JOBNAME: Mcgraw−crowsemmens PAGE: 5 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 3BC243EA
Level 1 corresponds to a study where there is a correlation between,

for example, a crime prevention programme and a measure of crime
or risk of crime at a single point of time (which cannot give very
much, if any, indication that the programme has an effect on crime).
Level 2 shows that there is a temporal sequence between the
programme occurring and the outcome measure of crime (i.e.
following implementation of a programme, crime levels change, but it
cannot be shown that this is due to the intervention). At this level there
may be a comparison group or area, but it cannot be demonstrated
that it is truly comparable with the intervention group or area.
Level 3 indicates that there is a comparison between two units
(e.g. neighbourhoods) one of which has had the programme and the
other hasn’t (more indicative that if there is any change that it was the
programme that was responsible).
Level 4 compares multiple units or groups with and without the
programme, controlling for other factors, or using comparison units
that show only minor differences from the intervention units (e.g.
taking several places where programmes were implemented and comparing
them with several other comparable areas where they were not).
Level 5 Random assignment of comparable units of analysis to
intervention or comparison group (e.g. take a couple of dozen compa-
rable neighbourhoods, give half of them, randomly chosen, a programme,
and the other half not: the most powerful way of evaluating impact, but
difficult to do in practice).
Table 4.2 illustrates the importance of these elements, showing
the extent to which each level controls for various threats to internal
validity.
The threats are:
Table 4.2 The Maryland Scale of Scientific Methods (b)

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 5034B03A
+ Causal direction: whether the intervention caused the observed

level of crime or the level of crime brought about the interven-
tion.
+ History: The passage of time causing any change rather than the
intervention.
+ Chance factors: Unforeseen events such as a police operation, or an
outbreak of disturbances.
+ Selection bias: Characteristics of the unit of intervention (e.g. a
neighbourhood) which influence the chances of it being selected for
intervention.
It can be seen that Table 4.2 is almost a mirror image of Table 4.1.
The removal of each threat to internal validity increases the
likelihood of a causal link between intervention and crime reduction.
While the Maryland Scale has refined the basic OXO model, it
still considers the random controlled trial to be what is referred to as
the ‘gold standard’ for programme evaluation. Other researchers have
criticized such a purely outcome-oriented approach, and here we
need to more fully explore the relationship between evaluation and
theory.
Evaluation and theory
In Chapter 2 we looked at two different kinds of theory, crimino-

logical theories and epistemology. The same applies in relation to
evaluation: there is evaluation as a way of examining theories about
crime, and there are also theories about evaluation as a methodology.
Looking at evaluation as a methodology, the OXO model and the
Maryland Scale can be seen as essentially positivist in nature (see
Chapter 2), in attempting to apply the methods of the physical
sciences in a criminological context. Other researchers (and we
would include ourselves in this) have criticized such a purely
outcome-oriented approach. One criticism is that such a model is
‘black box’ research. That is, it is principally interested in the end
products, paying little attention to the processes that take place
during the course of intervention which make the relationship
between intervention and outcome comprehensible. There is more to
it than whether something ‘works’ or not. We also need to know
how and why something has the effects that it does in order to
generalize the results, otherwise we are disregarding the theoretical
basis for intervention, which also requires explanation. Two crimi-
nological researchers in particular, Ray Pawson and Nick Tilley, have
gone even further and during the 1990s mounted a wholesale assault
on the OXO model, advancing a different model based on scientific
realism. This became known as the ‘paradigm wars’, involving

JOBNAME: Mcgraw−crowsemmens PAGE: 7 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 5709ACE4
exchanges in academic journals between Pawson and Tilley, on the

one hand, and Trevor Bennett and David Farrington of the Institute
of Criminology in Cambridge, on the other.
In an article in the British Journal of Criminology Pawson and
Tilley (1994) argued that the traditional OXO model had resulted in
moribund evaluation. Although they claimed to be raising funda-
mental objections to the traditional model, their main criticism was
essentially the familiar criticism that such an approach fails to
comprehend how a programme works (ibid.: 294), and they submit-
ted to critical scrutiny an evaluation by Trevor Bennett of a police
initiative for reducing fear of crime (Bennett 1991). Pawson and
Tilley recognized this to be a well-executed study in OXO terms,
but said that it failed to explain how police activities might bring
about change in a community (Pawson and Tilley 1994: 297). It was
suggested that what was needed was to understand a programme’s
mechanisms and the context in which it takes place. Pawson and
Tilley then put forward their solution, citing a study by Nick Tilley
of the operation of closed circuit television in car parks as a model
of such an approach. This approach is what they term a ‘scientific
realist’ evaluation. It is explained that for a realist evaluator,
Outcomes (O) are understood and investigated by bringing to
the centre of investigation certain hypotheses about the mecha-
nisms (M) through which the programme seeks to bring about
change, as well as considering the contextual conditions (C),
which are most conducive to that change. (Pawson and Tilley
1994: 300)
The original article was the subject of subsequent exchanges in the

British Journal of Criminology between Pawson and Tilley and Trevor
Bennett (Bennett 1996 and Pawson and Tilley 1996). Bennett
disputed that researchers who use quasi-experimental designs ‘over-
look … mechanisms and contexts’ (Bennett 1996: 568), citing Cook
and Campbell’s work on quasi-experimentation in support (Cook
and Campbell 1979). He also denied that competent evaluators
merely seek associations between treatments and outcomes. In
defence of his own work, Bennett said that he did consider causal
mechanisms and the areas in which the research took place. He
suggested that, although there is a need to refine evaluation
methods, Pawson and Tilley offer ‘little that is new and nothing that
is useful in moving towards this goal’ (Bennett 1996: 572).
The debate continued in the pages of the journal Evaluation
where it focused on a community safety programme developed
initially in the United States, but since implemented in the UK and
elsewhere, Communities That Care (CTC; see Crow et al. 2004 and
Chapter 11 of this book). Farrington recommended the comparison
of experimental and control communities, and taking measures of

JOBNAME: Mcgraw−crowsemmens PAGE: 8 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 4F5B999A
key outcome variables (crime, delinquency, substance abuse and

adolescent problem behaviour) before and after the intervention
package (Farrington 1997). Pawson and Tilley responded by criticiz-
ing Farrington’s design for not sufficiently taking account of the
complexity of the Communities That Care programmes. They
suggested that the design favoured by Farrington would miss the
opportunity to look at ‘what works for whom in what circumstances’
(Pawson and Tilley 1999: 83). Instead they stressed the importance
of studying the theories that underlie a programme such as CTC,
and suggest that a ‘scientific realist’ study would transform the
evaluation ‘so that the test of the programme becomes a test of
whether its theories come to pass’ (ibid.: 85). Accompanying Pawson
and Tilley’s article was a response from Farrington in the same issue
of the journal (1998) suggesting that the main difference was that
whereas he had seen CTC as a ‘well-developed technology waiting
to be tested’, Pawson and Tilley assumed it to be an ‘iterative,
evolving, ill-defined, highly variable procedure that essentially
required further exploratory and developmental work’.
One of the reasons why the debate between Pawson and Tilley,
on the one hand, and Bennett and Farrington, on the other, became
such a prolonged, and indeed heated, one is because it concerns
different philosophical approaches to evaluation, hence the reference
in one of the articles to the ‘paradigm wars’ (Pawson and Tilley
1999: 73). Pawson and Tilley make a distinction between two modes
of causative explanation:
1 ‘successionist’ causal thinking (more frequently referred to as

deterministic causality), which describes a constant conjunction
between events (‘the action of billiard balls is archetypally describ-
able in these terms’; Pawson and Tilley 1994: 293). This is the
approach, they say, of quasi-experimentation.
2 the ‘generative’ conception of causation, which they explain as
describing the transformative potential of phenomena, rather like
gunpowder whose potential to explode inheres in its chemical
composition. This is the approach of the school of thought that
they espouse, ‘scientific realism’.
In his response to the original Pawson and Tilley article, Bennett

denied that quasi-experimentation is necessarily deterministic, and
said that in the social sciences probabilistic causality is generally
accepted to be more appropriate. In their response to Bennett’s
criticism of their article, Pawson and Tilley said that Bennett had
failed to grasp the basic difference between the principles of
successionist and generative causation (Pawson and Tilley 1996:
574), and that while Bennett’s own research may have incorporated
some attempts to describe what was happening alongside his

JOBNAME: Mcgraw−crowsemmens PAGE: 9 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 4EF0BEE9
outcome evaluation, these did not constitute a realist approach to

evaluation (Pawson and Tilley 1996: 576).
A key passage in the exchanges occurred when David Farrington
referred to Pawson and Tilley’s contention that mechanisms provide
the ‘reasons and resources’ to change behaviour. Farrington’s com-
ment is that it is not clear how reasons could be investigated:
Many psychologists are reluctant to ask people to give reasons
for their behaviour, because of the widespread belief that
people have little or no introspective access to their complex
mental processes … Hence, it is not clear that reasons in
particular and verbal reports in general have any validity, which
is why psychologists emphasize observation, experiments, valid-
ity checks, causes, and the scientific study of behaviour.
(Farrington 1998: 207)
This passage epitomizes the difference between the positivist

approach to inquiry, and the interpretive approach. It is therefore
not surprising to find that Pawson and Tilley take issue with this in
their rejoinder:
Suppose a woman locks a door and is asked why she is doing
so. If she says that it is a precaution against intruders, there
seems no reason to us not to take her reply seriously. To be
sure, and as we have argued repeatedly, such reasoning may
have unacknowledged conditions and unintended consequences,
but these do not place the ideas beyond investigation. (Pawson
and Tilley 1998b: 211)
Pawson and Tilley’s approach has been set out in more detail in
their book Realistic Evaluation (1997), which is recommended for
further reading.
These exchanges give the impression of a polarization of views on
evaluation models. There are, however, other perspectives, which are
not so extreme. For example, it has been suggested that the danger
with the OXO model is that it becomes reduced to a purely
technical exercise focused on outcomes, forgetting that evaluation
needs to be seen in a theoretical context (Crow 2000). The OXO
model represents only a small part of the research process referred to
in Chapter 1 (Figure 1.2). What is missing is the fact that all
interventions are essentially a test of a theory. For example, if one is
evaluating an employment training programme for offenders with
the aim of reducing their likelihood of re-offending, then implicit
within this is a theoretical position about the relationship between
employment and criminality. Similarly, a project for drug users
designed to reduce drug-related offending is based on assumptions
about the relationship between drug misuse and crime. Not to

JOBNAME: Mcgraw−crowsemmens PAGE: 10 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 2E2B3F3E
labour the point too much, but a third example might be a project
designed to provide sporting activities for young people at risk of
getting into trouble, which is likely to encapsulate theories about
youthful behaviour and offending. It is worth noting that in
referring to theory we are not necessarily envisaging anything
profound about the nature of crime and criminality; it is more a case
of unravelling the thinking that is informing action. It is also worth
noting that medical trials are also based on theories. Quite often
such theories are not made explicit by the programme itself, or are
poorly articulated, and part of the task of the researcher is to explore
the theoretical assumptions implicit in any initiative. Seen in this
context, therefore, an evaluation is an empirical test of a theoretical
proposition, and the results of any evaluation should be seen not
simply in terms of ‘does it work?’, but ‘what does this tell us about
the theory underpinning the initiative?’ Thus:
Theory ⇒ Intervention ⇒ Outcome ⇒ Modified Theory
The OXO model referred to above should therefore really be located
within the broader research process, outlined in Chapter 1, as shown
in Figure 4.3. Similar observations have been made by Carol Weiss,
who refers to a programme’s ‘theories of change’ (Weiss 1998). She
describes the theory of change as having two components: pro-
gramme theory and implementation theory. Programme theory refers
Figure 4.3 The research process and the OXO model

JOBNAME: Mcgraw−crowsemmens PAGE: 11 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 54BBE638
to ‘the mechanisms that mediate between the delivery (and receipt)

of the program and the emergence of the outcomes of interest’
(ibid.: 57). It is a set of hypotheses on which people build their
plans for a programme, and one of the jobs of the evaluator is to
find out (‘surface’) what these hypotheses are. Note that in referring
to mechanisms Weiss echoes the views of Pawson and Tilley.
Implementation theory refers to the fact that if the programme does
the things it intends to do, then the desired outcomes should occur:
‘Implementation implicitly incorporates a theory about what is
required to translate objectives into ongoing service delivery and
program operation’ (ibid.: 58).
One feature that Pawson and Tilley and others have in
common is their insistence on evaluation consisting of both process
and outcome research, the process research being likely to incorpo-
rate qualitative components intended to explore programme theories
and how programmes are implemented. In an attempt to bring
together the various perspectives on evaluation that developed
between 1994 and 2004 Home Office researchers have produced
what they consider to be an integrated model of evaluation for
criminal justice interventions (Friendship et al. 2004) (Figure 4.4).
This is primarily concerned with the treatment of individual
offenders.
Friendship et al.’s model incorporates some quite specific sugges-
tions for what needs to be done in relation to the various
components of the model, and at this point, having looked at some
of the principles underlying evaluation designs we go on to consider
what criminological evaluation needs to involve in practice.
Doing criminological evaluation
Identifying the aims

From what has been said above it should be apparent that a first
priority for the evaluator is to find out what the aims of any
intervention are. This is important in order to understand what
theories, explicitly or implicitly, underlie the initiative, and to
determine what outcome criteria are relevant. Identifying the aims of
a project may at first appear to be fairlystraightforward, but this is
not always the case. In criminology it is quite likely that an initiative
will feature crime reduction as one of its aims in some form or
other. But this may not be the only objective. For example, training
and education programmes for offenders may hope to reduce the
chances of re-offending, but perhaps the first aim is to get people
qualifications and jobs. It may be that a programme is directed at
individuals, such as drug users, where the initial concern is to reduce
drug consumption, but that the main concern is with reducing levels

JOBNAME: Mcgraw−crowsemmens PAGE: 12 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 1A420C21
Figure 4.4 Friendship et al.s’ integrated model to evaluate the

impact of interventions
of crime in an area. Some types of evaluation cover developments

with quite complex aims: if you were seeking to evaluate the impact
of a new Criminal Justice Act, then a variety of ideas may have gone
into framing the legislation at various stages.
It may also be the case that there are differing views about what
the aims are. For example, a sponsoring agency or management
board may have set out what the aims of a project are in formal
terms, but when you talk to the staff delivering the service on the
ground, they see things quite differently. This is even more likely to
be the case where, as often happens, more than one agency is
involved. One agency may have one agenda, while another agency is

more concerned with other goals. The eventual aims of a programme

may be a compromise that has come about as a result of a series of
negotiations.
What all this says is that the researcher should take nothing for
granted. Evaluation should start with a careful consideration of
where the initiative has come from, who is involved and what each
party hopes to get out of it. This is likely to involve the use of
interviews, and where a large number of parties or people are
involved, the use of questionnaires may be justified. For example, if
a new initiative is being brought in that will affect the way courts
work, then in addition to interviewing key players it may be
instructive to do a survey of the magistrates who sit in a court (or
courts), which asks them what they think the initiative is about (see
Chapter 10 for an example of such a project). It may also be
relevant to identify primary, secondary, and even tertiary aims. Thus,
if a programme of sporting activities has as one of its aims reducing
the likelihood of young people getting into trouble, is this its
primary aim, or something that depends on other things being
achieved that are more important to the project?
Implementation and when to evaluate

Another key element of any evaluation is to determine whether an
initiative is in fact doing what it sets out to do; this is known as
‘programme integrity’. A programme may end up doing something
different from what was intended. It may be that the original
intentions were found to be too ambitious, or could not be
implemented in their original form. This raises the question of when
to evaluate. It is often assumed that evaluation should commence
when an initiative first starts. This certainly has advantages because
the researcher can then observe how the intervention is put into
operation. But the early months, or even years, of a new initiative
can also be a time when it is settling down and finding the best
ways of operating. If outcome evaluation starts too early, it may not
reflect what the programme eventually becomes.
However, another kind of change in implementation can occur
when an initiative has been going for some time, and there is
‘programme drift’. This can come about as a result of changes in the
external environment in which the programme is operating, or
because the key people who originally set up a project have moved
on and new people want to bring in their own ideas – or just
through tedium. New projects attract a lot of attention (not least
from researchers), but after a while what the programme was
originally supposed to do becomes lost.
It is when examining implementation that researchers are most
likely to get drawn into an initiative. They may then influence what

JOBNAME: Mcgraw−crowsemmens PAGE: 14 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 4D88125D
happens because initiatives are modified as a result of their enquir-

ies. Views vary about how acceptable this is. Should the evaluator
stand aloof while possibly holding information about how an
important and expensive project is failing disastrously? This illus-
trates the difference between natural and social sciences, since most
researchers would regard doing nothing as both unethical and
prejudicing the likelihood of learning from the results of the
intervention. Evaluators will often provide feedback at certain stages
in the investigation, and this is defensible. What is most important
is being transparent about the role the researcher plays.
Another aspect of evaluating programme implementation may be
to consider whether the initiative was delivered to an appropriate
target group. It is all too easy for a training project, for example, to
take people who are most likely to boost its success rates rather than
those most in need. Is a programme aimed at young people at risk
of getting into trouble actually getting those most at risk, or is it
getting young people who might not otherwise get into trouble
anyway? This is likely to be particularly important if the initiative
involves referrals being made by other agencies or individuals. It may
be that those agencies are referring to the project those who they
don’t want to deal with themselves, rather than the people at whom
the initiative was aimed.
Outcome criteria
In criminological evaluation the outcome criteria are likely to centre
around a reduction in re-offending in the case of individuals, and a
reduction in crime rates in the case of geographical areas or
collections of people. However, care should be taken to check
whether these are necessarily the appropriate criteria given what an
intervention is trying to do, and what it is actually doing. Many
initiatives, such as employment schemes, community programmes, or
leisure activities have other criteria by which they might be judged
(and on which their funding is based). Even where re-offending or
crime reduction is a relevant criterion, there may be other outcomes
that need to be measured, such as cost, whether the intervention has
reduced the likelihood of imprisonment, or a reduction in concern
about crime in the local community.
It is also worth considering who will be the parties to any
outcomes. It may be offenders, or communities, or perhaps the
victims of crime. A good example of the last outcome criterion
comes from restorative justice interventions. Here an important
outcome measure might be the degree of satisfaction felt by victims.
The criteria by which an intervention is to be evaluated are
largely determined by its aims. But it is not unusual for the criteria
to be redefined by the project so that they are more achievable. For

JOBNAME: Mcgraw−crowsemmens PAGE: 15 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 4EF6FD30
example, having failed to reduce offending a project may be tempted

to fall back on ‘softer’ outcome measures based on changes in
attitudes or perceptions rather than behaviour. This may occur
because of the need to be seen to achieve something in order to
secure continued funding, or simply because those involved feel their
credibility is on the line. Researchers should be on the look-out for
such outcome shifts.
Crime and offending as an outcome

While the possible outcome criteria for an intervention should be
considered with an open mind, crime reduction or reduced offending
is likely to feature as one of the main measures. Re-offending is
usually measured by reconviction information. This is often done by
comparing actual reconvictions with the expected rate of reconvic-
tion for a given group of offenders using the Offender Group
Reconviction Scale (OGRS), which uses known predictors of
re-offending to produce a predicted probability of reconviction. The
problems of measuring crime and re-offending are one of the main
issues in criminology but we will not discuss the difficulties at
length here. However, we will briefly highlight some of the main
considerations as far as evaluation research is concerned (for broad-
based discussions, see Lloyd et al. 1994; Friendship et al. 2002).
First, reconviction information, and crime rates can be unreliable,
depending on factors such as detection and successful conviction.
They can be affected by variations in reporting, and in police
practices and policy, and by recording errors. The use of OGRS for
evaluating reconviction has its limitations because it depends on
prediction data which may be unreliable or not appropriate for
certain groups (Kershaw 1998).
However, simply measuring whether or not offenders are recon-
victed, or crimes rates go up or down are, by themselves, very crude
measures. Looking first at reconvictions, there can be relative
success. Following intervention, offenders who have frequently
re-offended in the past may go for a longer period than they have
previously before being reconvicted. This suggests that the interven-
tion may have had some effect, even if its influence was only
temporary. Someone may also be reconvicted for a less serious
offence than previously, indicating at least a shift from violent
robbery perhaps to shoplifting. The same can be true if you are
looking at whether or not crime has gone down following a crime
prevention initiative; patterns of crime might become less worrying
to residents or businesses if they involve less violence or damage.
Another consideration is the length of follow-up. Where indi-
vidual offenders are concerned, a two-year follow-up period has
traditionally been used. However, this may need to be adjusted to

JOBNAME: Mcgraw−crowsemmens PAGE: 16 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 5516C760
take account of particular circumstances. For example, young people

in their teens tend to re-offend more rapidly than older people, so
changes over six months or a year may need to be examined.
Research on sex offenders has shown that such offenders need to be
followed up for much longer periods than the usual two years
(Ditchfield and Marshall 1990; Marshall 1994; Hedderman and
Sugg 1996). Where you are dealing with crime rates, rather than
measures of reconviction of individual offenders, it is also relevant to
consider over what period changes in offence patterns should be
studied. This is because it is possible that more than one follow-up
is needed in order to see whether an initial reduction in crime rates
has been maintained in the long term.
It is also necessary when using reconvictions as a measure to take
account of ‘false positives’: convictions for offences committed prior
to intervention, which occur following the offence that led to
intervention. In the same way, it is necessary to take account of the
custodial and non-custodial periods ‘at risk’. Periods when someone
is in prison do not count for follow-up purposes.
Because of the difficulties posed by actual reconvictions and crime
data, some researchers have resorted to secondary ‘offence-related’
measures to supplement crime data. In the case of individual
offenders this might include whether or not someone’s general
behaviour or attitudes change, referred to as ‘non-reconviction
benefits’ (Friendship et al. 2004: 13). It might also include the use of
unofficial sources of data such as police intelligence, and information
from probation, social and health services (e.g. Falshaw et al. 2003).
The problem with such measures is that they may be unreliable
indicators of potential behaviour, or based on the subjective judge-
ments of others. In situations where changes in crime rates cannot
be demonstrated, then changes in measures such as the fear of crime
might be substituted (Ditton, 2000, is a useful discussion of the ‘fear
of crime’ as a crime survey measurement). The use of such measures
can be controversial. On the one hand, it might be argued that these
are not true measures of outcome, and rely on subjective perceptions
to obtain a less satisfactory (and perhaps more amenable) result than
it is reasonable to expect. On the other hand they could be defended
on the basis that more objective measures cannot be obtained, or are
also likely to be unreliable, and that more qualitative measures do
have value in their own right. Whether they are justified is likely to
depend on the circumstances of a particular evaluation.
Where interventions with individuals are concerned, it is impor-
tant to consider completion rates. It is not unusual for participants
in a programme to drop out before the end. To take one instance,
this has been found to be the case for Drug Treatment and Testing
Orders (DTTOs; Hough et al. 2003). What is important here is
that those who drop out may be different from those who complete

the programme in certain ways. For example, they may be the most
recalcitrant individuals who are most likely to re-offend, and
therefore the results are not a true test of the programme. The best
that can be said is that the intervention works with the kind of
people who complete it – which rather limits its applicability.
Finally, do we know what happened – why a project did or did
not work? This is especially important where a programme consists
of several different kinds of intervention, which we consider next.
Complex evaluations
We have been concerned so far with the evaluation of a single
intervention, which might be a single treatment programme, a crime
prevention initiative in one area, or a community-based project.
However, evaluation can be more complex than this. For a start, an
intervention might be what is referred to as ‘multi-modal’, consisting
of several components. Schemes for ex-prisoners may need to
address several needs, such as accommodation, literacy and social
re-integration. Research has suggested that reducing re-offending is
best done, not by addressing offenders needs one by one, but by
looking at the ways in which their needs are inter-related, requiring
an integrated approach which may involve several agencies (Harper
and Chitty 2004).
Sometimes there may be a number of interventions which are part
of a broader programme. This was the case with the UK Safer
Cities Programme which ran from 1988 to 1995, costing £30
million. Just over 500 schemes were set up to prevent domestic
burglary. Most schemes upgraded physical security in British cities,
although some mounted community-oriented initiatives as well. The
schemes usually centred on local neighbourhoods or estates, and an
evaluation of the programme reported results based on nearly 300
schemes (Ekblom et al. 1996).
The evaluation strategy involved comparing changes in the risk of
burglary in a number of areas with differing levels of anti-burglary
action, against areas with no action. The data collection operation
was large-scale and complex. The amount of input (‘scheme inten-
sity’) was measured by the amount spent per household in the
scheme area. Outcome was measured by means of before and after
surveys in 400 high crime neighbourhoods, and local police statistics
before and after implementation. Evidence from the surveys and
police figures showed that Safer Cities reduced the risk of burglary
in the areas covered (Table 4.3).
The researchers then had to consider whether there were expla-
nations other than the impact of Safer Cities intervention that
might be responsible for the observed results. These included
selection effects (whether areas with temporarily high crime levels

JOBNAME: Mcgraw−crowsemmens PAGE: 18 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 31706BDA
Table 4.3 Burglary prevalence by intensity of burglary action
had been targeted, which would have resulted in a downturn in

crime regardless of whether the action itself worked), the influence
of other Safer Cities activities going on to tackle other problems in
the areas, and the influence of action other than the Safer Cities
programme in the areas. A statistical technique known as multi-level
modelling4 was used to analyse the effect of background trends in
crime and demographic differences across the areas and between
survey respondents. The study also took account of cost-effectiveness
and displacement effects.
As a result of these further analyses it was concluded that the
schemes’ impact had two distinct components:
+ setting up any action reduced burglary risks;
+ the more intense the action, the greater the additional decrease in
burglary risk.
In a different way a number of evaluation studies have been
generated by the UK Government’s programme to discover ‘What
Works?’ in reducing offending. The ‘What Works?’ movement
developed in the late 1980s and 1990s as a reaction to the ‘Nothing
Works’ doctrine which had dominated criminology and criminal
justice since the mid-1970s. The ‘What Works?’ agenda was taken
up by an incoming Labour administration after 1997, resulting in
numerous evaluation studies. Although these studies were not part
of a single co-ordinated piece of research, their findings have been
cumulative and have tended to focus on the use of cognitive
behavioural therapy (CBT) as a means of changing offenders’

JOBNAME: Mcgraw−crowsemmens PAGE: 19 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 56D73CBC
behaviour. This illustrates the way in which evaluations add to, and
become part of, developing ideas and practice in criminology, again
making the point that evaluation is about examining the theories
that underlie interventions. (For further information about the
‘What Works?’ programme see Crow 2001, Chapter 4; Mair 2004,
especially the first couple of chapters; and Raynor and Robinson
2005, Chapter 6.)
The politics of evaluation

Evaluation often takes place in a political context, of which the
researcher needs to be aware. This can range from ‘politics’ on a
small scale, to the larger-scale politics of government programmes.
Even when you are evaluating a local community-based project, its
future is of great importance to a number of people and the
community itself, and this means that the evaluator may be
conscious of the aspirations and the jobs that might depend on him
or her. When evaluating larger-scale programmes there is a danger
that evaluation may be used to justify policies. In both instances
there can be pressure on evaluators. It is therefore worth noting that
however ‘scientific’ and ‘objective’ evaluation research is supposed to
be, power relationships are involved, and the evaluator is part of this.
It is not just a matter of producing the desired results. Practitioners
and policy-makers work to demanding time frames, whereas a final
evaluation is likely to take some time for many of the reasons
outlined earlier (allowing time for an intervention to settle down,
doing follow-up, etc.). The result may be that by the time the results
are produced the political agenda has shifted, or the environment in
which the intervention was introduced has changed.
Evaluation research can also affect the way criminological research
progresses. Evaluation is one of the main sources of funding
available to criminologists. This means that while criminologists
have the opportunity to influence social policy, the criminological
agenda can be affected by what needs to be evaluated. The arrival of
a new government in the UK in 1997 had a big impact on the
amount of funding available for criminological research. However, it
was of a particular kind. The Government was particularly keen to
promote its ‘evidence-led’, ‘What Works?’ programme (referred to
above), and this meant that from the late 1990s into the early years
of the twenty-first century much of the funding available from the
Home Office was for the investigation of programmes for treating
individual offenders, and in particular leant towards psychologically-
based initiatives such as CBT. Criminological researchers needing to
think about getting funds in order to satisfy the needs of their
university’s Research Assessment Exercise (RAE) rating may be
more inclined to take up such funding than pursue studies in other

JOBNAME: Mcgraw−crowsemmens PAGE: 20 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 50C705BF
Box 4.1 The Future for Youth Project

Among many of the problems faced by Youngtown is how to
counteract the reputation that the area has for being a focus of drug
dealing. Whether this is true or not, the fact remains that for many
young people being part of this culture is irresistible. For many it
begins as an adventure and excitement, with little awareness of the
dangers.
The project has targeted a core group of such young people and has
a room in a local community centre which is being converted into
an open school forum area. The objective will be to create a space
where workers can work with the young people in a personalized
and intensive way. The project has links with local statutory and
voluntary agencies.
Central to the project’s way of working is linking the young people
with a group of volunteers who are themselves part of the
community. The volunteers will act as mentors, and will have
received training to enable them to fulfil this role. Some of the
volunteers are ex-offenders and ex-drug users.
The key features of the project are:
+ it focuses on an area of poverty;
+ it involves the local community and other agencies in developing
a community-based approach to community safety;
+ it attempts to reduce crime and the fear of crime in the area;
+ it promotes an alternative lifestyle for young people;
+ it lays the foundation for further developments;
+ it targets young people who are vulnerable to becoming part of
an ever-growing drug scene in Youngtown.
criminological areas, which might be more questioning of current

policy initiatives. All this makes evaluation one of the most difficult
and demanding kinds of research design that a criminologist is likely
to encounter.
We will end this chapter by giving an example of the issues raised

by the evaluation of a fairly small-scale project. It is based on a
real-life project proposal, which we will call the ‘Future for Youth
Project’. The proposal reads as follows:
Faced with such a project being implemented, how might one set
about evaluating it? First, what are the aims? Well, there seem to be
two main aims. One is to reduce the likelihood of young people
becoming involved in drug misuse and the other is, as a result, to
reduce drug dealing in the area, so there are aims relating to
individuals and to the community as a whole. However, this
interpretation needs to be checked with those running the project. If

JOBNAME: Mcgraw−crowsemmens PAGE: 21 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 4FB49418
one were proposing to evaluate the project, one would want to talk
to those running the project about what they are doing. The very act
of doing this is likely to result in some things being changed.
Perhaps as a result of asking about the project’s aims, the organizers
decide to clarify or modify them.
The inputs are likely to include the resources needed to staff and
run the project, the volunteers, and the activities that are carried out.
As far as the processes are concerned, ‘The objective will be to
create a space where workers can work with the young people in a
personalized and intensive way’. However, exactly what these activi-
ties are, and how they are supposed to affect involvement in drug
use are unclear. Will they divert young people from drug use by
providing them with something more interesting to do? Will they
occupy time that might otherwise be taken up hanging around
locations where drugs are dealt? Will the volunteers act as potential
role models for young people? Or perhaps there is some idea that all
these things will happen? These are all hypotheses waiting to be
explored. One might also want to know more about the young
people themselves. How are they to be identified and recruited to
the project? Furthermore, the implementation of the project needs to
be monitored by collecting information of various kinds regarding
both the inputs, and the outputs, such as the number of young
people attending, and how frequently, and what activities take place
when. The involvement, recruitment and monitoring of volunteers
may also need some consideration.
However, none of this constitutes evaluation. In addition, outcome
measures also need to be identified. What criteria are relevant? Drug
arrests? Other crimes? Residents’ perceptions and fear of crime?
These might involve interviews with young people regarding their
activities, police data on drug arrests, and before and after surveys of
local residents about their perceptions. In order to attribute any
changes in these measures to the project, some comparisons need to
be made. These might involve comparisons with other young people
and/or with another area. We are not saying that there is a single
‘right’ way of doing such an evaluation; much will depend on the
specific circumstances. But the proposal above is not dissimilar to
many others that have been developed, and we have highlighted
some of the issues that need to be considered, and would once again
make the point that research is often a case of balancing a number
of factors.
Further reading
The literature on evaluation has become a big market in recent

years, but most of it deals with evaluation designs in general.

JOBNAME: Mcgraw−crowsemmens PAGE: 22 SESS: 27 OUTPUT: Tue Aug 28 15:47:02 2007 SUM: 21F10FCB
Regarding its use in criminology, we have already mentioned Pawson

and Tilley’s (1997) book, Realistic Evaluation. Chapter 1 of the book
is entitled ‘A History of Evaluation in 281⁄2 Pages’, and it is just
that, although it has to be said, reflecting their perspective. It is also
worth reading Chapter 1 of the Home Office Research Study 291,
The Impact of Corrections on Re-Offending: A Review of ‘What Works’
(Friendship et al. 2004). Although primarily concerned with the
substantive issue of what policies and practices are most effective in
dealing with offenders, the first chapter looks at the methods used
for assessing that evidence, with a particular Home Office approach
to such matters. There is also a good collection edited by George
Mair on Evaluating the Effectiveness of Community Penalties (1997).
Carol Weiss’s book, Evaluation: Methods for Studying Programs and
Policies (1998), on theories of change is also recommended; Chapter
3 is a useful summary of her model.

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 120448E6
Part II
Collecting and analysing
material

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 043BF382

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 4439CB98
Introduction
We gain our knowledge about crime through the processes of

acquiring and interpreting different types of information from a
range of different sources including personal experience, gossip,
political rhetoric, scientific evidence and the media – this is as true
for ordinary members of the public as it is for us as researchers!
In Part II of the book, we will be exploring the different methods
we use to build knowledge about crime and we will do this by
focusing separately on how information is gathered and analysed by
researchers through ‘reading’, ‘seeing’ and ‘talking/listening’ about
crime. When we begin our work as researchers, a lot of our
authoritative knowledge about crime is derived from things we have
read, and that is an obvious starting point for this section of the
book. In this chapter, we are going to consider methods of research
which involve the use of research material that has been documented
and made available by other people. This material might be
produced expressly for the purpose of building knowledge about
crime (for example, studies, reports, statistics, and books produced by
academics or government departments) but we also wish to consider
how we might use material which is produced for other purposes
(e.g. news, entertainment) which may, nevertheless, contribute to our
understanding of crime.
There are three different approaches to research by reading that
we cover in this chapter. We begin the chapter by thinking about
how we utilize the material that has been produced expressly for the
purpose of building knowledge about crime. First, the process of
reviewing the literature in an area of study. Second, the process of
reading and using other people’s research data. Finally, we will
conclude the chapter by considering how documents which have not
been specifically produced for research purposes, such as newspaper
reports or historical documents, may be analysed through a method
called content analysis.

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 540CD2F6
Reading and reviewing the literature critically
It could be said that the most important part of any research project
is the literature review. Indeed, it would not be too strong for us to
state from the outset that it is the responsibility of any researcher to
have a very good understanding of the literature which exists in
his/her field of study before embarking on a research project.
However, although the literature review is the essential first step in
most research projects, it is important to recognize that it is not
merely a preliminary exercise in research design. It may very well be
necessary to consult and expand the literature review at various
points in the research process – before a topic is selected, during the
process of designing a research question, at the stage of methodol-
ogy design and during data analysis. As researchers, we also need to
be mindful of the possibility that new material may be published
during the life of a project which overlaps, contradicts or steers the
direction of the work in progress.
Perhaps it is worth pausing a moment to consider what exactly we
mean by ‘literature’ in the research context? Put very simply, most of
the work used by researchers is published in the form of books,
reports and articles in academic journals. However, there is an
additional type of source which O’Leary (2004) calls ‘grey literature’
and includes sources such as unpublished conference papers, news-
paper articles, and student theses. For the criminological researcher,
the literature base is particularly broad due to the multi-disciplinary
and policy-driven nature of our discipline. We must become familiar
with relevant research produced by a range of people including
academic researchers, government departments, public and private
sector agencies and interest groups and it is also often necessary to
delve into literature from other disciplines such as sociology, psy-
chology, law and geography.
The potential range of sources, then, is extremely wide and the
process of searching for and reviewing the literature can be a
particularly challenging and time-consuming one. A researcher will
have to invest a significant amount of time and effort in reading and
critically analysing the work of other researchers in the field and this
must be factored into the research design. It is worth pointing out
that this can be an isolating task as a lot of time is spent working
alone in libraries and/or at a computer and so it is important to
work some variety into the working day and to make efforts to talk
to colleagues about recent discoveries or research problems. It also
requires a good level of organization as several different lines of
enquiry will need to be followed simultaneously in order to build a
complete picture of the work in an area.
There are different ways of approaching a literature review and it
is important to have a clear strategy from the outset. Often, the

JOBNAME: Mcgraw−crowsemmens PAGE: 5 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 57A95903
Researching by reading 83
most difficult part is getting started, but once you have identified
some initial sources, new sources will start to emerge and accumulate
(rather like a snowball effect). It may be necessary to adopt several
different search strategies to start the ball rolling. For example, a
good way to identify journal articles is to use a bibliographical
database (examples include the ISI Web of Knowledge and the
Social Sciences Citation Index), many of which are now available
online, whereas for books, it may be better to search library
catalogues and bookshop databases for books in the area. Further-
more, in order to find reports and studies which are published by
public and private sector agencies and interest groups, the best
approach might be to search the online sites of these organizations
or, if necessary, to use an internet search engine. The key to success
in all of these approaches to searching is to develop a good set of
search terms which can be derived from the topic, subtopics, key
concepts, theories and named authors. So, for example, if you were
to be searching for literature relating to methods used by credit card
fraudsters, your search terms might start very simply: ‘fraud’ and
‘credit card’. Then, as you get to know the literature in the area, you
would discover that the terms ‘payment card’ and ‘plastic card’ are
often used and you would repeat your searches using these new
terms. As the search continues, yet more search terms will emerge –
continuing our example here, the concepts of ‘identity fraud’,
‘card-not-present fraud’, ‘skimming’ and ‘counterfeit’ will become
apparent as methods of credit card fraud. It may also be the case
that theories can be identified or that authors’ names can be used as
search terms (for example ‘opportunity theory’ or ‘Levi’).
Once each item of literature has been identified and found, the
researcher must manage and evaluate the information. As we said
earlier, it is usual for the researcher to accumulate a large literature
collection for a project and there are two issues to resolve – first,
how to catalogue the sources and, second, how to record the
information contained within them.
In years past, it was necessary to use manual filing systems and
card indexes to build a catalogue or bibliographic database but this
task is made easier now by the use of a computer. There are a
number of good bibliographic computer programmes (such as
Endnote and Reference Manager) which are particularly useful because
they are automatically formated and particularly easy to search, but
many people prefer to use their own personal system of cataloguing.
Whichever approach you use, the most important thing to remem-
ber is that you should ensure that you record a full reference for
each item at the time of acquisition – there is nothing worse than
discovering that you have not recorded a reference and you cannot
remember where or how you found the item (probably many months
ago!). You will need to decide which details are important to record

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 54F7F8C3
(such as author name(s), year of publication, title of article/book,

journal title, page numbers, ISBN/ISSN, http address, date source
was recorded, where it was found/how it was acquired and whether
a copy is held on file) and choose which category of information you
will use to organize the materials (for example, chronologically by
date or alphabetically by author).
Having located and accurately recorded an item in the catalogue,
you can finally sit down and read it! This is perhaps the most
challenging part of the literature review process because it requires
critical analysis. It is, of course, impossible to commit everything
which has been read to memory and therefore it is necessary to keep
some kind of record of what is contained within each source. But it
is neither efficient nor productive to record a simple description of
the content of the source – such descriptions tend to be lengthy yet
mean little when utilized later in time. Instead, it is better to
critically analyse a source and distil its ‘contribution to the field’. By
doing so, the researcher is able to build a meaningful picture of
knowledge in a particular topic and identify gaps or inconsistencies
which can be developed into research questions. However, the
process of critical analysis is not easy and requires confidence and
practice. We do not intend to give any lengthy instructions here on
how to conduct analysis but O’Leary (2004) provides some excellent
guidance, especially for those who are new to research. We also
include at the end of this chapter, a ‘Guide to Dissecting Articles’
which we use when teaching our own students how to critically
evaluate published research, (Appendix on p. 112).
We also think it is necessary to make one final point about
context, which is particularly important for those of us who research
in the areas of crime and crime policy. It is important to step back
sometimes and consider the contexts in which the literature has
evolved. As researchers in crime, we are usually dealing with
contemporary social problems which have arisen from specific
political, economic and legal contexts. An example would be the
debates surrounding ASBOs: the emergent literature in this area is
best understood in the context of the public debates taking place
surrounding binge drinking, knife possession and the ‘respect
agenda’. Similarly, the controversial proposals to introduce identity
cards are best understood in the context of post-9/11 concerns about
terrorism and the situation regarding illegal immigration. We can
imagine that researchers looking back on today’s literature on
anti-social behaviour or identity cards in ten years time might not
fully understand the meaning of its conclusions and fully appreciate
its contribution to criminology and policy if they were not aware of
the wider context.
To conclude, then, the process of finding and reviewing the
literature in an area can be onerous and time-consuming. It should,

JOBNAME: Mcgraw−crowsemmens PAGE: 7 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 599DEAC6
however, be an interesting, stimulating, thought provoking, challeng-

ing and forward-looking exercise. If you find yourself struggling,
Hart (1998) gives excellent guidance on maintaining momentum
and helping to make the process a positive experience.
Analysing other people’s data
In the preceding section, we discussed how published research

(articles, books, reports, etc.) should be analysed in order to build a
critically informed picture of the proposed area of research. Often,
you will be reading and reviewing data which has been analysed by
researchers and presented in a summarized form (there is, after all,
a restriction on the number of points which can be made in a single
publication!). There will be times when you wish that you had access
to the data. Perhaps you want to know more about the relationships
between sets of variables which are of interest to you, or perhaps you
can think of different avenues for analysis than those which are
presented. If the latter case is true, you may be able to envisage an
entire research study which could be carried out using data which
has already been collected by somebody else! In some situations, for
example where the data presented comes from a government
department, major survey series (see below) or a project funded by a
Research Council, it may be possible to get access to other people’s
datasets and actually conduct fresh analysis on the raw data. This
approach is referred to as secondary analysis and is a common
method in criminological research.
Put simply then, secondary analysis involves an ‘independent’
researcher (i.e. who has not been involved in the data collection or
who was not originally envisaged as a primary ‘user’ of the data)
conducting analysis on the data and drawing conclusions from that
data. Usually, the secondary analysis involves a new interpretation of
the data and thus can be a novel and exciting approach to
researching. It is also attractive as a research method for a number
of practical reasons (for a good overview, see Dale et al. 1988). First,
the process of collecting and analysing empirical data is often
time-consuming and expensive, so there is a clear incentive to utilize
data which has already been collected by someone else. More time
can be spent on analysing data, a luxury not always available in
empirical work. Good quality datasets are also increasingly easy to
access. This is certainly the case in the UK, but access to interna-
tional data is also becoming easier. In the UK, we are fortunate to
have the UK Data Archive (based at Essex University) which acts as
‘curator’ of the largest collection of digital data in the social sciences
and humanities in the UK. The archive can be accessed freely by
researchers through its webpages (http://www.data-archive.ac.uk/)

JOBNAME: Mcgraw−crowsemmens PAGE: 8 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 4BF9BFE2
and both qualitative and quantitative datasets are available for

download (sometimes there is a small administrative fee and some
special conditions may be imposed on publication rights).
Secondary analysis of qualitative data

Secondary analysis of qualitative data is becoming more popular,
especially as data is being made more widely available through the
Economic and Social Data Service (ESDS) provided by the UK
Data Archive (UKDA) at Essex University.1 The service holds
digital data collections from purely qualitative and mixed methods
contemporary research, including all relevant studies funded by the
ESRC. General guidance and a dedicated advisory service for data
creators and depositors on research project management, issues of
confidentiality and consent, and data documentation of data for
archiving are provided.
It can be difficult to analyse qualitative data which has been
collected by someone else. As we shall see later (in Chapter 8),
qualitative analysis is not governed by as clear a set of rules and
procedures as is quantitative analysis. Thus, it is essential for the
researcher to develop an analytic strategy of his/her own, ensuring
high standards of reliability and validity throughout the process.
This requires an in-depth knowledge of the project’s methodology
and an understanding of the contexts in which the data were
collected. It is, indeed, argued that secondary analysis of qualitative
data is simply not feasible because it is impossible for someone who
was not part of, say, the interview or observation to understand the
data fully. However, the possibility that a new analyst may bring a
valuable objective perspective to the data, combined with the very
fact that there is so much room for exploration and conceptual
development, seems to outweigh the potential pitfalls (Bryman
2004).
Secondary analysis of quantitative data

The range of quantitative datasets available to the secondary analyst
are more plentiful and consist of both major survey series conducted
by research organizations and official statistics collected and pro-
duced by government agencies. A large proportion of these quanti-
tative datasets are held by the UKDA and are available for secondary
analysis but some are administered by dedicated research institutions
(see below). The data tends to be of particularly high quality, drawn
using well-developed sampling procedures by large, well-established
research organizations. Often, the datasets are repeated (annually or
less often) and may then be suitable for longitudinal analysis or the
analysis of trends.

There are, however, a number of disadvantages too. Because you

have not been involved in the operationalization of the measures, the
collection of the data and the preparation of the dataset, you have to
spend a lot of time understanding the data before you can begin to
analyse it. This is particularly problematic in the case of large
datasets such as the British Crime Survey (see Chapter 12 for an
illustration). There will also be a number of restrictions on what you
can do with the data. You may, for example, have data which
operates at only the individual level (rather than the household, local
or national level). There may also be some key variables missing,
either because they were not originally included or they have been
removed for the purposes of anonymizing the data (for example, the
postcode variables are sometimes removed from datasets which can
be frustrating when you are intending to attempt area-level analysis).
Finally, Arber (2001) makes the excellent point that secondary
analysis requires both sociological imagination and the construction
of theoretically informed research questions.
There are two types of quantitative data which will be of use to
you as a criminological researcher. First, we will briefly consider the
major survey series which are available. Although these are not
dedicated ‘crime surveys’, they tend to contain information about
social characteristics and indicators of social change that criminolo-
gists are interested in (for example, poverty, unemployment, health
and political attitudes). Next, we move on to consider the series of
data collected by the government, information that we usually refer
to as Official Statistics.
Major survey series

There are a number of major surveys and longitudinal studies which
have been carried out for a period of years, which provide a fuller
and more detailed picture on various topics. Some of these are
government surveys and some are funded and carried out by
independent agencies. They include:
+ The OPCS Longitudinal Study – based on records of vital events
relating to 1 per cent of the population of England and Wales
born on each of four dates each year, starting in 1971 (about
500,000 people). It represents a continuous sample of the popu-
lation rather than a sample taken at a single point in time. More
information is available through the Office of National Statistics
(ONS): http://www.statistics.gov.uk/about/data/methodology/
specific/population/LS/
+ The National Child Development Study (NCDS) – a longitudinal
study of all those living in Great Britain born between 3rd and
9th of March 1958 (17,000 people). Originally designed to
investigate perinatal mortality and infant health. To date, there

JOBNAME: Mcgraw−crowsemmens PAGE: 10 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 500EC430
have been seven main follow-ups to monitor the sample’s devel-

opment, the latest in 2004 carried out by the Social Statistics
Research Unit (now known as the Centre for Longitudinal
Studies (CLS)), when the sample was 46 years old. More
information is available at: http://www.cls.ioe.ac.uk/
+ The 1970 British Cohort Study – similar to the NCDS, based on a
cohort of 17,000 people born in the UK between 5th and 11th
April 1970. Also being carried out by SSRU and was last
followed up in 1999/2000. More information is available at:
http://www.cls.ioe.ac.uk/
+ The Millennium Cohort Study – The sample population for the
study was drawn from all live births in the UK over 12 months
from 1 September 2000 in England and Wales and 1 December
2000 in Scotland and Northern Ireland. The focus is to study the
all-important first year of life and potentially resolve many of the
issues about its long-term impact. These include issues of central
policy interest such as the foundations of social capital and
cohesion. There have been three sweeps of the survey to date, the
most recent occurring in 2006 when the children started primary
school. More information is available at: http://www.cls.ioe.ac.uk/
+ Labour Force Survey – a quarterly survey comprising 80,000
people which mainly focus on employment and training activity.
Unlike the monthly claimant count, its definition of unemploy-
ment is based on an internationally recognized definition of
unemployment, and its results are often compared with the
claimant count. More information is available through the Office
of National Statistics at: http://www.statistics.gov.uk/STATBASE/
Source.asp?vlnk=358&More=Y
+ General Household Survey – an annual survey started in 1971
covering about 10,000 households used to monitor the effect of
various areas of government policy over time. More information is
available through the Office of National Statistics at: http://
www.statistics.gov.uk/ssd/surveys/general_household_survey.asp
+ British Social Attitudes Survey – an annual survey charting
continuity and change in British social, economic, political
and moral values in relation to other changes in society. Topics
include newspaper readership, political parties and trust, public
spending, welfare benefits, health care, child care, poverty, the
labour market and the workplace, education, charitable giving, the
countryside, transport and the environment, Europe, economic
prospects, race, religion, civil liberties, immigration, sentencing
and prisons. More information is available through the National
Centre for Social Research at http://www.natcen.ac.uk/natcen/
pages/or_socialattitudes.htm

JOBNAME: Mcgraw−crowsemmens PAGE: 11 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 51C81135
Official Statistics
When we talk about Official Statistics, we are referring to a wide
range of statistics which are produced by government departments
and agencies. The government collects a vast amount of data relating
to its citizens and, if you think about it, most of the major events
and activities in a person’s lifetime are recorded by government
agencies. From the moment a birth is registered, an individual’s
interactions with state agencies such as the NHS, Benefits Agency,
Passport Office, HM Revenue and Customs, Land Registry and
criminal justice system are recorded. This, of course, is increasingly
true in the modern era of powerful electronic data storage and
exchange systems. Each government department collects and pro-
duces statistics on areas within its remit, but since 1996 the Office
for National Statistics (ONS) has been the central government
agency responsible for co-ordinating government statistics. There is
an ONS website (http://www.statistics.gov.uk/) which gives access to
a wide range of government statistics, but it is often better to go to
the website of a particular department to seek out published
information.
It is, of course, the Crime Statistics (a sub-set of the Official
Statistics) which are of most interest to the criminological
researcher. As we saw earlier in this chapter, when you begin to
investigate a specific topic or research idea, it is necessary to build a
picture of what is already known about that issue. It is common to
include an overview of the relevant official statistics in the literature
review and this can be particularly useful for developing the context
and theoretical framework. You might be interested in establishing
answers to the following questions: How much crime is there? Is
crime going up or going down? Are the police doing their job? How
do criminals get punished for their crimes? These are the kinds of
questions of interest to policy-makers, the public, the media and
researchers and the data has been used as a kind of barometer of the
moral health of the nation for more than a century (Maguire 2007).
It is also possible for researchers to gain access to the datasets and
conduct secondary analysis on the data. The British Crime Survey,
for example, is made available through the UKDA and the Criminal
Statistics through the Home Office (http://www.homeoffice.gov.uk/
rds).
The Home Office is currently responsible for the collection and
publication of crime statistics for England and Wales (data for
Scotland and Northern Ireland are collected independently by the
Scottish Executive and the Police Service of Northern Ireland
respectively). The statistics have traditionally been published in three
main statistical series/reports and the strengths and weaknesses of all

JOBNAME: Mcgraw−crowsemmens PAGE: 12 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 54049C0A
three are well documented (for a good overview, see Maguire 2007).
We present a summary here, highlighting the key points of interest.
+ The Criminal Statistics: The Criminal Statistics have been pro-
duced for more than 150 years by the government. They are a
national compilation of records produced at a local level by the
police and the courts, focusing on notifiable offences recorded by
the police, the number of cautions given and the number of
offenders found guilty of specific offences. The main reports and
accompanying data spread sheets are available through the Home
Office webpages (the most up-to-date at the time of writing are
for 2004; http://www.homeoffice.gov.uk/rds/crimstats04.html). For
many decades, the Criminal Statistics were the key measures used
to monitor crime change in the country. In the 1970s, however, a
change in criminological thinking prompted demands for new
kinds of information. Police and court records were heavily
criticized for not being a neutral set of facts about crime but
more of an indicator of institutional needs, and new interests in
victims and situational crime prevention sparked off a whole new
range of empirical studies. Whereas it was once the case that the
Criminal Statistics provided the one official picture of crime, now
they are simply part of a whole ‘kaleidoscope’ (Maguire 2007).
+ The Recorded Crime Statistics: The Recorded Crime Statistics
(RCS) are a national compilation of crimes recorded by the police
which has been administered since 1876. The Home Office lays
down a set of rules which determine how and when crimes are
recorded and are responsible for collating and publishing the
results provided by individual forces. The statistics are useful in
providing overall totals of all crime which is recorded and can
highlight changes in crime problems through trends and compari-
sons of crime types (so, for example, we can see whether auto
crime is on the rise or whether property crime is more common
than violent crime, etc.). However, the RCS have received
significant criticism regarding validity and reliability and you
therefore have to interpret the data with great caution. Maguire
notes that there are three main factors which need to be taken
into consideration:
+ Conceptual consistency: As new legislation and policies are
introduced over the years, definitions of offences are subject to
change. You have to be careful, therefore, that the crime you
are examining now meant the same to people working in, say,
the 1950s. The last major changes were made in 1998 in an
attempt to bring consistency between police forces, implement-
ing a whole new range of recordable offences and changing the
rules for recording multiple offences. The effect was to drasti-
cally boost the recorded crime figures and this has made
comparisons across time very difficult.

+ Consistency in recording practice: Although the Home Office

provides rules which govern the practice of recording (the
National Crime Recording Standard was introduced in 2002),
there is still room for police discretion and therefore a risk of
discrepancies between the approaches taken by individual
forces. It is often argued that police attitudes and practices can
significantly influence the statistics.2 Although changes have
been made to help ensure consistency, it is still a concern that
the police have the ability and, in a growing culture of
performance and managerialism, the incentive to hide some
offences and/or over-record others. It is still the case, then, that
you have to be wary of possible variations within and between
forces.
+ Reporting of crime by the public: We know that people choose
whether to report crimes or not to the police for a range of
reasons including trust, apathy, access or ignorance. The RCS
then are subject to any changes in reporting behaviour and you
should always keep this in mind when interpreting the statis-
tics.
+ The British Crime Survey: The British Crime Survey (BCS) was
introduced in 1982 at a time when official statistics based on the
work of the police and the courts were receiving heavy criticism
and criminologists were becoming increasingly interested in the
victim perspective. The BCS is a national household survey which
seeks to establish whether anyone in the household has been a
victim of crime and, where a victimization has occurred, to gather
information about that experience. Thus it is said to provide a
more accurate picture of crime because it picks up all crimes
which are not reported to the police and/or end up in the courts.
It also includes questions which measure attitudes to the police
and the Criminal Justice System, as well as self-completion
questionnaires for drug taking and domestic violence. In its early
years it was conducted on a four-year cycle, but now is based on
a continuous annual cycle with a sample size of 40,000 respond-
ents per year. The questionnaire has not changed a great deal
since its original format and great emphasis is placed on main-
taining its content to allow for accurate trends analysis. However,
this has actually led to considerable criticism because it has
prevented changes being made to some measures which have been
recognized as poorly constructed or in need of conceptual
development (see Chapter 12 for an illustration). The BCS is also
limited to a relatively small range of crimes and this makes
comparisons with the RCS data quite difficult (according to
Patterson and Thorpe 2006, only 3⁄4 of BCS crimes are compa-
rable with 1⁄2 of the RCS crimes).

JOBNAME: Mcgraw−crowsemmens PAGE: 14 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 526B314B
For the past two decades, the RCS and BCS statistics have been the
figures most often cited by both researchers and the media as
indicators of crime. Traditionally, they were published separately as
distinct but complementary sources of data, but recently the Home
Office took the decision to merge the two in an annual report called
‘Crime in England and Wales’. This decision has been heavily
criticized because it has resulted in what appears to be ‘analysis with
political spin’ and not a neutral presentation of factual data (Maguire
2007). Perhaps the most damning criticism came in the 2006 review
of crime statistics conducted by the Statistics Commission (Statistics
Commission 2006) in which it was recommended that the distinc-
tion between BCS data and Recorded Crime Statistics is reintro-
duced as soon as possible.
Despite the obvious flaws in all three approaches, the Criminal
Statistics, the British Crime Survey and the Recorded Crime
Statistics are still arguably three of the most important sources of
information for criminological researchers. However, the crime sta-
tistics are reviewed periodically, and it is important for those
involved in criminological research to stay abreast of the changes
being implemented. Changes occur in terms of the volume and
range of data collected and the ways in which analysis is conducted
and disseminated. Indeed, Maguire (2007) refers to a ‘data explosion’
which has occurred in the field of crime statistics in recent years,
propelled in part by significant technological advances which have
transformed the ways in which we can store and analyse electronic
data but also by a new culture of managerialism and performance
management in government departments and agencies. He observes
that there is now more data (quantitative and qualitative) being
collected in a range of new areas including unreported and unre-
corded offences, the circumstances of the offence, the geographical
distribution of offences, better development of concepts (such as
street crime), more on hidden crimes and hidden offenders and on
victims and public perceptions. It also seems that there is now a
willingness to think more imaginatively about how crime is counted,
and a move away from ‘legal definitions’ of crime in favour of seeing
crime as a socially constructed phenomenon. The Simmons Report
proposed a ‘national index’ of crime which would focus on the most
serious and high volume crimes, for example. However, it would
appear that we are some way from developing a meaningful ‘index’
of crime.
It is clear, then, that it is possible to become confused when
working with the official crime statistics. Indeed, there has been
growing criticism that the statistics are presented to the public in a
confusing way and it has been suggested that the widespread
misunderstanding about rates and distributions of crime is eroding
public trust (Statistics Commission 2006). How these issues are

JOBNAME: Mcgraw−crowsemmens PAGE: 15 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 54FB25D4
resolved remains to be seen but it is perhaps worth concluding by

reflecting on the six key recommendations made by the Statistics
Commission aimed at improving the availability, usability and
accuracy of official crime statistics:
1 The structure of the Home Office should be changed to separate
the policy-making and information provision/publication func-
tions.3
2 Improved communication with users.
3 Better data at the small area level.
4 Furthering of research opportunities.
5 Improvements to the comparability of data (especially with
Scotland and Northern Ireland).
6 Further work on index measures of ‘total crime’.
Criminological research, policy and politics

Before leaving the topic of official crime statistics, we think it is
important to say something about the relationship between crimino-
logical research and politics. Crime statistics are actually used by a
variety of different people acting in different roles. The government
and local delivery organizations, for example, use statistics to develop
policy on criminal justice and crime reduction and to manage
performance and to communicate about crime with the public. The
public use crime statistics to understand the rationale for govern-
ment policy, to hold local delivery organizations to account and to
build knowledge about other people’s experiences and perceptions of
crime and disorder, and to use this information to assess risk in their
everyday lives. Consequently, there is a close relationship between
public concern about crime, the position of crime on the political
agenda and government investment in data collection (Maguire
2007) and this relationship is one which always needs to be
considered in the context of research.
It is hard to think of a time when criminological research would
not have had some political implications. During the 1970s, for
example, The New Criminology (Taylor et al. 1973) epitomized a
radical critique of a criminology that had previously been seen as an
extension of a criminal justice establishment. Nonetheless, there was
a time, prior to the 1980s when crime and the treatment of
offenders were not a party political issue (see Cavadino et al. 2000
for a fuller history). However, criminal justice policy, and conse-
quently criminological research have become increasingly politicized
in recent decades. Concomitant with this (we do not necessarily
suggest a direct causal link) criminological research has seen debates
emerging about its methods and their relationship to policy. These
have covered two main areas of criminological endeavour: evaluation
and criminal statistics. During the 1990s there was a debate about

JOBNAME: Mcgraw−crowsemmens PAGE: 16 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 5620BCA3
the nature of evaluation, the so-called ‘paradigm wars’ (see Chap-

ter 4). However, this was mainly an argument about different models
of evaluation, and not overtly political. The significance of evaluation
was, however, boosted in the late 1990s when the Labour Govern-
ment declared itself committed to ‘evidence-based’ research. This
commitment later became doubtful when it seemed that some of the
evidence did not provide what the Government desired, leading the
President of the British Society of Criminology, Tim Newburn, to
refer to ‘the rise and fall of evidence-based research’ (address to The
University of East London’s Crime Research Conference, Royal
Statistical Society, 27 November 2006).
One of the most trenchant critics of the Home Office has been
Tim Hope, once a Home Office researcher himself. In the context
of a controversial project in which he was involved, Hope discusses
the role of ‘scientific’ evidence in politics (Hope 2004). The
traditional Enlightenment model, he notes, sees science as some-
thing separate from policy. However, more recent thinkers, notably
Beck in his discussion of the Risk Society (1992), have argued that
as scientific evidence becomes the basis for policy decisions, so
‘science becomes itself a site for politics. … The value of science
ceases to be derived from its methodology alone and is now also to
be derived from its promise of applicability and utility’ (Hope 2004:
290–1). This promise has been augmented in recent years by the
appearance of what is claimed to be a new ‘crime science’, which
seeks to apply techniques in ways that might enhance the work of
criminal justice agencies, and sees research as essentially something
that is intended for policy utilization (ibid.: 291). This ‘reflexive’ role
of science is different from the previously accepted view that
criminology is concerned with understanding and explaining crime
and criminality, and that policy relevance is incidental.
Another element in the way that research links in with politics is
the role played by the Treasury. HM Treasury requires that Govern-
ment departments demonstrate the cost effectiveness of their pro-
grammes, based on guidance issued by the Treasury in the form of
The Green Book, which sets out a framework for project appraisal.
This is supplemented in the context of Government social research
by the Magenta Book of guidance for policy evaluation and analysis.
It is not surprising that a Government should want to demonstrate
value for the money it spends, but it can constrain research in terms
of both scope and timescale. Raynor and Robinson have argued that
The policy of recruiting as many offenders as possible to
programmes to meet Treasury targets has probably tended to
undermine the fit between offenders’ needs and programmes
and contributed to increasing attrition and non-completion,
which in turn reduces the overall impact of the programmes if
non-completers reconvict more. (2005: 116)

JOBNAME: Mcgraw−crowsemmens PAGE: 17 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 50BC4E95
In addition, one of the most common complaints among researchers

is that programmes which need careful planning and development
are rushed in order to produce results before they have really had
time to bed in.
To conclude, then, while the Home Office is not the only arbiter
of criminological research, it does play an important role in both
setting a research agenda and providing a significant amount of
funding. The point we wish to make here is that these matters
should make the criminological researcher aware that research does
take place in a context where values, disagreements and politics
occur, and these need to be considered when research is undertaken.
Analysing documents
The final approach to research through reading that we want to

mention in this chapter is the method of analysing documents
known as content analysis. Content analysis is often used as a
supplementary method in a multi-method strategy. It can involve
the collection and analysis of both quantitative data (for example,
the number of relevant words or paragraphs) and qualitative data
(the nature of language or style of pictures). In the criminological
context, most studies using a content analysis approach have focused
on printed news media and official policy documents.
We do not wish to discuss the technicalities of the process of
content analysis in depth here because we do so later in Chapter 8.
We should, however, say a little about content analysis here as a
method of research by reading. The process begins with the
identification of a focused research question and, where suitable,
developed hypotheses. The next stage is to develop a sampling
strategy. The sampling frame will depend entirely on the nature of
the research question, for example, if the aim is to test the
relationship between crime reporting in the news and perceptions of
crime rates, the sampling frame would be constructed of relevant
national and local newspapers and/or news programmes on television
and/or radio. The sample can be drawn according to a number of
different elements, for example, time, publication, author and refer-
ence to particular events; again, this will be determined by the
nature of the research question. Once the sample has been drawn,
the analysis can begin and the process of development of categories
for analysis and the definition of recording units is described further
in Chapter 8.
Like all methods of research, content analysis has its strengths
and weaknesses. On the plus side, the documents under scrutiny
have usually been produced for another purpose (i.e. not for the
purpose of the research) and content analysis has a clear strength in

being unobtrusive and non-reactive. However, this can also cause

problems for the researcher as the document can be structured in an
unhelpful way, leaving the researcher the tough task of organizing
the material. One of the major challenges is to ensure validity in the
research, especially where the analyst must distinguish between
‘witting evidence’ (information which the author of the document
intended to convey) and ‘unwitting evidence’ (any additional infor-
mation which the analyst can glean through his/her reading of the
document). Finally, content analysis is renowned for being a rather
laborious and time-consuming process although it is true that
computers have eased the labour to some extent.
Conclusion
In this chapter we have introduced a range of different approaches

to research in which you might use documents, publications and
other people’s data. We hope to have illustrated the vast amount of
information available to the criminological researcher and opened
your eyes to the opportunities for incorporating this information
into a research strategy. We have discussed how we utilize material
that has been produced expressly for the purpose of building
knowledge about crime, and considered how we might conduct
original analysis on other people’s data. Also, we have considered the
use of documents which have not been specifically produced for
research purposes, such as newspaper reports or historical docu-
ments. What we hope to have illustrated is that criminological
research is not simply about going out and harvesting new data. In
the majority of research projects, you will need to spend a lot of
time reading and it is important to recognize that this element of
research requires a range of skills and awareness of the wider
contexts in which information about crime is published.
Further reading
O’Leary’s (2004) book, although not specifically tailored to the

criminological researcher, provides some useful information about
the role of the literature review in the research process for students.
Hart’s text (1998) is more detailed and perhaps more appropriate for
practitioners or those conducting more advanced research projects.
In terms of the datasets available for secondary analysis, we would
recommend that you use the weblinks we have provided to gain
access to the most up-to-date information about the major survey
series and official statistics. The UKDA also has a great deal of
useful documents on its website. For those looking for a more

JOBNAME: Mcgraw−crowsemmens PAGE: 19 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 5261EF86
general text, both Bryman (2004) and Arber (2001) provide good
discussions of the secondary analysis of survey data. Dale et al.
(1988) is also a useful text, but a little dated. Maguire’s chapter on
crime statistics in the fourth edition of the Oxford Handbook of
Criminology (2007) is perhaps the most comprehensive overview of
the current state of crime statistics in England and Wales. It is a
‘must-read’ for anyone intending to use crime statistics in research.
Another useful document for anyone interested in the current debate
about the relationship between science and politics is the report by
the House of Commons Science and Technology Committee (2006).
Content analysis is discussed further in Chapter 8, but a compre-
hensive discussion is given by Holsti (1969).
Appendix: A Guide to Dissecting Research Publications
What is the point of the study?

The researcher should convey to the reader concisely why the study
was undertaken, what the field of study was, what the key questions
were which the study was trying to answer. Evaluate whether the
way the study is written up easily enables you to work out its
purpose and summarize it.
The objectives of the research could include:
+ the development of a particular area of theory
+ the collection or accumulation of a new body of information or
data
+ the development of research methods or techniques
+ a contribution to knowledge or understanding
+ a contribution to policy and practice in a particular area
+ a product such as computer software, patents or research facilities.
What kind of theoretical explanation is it using?

Describe briefly the kind of theoretical explanation which underlies
the study. This may be a particular school of criminology or
well-known theory (for example, positivism or subcultural theory) or
it may be a set of ideas or way of viewing the subject of the
empirical work. Remember that not all authors set out clearly what
their theoretical basis is. You will need to evaluate how clearly you
are able to understand the author’s theoretical position and how well
this is set out in the article. Remember that there is always a
theoretical underpinning to empirical work – if only the unstated
predisposition or ideological position of the researcher.
Evaluate as well whether the type of theoretical explanation is
suited to the purpose of the study. What other levels or types of

JOBNAME: Mcgraw−crowsemmens PAGE: 20 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 4C0635DA
explanation would be possible or have been used in previous research

cited in the article? Would any of these have been more suitable
than what the author actually used?
What methodological possibilities present themselves?

Given the purpose of the study and the kind of explanation, think
through for yourself what methods the researchers could have used
(for example, think through the different types of methods covered
in the course and see whether each one could have been helpful for
the topic being researched). Is the method the author actually chose
the most sensible? Why/why not? Do the researchers make it plain
why they chose this method?
How appropriate were the methodological choices related

to doing the research?
Given the choice of that particular method, you need to work
through the methodological choices which the researchers have then
made, at each point (briefly) setting out what the researchers did,
whether that was the most sensible choice and why/why not. If they
don’t tell you what they did, why they did it – and it’s not obvious
– then that’s a negative mark for them; you should comment upon
any lack of information as much as upon what is said in the article.
The particular questions you should ask yourself will vary with the
method chosen, but some likely ones are:
+ What key concepts are being used, are their definitions and
operationalization clear?
+ Was there just one method, or was there method/theoretical
triangulation?
+ What role is the researcher taking (scientist, ‘disinterested’
observer, participant observer, action researcher, analyst of previ-
ous data, etc.)?
+ What population was chosen on which to do the research and
how were people/records sampled? Was the site for the research
the most appropriate/practically feasible one?
+ What are the likely biases leading from those sampling choices of
population, sample and site?
+ How was access achieved?
+ What ethical/legal considerations are involved and how were they
dealt with (particularly focusing on confidentiality, disclosure of
purpose of research, learning about illicit behaviour)?
+ If an interview/questionnaire, were the questions appropriately
phrased? Does it look as though any precautions were taken to
guard against leading questions, socially desirable answers, etc?

JOBNAME: Mcgraw−crowsemmens PAGE: 21 SESS: 30 OUTPUT: Wed Aug 29 11:13:44 2007 SUM: 2B8F1E32
+ Hence, how reliable and valid were the research method(s)

chosen?
How effective were the methods of data analysis?

+ Is it clear how the data were analysed (whether a qualitative or
quantitative approach was used, the kinds of statistics used, etc.)?
+ Is the overall decision on which kind of data analysis was used
appropriate for the research method chosen?
+ Are the variables/topics analysed clearly defined and suitable for
the theoretical questions being asked in the study (for example,
how variables analysed emerged from open-ended questioning)?
The conclusions of the study and its presentation

Describe briefly the conclusions of the study. To what extent do you
think the conclusions drawn fit the results of the analysis (are there
any unjustified leaps of guesswork)?
Do the study’s conclusions relate in the way you would expect to
the original purpose and aims of the study? Has it managed to
create theoretical insights/test its hypotheses, etc?
How effective was the presentation of the study to the reader?
How could it be improved?

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 23 OUTPUT: Tue Aug 28 15:55:35 2007 SUM: 494D777B
6 Researching by looking
Introduction: watch and learn
If you think about it, the process of observation is one which we all
use instinctively. Every day we are learning about the world around
us by looking at our surroundings and interpreting our observations.
You might, for example, notice a sudden increase in traffic around
mid-afternoon and conclude that the problem is caused by parents
collecting their children from a nearby school. Or, you might
observe that two friends are acting frostily towards each other and
realize that they have had an argument. A lot of the time we are
making these kinds of observations subconsciously and using our
interpretations to inform our own decisions or actions.
It is possible to formalize these skills and utilize them in research,
either as a stand-alone method or in combination with other
methods such as those covered in the chapters on ‘talking’ and
‘reading’. Observation is a particularly useful approach in crimino-
logical research because we are so often concerned with understand-
ing behaviour, interactions and processes which both drive criminal
behaviour and facilitate the dispensation of justice. Often, observing
the actions of individuals or a group dynamic is a far better
indication of how people are feeling than asking them to try and
explain it themselves.
Observation is used in criminological research studies but there
are a number of issues which require careful consideration before it
is adopted as an approach. It is extremely time consuming and there
are a number of ethical and philosophical problems to be overcome.
We will discuss these issues in this chapter.
The science of observation? Epistemology and research design
Natural scientists who adhere to the principles of positivism (see

Chapter 2) accept that knowledge about the natural world is best

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 23 OUTPUT: Tue Aug 28 15:55:35 2007 SUM: 578E0FBB
Researching by looking 101
gathered through systematic, objective and repeated observations of

naturally occurring phenomena. So, for example, we might gain
knowledge about the physical properties of glass through a series of
experiments in which we heat a piece of glass to different tempera-
tures and record observed changes in colour and texture. Observa-
tional methods in this context, then, would be formal, structured and
objective and we would control the experiment using tools such as a
stopwatch, a thermometer and a pre-designed table for recording
results. But can such methods be used to study social phenomena?
Social observation methods were originally developed during the
colonial period of the British Empire. British anthropologists were
keen to study and understand the ‘exotic’ cultures which existed in
the newly established colonies and so developed an approach to
research which incorporated methods of observation, the collection
of documents and interviews. This approach became known as
ethnography. Although early anthropologists actually accompanied
scientific expeditions into new territories, an ‘armchair approach’ was
more typical as it became an established procedure. Rather than
working in the field, they would work from their libraries in London
constructing a picture of life in the colonies from notes sent back by
the teams of colonial administrators. The colonial administrators
were instructed what to do by a guidebook, initially published in
1874 (BAAS 1874), which provided the local district commissioners
with a set of questions to put to native informants.
After the colonial period, ethnographic methods were taken in a
new direction by the Chicago School of sociologists working in the
1920s and 1930s in the USA. Importantly, the Chicago School
scholars developed two key elements which added to the British
anthropological approach. First, they introduced an element of
pragmatism to what had previously been a very positivistic approach.
They accepted the view that social life is not a fixed, unalterable
phenomenon (the sort of thing that positivists like to study with
their fixed measuring instruments) but instead was dynamic and
changing. The second element was formalism. This is the idea that
although social relationships will always differ in content from each
other, they nevertheless represent underlying forms which can have
similarities. So the focus in any social setting will be the interactions
between people and not the people themselves.
Thus, the methods used by the Chicago School researchers were
strikingly different from the anthropological approach. Having
shunned the strict principles of positivism, they believed that
understanding cannot be derived from a formal process of testing
hypotheses. Instead, they argued that one must undertake meticulous
and detailed observations, and then generate understanding from the
data. The researchers were, then, encouraged to go out and plot the
contours of the city streets – not with clipboards and questionnaires

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 23 OUTPUT: Tue Aug 28 15:55:35 2007 SUM: 57FE39C3
(as these asked static, fixed questions) but with empty notebooks.
Their aim was to attempt to record and understand change in the
city.
Crucially, the Chicago researchers did not start with a specific
theory which they wanted to test. Instead, they ended up with a
theory after the data had been collected and analysed. This is one of
the key features of ethnography – the theory is generated from data
(rather than collecting data to illustrate a theory). It is usual then for
the researcher to re-formulate the research question as the research
progresses. Hammersley and Atkinson (1983) note, however, that
there is a constant interplay between ‘substantive’ or ‘topical’ theory
(i.e. theory which is developed for specific empirical inquiries) and
‘formal’ or ‘generic’ theory (i.e. broader conceptual theory). An
ethnographic study may move from the formal to the substantive
(see the study of classroom deviance by Hargreaves et al. (1975) in
which labelling theory was taken as a starting point and transformed
into a ‘shopping list’ of issues directly relevant to schools), or the
opposite way (a good example of this is Atkinson’s study of
industrial training units (1981) in which topical, concrete research
questions were related to generic concepts such as labelling and
stigma).
It is important for anyone considering an ethnographic approach
to realize that ethnography and observation are different – it is
possible to carry out an observation study without it being an
ethnography. It is perhaps most helpful to think of ethnography as a
research strategy and observation as a research method. An ethno-
graphic study, then, may incorporate a number of different methods,
including observation, documentary research and interviews. For
example, look at a study by Tomsen (1997) in which the researcher
combined observations in pubs/bars, interviews with bouncers and
the analysis of police data in his study of the culture of drinking
violence in Newcastle, Australia. The defining features of an eth-
nography are its theoretical/epistemological grounding (the genera-
tion of theory from data) and the ‘immersion’ of the researcher into
the social setting for an extended period of time. It is the case then
that a ‘grounded theory’ approach is adopted in which the observer
enters the environment completely open-minded as to what might
emerge. This kind of approach is particularly well suited to cases in
which the researcher intends to explore and understand an unknown
phenomenon or situation. By adopting an open observation
approach, the observer assumes a neutral starting point and opens
his/her mind to the environment and relationships which evolve in
front of him/her. There is no need to make a list of concepts for
observation, nor to construct hypotheses in advance of the observa-
tion. Instead, the concepts, hypotheses and theories emerge as the
process unfolds. The observer is able to determine what information

is noted and recorded from the outset and has complete freedom to
allow the observation process to evolve.
However, as we emphasized earlier, observation is a research
method in its own right and does not have to be carried out as part
of an ethnography. An observation can be integrated into different
types of research strategies, including those with a strong positivistic
leaning. There are, indeed, research contexts which are suited to a
more prescriptive and structured approach to observation. If, for
example, the aims of a research project included the testing of
specific hypotheses and measurement of pre-identified concepts, it
would be necessary for the observer to pre-determine the concepts
which are to be measured and to develop a research tool or
instrument to be used in the observation process. In these kinds of
studies, it is common for the observer to take on the role of a
detached/non-participant observer (see later) and there are a number
of good examples of this approach in criminology, including a street
observation study by Shapland et al. (1994) (discussed later in this
chapter) and observation of the Youth Court by Allen et al. (2000)
(discussed in detail in Chapter 10).
While there is a clear contrast between the open style of
observation found in ethnographic work and studies which use a
structured observation approach, an important feature common to
both approaches is the process of analytic induction. If you recall
from the discussion in Chapter 2, this process begins with the
development of a theoretical framework (in a grounded theory
approach the theoretical framework will emerge as the study
progresses, rather than being constructed at the start). Using this
framework, hypotheses are constructed and are then tested, and
usually redefined as the observation process progresses (Denzin
(1978), see also Cressey’s study of embezzlers (1950)).
Methodological decisions
Sampling
The three main sampling dimensions in observational research are
time, people and setting or context (Hammersley and Atkinson
1983). In terms of time, you have to decide when observations are
most appropriate – what time of day, which days of the week, which
months of the year, which seasons (taking into account events like
Christmas or Bank Holidays), and so on, as well as how long the
period of observation should last (a couple of hours, a whole day or
the duration of an event such as a meeting). When it comes to
people, you have to think in terms of sociological demographics
(gender, age, ethnicity, etc.) but also about the roles or categories
one has identified as existing in the observed group (so, for example,

JOBNAME: Mcgraw−crowsemmens PAGE: 5 SESS: 23 OUTPUT: Tue Aug 28 15:55:35 2007 SUM: 516A7ADD
workers, middle managers and executives). Importantly, roles can be

‘member identified’ or ‘observer identified’ (Loftland 1973). You also
need to consider the different settings or cultural contexts in which
the phenomena under examination might occur. For example, there
may be a difference between the way police officers act when in
public and when in the work’s canteen. Finally, it might be the case
that the phenomenon can be studied from any number of theoretical
angles and you must decide which of these angles to take.
An example may help to illustrate the kinds of sampling decisions
which are necessary in observational studies. Suppose that a research
team studying the experiences of witnesses in the Crown Court plan
to conduct an observation as part of the study. You can imagine that
the sampling design discussions might include consideration of the
following questions:
+ How long should the overall period of observation be?
Weeks, months? What time of year (if it is over Christmas, for
example, how will the holiday period affect the court timetable)?
+ How often will the observer go to court and how long will s/he spend
there?
How many times a week? Which days? Should it be the same
days each week or different days? On Mondays, are the majority
of cases likely to be alcohol-related incidents from the weekend –
how will this be dealt with? Should the observer attend all cases
during the day, or only some? If cases are adjourned, should the
observer return to hear the rest of the case on another day?
+ Which types of witnesses should be selected for observation?
Are we interested in distinguishing between witnesses of different
genders, ages and socio-economic status? If so, how many of each
demographic group should we aim to include? Are there any
other characteristics which might be of interest (perhaps the legal
teams tend to distinguish between reliable and unreliable wit-
nesses or perhaps between confident and nervous/reluctant wit-
nesses)?
+ Where will the observations take place?
Should the observer focus simply on the role of the witness
during the formal proceedings in the courtroom or should s/he
also examine the ways in which other members of the courtroom
behave towards the witness? Should s/he also observe the activi-
ties and interactions outside of the courtroom (for example, in the
waiting area, in interactions with the Witness Service or when
being briefed by the lawyers for the prosecution)?
The answers to many of these questions will, of course, depend upon
the research questions which have been identified previously. Also, a
lot will depend upon the time and resources available to the
researcher.

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 23 OUTPUT: Tue Aug 28 15:55:35 2007 SUM: 596FCC93
Access
As we have seen in previous chapters, one of the major challenges in
criminological research is getting access to the site of interest. This
is especially true for observation studies because it requires a
significant level of intrusion on the part of the researcher – it is not
just like having someone dropping in to do a few interviews and
leaving again, it is a much more invasive process. So, the researcher
must have excellent communication and negotiation skills and must
be able to strike up good relationships with research subjects.
Generally, there are two kinds of setting to which the researcher
can seek access. The first kind of setting is a group or institution
which has a clear organizational structure. This may be an official
organization (such as a prison or a school) or it may be an informal
group (such as a street gang or criminal network). It may be the case
that the observer already has access to this group/organization due
to an existing position (for example, Holdaway, in his (1983) study
of the police, was already in position as a police officer). It is more
likely, however, that the observer will not already be part of the
setting and needs to negotiate access. The second kind of setting is
a public environment in which the observer can access freely. Usually
it will not be necessary for the observer to seek permission to
infiltrate but there may be practical and ethical issues to consider.
The process of gaining access to an organizational/group setting
often involves seeking permission from ‘gatekeepers’. Gatekeepers are
individuals (or sometimes committees) who have a formal responsi-
bility to control the boundaries of the setting, particularly where that
setting is an organization or institution of some kind (for example,
a police force, government department or a prison). Where the
setting does not have an official organizational structure, an ‘infor-
mal sponsor’ may be used instead of a gatekeeper. A sponsor is
usually someone who is part of the setting and who is prepared to
brief, train and vouch for the observer (see, for example,
Whyte’s 1981 study of street corner society in an Italian slum in
which a character called Doc famously coached the researcher in
order to help him to blend in). This person should be contrasted
with the ‘gatekeeper’ who is a more formal guardian of the
boundaries (Hammersley and Atkinson 1983). Access can even be
hard in public space settings because the observer still has to
somehow account for his/her loitering. In some very closed situa-
tions, it may even be necessary for the researcher to negotiate access
with the gatekeeper and to secure a relationship with an informal
sponsor.
It is essential for the researcher to define a good working
relationship with the gatekeeper. When approaching a gatekeeper,
s/he needs to think carefully about how s/he describes the proposed
hypotheses since the gatekeeper is likely to feel responsible for

JOBNAME: Mcgraw−crowsemmens PAGE: 7 SESS: 23 OUTPUT: Tue Aug 28 15:55:35 2007 SUM: 563E97B9
protecting his/her institution and having it portrayed positively. In

addition to appropriately presenting the aims of the study, the
researcher has to think also about how the gatekeeper perceives the
role of the researcher him/herself. Depending on the situation, s/he
will be perceived as either an ‘expert’ (who is able to sort a problem
out) or a ‘critic’ (there to evaluate) and either perception needs
careful management (Hammersley and Atkinson 1983).
Inevitably, though, however good the relationship between the
gatekeeper and the researcher is, the gatekeeper will have a signifi-
cant hold over the direction of the research and complete independ-
ence for the observer is unlikely. This may actually be a problem for
some researchers, especially where the setting is closed or where the
observer needs to uncover information which is particularly sensitive.
There is, therefore, a possibility that the researcher may choose to
avoid these problems by resorting to covert or secret research. Of
course, this means that the researcher must deceive all gatekeepers
and participants within the setting and, as well as raising obvious
ethical questions, this kind of approach can be very stressful for the
researcher. It is, however, an approach which has been adopted in
criminological research but is not now common. Holdaway (1983),
in his study of the police, opted to carry out covert research, arguing
that to misrepresent his research would be dishonest and to seek
permission would be unrealistic as it would be refused. A completely
covert approach such as this is obviously highly controversial,
especially in the modern climate of formalized research ethics
procedures and not a decision to be taken lightly by the researcher.
It does, however, lie at the extreme end of the deception scale and
it is more common for the researcher to strive to achieve a balance
between presenting any details which may be perceived as contro-
versial or threatening in a positive light while avoiding any unjusti-
fiable deception when negotiating with gatekeepers.
Once you have gained access, you then have to actually enter the
setting. This can be difficult because you often have to immerse
yourself into an already established group context. There are three
main challenges to overcome in this regard. First, members of the
group are likely to be wary of a stranger who they know to be
‘researching’ and you have to think carefully about disclosing the
nature of the research being undertaken. If subjects are unfamiliar
with the whole notion of social research, they are likely to see the
researcher as an ‘official’ and behave accordingly. If they are,
however, familiar with the research process, they may feel defensive
(depending on the critic/expert roles described above). Second, the
social characteristics of the observer, especially gender, age and
ethnicity can also have an effect on the way in which s/he is
perceived. Third, you must think carefully before interacting too
‘honestly’, it may be necessary to avoid sharing your personal beliefs.

JOBNAME: Mcgraw−crowsemmens PAGE: 8 SESS: 23 OUTPUT: Tue Aug 28 15:55:35 2007 SUM: 53ABB793
The skilled observer, then, will be adept at what Hammersley and

Atkinson (1983) call ‘impression management’ – dressing and
carrying yourself like the subjects, without taking it so far that you
appear completely uncomfortable and stand out like a ‘sore thumb’.
To conclude, access is a serious practical issue in observational
research and there are a number of important decisions to be made.
Interestingly, the process of obtaining and negotiating access can
give a valuable insight into the social organization of the group or
setting and it should form part of the observational process (rather
than being seen merely as a preliminary administrative procedure).
The role of the observer

Perhaps the most important decision you need to make in an
observation study is what role the observer will take during the
observation. Broadly speaking, there are two types of role: one where
the observer is a participant in the group and setting and one where
the observer is not a participant and is detached from the setting.
Robson notes that the key difference between the two roles is that
where the non-participant observer typically uses an observation
instrument of some kind, or where the participant observer is the
instrument (2002: 313).
The choice will depend upon a number of things including the
nature of the research question and the epistemological approach
adopted by the researcher. In ethnographic work, the participant role
is the norm. In other situations, especially where the avoidance of
‘observer effect’ is paramount, a non-participant role is adopted. The
observer will have decided whether to adopt a covert or overt role.
As we saw earlier, the decision to adopt a covert role is not one to
be taken lightly since there are clear ethical issues. The advantage is
that no one knows they are being observed and are therefore more
likely to act naturally.
Although it is common to see the simple distinction between
participant and non-participant observation, in reality, there exists a
spectrum of observational roles which lie between the two extremes
of full participation and complete detachment. Gold’s scheme of
classification of participant observer roles is often cited in textbooks
as the most useful analysis of observational roles (1958). Within it,
he identifies four key roles:
1 Complete participant. The complete participant is a fully partici-
pating member of the setting and his/her role as an observer is
not known to the other members of the group. This, of course, is
easier where the observer has an existing role in the group. A
good example is Holdaway’s work referred to earlier in the
chapter (Holdaway 1983). In his study, Holdaway was already a

JOBNAME: Mcgraw−crowsemmens PAGE: 9 SESS: 23 OUTPUT: Tue Aug 28 15:55:35 2007 SUM: 54E7004D
police officer and was able to conduct his research without his
colleagues knowing about it. The advantage here is that the
observer is able to fully immerse him/herself in the ‘true’ social
setting and does not have to worry about the effects that the
research might have on the behaviour of group members. While,
however, this might sound like the ‘ideal’ research environment, it
does require a high level of deception and this can be an
extremely stressful experience for the observer, both mentally and
physically. There is also a danger of ‘going native’ – a term which
is used to describe the situation where a researcher becomes so
emotionally involved in the setting that s/he is unable to retain
his/her sense of objectivity, thus threatening the validity of the
research. Examples of this are not easy to find in the literature
(after all, if the researcher has ‘gone native’ it is unlikely that the
research would be published!) but examples can be found of
researchers struggling to maintain their objectivity (see, for exam-
ple, Hobbs’ (1988) study in which he struggled to maintain his
role as ‘researcher’ during pub visits with the entrepreneurs he was
studying).
2 Participant-as-observer. In this role, the observer fully participates
in the group in the same way as the complete participant but here
the other group members are aware that s/he is carrying out
research. It is not necessary for the group members to know what
the researcher is studying, in fact, it is often preferable for the
topic of observation to be kept secret. An example of this role can
be found in Ditton’s study of fiddling and pilfering in a bakery in
which his co-workers knew he was a research student but they
did not know that he was observing their dishonest behaviour
specifically (Ditton 1977). Had they known, they may have
changed their behaviour.
3 Observer-as-participant. The observer-as-participant does not
actually participate in the group activities. S/he can be present in
the setting but does not fully participate as a functioning group
member. As Bryman (2004) notes, in this role the researcher is
effectively acting as an interviewer. Since s/he does not have a
role to play in the setting itself, s/he must ask people questions
about why they act in certain ways or make certain decisions as
those acts/decisions occur. An example is Punch’s research on
police work in Amsterdam. Since he was not a trained police
officer, he was not able to participate fully in police activities, but
he was able to accompany officers on their patrols and ask them
questions (Punch 1979). However, Robson (2002) questions
whether anyone who interacts with group members in this way
can really be described as ‘not taking part’ in the setting, and
therefore one must be careful to take this into account when
assessing the validity of the results.

4 Complete observer. The complete observer lies at the opposite end

of the spectrum to the complete participant. S/he is completely
detached from the group and does not interact with the group
members in any way. A classic example is Loftland’s study of
public behaviour in the street through a window (Loftland 1973).
An interesting modern approach to this role was taken by
Sampson and Raudenbush (1999) who observed disorder on the
streets of Chicago by driving slowly and systematically through
the streets, discreetly filming their surroundings with two cameras
attached to each side of the car. They were able to collect and
analyse their data without intruding into the settings under
observation.
There are clearly advantages and disadvantages to all four roles and
the researcher must consider these in detail when making the
decision as to which role to adopt. Interestingly, Hammersley and
Atkinson (1983) note that the complete participant and complete
observer roles share some advantages and disadvantages, even though
they lie at different ends of the spectrum. Both approaches minimize
reactivity as the researcher does not interact with the subjects in
his/her role as a researcher. Yet, direct questioning is improbable so
the researcher must rely entirely upon his/her own interpretations of
the observed behaviour, conversations and events. They conclude
that it can be difficult to generate and test theory in a rigorous way
when a role at the extreme ends of the spectrum is adopted. They
also question, however, whether there is a meaningful distinction
between the participant as observer and observer as participant and
this is a valid concern.
Data collection
In the previous sections, we have discussed the various epistemologi-

cal, methodological and ethical decisions which need to be made
before an observation study is undertaken and highlighted the
importance of a well-thought-out research strategy. Once these
decisions have been made, the observation may finally begin. There
is, however, one final issue to be resolved – how will the data be
recorded and analysed? The answer to this question will depend
upon the nature of the observation (structured or open) and the
nature of the data (quantitative and/or qualitative). In all observa-
tional work, it is essential for the observer to regularly record his/her
observations because it is extremely difficult to reflect back on an
event days, weeks or months later and recall all of the important
occurrences. But there are different methods of recording data and
these are described below.

JOBNAME: Mcgraw−crowsemmens PAGE: 11 SESS: 23 OUTPUT: Tue Aug 28 15:55:35 2007 SUM: 529BC4C9
Open/unstructured observation studies

In unstructured, open observation studies, the observer’s essential
tool is the notebook. Note-taking is a skill which the observer must
perfect and there is a careful balance to be struck between too many
and too few notes. On the one hand, as Hammersley and Atkinson
note, (1983: 146)
[A] research project can be as well organised and as theoreti-
cally well informed as you like, but with inadequate note
taking, the experience will be like using an expensive camera
with poor quality film. In both cases, the resolution will prove
unsatisfactory and the results will be poor. Only foggy pictures
result.
It is, however, impossible to record everything. This might seem an

obvious statement, but it is an important point. Try to conduct a
10-minute observation of a television programme (soap operas or
dramas work best) in which you record everything you see – the
setting, interactions, and events. Use a stopwatch to keep track of
the time and aim to record your observations in a notebook. At the
end of the 10-minute period, you are likely to feel exhausted and
your notes are probably lengthy and confused. Ask yourself if your
notes would make any sense to you six months from now.
Most experienced observers will agree that the best time to record
events is as soon as possible after the action/event has occurred, even
if things are not immediately understood, it is important to write
them down. If the observation is covert, this is of course not easy to
achieve as suspicions are sure to be raised by the incessant scribbler
(unless writing is part of the role being adopted, perhaps). Although
open note-taking is easier in an overt situation, it is still necessary to
make sure that note-taking does not appear odd or distracting in the
setting.
What should be recorded and what should be omitted from notes
will depend upon the study in question, but there are general
guidelines to follow. Broadly speaking, fieldnotes should contain
concrete descriptions of both social processes and contexts. Robson
(2002) suggests that in observation there are different types of
question for the observer to raise: what, how and why. The what
factors are simple facts and contexts which may be recorded during
the observation but can also be found out from other evidence (e.g.
documents). So, for example, if we were to conduct an observation
in a courtroom, the what questions might include a description of
the court personnel (the names of judges and barristers), what cases
are listed to be heard that day or how many witnesses are called
during a specific case. The how questions relate to interactions and
events which occur during the observation process. To use the

JOBNAME: Mcgraw−crowsemmens PAGE: 12 SESS: 23 OUTPUT: Tue Aug 28 15:55:35 2007 SUM: 5496F04C
courtroom example again, we might then ask how the judge

addresses the defendant (words used, tone adopted, facial expression)
or how barristers present a legal argument (using cases and statutes
to support their case). It is often necessary to take things a little
further and explore the reasons for the behaviours being observed.
The why questions are, perhaps, the more challenging since they
require the observer to draw inferences from what is said/done.
Where the observer has adopted a participatory role, s/he will be
able to build on these inferences by actually asking the participants
questions about the situation.
Structured observation studies

In a structured observation, it is usual for the researcher to use an
observation schedule or coding scheme to record his/her observa-
tions. A coding scheme is simply a set of pre-determined factors
which will be counted or rated during the observation. The coding
scheme can involve simple counting of objects/events or more
complex recording of sequences of behaviour. An example of a
simple structured scheme for use in a Youth Court can be found in
Chapter 10.
Coding schemes do need to be designed and used carefully in
order to ensure validity/reliability. It is not unusual for researchers to
adapt existing coding schemes which have been used in previous
studies. The categories used should be focused, clearly defined,
objective, exhaustive, mutually exclusive and easy to record. A good
example is the street observation by Shapland et al. (1994) men-
tioned earlier in the chapter. In this study, the researchers wished to
explore the crime patterns in different areas of a city. They wished
to examine these patterns in the context of the physical environment
in which they occur, thus they argue that it is important to establish
how people use different parts of an area and what crime opportu-
nities are present. So, in this study they used street observations in
which an observation schedule was used. The results of the obser-
vation were used in conjunction with interviews and analysis of
police data to target crime reduction initiatives.
In the first stage of the observation, the process of mapping an
area allowed the observers to identify potential crime targets. They
focused in particular on mapping residential, commercial and indus-
trial property, public space/property, routes (pedestrian and vehicles),
parking, possible crime hotspots (car parks, pubs, amusement
arcades), possible problems (litter, roads in bad repair, unused
buildings, graffiti) and crime prevention levels (burglar alarms,
security lighting, CCTV, etc.). The next stage of the observation was
a series of 10-minute observations of street activity which took place
at different times of day. Before the observation began, observers

JOBNAME: Mcgraw−crowsemmens PAGE: 13 SESS: 23 OUTPUT: Tue Aug 28 15:55:35 2007 SUM: 53C5A548
were required to note the general environmental conditions: weather,

lighting, general feel of the area, any activities happening, etc.
During the 10-minute observation period, observers were required to
note down everything that moved, including traffic flow and social
use of the area (totals of each kind of vehicle and pedestrian
counted). The observation schedule was designed so that each item
of interest, be it vehicles (cars, taxis, buses or lorries) or pedestrians
(women alone, families or groups of children), was clearly and
individually logged.
In both stages of the research, then, carefully designed observation
schedules were used and this allowed the collection of useful
quantitative data. A coding scheme/schedule does not, however,
necessarily collect purely quantitative data and can be useful in a
range of different contexts. It is particularly useful for ensuring
consistency between observations, especially where there are many
different sample periods or where a team of observers are employed.
Alternatives to note-taking
The most effective way to keep an accurate record of situations
where a lot of speech is used is to use a tape recorder. However,
there are a number of disadvantages with tape recording. The
presence of a tape recorder can make people feel uneasy; indeed,
some people may not consent to it in the first place. This approach
also produces a great deal of data which needs to be carefully filtered
through at a later stage. If an observation study is likely to be
carried out over long periods of time, it may be impractical to tape
record and transcribe everything. It may be more appropriate to
select only specific formal meetings or interviews for taping and
select a different method of recording for other, more informal,
situations (Hammersley and Atkinson 1983). Video-taping is
another option and it has been used successfully in some research
studies (see Sampson and Raudenbush (1999) in which the research-
ers drove through the streets of Chicago video-taping them as they
went to develop scales of social disorder, and Shrum and Kilburn
(1996) who video-taped the Mardi Gras festival in New Orleans to
examine episodes of nudity and ritual exchange). It is also recom-
mended by some researchers that a fieldwork journal is used as a
way of recording the process of research. It enables the observer to
record methodological problems and developmental ideas and can be
used in ‘de-briefing’ sessions as a way of reducing stress (Johnson
1975).
Data analysis
Once the process of data collection has begun, you must then think
carefully about the storing and cataloguing of that data. There are

JOBNAME: Mcgraw−crowsemmens PAGE: 14 SESS: 23 OUTPUT: Tue Aug 28 15:55:35 2007 SUM: 4F40C3E5
different ways of approaching this. For example, you could choose to

catalogue data chronologically. However, it is often better to develop
a system of cataloguing based on theoretical themes and concepts
identified before (and during) the study. So, for example, you might
organize according to people or places, or by events. There is also a
decision to be made with regards to whether one keeps manual or
computerized records. It will be necessary to create indexes, annotate
files and sometimes (where observations fall into more than one
category) duplicate entries. For some people, a manual system is
easier to control; for others, a computerized approach makes it
easier.
Unlike other methods, the process of analysis in observation
studies is not a distinct stage. Instead, analysis takes place through-
out the research process. The precise methods of data analysis in an
observation study will vary, depending upon the amounts of quan-
titative and qualitative data collected, but there are some basic
principles which should be followed. These are discussed in detail in
Chapter 8 but it is appropriate to give a short overview here. All
observation studies should start with the process of descriptive
observation – a ‘painting the picture’ exercise (Robson 2002). We
saw an example of this in the street observation study where the
observers ‘mapped’ what they could see before the 10-minute
observation period began. The next stage is to go beyond the
descriptive and identify concepts and develop theoretical frame-
works. To give an example, in a study of the ways in which the
police deal with disputes, Kemp et al. (1992) identified a series
of different concepts which together form an ‘incident’: the arrival of
the police at the scene, contact (when the police physically restrain
someone), information-gathering, decision-taking and implementa-
tion, the conclusion and the aftermath.
By thinking about how different concepts relate to each other and
how emergent ‘problems’ might be defined, you should be able to
draft some hypotheses. The next stage of analysis is to check on the
frequency and distribution of these observed concepts and test the
preliminary hypotheses. If the hypotheses are disproved (as they
often will be), then they are reformulated and re-tested. The
reformulation of hypotheses continues until successfully proven. The
final analysis involves the more complex modelling of relationships
(and relationships between relationships). This involves the construc-
tion of ‘roles’ and the creation of relations between roles which can
finally be developed into full descriptions of lifestyles and social
worlds.

JOBNAME: Mcgraw−crowsemmens PAGE: 15 SESS: 23 OUTPUT: Tue Aug 28 15:55:35 2007 SUM: 576754B3
The disadvantages of observation
Observational methods have received some significant criticisms over

the years but still remain popular in criminological research. It is
important, therefore, for you to be aware of the potential criticisms
before undertaking any observational work. First and foremost, the
process of observation is time- and resource extensive. Goffman said
that the minimum period of time for a participant observation study
should be one year (1961). We would not wish to set absolute
maximum or minimum time frame here, but there is a careful
balance to be struck between the need for adequate data collection
and the limits placed by finite resources. Some commentators have
also raised grave concerns about the ethical issues raised by both
covert observation (Bulmer 1982) and overt observation (Norris
1993). Another common criticism is that observation can be highly
subjective and so there is a significant risk that the researcher’s own
views may influence the outcome of the study. In the case of
structured observations, this happens where a completely inappropri-
ate conceptual framework is put onto the group or setting that you
are observing. This is particularly problematic where the study is
intended to be exploratory and uncover new material. Bryman
(2004) suggests that a possible solution to this problem is to start
with some unstructured observations and to use the results to
develop the observation schedule. However, this is not to say that
unstructured observations are more immune from subjectivity. It may
be argued that, especially in the case of grounded theory studies,
analytic induction serves as a useful tool in theoretical development,
but we would join critics in questioning whether it is really possible
for a researcher to approach a research study without bringing any
theoretical ideas to it at all (for further discussion, see Chapters 2
and 8). Thus, we think it important to emphasize how crucial it is
for the researcher to build validity checks into the research meth-
odology wherever possible.
It is also important to recognize the limits of what can be
understood through observation. One of the major restrictions with
structured observation is that you are restricted to recording only the
‘directly observable’. This means that you may overlook the context
in which an event or interaction occurs and that you may not be
able to extract much information about people’s intentions or
motivations. Participant observation in particular has been heavily
criticized as, due to the significant access issues, it is mostly
restricted to working-class settings and therefore cannot generate
any real analysis of power relations. For the criminological researcher
this is particularly frustrating because interesting avenues of research
are closed off (for example, finding out how senior personnel in the

JOBNAME: Mcgraw−crowsemmens PAGE: 16 SESS: 23 OUTPUT: Tue Aug 28 15:55:35 2007 SUM: 3CEC9F3E
criminal justice system do their jobs, or how crime and corruption

are committed by executives in multinational corporations).
Conclusion
Ultimately, the decision whether to use observational methods will

depend upon the nature of the proposed research questions. Ham-
mersley and Atkinson (1983) emphasize the logic of the ethno-
graphic approach and argue that its reflexivity eases the tension
between positivism and naturalism. Yet, observation is not useful for
understanding ‘past or fast’ events, nor is it possible to move beyond
a local, restricted level of explanation and generalize our conclusions.
Why, then, would anyone bother engaging in this kind of research?
Well, the attraction lies in the ability to discover things which are
new, often things which are deeply buried. Observation can reveal
surprising phenomena and that is its major value.
Further reading
Both Robson (2002) and Bryman (2004) contain useful illustrations

of schedules for structured observation and we provide our own
example in the Appendix to Chapter 10. One of the most useful
texts on ethnography is by Hammersley and Atkinson (1983)
although it is now a little dated and might be best supplemented
with a more modern text such as Coffey et al. (2007). We do
believe, though, that in order to get a real feel for the advantages
and disadvantages of participant observation, you really have to read
one in its original form! These kinds of studies tend to be both
accessible and engrossing to the novice researcher and we have
found that our own students particularly enjoy reading Ditton’s
bakery study (1975), Holdaway’s ethnography of the British police
(1983) and Becker’s classic, Outsiders (1963).

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 4567334D
7 Researching by asking and

listening
Introduction
The method that has produced probably more criminological

research than any other is that of listening to what is said. We use
the word listening here in a colloquial sense because we include
questionnaires which may sometimes, but not always, produce
non-verbal responses, but we think that using written responses to
questionnaires is a way of listening to what you are being told. The
three main types of listening covered here are semi-structured
interviews, questionnaires and the kind of group interviews usually
called focus groups. What they all have in common is that there is
some form of interaction between researcher and participants. This
is important because a two-way relationship is involved. It is easy for
the researcher to be concerned only with the information that they
require in order to pursue their studies, and to see the people they
are relating to as passive partners who are there to provide what they
want. This would be a serious mistake for several reasons. One is
that it involves the objectification of participants, and once this
occurs, unethical research practices can happen more easily. The
second reason is a more pragmatic one: the more aware you are of
how others respond, the more likely you are to gain reliable and
meaningful insights.
It is also important to be aware of how participants might use
researchers. This can happen in various ways. One is that the
researcher can become an ad hoc, and totally unqualified, counsellor.
Research gives people an opportunity to be listened to that they
might not otherwise have, to say what they think, to say what they
feel, perhaps to talk about some very personal matters. We have
already referred, in Chapter 3, to the ethical concerns that such a
situation can raise. A good and experienced researcher will acquire

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 5045ACB9
Researching by asking and listening 117
the skills to draw information out of people, sometimes victims,

sometimes offenders, sometimes members of the public or the
judiciary. A victim may be opening up to someone for the first time
about what they have experienced. An offender might be revealing
things that they have done that they wish they had not. It is in this
sense that the researcher is being used as, almost literally, a listening
post. This is perfectly reasonable, but it is important to remember,
and remind respondents of, what your real role is, and what the
limitations of that role are lest unrealistic expectations occur. When
talking to people who represent various parts of the criminal justice
system they may be using the researcher to get their views taken
note of by the consumers of the research, whether that is the Home
Office funding the project, or the wider readership of the research
output. For example, if you interview court officials about the effects
of new legislation or criminal justice procedures, you need to be
aware that what you are getting may reflect the agency’s ‘party line’
on the issue. It may be that you are happy to find out what the
agency’s ‘party line’ is, and accept it as such, or it may be that this
is masking what is really happening. The skilled researcher has to
make many judgements about the kind of information they receive,
and how to interpret it in the light of their own intervention.
Listening can occur in different methodological contexts. Ethnog-
raphy and participant observation are likely to involve the least
structured kinds of research encounter, whereas a survey will usually
be much more structured. A research project might use more than
one kind of listening technique, combining structured questionnaires
with semi-structured interviews. This might be because different
techniques are appropriate to different situations, or it may be that
different techniques feed into one another. For example, a study of
patterns of crime and law enforcement in a particular area might
start with some relatively unstructured discussions with various
groups of people and key individuals. What emerges from these
might be incorporated into a more structured questionnaire given to
a larger number of people to see whether what came out of the
small-scale explorations is reflected on a larger scale.
Interviewing
Interviewing can range from a very minimal structure to the much

more highly structured interview conducted using a questionnaire
(which we will return to in the next section). Denzin (1988, Ch. 6)
defines three types of interview:
+ the schedule standardized interview;
+ the non-schedule standardized interview;

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 4F8E9C4F
+ the non-standardized interview.

The schedule standardized interview is characterized by being highly
structured, the wording and order of all questions being the same for
every respondent. However, careful preliminary investigation, devel-
opment and pre-testing (probably involving less structured forms of
interviewing initially), are required. The non-schedule standardized
interview, also referred to as the focused interview, consists of a list
of information required from each respondent, but the phrasing and
order of questions may vary. It recognizes the special nature of each
interviewee’s contribution, and requires skilled interviewing. In the
non-standardized interview, questions are not pre-specified. The
interviewer probes various areas, and can explore a range of
hypotheses as the interview progresses. It also means that the
interviewee has the freedom to take the interview into areas that the
researcher had not previously considered. Where unstructured, non-
standardized interviews are used, the aim is usually to elicit material
that comes from respondents with the minimum of input from the
researcher. Even here, however, there is likely to be an agenda that
the researcher will follow. The researcher will usually bring to a
situation some expectations, and emergent hypotheses.
The three types of interview serve different functions. For
example, the non-standardized interview is best for exploratory
work, ascertaining the meanings and definitions used by respond-
ents, whereas the standardized interview enables more generalizable
hypothesis testing. It is important to be clear about which function
is being used at different points in the investigation. Each interview
type has shortcomings, and their use depends on the purpose for
which they are intended, the expertise of the interviewers, and the
time and resources available.
We have already considered the listening that takes place in the
context of ethnographic research, so we focus first on semi-
structured interviewing, which comes closest to Denzin’s non-
schedule standardized interview. This is an interview where there is
an interview agenda, usually with certain pre-determined questions,
but with a degree of flexibility in the way in which the interview
proceeds. Following this, we consider group interviews and focus
groups, and finally consider the most structured form, the question-
naire, which may be used in interviews, but may also be used as a
self-completion tool.
Semi-structured interviews
When is semi-structured interviewing appropriate?
In our experience, this is a method that is popular in socio-legal
research,1 especially with those who do not have a background in

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 53A12DDD
social research training. It is appealing because it does not appear to

be as demanding as designing a questionnaire, but it should be used
for the right reasons – because it is appropriate rather than merely
convenient. The decision to employ such a data collection technique
should be primarily determined by the theoretical requirements of a
study rather than just as a way of ‘fishing’ for information. You need
to be knowledgeable enough about the topic as a result of prelimi-
nary investigation and informal enquiries to get the most out of an
interview without appearing naïve. Don’t rely on semi-structured
interviews to enable you to get to know about an issue.
Semi-structured interviews may be used where depth of study is
of foremost concern, especially where it is important to know what
meaning certain things have for people, or where one wishes to
explore a process or sequence of events. Semi-structured interview-
ing is likely to be particularly helpful where you are looking at
complex phenomena that may have developed over a period of time.
For example, if you were attempting to unravel the relationship
between employment and offending, it might be useful to take
people through a period of time, since they left school, for example,
and ask them what they were doing at particular periods in time
(this was done in a study in Northern Ireland: Gormally et al.
1981). In a similar way, a study of how fine default is related to
people’s other debt problems may need to help respondents recall
their debt situation by means of a patient and careful reconstruction
of their circumstances (Crow et al. 1993). Another example of the
use of semi-structured interviews might be to examine the experi-
ences that people from ethnic minorities have of the criminal justice
process with which they have been involved. In this instance it may
be valuable to take people through various stages of the process (see
Smellie and Crow 1991).
Another example of situations where semi-structured interviews
are appropriate is when you are doing what are referred to as ‘elite’
studies (Moyser and Wagstaffe 1987). Much social research is
directed at ‘ordinary’ people, or at the least well-off and least
powerful in society. Elite studies involve the more powerful members
of society, such as judges, magistrates, senior police officers, and
senior members of organizations. Such people, if you gain access to
them at all, are likely to regard filling in a questionnaire as not for
them. It is both a matter of making the most of an opportunity to
engage with powerful elites, and of them wanting to be accorded
some status in the research process. These are likely to be situations
requiring careful handling by the researcher.
Another situation where semi-structured interviewing is applicable
is where you are studying sensitive issues – something that is quite
common in criminology. These may include both offending and
victimization experiences, situations of abuse and mental ill health.

JOBNAME: Mcgraw−crowsemmens PAGE: 5 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 509689F3
This is not an exhaustive list of circumstances where semi-structured

interviewing is likely to be advantageous, but gives an indication of
some of the circumstances where it can be appropriate.
The interview schedule

We will come on to question construction in the context of
questionnaire design shortly, but the significant feature of the
semi-structured interview is that questions, and the order in which
they are asked, are not completely pre-determined. Thus, primary
questions may be followed by pre-determined ‘prompts’, or the
interviewee may encourage the respondent to develop his or her
thoughts on a matter with such general prompts as, ‘Can you tell me
what you mean by that?’ or ‘Could you tell me more about that?’.
This does not mean that semi-structured interviews are completely
free-flowing. The interview will be shaped by the hypotheses that
have been developed as a result of the research process outlined in
Chapter 1, but the interview is likely to contribute to hypothesis
development as much as hypothesis testing. The extent to which this
takes place depends on the generalizability of the interviews. In
some situations, the interviewees may be part of a wider group of
people, a bench of magistrates, for example, and the interviews tell
you something about the group. But unless the interviewees consti-
tute a reliable sample then they will not enable you to determine
whether what you hear from interviewees is true generally. A more
systematic process of data collection, such as a questionnaire to all
magistrates, may be a necessary further stage for this to happen. If,
on the other hand, you were interviewing individuals who play a
pivotal role in the issue being researched, such as a chief constable,
a Crown Court judge, or a senior civil servant responsible for an
area of criminal justice policy, then what they say may be more
definitive.
When working on the questions to go in a semi-structured
interview, it is a good idea to frame them in such a way as to
encourage the respondent to open up, rather than something that
will simply elicit a ‘yes’ or ‘no’. So rather than ask a probation officer
‘Do you think the proposals for a new sentence for offenders are a
good idea?’, ask, ‘What do you think of the new sentencing
proposals?’. The studies of the unemployment and offenders
described in Chapter 9, and the Youth Court (Chapter 10) also used
semi-structured interviews.
Conducting the interview

The way the interview is conducted will vary depending on
circumstances. If the interviewee is a member of a criminal justice
agency, or some other organization, then the interview is likely to be

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 507D4A5E
in his or her office. If they are a member of the public, a victim, or

an offender, then it might be in their home, in your office, or it
might be in some mutually agreed public space. Whatever the
setting, it is important to try to arrange for privacy, and for the
interview to be uninterrupted (e.g. by telephone calls). Bars and
cafés are not a good idea. The interviewee needs to be told things in
advance, such as how long the interview is likely to last. The length
of interview will depend on circumstances, and there may be
situations where more than one interview with the same person is
desirable and possible (if trying to reconstruct someone’s life history,
for example). If interviewing a busy criminal justice practitioner,
then 45 minutes to an hour may be as much as you can expect. If
the respondent is a teenager who has been to court, then one might
be lucky to get 20–30 minutes. While people don’t like the prospect
of an interview lasting too long, it is not unusual for the interviewee
to become involved in the topic and talk for longer than intended,
so be patient and be prepared to listen to some things that may
subsequently prove to be irrelevant.
An important consideration is how the interview is to be
recorded. Ideally a recorder will be used, but this will depend very
much on the interviewee’s agreement and, if possible, ask for
permission in advance rather than put them on the spot at the time
of the interview. If a recorder is being used, then ensure the
recording will be clear and free from background noise. Check that
batteries are topped up (or that a mains supply is available), and that
whatever recording medium is being used, there is enough tape, disc
or memory for each interview. These may all appear to be obvious,
but good interviews have been spoiled by batteries running down or
tapes running out.
If the interview is not being electronically recorded, then taking
notes is the next best option, but it is difficult to handle the rapport
needed to conduct an interview, listen carefully to answers, and write
notes at the same time. Respondents are unlikely to open up if they
have to keep waiting for you to write things down, so some very
brief aide-mémoires may be best, and if you don’t know shorthand,
try to develop your own shorthand codes. It is sometimes possible to
do an interview with a fellow researcher in attendance, which makes
it easier to share the tasks of interviewing and writing notes, and to
compare notes afterwards. In certain situations even note taking is
not possible and the best you can do is write down as much as
possible as quickly as possible after an interview. This has happened
to one of us when doing research involving drug users and dealers.
Even when notes have been taken and the interview recorded, it is
still a good idea to write it up as soon as possible – certainly the
same day. Straight after an interview phrases and ideas remain in the
mind which fade after a night’s sleep. When a series of interviews

JOBNAME: Mcgraw−crowsemmens PAGE: 7 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 50E132A3
are being conducted it is also worthwhile drafting a synopsis of the

results of a group of interviews at regular intervals, and it is
important to reflect on what you are learning and its relevance to
the hypotheses you are investigating.
Your design may require that the interview agenda or schedule
remains the same on each occasion. This enables you to make
generalizable statements covering all respondents, but equally this
can be frustrating because after you have interviewed a few people
you may think of further questions that you would like to ask, but
if you keep altering the interview, then you have no way of knowing
how earlier interviewees might have responded.
Analytic induction
There is an alternative option and this is the technique known as
analytic induction (also mentioned in Chapter 8 in relation to the
analysis of qualitative data). Using this approach, instead of covering
the same or similar questions for all respondents, the aim is to
progressively develop your knowledge and ideas from one respondent
to the next. To do this you start with a preliminary hypothesis and
interview the first respondent on the basis of this hypothesis. You
then consider what has been learned from this respondent, and
modify your hypothesis accordingly. The next respondent is inter-
viewed on the basis of this modified hypothesis and the second
interview is used to refine the hypothesis further so that it can apply
to both respondents. This process is repeated until diminishing
returns are achieved. Analytic induction is put forward by Denzin as
participant observation’s way of demonstrating causality (Denzin
1988: 194–9). While quantitative analysis using survey techniques
can manipulate variables statistically to control factors that might
affect the relationship between the dependent variable and inde-
pendent variables, the participant observer seeks to disprove a
hypothesis until all the cases in the study can be explained.
The technique was used by Lindesmith in explaining the onset of
opiate addiction in 1947. As an example, one of us used this
technique some years ago when doing some exploratory investiga-
tions into why some people were having bad experiences when using
certain illegal drugs such as cannabis and LSD. An initial hypothesis
might be that the drug was contaminated in some way. The first
interview reveals that while the respondent had a bad experience,
friends who had taken the same drug had not. Thus the first
hypothesis was rejected, and this led to the hypothesis that it was
some of the friends he was with who had made the experience a
difficult one. Further interviews revealed a series of other factors,
including the setting in which the drug was taken, and personal
problems being experienced at the time. After a couple of dozen
interviews it emerged that the bad drug experience could result from

JOBNAME: Mcgraw−crowsemmens PAGE: 8 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 4B9B5351
a particular combination of factors which might vary from individual

to individual, enabling advice to be given to users about how to
reduce the risk of harm occurring.
While analytic induction can be a useful method, it is important
to bear its limitations in mind. The first concerns the selection of
cases, and as ever in social research the question is how far one can
generalize from the cases interviewed. Denzin suggests that the
principle covering the selection of cases to test a theory is that the
chances of discovering a decisive negative case should be maximized
(1988: 194). One case is sufficient to refute or modify a hypothesis.
However, this depends very much on the context in which one’s
study is taking place. Analytic induction can be very useful for
developing and refining hypotheses which are then validated on a
larger and more generalizable sample. It also has the advantage of
enabling the researcher to understand the processes that shape the
relationships between variables in a way that purely correlational
analysis does not.
Writing up
As mentioned above, once completed, interviews should be written
up as soon as possible afterwards, preferably the same day while
words and phrases are still fresh. How this is done will depend on
the circumstances and recording medium, but even an interview that
has been recorded and is to be typed out verbatim should be
summarized in order to distil the key issues. Where a series of
interviews, perhaps comprising different groups of respondents are
involved, it is worth reflecting on how each interview adds to the
overall picture that is emerging. We discuss the analysis of material
further in Chapter 8.
Group interviews and focus groups
The second method for collecting material by asking and listening is

that which involves groups of people. This takes two main forms.
One is using group interviews, and the other is focus groups, so it is
useful to start by defining the difference.
Put at its most simplistic, group interviewing is a way of doing
interviews with people more than one at a time. You can use the
kind of schedule that one might use for a semi-structured interview,
responses can be recorded on the basis of what each individual said,
and the results can be analysed both quantitatively and qualitatively.
It is more complex than doing interviews one by one, but the key
features are:
+ group interviews are a way of saving time and resources;

JOBNAME: Mcgraw−crowsemmens PAGE: 9 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 4BE0B908
+ several topics are usually discussed;

+ the researcher is interested in the responses of the individuals
involved.
Focus groups, on the other hand, are usually used to explore a
specific topic in some depth, with the main interest being in the
interaction between group members. The use of the two methods is
similar in some respects, but we will concentrate mainly on focus
groups, referring to group interviews where appropriate.
When are focus groups appropriate?

Focus groups were developed by the sociologist Robert Merton in
the 1940s as a way of evaluating the effects of wartime propaganda
(Merton et al. 1956). They came into prominence again during the
1990s both as a research method, and as a tool of social and political
policy (Morgan 1992; Morgan and Krueger 1993). Questionnaires
and semi-structured or unstructured interviews involve asking and
listening to individuals. But people’s views and perceptions, and what
occurs in organizations, don’t happen in isolation from others; they
develop in a social context. Even people’s most personal experiences
are often interpreted by them in relation to others. What people
think, and how they frame their thoughts, takes place in interactions
with their neighbours, work colleagues, family and friends. Therefore
if we want to understand how people feel about crime and safety in
their neighbourhood, for example, one way of doing this is to get a
group of them together. This enables you to get closer to the
to-and-fro social exchanges that take place in everyday life. You also
get a more dynamic picture. One-to-one interviews produce static
results, as though this is what a person has always done or thought,
but people’s knowledge, interpretations and conclusions change
through exchange. In a (non-scientific) way people are constantly
producing their own hypotheses and testing them out in relation to
the world around them. Focus groups are a way of getting some
insight into this process, albeit a limited one. Focus groups are
therefore useful when we are interested in finding out how people
reason through problems or make decisions about things. They help
us to understand why people feel the way they do, and they enable
us to find out how groups of people interpret a phenomenon and
construct meaning.
Limitations of focus groups

On the other hand, the views that people might express by
themselves are liable to be challenged when mentioned in a group,
and if the topic is particularly sensitive, people may feel uncomfort-

JOBNAME: Mcgraw−crowsemmens PAGE: 10 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 521EDF18
able. One of the drawbacks of collecting material from groups of

people is that they may not say what they really think in front of
others, and they may not want to talk about experiences, such as
victimization, with other people present. As another example, if we
were wanting to explore how people’s views and experiences of the
way the criminal justice system deals with people from different
ethnic backgrounds, then it is less likely that people will be honest
if they are in a mixed race group. Similarly, there are things that
women and men may be less likely to want to talk about in a mixed
sex group.
Furthermore, the dynamics of a group and the way they interact
can affect how the discussion goes. If ten people are allocated to two
groups on the basis of the first ten in one group and the second ten
in another group, we might get a very different picture to that
which would emerge if the first, third, fifth, seventh and ninth
people were in one group, and the second, fourth, sixth, eighth and
tenth people were in another group.
Interviewing in groups, whether focus groups or group interviews,
is an altogether more complex undertaking than one-to-one data
collection. The researcher has less control than s/he would have over
a single interview and has to have good interpersonal skills. There is
a delicate balance to be maintained between having enough control
to ensure the group doesn’t take over and stray away from the topic,
and exerting too much control on the other. If the latter happens,
then the group’s responses may be dictated too much by the
researcher’s influence. The material is also more difficult to record
and analyse when several people are involved.
Forming the groups

It is usually necessary to do more than one focus group because you
have to be sure that the discussions emerging are not unique to that
group. How many groups you do will depend on the resources
available, but there is a general rule that you should keep running
groups until you think you can predict what kind of things a group
is going to say, known as ‘category saturation’. Typically, a group will
consist of 6–8 members. A common problem is predicting the
number of people who actually turn up on the day. How people are
selected will depend on the topic. In theory, the same principles of
sampling apply to focus groups as to any other form of social
research: you try to ensure that they are as representative as possible
for the topic being studied. However, one of the great strengths of
focus groups is that they engage people in discussing common
interests (e.g. magistrates talking about sentencing or sex workers
talking about things they would not want to discuss with others).
Sometimes it is necessary to sample very purposively, such as asking

JOBNAME: Mcgraw−crowsemmens PAGE: 11 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 5318803F
a group of young offenders about their experiences of the criminal

justice process. Indeed, Morgan (2006: 121) argues in favour of
homogeneous group composition because such groups can discuss
topics that would be considered taboo in other contexts.
Running the groups

As mentioned already, listening to what people have to say can come
from using a structured set of questions, or from having just a few
broad questions as a starting point, which can then be followed up
with subsidiary questions and prompts (‘What do you mean when
you say that?’, ‘Can you give me any examples?’). The same is true of
focus groups. Even if the questions are not structured, it is
important to have an agenda, both to ensure that participants don’t
stray too far from what you need to cover, and to ensure that you
cover the same ground when several groups are being used.
The role of the moderator is crucial. He or she may be a
researcher, but if resources permit, it may also be worth having
someone whose role is purely to facilitate. This leaves the research-
er(s) free to concentrate on what is going on, both verbally and
visually. The down-side of this is that a non-researcher facilitator
may not be so well attuned to the requirements of the research, and
less likely to steer participants in the desired direction. If there is no
independent facilitator, it is desirable to have two researchers present
– one to moderate and one to record, although this is less crucial if
the session is being filmed.
The moderator should not be too intrusive, achieving a balance
between enabling discussion and keeping focused on those matters
which will permit the research hypotheses to be explored. He or she
should introduce the participants, explain the purpose of the
research, and tell participants about the format of the session,
recording techniques, and of course about confidentiality and what
people can do if they have any concerns. The moderator will also
conclude the session, thanking people, and explaining what will
happen to the data. One of the biggest concerns about focus groups
is ensuring an adequate balance of participation, so that more
assertive people don’t become too dominant, and bringing more
retiring members into the discussion. However, there may be
situations when sensitive topics are being discussed (e.g. domestic
abuse) when it is proper to give participants the option to remain
silent.
It is also worth commenting on the fact that a focus group may,
whether intentionally or not, affect participants in some way,
especially since criminological research is liable to deal with uncom-
fortable issues relating to victimization and offending. Sessions may
evoke changes in attitude and perception, or re-awaken memories.

JOBNAME: Mcgraw−crowsemmens PAGE: 12 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 52966BB2
As a matter of good ethical practice there should be some provision

for dealing with this by following through any consequences, and
referring participants for help if appropriate.
Focus groups should ideally be recorded using a voice recorder,
and if possible filmed. It is important to remember that in addition
to what people say, a lot can be learned from visual responses and
reactions. Do people look angry, concerned, disgusted? Eye contact
and body language can also be revealing. This is particularly
important where groups of a particular type are being used. For
example, if you were bringing together a group of criminal justice
personnel such as magistrates, police and probation officers to talk
about how crime problems in an area are being dealt with, do some
members appear more confident than others? Similarly if the focus
group was young people, then it may be interesting to see whether
some of the group are more assertive than others. This may tell you
a lot about peer group influences.
We deal with the analysis of data in Chapter 8, but it is worth
noting here that the analysis of group interviews and focus groups is
likely to be more complex than other, one-to-one forms of data
collection. But it is the very richness of the material that gives such
group data collection methods their value to researchers.
Illustrative examples
Having written about focus groups in the abstract, it is probably
useful to end with some examples. A good example can be found in
an article by Överlein et al. (2005), which discusses the focus group
as a method for engaging young women aged 15–20 at a detention
home in talking about sexuality. They show how the focus group
approach can enable a quite troubled group to engage in discussing
a ‘high involvement’ topic with their peers in a way that might not
happen if they were just responding to a researcher. In this study,
articles from popular magazines for young women were used as
discussion material in five one-hour long focus groups. The use of
some kind of presentation, video, or reference to material is helpful
in focus groups, since it acts as a stimulus to discussion. They
explored topics such as femininity and body image, starting with less
sensitive topics such as ‘female role models’ before moving on to
more sensitive topics such as ‘when to say no to sex’. The researchers
found that participants provided mutual support to each other and
concluded that the method had provided a rich and varied set of
data that could not have been obtained by individual interviews.
They also found that the groups enabled the young women to
pursue matters that they had been concerned about previously, but
not felt able to discuss, and to this extent focus groups can also act
as a form of action research.

JOBNAME: Mcgraw−crowsemmens PAGE: 13 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 4C8BFF54
Another example is a study that was done by one of us

concerning the ways in which magistrates and social workers
respond to youth crime. Eight group sessions were arranged, three
involving magistrates, two with justices’ clerks and three with social
workers, ranging in size from four to six people. The sessions lasted
approximately one and a half hours.
Three main areas of discussion were identified for the groups:
1 Were there different views among magistrates and social workers
about the punishment and care of young people in general?
2 What perceptions did magistrates, court clerks and social workers
have of each other?
3 What were the views of magistrates, court clerks and social
workers in relation to certain current local and national issues,
such as the closure of a remand centre, the introduction of
remands with foster parents, locking up young people, the
introduction of a Secure Training Order, and the problems posed
by certain types of young offender, especially the young persistent
offender?
Participants were given a sheet of questions covering the following
issues:
+ any particular types of crime that caused them concern;
+ particular groups of offenders that caused them concern (e.g.
certain age groups), and what they thought could be done about
them;
+ their views about what the main causes of youth crime might be;
+ what they thought the main objectives of the magistrates’ court
and of Family and Community Services were in relation to youth
crime;
+ their views regarding such matters as cautioning, remand arrange-
ments for young people, and the sentencing of young offenders.
It was stressed that participants were also free to introduce other
matters related to youth crime which they felt were relevant. There
were always two researchers present at each group, which was tape
recorded, and researchers’ notes were compared and written up soon
after the session.
The study showed that magistrates, clerks and social workers
shared some concerns about persistent offenders, drug use and
offending while on bail. However, it also showed how magistrates,
clerks and social workers differed in their perceptions of what led
young people to get involved in offending and what should be done
about it. Those from the court emphasized their role in trying to
protect the public, whereas the social workers, while concerned to
protect the public, also wanted to protect the young people from
themselves and others. The research laid the basis for subsequent

JOBNAME: Mcgraw−crowsemmens PAGE: 14 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 565EA2B0
discussions between the court and Family and Community Services

about how to improve their working relationship and build on what
they had in common (Crow 1996).
A third example of the use of focus groups in criminology is a
study of women’s fear of crime by Esther Madriz (1997a, 1997b).
The aim of her research was to explore the attitudes of women
towards crime, with a view to untangling the apparent paradox that
women have higher levels of fear of crime, even though their levels
of victimization are lower than men. She developed a theory that
women’s fears are exacerbated by stereotypical images of criminals
and victims and set out to examine how popular images of criminals
and victims were interpreted by a group of women of different ages
and backgrounds.
In her research, she used a combination of 18 focus groups and 30
in-depth interviews with women living in the New York area. This
is a good example of a research study in which focus groups are
combined (or triangulated) with other methods in order to build a
more detailed and valid picture of the phenomenon being studied.
Another interesting aspect of this study is the approach taken to the
sampling. The researcher made a deliberate decision to include more
women of colour and teenagers in the sample than were represented
in the actual population. This was for two reasons: first, because
these groups had not been studied in detail in the existing research
on fear of crime; second, because the researcher hypothesized that
racial diversity, socio-economic status and age may be important
aspects of the images of criminals and victims. The participants were
recruited using a snowballing technique (see the section on sampling
in Chapter 3). The researcher contacted community organizations
and began the process of collecting participants through ‘word of
mouth’. This meant that the final sample consisted of a range of
different women from different socio-economic backgrounds and
from different ethnic groups but could not be said to be representa-
tive or free from potential bias. However, this was recognized by the
researcher and justified as an approach which fitted the aims of the
study.
There were 18 focus group sessions with a total of 140 partici-
pants (thus approximately seven participants in each group) and each
session lasted between one and two hours. The researcher describes
how she used a ‘discussion guide’ which was structured enough to
keep some level of consistency between groups, but also flexible to
allow for the discussion to flow. The result was that several themes
emerged from the discussions. Regardless of the race and socio-
economic background of the women, the images of criminals tended
to be strongly racialized with Black and Latino men being the most
feared. Immigrants were also perceived to be threatening and
criminals were often described as being savages or monsters. The

JOBNAME: Mcgraw−crowsemmens PAGE: 15 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 509F389A
women also had a common image of victims. However, although the

stereotypical victim was commonly perceived as a white middle-class
female, there were differences between some individuals’ ideas of the
race, socio-economic background, size and appearance of victims.
Interestingly, most discussions contained images of women as victims
of sexual assault and murder rather than of other, more common
types of crime. The researcher concluded that female fear is
exacerbated by images of crime contained in the prevailing ideology
of crime. Although the images are not consistent, some common
themes emerged from the participants’ discussions. It is clear from
reading about the research that some very interesting themes
emerged from the focus group discussions, themes which may have
been missed if the researcher had relied upon interviews with
individuals. The focus group format allowed the women to share
their ideas and experiences and express themselves more fully.
Questionnaires
When are questionnaires appropriate?

When people think of social research, a questionnaire is often the
first thing that comes to mind. However, this is only one way of
asking questions and gathering information, and it may not be the
best one. Questionnaires are usually highly structured, with limited
options for responses and involves ticking boxes and scales, which
will be coded and analysed numerically. It is important that every
respondent is asked the same question, in the same way, and
presented with the same response choices. They involve quite a lot
of preparation if the options for administration, response and
analysis are to produce worthwhile results, and they have significant
limitations. Therefore it is important to think about what you hope
to get out of a questionnaire, and whether some other form of data
collection might be more appropriate before preparing one. Another
common belief is that questionnaires are just a matter of common
sense, which anyone can undertake regardless of whether or not they
have prior experience or training. Both of us can attest to having
been asked to rescue someone who thought that doing a question-
naire was easy, who ends up with a large pile of material that is of
little real value!
Strengths and weaknesses of questionnaires

First, they make it possible to collect information about a large
number of people, or whatever you are studying, relatively easily and
quickly.2 It is possible to collect hundreds of completed question-

naires in a relatively short time by getting people to fill them in at

meetings or conferences, in schools or universities, or in some other
setting where people are assembled. The same can happen over a
matter of days by sending questionnaires to people. Consequently,
you can have a large amount of information at your fingertips in a
short space of time. This apparent simplicity can, however, be
deceptive since your chosen means of administration is not guaran-
teed to produce reliable results, or a high response rate.
Second, once you have got your completed questionnaires, pro-
ducing at least some of the results can be relatively quick and easy.
As we explain in Chapter 8, numerical analysis can be performed
using a software package. If the data is to be entered by hand, this
can take time, but these days it is relatively easy to enter data
directly into a computer, or set up a pre-coded questionnaire so that
the responses can be scanned in to the computer. Again, a note of
caution, since problems can be encountered with errors in the data,
and more complex and worthwhile results often require data
manipulation and an advanced knowledge of statistics.
Third, as long as the sample is a well-founded probability sample
(see Chapter 3) and there is a respectable response rate (see below),
generalizations can be made to a wider population, and this is one
of the great strengths of the survey compared with other means of
inquiry.
For these reasons questionnaires have become a popular and
extensively used means of inquiry. But they also have their draw-
backs. One is that precisely because they have become so widely
used they can engender ‘questionnaire fatigue’ and consequently a
reluctance to complete them.
Furthermore, they are a very stilted form of inquiry. The very
qualities that make them relatively easy to complete also mean that
they tend to simplify things. While they may be a convenient way of
collecting factual data, gathering information into categories and
producing rankings and scales, a questionnaire is less useful for
looking at shades of meaning, and complex processes. There is also
a problem of validity: it is difficult to be sure that all the
respondents understand a question in the same way.
So questionnaires are probably best used when you have a good
idea what it is you want to know, possibly from previous fieldwork,
and you want to be able to generalize the results to a much wider
population. Because of their limitations questionnaires are best
produced only after a reasonable amount of preliminary work and
preparation has gone into them, and it is best not to try to use them
to achieve something that they cannot do. Examples of when a
questionnaire might be appropriate in criminology are:
+ crime surveys, where a study is being undertaken of people’s
experiences of, or views about, crime-related matters;

JOBNAME: Mcgraw−crowsemmens PAGE: 17 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 413AF9E9
+ questionnaires for criminal justice personnel about some aspect of

their work.
If a questionnaire is appropriate, before developing one it is
important to check that this really is the only way of getting the
information needed. Is the information already available elsewhere,
in earlier surveys, official sources or agency records? For example,
when doing a study of debt among a sample of fine defaulters, quite
a lot of information is likely to be already recorded. Therefore
(observing proper ethical practices for obtaining access to these
records), it is possible to avoid asking people in the sample questions
about information that is already available elsewhere, which could be
obtained by means of an abstraction pro forma.
Collecting the data

The term questionnaire actually covers a variety of information
collection techniques. These vary depending on the method of
collection, and who provides the information (Table 7.1). When
people think of a questionnaire survey the image that comes most
readily to mind is that of a researcher with a clipboard interviewing
a respondent from a sample face-to-face. Although quite a common
method of administration, it is by no means the only one. It is one
of the best because there is both verbal and visual interaction in
which the persons concerned can express themselves most clearly.
A common alternative is the telephone interview, which is rela-
tively cheap and quick. One of the drawbacks of telephone inter-
viewing used to be that not everyone had access to a telephone,
leading to a biased sample. The spread of land-line telephones
reduced this problem, but in recent years the spread of mobile
telephones has complicated matters, since many people are more
likely to rely on their mobile telephone which is harder for a
researcher to contact. Since telephone surveys tend to rely on
land-lines, they are likely to exclude certain groups who don’t have
access to a personal land-line, such as those who are not living in
Table 7.1 Data collection

JOBNAME: Mcgraw−crowsemmens PAGE: 18 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 574A8D32
their own accommodation, in a hostel, or without a home at all.

Furthermore, telephone interviewing is less personal than face-to-
face interviewing and has become devalued as a result of its use as
a marketing tool. But it could still be useful for interviewing people
as representatives of businesses or organizations.
The internet has also been used as a way of getting people to
complete questionnaires online, and has the advantage of providing
anonymity. It is likely to be particularly useful for an organization
that wants to get information from its staff, students, clients, and so
forth – in other words, where the enquirer has access to a database
or sampling frame. It is, however, of limited use for surveys of the
general population, since a viable sampling frame is more problem-
atic. Furthermore, research in 2003 found that internet polls do not
necessarily produce results that are representative of the population
as a whole, ‘even after very considerable weighting of the results has
been undertaken or care exercised to ensure that those who are
asked to complete an internet poll are demographically and politi-
cally representative of the whole population’ (Baker et al. 2003: 27).
This could, however, change as technology develops.
Postal questionnaires are a relatively cheap and convenient way of
covering a large number of people or organizations. But response
rates can be low, especially if the recipient gets the questionnaire ‘out
of the blue’. The researcher also has very little control over the
circumstances in which the questionnaire is filled in, and there is no
opportunity to explain any questions that might be interpreted
ambiguously by the respondent.
Who produces the information?

Another important consideration is who is involved in the collection
process. Self-completion means respondents provide the data directly,
but care has to be taken to ensure that they understand what is
required of them; very precise instructions are needed, and you have
less control over the interpretation of questions. If the collection is
by interview, then by whom? If assistants are being used, then they
need to be trained to ensure consistency of collection. Sometimes
information can be collected or provided by the workers in an agency
dealing with the people who are the focus of inquiry. For example,
if you are undertaking a study of a victim–offender mediation
service, the mediators are likely to be interviewing referrals anyway
and can explain that they want to ask questions that will also be
used for research purposes. Similarly, in a study of offenders on
employment and training programmes, much of the data may be
supplied by the project staff completing forms when people start and
leave the scheme. However, in such situations it is necessary to
consider that the relationship between the worker and client is not

JOBNAME: Mcgraw−crowsemmens PAGE: 19 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 543EAB3A
the same as that between researcher and respondent, and people may
be reluctant to tell project workers certain things. It also has to be
borne in mind that while such workers may be practised at what
they do, they are not necessarily trained researchers, sensitive to the
requirements of research.
A similar principle applies to another form of data collection,
where information is abstracted from agency records. Information
collected for court or hospital files is collected to meet the
requirements of the agency concerned, not those of researchers (see
Garfinkel 1967, Chapter 6, ‘Good organizational reasons for “bad”
clinic records’, for a good illustration of this). Much sifting of
material may need to be done and certain key items of information
may not be available, or may not be in a form that is most
convenient for researchers. For example, in a study of sentencing in
magistrates courts one vital piece of information may be the previous
convictions of offenders. However, it is quite possible that this
information will not be in court files, having been retained by the
police or Crown Prosecution Service after the case was dealt with.
Hence these agencies may also need to be approached.
In addition to the foregoing, it is also worth considering the
circumstances in which data is collected. Will it be in the street, on
the doorstep, in offices? Who will be interviewed if, for example, a
family or an agency or a business is involved? The mode of
collection has implications for the reliability and validity of the data,
and for the likely level of response. Respondents need to be assured
of privacy and confidentiality or the information is unlikely to be
valid. If questions are asked in different ways by different people,
then the information is unlikely to be reliable.
Response rates are important because those who do not respond
to a questionnaire may well be different in some way from those
who do, and this therefore biases the results. A low response rate
reduces your ability to generalize from the sample you have selected
to the population from which it is drawn. One example of this is
responses to surveys of voting intentions. If those likely to vote for
one particular party also don’t like being asked how they will vote,
then this means that any extrapolation from the survey to likely
election results may be wrong – which is what happened during the
1991 General Election. Inevitably those who have committed
offences are reluctant to disclose this to a stranger, which makes it
difficult (but not impossible) to do accurate surveys of offending
rates based on self-report. Postal questionnaires are particularly
prone to low response rates if people are not prepared for them.
Efforts to overcome the response rate problem have included
offering to pay respondents or to offer vouchers or prizes to those
who participate. This lends itself to accusations that those who take
up such offers will produce unreliable responses because they are

JOBNAME: Mcgraw−crowsemmens PAGE: 20 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 4E99E155
doing it for the ‘wrong’ reason, or may give answers that they think
will please the piper who is calling their tune.
The inevitable question that follows is, ‘What constitutes an
acceptable response rate?’, and the inevitable reply is that there is no
definitive answer. There is a tradition in social research that in a
survey of the general population, using personal contact based on a
sound sampling frame, somewhere around 60 per cent would be the
minimum that would be regarded as acceptable (Marshall 1994:
358), but this means that four out of ten people in the sample are
ignored, and this is a substantial amount of uncertainty. In crimi-
nology, the conditions for obtaining a good response are sometimes
not ideal because you may be dealing with populations with whom
it is difficult to establish contact. However, the answer to this is not
to accept a poor response rate but to take steps when designing the
research to minimize the likelihood of non-response. This may mean
contacting people through an intermediary, such as an agency that
they are in contact with. For example if you wanted to contact
known offenders, this can be done through the probation or prison
services. Alternatively agencies who work with victims or offenders
may be a point by which access can be obtained.
Drafting the questionnaire
Content
As we have emphasized in earlier chapters, in any empirical
investigation, what goes into a questionnaire is determined by the
theoretical basis of the inquiry, and the hypotheses that are being
explored or tested. So you start by asking what information is
needed to test or develop hypotheses. Ideally each question consti-
tutes a variable,3 some of which will be dependent variables, some
independent variables, and some intervening variables (see Chap-
ter 3). Some variables will be demographic or classificatory data,
such as age, sex, ethnic group, marital status, and so on. These are
likely to form the independent variables against which dependent
variables are compared. Other questions will cover the things that
you are most interested in exploring, and these are likely to
constitute the dependent variables. So, for example, a very simple
survey might be interested in people’s views about various aspects of
crime and related issues. This might lead to hypotheses such as:
+ whether younger people are more in favour of legalizing cannabis
than older people;
+ whether people from higher income groups are more liberal
towards convicted offenders than those from lower income
groups;

JOBNAME: Mcgraw−crowsemmens PAGE: 21 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 50987E3F
+ whether men have more confidence in the police than women.

In each of these examples the characteristics of the respondent (age,
income, gender) form the independent variables, and their views
about various topics the dependent variables. There may also be
some intervening variables that might be taken into account. For
example, in comparing people’s views about whether cannabis should
be legalized, it might be important to consider whether or not
respondents have themselves used cannabis, and look at the effect
this has alongside variations in age – a three-way analysis. It is also
worth noting that an item might be an independent variable in the
context of one hypothesis, but a dependent or intervening variable in
relation to another hypothesis. For example, in the second hypoth-
esis above about income and views about offenders, income is the
independent variable. But if our hypothesis were that in general men
earn more than women, gender is the independent variable, and
income is the dependent variable.
In constructing a questionnaire, consider why each question is
there. Whatever the subject matter, it is always a sound principle to
go through the final draft and ask yourself why you are asking each
and every question, how you are going to use it in the analysis, and
what you are going to do with the results. This is partly a theoretical
matter (every question should have a theoretical justification some-
where along the line), partly a pragmatic matter (there is little point
in having an unnecessarily long questionnaire where much of the
material is wasted), and it also has an ethical component (why are
you asking somebody something unless it is really necessary?).
Sequence
When drafting a questionnaire, it is quite important to think about
the order in which questions will be asked. There are a variety of
reasons for this:
1 One of them is quite simply to think about the order that will run
most naturally and be most comfortable for respondents. One of
the problems here is that ideally you don’t want to start by asking
people to whom you are likely to be a stranger a lot of personal
questions. On the other hand, it may be important to enquire
about people’s age and background in order to ascertain whether
they fall into certain categories that are important for what
follows. Furthermore, it’s not a bad idea to get some key details
recorded in case the questionnaire is not completed for some
reason.
2 Another reason for giving careful thought to the sequence of
questions is what is referred to as routeing. Some questions may
be not applicable to all respondents and they need to be directed
to the next question that is appropriate for them, for example:

JOBNAME: Mcgraw−crowsemmens PAGE: 22 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 4C5C1B41
Have you been burgled within the last six months?

If yes, go to Question X; If no, go to Question Y.
These are filter questions which involve loops. It may be useful to

draw up a flow chart to check that respondents will be taken
through the questionnaire in the right sequence and not find
themselves with nowhere to go.
3 The timing of events. If, as an example, you are interviewing
someone about their history of employment and offending, it may
be necessary to have a question sequence that takes them through
a sequence of time, such as since they left school, or over the past
three years. This could be done diagrammatically by showing
them a dated timeline. Remember that you also need to think
about how it is going to be analysed subsequently.
4 Multiple events. Some people may have experienced something
several times. For example, if you are inquiring into people’s
experiences of being victimized or a series of fines that led up to
them becoming defaulters, some people may have had several
such incidents. If several experiences are to be recorded, you may
need to take someone through the same sequence of questions
several times, once for each occasion. It may be worth considering
whether details are needed about all the incidents, or perhaps
those within a given period of time (such as within the past year),
or the last three such incidents, or just the most recent one. If
you have a good, sizeable probability sample, then although you
may miss important experiences for particular individuals, the
picture will be representative of the sample as a whole. Again,
thought needs to be given to how a series of incidents will be
analysed.
5 Finally, there is the fact that the order in which questions are
asked might affect the way people respond to them. For example,
if you were to ask people about the effects that burglaries have on
their victims and then ask them about what sentences they think
burglars should be given, you may well get different responses if
you had first asked people what sentences burglars should get and
then asked them about how they think burglary affects victims
(see Schuman and Presser 1996, for more on such matters).
Question construction
There are a number of important rules about question composition:
1 Avoid leading or biased questions. Seldom, if ever, is a research
inquiry ‘objective’ in the pure sense, since the very fact that some
questions are asked rather than others means that you are
approaching a topic from a particular perspective. However, every

attempt should be made to make questions themselves as neutral

as possible. Thus, ‘Are you for or against hanging?’ is to be
preferable to ‘Are you in favour of hanging?’.
2 One idea per question. A common fault when trying to keep
questionnaires as compact as possible is trying to get too much
out of each question. The result can be an overly complex
question. For example, don’t ask, ‘How do you usually get to and
from work?’ People may have different ways of getting back from
work than when going to work. Two questions are needed rather
than one here.
3 Use simple sentence construction, clearly defined terms, and avoid
jargon. The more complex a question is, the more chance of a
misunderstanding occurring.
4 Beware of ‘social desirability’ effects. For example, ‘Are you in favour
of recycling?’ is unlikely to elicit a negative response, but may not
reflect people’s true views or behaviour. One study has found that
children develop the ability to control their responses so as not to
express their ethnic prejudices publicly (Rutland 2005; Rutland et
al. 2005). Similarly, avoid questions with ‘obvious’ answers, such as
‘Are you afraid of crime?’ or ‘Are you against cruelty to children?’
A better way to approach such matters is to be more specific, or
ask hypothetical questions, or you could put such topics in a list
of things that might concern people and ask them to say which
they regards as most significant.
It is a good idea to put yourself in the position of the respondent
and think about how you might answer a question.
Response options
As mentioned earlier, one of the main advantages of questionnaires
is that they are a way of getting a lot of information as efficiently as
possible. The most common way of obtaining responses to question-
naires is to ask people to tick boxes or circle options which have
been pre-determined and pre-coded. But there are in fact a variety
of ways in which responses can be recorded, see Box 7.1.
Each of these response types has its advantages and disadvantages.
One thing that is important for all of them is to make sure you tell
respondents what you want them to do: ‘tick one only’, ‘tick any that
apply’, ‘rank in order’. Where you want people to tick only one
option (as in number 2) ensure that the categories are indeed
mutually exclusive. When given the option of ticking multiple
responses (number 3) note that each option becomes a variable,
which will be coded as ticked or not ticked (usually coded 1 or 0).
The ranking option (number 4) can produce some confused
responses and be more difficult to analyse. For example, how are you

JOBNAME: Mcgraw−crowsemmens PAGE: 24 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 08125AB2

JOBNAME: Mcgraw−crowsemmens PAGE: 25 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 1F28757A
going to compare someone who has ranked three of the options to

someone who has ranked all six (and perhaps added the ‘Other’)?
The scale or grid (numbers 5 and 6) are generally a much better way
of gauging the strength of feeling because you then have an ordinal
scale which enables you to compare scores quite easily. These are
usually used for exploring attitudes, which often form the dependent
variables in a survey. A grid is a way of integrating several scales,
and more will be said about attitude measurement shortly.
Response order effects

Just as the order in which questions are asked can affect the kind of
responses you might get, so the order in which response options are
presented can have a similar effect. Some research suggests that
response options that come towards the start of a list tend to get

favoured more, a primacy effect, whereas other research suggests a

recency effect, where options that are given towards the end of a list
are more likely to get chosen (Schuman and Presser 1996). The
latter effect is more likely to happen where there are just two
options.
Open-ended questions
Questionnaires are usually quite highly structured instruments. But
there is often a wish to include open-ended questions, where the
respondent can say what they like. It is perfectly reasonable to
include such questions, as long as you think about why they are
there. They give respondents a feeling that they have been able to
get across what they want to say in their own words rather than just
being treated like a ticking machine, and they may give the
researcher the reassurance that they have not overlooked some
important view or experience. However, they need to be used with
care. If you feel the need to put a lot of them in a questionnaire,
then you should be asking whether a structured questionnaire is the
right tool to be using at all; perhaps a less structured interview is
more appropriate.
It is also important to be clear about what you are going to do
with the answers to open-ended questions. Are you just going to sift
through to check for responses that haven’t been included under any
of your pre-coded categories? Are you going to go through and
post-code responses into categories? Are you going to transcribe and
qualitatively analyse the information, or perhaps use some of the
responses as quotations to illustrate points that will be made in the
report? All too often what actually happens is that open-ended
responses are simply never analysed and end up getting left on one
side until the questionnaires are shredded. Similar considerations
apply to the ‘Other’ option used with more structured responses:
what are you going to do with what you get?
An illustration of the kind of effect that the form of the question
can have on responses is given in a report of an experiment
conducted in America in 1974 (Schuman and Presser 1996), (see
Table 7.2). The columns on the left show the percentage of
responses received to options that respondents were presented with
by the researcher. The columns on the right show the percentages
for various responses when participants were asked an open question.
Thus, when presented with prepared options, crime and violence
were regarded as the biggest problem by far, but when respondents
were allowed to choose their own responses, unemployment was the
problem that most concerned people, and there was a more even
spread of opinion.
Apart from the response types described above, there are a
number of other things to think about when constructing a

JOBNAME: Mcgraw−crowsemmens PAGE: 27 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 07FF9975
Table 7.2 Question form

JOBNAME: Mcgraw−crowsemmens PAGE: 28 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 50F64D0E
questionnaire. This includes situations where a particular question is

not answered for some reason, or where the answer is ambiguous:
+ Not Applicable. There are various reasons why you may need to
include this in your response options. The most common is when
you have used a preceding filter question. ‘Have you been burgled
in the last year? If no, go to Question X.’ All the questions
between this one and Question X need to have a Not Applicable
option for those who have not been burgled. Fairly obvious, but it
can be overlooked.
+ Don’t Knows. Not surprisingly some people can’t make up their
mind, can’t remember, or just don’t know the answer. It is
important not to ignore them because sometimes the ‘Don’t
Knows’ can tell you a lot. For example, if you are comparing
people’s attitudes towards sentencing with their knowledge of the
criminal justice system, and you ask questions such as, ‘How likely
do you think it is that a burglar will be sent to prison?’, then the
proportion of people who don’t know can be quite telling. People’s
knowledge about criminal justice (or the lack of it) may be related
to their attitudes. Apart from examples like this, if you get a lot
of ‘Don’t Knows’, then it is probably a weak question and you
really need to reconsider it. This is one reason why it is a good
idea to do a pilot study with a group of people who are similar to
those who will take part in the main study, with enough cases to
do a pilot analysis (referred to again later). If you don’t offer a
‘Don’t Know’ category in an attempt to force people to make up
their minds you might get a lot of cases where people don’t
answer at all.
+ No Answer. What do you do if someone simply does not give an
answer? This is more likely to happen with self-completion
questionnaires, but even if you or others are administering the
questionnaire you need to know if someone chooses not to answer
a question. In order to decide what to do, you have to be able to
interpret what the lack of an answer means. It could be another
way of saying ‘Don’t Know’, or that it is ‘Not Applicable’, or that
they simply couldn’t be bothered to answer the question.
With all these kinds of responses you have to be careful how you
code them. If you leave them blank, data analysis packages such as
SPSS will treat it as missing data, and sometimes it may not be. If
you code zero, then make sure this does not conflict with a genuine
zero value, e.g. how many times did you go to the cinema last year
– zero or blank could be not at all or don’t know. In the case of
missing data or ‘Don’t Knows’ you may want to exclude them from
the analysis for that question. But bear in mind that if you do so it
will have an effect on the remaining data. Thus, if 10 per cent of the
cases have said ‘Don’t Know’ and they are excluded from analysis,

JOBNAME: Mcgraw−crowsemmens PAGE: 29 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 3E3D3E98
the percentages for other categories will go up correspondingly, and

the analysis of the relationship between the remaining answers and
other variables is more likely to be strong and statistically signifi-
cant.
The middle position

Questions that ask respondents to rank something, or scales which
range from ‘Strongly Agree’ to ‘Strongly Disagree’, often have a
middle position. But this middle position can be problematic. First,
as with ‘Don’t Knows’, having a middle position at all can affect
how people respond to the question. Not having a middle position
means that people will be forced into making a choice or not
answering at all. If there is a middle position then it can mean:
+ a lack of opinion: a middle position can be a form of ‘Don’t
know’;
+ a positive wish to adopt a position of neutrality;
+ a conservative opinion indicating ‘No Change’, ‘About Right’ or
‘Stay As It Is’.
So it may be quite important how you label the middle position. It
could say ‘Neither Agree nor Disagree’, and one could have a neutral
position and an additional box at the end labelled ‘Don’t Know’.
Researching attitudes
An attitude is a state of readiness, a tendency to act or react in
a certain manner when confronted with certain stimuli.
(Oppenheim 2000)
An orientation towards a person, situation, institution or social
process, that is held to be indicative of an underlying value or
belief. (Concise Oxford Dictionary of Sociology, definition of
‘attitude’ 1994)
Like several of the topics dealt with in this book, attitudes and the
study of attitudes deserve much more space than we can give them.
They are an important branch of social psychology, and we can only
give the kind of brief introduction that deals with the ways that they
can be used in researching criminology. Attitudes have several
components:
+ beliefs – a cognitive (thought) component;
+ feelings – an ‘affective’ (emotional) component;
+ behavioural – a potential for acting in a certain way.

JOBNAME: Mcgraw−crowsemmens PAGE: 30 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 419AC1A3
They can also range from being very ephemeral to longer-lasting

and more deep-seated. Thus, there can be said to be spectra of
attitude, from
+ specific to general
+ changeable to stable
+ superficial to profound
+ opinions to ideology.
When undertaking criminological research, we often want to know
about what people think about such things as crime in general,
criminal justice policy, and specific aspects of criminal justice. This
can be done in a qualitative manner by open-ended, semi-structured,
and group interviews. However, one of the most economical ways of
enquiring about people’s attitudes is to use some kind of a scale,
which can form part of a questionnaire. This is what we focus on
here.
Types of attitude scale

Inferences about peoples’ attitudes are most usually made on the
basis of measuring responses to questions or statements that require
respondents to indicate whether they have positive or negative
feelings about the attitude object: do they favour or oppose, prefer or
not prefer, agree or disagree? A scale is usually a list of such
statements, or adjectives, which people respond to. Only a brief
consideration of some of the best known attitudes scales is given
here:
+ Bogardus social distance scales
+ Osgood semantic differential scales
+ Guttman scales
+ Thurstone scales
+ Likert scales.
Bogardus social distance scales

These involve statements about how close respondents feel to certain
people, objects or ideas, for example:
1 I wouldn’t be seen dead in a Skoda.
2 I might be prepared to be given a lift in a Skoda in an
emergency.
3 I’d be happy to be a passenger in a Skoda.
4 I’d be prepared to drive a Skoda.
5 I would be happy to own a Skoda.
If asked to tick one of these, the higher the number you tick, the
lower the distance between you and a Skoda.

JOBNAME: Mcgraw−crowsemmens PAGE: 31 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 4C5DD638
Osgood semantic differential scales

These consist of bipolar adjectives such as good–bad, hot–cold,
dark–light, fast–slow, with a range of points between the two
extremes. They give a profile of your orientation towards someone or
something. For example, do you see the leader of a particular
political party as warm or cold, generous or mean, active or passive,
etc?
Guttman scales
Also known as scalogram analysis, these are scales indicating an
increasingly positive attitude towards the object under investigation.
Items on a Guttman scale contain statements which show an
increasingly positive attitude towards the object under investigation.
Their distinguishing characteristic is that an endorsement of any
statement implies the endorsement of every less positive statement.
The items on a Guttman scale have the properties of being ordinal
and cumulative.
Respondents indicate whether they agree or disagree (endorse, do
not endorse) a series of both positive and negative statements. The
score is the number of statements endorsed. For example, in
arithmetic, addition, subtraction, multiplication and the extraction of
square roots are increasingly more difficult things to master. It is
highly likely that anyone who can multiply can also add, and anyone
who can extract square roots can both multiply and add. If we think
of a number of items ranked in order of difficulty, then many
respondents will endorse the early ones, for example, indicating that
they know how to add, subtract, multiply, etc. However, sooner or
later they will ‘cross over’ and fail to endorse the remaining items,
for example, solving differential equations or carrying out integra-
tions. This cross-over point is their individual score.
The advantage of a Guttman scale is that the score provides
detailed information on the total pattern of results. Because of this
it is possible to measure any change in attitude over time, because
change becomes an additive score. A higher number than before is
more positive, a lower number less positive.
The main disadvantage of a Guttman scale is the large amount of
quite difficult work involved in producing it. Items are measured on
the basis of a ‘coefficient of reproducibility’ and a ‘coefficient of
scalability’. A large pilot test must be carried out to gather enough
information to compute the two coefficients. Without going into the
statistics involved, one coefficient checks that the items can be
graded into an increasingly positive manner, so that everyone who
responds positively to the 3rd item also responded favourably to the
1st and 2nd items. The other coefficient measures the possible error
for reproducing the result.

JOBNAME: Mcgraw−crowsemmens PAGE: 32 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 4A93C79C
Thurstone scales
Thurstone scales cover all the points along a continuum from
strongly positive to strongly negative, with equal intervals between
the points on the scale. Thurstone’s preoccupation was with the
problem of equal, or rather equal-appearing, intervals. To develop a
proper Thurstone attitude scale requires a lot of development work
and statistical knowledge. This involves getting a panel of people to
compare attitude statements two at a time and deciding which of
the pair is more positive or negative.
Thurstone equal interval scales are formed by the following
process:
1 Having a panel of judges who sort several hundred statements
about a topic into eleven piles ranging from the most positive to
the most negative.
2 Identifying a smaller number of statements (usually between 10
and 20) that are characterized by a high level of agreement
among the judges and are spread evenly across the 11 piles.
3 The respondents are instructed to check the two or three
statements that are closest to their own positions.
In this way, the respondents are ordered according to the average
value of the items they have endorsed. This method of scale
construction is relatively laborious and is not widely used by social
researchers.
Likert scales
Likert scales are probably the most common type of attitude
measure. However, many scales today are only loosely based on the
ideas of the Likert scale. Like Thurstone scales, they cover all the
points along a continuum from strongly positive to strongly negative.
But Likert’s primary concern was with uni-dimensionality – making
sure that all the items would measure the same thing.
Items are prepared which can be responded to as either favourable
or unfavourable towards the object or idea concerned. Usually a
five-point response pattern is used: strongly agree, agree, not sure,
disagree, strongly disagree. A numerical value is assigned to each
response, such as –2 for strongly agree and +2 for strongly disagree,
with the mid-point being scored zero.
An 8-step procedure is used for constructing a full Likert scale:
1 Items favourable and unfavourable to the issue are written.
2 Independent judges examine each item to classify whether items
are negative, positive, or neutral.
3 Statements unclassifiable or neutral are excluded.
4 Remaining items are placed in random order on paper.

JOBNAME: Mcgraw−crowsemmens PAGE: 33 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 51FE2062
5 Items are piloted tested with a sample group representative of the

population under investigation.
6 Statistical correlations are made between the scores of each item
and the total summated score.
7 For an item to remain in the scale, the correlation between it and
the total scale score must be statistically significant. All other
items are dropped.
8 The final form of the scale is prepared with the remaining items.
A common modification that has evolved over the years is to use
attitude statements in a questionnaire, but often these statements are
unrelated to each other and one statement is given the job of doing
all the assessing of an attitude. No matter how well written or well
presented, one statement is rarely going to encompass all the twists
and turns of people’s attitudes, so Likert-style attitude statements are
not the same as using a properly formed Likert scale.
Likert scales, or Likert-type scales, are popular because they are
fairly easy to construct and easy to complete. The main difficulty is
that because it is a summated score you can get the same total score
with several different response patterns. In that sense Likert scales
provide a good measure of consistent attitudes but are not capable of
measuring attitude change to the same extent that Guttman and
Thurstone scales do. In practice, Likert scales tend to perform well
when it comes to a rough, but reliable ordering of people with
regard to a particular attitude. These scales provide more precise
information about a respondent’s degree of agreement and disagree-
ment. Also it becomes possible to include items that are not
obviously related to the attitude in question, enabling subtler and
deeper ramifications of an attitude to be explored.
As a guide to a shortened form of developing a Likert-type scale,
first produce preliminary hypotheses. Next, explore the topic area by
studying the literature, and undertaking some in-depth interviews to
get the actual words that people use. It is also worth using items
that have already been used by other researchers: (1) to reduce the
amount of work that you have to do to create a valid scale; and (2)
to enable comparisons. Next, compose an item pool. These are
questions or statements derived from exploratory work that appear
to measure a particular theoretical construct. Finally, items are
selected for inclusion on the basis of how strongly they contribute to
a common underlying factor. Box 7.2 shows how a simple Likert-
type scale might look in a questionnaire.
In the questionnaire all the statements would be in the same
typescript. However, three of them have been written here in italic.
You don’t have to think about it too long to realize that the
statements in italic are about what could be done to reduce burglary,
while the statements in ordinary script concern what might be done

JOBNAME: Mcgraw−crowsemmens PAGE: 34 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 2C39B5AD
with burglars. We have shown it this way in order to demonstrate a

common feature of such scales, which enables them to be used to
explore different attitude ‘dimensions’. In analysis it might be
interesting to explore the hypothesis that one dimension is related to
the other.
Presentation and administration of questionnaires

The most important aspects of a questionnaire are its theoretical
basis, the hypotheses it will test and where it sits in the research
design as a whole. Nonetheless, there are some minor issues that can
have a big effect on how it is answered, whether it is answered, and
what kind of results you get.
Length
A questionnaire needs to be long enough to do its job, and on the
whole, having agreed to answer a questionnaire, most people will go
through with it. But it is also the case that most people have a
limited amount of time, and a limited attention span. So always
keep questionnaires as short as possible. There is no hard and fast
rule about how long a questionnaire should be. A survey where
people are stopped in the street needs to be no more than a few
minutes, whereas a pre-arranged questionnaire interview with some
who works in a criminal justice agency could be up to an hour. An
illustration of what can happen if a questionnaire is too long is given

by a survey of schools done for a community safety programme

known to us. The pupils were asked to complete the questionnaire
in class, so apart from those absent that day the response rates were
very good, in the region of 85–90 per cent – for questions at the
start of the questionnaire, that is! Unfortunately the questionnaire
was so long and complicated, with lots of questions asking much the
same kind of thing, that by the time pupils got to the end of the
schedule, up to 15 per cent of them were no longer responding to
the questions towards the end.
Layout
Use a clear layout and consistent format. Allow plenty of space so
that people don’t miss a question and have room to write in their
replies. Give clear instructions about what they need to do.
Explanation
Introduce the questionnaire properly. A written covering document
should be given to every participant, in advance if possible. Its
purpose should be as follows:
This should tell people what the research is about.
+ It should explain that you need their consent to do the question-
naire. Whenever possible, there should be a consent form which
they sign. At the very least the document should tell them that if
they do agree to complete the questionnaire, then this will be
taken as having given consent to participate in this part of the
research.
+ They should be told that the information is confidential, that they
will not be identified, that the information will be kept securely,
and will be destroyed after analysis.
+ They should be given details of who you are and, if a face-to-face
interview, your identification should be available.
+ They should be told who to contact if they have any queries or
concerns, and how.
+ It may be advisable to tell respondents that the project has been
approved by an ethics committee (and give details of the
committee).
+ If identifying details (such as name, date of birth, address) are
obtained (e.g. to enable a follow-up survey later), participants
should be told that they will be kept separately from the
questionnaire itself, using a cross-reference for the two.
+ They should be told how the results will be used, and wherever
possible participants should be given feedback on the results, or
told where they can find them.

JOBNAME: Mcgraw−crowsemmens PAGE: 36 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 53D0D0D8
Finally, remember that you are dealing with people, not data. Be
respectful. If possible, give respondents an opportunity to ask any
questions they may have, and thank them.
Piloting and analysis

We have mentioned several times that when questions are being
drafted you need to think about how they are going to be analysed.
Although we consider analysis further in Chapter 8, it is worth
emphasizing that this is something that needs to be considered at
the start of an inquiry, not left until there is a pile of material to be
dealt with, only to find it would have been better to have asked the
questions a different way. The data needs to be collected in a form
that is suitable for whatever analysis is to be carried out. In the case
of quantitative data, this means thinking about the appropriate
coding categories before the data is collected.
The matter of analysis, along with a number of the other issues
considered here, often comes most sharply into focus once the actual
process of inquiry proper has started. However, by then it may be
too late to alter something. You only get one go at data collection –
nobody wants you coming back again because you got it wrong the
first time. For this reason one of the most important parts of any
inquiry is piloting. Ideally a pilot study is a full ‘dry run’ in which all
the measures to be used are tested. It should involve people or areas
similar to those which will constitute the main study, and the results
should be analysed to make sure that the various codings and
statistical procedures produce the desired outcomes. If for some
reason this is impossible, then at the very least questionnaires and
interview schedules must be tried out. There is no such thing as a
perfect questionnaire, but some errors can be avoided by piloting.
Further reading
Several of the general social research methods books have chapters

on questionnaires including Bryman (2004), Burton (2000), Gilbert
(1993), and Kane (1985). However, the classic text on questionnaire
design and attitude measurement is that by Bram Oppenheim
(2000). Although the first edition was published in 1966, a more
recent version is available (1996) and is also very readable. Schu-
mann and Presser’s book on attitude surveys is very helpful on the
best wording for questions. Pat Mayhew’s chapter on ‘Researching
the State of Crime’ in King and Wincup, Doing Research on Crime
and Justice (2000), is a good introduction to the use of crime surveys
as she brings her own extensive personal experience to bear on this
topic.

JOBNAME: Mcgraw−crowsemmens PAGE: 37 SESS: 26 OUTPUT: Tue Aug 28 15:58:10 2007 SUM: 1398F30D
Regarding focus groups, Alan Bryman (2004) has a chapter, and

David Morgan’s work is also worth referring to, a 1996 article in the
Annual Review of Sociology being most useful. It is also well worth
reading the (2005) article by Överlien et al. in the International
Journal of Social Research Methodology, both for seeing how it was
used in relation to a group of young female delinquents, and for
their description of the specific techniques used.

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 40E55BB9
8 Analysing criminological
research
About analysis
The analysis of material collected in the course of a study can take

various forms, and involves a certain amount of detailed knowledge
of how to carry out particular kinds of analysis. For example,
statistical analysis requires not only a knowledge of the principles
and procedures required to undertake statistical tests, but a knowl-
edge of how to operate whatever software package is to be used.
Computer packages have also become increasingly employed in the
analysis of qualitative information. To master these techniques
involves becoming acquainted with statistics and with the procedures
needed to analyse non-numerical material. In this book we do not,
and cannot, offer a complete guide to statistics, to the use of SPSS
(one of the main means of analysing numerical information), or
software such as NUD*IST and Atlas (packages for analysing
non-numerical material acquired from documents or interviews).
This would require additional volumes.
In order to give readers an understanding of what is involved, and
how to go about data analysis, we will consider the general
principles and procedures of data analysis, and provide some illus-
trative examples. There are other texts which deal with statistics and
qualitative analysis in more detail (see Further Reading). Our aim is
to provide a grounding which will enable readers to progress towards
the more detailed study required to become familiar with specific
techniques.
One of the points we have made throughout this book is the
importance of thinking about the analysis of material at an early
stage in the research, and making appropriate provision for it. A
distinction is commonly made between analysing numerical infor-
mation (quantitative) and information in the form of words (usually

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 52102697
referred to as qualitative). However, we would like to start by

cautioning against making too clear-cut a division. Numbers often
hide a rich diversity of meaning, as, for example, when examining
people’s attitudes. What does it mean to say that a certain
percentage of a sample of the public favour being ‘tough’ on crime?
Conversely, documentary or interview information can also involve
counting things, such as the number of times the word ‘tough’ is
used. Words and numbers often complement each other in crimi-
nological research, and the competent criminologist will be adept at
analysing both.
One distinction that is worth making is that between descriptive
and inferential analysis. Whether words or numbers are being used,
there is a certain aspect of any study that involves describing what
happened, what was said, or how many times something occurred.
This is a necessary and valuable part of any inquiry. Beyond that,
however, there is the interpretation of information, and this is where
inferential analysis comes in. Inferential analysis involves using the
data collected to consider its wider implications. Usually these will
be related to hypotheses that are the subject of inquiry. Where
quantitative data are involved, then statistics are used to make
inferences about a population based on a sample drawn from that
population. For example, a descriptive statistic might tell us that the
mean number of previous convictions of a sample of offenders
convicted at a particular court is five. As long as the sample was
randomly drawn, it can be inferred that the mean number of
previous convictions of all those appearing at the court is also going
to be about five, give or take a bit. It is unlikely to be exactly five
because any statistical inference is an estimate based on a probability
that allows for a certain margin of error. This is known as the
confidence limit. Making inferences is about making generalizations.
Any criminological study hopes to be able to add to the pool of
knowledge about a topic. It therefore has to have regard to where
the information came from and the extent to which any generaliza-
tions are valid, and this will usually depend on what kind of sample
it is based. The extent of inference can be estimated numerically
where quantitative data are concerned, but this is not possible where
non-quantitative material is involved, nonetheless the same princi-
ples apply.
Analysis can take the form of either primary or secondary
analysis. Analysis is frequently thought of in terms of analysing data
collected by the researcher(s) themselves. This will often involve the
analysis of interviews, questionnaires or observations. However, as
we mentioned in Chapter 5, valuable research can be done using
existing data and sources. It might involve using official sources of
data, such as the criminal statistics and subjecting them to further
analysis than that which has been done already. It can also involve

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 535DC7D1
Analysing criminological research 155
drawing on records and archives, such as court records and historical

archives. Secondary analysis may also involve taking information
collected by other researchers and analysing it in a different way. Of
course, a study can do both, analysing some material already
available, and also adding to it by collecting further primary data.
Secondary analysis might involve using the ESRC’s Economic and
Social Data Service, based at the Universities of Essex and Man-
chester1 (also referred to in Chapter 5). This is an archive of social
science studies which the original researchers have deposited with
the ESDS, and requests can be made to re-analyse particular
datasets. The archive includes both quantitative and qualitative
material. The decennial census is another large source of information
which can be used by bona-fide researchers. For criminologists a
considerable amount of information is collected by the British
Crime Survey (BCS). The Home Office publishes and commissions
reports using this information, but cannot examine every facet of the
material collected. Chapter 12 describes a study that involved using
the BCS. Before assuming that new data needs to be collected, it is
well worth any criminologist’s while to consider what material might
already be extant that could be analysed to cast light on their own
research question.
We would also like to take the opportunity to nail what is
probably the most commonly used aphorism about statistics, that
there are ‘lies, damn lies and statistics’2. Statistics don’t lie; it is the
people who use them who either knowingly misuse them, or don’t
know how to use them. Wrongly applied statistics can be used to
support whatever argument you want. Properly applied, data can
help to evaluate competing claims. The criminologist needs to
acquire good statistical practice, and to recognize when bad practice
occurs in the work of others. This does not mean getting to know
lots and lots of different statistical procedures, but does involve
understanding the principles of good statistical analysis.
Analysing quantitative material: An introduction
The word statistics makes some people nervous, perhaps summoning

up memories of struggles experienced in school maths lessons.
However, statistics are an important part of criminology, and need
not be daunting. Those researching criminology do not necessarily
need to be experienced statisticians, but some appreciation of how
numerical data is analysed is important in order to understand the
possibilities and limitations of quantitative material – both your own
and that of others. We have already said something about official
statistics in Chapter 5. Here we look mainly at the primary analysis
of your own data. In doing this it is necessary to say something

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 3EBE3E76
about probability and statistical inference, because they are at the

heart of analysing quantitative material. However we will come to
this shortly. In a Glossary of key terms at the end of this book we
have included short descriptions of some of the main concepts
applied in doing quantitative analysis.
The Statistics Package for the Social Sciences (SPSS)

Although it is possible to do some analysis using just a calculator,
most numerical analysis is done on a computer using a statistics
package, such as SPSS, which is the one we will refer to. As we
have already said, we aim only to give an introduction to data
analysis and therefore do not set out to produce an SPSS user’s
manual. There are several texts available (see Further Reading at the
end of this chapter) which explain procedures in detail.
Like many computer programs, SPSS has menus and buttons at
the top of the screen which enable the user to perform various
operations, and when first started it shows a grid of rows and
columns. The rows contain individual cases, and the columns contain
variables. Let’s suppose we have been to the local court and
extracted from the files some information about 56 cases dealt with
in the last month, consisting of information about age, gender,
ethnicity, the offence, number of previous convictions and the
sentence. This information is shown in full in the appendix on p.
220. The variables have been simplified so that there are just three
broadly defined ethnic groups (1 = White, 2 = Black, 3 = Asian),
three offence categories (1 = theft, 2 = criminal damage, 3 =
violence), and three sentences (1 = a fine, 2 = community sentence,
3 = custody). For convenience the first six cases are shown in Table
8.1.
Thus, the first case is a 21-year-old male, white, who committed
an offence of theft, had one previous conviction (‘precons’ is the
number of previous convictions), and was given a fine. The last
Table 8.1 Court abstract data (first six cases)

JOBNAME: Mcgraw−crowsemmens PAGE: 5 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 51EA43E5
variable, ‘sscore’, is a calculation based on previous variables which

reflects the seriousness of the case. We will refer to this data again
in due course.
Clicking on a tab at the bottom of the SPSS screen brings up the
Variable View, which enables you to insert and look at details of the
individual variables. When an operation has been performed a new
screen will appear, the Output View, which gives the results of the
analysis. This output can be confusing at times because it often
contains a lot of information, and although it is all relevant as far as
the practised statistician is concerned, it can sometimes be difficult
for the novice to know which bits are the important ones to use in
a given situation. A good textbook will help to decipher this output
(see Further Reading at the end of this chapter).
We will give an overview of three main kinds of numerical analysis.
These are:
+ univariate analysis – looking at a single variable;
+ bivariate analysis – looking at two variables;
+ multivariate analysis – more than two variables.
Analysing a single variable
Analysing a set of data often starts by looking at individual variables
such as what the mean (average) age of the sample is, what
percentage are female and male, how many come from different
ethnic groups, and so on. However, there are important differences
between the types of variable involved, and this depends on the kind
of measurement involved. There are four types of measurement:
nominal, ordinal, interval and ratio.
Nominal variables are mutually exclusive categories such as gender
(male, female). In the court abstraction data referred to above, four
of the variables are of this type (gender, ethnicity, offence, sentence).
When analysing variables of this type it is usual to look at the
number of cases in each category (the frequency) and the percentage
of cases in each category. For example in the first six cases in the
court abstraction table there were five males and one female, which
means the percentage of females was 1 ÷ 6 × 100 = 16.7 per cent.
The sentencing for the first six cases can be described as shown in
Table 8.2.
Ordinal variables are ranked in order. For example: Unsatisfactory,
Satisfactory, Good, Excellent. In the example above, the three
sentences fine, community sentence, custody could be seen as an
ordinal scale of severity. These are often referred to as scales or
scores and assigned numerical values ranging from 1 for the
‘weakest’ to 2, 3, 4, 5, etc. for ‘stronger’ items. It is quite common
for responses to attitude questions to take the form of an ordinal
scale:

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 3B2BF108
Table 8.2 Frequencies and percentages
Q: Are you worried about going out alone at night?

A: Not at all – A little – A lot – Very
However, it is important to note that ordinal scales are very

imprecise measures, and the intervals between different points on
the scale are not necessarily the same; they are not equal intervals.
Thus, there may not be much difference between ‘A lot’ and ‘Very’
in the example above, but there might be a much bigger distance in
people’s minds between being a little worried and a lot worried.
Similarly many might regard the difference between getting a fine
and a community sentence as being not as great as that between
getting a community sentence and a custodial sentence.
Interval variables, on the other hand do have equal intervals
between each point on the scale (e.g. temperature, time), and ratio
variables have equal intervals, with a zero point. Examples here are
money, age, height, weight, and so on. The fact that there are equal
intervals means that more precise interpretations can be made: £20
is twice as much as £10; being sent to prison for 12 months is twice
as long as a prison sentence of 6 months. In the court abstraction
table shown earlier, there are two interval level variables, age and
number of previous convictions. When doing descriptive analysis of
these variables you can look at two main attributes:
+ where the central point is: a measure of centrality;
+ how the cases are spread out between the lowest point and the
highest point: their distribution.
Both are important in describing the characteristics of interval and
ratio level variables.
Measures of centrality
There are three main ways of describing where the centre of a
distribution is. The mean is the average of all values. It is the sum of
the values for all the cases divided by the number of cases. It is

usually shown in calculations as an X with a bar above it: X. Thus,

if we look at the average age of the first six cases given in the court
abstraction table we get:
X = total of all values = 21+17+24+30+19+25 = 136 = 22.67
no. in sample 6 6
The mean is the most commonly used measure, but it has the
drawback that a few cases with extreme values will have a big effect.
For example, the fines imposed by magistrates courts are generally in
the region of £100 to £150, with a mean of £125. However, every so
often courts will feel that, while a fine is the right disposal, there is
something about a particular case that warrants a heavier fine to
mark the gravity of the offence. So they may well impose a fine of
£500 or more. This will considerably increase the mean, even though
the great majority of offenders are still within the £100 to £150
range.
The median avoids this because it is the value above which half
the cases lie, and below which the other half of the cases lie (if there
is an uneven number of cases it will therefore be the one in the
middle). This can be illustrated by taking another example. When
awarding a degree classification, an examinations board looks at the
mean marks achieved by a student in their second and third years.
In borderline cases a wise exam board will also look at the median
mark, because one poor result could have an adverse effect on a
student’s overall pattern of performance, as shown in Table 8.3.
Six subjects are taken in each year. Marks for Year 3 are double
weighted to take account of improvement. Had the classification
been based solely on the mean, the student would have received a
lower second degree because of the mark of 30 in Year 3, but in fact
received an upper second when the median was taken into consid-
eration.
The mode is simply the most frequently occurring value. Although
it may not tell us much on its own, it does give an indication of
where most of the cases congregate.
Measures of dispersion
Another way of describing variables with equal intervals is how the
number of cases is distributed. Kurtosis is a measure of how closely
cases cluster around a mean. If there are a lot of cases close to the
mean then the distribution will be tall and thin and is called
leptokurtic. If the cases are widely spread, the distribution will look
much flatter and is called platykurtic (see Figure 8.1).
Skewness describes the extent to which a distribution is asym-
metrical. If most of the cases are to the left of the mean, then the
distribution is said to be positively skewed. This would be the case if,

JOBNAME: Mcgraw−crowsemmens PAGE: 8 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 2B3E2580
Table 8.3 Comparing mean and medium
for example, one was examining the age distribution of a sample of

offenders and there were a lot of young people in the sample. If
more of the cases are to the right of the mean, then the distribution
is said to be negatively skewed (see Figure 8.2).
If a distribution is very evenly spread, then it is said to be
normally distributed (see Figure 8.3).
In real life, it is rare to find a distribution for a variable that is
like this. The normal distribution is an ideal distribution, which has
a symmetrical, bell-shaped, curve. It is important in statistics,
however, because it has certain properties. The attributes of areas
under the curve are known, and can therefore be used as a basis for
making comparisons.3
Two commonly used measures of spread are the variance and the
standard deviation. If you subtract the value of every case from the
mean for the distribution, square this value, add them all up, and
divide by the number of cases minus one, then you get the variance
for the distribution. If you take the square root of this value, you get
the standard deviation. These measures are often used in performing

JOBNAME: Mcgraw−crowsemmens PAGE: 9 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 553D0C26
various statistical tests, and SPSS will do the necessary calculations.

An example is given in Appendix 2 on p. 222, and more details can
be found in one of the books on using SPSS mentioned in the
Further Reading at the end of this chapter.
In the curve in Figure 8.3, the solid central line represents the
mean. The pair of dashed lines either side of this represent one
standard deviation either side of the mean. The outer pair of dotted
lines beyond the first pair represent two standard deviations either
side of the mean. In a normal distribution approximately two-thirds
of cases (68.26) lie between the mean and one standard deviation
either side of the mean. Approximately 95 per cent of cases (95.44)
are between plus and minus two standard deviations of the mean.
This provides a common standard with which to compare scores
taken from different distributions.
Although this is taking us into some detail, these distributions
and measures are very important in understanding how statistics can
be used in criminology and other areas of social policy; they are the
building blocks of subsequent quantitative analysis.
It is also important to appreciate the significance of these
different types of variable, because they affect what you can do with
the information they represent. In essence, it is possible to do more
with equal interval variables like age, and money and time. There are
more different ways of examining them, and they meet the require-
ments for more complex analyses. There may be times when it is
necessary to ask respondents for their age group (20–29, 30–39, etc.)
or income group (less than £10,000 per annum, £10–14,999 p.a.,
etc.) for reasons of sensitivity or convenience, but it is generally
better to get actual ages, times, and money whenever possible. The
actual values can always be grouped later. Ordinal variables, in
particular, need to be treated with care because they look as though
they can be used like interval level variables, for example using the
mean of a scale, when in fact this is not a ‘real’ mean.4
Inferences from single variables

Referring to the distinction we made earlier between descriptive and
inferential analysis, the examination of single variables may be purely
descriptive. However, when you then seek to apply what has been
found to a wider population on which the data is based, this
involves inferences from a sample to a population. Sometimes such
descriptive material can be important and revealing in its own right.
For example, from time to time there have been surveys of the
prison population (Walmsley et al. 1992; Niven and Olagundaye
2002). These tell us such things as the proportion of people in
prison who have been brought up in local authority care, those who
are likely to have accommodation problems on release, those who

JOBNAME: Mcgraw−crowsemmens PAGE: 10 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 0BDD0575
Figure 8.1 Kurtosis
Figure 8.2 Skewness
Figure 8.3 Normal distribution

have reading and writing difficulties, and so on. Such information

has important policy implications.
Making inferences about a population based on a sample drawn
from the population involves an element of uncertainty. If you were
to draw several samples from the same population, they wouldn’t all
have the same characteristics, even if they have all been drawn at
random. This is partly a matter of chance (probability) and also
because there is always the possibility of errors creeping in one way
or another. So any extrapolation from a sample has a margin of
error. This is allowed for in any numerical analysis and the margin
of error can be estimated from a random sample. It is referred to as
the standard error. This is something that is used in quite a few
statistical calculations.
Let’s suppose that we go for a walk one day. When we get back,
we are asked how far we walked, and we say ‘six miles, give or take
a mile’. This ‘guesstimate’ means that we are saying that we may
have walked as little as five miles or as much as seven miles. This
means that we have what is referred to in statistics as a confidence
interval of plus or minus one mile. When doing proper quantitative
analysis we can be more precise. The confidence interval is plus or
minus two standard errors either side of the mean. The example
given in Appendix 2 on p. 222 shows how this would be calculated
using the example of the number of previous convictions of a sample
of offenders. In the example, the mean number of previous convic-
tions of a sample of 294 offenders was 1.2. By calculating the
standard error of this mean, we can be 95 per cent certain that the
true mean of the population from which this sample was drawn lies
somewhere between 1.06 and 1.34 previous convictions.
Statistical inference can also be used to estimate whether a sample
is representative of the population from which it is drawn. Let us
say that it is known that for the prison population nationally the
mean number of previous convictions is 4.8. If we take a random
sample of cases from a local prison, we would not necessarily expect
that the mean number of previous convictions for those in that
sample would be exactly 4.8. It might be a bit more, or it might be
a bit less. But exactly how much is a ‘bit’? Let’s say that the mean
number of previous convictions for our sample is 6.3. Is our sample
mean such that it is not significantly different to prisoners generally,
or is our sample significantly different, for whatever reason (e.g.
perhaps it is a prison that has a particularly hardened bunch of
criminals in it)? By calculating the standard error for the mean
number of previous convictions of our sample, and the confidence
interval for this mean, we can determine whether the national figure
lies outside these limits, and therefore whether our sample is
significantly different. If it is, we might then be interested in doing
further investigations to find out why this might be.

JOBNAME: Mcgraw−crowsemmens PAGE: 12 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 34A87C68
Analysing two variables

So far we have been looking at the analysis of one variable at a
time: univariate analysis. However, much of the most interesting
research comes from looking at the relationships between two or
more variables. In this section we will briefly consider two variable
analysis. This mainly involves two kinds of question. Is there a
difference between one thing and another, and is there a relationship
between one thing and another?
Nominal variables
Here again the distinction between different types of variable is
important. The analysis of variables that have categories (nominal
variables) is different from those that have equal intervals. Ordinal
variables also need to be treated differently. Let’s look first at
categories, and ask the question, do some groups of people get
sentenced differently from other groups of people? Do women get
sentenced differently from men? Do people from different ethnic
groups get different sentences? Do those who have jobs get different
sentences to those who are unemployed (something discussed in
more detail in Chapter 9). Sentence, gender, race, and employment
status are all categorical variables. Using the court abstraction data
referred to earlier, let’s look at race and sentencing for the 56 cases.
Table 8.4, produced by SPSS, is a simple descriptive cross-
tabulation of the two variables. It tells us, among other things, that
six of the 28 White people in the sample got a custodial sentence,
ten of the Black people, and none of the Asian people.
Table 8.4 Sentence by race cross-tabulation

JOBNAME: Mcgraw−crowsemmens PAGE: 13 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 3739DD34
Table 8.5 Sentence by race, with percentage
It becomes somewhat more informative if we add in the percent-

ages (see Table 8.5). We can now see more clearly that less than a
quarter (21.4 per cent) of the White offenders got custody, com-
pared to exactly half of the Black offenders. However, what we now
need to know is whether this is a difference that could have
occurred by chance, or is the difference such that it needs to be
explained in some other way, because it is not a statistical artefact.
To explain this a bit more, we might be surprised if exactly the
same percentage of people always got a custodial sentence, regardless
of what group they were in, or whatever the sample that was drawn
randomly. Tossing a dice 36 times, we would not necessarily expect
to get exactly six 6s, six 5s, six 4s, and so on. There is an element of
chance. We might get four 6s or seven 6s. But if we got 12 or more
6s we might think, ‘Hang on a minute, someone’s been fiddling with
this dice!’ Similarly if we found that 47 per cent of White offenders
got a custodial sentence compared with 53 per cent of Black
offenders, we might be unsure whether this is something that could
occur by chance, or that we ought to be considering alternative
explanations. Statistics enables us to do this by using a test for
comparing the proportions within different categories. This is called
the chi-square test (written algebraically as χ2). What it does is to
compare the observed values for each cell of the cross-tabulation
with what the expected values would be if there were no differences
between the categories. This gives us a chi-square value that can
then be looked up in a table of probabilities to see whether it could

have occurred by chance. In practice, SPSS will perform this test for
us and tell us the results. We are not going to explain the chi-square
test in detail, and recommend referring to one of the statistics texts
given in the suggested reading at the end of this chapter. Apart from
knowing how a test is done, it is just as important to know when to
use it and how to interpret the results.
In the example given, an SPSS analysis tells us that for the
cross-tabulation above, chi-square equals 10.33, and that the likeli-
hood of getting a value that large will only occur three times out of
every hundred times these groups are studied. This is very unlikely,
and we can therefore conclude that such a difference between the
sentences the different groups received did not occur by chance. Five
times out of every 100 (or 5 per cent, or a probability of .05) is
traditionally regarded as the point at which we reject the hypothesis
that there is no difference between the groups. This is referred to as
the significance level. If we have a probability lower than .05, then
we can say that the difference is statistically significant. In everyday
language the term ‘significant’ is used loosely to suggest that
something is meaningful. However, in statistics referring to a
significant difference means something more precise. It means that a
certain level of statistical probability has been reached. The
1 per cent level, or probability of .01, is an even more stringent test
of significance, and means the likelihood of a difference occurring by
chance is one time in a hundred.
It is important to note that what we have done so far is to
conduct the test for the whole table, comparing three different
sentences for three ethnic groups. While the chi-square test tells us
that the differences in sentencing between the three groups is a real
one and not just a matter of chance, we still have to consider the
differences further. For example, the large chi-square value could
reflect not just the fact that a higher proportion of Black offenders
get custody than the other two groups, but that the Asian offenders
seem to be more likely to get a fine than either of the other two
groups. If we specifically want to know whether Black offenders are
more likely to get sent to prison than offenders from the other two
groups, then we need to do a more specific analysis, comparing
Black with other ethnic groups in terms of the likelihood of a
custodial rather than a non-custodial sentence. We can do this by
means of a bit of simple recoding, which produces the following
result, shown in Table 8.6.
Table 8.6 tells us that the proportion of Black offenders getting a
custodial sentence is much greater than any other group, and that
this is statistically highly significant. The material in the note is
statistical shorthand for saying that the value of chi-square is 7.0,
and that the probability of this occurring by chance is less than
eight times out of every thousand that these particular groups are

JOBNAME: Mcgraw−crowsemmens PAGE: 15 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 37A62BF9
Table 8.6 Custody by race
examined. We can therefore say that there is a real difference:

16.7 per cent compared with 50.0 per cent may seem to be such a
big discrepancy as to convince us that this cannot be explained by
chance, but even so we need to check.
Having done this, it is important not to jump to conclusions
about why this may be the case. Theoretically, various explanations
are possible. The Black offenders in the sample might have
committed more serious offences than the other two groups, or been
more likely to have already received non-custodial sentences, or the
sentencers might be racially prejudiced. Each of these possibilities
has to be examined to see whether it can be eliminated as an
explanation. This brings other variables into play in addition to race
and sentence. We will say more about this shortly, but before we do
we want to say a bit more about bivariate analysis.
Interval variables
In looking at race and sentencing we have been looking at two
nominal variables, where the data is in categories. But suppose we
are looking at continuous variables, where the data are in equal
intervals, or a mixture of nominal and interval variables, what then?
Let’s say we want to look at an interval variable, such as age, and a
nominal variable, such as gender. Are the women in a sample
significantly older or younger than the men in the sample? Taking
our data from the court abstraction files we find that the mean age
of the men is 21.3 and the mean age of the women is 21.8. To find
out whether this is a chance difference or a significant difference we
would do a Students t test. Using SPSS to do the work for us, we
find that t = – .376, and that the probability of this occurring by

JOBNAME: Mcgraw−crowsemmens PAGE: 16 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 552585C6
chance is .709 (or 71 times out of a hundred) which is much bigger

than 5 per cent, so we can conclude that there is no significant
difference between the mean ages of the men and the women in the
sample.
As a further illustration, let’s go back to race and sentencing,
where we found that Black offenders were more likely to get a
custodial sentence than either of the other two groups. Could this
be because the Black offenders in our (fictitious) sample have more
previous convictions? A t test here tells us that the White offenders
have 2.32 previous convictions on average, whereas the Black
offenders have 1.8 previous convictions on average. This alone tells
us that the greater likelihood of a custodial sentence for Black
offenders cannot be explained by their having more of a history of
offending, but the t test result that t = 1.0 and p = .322 (again much
bigger than .05) also tells us that there is no significant difference
between the number of previous convictions of Black and White
offenders. For the sake of completeness we should also compare the
mean number of previous convictions for Black and Asian offenders,
and perhaps for White and Asian Offenders. This could be done by
doing two more t tests. This is because a t test can only compare
two groups. If we want to compare the means of more than two
groups at once, then we use something called an analysis of variance.
As you can see, the number of different statistical tests that a
criminologist needs to know about is growing. But this need not be
a worry. It takes time to get to know about each, and what it does
mean is that we have a range of options to meet different situations.
Another statistical procedure, which many people have heard of at
some time or other, is correlation. The test used for interval level
variables is a Pearson’s correlation coefficient, and designated by the
letter r. Here we are again looking at two variables, this time, two
interval level variables. However, with correlation we are not looking
at differences, but at whether there is an association between two
variables. The value of the correlation coefficient can vary from +1
to –1. Plus 1 is a perfect positive correlation and means that for
every unit increase in one variable there is a corresponding unit
increase in the other variable. Conversely, –1 means that for every
unit increase in one variable there is a corresponding unit decrease in
the other variable: as one goes up, the other goes down. A
correlation coefficient of zero means that there is no relationship at
all. Looking at our artificial court abstraction dataset, and using
SPSS to perform the calculations for us, we find that there is a
correlation between the age of the offenders in our sample and the
number of previous convictions they have of r = 0.269. This is not
a terrifically close relationship, but SPSS also obligingly tells us that
for this particular sample it is statistically significant at the

JOBNAME: Mcgraw−crowsemmens PAGE: 17 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 3B18ECF2
5 per cent level, so not surprisingly we can conclude that in general

as people get older, they also have more convictions.
Using SPSS we can also look at the correlation between several
variables at once, producing a correlation matrix. For example, let’s
suppose that we are doing a study of crime in 20 local areas, and
collect information on conviction rates (per 1000 population),
unemployment rate, and household occupancy levels (often regarded
as an indicator of socio-economic well-being).5 We could do three
separate correlations and find that:
+ r for conviction rate/household occupancy was .656, p = .002;
+ r for conviction rate/unemployment was .612, p = .004;
+ r for occupancy/unemployment was .163, p = .493, n.s.
However, we can put all three variables into the analysis at the same
time and get the following correlation matrix, shown in Table 8.7.
Notice that what you get is a mirror image, with each of the three
correlations appearing twice, either side of the cells that form the
diagonal. The diagonal cells consist of 1s because obviously there is
a perfect correlation between conviction rate and conviction rate.
The double asterisks show that for the two significant correlations
the results could only have occurred by chance less than one time in
a hundred (p < .01). In fact, in one instance the probability is two in
a thousand and in the other four in a thousand. One thing to notice
is that although correlation matrices include three or more variables,
we are still essentially doing bivariate analysis, because we are only
comparing one variable at a time with one other variable.
Table 8.7 Correlation matrix of convictions by occupancy by unemployment

JOBNAME: Mcgraw−crowsemmens PAGE: 18 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 36D9FE31
What matters is whether you are looking at the relationship between

the variables. We will come onto genuinely multi-variable analysis
shortly.
One of the features of correlation is that it is a symmetrical
relationship. Variable X is correlated with Y and vice versa; we are
not taking any view about which is the dependent variable and
which the independent variable. However, in another form of
bivariate analysis, we use one variable to predict another variable.
This is linear regression, where the fact that there is a correlation
between two variables is used as a basis for estimating the value of
one (usually designated as the Y variable) from the other (usually
designated as the X variable):
Dependent variable (Y) ← Independent variable (X)
(e.g. conviction rate) (e.g. occupancy rate)
In the example used here, where we know that unemployment is

related to conviction rates, if we know that the unemployment rate
in an area is 1.5, a regression analysis enables us to calculate that the
likely level of convictions will be 27 per 1000 in the population.
This kind of information can be used, for example, to target
resources in certain areas – although whether to put more police
into an area or to do more to tackle unemployment is a political
decision.
Ordinal variables
We’ve now considered the analysis of two variables where one or
both are nominal or interval level variables. What about ordinal
variables? Take a series of questions to secondary school children
such as the following, where pupils are asked to tick one of the
boxes (see Box 8.1).

JOBNAME: Mcgraw−crowsemmens PAGE: 19 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 52ACA44F
The likelihood is that when coding the responses to each statement,

we will give each box a score such as 3 = Very wrong to 0 = Not
wrong at all. Suppose we want to see whether there is a relationship
between thinking that it is all right to play truant, and all right to
steal. Or suppose we want to see whether the attitudes of girls
towards these statements are different from those of boys. We can’t
do the same kind of correlation as for age, or number of previous
convictions. We can’t use a t test to compare means because these
are not equal intervals. We could treat each of the boxes as
categories and analyse them as though they were nominal variables,
comparing the percentage of girls ticking each box with the
percentage of boys. There are also certain kinds of statistical test
that can be used with such variables. They are called non-parametric
statistics, and make fewer assumptions about the type of data for
which they can be used. So instead of using a Pearson’s r correlation
coefficient we can use a Spearman’s rho correlation coefficient, and
instead of using a t test we can use a Mann-Witney U test.
When analysing two variables it is often important to think about
which is the dependent variable and which is the independent
variable (something that was explained in Chapter 3). In the
example above, if we were interested in comparing the attitudes of
boys and girls then gender would be the independent variable and
the scores for the statements would be the dependent variables – the
results that we are interested in. In the example used earlier
regarding race and sentencing, ethnic group is the independent
variable and sentence is the dependent variable. When analysing
research results, it is often the case that the characteristics of the
sample, such as gender, ethnic group, occupation, age, etc. will be
the independent variables, and things like what happened to them,
their responses to attitude statements, and so on will be the
dependent variables. This is not always so because there may be
times when we want to know, for example, whether the men in a
sample are older or younger on average than the women in the
sample. The identification of variables as dependent, independent, or
intervening variables becomes more important when we embark on
multi-variable analysis.
Analysing three or more variables

Multi-variate analysis involves three or more variables, and can take
different forms. We’ve already had a taste of it earlier when looking
at the relationship between race and sentencing. Having found that
Black offenders were statistically more likely to receive a custodial
sentence than the other two groups, we then looked at previous
convictions with reference to race to see whether this might explain
why this happened – it didn’t. In this situation we were using the

JOBNAME: Mcgraw−crowsemmens PAGE: 20 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 34B2D2A2
causal model referred to previously in Chapter 3. Here race was the

independent variable, sentence the dependent variable, and previous
convictions the potential intervening variable (Figure 8.4).
In this situation we are taking account of, or controlling for, a
third variable. In the race/sentencing/previous convictions example,
we did it by looking at the relationships two variables at a time.
However, there are techniques that enable us to look at three or
more variables at the same time. One of these is an extension of
correlation called partial correlation. This requires three variables, all
of which, as in the case of ordinary correlation, are interval level
variables.
When we looked at correlation we used the example of conviction
rates, occupancy rates and unemployment rates, correlating each
variable with each of the other two. But we can also find out to
what extent the relationship between two of the variables is affected
by the third. Suppose we are doing a study of a particular town and
we are interested in the relationship between the households and the
conviction levels in the town. We now know that there is a
significant relationship because the correlation between conviction
rate and occupancy rate was .656, significant at a probability level of
.002. But we also know that unemployment is related to conviction
rates. What happens to the relationship between occupancy and
conviction rates if we control for the effects of unemployment
(‘partial it out’ in statistical language)? If we get SPSS to do a
partial correlation for us, it gives us the following output as shown
in Table 8.8.
Figure 8.4 Three variable relationships

JOBNAME: Mcgraw−crowsemmens PAGE: 21 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 38C6691C
Table 8.8 Partial correlation
Table 8.8 shows us that in taking account of the effects of

unemployment on the relationship between convictions and occu-
pancy rates, the correlation between convictions and occupancy has
now increased to 0.713, and is significant at a probability of 0.001
(i.e. it could only occur by chance one time in a thousand, compared
with two times in a thousand previously).
Regression can also be extended to take account of three, and
indeed more than three, variables. Using simple regression, we could
examine the ability of one independent variable (e.g. occupancy rate)
to predict a dependent variable (e.g. conviction rate). By using
multiple regression we can achieve a stronger prediction by incor-
porating several independent variables which contribute towards
explaining the dependent variable:
Dependent variable ←independent variable 1 + independent
variable 2
(e.g. conviction rate) (e.g. occupancy rate) (e.g. unemploy-
ment rate)
Logistic regression can be used where the dependent variable is a

dichotomous nominal variable (i.e. it is Yes or No, Black or White,
Male or Female).
Where all the variables are nominal, we can do cross-tabulations
that show three variables at once (e.g. gender by ethnic group by
sentence), but tables with more than three nominal variables will get
very confusing to read and, unless there is a very large sample, the
numbers in each cell quickly become so small as to be not much
use. The statistical procedure for examining several nominal variables
is log linear analysis.

JOBNAME: Mcgraw−crowsemmens PAGE: 22 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 4C752CE4
Summary
In this section we have attempted to give an introduction to some of
the possibilities for analysing numerical criminological information.
In doing so, we have:
+ referred to the different types of numerical variable (nominal,
ordinal, interval, ratio);
+ introduced the notion of statistical probability;
+ considered the analysis of single variables, two variables in relation
to each other, and situations which take account of three
variables;
+ illustrated the distinction between dependent, independent and
intervening variables.
We have not, however, attempted to explain in detail how all these
procedures work, nor how to use a computer to carry them out. This
would take many chapters and it is far better to read texts that are
dedicated to doing this. In doing so it will be found that a lot of
qualifications need to be added to when and how to use various
tests. In order to use multiple regression, for example, a number of
assumptions about the nature of the data have to be met. Further-
more, it is one thing to learn how to do a chi-square test, a t test,
or a correlation, but it is quite another to have a set of data from a
questionnaire or abstraction schedule in front of you and know what
to do with it, and when to use a particular test. There are a large
number of statistical tests, which meet a wide variety of situations,
and knowledge of what they are and when to use them will only be
acquired over a period of time. We hope that this section gives
readers an overview of numerical analysis which will enable you to
explore statistical procedures with an understanding of where they fit
in to the process of analysis overall.
It may help to summarize the material above, setting out what we
have covered by showing the procedures appropriate for nominal and
interval level variables depending on whether univariate, bivariate, or
multivariate analysis is to be used (see Table 8.9).
Finally, it is important to say something about how quantitative
analysis is used. It is easy to become immersed in the minutiae of
statistics, or be fascinated by some of the patterns and relationships
that emerge. But it is important to use statistics in the context of a
theoretical background, the research question being addressed, and
the hypotheses being investigated. For example, earlier on we
explained how it was possible to produce a correlation matrix in
which a series of interval level variables could be correlated with
each other. Because this can be done at the press of a button, it is
tempting to throw every variable in at once and see what comes out.
While we would not say that this should never be done (it may

JOBNAME: Mcgraw−crowsemmens PAGE: 23 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 3742C22B
Table 8.9 Examples of analysis by type of variable and number of variables
throw up some interesting questions worth pursuing), this is some-

thing of a ‘shotgun’ approach to analysis. In general, it is much
better to be clear about what one is looking for and why. So, for
example, if you are looking at the relationship between unemploy-
ment and crime rates, there is a clear rationale for the analysis we
did earlier, and for then going on to consider what other factors
(such as occupancy) might be involved. Similarly, when using the
court abstraction data, the question that needs to be asked is why
this data was collected. We have created the dataset to illustrate
certain ways of analysing data, but if it were a real dataset, then the
likelihood is that these particular pieces of information would have
been abstracted from court records in order to investigate what
factors influence sentencing. Specific hypotheses would then be
developed such as ‘Some ethnic groups are sentenced differently to
others’. In doing this, dependent and independent variables would be
identified. Essentially, therefore, the kind of analysis one does needs
to be anticipated when planning criminological research, and the
results of analysis need to relate to, and inform the research process
outlined in Chapter 1.
Analysing qualitative material: An introduction
We move on now to consider how we analyse qualitative data. We

have seen in earlier chapters (especially Chapters 5, 6 and 7) that
qualitative data can be collected in a range of different ways.

JOBNAME: Mcgraw−crowsemmens PAGE: 24 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 506D6CE1
Interviews, observations, documentary analysis and even surveys

(where open-ended questions are used) can all produce qualitative
data. Whenever you include these methods in a research study, you
soon find yourself in possession of a large, unwieldy database of
transcripts, fieldnotes and/or documents and the aim of this discus-
sion is to explain how these data can be managed and analysed.
However, unlike quantitative data, there are no clear-cut and widely
accepted ‘rules’ or procedures for qualitative analysis. There are some
broad guidelines and we will explain these in this part of the
chapter.
Although there is a range of computer packages available for
qualitative analysis, we do not refer to them specifically in this
discussion. The problem is that there are too many different
packages, each capable of different functions, and we cannot refer to
each in the same way that we could earlier to SPSS. An excellent
overview of the packages that are available (including CISAID,
SIMSTAT, NUD*IST and ATLAS) is provided by Lewins (2001)
and we refer you there for more detail. You will find that the
purpose of these packages is primarily to facilitate the very processes
we describe in this chapter.
The skills of qualitative analysis are notoriously hard to master
and it is unlikely that a simple reading of this chapter will equip you
adequately to conduct analysis of your own. Robson (2002) makes
an important observation when he says that traditionally these skills
have been taught using an ‘apprentice model’ with new researchers
being guided by more experienced supervisors. We aim simply to
provide a detailed introduction to the key processes in qualitative
analysis and to explain some of the most commonly used qualitative
approaches in criminology. We make some recommendations for
further reading at the end of this chapter.
The process of qualitative analysis

We have already seen that quantitative analysis takes place after all
the data has been collected and is carried out according to
pre-determined principles and statistical procedures. In contrast,
qualitative analysis begins soon after the data collection starts and
continues throughout the period of study and, as we mentioned in
the Introduction, the procedures are far more flexible and adaptable.
That is not to say, however, that qualitative analysis does not need
careful planning and organizing. In this section, we will explain the
basic principles of qualitative analysis and then will go on to
describe how these principles have been developed for use with the
most common qualitative methodologies in criminology (grounded
theory, ethnography, case studies, content analysis and conversation
and discourse analysis).

JOBNAME: Mcgraw−crowsemmens PAGE: 25 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 56594DA2
Qualitative analysis always involves two key tasks: coding and

fragmenting the data. The task of coding starts almost as soon as
the first batch of data has been collected. The first stage is often
referred to as ‘Basic Coding’, and because little analysis takes place,
it is simply a labelling process. You begin by reading the transcripts
or fieldnotes and reflecting on them. It is a good idea to resist the
temptation to write notes on the first reading because it allows you
to concentrate on understanding the data as a whole before
embarking on the process of splitting it up. On the second reading,
you can work through the text and start making a list of the
concepts which emerge (the coding process should now be easier
because you have a good idea of what is contained in the whole
document and this will facilitate the creation of a comprehensive
index of concepts). Interestingly, there is no clear definition of what
a ‘concept’ is in this context and often students and new researchers
find this frustrating. Bryman (2004) suggests that a good way to
start is to list questions such as: What is this about? Who is
involved? What question does it raise? What is happening? What
are people doing? What do people say they are doing? In the early
stages of analysis, the concepts will tend to be descriptive and poorly
developed and it should not be a concern if your index becomes
rather large and unwieldy. The best approach is to pursue all the
different lines of thought and refine the index at a later stage.
When you have developed a well-formed index of concepts, you
can begin fragmenting the data. You need to create a file for each
concept and then extract all the sections of text (a quote perhaps or
a paragraph or longer) which represent that concept for storing in
the file. In the days before computers, this was done quite literally
by copying, cutting up and gluing together chunks of documents.
Today this is made much easier because computers allow you to
copy and paste text without scissors or glue! For this, you can use
either a simple word processing package or one of the qualitative
analysis packages. It is very important that an unviolated version of
the original document is kept safe and that excerpts are clearly
identified according to their original source. Indeed, it is important
to keep in mind that when the data are broken down into small bits,
there is a real danger of losing the context once a certain quote or
observation has been extracted. It is important, therefore, not to lose
sight of the original context.
The next stage of the coding process, often called ‘content
coding’, requires you to identify themes in the data. In identifying
themes, you should be able to create ‘categories’ out of your
concepts. Again, there is no accepted definition of a ‘category’ and
this is less than helpful, but perhaps a good way of getting to grips
with the idea (especially if you do not have a supervisor or colleague
who can guide you) is to read some published work and see how

JOBNAME: Mcgraw−crowsemmens PAGE: 26 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 17BAB125
others have done it. There are worked-through examples in several

of the methods texts. Although they are too detailed for reproduc-
tion here, we do refer to them in the Further Reading section at the
end of the chapter. A useful criminological example is Tomsen’s
(1997) ethnographic study of violence in drinking establishments.
Although Tomsen does not explain how he developed his concepts
and categories, it is a useful exercise to attempt to identify them (as
we have done in Box 8.2).
Box 8.2 An analysis of Tomsen (1997)

JOBNAME: Mcgraw−crowsemmens PAGE: 27 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 58C3A7D3
There comes a point when it is possible to look for initial

connections between the concepts and start to draw up hypotheses
for testing. In this way, the themes/categories are related to the
broader analytical themes. We noted earlier that the process of
analysis usually begins early in the data collection and continually
feeds back into the collection process itself and this is done through
the method of analytic induction. You develop a hypothetical
explanation (a set of hypotheses) for the social situation or phenom-
enon you are to explore and have a look to see whether the data
confirm your hypotheses. If not, you must redefine your hypotheses
and start the process again. Once you reach the point where you
have confirmed your hypotheses and no deviant cases or examples
emerge, the data collection ceases and the final conclusions can be
drawn. It is particularly important to ensure that there are no cases
which contradict your hypothesis, an approach called negative case
analysis. A useful example of this process in action is Cressey’s
article in which he describes the analysis he conducted on embez-
zlers (1950). This is a particularly good example because he
specifically describes how his hypothesis (originally derived from the
literature) was reformulated a number of times throughout the study.
He notes how his initial hypothesis was ‘abandoned almost imme-
diately’ (ibid.: 741) and shows exactly how each subsequent hypoth-
esis had to be revised. He concludes that his eventual successful
hypothesis is as sound as it can be, but is open to the idea that it
might be disproved in the future. He has also included a range of
cases in the existing literature as part of his negative case analysis,
which is a common approach.
Cressey’s article is useful because he does explain how he has
carried out the process of analytic induction to the reader. However,
often researchers do not take this approach and the reader is left to
guess how the hypotheses have been developed and the final
conclusions drawn. In order to demonstrate how this may be done,
we refer you again to Box 8.2 in which we have listed the
hypotheses and subsequent theory which we think we can see in
Tomsen’s study. The key point to realize is that we could, of course,
be wrong in our analysis and that is why it is so important for
qualitative researchers to explain how they reached their conclusions
(a point we take up again in the final section).
We have now introduced you to the process of qualitative analysis
in simple terms. Remember that these are merely guidelines and that
the actual process of qualitative analysis is complex, time-consuming
and requires considerable skill. We will move on now to see how
these basic principles have been used in different qualitative tradi-
tions which use a range of methods. You will notice many common
themes and approaches but there are also differences, most of which
are shaped by the epistemological basis of the approach and the type

JOBNAME: Mcgraw−crowsemmens PAGE: 28 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 41BF0993
of data collected. It is often acceptable to utilize elements of

different qualitative approaches, but remember that it is always
important to ensure that you are maintaining validity and reliability
in your research.
Analysing open questions in surveys

It is common to see a combination of both closed and open
questions in questionnaires and interviews. Usually, a positivist
approach will have been adopted and it will be desirable to combine
the quantitative and qualitative data together for the purposes of
hypothesis testing. So, in order to analyse the quantitative and
qualitative data together, it is necessary to convert the open
responses into numbers. The process for doing so is best illustrated
through an example and we shall use this excerpt from an interview
with local residents about their local area:6
Q: What do you think are the most serious problems faced by
people living in this area at the moment?
A: We are having a lot of trouble with kids. They are always
on the corner of Bates Street smoking and drinking at night.
They make a lot of noise all night and leave chip-shop
wrappers and smashed bottles all over the paths. We also have
a problem with traffic. People who work at the hospital park
on both sides of our street and as everyone uses it as the short
cut through to the supermarket, there’s a lot of traffic. It makes
it hard for cars to pass and there have been crashes in the past.
The first thing you will need to do is make a long list of all the
problems mentioned by all the respondents and give each problem a
unique numerical code. So, our list here would look like this:
1 Youths hanging around at night
2 Youths making noise at night
3 Youths leaving litter
4 Parked cars obstructing road
5 Heavy traffic.
The list would grow as more interviews are analysed and additional
problems mentioned by respondents. You then need to decide how
to apply the codes to each case and record them in the data file.
There are three possible ways to do this:
1 Create a single variable (PROBLEM) and record only the first
problem mentioned by the respondent (so, for our respondent in
the example, PROBLEM = 1).

JOBNAME: Mcgraw−crowsemmens PAGE: 29 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 495D28F0
2 Create several variables (PROBLEMA, PROBLEMB, PROB-

LEMC, etc.) and record the first three problems mentioned by the
respondent (so, for our respondent in the example, PROBLEMA = 1 ,
PROBLEMB = 2 , PROBLEMC = 3).
3 Create a variable for each of the problems on the list
(PROBLEM1–PROBLEM5). If a respondent mentions the
problem, insert the code ‘1’. If the respondent does not mention
the problem, insert the code ‘0’. (So, for our respondent in the
example, PROBLEM1–PROBLEM5 would each be coded as ‘1’
but PROBLEM6–PROBLEM10 would be coded as ‘0’ because
they were not mentioned.
Each of these three approaches is valid but the important thing to
realize is that a clear decision needs to be made before the analysis
begins. Open questions can be very useful in interviews and
questionnaires, but they do need to be handled with care. You need
to resist the temptation to insert open questions and just wait to see
what happens – the risk here is that you end up with a batch of
wasted data. It is important, then, to test out the coding process at
the stage of piloting to ensure that the coding strategy is suitable.
Fielding (2001) also suggests that a Code Book is kept, in which
you record exactly how each of your variables are coded and
constructed.
Grounded theory studies

The most widely used framework in qualitative analysis is the
grounded theory approach. This is where you enter a research
situation or environment with no preconceived ideas about it and no
hypotheses to test. Instead, you allow the key features and relation-
ships to emerge from the situation and record them accordingly.
Thus, the process of data collection is controlled by the emerging
theory.
The grounded theory approach utilizes three types of coding
(Strauss and Corbin 1998):
1 Open coding. Data are separated, concepts and categories are
identified and given labels.
2 Axial coding. The concepts and categories are interconnected to
make theories.
3 Selective coding. One core category is selected for in-depth
consideration.
Grounded theory studies have some additional special features and
there is an extensive methodological literature which we cannot go
into here (although there are recommendations for further reading
later). Instead, we will simply highlight three of the most important

JOBNAME: Mcgraw−crowsemmens PAGE: 30 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 4D7B3937
elements. First, you utilize a process of theoretical sampling in

which the whole process of data collection is controlled by the
emerging theory (what to look for, who to speak to, etc.). Second,
an approach called ‘theoretical saturation’ is used which requires that
coding and analysis continue until no new categories emerge and the
data no longer illuminates aspects of the concept under develop-
ment. Finally, the ‘constant comparison method’ ensures that the
relationship between concepts and categories is continually examined
by the researcher. Glaser and Strauss (1967) suggest that the
researcher uses a tool called a ‘memo’ to facilitate the constant
comparison approach. This is a kind of journal or diary in which the
researcher records the analytic decisions s/he makes and explains
how concepts and categories have been developed.
There are, however, a number of criticisms of the grounded theory
approach which you need to be aware of before beginning this kind
of research. Aside from the fact that it is time-consuming and
resource-intensive, perhaps the biggest question is whether it is
possible to maintain the absolute level of objectively that the
approach requires. Robson (2002) notes that this is a particular
problem in reality because in order to make proposals for projects
(especially where you are bidding for funding) you need to have a
clear research question from the outset and to consider the possible
implications and uses of your research. There is also a question as to
whether this kind of research can actually make a useful contribution
to theory since the focus is on generating concepts rather than
seeking explanations.
Analysing case studies

This approach, developed by Miles and Huberman (1994), has a
strong realist thread and the aim is to look for the mechanisms and
processes which can provide causal explanations for events and
situations. It is best understood as a qualitative approach with some
in-built positivist elements for the purposes of making the research
more ‘scientific’. This approach has three stages but all form a
continuous iterative process. At the data reduction stage, descriptive
data are summarized in notes and memoranda. At the data display
stage, matrices, charts and networks are used to create structured
analysis. There are different types of tools to use and it is worth
noting that some of these tools are useful in non-case study
approaches too:
+ context charts (showing the relationships between organizations,
groups and roles);
+ event flow networks (showing incidents and events in time
sequence);

JOBNAME: Mcgraw−crowsemmens PAGE: 31 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 4E112F35
+ flow charts (showing how decisions are made);

+ tree diagrams (showing how phenomena are broken down into
categories and concepts);
+ cognitive maps (showing a person’s belief system or thought
process);
+ casual networks (showing how different concepts relate to each
other).
The final stage is conclusion-drawing and verification. It is impor-
tant to be looking for reliability and validity throughout the process.
There is a range of different approaches available and we discuss
these later in the section on interpretation, conclusions and presen-
tation.
Analysis in ethnographies
Ethnographic work, you may recall, usually involves a combination
of methods and as a result, different types of analysis may be
appropriate. Sometimes, a grounded theory approach is taken and
sometimes a case-study approach is more suitable. Either way, it is
likely that you will need to tailor the analysis to the individual study
and this may involve taking different aspects from different
approaches to analysing qualitative data.
Wolcott (1994) suggests that there are three stages of analysis in
ethnographic work. The first is the ‘thinking stage’. This requires
you to get to know your data extremely well and to reflect on it
fully. This stage should not be rushed and really does require an
unhurried, skilled approach. The second stage is the ‘categorizing
stage’ in which categories are created and typologies developed (this
is a common approach in ethnographic work: Hammersley and
Atkinson 1995). Categories start off poorly defined and descriptive
and take time to be developed fully. The third stage is the
‘progressing focusing stage’ in which the research question is defined
and clarified.
Content analysis
If you recall from our discussions in Chapter 5, the name given to
the method of analysing documents is content analysis. Content
analysis is often used as a supplementary method in a multi-method
strategy. It can involve the collection and analysis of both quantita-
tive data (for example, the number of relevant words or paragraphs)
and qualitative data (the nature of language or style of pictures). In
the criminological context, most studies using a content analysis
approach have focused on printed news media and official policy
documents.

Content analysis begins with the identification of a focused

research question and, where suitable, developed hypotheses. The
next stage is to develop a sampling strategy. The sampling frame will
depend entirely on the nature of the research question. For example,
if the aim is to test the relationship between crime reporting in the
news and perceptions of crime rates, the sampling frame would be
constructed of relevant national and local newspapers and/or news
programmes on television/radio. The sample can be drawn according
to a number of different elements, such as time, publication, author
and reference to particular events. Again, this will be determined by
the nature of the research question.
Arguably, the most crucial stages in content analysis are the
development of categories for analysis and the definition of record-
ing units. These categories are usually defined before the analysis
begins when the hypotheses are formed. It is not always easy to
identify categories effectively, but Holsti (1969) gives an excellent
and in-depth explanation of the processes involved in creating
‘category systems’. The categories for analysis may be derived directly
from the key concepts which, when put together, build the hypoth-
eses and may be quantitative or qualitative. Moreover, categories can
be ‘manifest’ (physically present) or ‘latent’ (inferred/interpreted by
the analyst) (Robson 2002). As in the other approaches to qualita-
tive analysis, these categories must be exhaustive and mutually
exclusive. So, if a study were to focus on the relationship between
the presentation of crime events in local newspapers and public
perceptions of local crime rates, the categories for analysis might
include those laid out in Table 8.10.
Once the categories have been designed, they have to be opera-
tionalized in accordance with the selected recording units. The most
common recording unit is the number of occurrences of words or
phrases in a document, but it is possible to use alternatives such as
paragraphs, column inches, number of stories, size of headline,
colour of text and position on page. The coding scheme should be
tested on samples of text and it is advisable to use more than one
person in the process so as to assess inter-observer agreement as a
measure of validity/reliability. This is especially important where a
category system is heavily based on latent content because these
categories are more prone to subjectivity or bias.
Finally, once the categories for analysis and recording units are
validated, the analysis can begin. Actually, the method used is not
dissimilar to using a structured observation schedule where occur-
rences of the pre-determined categories are counted and recorded
using a schedule. Bryman (2004) gives an excellent working example
of a coding schedule (including copies of the newspaper articles
used). Importantly, the data drawn from the content analysis can
then be related to ‘outside variables’. For example, the level of crime

Table 8.10 An example of categories for content analysis
reporting in a newspaper could be analysed to test a relationship

with the readers’ perceptions of crime rates which are measured
using a survey (for an example, see Williams and Dickenson 1993).
Like all methods of research, content analysis has its strengths
and weaknesses. On the plus side, the documents under scrutiny
have usually been produced for another purpose (i.e. not for the
purpose of the research) and content analysis has a clear strength in
being unobtrusive and non-reactive. However, this can also cause
problems for the researcher as the document can be structured in an
unhelpful way, leaving the researcher the tough task of organizing
the material. One of the major challenges is to ensure validity in the
research, especially where the analyst must distinguish between
‘witting evidence’ (information which the author of the document
intended to convey) and ‘unwitting evidence’ (any additional infor-
mation which the analyst can glean through his/her reading of the
document). Finally, content analysis is renowned for being a rather
laborious and time-consuming process. It is true that computers
have eased the labour to some extent. However, one has to be
careful because computers are able to analyse text in a restricted
mechanical way. For example, it is easy for a computer program to
count frequencies of words but it cannot be relied upon to draw

JOBNAME: Mcgraw−crowsemmens PAGE: 34 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 4FBC23DA
inferences, interpret or allow for errors or contexts. As long as one is

realistic about the limitations of the computer, it can be a very useful
tool in content analysis.
Analysing language – conversation analysis and dis-

course analysis
It is important to be able to understand language in criminological
research, especially where interviews or observations are being
conducted, because people use words in particular ways and we need
to understand what they actually meant when they used them. This
is particularly important where a special vocabulary or ‘argot’ is used
by members of a group or culture, such as drug users. There are two
main types of language analysis – conversation analysis and discourse
analysis. They are similar in approach but different in terms of the
type of communication they can be applied to. Whereas conversa-
tion analysis is used to analyse verbal communications between
people, discourse analysis can be used to analyse both verbal and
written forms of communication.
The method of conversation analysis emerged out of ethnometh-
odology, which is concerned with understanding the methods used
by people in their everyday lives to interact, function and accomplish
social order (Garfinkel 1967). The analyst seeks to understand the
structures which underlie interactions between people and which
contribute to the attainment of social order. So it is not just about
what people say, but about how people talk to each other in natural
situations. Unlike other kinds of qualitative analysis, conversation
analysis does have a relatively clear set of procedures to follow.
Indeed, conservation analysis is an interesting mix of principles
usually associated with both quantitative and qualitative analysis.
There are three assumptions made by conversation analysts
(Heritage 1987):
1 Talk is structured and there are unwritten rules of conversation
which we follow; we take it in turns to speak and use words or
prompts to progress the conversation.
2 Talk occurs in a context and in order to understand the
conversation we have to analyse it in that context.
3 Conversation has to be understood entirely from the data, not
through pre-theorizing and constructing hypotheses.
Put very simply, the process of conversation analysis requires going
over the transcript of a conversation with a fine-toothed comb,
interpreting what has been said and using special symbols to
indicate where a person pauses to reflect, hesitate or emphasizes
specific words. There are various well-established tools of conversa-
tion which are used to conduct the analysis (Wooffitt 2005):

JOBNAME: Mcgraw−crowsemmens PAGE: 35 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 55669AE8
+ Turn-taking: The idea here is that conversations depend on people

knowing when to talk and when to listen. Signals are used to
indicate when one person has finished talking and the next
person should respond. These signals include greetings, questions
and invitations and we all have an innate understanding of how
and when they should be used. The analyst seeks to identify if
and when they are used in conversation and to what effect.
+ Adjacency pairs: The questions, invitations and greetings used in a
conversation require appropriate responses. A question, for exam-
ple, should be met with an answer, an invitation with an
acceptance or rejection, a greeting with a greeting. Here the
analyst will look for the expected response and analyse the effect
it has on the progress and content of the conversation.
+ Preference organization: People who raise questions, invitations and
greetings will have a preferred and non-preferred response in
mind. So, for example, if I invite a friend to join me for lunch,
my preferred response is likely to be ‘yes’. If my friend cannot
come for lunch, s/he will have to give me my non-preferred
response. The analyst will examine how both parties manage the
delivery of the preferred or non-preferred response.
+ Accounts: Accounts are reasons given for not being about to give
the preferred response. So, my friend may look for an appropriate
way of turning down my lunch invitation without offending me.
The analyst will assess whether the account has be used positively,
to enforce the value of the original invitation or question, or not.
+ Repair mechanisms: Of course, conversations do not always follow
a strict turn-taking structure. Sometimes, people interrupt each
other or start talking at the same time. In these situations, one
person will stop speaking or perhaps it will be necessary for the
person to repeat a question or invitation. The analyst will examine
how the conversation is repaired.
This short summary should give you a flavour of conversation
analysis but it is, of course, much more complex than we suggest
here. It is a skilled approach to analysis which would probably
require some specialist training but it can be useful, especially in
ethnographic work. Where a video camera has been used to record
a conversation, it may also be interesting to analyse body language.
Let us move on now to consider discourse analysis. As we said
earlier, discourse analysis can be applied to written forms of
communication as well as spoken forms. It is used in a wide variety
of disciplines and there are different variations of how it should be
conducted. The focus is slightly different than that for conversation
analysis. Here, the assumption is that when people use discourse,
they are usually trying to achieve something. Discourse analysis
focuses on understanding the tools used as part of this process.

The aim in discourse analysis is to identify the ‘interpretative

repertoire’ (Wooffitt 2005). This is a concept which characterizes the
way in which people engage in discourse, so it might be an attitude
or belief system. It is common to find more than one repertoire in
a study because people may represent their ideas differently in
different contexts and to different audiences. Discourse analysis is
also used to look at the ways in which facts are presented by
different people. Here, the idea is that by analysing the discourse, we
come to understand how the originator has attempted to construct
an argument and convey facts. These are referred to as ‘rhetoric
strategies’ and can be identified by looking for variations in
expression which can alter the impact a series of statements has,
evidence that counter-arguments have been anticipated and
accounted for and attempts to present accountable arguments.
Unfortunately, it is not possible to go into further detail about the
methods of conversation and discourse analysis. It is clear to see
how both approaches might be useful in criminological research but
neither method is commonly used. It is worth exploring these
approaches, however, because even where a true conversation or
discourse analysis approach is not adopted, it can be useful to
understand more about how conversations work and how they can
reveal a great deal about relationships and group contexts.
Interpreting results, drawing conclusions and present-

ing findings
Finally, we will briefly consider how qualitative data are interpreted
and presented. Remember, as researchers we are seeking to produce
research which is both valid and reliable, and therefore need to think
carefully about how our data are analysed and interpreted to avoid
unjustifiable conclusions. There are some very significant potential
pitfalls for the qualitative researcher to fall into, all of which have
been discussed throughout the preceding chapters, and it is impor-
tant to consider using a range of methodological and analytical tools
to help build a defensible methodology. We summarize these in
Table 8.11.
Students of criminology are often frustrated because it is unusual
for researchers to explain the details of the analysis process and the
methods of interpretation in articles. How, then, can we be sure that
the research can be trusted? It can, indeed, be hard to judge whether
data has been interpreted correctly by a researcher and this is a
significant problem in qualitative research. We would suggest,
therefore, that whenever you conduct qualitative work and publish it,
you should keep this in mind. You should explain how the process
of analytic induction has been carried out, how concepts and
categories have been developed and how conclusions have been

JOBNAME: Mcgraw−crowsemmens PAGE: 37 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 37EAF9FE
Table 8.11 Pitfalls and solutions
drawn. Quotes are often used to illustrate how the data have been
interpreted, but you should always ensure that you use them
correctly and in context. Only say they are representative of the
kinds of responses you had, if they were (this may seem an obvious
thing to say but it is very important). Finally, as is almost always the
case with qualitative research, there is a problem with generalizabil-
ity and it is important for researchers to be able to distinguish
between the conditions that were sufficient for a phenomenon to
occur in this research, but not necessary in all situations and to
recognize that more factors may be relevant in other contexts.
Further reading
Books on quantitative analysis vary from the more traditional

statistics textbooks, to ones that aim to reach out and communicate
with those who find numbers off-putting. Among the more friendly
books on statistics for the non-statistician are: David Rowntree’s
Statistics Without Tears: a primer for non-mathematicians (1991);
Frances Clegg’s Simple Statistics (1990); and Kranzler and Moursund,
Statistics for the Terrified (1995).
A fuller knowledge can be gained from: Fielding and Gilbert,
Understanding Social Statistics (2000); Daniel Wright, Understanding
Statistics: An Introduction for the Social Sciences (1997); and Fink, How
to Analyse Survey Data (1995).

JOBNAME: Mcgraw−crowsemmens PAGE: 38 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 60DDDB91
David de Vaus’s book, Analysing Social Science Data (2002), also

gives a fuller understanding of data analysis, but instead of just
writing about various methods takes the approach of telling you how
to do different things.
There are also several books about how to use SPSS, which also
tell you about the various statistical techniques as well. SPSS
changes fairly regularly and the main thing to watch out for is
which version of SPSS a particular book covers.
Julie Pallant’s SPSS Survival Manual (2001) is one of the most
user friendly for beginners, which helps you to not only use SPSS
but interpret the sometimes confusing output that it produces.
Others include: Foster, Data Analysis Using SPSS for Windows: A
Beginner’s Guide (1998); and Norušis, SPSS for Windows: Base System
User’s Guide (1992), produced by the SPSS Corporation itself.
Probably the best all round guide is Andy Field’s Discovering
Statistics Using SPSS (2004). It covers a wide range of statistical
applications, explains how to apply them using SPSS, and is also
very accessible and well written.
There is a vast literature on qualitative methods, consisting of
contributions by researchers in a range of disciplines (including
sociology, psychology, history, information studies and communica-
tion science). Most textbooks contain quite detailed overviews of the
available methods and our students find the texts by Robson (2002)
and Bryman (2004) give useful introductions. Fielding (2001) is a
good introduction to coding and provides some very useful examples
to illustrate the techniques and Miles and Humberman (1994)
present their analytic tools very well. For conversation and discourse
analysis, Wooffit (2005) is extremely useful and accessible.
Appendix 1: Court sentencing project
Abstration Pro Forma

Code
Case Number : ❑❑❑1
Defendant’s Age : ❑❑2
Gender : male = 1 ❑3
female = 2

JOBNAME: Mcgraw−crowsemmens PAGE: 39 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 2CC2A10F
Race : White = 1 ❑4
Black = 2
Asian = 3
Offence : theft = 1 ❑5
damage = 2
violence = 3
Number of Previous Convictions : ❑6
Sentence : fine = 1 ❑7
community = 2
custody = 3
This is fictional data abstracted from court records at Escafeld

Magistrates Court. The variables are simplified. There are 56 cases
and 8 variables. These show a case reference number, age, gender,
ethnic group, offence, number of previous convictions, sentence, and
a score representing the overall seriousness of the case.

JOBNAME: Mcgraw−crowsemmens PAGE: 40 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 9EBDA4C0
case age gender ethnic offence precons sentence sscore

1 21 1 1 1 1 1 60
2 17 1 2 1 1 2 80
3 24 2 1 1 4 1 70
4 30 1 1 3 5 3 120
5 19 1 3 1 0 1 50
6 25 1 1 1 3 2 100
7 21 1 2 3 1 3 130
8 34 1 1 3 5 3 140
9 17 2 2 1 0 1 50
10 18 1 2 2 1 2 100
11 25 1 1 1 2 1 60
12 27 2 1 2 3 1 80
13 17 1 1 2 0 1 60
14 23 1 1 3 1 2 110
15 18 2 2 1 1 2 90
16 20 1 1 1 4 2 100
17 16 1 2 2 0 1 60
18 16 1 3 1 0 1 50
19 19 1 1 2 2 1 70
20 21 1 2 1 1 2 90
21 22 2 3 1 0 1 50
22 19 1 1 2 4 2 110
23 19 1 2 2 2 3 120
24 32 2 1 1 5 2 100
25 17 1 2 3 0 3 120
26 28 2 2 3 2 3 130
27 30 1 3 1 0 2 80
28 25 1 1 1 0 1 50
29 17 1 1 2 2 2 100
30 21 2 2 1 5 3 120
31 18 1 2 1 2 2 90
32 16 1 1 1 0 1 50
33 21 2 1 1 6 2 100
34 22 1 1 3 5 3 140
35 16 1 3 2 1 2 100
36 18 1 1 3 0 2 100
37 33 1 1 2 1 1 70
38 16 2 2 1 1 2 90

JOBNAME: Mcgraw−crowsemmens PAGE: 41 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 5E5E7B49
case age gender ethnic offence precons sentence sscore

39 17 1 2 2 3 3 130
40 24 2 1 1 1 2 90
41 19 1 1 2 3 3 130
42 26 1 3 1 0 1 50
43 19 2 2 1 2 3 110
44 20 1 1 1 1 1 60
45 21 1 1 3 3 3 140
46 16 1 2 1 1 2 90
47 22 2 2 1 6 3 120
48 17 1 1 2 0 2 90
49 19 2 3 2 0 1 60
50 28 1 2 2 1 1 70
51 30 1 1 1 0 1 50
52 17 1 2 2 4 3 130
53 19 2 3 1 0 2 80
54 23 1 2 3 2 3 130
55 20 2 1 1 3 2 100
56 25 1 1 3 1 3 130
Appendix 2: Calculating measures of a distribution
In this appendix we give an example of how the various features of

an interval level variable would be calculated. The variable is the
number of previous convictions which a sample of offenders have. In
practice these calculations can be done by SPSS, but it is also useful
to see how the particular measures are derived.
Suppose that in a sample of 294 offenders drawn from a
magistrates’ court,
103 had no previous convictions

95 had one previous conviction
50 had two previous convictions
33 had three previous convictions
5 had four previous convictions
8 had five previous convictions

JOBNAME: Mcgraw−crowsemmens PAGE: 42 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: 3E322BF4
We will calculate:
(a) the mean number of previous convictions

(b) the sum of squared deviations from this mean
(c) the variance
(d) the standard deviation
(e) the standard error of the mean, and
(f) the 95% confidence limits for the true mean.
You need to know that:
X is the value of a particular variable

X means the mean of X
f is the frequency with which a value occurs
n is the number of cases
Σ means the sum of all the values.
The procedure
The first thing to note is that although you have a ratio level of
measurement, the number of previous convictions, you also have
frequencies for each step of that variable. You could write down a
long column of numbers starting with 103 zeros, followed by 95 1’s,
followed by 50 2’s, etc. (this is how it would appear for 294 cases on
an SPSS data sheet). To calculate the answers by hand, however, it
is much simpler just to multiply all the zeros by 103 (answer = 0, of
course), the 1’s by 95 ( = 95) and the 2’s by 50 ( = 100), etc.
Add all these up to get the total sum (354), and divide by 294 to
get the mean ( = 1.2). You now have the basis for doing the ‘sum of
squares’ calculation by taking each case from the mean, squaring the
result and adding all these up (multiplying (X – X)2 by the
frequency for each step of the variable) to get
∑ (X − X̄)
2
= 445.76.

JOBNAME: Mcgraw−crowsemmens PAGE: 43 SESS: 38 OUTPUT: Wed Aug 29 08:37:28 2007 SUM: DD04D8F2
Laid out as a calculation this looks as follows:
_ _ _
X f fX (X - X) (X - X)2 f(X - X)2
0 103 0 - 1.2 1.44 148.32
1 95 95 - 0.2 0.04 3.80
2 50 100 0.8 0.64 32.00
3 33 99 1.8 3.24 106.92
4 5 20 2.8 7.84 39.20
5 8 40 3.8 14.44 115.52
n = 294 Σ = 354 445.76
X = 1.2
The variance is the sum of squares divided by the number of cases
minus 1, which is as follows:
445.76
s2 = 293 = 1.52
The standard deviation, the most commonly used measure of the
spread of a variable, is the square root of the variance, thus:
s = √1.52 = 1.23
The standard error of the mean, a way of estimating the extent to
which the sample mean may vary from the population or true mean
is obtained by dividing the standard deviation by the square root of
the number of cases, thus:
SEX = √294 = 0.07
By multiplying this by two and adding and substracting the result
from the sample mean we can be 95% confident that the true mean
lies between the resulting two boundaries. These are known as the
confidence limits, and the difference between them is the confidence
interval. Thus,
1.2 + 2 * .07 = 1.34 and 1.2 – 2 * .07 = 1.06
So we can be 95% confident that the true mean of the population

lies somewhere between 1.06 and 1.34. (This is because in a normal
distribution 95 out of 100 sample means lie within approximately 2
[1.96 to be precise] standard errors of the true population mean.)


Part III
Real-world research

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 24 OUTPUT: Wed Aug 29 08:47:52 2007 SUM: 0434816B

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 24 OUTPUT: Wed Aug 29 08:47:52 2007 SUM: 3F60C84A
9 Researching offenders
and employment
When unemployment rises there is often debate about the relation-

ship between unemployment and various other phenomena such as
ill health, and crime. During the 1980s, the United Kingdom
experienced the highest levels of unemployment since the Second
World War. This led to much writing about the possible impact on
offending, drawing both on previous economic and social research
(Brenner 1976) and new studies (Gormally et al. 1981) and analyses
(Farrington et al. 1986). While the main focus of interest was on the
relationship between unemployment and crime (Box 1987), there
was less awareness of other possible implications. For example,
higher unemployment means that it is more difficult for those with
a criminal record to get jobs. Since known offenders are more likely
to re-offend if unemployed (see Crow et al. 1989: 79), then this
means higher rates of re-offending, and of course this in itself
becomes an element in the link between higher unemployment and
crime. But another important element is what impact unemployment
may have for the way offenders are dealt with. In the 1980s the
NACRO Research Unit carried out various studies into the impli-
cations of unemployment for offenders. Some of these were con-
cerned with employment and training schemes for young and adult
offenders, but the project considered here looked at the relationship
between employment status and sentencing.
What the chapter shows is:
+ the conceptual basis for the study;
+ how this was then turned into a series of testable hypotheses;
+ the kind of research design that was employed in order to take
account of important variables;
+ the use of mixed methods;

+ the way the data was analysed;

+ the implications of these results for the hypotheses;
+ the consequent impact on thinking about unemployment and
sentencing;
+ the implications for practice and policy that followed.
In other words it very much illustrates the application of the
research process outlined in Chapter 1, from theory through hypoth-
eses to operationalization, data collection, analysis and interpreta-
tion.1
Theoretical context
As we explained in Chapter 1, a criminological inquiry should have
some theoretical connections, to enable the results to be interpreted
in relation to existing knowledge. This does not necessarily mean
that the study is putting some grand criminological theory to the
test; it may be several steps removed from theories that have been
advanced to explain crime and deviance. Nor need it necessarily have
a clear conceptual formulation from the start; this may become
clearer as the study progresses. However, to pretend that any enquiry
is atheoretical is to delude ourselves. Every question has some
underlying basis in theory, and the investigator needs to work at
making this explicit and exploring its implications. In the case of the
study described here, there were three stages of thinking about the
consequences of unemployment for offenders: (1) worklessness and
criminal justice; (2) worklessness and the way offenders are dealt
with; and (3) worklessness and sentencing.
The context for the first of these was a long-standing association
between worklessness and the law, going back to the Statute of
Labourers of 1349, and the Vagrancy Acts of 1824–35, which were
designed to control surplus labour, with references to ‘rogues and
vagabonds’, ‘idle and disorderly persons’, and ‘not having any visible
means of support’ (Home Office 1974). Offending has long been
linked with a background where work is in short supply, intermittent
and poorly paid, and criminological theories have included those
which argue that the deprivation brought about by little or no job
opportunities is at least a factor in explaining criminality. However,
our specific interest was in whether the unemployed were likely to
be dealt with differently to those who had jobs, and in what ways.
Again, there is much material to suggest that a defendant’s employ-
ment status plays a role at various stages in the criminal justice
process. It may be taken to reflect on a person’s character, such that
having a job and a steady work record will count in his or her
favour. Conversely, the lack of stable employment is regarded as an

Researching offenders and employment 201
adverse ‘risk factor’. Engels portrayed this graphically in his

nineteenth-century survey of the English working class:
But if a poor devil gets into such a position as involves
appearing before the Justice of the Peace … he is regarded
from the beginning as guilty; his defence is set aside with a
contemptuous ‘Oh! We know the excuse’, and a fine imposed
which he cannot pay and must work out with several months
on the treadmill. And if nothing can be proved against him, he
is sent to the treadmill nonetheless, as a ‘rogue and a
vagabond’. (Engels 1969: 306)
The second stage was to try to be more specific about worklessness

and the way offenders are dealt with. Much of what had been
written about this concerned the relationship between levels of
unemployment and the prison population (Box and Hale 1982).
However, the studies documenting this were mainly aggregate
studies based on trends in criminal justice and economic data (e.g.
Brenner 1976), and there was little understanding of why it should
happen. One possible explanation for this is that the prison
population rises at times of higher unemployment because there is
an increase in crime, which in itself results in more imprisonment.
But this depends on a variety of other factors, such as whether
enforcement and incarceration practices alter. One possible explana-
tion might be that societies, and their criminal justice systems,
become less tolerant and more repressive at times of economic
hardship (UNSDRI 1976). Our view was that there was probably
more to it than this, and we therefore used decision-making theory
as the basis for our inquiry. This posited that unemployment restricts
the options both of the unemployed and sentencers to impose
financial penalties, and meant that both courts and offenders had
fewer possibilities open to them than previously. So we did not just
want to establish whether there was an association between employ-
ment status and sentencing; we wanted to understand something of
the mechanisms that might be operating.
If fluctuating crime rates are taken out of the equation, then there
are still various ways in which employment status might affect the
likely use of prison. These are threefold:
1 Remand. The Bail Act 1976 referred to the relevance of ‘commu-
nity ties’, including whether the defendant had a steady job, as a
consideration influencing whether to remand a defendant in
custody or release him or her on bail.
2 Sentence. To what extent does the court take into account an
offender’s work status and record when deciding an appropriate
sentence?

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 24 OUTPUT: Wed Aug 29 08:47:52 2007 SUM: 4B2F9F25
3 Parole. Again, a prisoner’s ‘community ties’, including employment

prospects, may have been considered in deciding on early release.
It is seldom possible when undertaking a research study to cover all
the ground you would wish to. We therefore decided to focus on
what might be most likely to affect the way that offenders were
dealt with, and this was sentencing. This was reinforced by an article
in Justice of the Peace (1982: 700) which had stated that, ‘A man’s (sic)
employment situation has long been considered to be a relevant
factor in sentencing’. It also gave us the opportunity to look not just
at the use of imprisonment, but at whether unemployment affected
other disposals such as fines, community service, and probation. The
research question with which we started therefore became, what
impact does rising unemployment have on sentencing, and what are
the consequences for the individuals concerned, and for criminal
justice generally? We could discuss the topic itself at much greater
length, but what we want to highlight here is the way in which a
topic of general interest at the time became progressively refined
into a manageable, researchable question.
The research
Hypotheses
Having defined our terrain, the next step in the research process was
to develop hypotheses which could be put to the test. We developed
21 specific hypotheses, which fell into five main groups. These can
be summarized as follows:
1 Employment information is important to courts in making
decisions about offenders. E.g. Hypothesis 1: ‘Employment status
and history will generally feature as an item of information
presented to the courts at various stages of the process.’
2 Unemployed offenders are more likely than those with jobs to be
the subject of pre-sentence reports, and to receive different
recommendations. E.g. Hypothesis 14: ‘Unemployed offenders are
more likely to be the subject of an SIR.’ At the time of the
research, pre-sentence reports were called social inquiry reports
(SIRs). Whether they were asked for, and how they were drafted,
was more variable then than now.
3 Other things being equal, unemployed offenders are likely to
receive different sentences to those who are employed. E.g.
Hypothesis 4: ‘Unemployed offenders are more likely to receive a
custodial or potentially custodial disposal (such as suspended
sentence or committal for sentence) than employed offenders.’

JOBNAME: Mcgraw−crowsemmens PAGE: 7 SESS: 24 OUTPUT: Wed Aug 29 08:47:52 2007 SUM: 4ED16A0E
4 Unemployed offenders are likely to receive different amounts of

sentences to those who are employed. E.g. Hypothesis 10: ‘Unem-
ployed offenders given probation orders will be more likely to be
given longer probation orders than comparable employed offend-
ers.’
5 Local circumstances and sentencing traditions affect the hypoth-
esized disparities in court decisions between employed and unem-
ployed offenders. E.g. Hypothesis 18: ‘Employment status will be a
more important factor in an area of low unemployment than in
an area of high unemployment, and intermediate in a transitional
area.’ This was included because of previous research (Hood 1972;
Tarling and Weatheritt 1979) showing that in magistrates’ courts
there tended to be a ‘bench tradition’ of favouring some disposals
more than others. It was also necessary to take the local social
and economic conditions into account. More will be said about
these factors below when describing the research design.
Note the way the hypotheses were expressed. This is in the form of
an assertion that such-and-such is the case. The purpose of the
empirical inquiry is then to put these assertions to the test; in other
words, to see whether the hypotheses can be refuted. If they are,
then alternative formulations need to be considered. This principle
of refutation is based on the proposition that in the social sciences it
is seldom possible to prove something categorically. Because many
factors are usually in play, and alternative formulations are generally
possible, one’s task is to eliminate possibilities.
Research design
The next stages in the research process involve developing a project
to test the hypotheses, and this means devising a suitable research
design, defining terms, operationalizing variables, and deciding on
what methods to employ. (For anyone who is tempted to think that
research means going out and interviewing people, or analysing
statistics, it is worth noting just how far along in the research
process the selection of specific methods occurs.) Of course, the
hypotheses are framed in the knowledge that this is what has to
happen, so they are always likely to be drafted with these consid-
erations in mind. It is not simply a matter of working your way
through the process one stage at a time; each part of the process has
to be done with the consequences for other stages in mind.
Back to the research design. If you wish to look at the
relationship between unemployment and sentencing, what is the best
way to do it? Here again some choices have to be made. It is not
possible to study every court in any depth. Even a sizeable sample of
courts could only be studied at the most superficial level, since the

main task would involve having to go through a significant number

of cases in order to establish the employment status of defendants.
Although we wanted to determine whether and what kind of
relationship there was between employment status and sentence, we
also wanted to understand what was happening, and what the nature
of any relationship was, and this meant studying courts in some
detail. We also had to limit our enquiries to magistrates’ courts,
because obtaining access to Crown Courts at that time would have
been very difficult, if not impossible (Ashworth et al. 1984). In
addition, the factors operating at Crown Courts may well have been
different to those operating at magistrates courts, and would have
required a separate study.
The study was defined in two other ways as well: it would be
restricted to adults and to men. This is because very different
considerations are likely to apply to young people, and also to the
relationship between sentencing and the employment situation of
women. The reason for this sequence of limitations was in essence
that it was important to control for as many variables as possible.
Many factors affect sentencing, and although it would have been
highly desirable to cover a full range of situations and all types of
defendant, not only would an exhaustive study have demanded
massive resources, but even then it would have been difficult to take
account of all the variables. In effect, one would have had to mount
several different studies, and it is generally better to address one
aspect of an issue well than to cover all aspects ineffectually. Of
course, defining a study in this way also places limits on how far the
results can be applied, but the complete picture regarding a topic
will usually depend on a number of researchers undertaking studies
on different aspects, and there have been several studies of the
impact of unemployment on offending and criminal justice, some of
them referred to here. Our study contributed something to what was
already known, but also left considerable scope for others to
contribute.
We also needed to take account of the fact that although
unemployment at the time was high, it did vary from one part of
the country to another. As explained above, we needed to take into
consideration the sentencing perspectives of different benches, and
the possibility that the relationship between employment status and
sentencing may be different at courts with a more punitive approach
to sentencing than at courts which favoured a less punitive
approach. Given the resources likely to be available to us we
therefore proposed a study which involved six magistrates courts as
shown in Table 9.1.
Two courts were selected in the North-East of England to
represent an area with historically high levels of unemployment (H1
and H2), where it might be supposed that the courts were used to

JOBNAME: Mcgraw−crowsemmens PAGE: 9 SESS: 24 OUTPUT: Wed Aug 29 08:47:52 2007 SUM: 250D513F
Table 9.1 Research design
dealing with an above average proportion of unemployed offenders.

To contrast with these, two courts were selected in the South-East
of England where, even at a time of high unemployment nationally,
levels of joblessness were lower than elsewhere (L1 and L). London
was excluded because of the different nature of London courts, and
because there was a wide variation in levels of unemployment within
London. Finally, two courts were selected in the West Midlands
(LH1 and LH2). This was an area that had undergone a transition
from relative affluence and low unemployment, to severe recession
and high unemployment during the 1980s, and it was therefore an
interesting area to study to see what effect this might have had. The
levels of unemployment and sentencing patterns were checked by
reference to national statistics held by the Government departments
concerned.
The research methods

At each of the six courts three methods of data collection were used,
summarized in Table 9.2.

JOBNAME: Mcgraw−crowsemmens PAGE: 10 SESS: 24 OUTPUT: Wed Aug 29 08:47:52 2007 SUM: 36233CCE
Table 9.2 Research methods used
+ Case records: A form was used to abstract information from the

case files of over 500 adult males at each court who had been
sentenced for property offences, mostly burglary, theft and crimi-
nal damage. Again, the selection of property offences aims to
limit the amount of variability likely to occur, since other types of
offence might be dealt with differently. The information extracted
covered the man’s offence and criminal record, employment status,
the sentence given, any reports that there were, and any other
relevant background information.
+ Observations: These were carried out in order to get a better
understanding of the process leading to court decisions, who said
what to whom, and where employment status stood in relation to
the decision-making process.
+ Interviews: Semi-structured interviews were undertaken with key
people, not in order to obtain a representative sample, but because
they occupied an important role with regard to the court process
and were able to offer insights and perspectives on how employ-
ment status might affect the way the court made decisions.
The methods constituted a form of triangulation (Denzin 1988).
The hypotheses could therefore be examined from three vantage
points, and support for any hypothesis would be strongest if
provided by consistent evidence from all three methods.
Operationalizing the design and conducting the study

The next stage in the research process involved identifying depend-
ent, independent and intervening variables (see Figure 3.1 on p.41)

JOBNAME: Mcgraw−crowsemmens PAGE: 11 SESS: 24 OUTPUT: Wed Aug 29 08:47:52 2007 SUM: 2FC3DBAB
and deciding how they were to be measured. In one sense this was
comparatively easy. The dependent variable was clearly the disposal
in terms of the sentencing decision, and the independent variable
was employment status. However, it is also clear that the sentence a
person receives is determined by considerations other than their
employment status, in particular by the seriousness of the offence,
and their criminal history. It is important to take these into account
(control for them) when examining the extent to which employment
status plays a part in sentencing. Although one can never be sure
that one has taken every facet of a complex decision into account,
earlier research offered an insight into the main factors that
influenced sentencing (Thorpe 1978; Phillpotts and Lancucki 1979).
We therefore developed a composite scale, the ‘Offending Score’, to
take account of these other factors. This score included the current
offence, the number of charges, the value of the property involved,
the number of previous convictions the offender had, the interval
between convictions, the similarity of previous offences, and previous
sentences. It ran from a minimum of three points for a first offender
convicted of minor criminal damage, to a maximum of 19 points for
a serious and persistent burglar. The sentencing outcome was based
on a tariff of disposals running from conditional discharge to
immediate custody. Thus, the definition of the variables was as
shown in Figure 9.1.
Figure 9.1 Variables used in the study

JOBNAME: Mcgraw−crowsemmens PAGE: 12 SESS: 24 OUTPUT: Wed Aug 29 08:47:52 2007 SUM: 4B13715E
The fieldwork took place over a period of two years, including a

pilot study conducted in a different part of the country from those
to be covered by the main study. It involved spending two months at
each court. An abstraction form was used for collecting the material
from court files. This was not a straightforward process because,
while certain information such as the sentence could be easily
identified, it often took time reading through the various case files
to find other information, such as that relating to employment
status. Systematic notes were taken of court observations and
interviews, and the interviews were also recorded for subsequent
reference.
The analysis of the observations and interviews was largely
descriptive, although the hypotheses acted as the main point of
reference in recording observations and conducting interviews, and
in identifying themes in the interview material. These methods
yielded information on attitudes towards the unemployed, and
importantly on what information the bench received. The data from
the court records was analysed quantitatively, using multiple regres-
sion. This attempted to predict, for each type of sentence, the
probability of an offender receiving that sentence, as against all
others, taking into account, first, the offending score, then employ-
ment status, as independent variables. The data did not meet all the
statistical assumptions required for multiple linear regression, but
further analysis using a multinomial logit model confirmed the
general pattern of results.
Results
The main focus of this chapter is on how the research was done,
rather than on the substance of the study. However, it might be of
some interest to briefly summarize what did come out of the project.
One of the first points to emerge, mainly from the observations, was
to highlight the important role that information of any kind plays in
the court process, which underlines the necessity of getting the best
possible information to courts.
More specific conclusions were related to the hypotheses on
which the study was based. As outlined earlier, these fell into five
main groups, and the findings can therefore be summarized in
relation to these:
1 Employment information is important to courts in making decisions
about offenders. This was supported although, as might be
expected, information about employment was subsidiary to details
about the offence and offending history of the offender.

JOBNAME: Mcgraw−crowsemmens PAGE: 13 SESS: 24 OUTPUT: Wed Aug 29 08:47:52 2007 SUM: 4F9688D2
2 Unemployed offenders are more likely than those with jobs to be the
subject of pre-sentence reports, and to receive different recommenda-
tions. There was some support for the hypothesis that unemployed
offenders were more likely to receive social inquiry reports.
Looking at whether courts followed the recommendations of SIRs
or not, there was no support for the hypothesis that where courts
don’t follow recommendations, sentencing was more severe.
3 Other things being equal, unemployed offenders are likely to receive
different sentences to those who are employed. There was a significant
tendency for the unemployed, at the end of the day, to be more
likely to go to prison than those in work, but the main impact of
unemployment on sentencing lay elsewhere in the range of
disposals available to courts at the time. In particular, the
unemployed were less likely to be fined and more likely to be
given community service orders (CSOs) instead.
4 Unemployed offenders are likely to receive different amounts of sen-
tences to those who are employed. This was supported as far as fines
were concerned: unemployed offenders were fined less than those
who were employed. However, there was no support for this
hypothesis in relation to other sentences, such as CSO and
imprisonment.
5 Local circumstances and sentencing traditions affect the hypothesized
disparities in court decisions between employed and unemployed offend-
ers. It was found that what one might call more traditional
attitudes towards the unemployed (e.g. that they were shirkers)
persisted where unemployment was low and/or custody rates were
high. In areas where unemployment was higher, there was some
sympathy for the plight of the unemployed, and in fact in one
area a number of the magistrates had experienced unemployment.
Although the study was limited in various respects, as outlined
above, it was possible to make some extrapolations about the impact
that rising unemployment was likely to have on the sentencing of
adult male property offenders. The most significant impact was in
the use of fines. During the period preceding the study there had
been a decline in the use of the fine nationally, and this study
helped to explain how that had come about. Unemployment trig-
gered a ‘flight from the fine’ to other disposals, especially towards
the use of community service orders. The research contributed to the
debate about the need for fines that were graduated more towards
the means of offenders, something which was implemented in the
Criminal Justice Act 1991, but subsequently rescinded in the
Criminal Justice Act 1993. However, the use of fines and the
problem of fine enforcement continues to be an issue to this day.
The criminological significance of the study lay in explaining exactly

JOBNAME: Mcgraw−crowsemmens PAGE: 14 SESS: 24 OUTPUT: Wed Aug 29 08:47:52 2007 SUM: 268D75FD
how social and economic changes and criminal justice are inter-
related at the level of decision-making rather than just as aggregate
phenomena.
Conclusion
So to some final points, to highlight the main features of this study

in terms of the practice of criminological research:
+ First, the study shows how a broad area of interest becomes
progressively defined in order to render it susceptible to empirical
investigation.
+ Second, it should be noted how the stages which this study went
through reflect the research process outlined in Chapter 1. As
explained there, this need not always be the way that crimino-
logical research proceeds, but elements of that model will be
relevant to criminological research at various points in its
progress, even if not in that exact order.
+ Finally, we would highlight the key role played by the use of
hypotheses in linking the parts of the study together. The
hypotheses made it possible to turn some general issues into
specifically observable events, and then made it possible for the
results to be analysed and related to the initial focus of inquiry.

10 Researching theYouth Court
The Youth Court is the successor to the juvenile court, which dealt
with children under the age of 16. The Criminal Justice Act 1991,
s. 70 replaced the juvenile court with the Youth Court, dealing with
a wider age range, from 10 to 17 years old. The study described here
was undertaken for the United Kingdom Home Office as part of its
plans to change the way that Youth Courts operate in England and
Wales. The study shows how a combination of different research
methods is used in conjunction with each other. It also illustrates the
importance of being able to work with a variety of different people,
representing different interests. We will start by describing the
background to the project, next explain how the study was carried
out, and then discuss the way in which criminal justice research
often takes the investigator into engaging with the social and
political context in which research operates.
During its first term, from 1997 onwards, the Labour Govern-
ment was keen to make a number of changes to the administration
of justice, especially as regards youth justice. For example, as part of
its pledge to speed up the way cases were dealt with, it introduced
a system of ‘fast-tracking’ persistent young offenders, so that they
would be dealt with more speedily (Crow and Stubbing 1999).
Shortly after coming to power the new Labour Government
published a White Paper, No More Excuses: A New Approach to
Tackling Youth Crime in England and Wales (Home Office 1997).
Many of the proposals in this White Paper were implemented in the
Crime and Disorder Act 1998. However, the Youth Court Demon-
stration Project (YCDP) sought to change the culture of the Youth
Court without the necessity for legislative change.
The specific impetus for the project was a speech by the former
Home Secretary, Jack Straw, that the work of the Youth Court was
like a ‘secret garden’.1 This reflected the fact that, while the Youth

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 26 OUTPUT: Wed Aug 29 09:11:00 2007 SUM: 4B2F86A7
Court is in some respects like a junior version of the magistrates’

court, many aspects of its work were not open to the public. The
court sits in private and there are restrictions on reporting its
proceedings, including the identities of those appearing before it.
The Youth Court operates with a Panel of magistrates who receive
special training. The result is a system that attempts to balance
judicial requirements with a regard for the age and limited experi-
ence of those who appear in court. But for those not familiar with
the Youth Court, there can be a lack of awareness about exactly how
it works and what ends it is seeking to achieve.
The aims of the project were twofold. First, it aimed to have an
impact on the offending behaviour of those appearing before the
court, in the hope of reducing future offending. Second, it sought to
increase the confidence of victims and public in the Youth Court.
Around the time that the project was taking place, a survey was
published showing that the Youth Court was an aspect of the
criminal justice system in which the public had little confidence
(Mattinson and Mirlees-Black 2000).
The project sought to address these aims through a series of
related initiatives. The first was to increase the extent to which
magistrates engaged with offenders. Prior to the project it was quite
common for young offenders to take little or no part in proceedings.
Much of the business was transacted between court officials and
lawyers. It was hoped that if magistrates addressed young people
directly, they would take more responsibility for their actions.
Second, it was decided to experiment with the layout of the court.
Many courtrooms are set out in a very formal fashion in such a way
as to ‘distance’ magistrates and officials from lay persons including
defendants, witnesses and the public. In some instances this separa-
tion is achieved by the fact that the magistrates are on a platform
elevated above others in the courtroom. The extent to which the
layout could be changed was limited by the architecture in some of
the older courts, but generally attempts were made to make the
court less formal, and to bring magistrates and offenders closer
together.
A third initiative was to try to open the court up more by
encouraging attendance by victims and the press, and to lift
reporting restrictions in certain cases. This last practice went under
the soubriquet of ‘naming and shaming’.
Finally, the project gave feedback to sentencers about sentencing
patterns in the courts and other aspects of the court’s work. This
was intended to enable magistrates to have a better idea of what
effects their decisions were having.

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 26 OUTPUT: Wed Aug 29 09:11:00 2007 SUM: 510B126F
Researching the Youth Court 213
Theoretical context
Although much attention has been devoted to youth crime and

young offenders by the media, politicians, the public, and criminolo-
gists, at the time this study was carried out relatively little research
had been done on the Youth Court itself, especially the perceptions
of the young offenders appearing there (Allen et al. 2000: 1).
However, like many public policy initiatives the YCDP touched on
a number of issues, but was not based on a particular, explicit theory.
Nonetheless, this does not mean that it was without any theoretical
foundations. There were, for example, clearly some implicit notions
about the ‘responsibilization’ (Garland 2001: 124–7) of young
offenders. The project also needs to be seen in the context of the
historical development of youth justice and the nature of the Youth
Court. The juvenile court had a more welfare-oriented approach, but
even before it was replaced by the Youth Court in the Criminal
Justice Act 1991, cases requiring ‘care’ had already been removed
from its jurisdiction by the Children Act 1989. The decline in
rapport and offender involvement reflected the replacement of the
juvenile court with a court which mirrored the adult court, and
placed more emphasis on just deserts. In the juvenile court there had
been more of a sense of ‘sitting round a table’ to resolve the
problems which underlay a young person’s offending. The moves to
change the layout of the court and to engage with young offenders
more could be seen as an acknowledgement of the value of such an
approach. However, it did not signal a return to a welfare-based
approach to dealing with them; the emphasis was very much on
addressing offending behaviour.
The YCDP also had a human rights element, which became more
important as the project progressed. There was concern about the
rights of young people when they appeared in courts, and this had
particular relevance since the Government was at that time commit-
ted to introducing a UK Human Rights Act. The project was taking
place when the European Court of Human Rights delivered a
judgment in the case of two boys convicted of killing the little boy
James Bulger in 1993, a case which had attracted considerable
publicity and public concern. The judgment stated that the trial of
the 10-year-old boys had placed them in a situation in which they
were unlikely to understand what was going on. As a result of this,
the Lord Chief Justice issued a Practice Direction which, although
directed at Crown Courts, related to measures implemented as a
result of the YCDP. This meant that in future all courts dealing
with young people in the UK would have to have regard to this
ruling.

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 26 OUTPUT: Wed Aug 29 09:11:00 2007 SUM: 4DE017F4
The research
Before introducing any changes on a national basis it was considered

advisable to try out some reforms to existing procedure in two areas
to evaluate their impact. Consequently, in the summer of 1998
courts around the country were invited to apply to take part in the
Demonstration Project. Two areas were selected by the Home
Office, and the Project started in October 1998. This means that in
effect the research design had already been chosen prior to the
research team being involved. One of the areas was a single court
located in the centre of a medium-sized town. The other area was a
county which had five courts, one of which was in the centre of the
main city of the county, and the other four in towns in the county.
The research on the project took place over an 18-month period,
and was essentially a process evaluation. This means it was con-
cerned with how changes were implemented and what could be
learned from the process of change. Its purpose was to look at how
the initiatives were implemented, what problems were encountered
and resolved, how the various parties involved responded, and what
could be learned prior to the changes being extended to other Youth
Courts. In other words, its main aim was to assess the viability of
the changes that were made. It was not intended to look at
outcomes in terms of any changes in sentencing, or whether the
projects affected reconviction rates.
Because of this it was necessary to gather a wide variety of
information, both quantitative and qualitative. The main methods
employed were observations, semi-structured interviews, question-
naires, and a small number of detailed case studies in each area.
These were supplemented by analysis of logbooks in which magis-
trates and other court users could write comments, feedback on
sentencing data supplied by the Home Office, and a telephone
survey of ten other courts in the country where similar developments
had taken place on the initiative of the individual courts.
Observations
Observations formed the backbone of the study and took two forms.
The first was attendance at key meetings, 36 of them in all. These
mainly involved meetings related to the project itself, such as
Steering Groups and Project Board meetings, and training sessions,
but also included 14 of the regular Panel and Court User Group
meetings, at which the project was discussed. These meetings
enabled the researchers to keep track of the progress of the project
and the issues that arose at every stage of its progress.
The other kind of observations were in the courtrooms them-
selves. The main purpose of these was to examine the interaction

JOBNAME: Mcgraw−crowsemmens PAGE: 5 SESS: 26 OUTPUT: Wed Aug 29 09:11:00 2007 SUM: 2DA911F3
between those in the court, in particular, whether the extent and

nature of engagement between young offenders and magistrates
changed as a result of the Demonstration Project. Altogether 1293
hearings were observed during 110 sessions, but in analysing the
data most attention was paid to the 30 per cent of hearings where
offenders were sentenced, since this was the occasion when the most
important exchanges between magistrates and offenders were likely
to occur. When observing interactions such as those taking place in
a court, the most important thing is to get as much information
down as possible in the limited time available. To do this a one-page
pro forma was used to record details of the case and a sociometric
diagram was completed to show who spoke to whom, and how
often. An illustrative example is shown in Figure 10.1.
This example shows quite a lot of interaction taking place
between the magistrates and the defendant and his support (in this
case, his mother). The blank boxes could be used if any other people
were present, such as a probation officer or social worker. Fuller
details can be found in the Appendix to this chapter, which includes
the pro forma and shows what kinds of exchanges were recorded.
Figure 10.1 Diagram of court interactions

Note: Arrow heads show how many times people speak to each other.

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 26 OUTPUT: Wed Aug 29 09:11:00 2007 SUM: 4F2D245C
Questionnaires
The other main kind of method used was to ask people questions.
Again, this took two forms: questionnaires and interviews. While
the main target for the questionnaires was the magistrates, a
questionnaire was sent towards the end of the first year of the
project to anyone known to be involved with the work of the Youth
Courts in both areas, including clerks and other court staff, lawyers,
and social workers who attended the court regularly. A follow-up
questionnaire was sent to the same groups of people during a
follow-up period about six months later.
The questionnaires asked people how much they knew about the
YCDP, how they viewed the project, and about the changes that had
taken place. The follow-up questionnaire covered similar ground, but
also asked whether things had changed in the preceding six months,
and explored whether people’s attitudes towards the changes had
shifted during that time.
An important consideration when using questionnaires is the
response rate. One of the main justifications for using questionnaires
in this study was that the research was done with the support of the
Home Office and Lord Chancellor’s Department, and with the
willing co-operation of senior magistrates and court staff. A good
response rate was therefore expected. However, the response rate to
the first questionnaire was 54 per cent and to the follow-up
questionnaire 51 per cent. As mentioned in Chapter 6, social
researchers generally look for a response rate of around 60 per cent
in order to have confidence in their results, and the rate in this
study clearly fell short of that target. There were reasons for
believing that the responses were not as disappointing as the basic
figure suggests. For example, there were a number of magistrates on
the register who did not attend court very often, and there were
others such as social workers, who only appeared at the Youth Court
infrequently. It could also be argued that every effort was made to
ensure that those who wished to had the opportunity to make their
views known, and that those who did not probably didn’t have
strong feelings about the demonstration project. Nonetheless this
illustrates the difficulty of achieving good response rates for ques-
tionnaires, even with considerable support and assistance. Anyone
contemplating using a questionnaire needs to think carefully before
embarking on something that could involve a lot of time and effort,
producing disappointing results. In the case of this project we felt
the nature of the enquiry was such as to justify using the results of
the questionnaires, but clearly the results need to be qualified.
In addition to the questionnaires, 26 semi-structured interviews
were conducted with key personnel such as the Chairs of the Youth
Court panels, and the chief clerks, and a selection of magistrates and
other court users at three stages of the project: before it started, after

JOBNAME: Mcgraw−crowsemmens PAGE: 7 SESS: 26 OUTPUT: Wed Aug 29 09:11:00 2007 SUM: 47B1F14C
it had been in progress for six months, and during the follow-up
period. The purpose of these interviews was to gain detailed insights
in to how the courts were operating.
Case studies
Twenty cases in each of the two areas were studied in some detail to
examine at greater depth how the project affected individual cases.
The progress of each case was observed through the court process to
sentencing. Afterwards offenders and their support, if in court, were
interviewed, as were the magistrates who sat on the case. Defence
solicitors were interviewed where possible for an additional perspec-
tive. One case study involved an offender who was named in the
local newspaper, so there was an additional interest here on the
impact of lifting reporting restrictions. Telephone interviews were
also carried out with the clerks of ten courts which had introduced
some procedures similar to those taking place in the YCDP of their
own volition about what had been done and with what results.
Feedback
Although magistrates’ courts produce information about their activi-
ties, in the past this had tended to focus on administrative and
procedural matters such as workload and time intervals in dealing
with cases. Before the project started, relatively few details were
available about such matters as court sentencing patterns, breaches
and reconviction rates. During the course of the project, discussions
took place about the type of feedback that magistrates might like to
have regarding youth justice matters at their court and, in conjunc-
tion with the Home Office Research Development and Statistics
Division, newsletters were produced and discussed at subsequent
meetings. This development was welcomed by magistrates and other
court users, who were particularly interested in receiving information
about reconviction rates and the extent to which sentences were
breached or completed.
Analysis
The material collected was analysed by a mixture of quantitative and

qualitative methods. Written material from interviews, observations,
and logbooks was analysed by hand, and ordered in such a way as to
reflect the views of various parties towards the different initiatives. It
was a source of quotations that could be used to illustrate the
various perspectives, such as the comment of a magistrate regarding

JOBNAME: Mcgraw−crowsemmens PAGE: 8 SESS: 26 OUTPUT: Wed Aug 29 09:11:00 2007 SUM: 5269637B
engagement: ‘Defence solicitors speak in well oiled grooves and it is

refreshing to hear from defendants; you are hearing something new
rather than the standard mitigation.’ This was mirrored by the
defendant who said, ‘At least they know that it’s coming from you,
and not something that the solicitor’s made up.’
Clearly it is important when analysing and using such material
not to be selective. It is useful to have a combination of quantitative
and qualitative material because they can be used to complement
each other. The qualitative material enables one to gain a better
understanding of what the figures mean, while the numerical data
gives one a broader view of the whole picture than is possible when
seeing just a few cases, or talking to just a few people.
The observations yielded both quantitative and qualitative mate-
rial. Quantitative analysis involved examining the number of times
magistrates talked to defendants over a period of months to see
whether there was a pattern of increasing engagement, whether they
asked more questions, and the number of times that defendants
themselves talked. It was also possible to look at whether the level
and type of engagement were affected by changes in the layout of
the court.
When analysing the questionnaires, distinctions were made
between the views of magistrates and those who occupied different
positions in the court, such as clerks, solicitors, ushers, guards and
police. Differences between the different court locations also had to
be considered, and when analysing the follow-up questionnaire
comparisons were made with the first questionnaire. Thus, the main
independent variables were court occupation, location, and initial
versus follow-up questionnaire. The main dependent variables were
those which asked people for their views and perceptions, often in
the form of Likert-type scales. Usually simple bivariate statistics
such as chi-square or t-tests were appropriate to establish whether
there were statistically significant differences between such things as
whether people with different roles in the Youth Court had different
views about the changes in the layout of the court.
Policy-makers like the allure of hard ‘facts’, and there is a danger,
where public policy is concerned, that the numbers will dominate
the results. It is therefore important to strike the right balance
between different methods. One way of achieving this is in the way
that the results are presented. For example, the results from this
study could have been reported by presenting the results from the
questionnaires, followed by the results from the observations, fol-
lowed by the results of the interviews, and then pulling together the
results from the different methods. Not only would this have
produced a rather tedious report but it might have given the
impression of a ‘hierarchy’ of results. Instead we opted for a thematic
presentation, using material from each method as appropriate. Thus,

JOBNAME: Mcgraw−crowsemmens PAGE: 9 SESS: 26 OUTPUT: Wed Aug 29 09:11:00 2007 SUM: 4FEDE68C
when reporting the results of attempts to encourage magistrates to

engage more with young offenders and their families, material was
drawn first from the interviews and questionnaires, then from the
case studies, then from the telephone survey of other courts, and
then from the court observations, with comments from the logbooks
also being used. Further details of how each kind of development
was evaluated using each method are given in the Appendix to this
chapter.
Results
Although the purpose of this chapter is to focus on the way the

research was done, it may be of some interest to note that there was
broad agreement with the aims of the project on the part of the
majority of those involved. Reservations were expressed by some
magistrates and court staff about some aspects of the initiatives, such
as court security and whether the changes might reduce the
authority of the court. One of the main points noted was the fact
that young people change a great deal during the course of the age
range covered by the Youth Court; there is a considerable difference
between dealing with a 10-year-old first time offender and a
17-year-old for whom a visit to the court has become a familiar
experience.
Magistrates engaging with defendants and their families was
regarded as one of the most welcome and successful aspects of the
Project, but it was noted that training and guidance were needed if
this was to be successful. It was also noted that some of the aims
conflicted somewhat: a better engagement between magistrates and
offenders could be inhibited by the presence of victims and the
press, and the prospect of being ‘named and shamed’. Although the
project was not seen as particularly concerned with changing the
way the Youth Court sentenced people, some magistrates said that
engaging in discussions with offenders had on occasions caused
them to reconsider the sentence they were initially inclined to
impose. In one of the areas there was a shift in sentencing patterns,
with a significant decrease in the use of fines, supervision orders and
being sent to a Young Offender Institution, and an increase in the
use of discharges and probation orders.
There were mixed reactions on the part of court personnel and
defendants to the changes in layout that were made. Some felt it
was possible to get used to the new arrangements and that they
facilitated greater engagement. Others felt that space became very
restricted, with people sitting uncomfortably close to one another.
The indications from a follow-up study were that those involved
became more used to the new arrangements with the passage of

time, but this aspect of the project was regarded as having had the
biggest impact on the culture of the Youth Court.
Attendance by victims was supported in principle, but there were
problems in ensuring that it operated satisfactorily. In practice, few
victims wished to attend court other than as a witness. What
seemed to be most important to victims was knowing about the
outcome of a case. Attendance by the press was difficult to ensure in
practice and could be counterproductive in unduly increasing con-
cern about youth crime, but the efforts at forging better links with
the local press did raise awareness of the need for courts to make
more effort to keep in contact with, and have a positive relationship
with the press.
Although it was felt that it was useful for magistrates to have the
power to lift reporting restrictions on occasions, there was broad
agreement that this was a power that would be used only occasion-
ally, in instances where the public needed to know that someone was
a serious and persistent menace. It only happened twice during the
period of the YCDP itself. There was concern that ‘naming’ a young
offender could be counterproductive and give the young person
involved an enhanced and undesirable status among his or her peers.
It is worth noting from both the policy and research point of view
that the YCDP also had a general ‘galvanizing’ effect. It produced
more communication, both in the form of such things as newslet-
ters, and in the interactions that took place, not just between
magistrates and offenders, but between courts and press, and to
some extent between the people involved with the courts. In one
area a special meeting was arranged between magistrates and young
people to talk about the ways in which each saw the other, and this
was an unusual and successful event. This galvanizing quality of
research is not unusual. People respond to the fact that what they do
is worth studying, but it also means that the researchers do not
necessarily get a ‘real’ (i.e. normal) picture. It also means that
whatever happens during a period of research is not necessarily
maintained beyond the study period.
Conclusion
If we think of this project in terms of the model of the research

process set out in Chapter 1, then at first sight it appears to have
only a slight resemblance to the process outlined there. The project
did not come about because a criminologist (or group of criminolo-
gists) decided that here was an interesting and important research
question, derived from theory, with testable hypotheses. It came
about because of a Government agenda which required the feasibil-
ity of certain developments to be tested before being used more

JOBNAME: Mcgraw−crowsemmens PAGE: 11 SESS: 26 OUTPUT: Wed Aug 29 09:11:00 2007 SUM: 55E827DF
widely. This is a common way for research done by criminologists to

occur; remember we did say that the model was an ideal type, and
not everything happens exactly that way. But the model is nonethe-
less relevant. As mentioned earlier, there were implicit theoretical
and conceptual issues, and there were implicit hypotheses requiring
empirical verification. It’s just that these hypotheses related more to
the viability of changes, the acceptability of changes to the parties
concerned, and what impact the changes had on court functioning.
Studies such as the one described here have some, but not all, of the
features of what is often referred to as administrative criminology
(McLaughlin 2001). This term is sometimes used in a pejorative
sense, but such research can play a role in advancing understanding
of the criminal justice process, and the way in which changes occur.
The YCDP research can be compared with studies that have
taken place in the past, which have evaluated intensive probation.
One of the classic studies of intensive probation was the IMPACT
study that took place in the early 1970s (Folkard et al. 1976). This
focused very much on the outcome of reconviction rates, using a
traditional clinical trial research design, with experimental and
control groups and random allocation. The largely (but by no means
totally) negative results helped to reinforce the perception gaining
ground during the 1970s that ‘nothing works’. By contrast, a study
of another attempt to develop intensive probation some 15 years
later took a very different approach (Mair et al. 1994; Mair 1997:
70). The research involved process evaluation, and concentrated
more on how the programmes operated. It focused on the viability
of such an initiative, whether it reached the target group of high risk
offenders, whether it diverted people from custody, and the views
and opinions of sentencers and offenders, more than whether it
reduced re-offending. It was as much about management as treat-
ment, and reflected a very different theoretical, political and meth-
odological environment to that which had prevailed at the time of
the IMPACT experiment. The YCDP research described in this
chapter is more like the second intensive probation study than the
first. The Government’s Crime Reduction Programme, which started
in 1998, sought to examine ‘what works’, and evaluations during the
early part of the twenty-first century once again tended to focus on
outcomes measured by reconviction data. However, there are advo-
cates of what is referred to as pluralistic evaluation who support a
diversity of measures (Smith and Cantley 1984; Gelsthorpe and
Sharpe 2006; Israel and Chui 2006). Clear (1997) and Israel and
Chui (2006), for example, have suggested that researchers should
consider the extent to which initiatives can benefit criminal justice
organizations. The study described in this chapter falls into that
category.

JOBNAME: Mcgraw−crowsemmens PAGE: 12 SESS: 26 OUTPUT: Wed Aug 29 09:11:00 2007 SUM: 7FFBE58A
Another point to note about the YCDP study is the mixture of

methods used, and the way in which they interweave with one
another. A one-dimensional evaluation concerned purely with recon-
viction rates may well need to have regard to some process material
to determine whether a project was implemented as intended, but is
more likely to place statistical material at the centre of analysis to
determine the results.
Finally, it is worth noting that the research carried out on the
Youth Court Demonstration Project required the researchers to
interact with a wide range of people in a variety of settings and
contexts. This is a skill that is often overlooked in textbooks on
research methods. The criminologist not only needs to know how to
design and implement a project, but needs interpersonal skills if the
research is to be successful. A 30-page, structured questionnaire has
minimal value if one is engaging with homeless, petty persistent
offenders who frequently have alcohol and/or other drugs in their
bloodstream. Similarly, talking to magistrates and judges requires a
certain decorum if one is to gain their confidence in getting them to
tell you about their work. The YCDP meant being able to relate to
magistrates and young offenders alike. In both cases it involved
gaining their trust that what you were told would not only be
treated in confidence, but would not disadvantage them. Inevitably
when one seeks to change structures, as the YCDP was threatening
to do, people feel defensive and guarded. Another party to this
research was the Government, or rather officials of the Home Office
and then Lord Chancellor’s Department who represented the inter-
ests of the Government. Organizations at the local level – the Youth
Courts in this instance – are often wary of the intentions of central
Government, and it can be hard for researchers to tread the line
between the two, especially if the central authority is the customer
paying for the study. It takes time to overcome people’s inhibitions,
without at the same time becoming partial to one party or another.
It helps to build in occasions when participants will receive feedback
on how the research is progressing, as we did in this study.
Appendix: Youth court observation form
Defendant and Hearing

Name: John Smith Sex: Male
Date of Birth: 23/10/84 (16) Ethnicity: White
Court: Ro Court 1 Date of Observation: 8/2/00
Time Started: 2.19 Time Ended: 2.34
List Number: 5 Number of Charges: 1

JOBNAME: Mcgraw−crowsemmens PAGE: 13 SESS: 26 OUTPUT: Wed Aug 29 09:11:00 2007 SUM: DBCED1BE
Type of Charge: Criminal damage – broken window

Number of Previous Hearings: 0
Purpose and Outcome of Hearing: Sentence – 100 compensation order
Date of Next Hearing ? N/a £
Rep. Rest. Application ? No Reason:
Accepted: ? Reason:
Court Layout: traditional 䡺 alternative 䡺 well of court 聺
Details: Defendant next to Mum
PSR: requested? 聺 used? 䡺
Number of retirements: 1
In Court (no. of each)
Magistrates __3_ Stipendiary ____
Clerk __1_ Defence solicitor __1_
Defendant/s __1__ Prosecution solicitor __1_
Parents __1__ Youth Justice __1_
details: Mum Bail support ____
Other family ____ Victim ____
details: Witnesses ____
Other support ____ Press ____
details: Probation officer ____
Guards ____ Other Prosecutor ____
Usher __1_ details:
Other ____
details:
Total No. _10_
Note: Rep. Rest Application=Application for reporting restrictions to be lifted

JOBNAME: Mcgraw−crowsemmens PAGE: 14 SESS: 26 OUTPUT: Wed Aug 29 09:11:00 2007 SUM: 12C2769C
Magistrates’ questions
Use of methods by topic

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 23 OUTPUT: Wed Aug 29 09:18:57 2007 SUM: 44699BB1
11 Researching a Community
Safety Programme
In this chapter we describe an evaluation of a programme, the aim

of which was to develop safer neighbourhoods. The specific aim was
to reduce the likelihood of certain kinds of problem behaviour
developing among young people, and to promote their personal and
social development. We have referred elsewhere to the fact that
evaluation is one of the most common types of criminological
research, but evaluations can take various forms. The one described
here is very different to the evaluation of the Youth Court
Demonstration Project described in the previous chapter.
We will start by explaining the programme itself. Communities
That Care (CTC) involved local people and local agencies working
together to apply their knowledge about what factors in a neigh-
bourhood are most likely to put young people at risk of developing
social problems such as delinquency, drug misuse, school failure, and
school-age pregnancy. The factors can include such things as a
disadvantaged neighbourhood, high residential turnover, poorly per-
forming schools, and poor parenting. They try to reduce those ‘risk
factors’, and promote the kind of ‘protective factors’ that will
encourage achievement and social commitment among young peo-
ple, such as opportunities for pro-social involvement such as avail-
ability of youth provision. Even before saying any more you will
realize that this is not something that can be achieved overnight, or
even over a few months. Determining whether a neighbourhood has
been transformed, and whether the children who grow up in this
changed environment are indeed different to those who grew up
there previously (or have grown up elsewhere), is going to require
years. So it is a long-term programme, and this makes it particularly
difficult to evaluate.

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 23 OUTPUT: Wed Aug 29 09:18:57 2007 SUM: 514C231A
Communities That Care originated in the United States. In the

mid-1990s, the Joseph Rowntree Foundation (JRF) decided to
investigate the prospects for introducing a similar scheme into the
United Kingdom, and funded a British-based version of the pro-
gramme, to be run independently by an organization called CTC
UK. Three pilot projects were established in the United Kingdom,
and in 1998 three of us were commissioned to undertake an
independent evaluation. A series of reports and articles have been
published reporting the results of this study (France and Crow 2001;
Crow et al. 2004; France and Crow 2005; Crow et al. 2006). As is
the case with the other examples included in this book, this research
involved a team of colleagues, whose work is acknowledged here.
Many initiatives that attempt to reduce crime, or change the
behaviour of offenders, have a specific programme or series of
activities set out in advance. For example, a programme for the
treatment of offenders might be based on cognitive behavioural
therapy, or a crime reduction programme might be based on physical
changes, such as better street lighting, or better locks and entry
systems for blocks of flats. CTC, on the other hand, sets out a
process to be followed if the intervention is to be implemented as
intended. The first stage involves a central agency (such as CTC
UK) assessing and selecting areas that have the kind of conditions
that are favourable to receiving a CTC programme (known as
‘community readiness’). A community’s state of readiness is liable to
vary, and affects the chances of success, which means that CTC may
not work equally well everywhere. The second stage involves local
people and agencies, working through a programme board, in
carrying out two audits. One assesses the kind of factors locally that
are likely to place young people at risk of growing up to develop
problem behaviours, and the other assesses what resources are
already in the community that can be used to reduce such risks, and
to build protective factors. In the third phase, an action plan is
developed. In doing this the local board draws on a guide to
programmes that have been shown to be effective in reducing risk,
called Promising Approaches (Utting 1999) to identify programmes
that might tackle risk problems in their area. The fourth stage,
implementing the action plan, includes changing working practices
and developing new initiatives where the audits suggest they are
needed. The final phase should involve a review and re-assessment
of risk and protection, so that local communities can evaluate the
impact of their work and set new targets.
Theoretical context
CTC does not set out to test a particular theory of juvenile

development, but it does have certain conceptual foundations. One

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 23 OUTPUT: Wed Aug 29 09:18:57 2007 SUM: 4B213DEA
Researching a Community Safety Programme 227
of these is the proposition referred to above that certain risk factors

within communities are associated with particular types of future
problem behaviour. This proposition is supported by empirical
evidence which has shown that certain risk factors within commu-
nities are associated with particular types of future problem behav-
iour (Hawkins et al. 1992; Farrington 1996). Risk factors include
such things as lack of discipline in families, poor supervision by
parents, under-achievement in primary school, lack of neighbour-
hood attachment, and having friends involved in problem behaviours
(Farrington 2000). A risk factor approach to social issues has
become widespread in recent years, but when CTC was being
developed in the United States, and subsequently in the UK in the
1990s, the risk factor model was still a relatively new development
in the social sciences. One of the main proponents of the model was
also instrumental in introducing CTC to the UK (Farrington 1996,
2000), so CTC was an early test of the viability of such an
approach. A simplified version of the model can be presented as
shown in Figure 11.1.
However, it is worth noting that this model is based on correlations,
rather than explanations; a theoretical framework is nonetheless
required to understand the correlations. It could be argued that in
fact CTC involves several theories, for example about the conse-
quences of poor parental supervision and discipline, about disadvan-
tage, about school disorganization, about alienation, and so on. In
effect, every risk factor reflects some kind of theoretical position.
Consequently, the proposition stated above of a causal relationship
between risk factors and causal behaviours begs the question, where
does the risk itself come from?
This takes us to the other conceptual basis for CTC, which is the
social development model, put forward by Catalano and Hawkins
(1996). They argue that child development is influenced by the
quality of the interaction between children and adults, and that for
children to grow up free from social problems, they need to be given
clear standards of behaviour and have positive social bonding with
adults. To aid this process, children and young people need to be
given opportunities to be valued by, and involved in their families,
schools and communities, to gain social and learning skills, and be
given recognition and praise, ensuring that their positive behaviour is
recognized. The social development model therefore sees the devel-
Figure 11.1 The CTC model

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 23 OUTPUT: Wed Aug 29 09:18:57 2007 SUM: 5041BC34
opment of pro-social factors as a means of protecting children from

the consequences of risk factors. The CTC approach therefore
advocates the reduction of risk factors and the promotion of
pro-social factors that will help children to manage their circum-
stances better (Pollard et al. 1999).
We are including the research on CTC in this book as an
example of an evaluation of intervention in a community, as opposed
to interventions involving individuals. As such, the CTC evaluation
not only constitutes a test of the social development model, but also
has international significance because CTC has now been widely
adopted in several countries, having expanded from its original base
in the USA to countries in Europe, and to Australia. However, the
evaluation of CTC has a wider criminological significance because
of the role it played in a celebrated dispute concerning the nature of
evaluation during the 1990s. It was referred to by those involved as
the ‘paradigm wars’, described in more detail in Chapter 4. The
evaluation of CTC featured prominently in that debate, even though
it had not yet commenced at the time that several of the articles
were written (Pawson and Tilley 1994; Bennett 1996; Pawson and
Tilley 1996; Farrington 1997; Pawson and Tilley 1998a; Pawson and
Tilley 1998b; Farrington 1998). An exchange of articles in the
British Journal of Criminology, and the journal Evaluation ended with
two of the protagonists, Pawson and Tilley, saying ‘The UK
evaluation of CTC will be under way by the time this exchange is
published. Let us see what happens!’ (Pawson and Tilley 1998b).
The research
Evaluating a programme such as CTC presents a variety of

problems, which we will touch on as we describe what happened.
The first thing to do, as described in an earlier chapter, was to
define the research question. In the case of CTC there were actually
three key questions. First, was CTC implemented successfully? As
mentioned above, CTC involves carrying out a series of tasks.
Clearly it was important to establish whether the programme had
been implemented in the way intended. If it had not, then it would
be unreasonable to attribute any failure to achieve the intended
outcomes to a failure of the CTC model. The second question was,
did any change occur? Since the evaluation only had a three-year life
span, this involved establishing to what extent, if any, medium-term
measures of risk and protection changed in the three areas
(explained further below). The third question was, if there was any
change, what might have caused it? To what extent could any
changes be attributed to the intervention of CTC rather than to
other factors?

Answering these questions involved two kinds of evaluation

research: process evaluation and outcome evaluation. It therefore
employed the kind of integrated model of evaluation referred to in
Chapter 4 (Friendship et al. 2004: 14–15). The process research was
important in examining the first of the research questions, and
mainly involved qualitative research, including observations and
attendance at meetings, analysing documentary sources, and inter-
views with the key personnel concerned.
The outcome research required to answer the second question was
more problematic. The eventual aim of CTC is to reduce four
problems among young people: failure at school, school-age preg-
nancy, drug abuse, and youth crime. Thus, the long-term success of
CTC is to be judged by reference to these four measures. However,
CTC is a programme which affects children as they grow up, and it
may therefore be ten years or more before its full impact can be
evaluated. Unfortunately researchers seldom have the opportunity to
be around so long. But since the CTC proposition is that these
behaviours can be influenced by addressing risk and protective
factors, in the medium term it is possible to look at the extent to
which there are changes in certain risk and protective factors
associated with these longer-term outcomes. The best way of doing
this was to see whether the perspectives of young people themselves
were changing. In the CTC model a school survey is used as an
indicator of community levels of risk and protection. Hence the
main tool of analysis was to use surveys of young people going to
schools in the areas concerned before the programme was imple-
mented, and again some time later: a before and after research
design.
The third research question involved drawing on a combination of
both process and outcome research to make inferences about
whether any observed changes could be attributed to the interven-
tion, and this is where the research design is crucial. In Chapter 4
we discussed the nature of criminological evaluation and the tech-
niques used. When applied to looking at communities rather than
individuals the principles are the same – it is important to compare
like with like – but in practical terms this is often more problematic.
Although an attempt was made to compare the neighbourhoods
where CTC was taking place with similar communities nearby, this
did not work out, since the three intended comparison communities
turned out to have important differences. Consequently, what we did
was to compare young people going to local schools who lived in the
areas where the CTC programme was taking place, with young
people from the same schools who did not live in those areas. This
also had the advantage of controlling for the schools variable – i.e.
the possibility that schools might themselves make a difference. We

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 23 OUTPUT: Wed Aug 29 09:18:57 2007 SUM: 40802B7F
referred to the three areas involved in the pilot study as Northside,

Westside and Southside, so our research design took the form as
shown in Table 11.1.
We now need to say more about the main method for
measuring outcome: the school surveys. In CTC, risk factors are
organized under four domains relating to the local community,
school, the family, and individuals and their friends. The community
risk factors include measures of neighbourhood disadvantage, com-
munity neglect, and a lack of neighbourhood attachment. School
factors are measured by such things as low school achievement,
bullying, truancy, and poor school organization. Family factors
include poor parental supervision, family conflict, and parents
condoning problem behaviour. Factors related to individual children
and their friends are alienation and a lack of social commitment,
early involvement in problem behaviour, and having friends who are
involved in problem behaviour. The protective factors cover rewards
and opportunities for pro-social involvement, such as being valued
by parents and teachers and being involved in community activities.
Consequently, the questionnaires included questions about pupils’
personal and social circumstances, their families, neighbourhoods,
and school experiences, the availability and use of alcohol, tobacco
and other drugs, delinquent and anti-social behaviour, and spare
time activities. Identical questionnaires were used for both before
and after surveys. Altogether, there were 195 separate items of
information. Apart from collecting demographic information, much
of the questionnaire was devoted to asking about young people’s
experiences of, and views about these topics. Their views were
mostly rated on a four point scale from ‘Strongly Agree’ to ‘Strongly
Disagree’. Their responses to the various questions were put together
to compose 16 risk factors and seven protective factors. Box 11.1
gives an illustration of five questionnaire items which together
constituted a scale showing pupils’ overall assessment of a risk factor
measuring what kind of neighbourhood they lived in.
Table 11.1 CTC comparative design

JOBNAME: Mcgraw−crowsemmens PAGE: 7 SESS: 23 OUTPUT: Wed Aug 29 09:18:57 2007 SUM: 34A47BFB
The questionnaire was administered immediately prior to the imple-

mentation of the CTC Action Plans in late 1999 and early 2000,
and just over two and a half years later in July 2002. Nearly 11,000
children answered the two school surveys (5516 the first and 5334
the second). The surveys were completed at school, in confidence.
Response rates varied somewhat between the areas and from school
to school, but in general were very good, and show the advantages
of being able to get questionnaires completed in such controlled
circumstances. In Northside the response rates were 95 per cent for
the first survey and 96 per cent for the second. In Westside they
were 82 per cent for the first survey and 84 per cent for the second,
and in Southside they were 79 per cent and 68 per cent.
The results were analysed by comparing the proportion of
children in each area scoring positive for risk and protective factors
between the first and second surveys, as a basis for determining
whether any changes had occurred. The more items a child scored
for a risk factor, the greater the risk for the child, and similarly for
protective factors. It should be noted that this measurement of
change was done at an aggregate level. In other words, we were
looking at changes within groups of children from CTC and
non-CTC areas, rather than looking at the extent to which CTC
might have had an impact on individual children.
Results
In the other examples we have used we have reported the results as

being incidentally of interest; our main focus has been on the study’s

JOBNAME: Mcgraw−crowsemmens PAGE: 8 SESS: 23 OUTPUT: Wed Aug 29 09:18:57 2007 SUM: 4BC45F39
methodology. As elsewhere we will report the results briefly, but in

the case of this study the results do illustrate some important points
about the methodology used.
It will be recalled that the first research question we addressed
was whether the intervention (CTC) had been implemented as
intended. In fact, only one of the three areas implemented CTC in
such a way as to constitute a valid test of the model. In the area we
referred to as Northside the project ran into problems about a year
after starting, partly as a result of losing the Co-ordinator who was
not replaced for some time, which resulted in little activity, and no
meaningful interventions. In Westside there was more activity
resulting in several initiatives. But these were undertaken alongside
other developments taking place in the area, and were not of a kind
that were in accordance with the CTC model. This highlights the
importance of undertaking a process evaluation in order to see
whether an initiative is doing what it set out to do.
Only in Southside was there implementation in accordance with
the CTC model. Here 14 out of 21 risk and protective factors
examined for this project showed a positive change for children who
lived in the CTC area. Only one of these changes was statistically
significant, so it is possible that the changes are due to chance, but
the overall trend was positive. The analysis of the surveys in all areas
showed an overall trend towards increased risk overall, and the
indications were that CTC in Southside may have had an inhibiting
effect in keeping risk lower than it might otherwise have been.
Thus, looking at the picture with regard to medium-term outcomes,
the indications were that CTC was of some benefit to Southside,
but the results were mixed rather than strikingly advantageous.
However, this does not tell the whole story.
It was also relevant to look at the results in relation to the
activities taking place in the area. The CTC model requires
communities to identify, on the basis of risk and resource audits, the
risk factors that it intends to target, and the actions it plans to take.
In Southside, most of the interventions implemented were targeted
at family factors, and this was where the most encouraging results
were found. There was also evidence regarding the reduced availabil-
ity of drugs that could have been attributed to CTC-initiated
activity. On the other hand, intended action relating to school
factors was not implemented for various reasons, and the results for
the school factors were not encouraging. This holds out the
possibility that had more been done in relation to school factors
targeted at CTC children, the results for Southside could have been
better. In essence, our research showed that, while CTC may indeed
be a line of development worth pursuing, it was as yet unproven.

JOBNAME: Mcgraw−crowsemmens PAGE: 9 SESS: 23 OUTPUT: Wed Aug 29 09:18:57 2007 SUM: 55C492EB
Conclusion
To summarize: a long-term programme, intended to change com-

munities in order to reduce the chances of children who grow up
there developing certain social problems was imported from the
United States, and tried out in three experimental areas. The
initiative was evaluated over a three-year period by studying the
process of implementation, and by using a before-and-after survey of
children attending secondary schools in the areas concerned to
measure the medium term outcome. Their responses were scored
according to a series of risk and protective factors which have been
found to be related to the probability of children developing social
problems, in order to see whether the scores of the children who
lived in the CTC areas had improved between the first and second
surveys, compared with the children who did not live in those areas.
The results showed that the programme was not implemented as
intended in two of the areas, and that in the third, while there were
promising indications, it was not possible to conclude that the
programme had so far demonstrated significant success.
Several issues are raised by this study, some regarding the project
itself, others relating to the nature of the research, and inevitably
they overlap. To start with, this is a complex programme requiring a
number of events to happen over quite a long period. It is more
difficult to measure change affecting whole communities over a
period of time than it is to evaluate interventions directed at
individuals (not that that is easy). Then there is the school survey
itself. As we have said, in the medium term, this survey of a large
number of schoolchildren is the nearest the developers of CTC have
been able to get to measuring what is happening in the neighbour-
hood that affects them. But like all surveys, we can never be certain
that it reflects exactly what is happening. There are, however, various
checks within the survey to test for reliability. For example, pupils
who claimed to have knowledge of a drug called Derbisol were
eliminated from the survey – because it doesn’t exist; it was put in to
see if children were exaggerating.
It is also important to point out that the evaluation described here
occurred during the formative stages of these three projects, and of
the use of CTC in the UK generally. There are different views
regarding when it is best to evaluate a new initiative. On the one
hand, it could be argued that programmes should not be evaluated
until they are well established, because only then can there be a valid
test of their operation. On the other hand it could also be argued
that it is much better to study programmes from the outset because
one can observe what is happening and why; if evaluation were
delayed much valuable information would be lost. Generally it is the
latter view that prevails, but concomitant with this must go the

JOBNAME: Mcgraw−crowsemmens PAGE: 10 SESS: 23 OUTPUT: Wed Aug 29 09:18:57 2007 SUM: 4DD17D12
recognition that the results may not truly reflect what a programme
is capable of. The fact is that, as noted in an earlier chapter,
programmes change throughout their lives, perhaps because people
leave and arrive, or circumstances change, so it is probably best to
regard any initiative as a dynamic entity, and any evaluation as an
analysis at a particular point in time. This is one of the reasons why
the evaluation of criminological and other social initiatives is so
different to, say, testing a new drug or certain types of medical
treatment: the ingredients are more variable.
Alongside this is the importance of recognizing that the very
process of being evaluated may have had some effect on the projects
concerned, what is referred to as the ‘Hawthorne effect’ (from an
early sociological study of an organization). While the researchers
did make it clear that we were independent of the projects and there
as observers only, we did give some feedback and produce an interim
report (France and Crow 2001), and this was bound to have some
effect. In real life, it is difficult to stand back and say nothing
whatsoever to people with whom one has close contact over a period
of time.
It can also be argued that there were too few instances of the
CTC programme in the evaluation for it to have been adequately
tested. Two other researchers with experience of community evalu-
ations point out that,
If the community is the unit of analysis, then the number of

communities will be our sample size … Using a unit of analysis
of this size might make it more difficult to reach a sample size
adequate for effective statistical inference. (Hollister and Hill
1999: 130)
This raises the question of how many projects are needed to

constitute an adequate test of the model, but one of the proponents
of CTC has argued that a large number of projects are necessary to
evaluate CTC effectively (Farrington 1997). An earlier community
safety programme in the UK, Safer Cities, based its analysis on 96
schemes, rather than the three available to us (Ekblom et al. 1996).
It will be recalled that in the model outlined in Chapter 1, the
final stage of the research process was defined as interpretation.
Regarding the results of this study there are three possible interpre-
tations:
1 Measurement failure. This means that the intervention may have

been more (or less) successful than described, but that flaws in
the research failed to identify this. As we have noted, like any
research, this study had its limitations, but we do not think that
it was so badly inaccurate that the conclusions are misleading.

JOBNAME: Mcgraw−crowsemmens PAGE: 11 SESS: 23 OUTPUT: Wed Aug 29 09:18:57 2007 SUM: 288DC55B
2 Implementation failure. This is liable to occur in any evaluation

study, and clearly did take place as far as two of the projects were
concerned.
3 Theory failure. This is the key issue. In terms of the model we
outlined in Chapter 1, criminological inquiry derives from the
examination of some theoretical proposition, and the main pur-
pose of empirical investigation, such as an evaluation study, is to
see what conclusions can be drawn about the theory underpinning
the inquiry. As we described earlier, CTC does have a particular
conceptual basis. However, what we have to infer from the
present study is that the results do not allow us to conclude
whether the theory underpinning CTC is a failure or not. This
happens in research, through no particular fault of the researchers.
What it means is that the case is ‘not proven’. Lessons can still be
learned, both about the interventions and about the way the
research was done, but the results of this study showed that more
needs to be done before it can be concluded whether CTC, and
thus the ideas underlying it, are a success or a failure. Although
this particular study has ended, other evaluations of CTC have
continued in the UK and elsewhere.

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 4EDAC499

In this chapter, we will describe a research project which focused on
one of the most heavily researched topics in criminology – the fear
of crime. This project is a particularly good example of how the
research process can benefit from a collaboration between researchers
who are working for different types of organization. Indeed, it is not
uncommon for criminological research to involve a collaboration
between an academic researcher or research team and a government
department, criminal justice agency or community partnership.
Sometimes, agencies will approach academics due to their expertise
in a particular field and ask for help with their research; at other
times an ‘invitation to tender’ will be advertised, inviting researchers
to bid for the project. Collaboration is an attractive prospect for the
academic researcher not only because it can bring funding (often the
hardest part of getting a project underway) but also because it may
alleviate some significant access problems by giving access to
otherwise closed information. This was certainly the case in this
project which was facilitated by the development of a relationship
between a research team at the Home Office and a university-based
researcher. There are, however, potential problems and limitations
associated with collaborations like this and we hope that some of
these will be illustrated in this chapter.
The research project was carried out during the period 1999–2001
and it is essential to begin by explaining the context in which it
took place. In the 1990s, fear of crime had become a primary
concern for policy-makers. In his extensive literature review, Hale
(1996) had identified more than 200 articles, conference papers,
monographs and books on the topic and that number was escalating
with great speed. The vast majority of studies of fear of crime used
a survey approach and attempted to measure levels of fear by asking
people how safe or worried they felt in their everyday lives. This
was, more often than not, conducted using structured questionnaires
and the main hypotheses tended to focus on establishing who is
fearful, what causes fear and how fear manifests itself (all of which
are questions which need to be answered if the problem of fear is to
be addressed effectively).

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 4F501E65
Researching the fear of crime 237
There was, however, surprisingly little consensus reached as to

what factors influenced and explained fear levels. Gender and age
emerged as the most likely explanatory variables, but even these
were found to be unreliable predictors in many studies. The reason
for this incongruity seemed to stem from inadequacies in the
conceptualization and operationalization of fear of crime measures,
many of which had existed for many years and had been perpetuated
by an almost ‘blind’ acceptance of them as tried and tested measures
by researchers. As a result, the questions which had originally been
designed for the US National Crime Victims Survey and the British
Crime Survey in the late 1970s/early 1980s were being used as
standard, despite a growing body of literature which drew attention
to their weaknesses. The study described in this chapter was
designed at a time when researchers working in the field of fear of
crime had recognized the urgent need to address some of the
methodological questions which had emerged during the mid-1990s
and sought to make improvements in the conceptualization and
measurement of fear of crime.
Theoretical context
The research began with the lengthy process of reviewing the

literature on fear of crime. The process of organizing more than 800
items of literature was complex. Once each item had been read, it
was photocopied and filed alphabetically in a large cabinet for future
reference. Short summaries of each item were written in a word
processing package and organized according to topic (for example
‘studies about fear and gender’ or ‘studies about fear and victimiza-
tion’). Each summary contained details of the topic studies, the
methods used and the conclusions drawn, together with a critical
statement as to the usefulness or reliability of the work. In addition
to the summaries catalogue, a separate chronological catalogue was
constructed to try and trace how the concept of fear had been
developed over time. This was important because in order to
understand the methodological problems that existed, it was neces-
sary to understand how they had evolved.
During this period of reviewing the literature, a whole range of
interesting possible research questions emerged and at times the
need to restrict my ideas to one broad question was quite over-
whelming. There seemed to be so many unanswered questions and
possible avenues for making a contribution to the field that it was
easy to lose perspective on the research process itself. I was, though,
able to eliminate many of the questions through discussions with
other researchers and colleagues, many of whom were more experi-
enced than I was. Eventually, one question did emerge which

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 549B8509
seemed to prompt a curiosity in the reactions of colleagues which

was not present in previous discussions. This, for me, was the sign
that the question was probably well conceived and worth pursuing!
The question was: how worried are people about financial crimes?
The question itself had stemmed from my genuine surprise that
very few studies on fear of crime had included financial crimes in
their conceptual framework. I had read much about the perceived
impact of being burgled, mugged or having a car stolen but never
about the effect of losing money as a result of fraud. Repeatedly I
told myself ‘surely someone has researched this’ but there was very
little evidence that this was the case. The main reason for this
seemed to be that fraud was not seen as something that had a serious
or widespread impact on the lives of individual citizens and thus was
not a pressing political concern. In many ways, it was overshadowed
by the more ‘newsworthy’ and immediate crime problems that
policy-makers were usually concerned with. This was particularly
striking because, at the time, the financial industry had actually been
reporting record figures in levels of fraud on credit and debit cards
(plastic card fraud). Indeed, media reports were alerting the public to
the potential impacts of these crimes, describing how they can affect
the credit ratings and financial autonomy of victims. Moreover,
concerns were being expressed in the commercial world that ‘fear of
fraud’ was deterring consumers from engaging in e-commerce.
However, a lot of this discussion was informed by market research
studies and, although useful, they lack the theoretical rigour that
criminological research has. It seemed that the time was right to
start exploring the possibility of studying the fear and victimization
of plastic card fraud more formally, from a well-developed, crimino-
logical perspective.
The next stage of the research process, then, was to build a
theoretical framework which could be used to identify the key
concepts and develop the hypotheses. There were essentially two key
areas in need of theoretical development. First, it was necessary to
take the broad concept of ‘fear of crime’ and establish a theoretical
framework which utilized ‘crime-specific fears’. It was important to
know whether people had different levels of fear for different crimes
in order to construct some hypotheses as to levels of fear of plastic
card fraud. The second area of theoretical development focused on
plastic card fraud – how do people use cards? how do they manage
their accounts? where and when do they use their cards? and have
they any experience of theft or fraud?
Fear of different types of crime

Perhaps the first question to address here was what is fear of crime?
It was perhaps a little surprising to see that a concrete definition was

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 594DA818
far from forthcoming in the literature. Indeed, as mentioned at the

start of this chapter, it is this weakness in conceptualization which
has been cited as the primary cause for contradictory and inconclu-
sive research in the field. The first stage here, then, was to unpack
the concept of fear of crime and, through doing so, attempt to
re-build a theoretical framework. This involved breaking the concept
down into two constituent parts: fear and crime.
Whilst it is widely acknowledged that fear is a term used to
describe a whole range of perceptions and emotions from general
anxiety through to a state of terror, very little attention has been
given to the systematic conceptualization of fear itself. In studies of
fear of crime, fear has been loosely conceived in four ways and,
although far from comprehensive in conceptual terms, it is helpful to
explore the different perspectives that have developed over time.
First, as an emotional response to the perceived threat of victimiza-
tion. This approach might sound as though it has a solid psycho-
logical basis but, in reality, researchers have tended to focus only on
asking individuals about ‘feeling afraid’ or, more commonly, ‘worry-
ing about crime’ in their everyday lives. Only recently has the need
to explore the complex psychological and emotional dimensions of
fear been recognized. Second, fear is often conceived as a general
perception of safety (often referred to as a ‘global fear’). Typically,
crime surveys include a question asking the respondent how safe
s/he feels walking in the local area during the day and at night.
However, this kind of measure has been recognized as unreliable
because of its over-general nature and the need for respondents to
answer hypothetically. The third way of conceiving fear is as a
behavioural concept. This approach attempts to measure people’s fear
by their actions, that is, do they take security precautions or, for
example, avoid certain areas at certain times in order to reduce their
risk or fear? This is an interesting approach but one which is in its
infancy and in need of significant development. Finally, fear may be
conceived as a judgement of risk, that is asking people to judge their
likelihood of becoming a victim of crime in the future. The now
infamous ‘fear-risk paradox’ (the fact that those least at risk of
victimization seem to be the most fearful and vice versa) has been
the focus of a considerable amount of research and discussion, often
resulting in the conclusion that fear is somehow an irrational
response to the reality of crime.
The next stage was to address the crime dimension – what is it
that people are afraid of? Although fear of crime surveys do tend to
include measures of crime specific fears (usually burglary, vehicle
crime, mugging, personal attack and rape/sexual attack), it was noted
that most articles and research reports conclude with statements
summarizing the reactions to crime generally with very few recog-
nizing the implications of crime specific analysis. As Rountree states,

JOBNAME: Mcgraw−crowsemmens PAGE: 5 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 511A8B29
‘[s]tudies examining the crime-fear linkage tend to estimate the

effects of any type of victimisation or a total crime rate on fear.
Such an approach implies that there are no important differences in
the effects of various crime or victimisation types on fear’ (1998:
342). The literature review did, however, reveal some studies in
which some general distinctions were made between fear of property
and personal crimes. It became clear that property crime does evoke
reactions different to personal crime (LaGrange and Ferraro 1989)
but that the reasons underlying these differences are complex and
yet to be explained. At best, we can hypothesize that an individual’s
attitude towards property crime and personal crime may be influ-
enced by a range of perceptions including the perceived seriousness
of the crime, the perceived likelihood of that crime occurring, and
the perception of self-vulnerability (Skogan 1987; Rountree 1998).
However, this did not bring me particularly close to how fears of
financial crimes might manifest.
This process of deconstructing the concept of fear of crime was
useful, despite the fact that it raised more questions than answers! It
was important to understand that although not all of the conceptual
problems could be solved in one study, an awareness of the potential
threats to validity and reliability in the research was essential. This
part of the theoretical framework had been constructed and the next
stage of the theoretical development was to explore the concept of
plastic card fraud and make the essential theoretical links between
the two frameworks.
Plastic card use and misuse

There are characteristics or behaviours which can increase (or
decrease) an individual’s risk of becoming a victim of crime and
these risk factors have also, to varying degrees, been used to explain
different levels of fear of crime. These factors might include age,
gender, socio-economic status, health and fitness levels, place of
residence and lifestyles (whether an individual socializes, travels,
works, etc.). One of the central aims of crime surveys is to measure
the effect of these different factors and attempt to identify vulner-
able groups in society. It was necessary, then, to find out as much as
possible about the factors which can be associated with risk of card
fraud victimization and the fear of victimization, to build as
complete a picture as possible and inform the process of operation-
alization.
The starting point was to establish the different ways in which
card fraud can be committed in order to build an accurate concept
of the crime itself. The information produced by the Association of
Payment and Clearing Services (APACS) was helpful at this stage of
the research because the rates of card fraud are publicly available and

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 58CAC41A
routinely presented according to the source of the fraud. For

example, it is possible to compare the rates of fraud on different
types of card, the locations of frauds (in the UK or abroad) and the
method used by fraudsters. It was this latter category which was of
particular interest. Fraudsters use a variety of techniques to commit
their crimes and it is important to conceive their ‘target’ much more
broadly than just the plastic card itself. While a considerable
proportion of fraud is committed using physical cards (either genuine
cards which have been lost or stolen or counterfeit cards), fraudsters
can also utilise the card information (name, number, expiry date, etc.)
without being in possession of the card itself. Moreover, a significant
proportion of frauds are carried out during the application process, so
a fraudster may steal an individual’s personal details and apply for
cards or other forms of credit.
The next stage of building the theoretical framework was to
understand how and why cards are used by individuals. This was
important because it would provide the context in which the fear
and victimization of card fraud manifest. The first thing to become
clear was that it is important to distinguish between different types
of plastic payment cards. Credit cards, debit cards, store cards and
cashpoint cards are the most common card types and all have
different uses and features. Credit cards and store cards provide a
credit facility, allowing the consumer to borrow money in the short
or longer term; in contrast, debit cards and cashpoint cards are used
when the consumer wishes the money to be taken straight out of
his/her bank account. Because each card type is different in nature,
it is important to distinguish between them at a conceptual level.
The second point for consideration was where people use their
different cards. This allowed me to conclude whether people are
most at risk of fraud when they are using cards in shops, restaurants,
on the internet, to pay bills, over the telephone or at cashpoints.
Similarly, it was important to know how often cards are used to
establish whether the risk of victimization increases with card use.
All of this was important because it helped to build a picture of the
potential impact on a person’s life and well-being that fraud
victimization might have. So, for example, if a person becomes a
victim of credit card fraud, it may be the case that his/her credit
rating is affected. Where a debit card is targeted, the victim could
lose money direct from the bank account and, perhaps, incur bank
charges. In both situations, it is likely to take a lot of time and
effort on the part of the victim to sort out the problem.
Another important element of the ‘card usage’ framework was the
fact that individuals seem to manage their finances in different ways.
Some people, for example, may simply keep their money in a single
account and some have complicated financial interests distributed
across many different financial institutions. Where some people have

JOBNAME: Mcgraw−crowsemmens PAGE: 7 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 53FC2CC2
the ability to embark upon complex and profitable accounting

enterprises, some people have financial problems which restrict their
financial autonomy. Some people have special insurance to protect
their financial interests and some protect themselves from fraud with
preventative measures such as shredding receipts and only using
secure internet sites. The point is that individuals operate within
different financial contexts and these contexts may influence the risk
of victimization and, indeed, fear. There was very little literature on
these aspects, however, and so it was recognized that exploratory
work was needed in order to build this part of the theoretical
framework.
Theorizing the fear of financial crime

The final step in the theoretical development was to bring the two
frameworks (fear of crime and plastic card fraud) together. Due to
the limited amount of research carried out from the crime-specific
perspective in the field of fear of crime, it was hard to imagine
exactly what types of reaction financial crimes do prompt. These
could be direct emotional or behavioural responses or they could
contribute more indirectly to a general feeling of safety or unease.
So, I decided to explore an additional literature base, the literature
on white-collar crime, in order to seek guidance on any known
reactions and attitudes to financial crimes. A number of useful
sources were identified that confirm the serious impact that white-
collar crimes can have. Pearce and Tombs (1992) note that when
surveys do include questions on white-collar crime, often it has been
rated as being more serious than many other types of crime. Ganzini
et al. (1990) suggest that the victims of white-collar crime can be
compared with victims of violent crime in terms of their responses
and reactions to their victimization. Like the victims of violent
crime, white-collar victims tend to be older, more affluent and are
relatively more likely to be female. In terms of psychiatric outcome,
general anxiety disorder and major depressive disorder are the most
common psychiatric complications of both types of victimization.
Also, for both types of victimization, a previous history of psychi-
atric illness and the degree of victimization are important variables
in predicting the risk of psychiatric problems. Levi and Pithouse
concur, arguing that, ‘[f ]or these private victims, the victimisation
experience could be likened to a (comparatively mild) sort of rape’
(1992: 233).
Although it must be borne in mind that these studies were about
white-collar crime generally (thus including crimes like embezzle-
ment, tax evasion, consumer fraud and corruption in their remit), a
lot of useful ideas could be drawn from them. There was, in
addition, a small amount of literature which focused specifically on

JOBNAME: Mcgraw−crowsemmens PAGE: 8 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 52A801CE
the impact of financial frauds and frauds which involved the misuse
of personal information (of which credit/debit card details are an
important element). First, in a study of pensioners who were victims
of the Maxwell affair,1 Spalek (1995) found that their experiences
did indeed have a profound effect on their lives. In particular, their
mistrust of financial institutions and advisers left them in serious
doubt of their financial stability. Second, Grabosky et al. (2001)
argue that to lose control of one’s personal information is potentially
damaging to quality of life. They allude to a deep psychological
harm, ‘the loss of one’s private life is often accompanied by a decline
in spontaneity, creativity, and a diminished sense of self ’ (ibid.: 176).
Both studies were useful in building a picture of the potential
impact of plastic card fraud on the victims and, indeed, the impact
of the threat of victimization more generally. Finally, the theoretical
framework was strong enough to move on to the process of
conceptualization and hypotheses building.
The research
So, having explored the topic and identified an interesting gap in

the field, the research question was broadly conceived to be ‘to what
extent are people victims of and fearful of plastic card fraud?’. The
theoretical framework had been developed and the next stage was to
draw up some hypotheses. However, it was appropriate now to think
about the most appropriate research strategy to adopt. There were,
actually, several possible research strategies which might have been
suited to the aims of the research. As it was a completely new area
of research, it might have been appropriate to adopt an interpretiv-
istic approach, using open interviews or focus groups perhaps to
properly explore the experiences and attitudes of different groups of
people. It would, for example, be interesting to explore differences in
attitudes between those who had and had not been victimized and
this kind of data would be particularly useful both conceptually and
theoretically. However, a positivistic approach was another possibil-
ity, allowing for the testing of some simple hypotheses using a
survey approach. While there would be a reduced scope for
theoretical development using this approach, its value lay in the
ability to draw data from a larger sample and to make comparisons
with data about fear and victimization of other crimes and other
studies.
While either approach would have had been useful, there was one
factor which ultimately had to be taken into consideration. Because
this research was being carried out as part of a PhD thesis, there
was no funding available for the empirical work and the scope of the
study was limited to that which one person, in a short period of

JOBNAME: Mcgraw−crowsemmens PAGE: 9 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 4E7634BF
time could achieve. Fortuitously, at this early stage in the project, a

rare opportunity arose to form a working relationship with the
British Crime Survey team and to contribute some new questions to
the British Crime Survey 2000. This was a significant and exciting
development because it allowed for new questions to be developed
and tested on a well-established, national survey. However, choosing
this approach would mean that that there would inevitably be some
restrictions placed on the operationalization process and that the
analysis process would bring the many challenges associated with
analysing somebody else’s data (see Chapter 8). These issues will be
discussed in some detail, following a brief description of the research
itself.
Aims
One of the first things which needed to be established was the set
of aims and objectives which would serve the interests of both
parties to the research. This was achieved through a series of
informal meetings in which possible areas of research were discussed
and explored. It was important that everyone involved in these
meetings was honest about their own interests and motives and that
the boundaries of the research were clearly defined. It was also
necessary to negotiate the legal and ethical issues at this stage. So,
for example, it was agreed that the data would be anonymized
before being handed over for analysis. Also, I agreed to give the
Home Office team access to any work before being published. We
also had the opportunity to discuss the potential difficulties I might
encounter when trying to get to grips with the complex dataset and
we were able to arrange training sessions to help alleviate any
problems.
Since this was the first time that plastic card fraud would be
included in a crime survey, we finally reached agreement that the
aim of the research should be to test the feasibility of introducing
plastic card fraud questions to the BCS. It was, after all, a distinct
possibility that this relatively ‘new’ and potentially rare crime was
not suited to the survey approach. This meant that a lot of the more
complex theoretical and conceptual questions raised earlier in the
research process would remain unanswered by the research but
ultimately this was a sensible and more realistic decision. The
objectives of the research were, then, as follows:
1 To design and operationalize questions about plastic card fraud.
2 To compare plastic card fraud with the other BCS crimes (in
terms of fear and victimization levels).
It was hoped that by gathering this information about plastic card
fraud it would be possible to draw some useful conclusions about the

potential contribution which could be made to, (1) our understand-

ing of fear of crime through the introduction of a financial crime to
the concept; and/or (2) our understanding of plastic card fraud by
attempting to assess the extent of victimization and levels of fear
that exist.
Identifying key concepts and building hypotheses

Having developed the two theoretical frameworks, the process of
conceptualization and hypotheses building could begin. However, at
this point a significant design decision was taken by the British
Crime Survey team that restricted the scope of the research. It was
decided that due to space restrictions on the questionnaire, it was
necessary to confine the plastic card fraud section to two questions
only. The challenge, then, was to construct some effective hypotheses
which could be tested through two questions on the survey (plus, of
course, the data which would already be drawn from the existing
survey questions).
Two concepts were selected as the key concepts; ‘plastic card fraud
victimization’ and ‘worry about plastic card fraud’. ‘Worry’ was
adopted as the measure of fear here because it was identified as the
most reliable of the existing BCS fear measures (see below). These
were the only concepts needed to construct the primary hypotheses
which both focused very simply on the success of operationalization:
1 Card fraud victims are identifiable.
2 Worry about card fraud is measurable.
There were, also, some secondary hypotheses which were developed
in order to explore the nature of card fraud in some more depth.
There were two lines of development here. First, to make compari-
sons between the fear and victimization of card fraud with other
crimes. The focus here was on establishing whether card fraud was
similar to property crimes in terms of victimization and fear levels,
or whether it was more closely aligned with the personal crimes. For
the purposes of this analysis, burglary and vehicle theft were taken
to be ‘property crimes’ and mugging, violent attack and sexual attack
were taken to be ‘personal crimes’. Second, to test whether there is
a clear relationship between the fear and victimization of card fraud.
Accordingly, the following hypotheses were developed:
1 The incidence of card fraud is more similar to the incidence of
property crimes than to personal crimes.
2 The levels of worry about card fraud are more similar to the
levels of worry about property crimes than to levels of worry
about personal crimes.

3 Victims of card fraud are more fearful of card fraud than are
non-victims.
Methodology
In terms of methodology, there was little room for external input to
elements of the design such as sampling and interview style as the
BCS operates according to a well-established and rigid research
design (this is explained in depth by Kershaw et al. 2000). In a way,
this was a relief inasmuch as the burden of making complex
methodological decisions was removed. However, it was extremely
important that I developed a good understanding of how the data
was collected in order to be able to design effective questions and
conduct the analysis.
There was a great deal of work to be done in terms of question
design and operationalization. The earlier literature review had not
revealed any examples of survey questions used previously to
measure victimization or fear of plastic card fraud. The International
Crime Victim Survey included a question on consumer fraud (‘has
someone, when selling something to you or delivering a service cheated you
in terms of quantity or quality of the goods/services?’) but nothing on
plastic card fraud specifically. Thus, the key concepts needed to be
operationalized from scratch.
The first stage of question design produced the following two
questions:
1. People can steal money from other people’s debit and credit
cards, either by overcharging them or by copying down their
card details/PIN and using them to buy things or withdraw
cash. In the last year, has someone stolen money from you in
any of these ways?
a) Yes – overcharged
b) Yes – card details/PIN used
c) No
d) Don’t use cards
2. How much do you worry about someone overcharging you
on your credit/debit card or using your card details/PIN to
buy things or withdraw cash?
a) Very worried
b) Fairly worried
c) Not very worried
d) Not at all worried
e) Insured against losses
However, after initial pilots (informal pilots using a small random
sample of colleagues and students), these questions were rejected for

JOBNAME: Mcgraw−crowsemmens PAGE: 12 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 4BC2E728
two reasons. First, in an attempt to incorporate as many of the

conceptual considerations as possible into two questions, the ques-
tions themselves had become too complicated and in danger of
being misinterpreted. Second, it was decided that, in order to make
the results directly comparable with the other BCS crimes, the
wording needed to match (as far as possible) more closely the
wording of the existing BCS questions.
Therefore, the following two questions were redesigned as;
1. As far as you know, including anything we have already
talked about, since the first of January 1999 has anyone used
your credit card or bank card, or your card details, such as
your PIN, to buy things or withdraw cash without your
permission?
a) Yes
b) No
2. How worried are you about someone using your credit card
or bank card details, such as your PIN, to buy things or
withdraw cash without your permission?
a) Very worried
b) Fairly worried
c) Not very worried
d) Not at all worried
e) (N/A – don’t use cards)
Both questions were piloted as part of the formal BCS piloting
exercise which occurs before the sweep is conducted (Hales et al.
2000). The questions piloted well and no further changes were
necessary. The narrow scope of the questions reduced the range of
final conclusions which could be drawn but they did allow the
hypotheses to be tested. This ultimately reflected the overarching
aim which was, if we recall, to test the feasibility of introducing
plastic card fraud to the survey.
Results
The dataset was provided in SPSS format and had already been
‘cleaned’ by the data collection company to remove discrepancies and
any information which may have led to the identification of
respondents. This was helpful because it meant that the data were
ready for immediate analysis. On opening the dataset, the first thing
to strike me was the sheer size of the BCS dataset. It was larger
than any dataset I had ever encountered, containing data from more
than 20,000 respondents. The challenge, then, was to learn how the
dataset was constructed and how it should be handled. An essential

JOBNAME: Mcgraw−crowsemmens PAGE: 13 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 3990413E
tool here was the ‘Technical Report’, a large document which

contained all the information about the survey methodology and the
questionnaire. Without this, the analysis of the data would have
been almost impossible, for the reasons we outlined earlier in the
book in Chapter 8 (analysing official data).
The analysis was conducted in a systematic way, focusing on the
hypotheses which had been set earlier in the research process. The
results will be presented briefly below, dealing first with the primary
hypotheses and then the secondary hypotheses. The implications of
these results will be discussed in the section that follows.
Primary hypotheses
1 Card fraud victims are identifiable.

Table 12.1 shows that 2 per cent of respondents had been a victim
of card fraud in the last year (12 months from 1/1/99), on a par
with victimization levels for theft of a vehicle and mugging/robbery
for the same period. Victims of card fraud, it seems, are aware of
their victimization and do report the event to the interviewer.
2 Worry about card fraud is measurable.

Here we were concerned with the question of whether card fraud is
something people actually worry about, or would inclusion in the
survey be fruitless? Table 12.2 shows the distribution of respondents
across worry levels for card fraud. Just under half of respondents
(49 per cent) were worried about card fraud to some extent (very
worried or fairly worried). Indeed, it was encouraging, and some-
what unexpected, to find that ‘very worried’ levels (used by the
Table 12.1 Percentage of victims for each crime

Home Office as an indicator of fear, see Kershaw et al. 2000) for

card fraud were comparatively high (18 per cent of respondents were
very worried about card fraud). However, there were also a lot of
people who said that they were not at all worried about card fraud
and this was particularly interesting because the same observation
can be made for the personal crimes (when compared with the
property crimes).
Secondary hypotheses
If we recall, there were some secondary hypotheses that were
designed to make comparisons between the fear and victimization of
card fraud with other crimes in an attempt to explore the nature of
the phenomenon in some more depth. The focus here was on
establishing whether card fraud was similar to property crimes in
terms of victimization and fear levels, or whether it was more closely
aligned with the personal crimes. The secondary hypotheses were:
1 The incidence of card fraud is more similar to the incidence of

property crimes than to personal crimes.
As we saw earlier in Table 12.1, 2 per cent of respondents had been
a victim of card fraud. This figure was on a par with victimization
levels for theft of a vehicle and mugging/robbery, but lower than
rates for burglary and mugging. The conclusion here was that
victimization rates need to be interpreted independently and cannot
be contrasted on the grounds that they are ‘personal’ or ‘property’
crimes.
Table 12.2 Worry about crime

JOBNAME: Mcgraw−crowsemmens PAGE: 15 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 558791B1
2 The levels of worry about card fraud are more similar to the levels of
worry about property crimes than to levels of worry about personal
crimes.
In order to address this hypothesis, it was necessary to look more
closely at the distribution and dispersion of the worry variables for
all of the crimes. Beginning with the worry variables, Table 12.3
shows the distribution and dispersion values. It became clear that
worry about card fraud does not follow the same distribution
patterns as the other property crimes. First, looking at the average
values (mean and mode) for worry about card fraud compared with
the other crimes, we can see that the averages for card fraud are
closer to the personal crime averages, reflecting the lower levels of
worry for those crimes. Indeed, the distributions for worry about
burglary, theft of vehicle and theft from vehicle are positively skewed
but card fraud, like the four personal crimes, is negatively skewed
(but the skewness is notably smaller than for the other personal
crimes). Furthermore, card fraud has a higher value of variance and
thus, in terms of dispersion, is more similar to worry about rape and
attack than the other property crimes (which are less dispersed and
more clustered around the average).
To summarize, then, at a crime-specific level one can conclude that
there are distinct differences in the distribution and dispersion of
respondents across crimes. Card fraud has a similarly shaped
distribution to its property crime counterparts but the respondents
are well dispersed across the four worry categories, a feature more
strongly associated with the personal crimes (especially attack and
rape). We might interpret this to mean that worry about card fraud
is influenced by a range of factors, people may be more or less
worried for a range of different reasons. The analysis does not tell us
what those factors are, of course, but it does reinforce the need for
a complex theoretical framework in understanding fear of card fraud.
3 Victims of card fraud are more fearful of card fraud than are
non-victims.
The results showed that, for each crime, victims are significantly
more likely to be worried about that crime. So, for example, victims
of burglary are more likely to be ‘very worried’ about burglary than
non-victims of burglary. Similarly, non-victims are more likely to be
‘not at all worried’ about the crime. Table 12.4 shows that the
victims of card fraud have similar worry distributions to the victims
of the other property crimes – they are more clustered around the
top end of the worry scale and more likely to be very worried.
Victims of mugging and attack are less worried than victims of
property crime, differing little from their non-victim counterparts in

JOBNAME: Mcgraw−crowsemmens PAGE: 16 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 23F9B785
Table 12.3 Distribution and dispersion of worry variables
worry levels and variance across levels. Rape victims, however, are
considerably more worried than non-victims but, interestingly, more
dispersed across worry levels than the victims of the other crimes.
From this we might conclude that the victims of card fraud have a
heightened level of fear but we must be careful not to assume that
that victimization causes the fear. Further research would be needed
to establish a causal link.
Conclusion
There were a number of conclusions to be drawn from the analysis

of the data. Most importantly, this was the first attempt to introduce
plastic card fraud to a national victimization survey and it was
reasonable to conclude that it was a successful enterprise. Despite
fears to the contrary, the results reveal a significant level of

JOBNAME: Mcgraw−crowsemmens PAGE: 17 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 3861CCED
Table 12.4 Distribution and dispersion of worry for victims and non-victims of each
crime
victimization, together with an emergent ‘fear of plastic card fraud’

and the main objective of the research had been achieved.
If you recall, it had been hoped that by gathering this information
about plastic card fraud it would be possible to draw some useful
conclusions about the potential contribution which could be made to
(1) our understanding of fear of crime through the introduction of a
financial crime to the concept; and/or (2) our understanding of
plastic card fraud by attempting to assess the extent of victimization
and levels of fear that exist. Although this research only represents
the first step in building a more accurate picture of the phenomenon
of fear of crime, it was encouraging to gain some insight into the
contribution further survey work might make. In addition to
establishing that 2 per cent of respondents had been victims of card
fraud in the preceding year, it became clear that worry about card
fraud appears to be relatively high. This certainly suggests that
inclusion in future survey work has the potential to bring a new and
significant conceptual perspective to the fear of crime.
The research did raise a number of questions which clearly need
further thought. Reflecting on the original theoretical perspectives,
the most interesting question which arises is: why do so many people
worry about card fraud? Actually, individuals rarely suffer financial
loss in the event of fraud, it is usually the retailer or the financial
institution which bears the risk so there were two possible explana-
tions for this apparent paradox: either cardholders are under the
misconception that they, as individuals, are liable for the cost of the

JOBNAME: Mcgraw−crowsemmens PAGE: 18 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 526F23AF
fraud, or they are worried about some other aspect of the crime,
something beyond the simple loss of money. Either way, a great deal
could be learned by further exploratory research. Whether such
research will take place remains to be seen, however. Following its
successful inclusion in the BCS 2000 and a subsequent brief
appearance in the ‘Technology Crime’ module in the 2002 sweep of
the BCS, plastic card fraud has since been dropped from the survey.
Although this is a frustrating conclusion, it needs to be understood
in the context of policy-driven research. In this case, the Home
Office do have a responsibility to research areas of public concern
and it simply is not possible to measure everything. Anyway, future
developments may indeed be better suited to quite different meth-
odologies and the very fact that this research establishes a founda-
tion to be built upon should not be under-valued.
Final reflections on the research process itself were positive. Were
it not for the opportunity to collaborate with the British Crime
Survey team, a project of this scale would simply not have been
possible. In practical terms, I did not have to run extensive piloting,
train interviewers, organize access or even pay for the research! I was
able to avoid a lot of problems of the more complicated decisions
relating to sampling and questionnaire design, allowing me to
concentrate on the development of my small section of the ques-
tionnaire. However, there were also some disadvantages. The most
significant disadvantage was the restriction on the number of
questions I was permitted to use. This meant that many of the
conceptual distinctions I had made earlier in the research process
had to be set aside. Ultimately, then, my hypotheses were limited.
Also, the fact that I had no control over the methodology could be
seen as a disadvantage. When a researcher surrenders control of the
research design, s/he loses control over the issues of validity and
reliability. This was a particular issue when it came to wording the
new survey questions. There was a continual tension between
maintaining consistency within the questionnaire itself and allowing
for conceptual innovation. Finally, the amount of time it took to
learn to navigate the dataset could be seen as a disadvantage. It is
difficult at the best of times to use a dataset that someone else has
designed, but with a dataset as large as this one it was particularly
challenging. That having been said, I was provided with the
Technical Report and this was a great help.
We hope that this chapter has helped to illustrate a number of
important issues. It is, as we said at the start of the chapter, quite
common for criminological research to be conducted through a
collaboration of researchers or research teams. Such collaborations
can be extremely positive but they do require careful consideration.
Compromise and negotiation will be necessary and it is important
that you are clear from the outset what it is you wish to achieve out

JOBNAME: Mcgraw−crowsemmens PAGE: 19 SESS: 26 OUTPUT: Wed Aug 29 09:39:47 2007 SUM: 0B4B1A07
of the collaboration. You have to be very clear of the principles you

will prioritize because, at the end of the day, you want to produce
research that is valid and useful.

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 23 OUTPUT: Wed Aug 29 11:17:24 2007 SUM: 4D5DB2B8
/production/mcgraw−hill/booksxml/crowsemmens−11pt/v−concludingcomments
13 Concluding comments:Taking
it further
‘Research methods’ is sometimes thought to be a rather dry subject,
something that has to be endured before getting on to the more
interesting topics. Research is, in fact, a rewarding journey of
discovery. It is where substantive criminology comes from; it is the
production of knowledge, and what could be more exciting than
that? It also requires creativity and innovation. First, you need to
have ideas, and while these can be stimulated by reading and
discussion, it is your own experiences, observations and thoughts
that will produce the ideas for research. Second, you need to be
prepared to adapt and think laterally. Books such as this can give
you a basic knowledge, but inevitably the day comes when you have
to think it out for yourself. To give an example from one of our own
experiences, when you find yourself being asked to do some research
on a day centre for homeless alcoholics, you suddenly realize that a
30-page questionnaire has limited value and that you’d better think
of something else!
In this book we have sought to explain that criminological
research is part of an integrated process of inquiry. We have
described some of the methods that are used to further that inquiry,
and we have given examples of the ways in which inquiries can be
pursued. Inevitably there are many things that we have not done.
We could have covered some topics in more detail, and we could
have covered many others not touched on. For example, much
criminological research now has an international and comparative
dimension. This requires the consideration of a whole range of issues
that are beyond the scope of this book, although the same kind of
principles apply to comparative research as to other forms of inquiry.
We hope we have produced an introduction to researching crimi-
nology that can be used as the basis for further exploration. We end
by briefly indicating how you might extend your understanding of
what is involved in researching criminology, and start thinking about
conducting your own research.
The first thing to say is that the opportunities for researching
criminology are greater than ever before, for both students and
practitioners. There are more criminology courses available at uni-
versities than previously, and many of those (including our own) will
include the opportunity for students to do a research project at first
Kerrypress Ltd – Typeset in XML A Division: v-concludingcomments F Sequential 1

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 23 OUTPUT: Wed Aug 29 11:17:24 2007 SUM: 542AE2A6
hand. We believe that there should be a strong emphasis on

‘learning through research’ in criminology, encouraging students to
work individually or in groups on research projects under the
guidance of an experienced member of staff. What is exciting about
this approach is that those studying criminology can make a genuine
contribution to the discipline by means of online student journals,
student conferences, and sessions dedicated to work by students at
major conferences (such as the annual British Criminology Confer-
ence and the European Criminology Conference). There are also
opportunities to work with academic staff on projects they are
engaged in. In all this, of course, it is important to ensure that the
safety and well-being of students and participants are protected and
therefore student research needs to be carefully organized and
supervised.
For those people working outside of academia, there are also
many opportunities to contribute to the field of criminology (and
often have a more direct impact on crime policy). However,
becoming involved with criminological research in the longer term
requires developing a new range of skills and information. Perhaps
the biggest hurdle is the fact that you need money to conduct
research, so it is important that you know how to go about getting
it (and how to spend it once you have it!). Sadly crime and
criminality are a major cause of concern in society, but this does
mean that the need to know more about it produces a range of
resources for research. The most common sources of funding include
the following:1
+ Research Councils. The main research council covering the social
sciences, including criminology, is the Economic and Social
Research Council (ESRC). However, other funding councils may
also be relevant, including the Arts and Humanities Research
Board (AHRB) if one is interested in, say, historical research on
crime, or other forms of inquiry that do not include empirical
social science. The Engineering and Physical Sciences Research
Council (EPSRC) has funded a programme called ‘Think Crime’.
It is important to check out the research council’s requirements to
ensure that you choose the right one.
+ Government departments and agencies. A major source of fund-
ing for criminological research has been the Home Office, which
in May 2007 was reorganized to create a Ministry of Justice.
Other government agencies include the Statistics Commission,
and the Law Commission, and the Scottish Office funds research
north of the border.
+ Charitable foundations. A valuable source of funding for crimi-
nologists is a wide range of charitable trusts. The larger national
ones include the Nuffield Foundation, the Joseph Rowntree
Foundation, and the Leverhume Trust. However, there are others

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 23 OUTPUT: Wed Aug 29 11:17:24 2007 SUM: 50C57EFD
Concluding comments: Taking it further 257
which can be found in the Directory of Grant Making Trusts. If

you are considering focusing on a particular geographical area it is
also worth seeking out local funding agencies. The Barrow
Cadbury Trust, for example, can fund research particularly con-
cerned with the West Midlands area.
+ Specialist agencies. There are certain agencies working in criminal
justice which from time to time need research into issues that
they are addressing, such as Crime and Disorder Partnerships,
police forces, Victim Support, NACRO, and community groups.
However, they tend not to have research funding as such and it is
really a matter of working alongside them to understand what is
needed. If they are able to fund any research the amounts are
likely to be limited.
+ International funders. At the other end of the scale is interna-
tional funding, such as the European Union (EU), which has run
a series of Framework Programmes, usually quite wide-ranging in
scope, but often with the possibility of covering topics of interest
to criminologists.
In order to obtain funding you will need to produce a research
proposal. This is not the place to explain in detail how this is done,
but in effect a good research plan covers many of the topics covered
in this book. It needs to have clear aims, and a research design and
methods that address those aims. It also needs to have a timetable
and costing, and specify what the research outputs will be: reports,
conference papers, publications. It may also be appropriate to say
who will be most interested in the research apart from academics,
such as practitioners and policy-makers. Texts which cover the
writing of research proposals in more detail include Keith F. Punch’s
Developing Effective Research Proposals (2nd edition), and Nigel
Gilbert’s From Postgraduate to Social Scientist (2006). In all instances
it is not a bad idea to contact the potential funding source to discuss
your ideas informally before submitting a formal application.
The final point we want to make is the importance of research
skills. We have covered a wide range of methods and approaches in
this book and have highlighted the wide range of skills which are
needed. A good researcher needs strong interpersonal, communica-
tion and observation skills but that is not the full story! The drafting
and execution of a research project require skills that go beyond
knowing about research design, methods and analysis. Research
projects require management skills. The successful researcher will
need presentational skills, and be good at both time management
and the management of resources. By and large these are capabilities
that can be acquired. However, he or she will also need good
personal skills to manage relationships with staff, agencies and
participants, and be capable of taking on board constructive criti-
cism, and be adaptable and imaginative.

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 23 OUTPUT: Wed Aug 29 11:17:24 2007 SUM: 1F1242D5
If you are the kind of person who relishes such opportunities,

then you may wish to think about not only learning how crimino-
logical research is done, but how you could become someone who
contributes to it. Ultimately, we have sought to give you a good
grounding in the research process in this book to enable you to take
your interests further. We can only hope that we have managed to
convey our own enthusiasm about criminological research and
encourage you to make creative and confident decisions in your own
research. We finish by quoting one of our own students:
When you understand how research is carried out, you start to
look at criminology in a different way. I think I see the authors
of the things I read in a different light now – I understand
their motivations and inspirations. I suppose criminology has
come alive! I want to be part of it now … I am itching to
discover something new for myself. (Student (anonymous),
University of Sheffield, 2005–6)

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 22 OUTPUT: Wed Aug 29 09:44:09 2007 SUM: 3C847A64
/production/mcgraw−hill/booksxml/crowsemmens−11pt/w−glossary
Glossary
Various terms used in social research have specific meanings,
especially in relation to numerical analysis. These terms are
described in statistical texts, and there are dictionaries that are likely
to be particularly useful for the criminological researcher.1 However,
we felt that it would be useful for readers of this book to have a few
of the more commonly used terms ready to hand, and therefore offer
some brief descriptions, in alphabetical order.
Analysis by type of data and number of variables It is common
to make a distinction between variables that are analysed using
categories (e.g. sex has the categories ‘male’ and ‘female’), and those
that aren’t. This roughly corresponds to the distinction between
nominal data and interval or ratio level data. Information (usually
referred to as data) may be analysed for one variable at a time
(univariate analysis), two variables at a time (bivariate analysis), and
in situations which involve analysing more than two variables
(multi-variate analysis). Table G1 is a kind of ‘map’ showing which
kind of analysis can be done in particular circumstances.
Degrees of freedom Having carried out a statistical test and got a
result, you may have to look the result up in statistical tables in
order to determine whether the result is significant at a certain level
of confidence. In doing this you are quite likely to have to know
how many degrees of freedom were involved in the calculation. This
Table G1 Types of analysis
Kerrypress Ltd – Typeset in XML A Division: w-glossary F Sequential 1

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 22 OUTPUT: Wed Aug 29 09:44:09 2007 SUM: 4863FC45
refers to the number of cases that are free to vary, and will usually
be n – 1 or something similar, where n is the number of cases or the
number of pairs in your sample.2
Just to illustrate the concept, consider the following equation:
7 + 3 + ? = 12
What is the question mark? It is obviously 2. If the answer is 12,

then once the 7 and the 3 have been determined, the remaining
term in the equation must be 2 – it has no freedom to vary. For
example, we could have:
6 + 4 + ? = 12
The remaining term would still have to be 2. So, given the answer
12 and three terms in the equation, two have the freedom to vary,
the third does not. In other words there are two degrees of freedom,
or (where n equals the number of terms on the left-hand side of the
equation) n – 1 = 2.
Usually it will be made clear to you how many degrees of freedom
there are for a particular test, so all you have to do is look them up
in the appropriate row or column of a statistical table (or better still,
SPSS does it for you).
Hypothesis testing The simplest analyses are sometimes called
descriptive analyses because by and large all we are doing is saying
how many people there are in a particular category, what the mean
of a variable is, or how it is dispersed around the mean.
However, the main reason why we use statistics is to ask questions
– inferential statistics: if we know the mean age of a sample, can we
make any inferences about the mean age of the population from
which it is drawn? If we find that more people from one group are
sent to prison (or are fined more) than people from another group,
is this a real difference, or could it have occurred by chance? If we
observe that the children of those who have more money get better
exam results, how strong does this relationship have to be before we
say that there is a real association? (Note that being able to answer
these questions does not necessarily enable us to explain these
phenomena; for that we have to refer to the theories that caused us
to ask such questions in the first place.)
Answering such questions involves testing hypotheses to estimate
the probability of a particular event occurring by chance. The
examples given above are also examples of the main types of
hypothesis testing that are employed by social researchers, i.e.
+ whether a sample is representative of the population it is drawn
from;
+ whether there is a difference between one sample and another;

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 22 OUTPUT: Wed Aug 29 09:44:09 2007 SUM: 43FECC64
Glossary 261
+ whether there is an association between two phenomena.

Hypotheses are derived from the theories held by the researcher. In
order to test them statistically, they need to be stated in a particular
way.
Measurement ‘The assignment of numbers to objects or events
according to rules’ (S.S. Stevens, 1951, Handbook of Experimental
Psychology). There is a hierarchy of measurement which consists of
four main types of measure:
Nominal – classification by name
Ordinal – ranking items in order
Interval – equal intervals between points on the scale
Ratio – equal interval with zero point
The level of measurement that applies to a variable affects what you

can do with it: how it can be analysed and what kinds of statistical
test can be used. For example, in a class of students we might be
interested in students’ ages and which sex they are. We can count
how many are 20 years old, how many are 21 years old, etc. We can
also count up how many men there are and how many women.
These are the frequencies of particular ages and genders. However,
gender is a nominal variable, whereas age is a ratio variable. For age
we can calculate the average (or more properly the mean) age. But
we can’t have an average gender; all we can do is count the number
of people who are in a particular category. Having said that, it is
sometimes useful to calculate the proportion of people in a particular
category – i.e. what proportion in the class are women – and in this
way you can apply certain other statistical tests.
Scales are sometimes developed to measure hypothetical constructs
such as attitudes. For example, one might ask a sample of solicitors
whether they are happy with a particular piece of legislation and
score them on a scale:
1 = 2 = 3 = 4 = 5 =
very unhappy unhappy neutral happy very happy
Scores may also be developed to measure such things as seriousness

of offence, or severity of sentence. These may be comprised of
several variables. For example, to develop a score of how serious a
case is, one might combine scales reflecting the gravity of the
offence, the offender’s number of previous convictions, how long ago
it was since his/her last conviction, and other factors such as the
amount of damage caused or the amount stolen.

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 22 OUTPUT: Wed Aug 29 09:44:09 2007 SUM: 5314E86E
Null hypothesis The most common form of statistical hypoth-

esis is the null hypothesis: a hypothesis that there is no statistically
significant difference between your observations and what would be
expected, other things being equal. The null hypothesis is designated
as H0. For example, if you have a set of data concerning the
sentencing of black and white offenders, and you are interested in
finding out whether they are sentenced differently, your null hypoth-
esis is that there is no difference between the two groups. You then
carry out a statistical test to test this hypothesis, and you may then
reject the null hypothesis at a given level of confidence. Note that
this does not necessarily tell you that black people are discriminated
against in sentencing, because you then need to consider other
possible explanations.
One-tailed and two-tailed tests A test may be one-tailed, if you
are expecting a difference in a particular direction, or two-tailed if
you are simply looking at whether there is a difference, but not in
any particular direction. For example, you might be studying
whether support and counselling for the victims of crime have any
effect on their anxiety levels, and give them a scale to measure
anxiety before counselling and again afterwards. If you simply want
to see whether there is a difference, you would use a two-tailed test
to compare the scores before and after. But if you are expecting that
the level of anxiety will have decreased, you can use a one-tailed
test. Using the one-tailed test means that statistical significance is
more likely to be obtained. (This is because you are only using one
end of the distribution of scores, so you don’t have to split the
probability between both ends of the distribution of scores. This
may become clearer as one learns more about distributions and
probability levels.)
Populations and samples A population is the universe of cases
or units (people, things, events) with which we are concerned, and
about which inferences are to be made, e.g. all magistrates courts or
all the magistrates in a particular court. A sample is part of a
population acquired such that it is representative of the population.
The characteristics of a population are called parameters, while the
characteristics of a sample are statistics. Much depends on how a
sample is drawn; whether it is random or non-random.
Probability The concept of probability is central to statistical
analysis and is important. We can only test hypotheses at a given
level of probability. Derived from gambling in the seventeenth
century by Pascal and others, probability theory as a subject in its
own right stems from the publication of The Analytic Theory of
Probabilities by Laplace in 1812. Probability is concerned with the
analysis of quantities derived from observations of phenomena

JOBNAME: Mcgraw−crowsemmens PAGE: 5 SESS: 22 OUTPUT: Wed Aug 29 09:44:09 2007 SUM: 4EBBA48C
Glossary 263
whose occurrence involves a chance element. It is the basis for

hypothesis testing, calculating the chances of an event occurring out
of the possibilities available:
Number of “successes” Heads 1
P = Number of possible outcomes = Heads or Tails = 2 = .5
If a dice is thrown 36 times, you would expect to get 6 sixes. If you

get it less or more than this, is it just chance or do we conclude that
the dice is loaded, i.e. at what point do we reject the null hypothesis
that the dice is not loaded?
Similarly, if we pick at random a sample from the general
population then in theory half should be female and half should be
male. On the other hand perhaps we are studying sentencing at a
particular magistrates’ courts, and we know from Home Office
figures that 17% of those convicted of crime in a given year are
female. In our sample 18% are female. Do we conclude that more
females are dealt with at our court than nationally or is this
attributable to some slight and not significant variation? At what
point do we reject the null hypothesis that there is no difference
between our sample and what we would expect, other things being
equal: at 19%? 20%? 21%? To test such a hypothesis we calculate the
probabilities using an appropriate statistical test. (A good explana-
tion of the relevance of probability to statistical analysis can be
found in Cramer and Howitt 2004.) Such analyses can only be
applied to random samples, but there are non-parametric tests, which
are not based on random probability.
Significance Test A calculation made to see whether a result
that has been obtained varies more than would be expected by
chance. On the basis of a significance test one may accept or reject
a null hypothesis. Since it is seldom possible to be certain about
what the result should be, a significance test is usually expressed in
terms of the probability of its being correct, i.e. as a 95% chance
that your result is different from what you would expect, alterna-
tively expressed as p < 0.05 probability of being wrong. This is the
5% confidence level, i.e. we are willing to accept that 5 per cent of the
time we are wrong in rejecting the null hypothesis. 5% is the level
used most often, but there is nothing ‘magic’ about it; sometimes we
might be more stringent and apply a 1% level of certainty, but it is
in the nature of probability that we can never be totally certain.
In some circumstances we use confidence limits. If I go for a walk
in the countryside and when I get back someone asks me how far
have I walked I may say ‘about ten miles, give or take a couple of
miles’. In other words I estimate with reasonable confidence that I
have walked at least eight miles and no more than twelve miles.
These are my confidence limits, and two miles is my confidence

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 22 OUTPUT: Wed Aug 29 09:44:09 2007 SUM: 3FBCE041
interval, in other words 10 ± 2. In statistics, we are rather more

precise and after an appropriate calculation we might say that we are
95% confident that the distance walked lies between eight and ten
miles. Similarly, referring to the example used earlier where 17% of
those appearing in court nationally are female, we can calculate that
if the proportion of females in our sample lies outside the range
14–20% (the confidence interval being calculated as 3% in this case),
then we can reject the null hypothesis that there is no difference
between our sample and what we would expect, and we might
conclude that, for whatever reason, women are more likely or less
likely to be convicted at this court than in the country as a whole.
(In the example 18% of our sample were female, and since this is
within the limits we are prepared to accept our sample as being
within the confidence limits, and therefore, on this variable, one on
which we can base national generalizations.)
Type I and Type II errors If we only have a certain level of
confidence in our results then, as explained above, we have to accept
that sometimes the null hypothesis will not be rejected when it
should have been, or be rejected when it should not have been. The
first is called a Type I error and the second a Type II error (Table
G2).
Variables As the name suggests, a variable is a characteristic
that varies: age, height, gender, ethnic group, etc. Variables need to
be unitary and clearly defined.
Some appear only as categories (gender, race, occupation, etc.).
Others are more continuous in nature (time, money, length, etc.).
A variable is usually represented algebraically as x. Where there
are several variables, x1, x2, x3 etc. X is one value for a variable.
Variables take different forms: dependent, independent and interven-
ing.
Table G2 Type I and Type II errors

JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 33 OUTPUT: Wed Aug 29 09:46:19 2007 SUM: 516399BA
/production/mcgraw−hill/booksxml/crowsemmens−11pt/x−notes
Notes
1 For a good explanation of the different forms of positivism and the distinction
between positivism and the scientific method, see the entry on ‘Positivism’ by
Malcolm Williams in the Sage Dictionary of Social Research Methods (Jupp 2006).
2 The use of observations and attitude scales are discussed later in the book, in
Chapters 6 and 7 respectively.
3 See, for example, the July 2005 issue of the International Journal of Social
Research Methodology, Vol. 8, No. 3, and Bryman (2006).
4 We lay no claim to this being a particularly original formulation. It can be found
in one form or another in various texts. See, for example, Bryman and Cramer
(1990: 3).
5 Attempts to do this may involve some supposedly ‘objective’ measure, such as
having a certain socio-economic position, or by means of an attitude scale,
whereby respondents are asked to say to what extent they agree that certain
statements such as ‘I don’t feel that I belong to this society’ apply to them. More
is said about attitude scales in Chapter 7.
6 Observational notes are defined as ‘Statements bearing upon events experienced
principally through watching and listening’, involving as little interpretation as
possible. Theoretical notes ‘represent self-conscious, controlled attempts to derive
meaning from any one or several observational notes’, and are reflections on
what is observed. Methodological notes are ‘A statement that reflects an
operational act completed or planned; an instruction to oneself, a reminder, a
critique of one’s own tactics’, and are observational notes on the researcher him
or herself.
7 Organizations such as the Home Office, or the Joseph Rowntree Foundation,
who fund research will want to produce a 4–6 page summary of findings.
2 The principles of researching criminology
1 The earliest theories about crime emerged from the Classical School in the early
part of the nineteenth century. Both Bentham and Beccaria were part of this
movement, focusing on the development of theories of crime, criminal justice,
penology and explaining crime by human nature. They did not to base their
Kerrypress Ltd – Typeset in XML A Division: x-notes F Sequential 1

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 33 OUTPUT: Wed Aug 29 09:46:19 2007 SUM: 628FC2EF
conclusions on systematic empirical work, drawing instead on ‘common-sense

observations and armchair reflections’ (Bottoms 2000: 26).
2 The Chicago School led the way in these developments. They developed the
positivistic approach in their ecological analysis of crime in the urban context.
The city was their ‘laboratory’ and they advocated a scientific approach, but they
incorporated anthropological methods of observation which were distinctly
non-positivistic.
3 Designing criminological research
1 Defined here as the effect that one variable has on another: X brings about Y.
2 Of course, it is important to define what is meant by ‘well off ’, and arrive at an
appropriate measure of well-offness, which could be things like car ownership,
council tax bands, proportion of children qualifying for free school meals, and so
on.
3 A cross-tabulation is a table with one variable forming the columns and another
variable forming the rows.
4 If repeated random samples were drawn from the same population then the
mean of all the sample means gives an estimate of the true mean. This is what
is known as the Central Limit Theorem. The standard deviation of the sampling
distribution of the mean is referred to as the standard error of the mean. The
standard error of the mean (S.E.X) is obtained by dividing the standard
deviation by the square root of the sample size:
s
S . E . X̄ =
√n
The normal distribution has the property that 95 per cent of sample means fall
within an area enclosed by 1.96 standard errors either side of the population
mean. Thus the true mean will lie somewhere between ± 1.96 × S.E.X. Thus, if
we take a sample of offenders and find that their mean age is 21 years and
calculate that the standard error of that mean is 1.2, then there is a 95 per cent
probability that the true age of the population of offenders from which the
sample was drawn is somewhere between 18.7 years (21 – 1.96 × 1.2) and 23.3
years (21 + 1.96 × 1.2).
5 Several papers presenting contrasting views were published by the Survey
Methods Centre following a seminar on the topic (1994).
6 A brief, but useful discussion of snowballing can be found in the University of
Surrey’s occasional series on social research techniques (Atkinson and Flint
2001).
7 http://www.britsoccrim.org/ethical.htm.
8 This can be found at: http://www.esrcsocietytoday.ac.uk/ESRCInfoCentre/
Images/ESRC_Re_Ethics_Frame_tcm6-11291.pdf
9 The basis for these can be found in Beauchamp, T. L. and Childress, J. F. (2001)
Principles of Biomedical Ethics (5th edn). Oxford: Oxford University Press.
1 It is, of course, possible that the initiators are being entirely cynical and only
using evaluation to give some kind of scientific veneer to what was going to be
done anyway, but it is to be hoped that this never actually happens.

Notes 267
2 This applies just as much to medical situations. For example, it was reported
that a clinical trial of a drug (Tykerb) which might extend the life expectancy of
women with breast cancer had been stopped early because of good results. When
it became clear that progression of the disease was slowed compared with the
standard treatment, the trial was stopped on the basis that it would be unethical
to deny treatment to a comparison group that was keeping others alive longer
(reported in The Guardian, 4.4.2006, and at http://www.gsk.com).
3 Internal validity is defined as a study’s ability to determine cause and effect,
while external validity is the ability to generalize the findings of the study to
analogous situations.
4 A technique for analysing data that are hierarchically structured. For example,
we may wish to make comparisons involving regions, towns, districts, wards and
streets.
1 http://www.esds.ac.uk/qualidata/about/introduction.asp
2 Hall et al. (1978) famously illustrated that statistics were deliberately used to
manufacture moral panics and crime waves through his analysis of the treatment
of mugging cases in the 1970s.
3 There has been much criticism of the fact that the Home Office is responsible
for the administration of both the British Crime Survey and the Recorded
Crime Statistics and it is recommended that control of the BCS should be
transferred to the Office of National Statistics (the police statistics should
remain with the Home Office as they are an output of Home Office
administrative systems).
7 Researching by asking and listening
1 That is, social research relating to the law in general, rather than specifically
criminological research. It is also sometimes referred to as law in its social
context.
2 Often this will be individual people, but not necessarily. For example, to enquire
about what methods different magistrate’s courts use to chase up fine defaulters,
questionnaires could be sent to the clerks of all or a sample of magistrate’s
courts.
3 Although just to confuse matters, sometimes the responses to a particular
question need to be coded as separate variables. Types of response are mentioned
later, and the way such data is analysed is explained in Chapter 8.
8 Analysing criminological research
1 To quote the ESDS’s own description, it is ‘a national data service providing

access and support for an extensive range of key economic and social data, both
quantitative and qualitative, spanning many disciplines and themes. ESDS

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 33 OUTPUT: Wed Aug 29 09:46:19 2007 SUM: 500AB6B0
provides an integrated service offering enhanced support for the secondary use of
data across the research, learning and teaching communities.’ Access can be
obtained at www.esds.ac.uk.
2 Variously attributed to Mark Twain, Benjamin Disraeli, and others.
3 SPSS can be used to display graphs of variables, and there is an option called
the Explore command which will show the measure of skewness and kurtosis,
along with other attributes of each variable. You need to read one of the books
mentioned in Further Reading for a detailed explanation of how to use these
facilities.
4 There are special kinds of statistics, called non-parametric statistics, that do not
require the same levels of measurement as equal interval variables.
5 Because we are examining areas, not individuals, this is called ecological
correlation. It is important not to confuse the correlation of one kind of unit
(such as areas) with another kind of unit (individuals). The fact that there is a
relationship between unemployment rate and conviction rate across areas does
NOT mean that there would necessarily be a correlation between unemployment
and conviction among individuals.
6 This data comes from an unpublished local study conducted in 1999, looking at
fear of crime, using a postal questionnaire.
9 Researching offenders and employment
1 Full details of the study were published in Crow and Simon (1987), and a less
detailed account in Crow et al. (1989).
10 Researching the Youth Court
1 Speech to the Labour Party Conference, 2 October, 1997.
1 In the early 1990s, it emerged that the proprietor of the Mirror Newspaper,
Robert Maxwell, had used the company pension fund to finance his business
interests. After his death in an unexplained accident in 1991, campaigners for
the 30,000 Mirror Group pensioners mounted a three-year campaign for
compensation. Their funds were largely recovered thanks to a £100m govern-
ment payout and a £276m out-of-court settlement with City institutions and the
remnants of Robert Maxwell’s media group.
13 Concluding comments
1 All the sources mentioned have websites which can easily be found using a
search engine.

JOBNAME: Mcgraw−crowsemmens PAGE: 5 SESS: 33 OUTPUT: Wed Aug 29 09:46:19 2007 SUM: 1EADDBAF
Notes 269
Glossary
1 The Sage Dictionary of Criminology, compiled and edited by Eugene McLaughlin

and John Muncie (2001), The Sage Dictionary of Statistics, compiled by Duncan
Cramer and Dennis Howitt (2004), and The Sage Dictionary of Social Research
Methods, edited and compiled by Victor Jupp (2006).
2 This is because the laws of chance on which probability is based assume that
observations are independent and have the freedom to vary. However, when we
use deviation scores the sum of deviations is zero, so that one of the scores is
determined as soon as all the others are known, so only n – 1 of them are free
to vary. Therefore a correction has to be made for any loss of independence or
the estimate of the population parameter will be biased.


JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 18 OUTPUT: Wed Aug 29 11:32:10 2007 SUM: 4A21B656
/production/mcgraw−hill/booksxml/crowsemmens−11pt/y−references
References
Allen, C., Crow, I. and Cavadino, M. (2000) Evaluation of the Youth

Court Demonstration Project. Home Office Research Study 214.
London: Home Office.
Arber, S. (2001) Secondary analysis of survey data. In N. Gilbert,
(ed.) Researching Social Life. London: Sage Publications.
Ashworth, A., Genders, E., Mansfield, G., Peay, J. and Player, E.
(1984) Sentencing in the Crown Court: Report of an Exploratory
Study. Occasional Paper No. 10. Oxford: University of Oxford,
Centre for Criminological Research.
Atkinson, P. (1981) Transition from school to working life. Socio-
logical Research Unit, Cardiff (unpublished memorandum).
Atkinson, R. and Flint, J. (2001) Accessing Hidden and Hard-to-Reach
Populations: Snowball Research Strategies. University of Surrey,
Social Research Update, Issue 33.
Baker, K., Curtice, J. and Sparrow, N. (2003) Internet Poll Trial:
Research Report. ICM. Available at: http://www.icmresearch.co.uk/
reviews/2002/Internet-polling-paper-jan-03.htm
Beauchamp, T. L. and Childress, J. F. (2001) Principles of Biomedical
Ethics, 5th edn. Oxford: OUP.
Beck, U. (1992) Risk Society. London: Sage Publications.
Becker, H. (1958) Problems of inference and proof in participant
observation. American Sociological Review, 23: 652–60.
Becker, H. (1963) Outsiders: Studies in the Sociology of Deviance. New
York: Free Press.
Bennett, T. (1991) The effectiveness of a police-initiated fear-
reducing strategy. British Journal of Criminology, 31: 1–14.
Kerrypress Ltd – Typeset in XML A Division: y-references F Sequential 1

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 18 OUTPUT: Wed Aug 29 11:32:10 2007 SUM: 52F9C024
Bennett, T. (1996) What’s new in evaluation research? A note on

the Pawson and Tilley article. British Journal of Criminology,
36(4): 567–73.
Bottoms, A. (2000) The relationship between theory and research in
criminology. In R. D. King and E. Wincup, Doing Research on
Crime and Justice. Oxford: Oxford University Press.
Bowers, K. J., Johnson, S. D. and Hirschfield, A. F. G. (2003)
Pushing Back the Boundaries: New Techniques for Assessing the
Impact of Burglary Schemes. Home Office Online report 24/03.
Box, S. (1987) Recession, Crime and Punishment. Basingstoke: Mac-
millan.
Box, S. and Hale, C. (1982) Economic crisis and the rising prisoner
population in England and Wales. Crime and Social Justice, 17:
20–35.
Brenner, M. H. (1976) Time Series Analysis – effects of the
economy on criminal behaviour and the administration of crimi-
nal justice. In United Nations Social Defence Research Institute,
Economic Crises and Crime. Rome: UNSDRI.
British Association for the Advancement of Science (1874) Notes
and Queries on Anthropology. London.
Browne, D., Francis, E. and Crow, I. (1993) Black people, mental
health and the criminal justice system. In W. Watson and A.
Grounds (eds) The Mentally Disordered Offender in an Era of
Community Care. Cambridge: Cambridge University Press.
Bryman, A. (2004) Social Research Methods 2nd edn. Oxford: Oxford
University Press.
Bryman, A. (2006) Mixed Methods. London: Sage.
Bryman, A. and Cramer, D. (1990) Quantitative Data Analysis for
Social Scientists. London: Routledge.
Bulmer, M. (edn.) (1982) Social Research Ethics. London: Macmillan.
Burton, D. (ed.) (2000) Research Training for Social Scientists. Lon-
don: Sage.
Catalano, R. and Hawkins, J. D. (1996) The social development
model: a theory of antisocial behaviour. In J. D. Hawkins (ed.)
Delinquency and Crime. Cambridge: Cambridge University Press.
Cavadino, M., Crow, I. and Dignan, J. (2000) Criminal Justice 2000:
Strategies for a New Century. Winchester: Waterside Press.

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 18 OUTPUT: Wed Aug 29 11:32:10 2007 SUM: 4D32EEF7
References 273
Clear, T. (1997) Evaluating intensive probation: the American

experience. In G. Mair (ed.) Evaluating the Effectiveness of
Community Penalties. Aldershot: Avebury.
Clegg, F. (1990) Simple Statistics. Cambridge: Cambridge University
Press.
Coffey, A., Delamont, S. and Atkinson, P. (2007) Handbook of
Ethnography. London: Sage.
Cohen, A. (1973) Folk Devils and Moral Panics: The Creation of the
Mods and Rockers. London: Paladin.
Cook, T. D. and Campbell, D. T. (1979). Quasi-Experimentation.
Chicago: Rand-McNally.
Cooper, H. M. (1989) Integrating Research: A Guide for Literature
Reviews. London: Sage.
Cramer, D. and Howitt, D. (2004) The Sage Dictionary of Criminol-
ogy. London: Sage.
Cressey, D. (1950) The criminal violation of financial trust. American
Sociological Review, 15: 738–43.
Creswell, J. W. (2002) Research Design: Qualitative, Quantitative and
Mixed Methods Approaches, 2nd edn. London: Sage.
Crow, I. (1996) Approaches to Youth Crime: A Study of the Views of
Magistrates, Justices’ Clerks and Social Workers. Sheffield: University
of Sheffield, Centre for Criminological and Legal Research.
Crow, I. (2000) Evaluating initiatives in the community. In V. Jupp,
P. Davies and P. Francis (eds) Doing Criminological Research.
London: Sage.
Crow, I. (2001) The Treatment and Rehabilitation of Offenders.
London: Sage.
Crow, I., Cavadino, M., Dignan, J., Johnston, V. and Walker, M.
(1996) Changing Criminal Justice: The Impact of the Criminal Justice
Act 1991 in Four Areas of the North of England. Sheffield:
University of Sheffield, Centre for Criminological and Legal
Research.
Crow, I., France, A., Hacking, S. and Hart, M. (2004) Does
Communities that Care Work? An Evaluation of a Community-based
Risk Prevention Programme in Three Neighbourhoods. York: Joseph
Rowntree Foundation.
Crow, I., France, A. and Hacking, S. (2006) The evaluation of three
Communities That Care projects in the UK, Security Journal,
19(1): 45–57.

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 18 OUTPUT: Wed Aug 29 11:32:10 2007 SUM: 4C23FE77
Crow, I., Howells, G. and Moroney, M. (1993) Credit and debt:

choices for poorer consumers. In G. Howells, I. Crow and M.
Moroney (eds) Aspects of Credit and Debt. London: Sweet and
Maxwell.
Crow, I., Richardson, P., Riddington, C. and Simon, F. (1989)
Unemployment, Crime and Offenders. London: Routledge.
Crow, I. and Simon, F. (1987) Unemployment and Magistrates Courts.
London: NACRO.
Crow, I. and Stubbing, T. (1999) Fast tracking persistent young
offenders: to what effect? Liverpool Law Review, 21(2–3): 169–96.
Dale, A., Arber, S. and Proctor, M. (1988) Doing Secondary Analysis.
London: Unwin Hyman.
Delanty, G. and Strydom, P. (2003) Philosophies of Social Science: The
Classic and Contemporary Readings. Milton Keynes: Open Univer-
sity Press.
Denzin, N. K. (1988) The Research Act: A Theoretical Introduction to
Sociological Methods. Englewood Cliffs, N. J. : Prentice Hall.
De Vaus, D. (2002) Analysing Social Science Data: 50 Key Problems in
Data Analysis. London: Sage.
Ditchfield, J. and Marshall, P. (1990) A review of recent literature
evaluating treatments for sex offenders in prison. Prison Service
Journal, 81: 24–8.
Ditton, J. (1977) Part Time Crime: An Ethnography of Fiddling and
Pilferage. London: Macmillan.
Ditton, J. (2000) Crime surveys and the measurement problem: fear
of crime. In V. Jupp, P. Davies and P. Francis, Doing Criminologi-
cal Research. London: Sage.
Ekblom, P., Law, H. and Sutton, M. (1996) Safer Cities and Domestic
Burglary. Home Office Research Study 164. London: Home
Office.
Engels, F. (1969) The Condition of the Working Class in England.
London: Panther.
Falshaw, L., Friendship, C. and Bates, A. (2003) Sexual Offenders:
Measuring Reconviction, Reoffending and Recidivism. Home Office
Research, Development and Statistics Directorate Research Find-
ings No. 183. London: Home Office.
Farrington, D. P., Gallagher, B., Morley, L., St Ledger, R. J. and
West, D. J. (1986) Unemployment, school leaving and crime.
British Journal of Criminology, 23(3): 229–48.

JOBNAME: Mcgraw−crowsemmens PAGE: 5 SESS: 18 OUTPUT: Wed Aug 29 11:32:10 2007 SUM: 50E914F7
References 275
Farrington, D. (1996) Understanding and Preventing Youth Crime.

York: Joseph Rowntree Foundation.
Farrington, D. P. (1997) Evaluating a community crime prevention
program, Evaluation, 3(2): 157–73.
Farrington, D. P. (1998) Evaluating ‘Communities That Care’:
realistic scientific considerations. Evaluation, 4(2): 204–10.
Farrington, D. (2000) Explaining and preventing crime: the globali-
sation of knowledge. Key note Address to the American Society
for Criminology, 1999. Criminology, 38(1): 1–23.
Field, A. (2004) Discovering Statistics Using SPSS. London: Sage.
Fielding, J. (2001) Coding and managing data. In N. Gilbert (ed.)
Researching Social Life. London: Sage Publications.
Fielding, J. and Gilbert, N. (2000) Understanding Social Statistics.
London: Sage.
Fink, A. (1995) How to Analyse Survey Data. London: Sage.
Fischer, F. (1998) Beyond empiricism: policy inquiry in postpositivist
perspective. Policy Studies Journal, 26(1): 129–46.
Folkard, M. S., Smith, D. E. and Smith, D. D. (1976) IMPACT
Volume II. Home Office Research Study 36. London: Home
Office.
Foster, J. J. (2001) Data Analysis Using SPSS for Windows. London:
Sage.
France, A. and Crow, I. (2001) CTC – The Story So Far: An Interim
Evaluation of Communities That Care. York: Joseph Rowntree
Foundation.
France, A. and Crow, I. (2005) Using the ‘Risk Factor Paradigm’ in
prevention: lessons from the Evaluation of Communities that
Care. Children and Society, 19: 172–84.
Friendship, C., Beech, A. R. and Browne, K. D. (2002) Reconviction
as an outcome in research: a methodological note. British Journal
of Criminology, 42: 442–4.
Friendship, C., Street, R., Cann, J. and Harper, G. (2004) Introduc-
tion: the policy context and assessing the evidence. In G. Harper
and C. Chitty (eds) The Impact of Corrections on Re-Offending: A
Review of ‘What Works’. Home Office Research Study 291.
Ganzini, L., McFarland, B. and Bloom, J. (1990) Victims of fraud.
Bulletin of the American Academy of Psychiatry and Law, 18: 55–63.

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 18 OUTPUT: Wed Aug 29 11:32:10 2007 SUM: 4E0C7C79
Garfinkel, H. (1967) Studies in Ethnomethodology. Englewood Cliffs,

NJ: Prentice-Hall.
Garland, D. (2001) The Culture of Control: Crime and Social Order in
Contemporary Society. Oxford: Oxford University Press.
Garland, D. (2002) Of crime and criminals: the development of
criminology in Britain. In M. Maguire, R. Morgan and R. Reiner
(eds) The Oxford Handbook of Criminology. Oxford: Oxford Uni-
versity Press.
Gelsthorpe, L. and Sharpe, G. (2006) Criminological research:
typologies versus hierarchies. Criminal Justice Matters, 62: 8–43.
Gewirth, A. (1996) The Community of Rights. Chicago: University of
Chicago Press.
Gilbert, N. (ed.) (1993) Researching Social Life. London: Sage.
Gilbert, N. (2006) From Postgraduate to Social Scientist. London:
Sage.
Glaser, B. G. and Strauss, A. L. (1967) The Discovery of Grounded
Theory: Strategies for Qualitative Research. New York: Aldine
Publishing.
Goffman, E. (1961) Asylums: Essays on the Social Situation of Mental
Patients and Other Inmates. New York: Doubleday.
Gold, R.L. (1958) Roles in sociological fieldwork. Social Forces, 36:
217–23.
Gormally, B., Lyner, O., Mulligan, G. and Warden, M. (1981)
Unemployment and Young Offenders in Northern Ireland. Belfast:
NIACRO.
Grabosky, P., Smith, R. G. and Dempsey, G. (2001) Electronic Theft:
Unlawful Acquisition in Cyberspace. Cambridge: Cambridge Uni-
versity Press.
Green, A. and Preston, J. (2005) Speaking in tongues: diversity in
mixed methods research. International Journal of Social Research
Methodology, 8(3): 167–71.
Hale, C. (1996) Fear of crime: a review of the literature. Interna-
tional Review of Victimology, 4: 79–150.
Hales, J., Henderson, L., Collins, D. and Becher, H. (2000) 2000
British Crime Survey Technical Report. London: National Centre
for Social Research.
Hall, S., Critcher, C., Jefferson, T. and Roberts, B. (1978) Policing
the Crisis. London: Macmillan.

JOBNAME: Mcgraw−crowsemmens PAGE: 7 SESS: 18 OUTPUT: Wed Aug 29 11:32:10 2007 SUM: 4CCC56E0
References 277
Hammersley, M. (1992) What’s Wrong with Ethnography? London:

Routledge.
Hammersley, M. and Atkinson, P. (1983) Ethnography: Principles in
Practice. London: Tavistock Publications.
Hammersley, M. and Atkinson, P. (1995) Ethnography: Principles in
Practice. London: Routledge.
Hargreaves, D. H., Hesterr, S. and Mellor, F. (1975) Deviance in
Classrooms. London: Routledge and Kegan Paul.
Harper, G. and Chitty, C. (eds) (2004) The Impact of Corrections on
Re-Offending: A review of ‘What Works’. Home Office Research
Study 291. London: Home Office.
Hart, C. (1998) Doing a Literature Review: Releasing the Social
Science Research Imagination. London: Sage.
Hawkins, J. D., Catalano, R. and Miller, J. (1992) Risk and
protective factors for alcohol and other drug problems in adoles-
cence and early adulthood: implications for substance abuse
prevention. Psychological Bulletin, 112(1): 66–105.
Hedderman, C. and Sugg, D. (1996) Does Treating Sex Offenders
Reduce Reoffending? Research Findings No. 45, Home Office
Research and Statistics Directorate. London: Home Office.
Heritage, J. (1987) Ethnomethodology. In A. Giddens and J. Turner
(eds) Social Theory Today. Cambridge: Polity.
Hobbs, D. (1988) Doing the Business: Entrepreneurship, the Working
Class and Detectives in the East End of London. Oxford: Oxford
University Press.
Holdaway, S. (1983) Inside the British Police: A Force at Work. Oxford:
Blackwell.
Hollister, R. and Hill, J. (1999) Problems in the evaluation of
community-wide initiatives. In J. Connell, A. Kubisch, L. Schorr
and C. Weiss (eds) New Approaches to Evaluating Community
Initiatives. Vol. 1. New Jersey: The Aspen Institute.
Holsti, O. (1969) Content Analysis for the Social Sciences and Humani-
ties. Reading, MA: Addison-Wesley.
Home Office (1974) Working Party on Vagrancy and Street Offences
Working Paper. Appendix A. London: HMSO.
Home Office (1997) No More Excuses: A New Approach to Tackling
Youth Crime in England and Wales. Cm 3809. London: The
Stationery Office.

JOBNAME: Mcgraw−crowsemmens PAGE: 8 SESS: 18 OUTPUT: Wed Aug 29 11:32:10 2007 SUM: 4A11DC96
Home Office (2000) Review of Crime Statistics: A Discussion Docu-

ment. London: Home Office.
Hood, R. (1972) Sentencing the Motoring Offender. Cambridge
Studies in Criminology. London: Heinemann.
Hope, T. (2004) Pretend it works: evidence and governance in the
evaluation of the reducing burglary initiative. Criminal Justice,
4(3): 287–308.
Hough, M., Clancy, A., McSweeney, T. and Turnbull, P. J. (2003)
The Impact of Drug Treatment and Testing Orders on Offending:
Two-year Reconviction Results. Home Office Research Findings
184. London: Home Office.
House of Commons Science and Technology Committee (2006)
Scientific Advice, Risk and Evidence Based Policy Making. HC900–1,
7th Report of Session 2005–6 (Vol. 1). London: The Stationery
Office.
Hughes, J. (1990) The Philosophy of Social Research, 2nd edn.
London: Longman.
Hullin, R. (1985) The Leeds Truancy Project. Justice of the Peace,
488–91.
Israel, M. and Chui, W. H. (2006) If ‘Something works’ is the
answer, what is the question? Supporting pluralist evaluation in
community corrections in the United Kingdom. European Journal
of Criminology, 3(2): 181–200.
Johnson, J. (1975) Doing Field Research. New York: Free Press.
Jupp, V. (2006) The Sage Dictionary of Social Research Methods.
London: Sage.
Jupp, V., Davies, P. and Francis, P. (eds) (2000) Doing Criminological
Research. London: Sage Publications.
Justice of the Peace (1982) A job to go to Monday: notes of the
week. Justice of the Peace.
Kane, E. (1985) Doing Your Own Research. London: Marion Boyars
Publishing.
Kant, I. (1871) Critique of Pure Reason. Trans. J. M. D. Meiklejohn.
New York: Dover Publications Inc.
Kelle, U. (1997) Capabilities for theory building and hypothesis
testing in software for computer-aided qualitative data analysis.
The Data Archive Bulletin, 65, May 1997, 12–14.

JOBNAME: Mcgraw−crowsemmens PAGE: 9 SESS: 18 OUTPUT: Wed Aug 29 11:32:10 2007 SUM: 5003FD77
References 279
Kemp, C., Norris, C. and Fielding, N. (1992) Negotiating Nothing:

Police Decision-Making in Disputes. Aldershot: Avebury.
Kershaw, C. (1998) Interpreting reconviction Rates. In The Use and
Impact of Community Supervision, Research Bulletin No. 39,
Special Edition, 9–16. Home Office Research and Statistics
Directorate. London: Home Office.
Kershaw, C., Budd, T., Kinshott, G., Mattinson, J., Mayhew, P. and
Myhill, A. (2000) The 2000 British Crime Survey. Home Office
Statistical Bulletin 18/00. London: Home Office.
King, R. D. and Wincup, E. (2000) Doing Research on Crime and
Justice. Oxford: Oxford University Press.
Kranzler, G. and Moursund, J. (1995) Statistics for the Terrified.
Englewood Cliffs, NJ: Prentice Hall.
Lacey, N. (2007) Legal constructions of crime. In M. Maguire, R.
Morgan and R. Reiner (eds) The Oxford Handbook of Criminology,
4th edn. Oxford: Oxford University Press.
LaGrange, R. L. and Ferraro, K. F. (1989) Assessing age and gender
differences in perceived risk and fear of crime. Criminology, 27:
697–719.
Levi, M. and Pithouse, A. (1992) The victims of fraud. In D.
Downes (ed.) Unravelling Criminal Justice. London: Macmillan
Press Ltd, pp. 229–46.
Lewins, A. (2001) Computer assisted qualitative data analysis. In N.
Gilbert (ed.) Researching Social Life. London: Sage Publications.
Lloyd, C., Mair, G. and Hough, M. (1994) Explaining Reconviction
Rates: A Critical Analysis. Home Office Research Study 136.
Loftland, L. H. (1973) A World of Strangers: Order and Action in
Urban Public Space. New York: Basic Books.
Madriz, E. (1997a) Images of criminals and victims: a study on
women’s fear and social control. Gender and Society, 11(3): 342–56.
Madriz, E. (1997b) Nothing Bad Happens to Good Girls: Fear of Crime
on Women’s Lives. Berkeley, CA: University of California Press.
Maguire, M. (2002) Crime statistics: the ‘data explosion’ and its
implications. In M. Maguire, R. Morgan and R. Reiner, The
Oxford Handbook of Criminology. 3rd edn. Oxford: Oxford Uni-
versity Press.

Maguire, M. (2007) Crime data and statistics. In M. Maguire, R.

Morgan and R. Reiner, The Oxford Handbook of Criminology, 4th
edn. Oxford: Oxford University Press.
Mair, G. (ed.) (1997) Evaluating the Effectiveness of Community
Penalties. Aldershot: Avebury.
Mair, G. (ed.) (2004) What Matters in Probation. Cullompton,
Devon: Willan Publishing.
Mair, G. et al. (1994) Intensive Probation in England and Wales: An
Evaluation. Home Office Research Study 133. London: Home
Office.
Marshall, G. (ed.) (1994) The Concise Oxford Dictionary of Sociology.
Oxford: Oxford University Press.
Marshall, P. (1994) Reconviction of imprisoned sexual offenders.
Research Bulletin, 36. London: Home Office Research and Statis-
tics Department.
Mattinson, J. and Mirlees-Black, C. (2000) Attitudes to Crime and
Criminal Justice: Findings from the 1998 British Crime Survey.
Home Office Research Study No. 200. London: Home Office.
Mayhew, P. (2000) Researching the state of crime: local, national
and international victim surveys. In R. King and E. Wincup,
Doing Research on Crime and Justice. Oxford: Oxford University
Press.
McLaughlin, E. (2001) Administrative criminology. In E.
McLaughlin and J. Muncie (eds) The Sage Dictionary of Criminol-
ogy. London: Sage.
McLaughlin, E. and Muncie, J. (eds) The Sage Dictionary of
Criminology. London: Sage.
Merton, R. K. (1967) On Theoretical Sociology. New York: Free Press.
Merton, R. K., Fiske, M. and Kendall, P. L. (1956) The Focused
Interview. New York: Free Press.
Miles, M. B. and Huberman, A. M. (1994) Qualitative Data
Analysis: An Expanded Sourcebook, 2nd edn. Thousand Oaks, CA:
Sage Publications.
Morgan, D. L. (1992) Designing focus group research. In M.
Stewart (ed.) Tools for Primary Care Research. Thousand
Oaks, CA: Sage.
Morgan, D. L. (1996) Focus groups. Annual Review of Sociology, 22:
129–52.

JOBNAME: Mcgraw−crowsemmens PAGE: 11 SESS: 18 OUTPUT: Wed Aug 29 11:32:10 2007 SUM: 4D58D784
References 281
Morgan, D. L. (2006) Focus groups. In V. Jupp (ed.) The Sage

Dictionary of Social Research Methods. London: Sage.
Morgan, D. L. and Krueger, R. A. (1993) When to use focus groups
and why. In D. L. Morgan (ed.) Successful Focus Groups: Advancing
the State of the Art. Newbury Park, CA: Sage.
Moyser, G. and Wagstaffe, M. (1987) Research Methods for Elite
Studies. London: Allen and Unwin.
Niven, S. and Olagundaye, J. (2002) Jobs and Homes: A Survey of
Prisoners Nearing Release. Home Office Research Findings 173.
Norris, C. (1993) Some ethical considerations on field-work with
the police. In D. Hobbs and T. May, Interpreting the Field:
Accounts of Ethnography. Oxford: Clarendon Press.
Norusis, M. J. (1992) SPSS for Windows: Base System User’s Guide.
Chicago: SPSS Inc.
O’Leary, Z. (2004) The Essential Guide to Doing Research. London:
Sage Publications.
Oppenheim, A. N. (2000) Questionnaire Design, Interviewing and
Attitude Measurement. London: Continuum.
Överlien, C., Aronsson, K. and Hydén, M. (2005) The focus group
interview as an in-depth method? Young women talking about
sexuality. International Journal of Social Research Methodology, 8(4):
331–44.
Pallant, J. (2001) SPSS Survival Manual. Buckingham: Open Uni-
versity Press.
Palmer, C. and Hart, M. (1996) A PACE in the Right Direction?
Sheffield: University of Sheffield, Institute for the Study of the
Legal Profession.
Patterson, A. and Thorpe, K. (2006) Public preparations. In A.
Walker, C. Kershaw and S. Nicholas, Crime in England and Wales
2005/06. London: Home Office.
Pawson, R. and Tilley, N. (1994) What works in evaluation
research? British Journal of Criminology, 34(3): 291–306.
Pawson, R. and Tilley, N. (1996) What’s crucial in evaluation
research: a reply to Bennett. British Journal of Criminology, 36(4):
574–8.
Pawson, R. and Tilley, N. (1997) Realistic Evaluation. London: Sage
Publications.

JOBNAME: Mcgraw−crowsemmens PAGE: 12 SESS: 18 OUTPUT: Wed Aug 29 11:32:10 2007 SUM: 509672CC
Pawson, R. and Tilley, N. (1998a) Caring communities, paradigm

polemics, design debates. Evaluation, 4(1): 73–90.
Pawson, R. and Tilley, N. (1998b) Cook-book methods and disas-
trous recipes: a rejoinder to Farrington. Evaluation, 4(2): 211–13.
Pearce, F. and Tombs, S. (1992) Realism and corporate Crime. In R.
Matthews and J. Young (eds) Issues in Realist Criminology. Lon-
don: Sage.
Phillpotts, G. J. O. and Lancucki, L. B. (1979) Previous Convictions,
Sentence and Reconviction. Home Office Research Study No. 53.
London: HMSO.
Pollard, J., Hawkins, J. and Arthur, W. (1999) Risk and protection:
are both necessary to understand diverse outcomes in adolescence?
Social Work Research, 23(3): 145–58.
Popper, K. (1959) The Logic of Scientific Discovery. London: Hutch-
inson.
Punch, M. (1979) Policing the Inner City: A Study of Amsterdam’s
Warmoesstraat. London: Macmillan.
Punch, K. (2006) Developing Effective Research Proposals, 2nd edn.
London: Sage.
Raynor, P. and Robinson, G. (2005) Rehabilitation, Crime and Justice.
Basingstoke: Palgrave.
Robinson, W. S. (1950) Ecological correlations and the behaviour of
individuals. American Sociological Review, 15: 351–7.
Robson, C. (2002) Real World Research, 2nd edn. Oxford: Blackwell
Publishing.
Rose, G. (1982) Deciphering Sociological Research. London: Macmil-
lan.
Rountree, P. W. (1998) A re-examination of the crime-fear linkage.
Journal of Research in Crime and Delinquency, 35(3): 341–72.
Rowntree, D. (1991) Statistics Without Tears: A Primer for Non-
mathematicians. London: Penguin.
Rutland, A. (2005) The development and regulation of prejudice in
children. ESRC Society Today. available at: http://
www.esrcsocietytoday.ac.uk/ESRCInfoCentre/index.aspx
Rutland, A., Cameron, L., Milne, A. and McGeorge (2005) Social
norms and self-presentation: children’s implicit and explicit inter-
group attitudes. Child Development, 76: 451–66.

JOBNAME: Mcgraw−crowsemmens PAGE: 13 SESS: 18 OUTPUT: Wed Aug 29 11:32:10 2007 SUM: 4F8A3F51
References 283
Sampson, R. and Raudenbush, S. (1999) Systematic social observa-

tion of public spaces: a new look at disorder in urban neighbor-
hoods. American Journal of Sociology, 105(3): 603–51.
Sandel, M. (1982) Liberalism and the Limits of Justice. Cambridge:
Cambridge University Press.
Schatzman, L. and Strauss, A. L. (1973) Field Research: Strategies for
a Natural Sociology. Englewood Cliffs, NJ: Prentice Hall.
Schuman, H. and Presser, S. (1996) Questions and Answers in
Attitude Surveys: Experiments on Question Form, Wording, and
Context. London: Sage.
Shapland, J., Wiles, P. and Wilcox, P. (1994) Targeted Crime
Reduction for Local Areas. London: Home Office, Police Research
Group.
Sherman, L. W., Gottfredson, D. C., MacKenzie, D. L., Eck, J.,
Reuter, P. and Bushway, S. D. (1998) Preventing Crime: What
Works, What Doesn’t, and What’s Promising. Washington, DC: U.S.
Department of Justice, National Institute of Justice.
Shrum, W. and Kilburn, J. (1996) Ritual disrobement at Mardi
Gras. Social Forces, 75(2): 423–58.
Skogan, W. G. (1987) The impact of victimisation on fear. Crime
and Delinquency, 33: 135–54.
Smellie, E. and Crow, I. (1991) Black Peoples’ Experiences of Criminal
Justice. London: NACRO.
Smith, G. and Cantley, C. (1984) Pluralistic evaluation. In J.
Lishman (ed.) Evaluation, Research Highlights in Social Work
26. London: Jessica Kingsley.
Smith, J. K. (1983) Quantitative versus qualitative research: an
attempt to clarify the issue. Educational Researcher, 12: 6–13.
Smith, J. K. and Heshusius, L. (1986) Closing down the conversa-
tion: the end of the quantitative-qualitative debate among educa-
tional enquirers. Educational Researcher, 15: 4–12.
Spalek, B. (1995) Trust as an ideology: exploring fear of white collar
crime. Paper presented to the British Criminology Conference.
Loughborough University, 18–21 July 1995.
Statistics Commission (2006) Crime Statistics: User Perspectives,
Report No. 30. London: Stationery Office.
Stinchcombe, A. L. (1968) Constructing Social Theories. New York:
Harcourt, Brace and World.

JOBNAME: Mcgraw−crowsemmens PAGE: 14 SESS: 18 OUTPUT: Wed Aug 29 11:32:10 2007 SUM: 4AD50105
Strauss, A. and Corbin, J.M. (1998) Basics of Qualitative Research:

Techniques and Procedures for Developing Grounded Theory. Thou-
sand Oaks, CA: Sage.
Survey Methods Centre (1994) Quota Versus Probability Sampling,
Report of the First Cathie Marsh Memorial Seminar, Newsletter,
15(1).
Tarling, R. (1982) Unemployment and crime. Research Bulletin No.
14. London: Home Office.
Tarling, R. and Weatheritt, M. (1979) Sentencing Practice in Magis-
trates’ Courts. Home Office Research Study No. 56. London:
HMSO.
Taylor, I., Walton, P. and Young, J. (1973) The New Criminology.
London: Routledge and Kegan Paul.
Thorpe, J. (1978) Social Enquiry Reports: A Survey. Home Office
Research Study No. 48. London: HMSO.
Thrasher, F. M. (1947) The Gang. Chicago: University of Chicago
Press.
Tomsen, S. (1997) A top night: social protest, masculinity and the
culture of drinking violence. British Journal of Criminology, 37:
90–102.
United Nations Social Defence Research Institute (1976) Economic
Crises and Crime. Rome: UNSDRI.
Utting, D. (1999) A Guide to Promising Approaches. York: Joseph
Rowntree Foundation.
Vold, G. B. and Snipes, J. B. (2001) Theoretical Criminology. New
York: Oxford University Press.
Walmsley, R., Howard, L. and White, S. (1992) The National Prison
Survey 1991: Main Findings. Home Office Research Study 128.
London: HMSO.
Weiss, C. H. (1998) Evaluation: Methods for Studying Programs and
Policies, 2nd edn. Englewood Cliffs, NJ: Prentice Hall.
Whyte, W. (1981) Street Corner Society: The Social Structure of an
Italian Slum. Chicago: University of Chicago Press.
Williams, M. (2006) Positivism. In V. Jupp (ed.) The Sage Dictionary
of Social Research Methods. London: Sage.
Williams, P. and Dickenson, J. (1993) Fear of crime: read all about
it? British Journal of Criminology, 33: 33–56.

JOBNAME: Mcgraw−crowsemmens PAGE: 15 SESS: 18 OUTPUT: Wed Aug 29 11:32:10 2007 SUM: 1210D7D3
References 285
Wolcott, H. F. (1994) Transforming Qualitative Data: Description,

Analysis and Interpretation. Thousand Oaks, CA: Sage Publica-
tions.
Wooffitt, R. (2005) Conversation Analysis and Discourse Analysis.
London: Sage.
Wright, D. B. (1997) Understanding Statistics: An Introduction for the
Social Sciences. London: Sage.


JOBNAME: Mcgraw−crowsemmens PAGE: 1 SESS: 10 OUTPUT: Thu Oct 4 16:18:43 2007 SUM: 76F41E84
/production/mcgraw−hill/booksxml/crowsemmens/z−index
Index
Locators shown in italics refer to boxes, figures and tables.
access, researcher attitudes, personal

decisions necessary when carrying research process investigating,
out observation, 105–7 144–9
accidental samples, 49 availability samples, 49
accuracy, sample, 46 Bail Act (1976), 201
administration, questionnaire, 149–51 Beauchamp, T., 52
aims and objectives, research Beck, U., 94
case study involving development Becker, H., 29
of, 244–5 behaviour, problematic
association between risk and
analysis, causal
problem behaviour, 226–8, 227,
as element of evaluation, 58–62, 58,
Bennett, T., 63–5
59, 60, 61
Bentham, J., 51
characteristics, 35–6, 35
bias, sample, 46
see also types eg experimental studies;
bivariate analysis, 164–71, 164, 165,
non-experimental studies;
166, 167, 169, 170, 175, 175, 259
quasi-experimental studies
Bogardus social distance scale of
analysis, data personal attitudes, 145
characteristics, process and role, British Cohort Study, The (1970), 88
16–17, 85–95, 153–89, 156, 158,
British Crime Survey, The, 10, 48, 91–3,
160, 162, 164, 165, 167, 169,
155, 237
170, 172, 173, 175, 178, 185,
British Household Panel Survey, 10
189
decisions necessary when carrying British Social Attitudes Survey, 88
out observation, 112–13 Bryman, A., 28–9, 108, 114, 184–5
guidelines for research publications burglary and burglaries
analysis, 97–9 Safer City programme statistics, 74
selection of methods of, 31–2 Cambridge Study of Delinquent
see also analysis types eg content Development, 39
analysis; conversation analysis; cards, financial transaction
discourse analysis; multivariate existence as influence on fear of
analysis; univariate analysis crime, 240–42
anti-positivism and positivism case studies
in relation to criminology research, Communities That Care UK,
25–6 226–35, 227, 230–31
influence of crime fears, 240–54,
anxieties fears and anxieties 248, 249, 251, 252
case study of crime fears influence, literature reviewing, 237–8
240–54, 248, 249, 251, 252 offender sentencing project,
definition of concept, 238–40 190–93, 200–210, 206, 207
Arber, S., 87 research aims and objectives, 244–5
Association of Payment and Clearing research design selection, 40–43,
Services (APACS), 240–41 41, 203–5, 205
Atkinson, P., 102, 109, 110 use of focus groups, 127–30
Kerrypress Ltd – Typeset in XML A Division: z-index F Sequential 1

JOBNAME: Mcgraw−crowsemmens PAGE: 2 SESS: 10 OUTPUT: Thu Oct 4 16:18:43 2007 SUM: 7FAB9336
case studies–contd content analysis

Youth Court Demonstration definition and characteristics, 95–6
Project, 213–24, 215 qualitative analysis of, 183–6, 185
cataloguing and storage content and sequence
of observation data, 112–13 of questionnaires, 135–7
centrality, measurement of, 158–9, 160 conversation analysis, 186–7
Chicago School of ethnography, 101–2 Corbin, J., 181
Children Act (1989), 213 correlation
Childress, J., 52 process of, 168–70, 169, 172–3,
Chui, W., 221 172, 173
Clear, T., 221 correlation studies, 37–8
closed questions covert research
role in questionnaire design, 141–4, ethical dilemmas within
142 criminological research, 52–3
cluster samples, 48
Cressey, D., 179
coding
of qualitative data, 177–8, 178 Criminal Justice Act (1991), 209, 211,
213
collection, data
characteristics and role, 16 Criminal Justice Act (1993), 209
decisions necessary when carrying Criminal Statistics, The, 90
out observation, 109–12 criminals and crime
selection of methods of, 28–9 case study of crime relationship
see also methods and sources eg with unemployment, 40–43, 41
information, abstracted; internet; case study of sentencing, 190–93,
interviews; observation and 200–210, 206, 207
observers; questionnaires; records, definition of concept, 238–40
case impact and case study of fear of
communities, safety of crime, 240–54, 248, 249, 251,
projects enhancing see 252
Communities That Care (CTC) reduction of as intervention
UK outcome, 71–3
Communities That Care (CTC) UK see also burglary and burglaries;
case study of, 226–35, 227, 230–31 white collar crime
history and characteristics of criminal justice, systems of
project, 225–6 case study of offender sentencing,
‘complete participant’ observation role, 190–93, 200–210
107–8 impact and relationship with
‘complete observer’ role, 109 concept of unemployment,
200–202
conclusions and findings, research recent policy developments, 211–12
interpretation and presentation,
188–9, 189 cross-sectional studies
writing up of, 123 case study of use of, 41–2
characteristics, 39
see also subject and name of studies of use
eg Communities That Care project; cross-tabulation, 164–7, 164, 165, 167
criminals and crime; sentencing, courts, youth
offender; Youth Court developmental pilot projects see
Demonstration Project Youth Court Demonstration
confidentiality Project
as ethical dilemma within data
criminological research, 53–4, 54 analysis see analysis, data
consent, informed collection see collection, data
as ethical dilemma within see also types eg qualitative data;
criminological research, 53 quantitative data

JOBNAME: Mcgraw−crowsemmens PAGE: 3 SESS: 10 OUTPUT: Thu Oct 4 16:18:43 2007 SUM: 82A12C6A
Index 289
deception evaluation, research–contd

as ethical dilemma within complexity of strategies and
criminological research, 53–4, processes, 73–5, 74
54, 55, 55 definition and characteristics, 57–8
Denzin, N., 11, 117–18, 122, 123 identification of aims and purpose,
Descartes, R., 24 67–9, 68
designs, research models of, 58–62, 58, 60, 61
case study of selection of, 40–43, of intervention/programme
41, 203–5, 205 integrity assessment, 69–70
definition and characteristics, 34–6, of outcomes criteria assessment,
35 70–73
types, 38–40 socio-political context, 75–7
see also evaluation, research; see also analysis, causal; feedback,
operationalization, research research; outcomes; process
Developing Effective Research Proposals evaluation
(Punch), 257 evaluation studies, 39–40
Dewey, J., 24 experimental studies, 36–7
discourse analysis, 187–8 Farrington, D., 63–5
dispersion, measurement of, 159–61, fears and anxieties
162 case study of crime fears influence,
distribution, measurement of, 193–5 240–54, 248, 249, 251, 252
Ditton, J., 108 definition of concept, 238–40
Drug Treatment and Testing Orders feedback, research
(DTTOs), 72–3 case study of use in Youth Court
Durkheim, D., 25 Demonstration Project, 219–20
Economic and Social Research see also evaluation, research
Council (ESRC), 52–3 filming
Economic and Social Data Service as observation method, 112
(ESDS), 86, 155 of focus groups, 127
empiricism, 23–4, 26–7 findings and conclusions, research
Engels, F., 201 interpretation and presentation,
epistemology 188–9, 189
role in research planning, 23–7 writing up of, 123
errors see also subject and name of studies of
sampling, 45–7 use eg Communities That Care
Type I and Type II, 264, 264 project; criminals and crime;
ESDS (Economic and Social Data sentencing, offender; Youth
Service), 86, 155 Court Demonstration Project
ESRC (Economic and Social Research focus groups
Council), 52–3 case studies of use of, 127–30
ethics and ethicality characteristics, limitations and
principles in relation to criminology process, 123–7
research, 50–55, 54, 55 fractions, sampling, 44
ethnography fragmentation
historical development and of qualitative data, 177–8, 178
characteristics, 101–103 frames, sampling, 44
qualitative analysis of, 183 fraud, financial
see also elements and decisions eg case study of impact on fear of
access, researcher; sampling and crime, 240–43
samples ‘freedom’, degrees of, 259–60
evaluation, research Freud, S., 25
and criminological theory debates, Friendship, C., 67, 68
62–7, 66 From Postgraduate to Social Scientist
case study of use in Youth Court (Gilbert), 257
Demonstration Project, 217 funding, criminology research, 256–7

JOBNAME: Mcgraw−crowsemmens PAGE: 4 SESS: 10 OUTPUT: Thu Oct 4 16:18:43 2007 SUM: 71809D06
Ganzini, L., 242 integrity, programme

General Household Survey, 88 assessment of as element within
Gilbert, N., 257 evaluation, 69–70
Glaser, B., 182 International Crime Victim Survey,
Goffman, E., 114 246
Gold, R., 107–9 internet
Grabosky, P., 243 as mode for questionnaire data
Green, A., 11 collection, 133
grounded theory interval variables, 158, 167–70, 169,
approach to qualitative analysis, 193–5
181–2 interventions
characteristics, 32 as purpose and outcome of
groups, focus criminological evaluation, 67–9,
case studies of use of, 127–30 68
characteristics, limitations and interviews
process, 123–7 case study of use in offender
Guttman scales of personal attitudes, sentencing research, 205–8, 206,
146 207
Hammersley, M., 102, 109, 110 characteristics and process, 117–23
Hargreaves, D., 102 see also focus groups
hypotheses interviews, telephone
building of in case study of crime as mode for questionnaire data
fear, 245–6, 248–51, 248, 249, collection, 132–3, 132
251 Israel, M., 221
case study of development of judgemental samples, 49
concerning offender sentencing, justice, criminal see criminal justice,
202–3 systems of
characteristics and role in Kant, I., 24, 51–2
criminology research, 13–15 Kelle, U., 14
development of as element of Kemp, C., 113
research methods, 29
Labour Force Survey, 88
testing of, 260–61
see also null-hypothesis language
qualitative analysis of, 186–8
Heritage, J., 186
Hill, J., 234 Levi, M., 242
Holdaway, S., 106, 107–8 Lickert scales of personal attitudes,
Hollister, R., 234 147–9, 149
Holsti, O., 184 Lindesmith, A., 122–3
Hope, T., 94 Loftland, L., 109
Huberman, A., 182 logistic regression, 173
induction, analytic Lombroso, C., 25
as applied to qualitative analysis, longitudinal studies
179–80 case study of use of, 42–3
characteristics, 122–3 characteristics, 39
inferences Madriz, E., 129–30
applicability to samples, 45–7 Maguire, M., 92
characteristics and role, 17, 153–5 Maryland Scale of Scientific Methods,
from single quantitative variables, 60–62, 60, 61
161, 163 measurement (concept)
information, abstracted definition, 261
as mode for questionnaire data see also element measured eg
collection, 134, 154–5 centrality; dispersion;
inquiry, research distribution
types, 9–12, 10, 12 Merton, R., 124

JOBNAME: Mcgraw−crowsemmens PAGE: 5 SESS: 10 OUTPUT: Thu Oct 4 16:18:43 2007 SUM: 803530FB
Index 291
methods, research observation and observers–contd

definition and characteristics, 34–6, case study of use in Youth Court
35 Demonstration Project research,
future development pathway, 255–8 214–15, 216, 222–4
selection and development of, historical development and
28–31 characteristics, 100–103
see also elements and outcomes eg aims role of observer, 107–9
and objectives, research; types and models, 110–12
evaluation, research; strategies, weaknesses of as method, 114–15
research see also process elements and decisions
see also type eg interviews; eg access, researcher; analysis, data;
observation and observers; collection, data; sampling and
questionnaires samples; storage and cataloguing
Miles, M., 182 ‘observer-as-participant’ role, 108–9
Mill, J.S., 51 offenders and offending see criminals
Millennium Cohort Study, The, 88 and crime
‘mixed method’ research, 11 O’Leary, Z., 82, 84
models ‘open’ observation, 110–11
of evaluation, 58–62, 58, 59, 60, 61, open questions
67–9 qualitative analysis of, 180–81
of research process, 12 role in questionnaire design, 141–4,
moderators 142
role in focus groups, 126 operationalization, research
Morgan, D., 126 characteristics and role in
multistage samples, 48 criminology research, 15–16
multivariate analysis, 171–5, 172, 173, experience of use in case study of
175, 259 offender sentencing, 206–8, 207
National Child Development Study involvement in case study of crime
(NCDS), 39, 87–8 fear, 246–7
National Crime Victims Survey (USA), Oppenheim, A., 144
237 opportunities, criminology research
neighbourhoods, safety of future outlook, 255–6
projects enhancing see ordinal variables, 157–8 170–71, 170
Communities That Care (CTC) Osgood semantic differential scales
UK of personal attitudes, 146[check
Newburn, T., 94 exact tit]
New Criminology, The (Taylor), 93 outcomes
nominal variables, 157, 164–7, 164, assessment of intervention
265, 167 outcomes as element within
No More Excuses: a new approach to evaluation, 70–73
Tackling Youth Crime (1997), 211 case study of use of outcomes
non-experimental studies, 37–8 evaluation, 228–31, 230–31
non-probability sampling, 30 Överlein, C., 127
non-random samples, 49–50 OXO model of causality evaluation,
non-schedule-standardized interviews, 59, 62–3
118 packages, computer see name eg
non-standardized interviews, 118 Statistical Package for the Social
null hypothesis, 262 Sciences
objectives and aims, research ‘paradigm wars’ concerning evaluation
case study involving development theories, 62–7, 66
of, 244–5 partial correlation, 172–3, 172, 173
observation and observers ‘participant-as-observer’ role, 108
case study of use in offender participants, research
sentencing research, 205–8, 206, unintended ‘manipulation’ of
207 researcher by, 116–17

JOBNAME: Mcgraw−crowsemmens PAGE: 6 SESS: 10 OUTPUT: Thu Oct 4 16:18:43 2007 SUM: 77DA2941
Pawson, R., 62–7 questionnaires–contd

Pearce, F., 242 modes of data collection, 132–4,
Peirce, C., 24 132, 154–5
pilot studies, questionnaire, 151 questions, research
Pithouse, P., 242 characteristics and importance,
planning, research 20–22
role of epistemology, 23–7 design for specific case studies,
politics 228–30, 237–43, 246–7
relationship with criminological design of in questionnaire surveys,
research process, 93–5 137–8
populations see also closed questions; open
definition of, 262 questions
positivism and anti-positivism see also processes and features
in relation to criminology research, involving eg hypotheses;
25–6 operationalization, research;
theories
pragmatism, 24–5, 26–7
quota samples, 49
presentation, questionnaire, 149–51
random samples, 47–8
Preston, J., 11 rates, response
primary data analysis in questionnaire surveys, 134–5
characteristics, 154–5 rationalism, 24, 26–7
probability (concept) Raudenbush, S., 109
definition, 262–3 Raynor, P., 94
samples and sampling, 30, 47–8 realism
process evaluation in relation to criminology research,
case study of use of, 228–31, 26
230–31 Realistic Evaluation (Pawson and
Promising Approaches (Utting), 226 Tilley), 65
prospective studies, 38–9 reasoning
publications, research inductive and deductive, 31
guidelines for analysis of, 97–9 Recorded Crime Statistics, The, 90–91,
Punch, K., 257 93
qualitative data recording, tape and video
case study of use in Youth Court as observation method, 112
Demonstration Project research, of focus groups, 127
217–19 see also writing up
process of analysis, 175–89, 178, records, case
185, 189 case study of use in offender
quantitative data sentencing research, 205–8, 206,
case study of use in Youth Court 207
Demonstration Project research, regression, 173
217–19 reliability
process of analysis, 155–75, 156, 158, in qualitative research, 189
160, 162, 164, 165, 167, 169, 170, relevance in research, 20
172, 173, 175 representativeness
quasi-experimental studies, 37–8 of samples, 45–6
questionnaires see also validity
case study of use in Communities research and researchers, criminology
That Care project, 230–31, 231 definition, characteristics and
case study of use in Youth Court process of, 7–17, 10, 12
Demonstration Project research, ethical dilemmas within, 50–55, 54,
216–7 55
characteristics, limitations and future development pathway, 255–8
design, 130–51, 132, 139, 140, manipulation by participants,
142, 149 116–17

JOBNAME: Mcgraw−crowsemmens PAGE: 7 SESS: 10 OUTPUT: Thu Oct 4 16:18:43 2007 SUM: 825C5912
Index 293
research and researchers, sampling and sample–contd

criminology–contd selection of methods of, 30
relationship with political agendas types, 47–50
and expediency, 93–5 see also elements and outcomes eg
role of researchers, 30–31 accuracy, samples; errors,
see also methods, research; planning, sampling; inferences;
research; results, research; skills, representativeness
research Sampson, R., 109
see also approaches to eg positivism scales, personal attitude, 145–9
and antipositivism; realism schedules, interview, 120
see also elements of process eg analysis, schedule-standardized interviews, 118
data; collection, data; induction, searches and searching, literature
analytic; operationalization, sources and processes, 82–5
research; questions, research secondary data
respondents case study of use in offender
unintended ‘manipulation’ of sentencing research, 205–8, 206,
researcher by, 116–17 207, 240–54, 248, 249, 251, 252
responses definition, characteristics and
options and order within process of analysis, 85–95,
questionnaire surveys, 138–41, 154–5
139, 140 self-completion questionnaires, 133–4
rates of in questionnaire surveys, semi-structured interviews, 118–23
134–5 sentencing, offender
results, research case study of, 190–93, 200–210,
interpretation and presentation, 206, 207
188–9, 189 services, postal
writing up of, 123 as mode for questionnaire data
see also subjects and name of studies collection, 133
eg Communities That Care project; Shapland, J., 111
criminals and crime; sentencing, Sherman, L., 60
offender; Youth Court significance tests, 263
Demonstration Project simple random samples, 47
retrospective studies, 38–9 skills, research
reviews, literature importance and need for, 257–8
case study of use of, 237–8 snowball samples, 49
sources and processes, 82–5 Spalek, B., 243
risks Statistical Package for the Social
association between risk factors and Sciences (SPSS), 156–7, 156
problem behaviour, 226–8, 227 statistics (as data) see quantitative data
case study of, involving statistics, official
Communities That Care, of social and crime trends, 89–93
228–35, 230–31 Statute of Labourers (1349), 200
Robinson, G., 94 storage and cataloguing
Robson, C., 21, 108, 110–11, 182 of observation data, 112–13
Rowntree Foundation, 226 strategies, research
Safer Cities Programme, 73–4 definition and characteristics, 34–6,
safety, neighbourhood 35
projects enhancing see selection of, 22–7
Communities That Care (CTC) see also methods, research
UK see also aims and objectives,
sampling and samples research; outcomes
decisions necessary when stratified random samples, 47–8
embracing observation, 103–4 Strauss, A., 181, 182
definition and characteristics, 43–5, Straw, J., 211–12
44, 45, 262 structured observation, 111–12

JOBNAME: Mcgraw−crowsemmens PAGE: 8 SESS: 10 OUTPUT: Thu Oct 4 16:18:43 2007 SUM: 5DAA205A
studies (research method) see name eg unemployment

experimental studies; longitudinal case study of unemployment
studies; prospective studies relationship with crime, 40–43,
41
surveys impact and relationship with
crime and social trends, 87–8 concept and reality of criminal
questionnaire see questionnaires justice systems, 200–202
systems, postal univariate analysis, 157–63, 175, 158,
as mode for questionnaire data 160, 162, 175(tab), 259
collection, 133 unstructured observation, 110–11
tape recording Utting, D., 226
Vagrancy Acts (1824–35), 200
as observation method, 112
validity
of focus groups, 127 in qualitative research, 189
Taylor, I., 93 relevance in research, 20
tests and testing see also representativeness
of significance, 263 variables
theories definition and characteristics, 157,
characteristics and role in 164–71, 164, 165 167, 169, 170,
criminology research, 12–13, 22 264
concerning evaluation-criminology inferences from, 161, 163
theory debates, 62–7, 66 video recording
as observation method, 112
Thurstone scales of personal attitudes,
of focus groups, 127
147 Weiss, C., 66–7
Tilley, N., 62–7 ‘What works?’ agenda, 74–5
time series analysis white collar crime
case study of use of, 42 impact on fear of crime, 240–43
Tombs, S., 242 Whyte, W., 105
Tomsen, S., 102, 178 Wolcott, H., 183
tools, research see methods, research worklessness see unemployment
trends, crime, 89–93 writing up
triangulation, 11, 29 of findings, 123
Youth Court Demonstration Project
Type I and Type II errors, 264, 264
(YCDP)
UK Data Archive, 85–6 case study of, 213–24, 215, 216
UK Safer Cities Programme, 73–4 history of, 211–12

Researching Criminology
R e s e a rc h i n g C r i m i n o l o g y
“research
This book provides an essential tool for undergraduate students embarking upon their own
projects in Criminology. It provides clear and informative guidance on a range of
research methods and designs to assist students in their own criminological endeavours.”
D J
R T
ACKI ,
APLEY SENIOR LECTURER IN CRIMINOLOGY, INSTITUTE OF CRIMINAL JUSTICE STUDIES, UNIVERSITY OF PORTSMOUTH
“institutions
…what makes the book stand out is the inclusion of real research into various criminal justice
that have actually been undertaken by the authors. In doing so, what is produced is a
book that offers research 'know how', stimulates interest and interjects research passion into
what can often be a difficult, and sometimes dry area of research. ”
DR TINA PATEL, LECTURER IN CRIMINOLOGY AND SOCIOLOGY, LIVERPOOL JOHN MOORES UNIVERSITY
❙ How do criminologists go about studying crime and its consequences?

❙ How are programmes for offenders and communities evaluated?
❙ How can you collect and analyse criminological material?
Research on crime and criminality is often referred to by the media, policy makers and
practitioners, but where does this research come from and how reliable is it?
Designed especially for students on criminology and criminal justice courses, and professionals
working in the field, Researching Criminology emphasises the importance of research as an
integrated process. It looks at the ways in which a mixture of investigative methods can be used
to analyze a criminological question.
Written by two experienced researchers and lecturers, Researching Criminology is a

comprehensive introduction to the aims, principles and methods of doing criminological
research. The book covers all the key topics that you will encounter when researching crime.
Individual chapters include:
❙ The research process ❙ Evaluation research

❙ Principles of researching criminology
❙ How to design criminological research
❙ Researching ethically
❙ A glossary of essential key concepts
Researching
Crow and Semmens
Structured in three parts, addressing the principles of criminological research, how to collect and
analyse material and providing detailed examples of real world research, Researching
Criminology
Criminology will be of benefit to all students of criminology and criminal justice, for practitioners
interested in criminological research, and for those undertaking criminological research for the
first time.
Iain Crow is Reader in Research Methods at the University of Sheffield, UK. His teaching covers
research methods and statistics, and research ethics. He has undertaken research on a wide
range of topics concerned with the treatment and rehabilitation of offenders and crime reduction.
Natasha Semmens is a lecturer in Criminology at the University of Sheffield, UK. Her research
interests include the fear of crime, crime survey methodology, white collar crime and cybercrime.
Her teaching includes research methods, quantitative
analysis, and white collar crime.
Cover design Hybert Design • www.hybertdesign.com

www.openup.co.uk

Researching Criminology

Uploaded by

Copyright:

Available Formats

Researching Criminology

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Researching Criminology

Uploaded by

Copyright:

Available Formats

What are the main topics covered in the book?

What are the main topics covered in the book?

How does the book help students and practitioners with criminological research?

How does the book help students and practitioners with criminological research?

Researching Criminology

❙ How do criminologists go about studying crime and its consequences?

Written by two experienced researchers and lecturers, Researching Criminology is a

❙ The research process ❙ Evaluation research

Cover design Hybert Design • www.hybertdesign.com

Kerrypress Ltd – Typeset in XML A Division: aplm_01-halftitle F Sequential 1

Kerrypress Ltd – Typeset in XML A Division: aplm_01-halftitle F Sequential 2

Iain Crow and

Open University Press

Kerrypress Ltd – Typeset in XML A Division: aplm_02-title F Sequential 1

Open University Press

and Two Penn Plaza, New York, NY 10121—2289, USA

First published 2006

Copyright © Author’s name 2006

A catalogue record of this book is available from the British Library

ISBN-13: 978-0-33-521546-1 (pb) 978-0-33-521547-8 (hb)

Library of Congress Cataloging-in-Publication Data

Typeset by Kerrypress, Luton, Bedfordshire

Kerrypress Ltd – Typeset in XML A Division: aplm_03-imprint F Sequential 1

www.kerrypress.co.uk - 01582 451331 - www.xpp-web-services.co.uk

About the authors

Part I: The principles of criminological research

Kerrypress Ltd – Typeset in XML A Division: aplm_03-contents F Sequential 1

Part II: Collecting and analysing material

Part III: Real-world research

Kerrypress Ltd – Typeset in XML A Division: aplm_03-contents F Sequential 2

Kerrypress Ltd – Typeset in XML A Division: aplm_03-contents F Sequential 3

Kerrypress Ltd – Typeset in XML A Division: aplm_03-contents F Sequential 4

About the authors

Iain Crow is a reader in Research Methods at the University of

Kerrypress Ltd – Typeset in XML A Division: aplm_04-aboutauthors F Sequential 1

Kerrypress Ltd – Typeset in XML A Division: aplm_04-aboutauthors F Sequential 2

Inevitably such a book reflects our own approach to criminological

Kerrypress Ltd – Typeset in XML A Division: aplm_05-acknowls F Sequential 1

Chapter 11 explains how a Joseph Rowntree Foundation-funded

Iain Crow and Natasha Semmens

Kerrypress Ltd – Typeset in XML A Division: aplm_05-acknowls F Sequential 2

Almost the first question we asked ourselves before embarking on

Kerrypress Ltd – Typeset in XML A Division: aplm_06-introduction F Sequential 1

www.kerrypress.co.uk - 01582 451331 - www.xpp-web-services.co.uk

the process, and many of the topics mentioned in the opening

Kerrypress Ltd – Typeset in XML A Division: aplm_06-introduction F Sequential 2

www.kerrypress.co.uk - 01582 451331 - www.xpp-web-services.co.uk

The first study was conducted at a time of high unemployment,

Kerrypress Ltd – Typeset in XML A Division: aplm_06-introduction F Sequential 3

www.kerrypress.co.uk - 01582 451331 - www.xpp-web-services.co.uk

Kerrypress Ltd – Typeset in XML A Division: aplm_06-introduction F Sequential 4

www.kerrypress.co.uk - 01582 451331 - www.xpp-web-services.co.uk

Kerrypress Ltd – Typeset in XML A Division: chap01 F Sequential 1

www.kerrypress.co.uk - 01582 451331 - www.xpp-web-services.co.uk

Kerrypress Ltd – Typeset in XML A Division: chap01 F Sequential 2

www.kerrypress.co.uk - 01582 451331 - www.xpp-web-services.co.uk

1 The research process

The term ‘research’ is used to refer to a wide range of activities.

Kerrypress Ltd – Typeset in XML A Division: chap01 F Sequential 3

www.kerrypress.co.uk - 01582 451331 - www.xpp-web-services.co.uk