University of South Florida

Scholar Commons
Graduate Theses and Dissertations Graduate School

2011

General Strain Theory and Juvenile Delinquency:

A Cross-Cultural Study
Wen-Hsu Lin
University of South Florida, [email protected]

Follow this and additional works at: http://scholarcommons.usf.edu/etd

Part of the American Studies Commons, Criminology and Criminal Justice Commons, Social
Psychology Commons, and the Sociology Commons

Scholar Commons Citation

Lin, Wen-Hsu, "General Strain Theory and Juvenile Delinquency: A Cross-Cultural Study" (2011). Graduate Theses and Dissertations.
http://scholarcommons.usf.edu/etd/3208

This Dissertation is brought to you for free and open access by the Graduate School at Scholar Commons. It has been accepted for inclusion in
Graduate Theses and Dissertations by an authorized administrator of Scholar Commons. For more information, please contact
[email protected].
 General Strain Theory and Juvenile Delinquency: A Cross-Cultural Study

by

Wen-Hsu Lin

A dissertation submitted in partial fulfillment

of the requirements for the degree of
Doctor of Philosophy
Department of Criminology
College of Behavioral & Community Sciences
University of South Florida

Co-Major Professor: Richard Dembo, Ph.D.

Co-Major Professor: Christine S. Sellers, Ph.D.
John K. Cochran, Ph.D.
Thomas Mieczkowski, Ph.D.

Date of Approval:
June 1, 2011

Keyword: strain, negative emotions, cross-cultural comparison, criminological theory,

criminology

©Copyright 2011, Wen-Hsu Lin

 ACKOWLEDGEMENTS

I feel blessed by God who brought many people into my life to help and support my

dissertation research. I thank my wife, Dr. Chen, for her dedication, without which I could not

complete my goals. Of course, the gratitude should be extended to my parents who continually

provide both financial and emotional support.

I acknowledge the hard work of my committee: Dr. Richard Dembo, Dr. Chris Sellers, Dr.

John Cochran, and Dr. Thomas Mieczkowski. I thank Dr. Dembo, my dissertation committee

chair, for helping me learn and adhere to a high standard of scholarship, and his support and

mentorship throughout my graduate education.

I thank Dr. Sellers for enlightening me in theory and research, and for her extra guidance

and help in exploring ideas that are reflected in this dissertation. I thank Dr. Cochran for his

guidance and scholarly advice on my research practice. I thank Dr. Mieczkowski for encouraging

me throughout my graduate education and agreeing to work with me. Each member contributed

to my development as a scholar, and I would not have been able to complete this project without

their individual effort and support. Special thanks should be given to Dr. James Schmeidler who

sacrificially gave sound advice on guiding my statistical journey.

 TABLE OF CONTENTS

LIST OF TABLES .......................................................................................................................... iii

LIST OF FIGURES ......................................................................................................................... v

ABSTRACT .............................................................................................................................. vi

CHPATER I: INTRODUCTION ..................................................................................................... 1

Purpose of the present study ............................................................................................... 3
Dissertation outline ............................................................................................................. 5

CHAPTER II: STRAIN THEORY AS AN EXPLANATION OF

DELINQUENCY ...................................................................................................................... 6
Review of classic strain theory ........................................................................................... 7
Durkheim‘s anomie theory .................................................................................... 7
Merton‘s anomie theory ....................................................................................... 12
Cohen‘s delinquent subculture theory ................................................................. 18
Cloward and Ohlin‘s theory of differential opportunity and delinquent
subculture..................................................................................................... 22
Criticisms of the classical strain theory ............................................................................ 26
Modern strain theory and Agnew‘s general strain theory ................................................. 30
Modern strain theory............................................................................................ 30
Agnew‘s general strain theory ............................................................................. 32
Strain ....................................................................................................... 32
Negative emotions .................................................................................. 35
Coping strategies .................................................................................... 36
Empirical assessment of general strain theory .................................................................. 38
Strain-delinquency relationship ........................................................................... 39
The mediating effect of anger .............................................................................. 42
Agnew‘s revision of GST ................................................................................................. 44
Empirical assessment of the revised GTS ......................................................................... 46
Characteristics of strain and delinquency ............................................................ 46
Other negative emotions ...................................................................................... 48
Summary and general limitations of previous studies ......................................... 50

CHAPTER III: CULTURE, DELINQUENCY, AND GENERAL STRAIN THEORY .............. 55

Culture and its impact on the stress/strain process ........................................................... 56
Dimensions of culture-individualism and collectivism ....................................... 56
Confucian ideology and its influence .................................................................. 62
The impact of culture on stress/strain .................................................................. 66

i
 The impact of culture on negative emotions ........................................................ 70
The impact of culture on coping strategies .......................................................... 72
Juvenile delinquency in Taiwan........................................................................................ 75
Taiwan-an overview ............................................................................................ 75
Juvenile law in Taiwan ........................................................................................ 76
Juvenile delinquency in Taiwan .......................................................................... 77
GST in other non-western countries and Taiwan.............................................................. 87
Cross-national study ............................................................................................ 87
Applying GTS in non-western countries ............................................................. 89
GST in Taiwan ..................................................................................................... 93

CHAPTER IV: RESEARCH QUESTIONS .................................................................................. 97

CHPATER V: METHODS ............................................................................................................ 99

Sample ............................................................................................................................. 99
U.S. sample .......................................................................................................... 99
Taiwanese sample .............................................................................................. 100
Survey preparation .......................................................................................................... 106
Measurement of variables ............................................................................................... 107
Delinquency ....................................................................................................... 108
Physical aggression ............................................................................................ 110
Strain .................................................................................................................. 110
Failure to achieve positively valued goals-disjunction between
Desired and actual outcome .......................................................... 114
Failure to achieve positively valued goals-unjust outcomes ................. 114
Loss of positive stimuli-negative life-events ........................................ 115
Presentation of noxious stimuli-victimization ...................................... 115
Negative emotion ............................................................................................... 116
Demographic variables ...................................................................................... 119
Analytic strategy ............................................................................................................. 122
Statistical method............................................................................................... 122
Missing data ....................................................................................................... 126

CHPATER VI: RESULTS........................................................................................................... 134

GST in the U.S. ............................................................................................................... 135
GST in Taiwan ................................................................................................................ 147
Multiple group analysis .................................................................................................. 155

CHAPTER VII: CONCLUSION AND DISCUSSION ............................................................... 171

Summary of findings ...................................................................................................... 172
Discussion of the findings ............................................................................................... 178
Strain and its characteristics .............................................................................. 178
Negative emotions ............................................................................................. 182
Similarities and differences in GST across culture ............................................ 188
Limitations of the present study and future study ........................................................... 193

REFERENCES ........................................................................................................................... 197

ii
APPENDICES ........................................................................................................................... 227
Appendix A: Survey questionnaire ................................................................................. 228
Appendix B: EFA tables ................................................................................................. 230

iii
 LIST OF TABLES

TABLE 1 Juvenile Offender/Suspect Situation from 2004 to 2009 ..................................... 78

TABLE 2 Descriptive Statistics of Strain and Negative Emotion Variables...................... 120

TABLE 3 Frequency Distribution of Demographic Variables, Aggression,

and Delinquency ................................................................................................ 121

TABLE 4 The Relationship between all Strain Variables, Delinquency and

Aggression in the U.S. ....................................................................................... 138

TABLE 5 The Full GST Model in the U.S. ........................................................................ 141

TABLE 6 The Indirect Effects of Strains on Outcome Variables through Anger and
Depression in the U.S. ....................................................................................... 144

TABLE 7 The Direct, Indirect, and Total Effects in the U.S. ............................................ 146

TABLE 8 The Relationship between all Strain Variables, Delinquency and

Aggression in Taiwan ........................................................................................ 148

TABLE 9 The Full GST Model in Taiwan ......................................................................... 151

TABLE 10 The Indirect Effects of Strains on Outcome Variables through Anger and
Depression in Taiwan ........................................................................................ 152

TABLE 11 The Direct, Indirect, and Total Effects in Taiwan ............................................. 154

TABLE 12 The Full GST Model-Multiple Group Analysis for Delinquent Acts ................ 157

TABLE 13 The Full GST Model-Multiple Group Analysis for Aggression ........................ 158

TABLE 14 The Full GST Model-Multiple Group Analysis with Constrains

(Damaging Property) ......................................................................................... 160

TABLE 15 The Full GST Model-Multiple Group Analysis with Constrains (Hitting
Someone) ........................................................................................................... 161

iv
TABLE 16 The Full GST Model-Multiple Group Analysis with Constrains (Alcohol
Use) .................................................................................................................... 165
TABLE 17 The Full GST Model-Multiple Group Analysis with Constrains
(Aggression) ...................................................................................................... 167

TABLE 18 The Wald test for Indirect Effect ........................................................................ 169

TABLE 19 Summary for the Basic GST model in the U.S. and Taiwan............................... 174

TABLE 20 Summary for the Full GST model in the U.S. and Taiwan ................................. 177

TABLE 21 Summary for the Indirect Effect of Anger and Depression in the U.S. and
Taiwan ................................................................................................................. 177

TABLE 22 Summary for the Tested Similarities and Differences in the Full
GST Model between the U.S. and Taiwan ........................................................... 180

v
 LIST OF FIGURES

FIGURE 1. The GST Path Model ......................................................................................... 122

FIGURE 2. The Iteration History Plot of Mean and Standard Deviation for the U.S.
sample ................................................................................................................ 132

FIGURE 3. The Iteration History Plot of Mean and Standard Deviation for the
Taiwanese sample .............................................................................................. 133

FIGURE 4. The Path-Analytic Model of Strain and Negative emotions in

the U.S................................................................................................................ 140

FIGURE 5. The Path-Analytic Model of Strain and Negative emotions in

the Taiwan.......................................................................................................... 149

vi
 ABSTRACT

General strain theory (GST) (Agnew, 1992, 2001, 2006a) is an established criminological

theory. Although the theory has been examined by many and enjoys empirical support, some

limitations of previous studies need to be addressed. First, previous studies have not incorporated

all major types of strain in their models; hence, the effects of these strains on delinquency are

unclear. Second, many previous studies did not include negative emotions and even negative

emotions other than anger. Finally, and the most serious limitation, many previous studies rely

heavily on samples from Western countries, mostly the U.S.; thus, possible cultural influences are

ignored. Although a few studies have moved forward by using subjects from Asia (e.g., China,

Korea), these studies only provide empirical results regarding whether GST is applicable in other

cultures. The lack of comparable samples from both Western and Eastern cultures hinders direct

comparison.

The present research contributes to the theoretical body of literature through addressing

the aforementioned three limitations. First, the study measures the major types of strain that are

mentioned by Agnew. Second, anger and depression are included in the analysis, which addresses

not only the limitations of previous studies but also the suggestions of Agnew (2006a). In

addition, the measure of anger is situational and consistent with GST. Thirdly, the present study

uses the same research instrument to collect comparable samples from both the U.S. (Western

country) and Taiwan (Eastern country). This enables a direct comparison across cultural

boundaries, and the similarities and differences can be empirically established.

vii
 Whereas the core propositions of GST are supported, the study finds some negative

results. In addition, most of the GST processes are found to be similar between the U.S. and

Taiwanese juveniles. However, some differences were also discovered. Explanation of these

similarities and differences from their cultural perspectives are offered. Furthermore, the results

from this study also raise some challenges to GST and point out that revisions of GST may be

required.

vii
 CHAPTER I:

INTRODUCTION

Psychologists and sociologists often refer to the period of adolescence as a time of

storm and turmoil. One must understand that the connotation of ―storm and turmoil‖ not

only points out the high risks involved in various antisocial behaviors during this period

(Goffredson & Hirschi, 1990; Moffitt, 1993), but also refers to the increasing stress and

the levels of negative emotions that occur during puberty. Studies from the stress

literature document that the juvenile period is fraught with struggles, distress, and

negative emotions (Agnew, 1997; Compas & Wagner, 1991; Larson & Asmussen, 1991).

Gore and Colten (1991, p.1) state ―[T]he concept of stress is an important tool for

organizing research seeking to understand development during the adolescent years.‖

DuRant and colleagues (1995, p.233) also suggest that ―[L]ife stress can have a

deleterious impact on the psychological adjustment of adolescents,‖ and the impact of

such stress has been related to various negative outcomes including delinquency (Vaux &

Ruggiero, 1983). Brandt (2006, p.58) concluded that ―the increase and decrease in

antisocial behavior are linked with increases and decreases in the levels of developmental

stress associated with adolescence.‖

Studying the effects of stress or strain on delinquency during the adolescent years

is important for two reasons. First, empirical studies have shown that there is a

1
relationship between strain and juvenile delinquency (Agnew, 2006a; Drapela, 2006;

Seiffge-Krenke, 2000; Sigfusdottir, Farkas, & Silver, 2004). Second, delinquency not

only causes immediate problems to juveniles (e.g., increasing victimization) (Kennedy &

Baron, 1993) but also increases the risk of later life maladjustment (Moffitt, 1993;

Sampson & Laub, 1993). In fact, scholars have found that delinquency during the

adolescent years is a risk factor for later criminal involvement and negative life

consequences (Elliot, 1994; Farrington, 1989; Nagin & Paternoster, 1991; Moffitt, 1993;

Sampson & Laub, 1993; Tolan & Tomas, 1995).

Agnew‘s (1992, 2001, 2006a, 2006b) general strain theory (GST) is not only an

important criminological theory (Cullen, Wright, & Blevins, 2006) but also used by many

scholars to examine the strain/deviance relationship. This theory refines key concepts of

classic strain theory (Cohen, 1955; Cloward & Ohlin, 1960; Merton, 1938), and provides

a rich framework for analyzing the underlying mechanisms that connect strain, negative

emotions, and delinquency in adolescents.

Although GST seems to be an important theory in explaining juvenile

delinquency, some mixed empirical evidence suggests that GST can still be improved.

Agnew (2001, 2006a, 2006b) revised GST from its original version in order to respond to

this mixed evidence and criticisms. While examining the revised GST seems to be an

important next step, without systematic examination of the basic and fundamental GST

model seems premature and unwise.

2
 Furthermore, extant tests of GST have relied almost exclusively on samples

drawn from the U.S. Froggio (2007) raise a question regarding the utility of GST in

explaining juvenile deviance in other countries. So far, only a few studies have applied

GST in non-Western cultural settings (e.g., China, Korea, Philippine). Cultural attitudes

and values influence one‘s definition of events and conditions as either typical or

stressful (Chun, Moos, & Cronkite, 2005). Hence, a strain in the U.S. may not be seen as

stressful in other cultures, which may not lead to subsequent negative emotions and

delinquency. This raises some questions regarding the generality of GST. A single study

that compares the similarities and differences in the GST process across nations is

virtually non-existent (see Botchkovar, Tittle, & Antonaccio, 2009 for an exception).

This is unfortunate because without comparative studies, both the generality of GST and

its cross-cultural validity are questionable. Moreover, scholars have argued that cross-

cultural studies could help to refine a theory so that such a theory is able to accommodate

cultural differences (Kim, Triandis, Choi, & Yoon, 1994; Kohn, 1987). For example,

Adler (1996) has argued that globalization offers a great opportunity to test and develop

criminological theory. Moreover, Karstedt (2001) indicated that comparative studies

―offer new insights, fresh theories and chances of innovative perspectives‖ (p.285).

Purpose of the present study

The present study addresses three gaps in the literature on GST. First, the present

study will examine the effects of the three major types of strain on delinquency and

negative emotions. Specifically, this study will measure the four different strains which

cover all three major types of strain. Second, while most previous studies focus on anger,
3
the present study will include anger as well as other negative emotions, namely,

depression as intervening variables in the strain-delinquency pathway. Moreover, the

present study will examine the basic GST model by using a path analytic approach,

which is useful in examining theoretically specified causal models and in exploring

mediating effects.

Thirdly, most previous empirical studies use almost exclusively Western samples

(e.g., American, Canadian). Only a few studies bring GST into Eastern cultural settings.

A study that compares the GST process across Western and Eastern cultures is non-

existent. The present study will fill this void by comparing the GST model as it operates

in both the U.S. and Taiwan. This should either further establish the general scope of

GST or demonstrate the need to further revise the theory to account for cross-cultural

differences in strain, negative affect, and illegitimate coping mechanisms. Moreover, path

analysis is able to implement multiple group analysis, a statistical method capable of

comparing and contrasting theoretical models directly across different populations and

providing statistically sound tests.

In sum, the present study is among the first that directly compares and contrasts

the GST process between Western (U.S.) and Chinese cultures (Taiwan) by using

comparable adolescent students from both the U.S. and Taiwan. In addition, the path

analytic approach used to examine the basic GST model provides a new look at this

model and provides stringent statistical tests on mediating effects (e.g., bootstrapping).

This study includes adolescents from both the U.S. and Taiwan; hence, any similarities

will not only support the core theoretical propositions of GST but also validate the GST
4
model in both cultures. Moreover, any differences found could help to expand and revise

GST to incorporate variation in cultures.

Dissertation outline

The dissertation comprises seven chapters in addition to the introduction. Chapter

Two introduces the background and theoretical framework for classic and general strain

theories. A thorough empirical review of GST will be included in Chapter Two. Chapter

Three examines three related topics. First, the characteristics of culture in Taiwan and the

impact of these cultural characteristics on strain/stress process will be reviewed, with a

specific focus on how culture can affect strain, negative emotions, and coping behavior.

Second, an overview of juvenile delinquency in Taiwan will be provided, which includes

discussion of some specific delinquent acts in Taiwan that may vary from those in the

U.S. due to environmental and cultural differences. Thirdly, a review of cross-national

studies on GST and studies using a GST approach that are conducted in Taiwan will be

discussed. Chapter Four gives a list of research questions that this study will address.

Chapter Five describes the study‘s research design and analytic strategy. A specific

discussion about cross-cultural research preparation will be given. Chapter Six describes

the analytic results, and Chapter Seven presents a summary of findings, discussion of the

findings, and limitations of the study.

5
 CHAPTER II:

STRAIN THEORY AS AN EXPLANATION OF DELINQUENCY

Strain theory has a long history in both sociology and criminology. It can be

traced back to Durkheim, although some have argued that Durkheim developed only a

conceptual theme of crime, rather than a full anomie/strain theory of crime (Paternoster &

Bachman, 2001). Nevertheless, Durkheim‘s analysis of the effects of anomie on suicide

set the foundation for later development of anomie/strain theory. Decades later, Merton

(1938) revised Durkheim‘s idea and proposed the influential anomie theory, which was

intended to explain crime in America. Following Merton, Cohen (1955) and Cloward and

Ohlin (1960) applied the concept of anomie/strain to subcultural delinquency (e.g., gang

culture). Parsons (1951) also employed a similar idea of strain to explain individual

deviant behavior and social control.

These leading scholars developed anomie/strain theory into a dominant paradigm

in sociology and criminology. Although some scholars have argued that the classic

anomie/strain theory is meant to explain only macro-level phenomena (e.g., social

structure and crime rates) (Bernard, 1987), others have suggested that strain theory

applies at the individual level and can be seen as an extension of Merton‘s anomie theory

(Agnew, 1987). Regardless, the theory of anomie is best conceived as a macro theory (in

6
Merton‘s words (1964), ―a sociological not an atomic theory‖), and strain theory, which

originated from anomie, is suitable for explaining individual level behavior (Paternoster

& Bachman, 2001).

Although anomie and strain theories seem to have dominated empirical research

in the 1960s, later critiques from other criminologists (Hirschi, 1969; Kornhauser, 1978;

Bernard, 1984) have dampened its dominant status in criminology. Agnew (1992; 2001),

in responding to the criticisms, outlined a revised strain theory, which he called general

strain theory (GST). Importantly, Agnew (1992, 2001) expanded the scope of the sources

of strain and delineated the underlying mechanism that leads strained individuals to crime

and delinquency.

This chapter first provides a review of classic strain theory, beginning with its

origins in Durkheim‘s macrosociological anomie theory. The contributions of three

principal strain theorists-Merton, Cohen, and Cloward and Ohlin- will be discussed along

with the extant empirical evidence and criticisms of the theory. After a review of this

historical background, general strain theory will then be thoroughly explicated and its

empirical evidence assessed.

Review of Classic Strain Theory

Durkheim’s Anomie Theory

Durkheim ([1897] 2006) indicated that a human being has two needs:

physical/organic needs and social needs. The former refers to material needs, such as

food and shelter; the latter pertains to the desire to pursue status and love through
7
developing relationships with others or society in general. These intrinsic needs1 are not

bound by any limit because, according to Durkheim, humans have the ability of

―reflection,‖ which pushes desire to another level whenever the current desires are

satiated. He further stated that ―[The] more one has, the more one wants to have, the

satisfaction one receives only serving to stimulate needs instead of fulfilling them‖

([1897] 2006, p.271).

Furthermore, Durkheim argued that ―if nothing comes from outside to restrict it [desire],

it can only be a source of torment for itself‖ (([1897] 2006, p.270). Hence, for things to

be otherwise, desires/needs must be controlled. For Durkheim, nothing inside the

individual could possibly set the limit for ever-craving intrinsic desires, so the regulatory

control must come from outside, which is society: ―Only society …can play this

moderating role, because it is the only moral power superior to the individual whose

superiority the individual accepts‖ (Durkheim, [1897] 2006, p. 272). Therefore, society is

the regulator of needs. And the relationship between societal regulation and individual

needs varies across two types of societies: mechanical and organic.

A mechanical society is formed based on the resemblance of members and social

groups in the society. The interdependence among members and groups is minimal and

each individual or social group is self-sufficient (Durkheim, [1893] 1947). In this type of

society, solidarity is achieved through the resemblance of members and the high overlap

1
Although Durkheim ([1897] 2006) argued that the physical needs (basic biological drives) are
automatically regulated by the organism, physical needs are, to some extent, insatiable. For example, it is
true one cannot eat or drink over the physical limits, but one can always crave for better (e.g., delicacy,
exquisite apparel).

8
of individual conscience and conscience collective. Consequently, social cohesion among

members of society is no longer an individual but a collective phenomenon. Mechanical

solidarity is only possible as long as individual conscience submerges into the collective

conscience. Moreover, regulation of societal members stems from this strong and defined

state of common consciousness (repressive sanction/penal law). Individuals who violate

the common consciousness are punished severely because the law holds the essential

meaning of society and, to some extent, threatens the existence of the society. The

repressive sanctions reaffirm and revitalize the conscience collective.

In contrast, an organic society is formed by diversity, so interdependence is high.

Social groups and members depend on each other in an organized system: ―[We] seek in

others what we lack in ourselves, and associations are formed wherever there is such a

true exchange of service—in short, wherever there is a division of labor‖ (Jones, 1986, p.

27). Durkheim argued ([1893] 1986) that an organic society, due to cooperative relations,

has a higher level of division of labor. In contrast to mechanical societies, (primitive), the

social order or solidarity in organic societies (advanced) is achieved through both

conscience collective and individual conscience:

For the organic solidarity to emerge, the conscience collective must leave

untouched a part of the individual conscience so that special functions …may be

established there; and the more this region of the individual conscience is

extended, the stronger is the cohesion which results from this particular kind of

solidarity.‖ (Jones, 1986, p. 34)

9
In organic solidarity, the main purpose of the law (restitutive sanction) is not to punish

the law breaker but to restore relations between individuals, or contractual parties, and to

achieve original states before the offense began. As Durkheim stated, the sanctions (law)

in an organic society consisted ―only of the return of things as they were, in the re-

establishment of troubled relations to their normal state‖ ([1893] 1947, p. 69, emphasis in

original).

According to Durkheim ([1893] 1947), social solidarity indicates that the society

integrates all individuals or social groups into a single entity. The degree of social

solidarity can be understood through two dimensions: integration and regulation. A well

regulated and integrated society, hence, achieves solidarity. But a society that deviates

from either dimension threatens the harmony of the division of labor2; therefore, the

regulation of individual appetite is weakened.

Although Durkheim did not focus his theory on crime, he did apply his theory to

one type of deviance: suicide. He argued that the phenomenon of suicide could not be

explained away by causes that lie within the individual, but instead can be attributed to

social factors ([1897] 2006). As mentioned earlier, Durkheim classified society based on

two dimensions of social solidarity (integration and regulation). An abnormal society in

which each dimension verges on the extreme has a higher rate of suicide. Accordingly,

four different types of abnormal societies exist, and each is conducive to a unique kind of

suicide. The first two types of abnormal societies are at two ends of the integration

2
Only organic solidarity (division of labor) is discussed because each society, according to Durkheim, is at
a developmental stage of a process toward division of labor.

10
continuum. On the one end, where society is tightly united and the individual is absorbed

by the collective, individuals commit suicide for social purposes. For example, in some

primitive societies elders kill themselves in order to reduce the burden on the tribe, and in

mordent society, soldiers might end their lives for the honor of the society. On the other

end of the integration continuum, where a society is disintegrated, the individual is

detached from the society that both gives him meaning and satisfies his social needs:

―[The] link that attaches him to society has itself been relaxed‖ (Durkheim, [1897] 2006,

p. 231). Consequently, social members commit egoistic suicide for the individuals‘

purpose. The third type of society, one polar on the regulation dimension, is fatalistic

suicide, which is a product of the over-regulated society3.

However, an anomic society—and its suicidal acts—is most related to the strain

paradigm. According to Durkheim ([1897], 2006), an anomic society refers to inadequate

regulation or normlessness, which is the other extreme on the regulation continuum. This

state is highly conducive to abnormal behavior such as suicide4. For example, during an

economic crisis, the suicide rate increases because of the anomic situation that

individuals encounter in the society. The norms that regulate individuals are no longer

appropriate because society fails to teach individuals to reduce their needs. Therefore,

individuals living under an economic crisis cannot meet their needs by using means that

3
Durkheim only mentioned this type of suicide as the result of a ―pitilessly blocked future‖ or ―excessive
physical or moral despotism.‖ And he recognized its rarity and assigned little importance to fatalistic
suicide in modern societies.
4
While Durkheim only related anomie to suicide, contemporary scholars have employed his idea to explain
other social problems, such as homicide (Pridemore, Chamlin, & Cochran, 2007) and economic crime (Cao,
2007).

11
they used during the normal economic situation, and this casts an individual into an

uncomfortable state. Durkheim claimed that ―[individuals] are not adjusted to this

condition and the very prospect of it is intolerable to them‖ ([1897], 2006, p. 276).

Economic prosperity also results in an anomic state because the regulation of needs is

broken, which leads people to attempt to satisfy limitless appetites. This condition pushes

individuals into an uncomfortable state since an individual constantly pursues unattained

goals, which makes him or her feel unhappy. Therefore, whether prosperity or crisis, both

situations lead to sudden social change, hence an anomic situation, which causes social

problems.

In sum, Durkheim argued that sudden social changes, whether caused by crisis or

prosperity, lead to the breakdown of social regulation, which in turn creates the state of

anomie where the old rules of regulating are inappropriate and new rules are not yet

formed. This anomic condition ―unleashes escalating needs that outstrip means and

ultimately pressure actors into committing suicide‖ (Cullen, 1984, p. 81).

Whereas Durkheim delineated the concept of the division of labor and various

forms of abnormal societies and their relationship to suicide, Merton‘s anomie theory had

more pronounced influence on later developments in strain theory. Merton (1938) revised

Durkheim‘s anomie theory and constructed his own anomie theory, which was

specifically developed to explain how inequality in the social structure and culture goals

produced deviant adaptations in the United States.

12
Merton’s Anomie Theory

In his ―Social Structure and Anomie‖ (Merton, 1938), Merton revised the legacy

of Durkheim‘s anomie theory and applied it to explain various deviant acts in America.

Like Durkheim, Merton intended to develop a sociological explanation for deviant

behavior in a society. Thus, he stated that ―our primary aim lies in discovering how social

structures exert a definite pressure upon certain persons in the society to engage in

nonconformist rather than conformist conduct‖ (Merton, 1938, p. 672). Later in his

writings, he proclaimed that ―our perspective is sociological‖ (1968, p. 186).

Although Merton and Durkheim both attempt to use sociological theory to explain

social problems, there are three major differences between them. First, Durkheim argued

that deviance arises because of a breakdown in social regulation, which governs innate

human impulse. In contrast, Merton stated that ―the aberrant behavior may be regarded

sociologically as a symptom of dissociation between culturally prescribed aspirations and

socially structured avenues for realizing these aspirations‖ (1968, p.188.). Second,

Durkheim referred to anomie as the failure of society to regulate or restrain goals and to

provide suitable norms to follow (normlessness), whereas Merton referred to anomie as

―a breakdown in the cultural structure, occurring particularly when there is an acute

disjunction between cultural norms and goals and the socially structured capacities of

members of the group to act in accord with them‖ (1968, p.216). Finally, Durkheim

applied his theory only to one type of social deviance – suicide. Merton, on the other

hand, is more ambitious in formulating a theory of different types of adaptations, which

include various kinds of deviance (Cullen, 1984).

13
 Merton (1938, 1959, 1964, 1968) conceptualized a social system as comprising

two elements: a social structure and a cultural structure. The cultural structure can be

defined as an ―organized set of normative values governing behavior which is common to

members of a designated society or group‖ and the social structure as the ―organized set

of social relationships in which members of the society or group are variously

implicated‖ (Merton, 1968, p. 216). The cultural element was further divided into two

subparts: the society‘s central goals or values (ends) and the institutionalized ways to

achieve such goals (means). The ―ends‖ referred to ―culturally defined goals, purposes,

and interests‖ that were ―held out as legitimate objects for all or for diversely located

members of the society‖ (Merton, 1959, p. 228); the ―means‖ referred to a cultural

structure that primarily defined, regulated, and controlled the normative modes of

realizing culturally defined goals (Merton, 1959, 1968).

On the basis of these two concepts, the cultural structure (means and ends) and

the social structure, Merton explained social systems in a systematic way. When there is

equilibrium between the cultural structure and the social structure, that is, when the

culturally approved goals could be realized by members of the collectivity via

normatively prescribed means within social structural relationships, an organized

collectivity or society is expected. In such an equilibrated society, where there is a

harmonious relationship between the cultural structure and the social structure,

individuals receive satisfaction both from achievement of goals and from striving to

realize goals via institutionalized modes (Merton, 1968). Hence, success is twofold: ―[I]t

is reckoned in terms of the product and in terms of the activities.‖ (Merton, 1959, p. 230).

14
 The focus of Merton‘s theory, however, was to explain non-organized society

(anomic society) based on his conceptions of the cultural structure (means and ends) and

the social structure. A society becomes unstable because of a malintegration between the

cultural structure and social structure and/or within the two elements of the cultural

structure (cultural goals and institutional means). The former occurs when the culturally

designated goals are universally applied but the access to the means is not equally

distributed within the social structure. The latter refers to the situation in which the

cultural goal is held at the highest position, while cultural means are relegated to a

relatively low position. In such a situation, satisfaction is likely come to individuals who

could not compete successfully either because access to the means is lacking or the

means are inefficient. Merton stated clearly that ―[I]f concern shifts exclusively to the

outcome of competition, then those who perennially suffer defeat will, understandably

enough, work for a change in the rules of the game‖ (1959, p. 230). Consequently, the

technically most effective means that lead to achieving the moral mandate will be

preferred whether it be ―fair means‖ or ―foul means‖ (Merton, 1968). Therefore, ―as this

process of attenuation continues, the society becomes unstable and there develops what

Durkheim called ‗anomie‘ or normlessness‖ (Merton, 1959, p. 231).

Based on the aforementioned conceptual scheme, Merton (1938) developed five

types of adaptations to strain based on the discrepancy between cultural goals and

institutional means to explain how an anomic social system induces deviant behavior.

The first adaption, conformity, is the most common in a society; if this were not so,

continuity and stability would not be possible for a society. Conformists are people who

15
not only accept cultural goals but also follow the institutional means to realize the goals.

Because Merton‘s primary purpose was to explain deviant behavior, he did not spend

much time on conformity.

The second type of adaptation, innovation, is of the most interest. This adaptation

occurs when people internalize cultural goals but do not at the same time assimilate the

institutional norms governing means to attain such goals. To Merton, this type of

adaptation is the most common deviant adaptation and is closely related to crime (1968).

Merton, then, employed this adaptation to explain the high crime rate among low SES

groups in America. People in low social strata absorb the cultural goals, such as

pecuniary success in America, but these individuals have little access to conventional

means for achieving success due to either little education or limited economic resources

or both. It is in these social strata that one can experience high levels of ―innovation,‖

namely, deviant behavior. Merton argued that ―successful‖ innovative behavior lessens

social norms, which intensifies the anomic situation for people in the system, and this, in

turn, leads others to deviate in order to adapt to the severe anomic situation.

Ritualism, the third type of adaptation, is characterized by retention of

institutional means but rejection of cultural goals. In Parsons‘ paradigm, this is seen in

individuals who either passively conform to social or compulsively acquiesce to social

norms (1951). For example, a government worker may realize that the opportunity of

advancing in the social system is slim; hence, he or she may go to work and follow the

rules just for the sake of ―doing it.‖ This kind of adaptation may not be highly deviant at

first, but Merton argued that in the Western cultural model, ―men are obligated to strive
16
actively…to move onward and upward in the social hierarchy‖ (1959, p. 246); therefore,

departure from such expectations is deviant. Merton expected (1959, 1968) that

innovation will be common in the lower class because of the prevailing emphasis largely

on cultural goals with limited conventional opportunities or means. On the other hand,

ritualistic adaptation should be heavily present in the lower-middle class because of

successful socialization along with limited opportunities to climb the social ladder.

Retreatism, the fourth type of adaptation, is seen in individuals who abandon both

cultural goals and institutional means. These individuals internalize both goals and means

but constantly face conflict between ideology and reality; that is, the promise of success

cannot be realized through conventional means. However, these individuals cannot adopt

―innovative means‖ nor can they give up the goals or without renouncing the ―supreme

value of the success goal‖ (Merton, 1959, p. 250). To resolve this conflict, they abandon

both the goals and the means. The escape is complete in that they are in the society but

not of it.

Finally, the rebellion adaptation consists of creating new goals and means in

addition to rejecting the original goals and means. Parsons described individuals who use

this form of adaptation as actively alienated from the social system (1951). Merton

argued that rebellion can be at two levels: the small and confined level and the endemic

level. The former provides an opportunity for the genesis of a subgroup alienated from

the community but unified within the group (Merton, 1959). Examples can be found in

both Cohen‘s (1959) and Anderson‘s (1999) studies, in which cultural goals and means

are replaced with ―new‖ goals and means. The latter (endemic level) mainly refers to
17
large-scale rebellion that intends to substitute the goals and means of society at large,

such as revolution.

In sum, Merton delineated the anomic society, within which the emphasis is

greater on the cultural goals than on the cultural means and the distribution of

institutional means to realize goals is unequal. In such a society, certain people will feel

strain and frustration, which in turn lead them to respond in deviant ways. It is the

explicit emphasis on structurally induced strain that make Merton‘s theory suitable for

both macro-level (Bernard, 1987; Messner, 1988) and micro-level analyses (Agnew,

1985). However, Merton did not develop a clear explanation of the strain-delinquency

relationship and ignored the presence of delinquent behavior in the subculture (Cohen,

1959). Merton also has been criticized for his neglect of the unequal distribution of

different means to realize different ends (Cloward & Ohlin, 1960). Cohen (1955) and

Cloward and Ohlin (1960) built on Merton‘s ideas to rectify these two limitations. The

following sections will discuss first Cohen‘s theory and then Cloward and Ohlin‘s theory.

Cohen’s Delinquent Subculture Theory

Cohen (1955) applied Merton‘s concept of anomie and strain along with societal

interaction theory to explain the formation of the lower-class, male delinquent subculture.

He argued that Merton‘s theory is valuable in explaining adult criminal acts or semi-

professional juvenile thieves but is less valuable in explaining the lower status, male

delinquent subculture. He also pointed out that Merton focuses on only one cultural goal-

monetary success-and ignored another important goal-middle class status. In addition,

18
Cohen (1965, p. 9) criticized Merton for ignoring the anomic processes ―whereby acts

and complex structures of actions are built, elaborated, and transformed.‖ In other words,

Merton should have considered the process of interaction between several individuals,

which may induce a deviant act (Clinard, 1964).

According to Cohen, all individuals are constantly involved in a series of efforts

to solve problems, which are defined as ―a certain tension, a disequilibrium and a

challenge‖ (1955, p. 59). Problems come from two sources: the situation that one lives in

and the reference frame that one employs. For Cohen, the most effective or satisfying

solution to any problem must ―entail some change in that frame of reference itself‖ (1955,

p. 53, emphasis in original). The reference frame is variously defined by different

subgroups, the most important of which are the ―reference groups,‖ which are more

effective in defining the validity of the individual‘s beliefs and which are more powerful

in providing incentives not to deviate from the established group norms (Cohen, 1955).

With the above conceptualizations, the core of Cohen‘s delinquent subculture

theory is that lower SES males encounter a ―status problem‖ or problem of adjustment in

school. Cohen (1955, p. 65) argued that ―status problems are problems of achieving

respect in the eyes of one‘s fellows.‖ Low SES males were not equipped with the middle-

class standards or ―middle-class measuring rods‖ (e.g., manners, nonaggressive behavior,

studying hard) that teachers or school officials use to evaluate students. Students who

study hard or behave well in class, for example, are more likely to gain ―status‖ from the

school system. The lower-status students, according to Cohen (1955), had not been

educated in such ways in their social milieu. As a result, the ―status‖ problem they
19
experience in school produces strain in these students, who were ―beset by one of the

most typical and yet distressing of human problems of adjustment‖ (Cohen, 1955, pp. 65-

66).

Many lower SES students lack the characteristics or capabilities of gaining status

in the larger society where they participate (e.g., school). One solution to status problems

is for individuals who experience the same problems to congregate together and jointly

establish a new standard of status that they could live up to. Accordingly, lower-class

juveniles who experience a similar status strain interact with one another to find a

solution to their common problem. After a long process of interaction (reaction-formation)

and ―joint elaboration of a new solution‖ (p. 60), these individuals come to form a new

subculture, which Cohen argued earlier was the most effective and satisfying solution

because it provides a new reference frame. This new reference frame, which entails ―the

kinds of conduct of which they are capable‖ (p. 66), satisfies the needs for ―status‖ and

thus reduces the strain. Although the subcultural solution to status strain is similar to

Merton‘s ―innovation‖ adaptation, or using ―new‖ ways to achieve goals, the subcultural

solution is a group solution, rather than a private one, because it defines status according

to one‘s fellows and the criteria of the group.

According to Cohen (1955), there are many important characteristics of the

delinquent subculture: ―hedonism,‖ ―group autonomy,‖ an orientation that is ―malicious,‖

―negativistic,‖ or ―non-utilitarian,‖ and ―versatility.‖ Hedonism refers to little interest in

long-term goals, planning activities or budgeting because the enjoyment of the here and

now is the key. Group autonomy refers to intolerance of restraint, with the exception of
20
the informal pressure to follow group rules (e.g., loyalty). A malicious orientation refers

to the enjoyment of the discomfiture of others and of defiance of social taboos (e.g.,

aggression). A negativistic orientation indicates the norms or standards in the delinquent

subculture are antithetic to the norms of the larger society. Cohen (1955) argued that

stealing is another way to achieve status, not merely an alternative means to acquire

objects. Finally, Cohen observed that the delinquent culture is versatile, which negates

the notion of ―specialization.‖ For Cohen, delinquents are involved in all kinds of

―activities‖ in order to attain status. This concept of versatility is consistent with that of

contemporary criminologists (Goffredson & Hirschi, 1990).

While the above paragraph describes how a delinquent subculture is formed, the

methods of solving the problem of status frustration are not limited to the youths who

interact together and form the delinquent subculture. Another solution can be found in

youths who neither fully commit to the delinquent subculture nor repudiate all middle-

class values. These youths, called ―stable corner-boys,‖ try to make the best of a situation

through middle-class rules, but are also involved in some minor delinquent acts. Finally,

the ―college-boys‖ are those who endorse not only the middle-class goals but also

middle-class rules. The differences between delinquent boys and college boys lie in the

fact that the former repudiate both middle-class goals and rules and create a new set of

rules to cope with the status strain, while the latter endeavor to achieve middle-class

goals by following conventional rules. Finally, the stable corner-boys, although they

commit delinquency, do not reject middle-class rules wholesale.

21
 In sum, Cohen (1955) argued that lower-class adolescents lack the resources and

social skills to succeed or gain status in school, where middle-class standards prevail, and

this lack provokes status frustration or strain. The adolescents who share this same

problem congregate together to create a new standard, that of the delinquent subculture,

within which they can succeed and gain status. This new standard is in conflict with

middle-class rules but provides a satisfying solution for these strained adolescents.

Whereas Cohen focused on status strain in the lower-class group and the relationship

between this strain and consequent delinquency and delinquent subculture formation,

Merton paid attention to the goals-means discrepancy and its effect on different kinds of

adaptation.

Cloward and Ohlin’s theory of Different Opportunity and Delinquent Subculture

Cloward and Ohlin (1960) explored the formation of male delinquent subcultures

in the lower social stratum, building on Merton‘s notion that the discrepancy between

goals and means creates strain, which in turn, leads to delinquency. However, they

argued that Merton ignored the fact that the distribution of illegitimate opportunities for

success is not available to everyone. They also built on Cohen‘s concept that adjustment

problems create strains for individuals, who then congregate together to find solutions,

which introduces delinquent subcultures and delinquency.

Cloward and Ohlin (1960) argued that individuals in a society make an effort to

meet or conform to social expectations or moral mandates and that such efforts ―often

entail profound strain and frustration‖ (p. 38). Furthermore, they stated that extending

22
socially approved goals under conditions in which conventional means are not available

is the precondition for deviance and delinquency, as responses to strain or adjustment

problems. To apply those basic themes to explain lower-class adolescent delinquency and

formation of delinquent subcultures, they pointed out that the problems of such strains are

disproportionally distributed in lower-class groups, are significant to these adolescents,

and are permanent to these adolescents (1960). In addition, they recognized that there are

barriers (structural and cultural) to reaching goals in this particular group, which makes

the strain bear these characteristics and seem to be even more stressful. With these basic

conceptualizations, Cloward and Ohlin continue by stating that lower-class youths who

face such strains interact with one another in a long and complex process, which may

provide ―encouragement for the withdrawal of sentiment in support of the established

system of norms‖ (1960, pp. 108-109). With the support of others who share the same

problems, individuals may devise or adopt illegitimate means to achieve success. Also

with such support, the anxiety and guilt associated with violating social norms is

generally reduced, because allegiance to the conventional means is set aside, which in

turn justifies the deviant means used.

While this group solution to a common adjustment problem is similar to the

solution described in Cohen‘s (1955) theory, the point of departure lies in the outward

attribution. Cloward and Ohlin (1960) argued that the formation of a delinquent

subculture as a solution to an adjustment problem is likely only when individuals

attribute their frustration to the unjust system rather than themselves. Lower class male

adolescents face a relative discrepancy between institutional expectations and possible

23
means of living up to them because of the various barriers to achievement, which causes

a feeling of discrimination; all these factors make the delinquent subculture solution, a

group adaptation, highly likely.

Having delineated the source and process of forming the delinquent subculture,

Cloward and Ohlin (1960) introduced another important concept: illegitimate means,

which help to distinguish between three different delinquent subcultures: criminal,

conflict, and retreatist subcultures. The means includes two things: the learning

environment for acquiring required skills and values to perform a particular role, and the

opportunity structure that enables individuals to fulfill the role (Cloward, 1959). A

―[C]riminal subculture‖ develops in ―integrated neighborhoods‖ where not only do

conventional values and delinquent values coexist but also the different age levels of

offenders are integrated well: ―Unless the carriers of criminal and conventional roles are

closely bonded, stable criminal roles cannot develop‖ (Cloward & Ohlin, 1960, p. 165).

As the criminal roles are established, adult criminals provide role models for adolescents

to emulate, and at the same time provide illegitimate opportunities to succeed.

Consequently, the ―criminal subculture‖ provides ―means‖ for these adolescents to

succeed, which in turn solves the problem of adjustment.

The ―conflict subculture‖ develops in disorganized communities, which are

unstable and transient. Such neighborhoods do not provide adolescents with legitimate

means to succeed, nor do the communities provide criminal means to achieve goals. In

addition, social control from both the conventional and the illegitimate sectors was

loosened. Because both conventional and criminal means are blocked, adolescent
24
delinquents are left on their own to solve the adjustment problem. The only thing they

can use to achieve ―status‖ or ―respect‖ is violent or physical conflict and, because of

weakened social control, violence intensifies.

Finally, the ―retreatist subculture‖ also emerges in disorganized communities. In

addition to the same conditions as the previous two subcultures, youths who participate in

this subculture lack both conventional and illegitimate means to success, whether

criminal acts or conflict. Cloward and Ohlin (1960) called these adolescents ―double

failure.‖ Because of the constant failure resulting from these restrictions, these youths

escape from society in order to deal with their strain and frustration. Cloward and Ohlin

(1960) argued that not all ―double failures‖ youths adopt the retreatist subculture; youths

might instead eventually become Cohen‘s ―corner-boys,‖ who live in accordance with the

lower-class lifestyle. The difference is that retreatist adolescents are ―incapable of

revising their aspiration downward to correspond to reality‖ (p.184).

In sum, the theory of Cloward and Ohlin (1960) argued that lower-class male

adolescents face adjustment problems or strains introduced by blocked legitimate means

to realize cultural goals. In contrast to assumptions of previous theories, they pointed out

that illegitimate opportunities are also not equally available to all adolescents who

experience such strain. Consequently, those who become involved in the criminal

subculture are those for whom illegitimate opportunities are available. For those who

react with severe violent acts, both the conventional and the criminal means are closed

but these adolescents possess physical ability or ―guts.‖ Finally, for those who lack all

these means, the ―double failure,‖ the retreatist subculture becomes attractive. The major
25
contribution of Cloward and Ohlin (1960) was to introduce the opportunity structure into

strain theory, to explicitly explain why some strained lower class adolescents become

involved in one or another kind of delinquency.

Criticisms of the classic strain theory

This section will review some general criticisms of classic strain theory as a

whole; however, some critiques of individual theorists will not be presented here5. There

are three general criticisms of the classic strain theory (Cohen, 1955; Cloward & Ohlin,

1960; Merton, 1938): conceptual problems, limited empirical support, and limited scope

of strain (e.g., focus only on goal-mean discrepancy) and delinquency (e.g., focus only on

lower class delinquency). Strain theorists suggest that the imbalance of emphasis on goals

over means in a society creates the pressure for its members to deviate; hence, it is the

imbalance of culture that creates the motivation (strain) to commit aberrant behavior.

Kornhauser (1978) argues that motivation to crime is not necessary because people

naturally want more, as Durkheim would argue, and she also points out that the source of

strain is due to ―weak culture,‖ not an imbalance of culture. The ―weak culture‖ is the

culture that fails to provide ―public reorganization of moral worth‖ to its members who

pursue their desired goals (Kornhauser, 1978, p.162). In her view, strain arises because a

culture or a society does not recognize different goals that individuals in a society may

pursue; hence, culture forces all members to live up to the same goal.

5
For example, Kitsuse and Dietrick (1959) provided an excellent critique on Cohen‘s theory.

26
 The second criticism, the lack of empirical support, is perhaps the most

detrimental to classic strain theory. Most empirical studies operationalize strain as the

discrepancy between aspiration/expectation of either high educational attainment or

prestigious occupation. These studies usually do not find support for strain theory‘s

prediction that those experiencing a large gap between aspirations and expectations

would commit more delinquency/crime (Agnew, 1984; Akers & Cochran, 1985; Burton,

1991; Hirschi, 1969; Kornhauser, 1978; Elliott & Voss, 1974; Eve, 1978; Liska, 1971).

Although most studies employing such an operationalization do not find support, Bernard

(1984) contends that the more theoretically consistent measures of strain should be the

discrepancy between aspiration/expectation of monetary success not the discrepancy

between aspiration/expectation of educational attainment. Bernard suggests that

education is only a means to meet the end; it is not the end in the classic theory. This

argument is later supported by Farnworth and Leiber (1989), who argue that educational

attainment is but one means to achieve economic success in the society. They find that

the disjunction between economic goals and educational means predicts delinquency,

especially serious utilitarian offenses. Similarly, Agnew and colleagues (1996)

operationalized strain as dissatisfaction with monetary status, and they find that this

variable strongly predicts income generated crime and drug use.

Although the common operationalization of strain is the discrepancy between

aspiration/expectation of educational attainment or job satisfaction, other studies have

focused on the blockage of opportunities to success. These studies usually measure strain

as individuals‘ perception of their chances of achieving culturally approved goals. Burton

27
and Cullen (1992) argue that such a measure is more closely related to classic strain

theory; however, the results from empirical studies are mixed (Burton & Cullen, 1992,

pp.15-16).

Finally, classic strain theory focuses only on a limited type of goal (Agnew,

1985a), and its explanation of delinquent behavior is narrow (Akers, 2000). For Merton

(1938, 1959, 1968) and Cloward and Ohlin (1960), the goal is monetary or material

accumulation, and for Cohen (1955), the goal is middle–class status. Agnew (1985a) and

others (Elliott & Voss, 1974; Elliott, Ageton, & Canter, 1979; Quicker, 1974) argue that

youths pursue various goals (e.g., popularity, good academic performance) rather than

limiting themselves to those goals that classic strain theorists have suggested (e.g.,

monetary success). In addition, the goals that adolescents recognize as important are not

necessarily long-term goals; rather, they may be immediate goals, such as popularity in

school.

Merton (1938, 1964) argues that the disjunction between goals and means in the

lower class of a society creates strain, which motivates individuals to act deviantly

(innovation). Cohen (1955) is interested in lower class adolescents who experience status

frustration in school and respond through delinquency and the creation of a delinquent

subculture. Cloward and Ohlin (1960) focus on strain, the lack of conventional and

criminal opportunities to achieve a monetary goal, of lower class male adolescents and

the consequent delinquent behaviors. All these theoretical arguments limit their scope in

explaining lower class delinquency and delinquent subculture. As such, Agnew (1991)

argues that the classic strain theory lacks the ability to explain the nature of middle class
28
crime. In addition, classic strain theory treats social class as a barrier that impedes

individuals‘ achievement of culturally prescribed goals whereas Agnew (1991) suggests

that other barriers might be at work (e.g., personality, skills).

In addition to the above criticisms, Agnew (1985a; 1991) argues that a related

limitation of classic strain theory is that it focuses on strain that is introduced by blocking

of the achievement of positively valued goals. However, another kind of strain, which is

the blockage of escape from an aversive situation, is also an important cause of juvenile

delinquency, because adolescents have relatively little power to change an aversive

situation and they have not yet developed fully the mature cognitive and problem-solving

skills and experience needed to cope successfully with these aversive situations (Agnew,

2003; DuRant et al., 1995). For example, whereas adults can move freely, within

financial limitations, away from an aversive situation, juveniles are bound to their family

and schools because they lack the means to move away (e.g., money, a car) and are

legally compelled to remain in these situations. Steinberg and Cauffman (1996) suggest

that psychological dispositions of the early adolescent (e.g., cognitive ability) lead many

youths to make immature decisions about coping. Hence, two different kinds of strain can

be identified; in the blockage of goal-seeking behavior, an individual is ―moving toward

a valued goal, but in the blockage of aversive avoidance, one is moving away from an

aversive situation‖ (Agnew 1985a, p.154, emphasis in origin).

On the basis of these criticisms, Agnew (1985a, 1992), among other scholars,

began to revise classic strain theory. First, the revised strain theory broadened the scope

of strain to including strains from aversive situations and goals that were immediate to
29
individuals (e.g., good school grades). Second, the revised strain theory was able to

account for delinquency in different social classes. This was in response to studies

concerned with the results of self-report delinquency, which often revealed that juveniles

from middle and high SES families also commit delinquent acts. Furthermore, the

broadened scope of revised strain theory included strains other than economic strain,

which were expected to be pervasive across SES levels. For example, students from

different SES families all struggled with gaining autonomy from their parents.

Consequently, parental control may be a common strain for youths. Third, the revised

strain theory included social psychological dimensions such as negative emotions. The

following section will briefly review ―modern‖ strain theory (Agnew, 1991), and a

detailed discussion of Agnew‘s general strain theory (1992) will then be given.

Modern strain theory and Agnew‘s general strain theory

Modern strain theory

The popularity of classic strain theory gradually waned in the 1970s because of

the aforementioned limitations, especially the serious shortcomings of limited scope and

lack of empirical evidence. Some scholars even suggested that this paradigm should be

abandoned (Hirschi, 1969; Kornhauser, 1978). In responding to these criticisms, while

still accepting the concept that failing to achieve desired goals through legitimate means

produces strain and motivation to delinquency, modern strain theory focuses on three

revisions: the characteristics of the desired goals, the barriers to realizing goals, and the

cumulative effects of strain on delinquency (Agnew, 1985a, 1991).

30
 First, modern strain theory proposes that the important desired goals of

adolescents are not limited to economic success or middle-class status but include many

other goals, such as good school grades or excellent athletic performance. In addition,

modern strain theory contends that the goals of adolescents may be immediate goals (e.g.,

friendship) rather than long-term goals (e.g., occupational aspirations). Although studies

indicate that adolescents pursue various goals and immediate goals (Agnew, 1984; Elliott

& Voss, 1974; Quicker, 1974), results of empirical tests of these ideas have not been

promising (Agnew, 1984).

Second, modern strain theory suggests that goal blockage increases the possibility

of delinquency, especially when it causes the adolescent to fail to realize most of his or

her goals. However, empirical tests of this proposition suffer from some limitations (e.g.,

measuring only some goal-blockage); they provide only mixed support (Agnew, 1984;

Greenberg, 1979). Hence, Agnew (1985a, 1991) concludes that the revised modern strain

theory, like its predecessor, receives only weak support. Furthermore, the various

revisions that modern strain theorists have proposed attempt to accomplish only

―patchwork‖ rather than providing a systematic explanation. For example, many of these

revisions limited their scope of strain to the strain that classic strain theories

conceptualized; that is, strain is induced because of failure to reach positively valued

goals. These revisions, hence, focused on including different sources of strain (e.g.,

different goals) in response to the common criticism that aspiration/expectation

discrepancy did not lead to, or was weakly related to, delinquency.

31
 Agnew (1985a, 1992) took up the challenge and developed a more advanced and

systematic explanation of the strain-delinquency relationship, which he labeled as general

strain theory (GST). In contrast to all his predecessors, Agnew (1985a) introduced a

different type of strain − failure to escape from aversive situations or stimuli.

Furthermore, he included the concept of anger and conditioning factors, although the

latter only implicitly. He found that individuals who could not escape from an aversive

environment were more likely to be involved in delinquency, interpersonal aggression,

and escape behavior directly and indirectly through anger; and the results were

significant even after controlling for other theoretical variables (e.g., delinquent peers,

attachment to mother). Agnew (1985a) concluded that this revised strain theory was able

to explain middle-class delinquency and sporadic juvenile delinquency.

Agnew’s general strain theory

After successfully introducing a new type of strain, ―avoidance of aversive

situations,‖ Agnew (1992) further revised classic strain theory. He added another type of

strain, provided a broader range of negative emotions, and introduced the concept of

conditioning factors and coping strategies. This more systematic explanation of the

strain-delinquency relationship was labeled general strain theory (GST). In GST, there

are three central components: strain, negative emotion, and coping strategies.

Strain

The focus of GST is on negative relationships with others: ―relationships in which

the individual is not treated as he or she wants to be treated‖ (Agnew, 1992, p.48). GST

32
defines three types of major strains: (1) relationships in which others prevent the

individuals from achieving positively valued goals, (2) relationships in which others

present or threaten to present negative stimuli, or (3) relationships in which others

remove or threaten to remove positively valued stimuli. The first type of negative

relationship includes strains from classic strain theories (e.g., monetary strain) and strains

from modern strain theory (e.g., doing well in athletics). The second type of negative

relationship includes various situations in which the individual feels uncomfortable.

Agnew (1985) pointed out that preventing individuals from escaping from an aversive

situation does indeed increase the possibility of juvenile delinquency. The third type of

negative relationship is commonly found in stressful life-events lists (e.g., death of family

members).

Although the first type of strain was similar to classic strain theory, Agnew (1992,

pp.51-53) further divided this type of strain (failure to achieve positively valued goals)

into three subtypes: ―the disjunction between aspirations and expectations/actual

achievements‖, ―the disjunction between expectation and actual achievements‖, and ―the

disjunction between just/fair outcomes and actual outcomes.‖ The first subtype was

consistent with classic strain theory, but Agnew did not limit himself to only monetary

goals and included other immediate goals (e.g., popularity). He argued that the second

subtype of strain, the discrepancy between one‘s expectations, which is more realistic,

and actual achievements was more distressing. Compared with aspiration, which is

idealistic and derived from one‘s cultural system, expectation is generated from one‘s

―past experience and or/from comparison with referential or (generalized) others who are

33
similar to the individual‖ (Agnew, 1992, p.52). For example, classic strain theorists

assume that a particular important aspiration for individuals in the United States is to

achieve monetary success. In contrast, the expectation for an individual may be to

achieve his or her parents‘ status. Hence, individuals who fail to achieve these more

realistic goals might have stronger motivation to seek other means to achieve them.

Finally, largely on the basis of the equity and justice literature, Agnew argued that

individuals not only pursue goals, whether aspired or expected; they might also expect

that fair or just rules will be followed in allocating rewards in each interaction.

Consequently, when unfair or unjust outcomes are encountered, individuals might feel

strain and have the desire to correct for such ―injustice‖ so that they could gain more

rewards, reduce their input, reduce others‘ rewards, or increase others‘ input. For

example, one might steal something from the employer (gain a greater reward) or be

uncooperative so as to increase the efforts that the employer must put into the job

(increasing others‘ input) in an effort to reduce the strain resulting from an unjust

promotion decision.

The second major type of strain refers to the presentation of negative stimuli.

Negative stimuli can be social (e.g., discrimination) or non-social (e.g., natural disaster,

illness). Individuals who experience such negative stimuli or aversive situations might

become involved in delinquency in order to escape from the situation (e.g., skipping

class), terminate or alleviate the negative stimuli (e.g., drug use), or seek revenge against

the source (e.g., aggression).

34
 The third type of strain derives from the removal of positively valued stimuli,

which can be social (e.g., friendship) or non-social (e.g., materials). For example, the

commonly used stressful-life event checklist in the stress literature usually includes items

such as ―loss of a boyfriend/girlfriend‖ or/and ―death of a relative.‖ Individuals who

experience such strain might try to prevent the loss, to retrieve the lost stimuli, to obtain

substitute stimuli, or to seek revenge against the source.

After describing the three major types of strain, Agnew (1992) specified the

characteristics of strain that made it more influential. He suggested that strain that was of

high magnitude, more recent, or of longer duration had stronger effects on consequent

negative emotion and delinquency. He also argued that strains closely clustered in time

had a particularly strong negative effect on individuals. For example, a bad score on one

exam may not be so stressful but it becomes a strain when students not only get a bad

grade but also have a fight with friends and lose a close relative. Although Agnew (1992)

did not give more detailed descriptions of these characteristics, he later pays attention to

these characteristics and elaborates upon them (Agnew, 2001).

Negative emotions

With three major types of strain having been delineated, GST explains the link

between these potential strains and consequent delinquency. GST argues that each type of

strain can lead the individual to experience an array of negative emotions, including

anger, fear, and depression. Among the various negative emotions, anger is the most

important to GST, because anger is very likely to increase an individual‘s level of

35
outward attribution of the injury, instigating the individual to act, motivating the

individual to take revenge, and lowering the individual‘s inhibitions. Hence, anger

influences individuals in various ways that are conducive to delinquency. Although anger

is the most criminogenic emotion, delinquency might still occur in response to other

negative emotions (e.g., depression). For example, anger might cause an individual to act

aggressively against other individuals, whereas depression might lead the individual to

take drugs in order to feel ―better.‖ Consequently, Agnew (1992, footnote 10)

distinguished between outer-directed negative emotion (e.g., anger), which increases the

likelihood of outer-directed actions (e.g., violence), and inner-directed negative emotions

(e.g., depression), which lead an individual to show inner-directed responses (e.g.,

substance use). In sum, ―[T]he experience of negative affect, especially anger, typically

creates a desire to take corrective steps, with delinquency being one possible response‖

(Agnew, 1992, p.60).

Coping strategies

The third element of GST is the coping strategies that the strained individual uses

to cope with strains and negative emotions. Agnew (1992) identified three major types of

coping strategies: cognitive, emotional, and behavioral. The cognitive coping strategy

mainly focuses on reinterpreting the strain in ways that minimize one‘s negative feelings.

The individual can use this strategy in three ways to deal with the adversity. First, the

individual can ignore/minimize the importance of the outcome. For example, one might

say that ―this is not important‖ or ―money is not important compared with family.‖

Second, one might maximize positive outcomes and minimize negative outcomes. For
36
example, Agnew (1985b) found that crime victims often stated that their victimization

helped them to learn from it, which in turn reduced the negative feelings they attached to

their victimization. Finally, individual might simply accept responsibility for the negative

results so as to manipulate the input and output of themselves and others in a relationship.

For example, one might claim that he or she did not work hard enough (minimize the

positive input) or that others worked harder than they did (maximize others‘ input). As

can be seen, this strategy is mainly a non-delinquent response.

The emotional coping strategies are responses that directly cope with the negative

emotions resulting from a strain. Agnew (1992) offers several examples of emotional

coping strategies, such as drug use, meditation, physical exercise, and various

psychological techniques (e.g., playacting). Most of these strategies are conventional;

however, drug use to reduce negative feelings could be antisocial (e.g., using an illegal

substance).

Finally, there are three major subtypes of behavioral coping strategies:

minimizing a negative outcome, maximizing a positive outcome, and taking vengeful

behavior. To minimize a negative outcome, an individual can reduce negative feelings,

terminate the cause of the negative outcome, or escape from the negative outcome.

Several delinquent behaviors are explicitly related to such strategies; examples are

substance use or skipping classes. To maximize a positive outcome, an individual can use

means to increase his or her gain in a relationship or to retrieve valued goals. For instance,

an individual may join a gang to gain support from gang members who will help the

person to achieve goals (Cloward & Ohlin, 1960). Individuals may take vengeful
37
behavior when they are strained in order to increase the inputs of others or decrease the

positive outcomes for others. For example, adolescents might act incorrigibly to force

parents and teachers to work harder to deal with them (increasing their inputs). Hence,

behavioral coping strategies, rather than emotional or cognitive coping strategies, are

closely related to delinquent acts.

In sum, Agnew‘s general strain theory (1992) advanced the classic strain

paradigm in several ways. First, GST broadened the scope of classic strain (goal blockage)

by including the discrepancy between expectations and real outcomes as well as

unjust/unfair outcomes. The scope of strain was further expanded to include the

presentation of noxious stimuli and loss of positive stimuli. Second, strain was seen as

leading to not only consequent delinquency but also a myriad of negative emotions,

which could also generate delinquency. Third, besides the cumulative effects of strain,

other characteristics were incorporated into GST (duration, recency, and clustering).

Overall, GST states that various strains can make an individual feel bad and want to do

something about it, and whether the ―something‖ is antisocial or conventional depends on

various conditioning factors. GST provides a more comprehensive account of the

strain−delinquency relationship than precedent strain theories. How this theory sustains

empirical scrutiny will be presented in the next section.

Empirical assessment of general strain theory

Studies that test the various theoretical propositions of GST have thrived since its

publication. While this body of literature is substantial, I focus on two empirical core

38
propositions of GST: (1) strains lead to consequent crime and delinquency, and (2)

negative affect mediates the strain−delinquency relationship.

Strain-delinquency relationship

Most research on GST focuses on the relationship between various strains and

consequent crime and delinquency. Early empirical tests conducted by Agnew (1985a,

1989) focused on presentation of noxious stimuli; he found a positive relationship

between aversive family/school environments and various delinquent acts in a male

adolescent sample. Agnew and White (1992) were the first research team to examine the

strain−delinquency relationship in a systematic way. By using a large sample of

adolescents (n = 1,380) from New Jersey, they found that various negative stimuli (e.g.,

negative life−events, neighborhood problems) had positive effects on delinquency and

drug use, whether such relationships were tested longitudinally or cross-sectionally.

Furthermore, this relationship remained significant even when rival theoretical variables

were incorporated into the model (e.g., attachment, delinquent peers).

After Agnew and White‘s (1992) study, the inclusion of negative life-events

became customary as a measure of negative stimuli, and the positive effects of negative

life-events on delinquency and substance use were found in numerous subsequent studies

(Aseltine & Gore, 2000; Aseltine, Gore, & Gordon, 2000; Broidy, 2001; Drapela, 2006;

Eitle, 2002; Eitle & Turner, 2003; Hoffmann & Cerbone, 1999; Hoffmann & Miller,

1998; Hoffmann & Su, 1997, 1998; Mazerolle, 1998; Paternoster & Mazerolle, 1994).

For example, Hoffmann and Cerbone (1999), using growth curve modeling, found that

39
experiencing a relatively high number of negative life-events over time was related to the

―growth‖ of delinquency among adolescents, and these results could be extended to

substance use (Hoffmann, Cerbone, & Su, 2000). Specifically, in adolescents who

experienced increasingly stressful life-events over time, the reported frequency of

delinquency and substance use also increased over time. In addition to negative life-

events, researchers have used other variables to measure negative stimuli, including

neighborhood/school problems (Johnson & Morries, 2008; Paternoster & Mazerolle,

1994), negative interpersonal relationships (Agnew & Brezina, 1997; Mazerolle, 1998),

maltreatment or victimization (Baron, 2004; Brezina, 1998, 1999; Eitle & Turner, 2002;

Harrell, 2007; Hay & Evans, 2006; Robbers, 2004), racial or gender discrimination (Eitle,

2002; Simons, Chen, Stewart, & Brody, 2003; Walls, Chapple, & Johnson, 2007), family

strain (Hay, 2003), and homelessness (Baron, 2004, 2006; Baron & Hartnagel, 1997), and

found these various negative stimuli were related to delinquency.

Studies have also investigated the relationship between failure to achieve goals

and loss of positive stimuli. Robbers (2004; Baron & Hartnagel, 2002) found that goal

blockage had effects on delinquency, and Ostrowsky & Messner (2005) found that

traditional strain (failure to achieve positively valued goals) affected both property crime

and violent crime. In contrast, Baron (2004) found that dissatisfaction about money was

related to property crime but not to other types of crime (e.g., violent crime). Paternoster

and Mazerolle (1994), who examined cross-sectional and longitudinal effects of

traditional strain (limitation of goal attainment) on delinquency, found that traditional

strain was related to delinquency in the cross-sectional model but only weakly or not

40
significantly related to delinquency in the longitudinal model. Finally, Broidy (2001)

used a college student sample to test GST and did not find support for direct effects of

failure to achieve positively valued goals on certain measures of delinquency (e.g.,

property crime, drug use).

With regard to the effects of removal of positive stimuli and of composite strain,

which combines various strains in a single measure, on delinquency, Mazerolle and

Piquero (1998; Mazerolle, Piquero, & Capowich, 2003), using a sample of college

students to test GST, found that removal of positive stimuli affected shoplifting whereas

an unjust strain, in this case an unfair grade, led to fighting but not shoplifting; however,

the effect of removal of positive stimuli became insignificant after controlling for other

variables. The same research team also found that a composite measure of strain had

direct effects on violence only (Mazerolle & Piquero, 1997; Mazerolle, Burton, Cullen,

Evans, & Payne, 2000); in contrast, others found that a composite measure of strain

affected different types of delinquent acts, such as property offenses (Piquero & Sealock,

2000), violence and substance use (Slocum, Simpson, & Smith, 2005).

In sum, a positive and significant relationship between strain and delinquency

appears to exist. Specifically, the studies reviewed above have found that various

stressors (e.g., negative life−event, unjust outcome, victimization) are positively related

to various delinquent acts (e.g., violent behavior, substance use), and this positive

relationship is found in both longitudinal and cross-sectional data. However, traditional

strain (goal discrepancy) usually fails to stand out as an important predictor of

delinquency.
41
 Although these studies generally find supportive results, they reflect some

limitations. First, not all of them examined all three major types of strain together. This

could lead to possible model misspecification. That is, some important strains that are not

included in the statistical model are treated as errors. Second, not all strains are related to

all forms of delinquency. For example, Aseltine et al. (2000) found that family conflict

directly affected juvenile marijuana use but not other forms of delinquency. Jensen (1995)

pointed out that the definitions or types of strains are too broad and hence are

unfalsifiable. Third, the magnitude of strain, such as severity or frequency of a strain, was

not incorporated into these early studies. This is understandable, because despite

Agnew‘s (1992) suggestion that the magnitude of a strain might make it more or less

influential, he did not elaborate much on it. However, he (2001, 2006a, 2006b) later

assigned great importance to these characteristics when he further revised GST.

The mediating effect of Anger

Another main proposition of GST is that strain not only has direct effects on

delinquency, it also has indirect effects on delinquency through negative affect. By

positing this, GST proposes that negative emotions will mediate the strain-delinquency

relationship. Negative emotions in GST include various inner-directed negative emotions

(e.g., depression, fear) and outer-directed negative emotions (e.g., anger), with anger as

the emotional reaction most critical to GST. As a result, empirical research has focused

almost exclusively on anger. Agnew (1985, 1989) found that aversive school and family

environments had significant effects on anger, which in turn had significant effects on

delinquency; and these results held up in both longitudinal and cross-sectional models.
42
Mazerolle and Piquero (1997, 1998) similarly concluded that strain had indirect effects

on delinquency through anger. However, these studies and many others found that anger

only partially mediated the strain-delinquency relationship; that is, strain affects

delinquency both directly and indirectly through anger (Agnew, 1993; Agnew &White,

1992; Aseltine, et al., 2000; Hay, 2003; Hay & Evans, 2006; Mazerolle &Maahs, 2000;

Perez, Jennings, & Gover, 2008; Sigfusdottir, Fakas, & Silver, 2004).

While the above studies found a partially mediating effect of anger, other studies

have found fully mediating effects (Broidy, 2001; De Coster & Kort-Butler, 20006; Ford

& Schroeder, 2009; Sharp, Brewster, Love, 2005). Broidy (2001) found that an unfair

outcome was significantly related to anger, which in turn was related to crime. When

both an unfair strain and anger were in the same model, only anger significantly predicted

crime. Besides the issue of full or partial mediating effects of anger on the strain-

delinquency relationship, studies also indicated some inconsistent results of the

relationship between anger and delinquency. Some studies have found that anger does

have effects on delinquency (Baron, 2004; Hay & Evans, 2006) whereas other studies

only find that anger is only related to outer-directed delinquency (e.g., fighting) (Aseltine

et al., 2000; Capowich et al., 2001). In contrast, Baron and Hartnagel (1997) did not find

any relationship between anger and delinquency (e.g., drug use, violent and property

crime) in their sample.

In sum, studies have usually found support for GST‘s mediating proposition that

anger mediates the strain−delinquency relationship, whether the mediation is a partial or

full mediating effect. However, some mixed results are also reported in the GST literature
43
that indicate that anger does not always have effects on delinquency or only has effects

on certain kinds of delinquency. These results threaten GST‘s credibility because

negative emotions constitute a central element of GST and one that distinguishes GST

from other leading criminological theories. For example, family strain increases crime,

which social control theory will argue is due to low attachment to parents. However, if

the variable affects criminal involvement through negative emotions, such as anger, one

can be more confident that the relationship between strain and delinquency follows

GST‘s theoretical prediction.

Agnew‘s revision of GST

As reviewed above, two general limitations of GST can be summarized. First,

although empirical evaluations generally support the strain–delinquency relationship, not

all strains lead to all forms of delinquency (Mazerolle & Piquero, 1998; Broidy, 2001)

resulting in a criticism that GST is ―unfalsifiable‖ (Jensen, 1995). Furthermore, research

has not specifically evaluated the effect of the characteristics of strain on delinquency. As

studies from the stress literature suggest, the characteristics (e.g., magnitude, duration)

affect which coping strategies one will use (Harnish, Aseltine, & Gore, 2000; Thoits,

1983). Second, empirical studies provide mixed support for the mediating effects of

negative affect on the strain-delinquency relationship. Some studies find that anger is

related to criminal acts (e.g., aggression, fighting) (Agnew, 1985; Baron, 2004; Hay &

Evans, 2006) but not other kinds of delinquency (e.g., non-violent acts) (Baron &

Hartinagel, 1997; Capowich et al., 2001), while others fail to find mediating effects of

other negative emotions (e.g., distress, anxiety) on the strain-delinquency relationship

44
(Aseltine et al., 2000; Broidy, 2001). Furthermore, whether one should expect to find a

total mediating or a partial mediating effect remains unclear. These limitations led

Agnew (2001, 2006a) to revise GST further.

Revised general strain theory retains the central proposition of GST, which is that

strain (the three major types of strain) leads to various negative emotions, which in turn

affect crime and delinquency. However, the revised theory contains two major changes

from this basic model. First, in regard to the problem with strain, Agnew (2001, 2006a)

argues that strain will be most likely to cause crime if the strain is seen as: high in

magnitude, unjust, associated with low social control, and creating an incentive for

criminal coping. As such, forms of strain with these characteristics (e.g., abusive peer

relations, negative school experiences, victimization) are regarding as criminogenic. For

example, criminal victimization is one such strain (Agnew, 2001, 2006a, 2006b) because

it is typically seen as unjust and high in magnitude, sometimes even traumatic (Kilpatrick,

Sanunders, Veronen, Best, & Von, 1987). Moreover, criminal victimization, which

usually occurs in peer groups where supervision is low or absent (Lauritsen, Sampson, &

Laub, 1991), briefly presents a criminal behavior model (Agnew, 2006a). In contrast,

strain that does not have these characteristics is less likely to cause delinquency.

Second, Agnew suggested that researchers should pay more attention to negative

emotions such as depression and fear, which Agnew (2006a) suggests as key negative

emotions, along with anger. Others echo such a suggestion (Capowich et al., 2001).

Agnew argues that what GST proposes is state emotion, not trait emotion. The former

refers to one‘s immediate experience of an emotion while undergoing strains. The latter
45
indicates one‘s propensity to experience a certain emotional states when facing strains

(Agnew, 2006b). While this revision focuses on the role of negative affect in GST,

Agnew (2006a) provides two other routes that explain how strain leads to delinquency.

The first is that strains may temporarily reduce levels of social control, which in turn

leads to delinquency. For example, negative parental treatment may temporarily reduce

an adolescent‘s bond to the parents, and this reduced bond may increase the likelihood of

delinquency. Second, strains may temporarily foster the social learning of crime that may

increase the occurrence of crime. For example, criminal victimization may briefly expose

individuals to a criminal model, which they might imitate. In summary, revised GST

argues that strains lead to crime and delinquency through negative emotions, low social

control, and criminal social learning of crime.

Empirical assessment of the revised GST

Characteristics of strain and delinquency

As the revised GST argues, some strains are more criminogenic than others, and

such strains possess certain characteristics: high in magnitude, unjust, related to low

social control, and association with criminal others and antisocial definitions. Empirical

studies can be divided into two categories. First, studies that test certain criminogenic

strains that were outlined by Agnew (2001; 2006a) generally found support. Specifically,

strains such as victimization or vicarious victimization (Agnew, 2002; Baron, 2009; Hay

& Evans, 2006; Harrell, 2007; Manasse & Ganem, 2009), discrimination (Eitle, 2002;

Eitle & Turner, 2003), unjust/unfair outcome (Mazerolle, Piquero, & Capowich,

46
2003;Piquero & Sealock, 2004), parental rejection (Agnew, 2005; Hay, 2003), child

abuse (Baron, 2004), homelessness (Baron, 2006, 2007), and a negative secondary school

experience (Agnew, 2005; Moon, Hays, & Blurton, 2009) increased the likelihood of

delinquency.

Second, there are only a handful of studies that directly test the effect of

magnitude of strain on consequent delinquency. Slocum, Simpson, and Smith (2005)

found that the best and most parsimonious model for explaining drug use included three

dimensions of magnitude (duration, clustering, and accumulation); only two dimensions

were significant in the model for violence (clustering and duration). They concluded that

there appears to be some redundancy between the various dimensions of magnitude, and

consequently researchers may not need to measure all of these dimensions in future

studies. Other studies have focused mainly on the subjective evaluation of strain6, which

could be seen as measuring the severity of strain. For example, Froggio and Agnew (2007)

found that adolescents from Italy committed more delinquent acts if they considered

school failure and a romantic relationship breakup as more negative to them, compared to

their peers who considered such strains as less negative. However, Botchkovar and

colleagues (2009) found that severity of strain (subjective strain) did not improve the

prediction of subsequent criminal responses among three European samples.

6
However, Agnew (2006a) argued that researchers should distinguish between objective and subjective
strain. The former refers to strains that are disliked by most people in a given group whereas the latter
refers to strains that are seen as aversive by the people who experienced them. However, most measures of
subjective strain asked individuals to indicate how big a problem (negativity) the strain was to them, which
could be used to measure severity. Therefore, in the present study, subjective strain is regarded as severity
of strain.

47
Other negative emotions

Only a few studies have investigated the effects of negative affect other than

anger on delinquent adaptations (Broidy, 2001; Ford & Schroeder, 2009; Hollist, Hughes,

& Schaible, 2009; Jang, 2007; Jang & Lyons, 2006; Kaufman, 2009; Sharp et al., 2005;

Sharp, Terling-Watt, Atkins, Gilliam, & Sanders, 2001; Walls et al., 2007). Some of

these studies have found that negative emotions other than anger increase deviant

behavior. For example, Ford and Schroeder (2009) found that university students who

experienced academic strain (disjunction between academic aspiration and actual

outcome) reported higher levels of depression, and such students were more likely to

engage in non-medical use of prescription stimulants. This particular study also found

that depression fully mediated the strain-delinquency relationship. In another study,

Sharp et al. (2005) found that negative emotions, combining depression, anxiety and guilt,

mediated the effects of strain on eating disorders. Capowich, et al. (2001) found that

negative emotions, such as feeling overwhelmed by life‘s demands, was related to

shoplifting and DUI but not to fighting, and others found depression was related to

substance use (Hoffmann & Su, 1998) and suicide (Walls et al., 2007). In contrast to

studies that test the mediating effect of negative emotions, Brezina (1996) took a different

view to testing the strain-negative emotion-delinquency proposition in GST. Specifically,

GST argues that strains cause various negative emotions that in turn lead to delinquency,

implicitly indicating that delinquency, as a coping strategy, would make the individual

feel ―better.‖ Brezina (1996) found that strain did increase the level of negative emotions

and that delinquency did help juveniles to reduce ―bad‖ feelings.

48
 In contrast, some studies have found null effects of negative emotions other than

anger on delinquency (Aseltine et al., 2000; Hollist et al., 2009 Piquero & Sealock, 2000,

2004). For example, Piquero and Sealock (2000, 2004) reported that depression was not

related to aggression and property crime, and Aseltine et al. (2000) found no support for a

relationship between delinquency and anxiety and depression. Notwithstanding these

mixed results, others who have examined the relationship between specific negative

emotions and specific crime and delinquency, have suggested that inner-directed

emotions (e.g., depression) affect inner-directed delinquency (e.g., drug use, social

withdrawal) more strongly than outer-directed delinquency (e.g., aggression) (Jang, 2007;

Jang & Lyons, 2006).

Another important issue related to negative affect is the different effects of trait-

like and state–like emotions on delinquent coping (Capowich et al., 2001; Mazerolle et al.,

2003). Trait–like emotion refers to one‘s tendency to experience a particular emotion

across different situations, whereas state-like emotion refers to one‘s feelings in a specific

situation. The majority of research has employed trait-like measures, which Agnew

(2006b) argues as responsible for the mixed results of studies of mediating effects.

Capowich et al. (2001), the first to measure the effect of situational anger on subsequent

delinquent adaptations, found that situational anger fully mediated the effect of strain

and other negative emotions (e.g., overwhelming feelings) on intention to fight.

Mazerolle and colleagues (2003) further explored the issue of trait– and state–anger.

They concluded that situational anger was strongly related to shoplifting and fighting

whereas dispositional anger was related only to assault. They also suggested that anger

49
should be measured as situational, not dispositional. Ellwanger (2007) used a different

method to measure state–like anger. He directly asked his subjects directly whether they

felt frustrated or angry when a driving-related strain (e.g., traffic congestion) happened.

He then combined strain with frustration and found this variable significantly affected

driving delinquency.

Other studies that use more situational measures of negative affect other than

anger have found some supporting results (Genem, 2008; Jang, 2007; Jang & Johnson,

2003). For example, Jang (2007) directly asked respondents what they feel when they

experience strain, in order to measure state–emotion. He found that situational distress

increased alcohol use as well as more conventional strategies of coping, such as religious

coping. Genem (2008), using a scenario method, found that situational strain led to fear,

which in turn increased the likelihood of cutting class. However, even with the scenario

method, she did not find that situational depression predicted drug use. A recent study

argued that because depression should be treated as a clinical disorder from the health

perspective, distinguishing between state– and trait–depression is problematic (Manasse

& Genem, 2009). This study suggested that depression measured by means of a clinical

symptom checklist should be regarded as trait–depression. Consequently, the relationship

between strain and trait-like depression is moderating rather than mediating (Manasse &

Genem, 2009).

50
Summary and general limitations of previous studies

GST states that strains/stressors increase the likelihood of negative affect, that

these negative emotions create pressure for correcting behavior, and that crime or

delinquency is only one possible outcome. Whether an individual copes with strain and

negative emotions in an antisocial fashion depends on several conditioning factors

(Agnew, 2001a, 2001b, 2006a, 2006b). GST builds on this theoretical framework by

describing three types of strain and four characteristics of criminogenic strain, negative

emotions, and conditioning factors. It also delineates three mechanisms through which

strain leads to delinquent acts. As reviewed above, empirical evaluation devoted to

assessing GST is substantial. However, three general limitations in this body of literature

require further investigation.

First, although previous studies have directly tested the effects of criminogenic

strains on delinquency, more studies are needed, especially those that test the effect of

unjust strains on delinquency. Equally important is that few studies have incorporated all

the major strains in one model. Doing this may provide insights into how different strains

affect delinquency simultaneously and the interrelationship between them.

Second, only a handful of studies have assessed the mediating effect of negative

emotions other than anger, and the results are mixed. Some studies find support for the

proposed negative emotions-delinquency relationship, but others do not. Still others find

that negative emotions other than anger affect some types of delinquency, which

indicates the possibility of a specific effect. In particular, inner-directed emotions (e.g.,

51
depression), affect only inner-directed behavior (e.g., substance use), whereas outer-

directed emotions, mainly anger, influence only outer-directed delinquency (e.g.,

aggression). This specific effect has not been studied fully. In addition, whether state-

emotion plays a mediating role whereas trait-emotion plays a moderating role remains to

be verified. Furthermore, the theoretical importance of the new pathways through which

strains lead to delinquency deserve more attention (Kaufman, 2009). Only two studies so

far have directly examined these newly proposed mediating effects. Such limitations

hinder further revision or refinement of the GST.

Finally, the most serious limitation is that most of the published studies have

employed samples from the U.S. (Froggio, 2007), which hinders the generalizability of

GST. Although generalizability could be explained as being able to account for various

types of criminal acts, it also connotes the applicability of the theory in different societies

or cultures. The GST process seems to be useful in explaining juvenile delinquency in the

U.S., as reviewed above. Scholars have argued that researchers need to adapt and test the

theory in other countries in order to increase generalizability and foster empirical

development of the theory (Piquero & Sealock, 2000). In addition, Hoffmann et al. (2000)

suggested that studies in other social settings are important for understanding fully the

issues related to stressful experiences (Hoffmann & Su, 1998).

To complicate the matter further, many different cultures exist in the world (e.g.,

Chinese, Latino, African, European culture). So if GST finds support in the U.S., this

may be replicated in other Western cultures, such as European countries (e.g., England)

and Canada, because the general cultural settings are very similar. Indeed, Froggio and
52
Agnew (2007) found support for GST in Italy, and Baron (2004, 2006), who used

Canadian street youth to test GST, also found similar results. However, Eastern cultures

are different in several ways from Western culture. For example, Markus and Kitayama

(1994) pointed out that in the United States, ―it is the emotional states that have the

individual‘s internal attributes (his or her needs, goals, desires or ability) as the primary

referent that are most commonly manifest‖ (p.101). Hence, anger is often caused by other

people who block or prevent individuals to achieve their goals. This assertion is

consistent with strain theory. In a Chinese culture, which focuses on harmony within

relationships and interdependence (Markus & Kitayama, 1991), negative emotions may

be caused by failing to maintain relationships or meet others‘ expectations. The response

to anger may also be different across cultures. For example, Tanzer and associates (1996)

found that even secret criticism of others is considered a manifestation of anger in a

Chinese culture because relationship harmony is so important; hence, even a subtle anger

response should be controlled. In contrast, in America, expression of anger may be

appropriate because it identifies individual needs and maintains identity (Markus &

Kitayama, 1994). Therefore, the influence of culture on GST processes in Eastern

cultures remains to be evaluated.

An even more valuable approach to addressing these limitations is to examine

empirically the GST model in a cross-cultural study, within which an Eastern sample and

a Western sample are included. By so doing, one can not only empirically evaluate the

revised GST but also compare and contrast effects of cultural differences on these issues

and on the GST process. As Kohn (1987) argued decades ago, cross-national studies

53
provide an efficient method for testing, generating, and further developing sociological or

criminological theories. The next section provides a review of culture and the differences

between a Western culture (U.S.) and an Eastern culture (Taiwan). This review also

provides an overview of crime and delinquency among adolescents in Taiwan. Finally, a

section is devoted to review the empirical studies of GST in non-Western countries (e.g.,

Taiwan, China, Korea) and some cross-national studies.

54
 CHAPTER III:

CULTURE, DELINQUENCY, AND GENERAL STRAIN THORY

Agnew‘s (1992, 2001, 2006a, 2006b) general strain theory (GST), which refined

key concepts of classic strain theory (Cohen, 1955; Cloward & Ohlin, 1960; Merton,

1938), has provided a rich framework for analyzing the underlying mechanisms that lead

strained adolescents to deviance. GST has been recognized by scholars as an important

criminological theory (Cullen et al., 2006) and has been used to examine the

strain/deviance relationship. However, it has relied heavily on studies in the U.S. and

other Western societies (e.g., Canada). This is unfortunate because, without comparative

studies, the generalizability of a theory and the validity of interpretations of the results,

based on research and theory from a single nation/culture, are questionable. Moreover,

even though some cross-national7 studies have been completed, these studies are usually

still limited to the same ―cultural frame‖ (Western culture). Cross-cultural studies have

shown many differences between Western cultures, mainly in the United States, Canada,

and some European countries (e.g., England, France), and Eastern cultures, mainly

Chinese culture (e.g., Taiwan, China, Singapore) (Hofstede, 2001; Triandis, 1995).

7
Although the present study uses cross-culture and cross-nation interchangeably, one must always keep in
mind that a nation can accommodate more than one culture (e.g., Native American cultures in the United
States). The present study mainly considers the dominant culture in a nation as representative of that nation,
such as Western culture in the United States and Chinese culture in Taiwan.

55
Consequently, to really test the generality of GST, one must test it across cultural

boundaries.

The present chapter focuses on three topics. First, culture, the important

dimensions of culture, and the differences between Western and Chinese culture are

discussed in order to help define the concept of culture and identify the important cultural

dimensions most responsible for cultural differences in the stress/strain process. Second,

juvenile delinquency in Taiwan, which is used as the sample nation for Chinese culture,

is discussed, focusing mostly on a description of juvenile delinquency trends and current

situations in Taiwan. In addition, some aspects of the cultural and juvenile justice system

background will be introduced. Finally, the current state of cross-cultural studies of GST,

especially those conducted in Asia, and of studies that examine GST in Taiwan will be

reviewed.

Culture and Its Impact on the Stress/Strain Process

Dimensions of culture–individualism and collectivism

Culture is one of the foundations of a society that affects the individual (e.g., how

one views what strain is), as well as the environment (e.g., sources of strain). However,

culture is both too broad and too abstract to be defined definitively. Kroeber and

Kluckhohn (1952) concluded that a consensus definition of culture could not be attained

after they had reviewed substantive literature and found 164 different definitions.

Notwithstanding the abstract nature of culture, Lonner (1994) outlined several common

ingredients in the definition of culture that can be summarized succinctly: (1) culture

56
provides settings within which various human behaviors can occur; (2) culture creates the

potential for individuals to react, and this potential changes over time and place, and (3)

culture contains values, beliefs, attitudes, and languages that emerge as adaptation to the

environment of a group of people (p. 234).Therefore, a definition of culture might be

considered suitable if it contains these important components.

Chun, Moos, and Cronkite (2005, p. 31) stated that a ―system [culture] of

meaning encompasses the norms, beliefs, and values that provide prescriptions for

behavior.‖ Kroeber and Parsons (1958, p. 583) arrived at a similar and cross-disciplinary

definition of culture as ―transmitted and created content and patterns of values, ideas, and

other symbolic-meaningful systems as factors in the shaping of human behavior and the

artifacts produced through behavior‖ (see Laungani, 2004, pp. 15-23 for further review).

In short, the present study will define culture to include shared norms, values, and beliefs

that guide the behavior of members in a group; it serves to distinguish the members of

one group from another (Hofstede, 2001) and plays a central role in affecting individuals‘

ideologies and behavior (Lam, 2007).

Among the many dimensions of culture, individualism and collectivism (Hofstede,

2001) have been recognized as important by scholars in various disciplines. For example,

Parsons (1951), a sociologist, proposed the pattern variable of ―self-orientation and

collectivity-orientation.‖ The former refers to the ―pursuit of private interests‖ and the

latter the ―pursuit of the common interests of the collectivity‖ (p. 60). Hsu (1983), an

anthropologist, not only recognized these two different dimensions of culture but also

contended that the two ideas are primary and defining characteristics of the Western and
57
Eastern worlds. In addition to holding a similar view of individualism and collectivism,

social and cross-cultural psychologists who have devoted a great deal of attention to this

dimension and its impact on human behavior have introduced many concepts similar to

individualism and collectivism, although these concepts relate more to the individual

level than to the cultural level, such as Schwartz‘s (1990) concept of the contractual and

communal society, Triandis‘ (1995) idiocentic and allocentric, Yang‘s (1986) individual-

oriented and social-oriented self, and Markus and Kitayama‘s (1991) independent and

interdependent self-construal.8

Although the definitions and connotations of individualism and collectivism are

varied, common features of these two concepts can be incorporated in a consensus

definition. Individualism refers to a society in which individuals are loosely linked and

are expected to be independent and look after themselves. Such a society places a higher

priority on self, and the individual is the central unit of society; consequently, self-

fulfillment, emotional independence, individual rights, and autonomy are valued.

Individualist societies emphasize ―I‖ consciousness, and members give priority to

personal goals over others‘ goals (Chun et al., 2006; Hofstede, 2001; Triandis, 1995). In

sum, an individualistic society promotes the ―independent self‖ (Markus & Kitayama,

1991), such that people place high value on independence, individual freedom, and

personal achievement.

8
Here, Triandis‘ idiocentric and allocentric, Yang‘s concepts of self, and Markus and Kitayama‘s self-
construal are the presentation of individualism and collectivism at an individual level. All these different
notations essentially deliver similar meanings. The present study will use individualism and collectivism,
and these micro-level concepts interchangeably.

58
 In contrast, collectivism, as a social pattern, refers to closely knit individuals and

strong expectations of mutual support and loyalty. Collectivistic societies are oriented

toward groups (e.g., family, nation), which are the central unit of society. Hence,

obligation, interdependence, and fulfillment of social roles are the focal points.

Collectivist societies stress a ―we‖ mentality, and members are willing to give priority to

the goals of the collective and emphasize group solidarity (Chun et al., 2006; Hofstede,

2001; Triandis, 1995). Collectivistic societies, then, cultivate an ―interdependent self,‖

such that individuals place a high value on cooperation, mutual support, and maintenance

of group harmony.

The major difference between the culture of the United States and that of Chinese

culture (Taiwan) is their different position on the continuum of individualism and

collectivism. The United States, as an individualistic culture, can be documented in

various historical accounts and other scholarly writing (Hsu, 1983; Bellah, Madsen,

Sullivan, Swidler, & Tipton, 1996). For example, Tocqueville (1969) commented that

America is a particularly ―individualistic‖ culture (as quoted in Bellah et al., 1996).

Similarly, Bellah et al. (1996) stated that ―individualism lies at the very core of American

culture‖ (p. 142). In contrast, Chinese culture has been described as a particularly

collectivistic culture (Ho & Chiu, 1994; Leung & Bond, 1982). Hofstede (2001), the first

researcher to systematically and empirically investigate the dimensions of culture

internationally, analyzed over 116,000 questionnaires collected across 53 countries and

identified four dimensions in which the cultures differed. Specifically, scores on

individualism for the United States (91) were the highest among all 53 nations, whereas

59
Taiwan‘s score (17) was lower than the score of all but nine other countries. The scores

for the United States and Taiwan also differed from each other for the three other

dimensions: power distance, uncertainty avoidance, and masculinity vs. femininity9. In a

nutshell, Taiwan, as a representative of Chinese culture, is more collectivist than the

United States, according to Hofstede‘s empirical research.

Besides the above discussed differences between Taiwan and the United States on

the individualism-collectivism continuum, the difference between these two countries is

also manifested in various psychological concepts. For example, King (1981) maintained

that an individual is treated as a psychological being in Western culture (United States),

which is more individualistic, and as a social being in Chinese culture, which is primarily

collectivistic. Similarly, Gabrenya and Wang (1983) found that Chinese from Taiwan and

Hong Kong are more likely than their American counterparts to endorse group-oriented

self-concepts. This particular result, along with others (Offer, Ostrov, Howard, &

Atkinson, 1988; Triandis, 1989), reveals the difference in ideological self between

Chinese and Western culture, which can be related to the cultural differences on the

individualism-collectivism continuum. King and Bond (1985) concluded that the

enduring prototype of the Chinese is the sense of belonging; in addition, Wilson (1970)

pointed out that group loyalties and the idea of loyal behavior differentiated Chinese

people from Westerners. In contrast, as Kim and Choi (1994) indicated, individuals in the

United States were strongly encouraged to separate from their ascribed relationships

9
In the present study, individualism versus collectivism is the focal point; however, there are three other
dimensions of culture, based on Hofstede‘s study: power distance, uncertainty avoidance, and masculinity
vs. femininity (the interested reader is referred to Hofstede, 2001).

60
(Bellah et al., 1996), such as family and relatives, and were encouraged to form other

relationships based on common goals and interests (e.g., accumulation of wealth).

In addition to the aforementioned differences in the impact of individualism and

collectivism, they also differ in their impact on how individuals interpret their ―ultimate

need,‖ self-actualization (Maslow, 1943). For example, Yang and Lu (2005) listed three

major differences between more individual-oriented and more social-oriented self-

actualization. They argued that for individuals from Western cultures, particularly in the

United States, self-actualization focuses on the internal-personal self; people want to

enhance personal potential and characteristics that increase individuality and autonomy,

and achieve personal rights and an egalitarian society. In contrast, for individuals from

Eastern cultures, such as Taiwan, self-actualization emphasizes the social-relational self,

self-cultivation, and self-improvement with regard to moral and personal skills, which

lead people to realize their obligations in relationships in society.

In conclusion, since the concept of individualism and collectivism were derived,

empirical research has thrived. The body of literature focuses largely on how

individualism-collectivism impacts an individual‘s behavior cross-culturally. A

systematic meta-analysis conducted by Oyserman, Coon, and Markus (2002) reached

several conclusions regarding the influence of individualism and collectivism on an

individual. First, although the U.S. is often singled out as more individualistic than most

countries, such as African and Latin American countries, such a stereotype must be

viewed with great caution. Second, Americans emerged as high in individualism and low

in collectivism, and the differences between Americans and Chinese (e.g., Taiwanese)
61
were large. The authors further suggested that the difference between the U.S. and

Chinese cultures on the individualism and collectivism continuum is more pronounced

than the differences between other countries (e.g., African countries). Based on this

discussion, this distinction is both theoretically and empirically sound. Nevertheless,

although Taiwan and the U.S. may be meaningfully separated through individualism and

collectivism, Confucianism ideologies, a salient cultural heritage unique to Chinese

culture, are also important. The Confucian philosophy not only provides a deeper

understanding of Chinese culture, but also provides a clearer view of the differences

between the United States and Taiwan.

Confucian ideology and its influence

Confucian ideologies are closely related to collectivism and have been rooted in

Chinese culture for centuries. The importance of the Confucian ideology in understanding

Chinese culture in general, and Taiwanese culture in particular, can be found in several

accounts. Bond and Hwang (1986) stated that center stage in almost all approaches to

Chinese social behavior is occupied by the teachings of Confucius. Gallois et al. (1996)

maintained that Confucianism provides the backdrop for the emergence of interpersonal

relationships, self-concepts, and communication styles. Hofstede and Bond (2001) argued

that the unique Eastern cultural dimension is Confucian dynamism. The present section

deals with the significant impact of Confucianism on Chinese societies, a discussion that

is suitably applied to Taiwan. Zhang (2003) argued that Taiwan is strongly influenced by

the philosophy of Confucianism because most residents in Taiwan, regular citizens as

well as those in government, are descendants of people from traditional Chinese culture.
62
The present review will focus on the immediate implication of the Confucian ethos for

Chinese social behavior.

A fundamental Confucian assumption is that man exists in relationship to others

and that harmony is the most treasured social value (Bond & Hwang, 1986; King & Bond,

1985). Moore (1967) further stated that in Confucian social theory, an individual is never

conceived as an isolated unity; rather, he or she is treated as an interactive being. The

proper way to maintain a relationship and to achieve harmony is prescribed by the dictate

li (propriety), a set of rules on how to interact properly in daily life. The implication of li

in daily life is exemplified in wu-lun, which delineate the proper interaction rules for five

cardinal relationships—those between sovereign and subject, father and son, elder and

younger brother, husband and wife, and friend and friend. Harmony is realized if each

member in a dyad conscientiously follows the requirements of li. Consequently, Chinese

people are commonly known to be peaceful and submissive, because of the emphasis on

relational harmony, and each individual in a relationship is entitled to both rights and

responsibilities. For example, parents receive their children‘s reverence and obedience; in

return, they provide love and meet their children‘s needs. These relational rules echo

Hofstede‘s (2001) results—Taiwan scored higher than the United States on ―power

distance.‖ As relational harmony is conceived as cardinal, it is no wonder that a

fundamental child-rearing practice among the Chinese, in general, is to teach youngsters

to inhibit emotional expression of hostility toward others (e.g., authority figures) and, by

extension, behavioral expression of aggression (Ho, 1986, 1996). This same

confrontation-avoidance socialization pattern has also been observed by scholars in

63
Taiwan (Ho, Chen, & Kung, 2008; Yang, 1995). For example, Yang (1995) commented

that the importance of relational harmony was instilled during early socialization, so that

maintaining social harmony was deeply internalized in the individual.

An extension of the relational rule is the emphasis on filial piety (Xiao), which

dictates intergenerational relations in the family. The family is considered the basic

functional unit in the Chinese culture (Yang, 1995). The central meaning of filial piety

for Chinese is to take care of parents, and on some occasions including senior extended

family members. According to Taiwan Civic Law, it is the children‘s responsibility to

support their parents, regardless of whether parents can support themselves.

To fulfill the mandate of filial piety, one not only takes care of parental needs, both

economic and emotional, fulfilling obligations and showing an attitude of love and

reverence, but also maintains the parents‘ ―mian zi” (face), which means one should

diligently pursue and maintain success in one‘s career–that is, success in the outside

world (Lin & Liu, 1999)–to bring honor rather than disgrace to the family name (Ho,

1994; Lin & Lin, 1999). Another way to show filial piety is to submit oneself to parental

wishes, especially the wishes of the father. Consequently, sacrificing one‘s own goals and

replacing them with familial goals is not uncommon in Chinese society. As a result, Ho

(1996) recognized that one of the characteristics of filial piety is acceptance of

hierarchical ranking and authority.

Filial piety, as an important moral foundation of the family, also influences other

interpersonal relationships. For example, certain significant social relationships, such as

64
master (teacher) and apprentice (student), operate on a simulated father-son basis (Lin &

Liu, 1999). Hence, students show similar reverence and attention to masters or teachers

as they would if the teachers were their biological parents. One famous Chinese saying

summarized this mentality nicely: ―yi ri wei shi zhong shen wei fu‖ (a teacher for one day,

a father for one‘s entire life).

Finally, the focus on educational attainment affects individuals‘ daily behaviors

and differentiates Chinese culture from Western culture. It is widely recognized that

Chinese parents attach great value and importance to education and academic

achievement (Ho, 1986; Sollenberger, 1968). Educational success allows one to pass civil

service exams, which, in turn brings honor and glory to one‘s family, a fulfillment of the

filial piety mandate. In addition, educational excellence allows one to climb the social

ladder and achieve a successful political, social, vocational, and family life (Gates, 1987;

Yang, 2004). Hence, educational attainment not only helps one to secure a share of the

limited social resources but also fulfills the responsibility of filial piety, bringing the

family mian zi. Shek and Lee (2007) found that Chinese parents place great emphasis on

their children‘s academic performance, and believe that academic achievement is very

important. Furthermore, parents are willing to invest greatly in their children‘s education

and help them attain the highest education level possible (Ho, Chen, & Kung, 2008; Yi &

Wu, 2004).

In sum, Confucian ideology affects Chinese society fundamentally, not only in the

daily relationships regulated by his philosophy, but also in the extended aspects of

65
personal life, such as career decisions. The close relationship between the Confucian

ethos and collectivism is a factor in the cultural difference between Taiwan and the U.S.

The impact of culture on stress/strain

Because self-achievement, personal rights, and autonomy are the primary values

in an individualistic society, struggles in these areas in life may be stressful in that society

(Chun et al., 2006). In contrast, individuals from a collectivistic society, where relational

harmony and interdependence are more important, may see problems related to pleasing

parents and fulfilling family goals as more stressful. Compared to American college

students, Asian students (Hong Kong Chinese, Korean, and Japanese) have reported

higher needs for affiliation and more sensitivity to social rejection (Hui & Villareal, 1989;

Yagmaguchi, Kuhlman, & Sugimori, 1995). Heine and Lehman (1995) found that

Japanese college students, members of a collectivistic culture, considered interdependent

events (e.g., ―Sometimes in the future you will do something that makes your family

ashamed of you‖) more stressful, and independent events (e.g., ―After growing old, you

will find that you never realized your most important dreams‖) less stressful, compared

with Euro-Canadian college students.

In addition to the impact of culture on an individual‘s perceptions of stress/strain,

differences in parenting are pronounced between individualistic cultures and collectivistic

cultures. Whereas Chinese culture, as mentioned earlier, stresses obedience and proper

relationship rules, Western culture focus not on these matters, but rather on individual

rights and self-realization. Authoritarian (strict and controlling) parenting has been

66
determined to have a negative impact on school performance for U.S. children, but is

related to a higher level of school performance for Chinese students (Steinberg,

Dornbusch, & Brown, 1993). It is very likely that authoritarian parenting is considered

illegitimate and unfair by American children but is perceived as showing love and

concern by Chinese children. Hence, stern parental rules or discipline may be seen as a

strain in the United States, whereas this same discipline may not be considered as a strain

or at least be considered a less serious strain, in a more collectivistic culture or a culture

with Confucian heritage, such as Taiwan. Consequently, coping behaviors in responses to

such impacts may vary between these two countries, as found by Steinberg et al. (1993).

Differences between individualistic cultures and Chinese culture may also be

found in interpersonal relationships. It should be clear by now that relational harmony is

important in Chinese culture; as Yang (1995) pointed out, Chinese people are prone to an

―other‖ orientation. That is, Chinese people worry about others‘ opinions, strongly

conform to others, care deeply about social norms, and have a high regard for reputation.

Students in the United States, especially graduate students, are encouraged to express

their own ideas, criticize the ideas of others, and actively participate in class (Cross,

1990); hence, standing out is generally not regarded as stressful or undesirable. In

contrast, Chinese students are often quiet and passive learners, and conforming to others

is expected and desired (Yang, 1995). Hence, by extension, American students may not

consider criticism from others as a strain that leads to negative emotions such as

resentment, because it is a common practice. However, such criticism may cause strain

67
and consequent negative emotions in students from collectivistic cultures, such as

Chinese students.

In addition, cultural differences have also been observed in influencing

communication style. Empirical studies have shown that people from collectivistic

cultures are more likely to use harmony-enhancing procedures—negotiating and

complying—to deal with conflict processing; in contrast, individuals from more

individualistic cultures are in favor of direct confrontation (Leung, Au, Fernandez-Dols,

& Iwawaki, 1992; Leung, & Li, 1990). Although direct confrontation may make an

interaction more stressful, it solves the problem more directly, making the strain

relatively short lived. In contrast, harmony-enhancing procedures may reduce strain at the

time of interaction, but they probably prolong the process and make the state of stress

relatively long lived.

While cultural differences may result in different definitions of strain, they may

also lead to similarities. In a collectivistic society, a verbal insult from an outside group

member may be sanctioned, based on a ―mind-your own business‖ rule (Bond, Wan,

Leung, & Giacalone, 1985); hence, such an insult is likely to cause conflict and strain. A

similar verbal insult in a more individualistic society invades personal identity, treasured

in such societies, and these may also cause strain and conflict. Therefore, the same

incident in different cultural settings may have the same results although the mechanisms

or underlying meanings may be different.

68
 While cultural differences affect individuals‘ perceptions and appraisals of the

meaning/stressfulness of a particular incident, these differences also generate unique

strains to the members of these cultures. As mentioned earlier, the Confucian heritage

deeply influences Chinese society, especially the emphasis on education, which creates

relatively high respect and high prestige for teachers at various levels of education.

Therefore, examination-related strain and the harsh, sometimes seemingly abusive,

discipline imposed by teachers is unique to students in Eastern cultures (e.g., Taiwan,

Korea). These education-related strains and subsequent pathological states in juveniles

have been documented in Korea (Morash & Moon, 2007; Moon, Blurton, & McCluskey,

2008), China (Bao, Haas, & Pi, 2007), and Taiwan (Li & Chiang, 2001; Xu & Hwang,

2004). For example, Li and Chiang (2001) reasoned that many juvenile delinquents may

be ―victims‖ of the educational system, which focuses too much on performance and

ignores other important issues, such as the student‘s well-being.

Besides the above reviewed studies that provide direct evidence of the differences

between individualistic and collectivist cultures, studies that obtained indirect evidence

by evaluating the well-being of individuals have also helped to reveal cultural influences

on strain. Oishi‘s (2003) cross-national study indicated that in more individualistic

cultures, such as the United States, autonomy is emphasized; hence, the relationship

between autonomy and life satisfaction is stronger in individualist nations than in

collectivist nations. By extension, failure to fulfill the goal of autonomy leads to an

unsatisfying life, which is stressful. To cope with such strain, according to GST,

delinquency is more likely if autonomy has not been achieved (Agnew, 2006a).

69
The impact of culture on negative emotions

Emotional states and the expression of particular emotions are heavily influenced

by culture. Although some scholars have identified a set of universal emotions (Ekman,

1999; Plutchik, 1980) and emotional responses (Mesquita & Frijda, 1992), the linguistic

concept of an emotion, the antecedent of an emotion, and the expression of an emotion

differ between cultures (Mesquita & Frijda, 1992; Russell & Yik, 1996). For example, in

the Chinese culture, with its emphasis on social harmony and collectivism, expression of

emotions, especially negative emotions, is prohibited or suppressed (Kleinman, 1986).

Tanzer and colleagues (1996) found that anger responses are different between

Singaporean Chinese women and their counterparts in Western societies. The expression

of anger is prohibited or censured (Bond & Hwang, 1986) in Chinese culture; the

maintenance of group harmony is so highly valued that even ―secretly and quietly

critiquing others‖ is considered anger expression.

Cultural influences with regard to negative emotions are not limited to expression

of such negative emotions; they also affect how individuals feel in response to the same

incident. In collectivistic cultures, communication is high-context and expressed

implicitly; alternatively, communication is more direct and low-context in individualistic

culture. As a result, training in the United States is more likely to focus on the speaker

and speaking skills (e.g., how to deliver an idea clearly), and unskillful communication is

likely to cause negative emotions (e.g., anxiety). Training in collectivistic cultures aims at

the receiver or audience of the communication; misunderstanding on the part of the

receiver leads to embarrassment or social criticism (Triantis, 1994, p. 185). Hence, the
70
same act, communication, may be a source of anxiety for the speaker in the United States,

whereas in Chinese culture it is more likely to lead the perceiver to feel some stress and

negative emotions.

As previously mentioned, authoritarian parental discipline may be considered a

strain in the United States, but not in Taiwan; this difference in response to a parenting

style may also cause different emotions in the two cultures. For example, one of GST‘s

assertions (Agnew, 2006a) is that authoritarian (e.g., harsh) discipline leads to

delinquency through increasing anger. In contrast, one in-depth study (Fung & Chen,

2001) found that Taiwanese parents explicitly and implicitly use shame to morally

educate their children. By extension, this type of authoritarian discipline in Taiwanese

culture will more likely increase the feeling of shame rather than anger, which may lead

to a different kind of response. Therefore, while authoritarian parenting may very likely

cause anger in the U.S., it is likely to elicit shame or other negative emotions in Taiwan.

From an independent-self perspective (Markus & Kitayama, 1991), self is more

central, and expression and realization of internal and private attributes are the goals. For

example, in the United States, emotional states that have the individual‘s internal

attributes (e.g., one‘s needs) as the primary referent are more commonly manifested. The

typical example from strain theories, both GST and classic strain theory, is that the

blockage of goals is one cause of negative emotions, such as anger or frustration, and that

expression of such negative emotion is not undesirable because it highlights the

individual‘s needs and internal attributes and is consistent with the cultural framework

(Markus & Kitayama, 1994). In contrast, in the cultural framework of Eastern Asia, the
71
goal is the alignment of one‘s reactions and actions with others (relational harmony). The

most common negative emotions, such as anxiety, shame, or even depression are more

likely to be related to relationships or a faltering of interdependence (Makus & Kitayama,

1994; Yang, 1995). For example, Yu (1996) argued that for the Chinese, failure to

achieve a goal is usually blamed on the self, which is more likely to lead to negative

emotions, such as anxiety, depression, or guilt.

The impact of culture on coping strategies

In addition to affecting the perception of stress/strain, consequent negative

emotions, and expression of these negative emotions, culture is also related to how the

individual copes with strain and negative emotions, because appropriate ways to cope and

the resources one can draw upon, whether from others or oneself, are all culturally bound.

The collectivistic culture and the Confucian heritage of Chinese society lead members of

such societies to employ more interpersonal resources and fewer individual resources to

cope with stress. In contrast, in more individualistic cultures, individuals are more likely

to rely on the self. Indeed, Mu (1991) determined that Taiwanese adults are more likely

than their U.S. counterparts to employ interpersonal social support to cope with stress.

The goal of coping may also be different in different cultural settings. Chun and

colleagues (2005) suggested that in individualistic cultures, the primary goal of coping

with conflict is to remove the barrier to a desired outcome, and to assert individuality and

autonomy. In contrast, for an individual with a collectivistic orientation, the goal is to

72
manage conflict in such a way that no one is shamed and interdependence is reinforced

and strengthened.

Individuals from individualistic cultures, compared to collectivistic cultures, have

higher individual strengths or resources, such as an internal locus of control or self-

esteem, and these in turn may lead to different coping strategies and outcomes. For

example, alcoholic patients from the U.S. who use a behavioral approach to coping

manifest lower severity of alcoholic problems (Chung, Langenbucher, Labouvie,

Panadina, & Moos, 2001) because they believe that they are in control and trust their own

abilities. These researchers also found that the patients who used fewer cognitive

avoidance coping strategies had fewer interpersonal and alcohol related problems. In

contrast, members of a collectivistic society are more likely to use avoidance strategies,

although these coping strategies are not associated with maladaptive outcomes because of

the belief that one is not in control (Chang, 2001; Yoshihama, 2002).

Self-esteem or mastery is usually called upon when one is under strain. As the

GST and stress literature theorize, these personal characteristics can help to reduce the

negative impact of strain on individuals‘ well-being (Agnew, 1992; Perlin, 1989).

However, this may be true in predominantly individualistic cultures, where self is the

focal concern, so that low self-esteem and a lack of the perception of control may

increase the likelihood of maladaptation, such as crime. In more collectivistic cultures,

self-esteem is different; some researchers (Lu & Yang, 2006) have argued that the self-

esteem of Chinese people may be group self-esteem, so that the Western concept of self-

esteem may not lead to positive coping behavior, whereas group-oriented self-esteem
73
may do so. Also, while a sense of control may be important, Confucian thinking

emphasizes self-cultivation and virtue; therefore, low mastery may promote positive

coping in the Chinese culture because it fosters perseverance.

Yu (1996) and Yang (1996) concluded that in Chinese society, success is usually

attributed to others and failure to self, whereas the opposite was more likely among

Westerners. Therefore, Chinese people are more likely to use self-directed coping

strategies, such as self-improvement or self-blame; on the other hand, Westerners are

more likely to use other-directed coping behaviors, such as escaping the situation or

creating a new standard (Cohen, 1955; Cloward & Ohlin, 1960). For example, Heine and

associates (2001) found that failing an experimental task is more likely to instigate

Japanese students‘ corrective efforts for self-improving– that is, they tend to work harder

on the same task. In contrast, Canadian college students are more likely to pay more

attention to the experimental task at which they succeeded. The cause of the opposite

result between these two cultures is that in a collective culture mian zi, or respect from

others, is very hard to obtain; hence, instead of working to achieve something, students

are motivated to work hard to improve their own shortcomings, so that their deficit will

not jeopardize the status quo of their group or family.

In sum, although the cultural differences between the United States and Taiwan

are many, they can be understood at least in part through individualism and collectivism,

as well as Confucian ideology. Moreover, these differences manifest at each stage of the

strain process. This section‘s discussion of the influences of culture on the strain process

provides a systematic method of understanding the complicated relationships between

74
culture and the strain process. The next section will review juvenile delinquency in

Taiwan and certain background aspects of Taiwan.

Juvenile Delinquency in Taiwan

Taiwan—an overview

Taiwan is located about 100 miles off the coast of southeastern China. Shaped

like a yam, the island is about 247 miles long and 90 miles at the widest point and has a

total area of about 14, 630 square miles. The total population is about 23 million. As of

2010, the population was largely composed of Chinese, numbering 23,162,123 (98%),

and only 512,701 aborigines (Directorate-General of Budget, Accounting, and Statistics,

Executive Yuan, R.O.C., 2011). Because of the population composition, Chinese culture

is the dominant culture in Taiwan.

For the past five decades, Taiwan has experienced great economic progress.

Major economic growth occurred in the 1970s, when annual rates of growth averaged

13.35%. It stabilized around 6.5% after the 1980s. The most significant change was the

average annual per capita income, which was around $500 before 1970s but which rose

to $16,432 during 2010 (Directorate-General of Budget, Accounting, and Statistics,

Executive Yuan, R.O.C., 2011). In addition to rapid economic growth, Taiwan

experienced political and social changes. For example, until the 1990s, when political

liberalization occurred, Taiwan was under strict government control as a police/military

state (Chu, 2000). During this long period of change, the family structure, considered the

75
foundation of Chinese culture, also underwent some changes. The double-income family

with only one or two children is the main prototype in modern Taiwanese society.

Juvenile law in Taiwan

In Taiwan, juvenile delinquents include those who are between 12 and 17 and are

subject to ―Laws and Regulations Concerning the Management of Juvenile Matters‖

(Juvenile Law, 1981). Juvenile Law is intended to regulate not only delinquents, those

who commit crimes, but also ―potential offenders‖ (similar to ―status offenders‖ in the

United States), those who are situated in an environment where future offending is likely.

Juvenile offenders are under the jurisdiction of the juvenile courts that operate at the local

level, and the procedures are similar to those of adult courts, although juvenile trials are

not open to the public (Chu, 2000).

According to the Juvenile Accident Act, two types of punishment can be given to

juvenile offenders. The first type is mainly to ―protect‖ youth, who are involved in minor

offenses or who are considered ―potential offenders‖: admonishment, weekend individual

or group counseling, community service, protective or probation control measures, and

reformatory school. Most judges are willing to sentence juveniles to these

services/programs of punishment because the philosophy of the juvenile law is that it is

better to teach juvenile delinquents rather than to punish them (Ho et al., 2008). This

―teaching rather than punishing‖ mentality is consistent with Confucian philosophy. The

second type of punishment, which is more serious and based on criminal law, includes

imprisonment, fines, and detention. Juvenile offenders in Taiwan cannot be sentenced to

76
death or life imprisonment, unless the youth committed homicide of a lineal relative (e.g.,

biological parents or grandparents) (Juvenile Law, 1981). In addition to juvenile law, the

Child and Youth Welfare Law, established and enforced in 2003, is designed not only to

protect children (ages 0 to 11) and juveniles, but also to regulate and establish proper

social welfare institutes. Such laws also emphasize parental responsibility for disciplining

their children10.

Within the current law enforcement structure, each police department has a

juvenile corps unit, and specific projects or guidelines are established in city and county

police departments, often known as the ―Juvenile Guidance Section.‖ The juvenile corps

unit in each police department is responsible for both enforcing Juvenile Law and

providing related services, such as referral to other resources and programs for enhancing

parent-child relationships (Introduction of Juvenile Guidance Sections, 1991).

Juvenile delinquency in Taiwan

Juvenile delinquency became a serious problem in Taiwan during the period of

rapid economic growth and social change, which included increased industrialization and

change in family structure. Although juvenile delinquency has increased during this

period (Chu, 2000, p. 211, Table 13.6), the number of juvenile offenders has decreased in

recent years (Xu, 2005, p. 267, Table 1). Indeed, recent official crime statistics indicate

that the total number of juvenile offenders/suspects was about 18,145 in 2003, but this

10
For details about laws and regulations related to child and juvenile welfare, the reader is referred to Kuo
and Wu (2003).

77
number decreased to around 11,283 in 2008 (National Police Agency, MOI, 2009) (see,

Table 1). Compared with other countries, Taiwan‘s juvenile delinquency is very low. For

example, juvenile offenders account for 9.9% of all criminal offenders in Taiwan in 2000;

the figure is 37.4% for Japan and 17.1% for the United States (National Police Agency,

MOI, 2002).

While recent official reports indicate that the number of juvenile delinquents is

decreasing, the problem of juvenile delinquency has not disappeared. For example, a

public poll conducted in Taiwan in 2002 revealed that the most worrisome serious crime

problem is juvenile delinquency (27%). Another government report painted the same

picture—Taiwanese citizens regarded juvenile problem/delinquency as a serious social

problem (Research, Development, and Evaluation Commission, Executive Yuan, 2001).

Table 1 Juvenile Offender/Suspect Situation from 2004 to 2009

Juvenile delinquency
Population of juveniles* Juvenile offenders**
rate (per 10,000)

2004 1,931,153 10,540 55

2005 1,948,681 9,620 49

2006 1,930,184 10,384 53

2007 1,944,062 10,881 56

2008 1,931,654 11,283 58

2009 1,936,831 10,762 55

*Juvenile population is based on age 12 to age 17 (Directorate-General of Budget,

Accounting, and Statistics, Executive Yuan, R.O.C., 2011)
**Juvenile suspects/offender are between age 12 to age17 (National Police Agent, MOI,
2009).
78
A more academic survey showed a similar result in that ―juvenile delinquency‖ was

ranked among the top five most serious social problems during 1985-2001 (Taiwan

Social Change Survey). The concern is not without empirical support. The early (1990s)

studies revealed that juvenile delinquency was on the increase, reaching an

unprecedented high in 1993 (Copper, 2003; Chen & Chen, 2000; Chu, 2000; Selya, 1995).

And, as recent scholarly investigations have indicated, while juvenile delinquency seems

to have decreased in recent years, the types of juvenile delinquency have changed. Chen

and Chen (2000) argued that, although the number of juvenile delinquency suspects

declined from 1990 to 2000, assaults and robberies more than doubled. Other scholars in

Taiwan have also revealed that juvenile delinquency has become not only more violent,

but also more prevalent (Hou, 2003; Kuo & Wu, 2003; Xu, 2005) than before.

Among juvenile delinquency acts, three are particularly serious and draw the most

attention from scholars and the general public in Taiwan—violent crime, sex and Internet

deviance, and substance use/abuse (Chen & Chen, 2000; Crime Investigation Bureau,

2008; Hou, 2003; Lee et al., 2009; Wu, 2001; Xu, 2005). Although total juvenile crime

has decreased, the violent crime11 rate increased from 15% of total crime in 1995 to 25%

in 2004 (Xu, 2005, p. 269, Table 3). This increase in violent crime is mainly due to

assault and an ―offense against sexual autonomy‖ (Ministry of Justice, 2008; National

Youth Commission, Executive Yuan, 2005). Two phenomena are related to these

increases—joyriding and school bullying. In Taiwan, motorcycles are popular because of

11
In Taiwan, a violent crime included assault, homicide, robbery, intimidation, and abduction. After 1999,
―offense against sexual autonomy‖ was included.

79
their mobility, swiftness, convenience, and economy, which make them very suitable for

traveling around dense cities and narrow streets. Almost every household has at least one

motorcycle, which results in Taiwan having the highest number of motorcycles in Asia

(Dai, 2005). Because motorcycles are easily available and fast, street motorcycle racing,

which provides an exciting experience, is appealing. In many cases, juveniles who are

involved in street motorcycle racing are in a small group (Xie, 1998), most of whom

clams that the small group protects them and most of whom carry weapons (e.g., knives,

bats).12 As a result, violent attacks on rivals and even on innocent bystanders have

become a serious social problem (Xie, 1998). For example, recently several joyriding

juveniles randomly attacked and killed a young man in a park and similar incidents have

commonly appeared in the news (see Xie, 1998, for a news review).

In addition, violence related to school bullying has increased and drawn attention

in recent years. According to the Campus Security Report Center (2006), compared with

previous years, in 2005, single assault incidents, which does not include group fighting,

increased over 1.1 fold; other violence and delinquency increased about 1.3 fold, and

intimidation and extortion increased over 2 fold. Another national self-report study

revealed that 30% of junior-high school students (grades 7-9) reported they had been

involved in some conflict with teachers or had fought with other students, and the figure

is much higher, 50%, for senior-high school students (grades 10-12) (National Youth

Commission, Executive Yuan, 2005). Other self-report studies have pointed out that large

12
In Taiwan, firearms are illegal; hence, most criminals, adults or juveniles, usually carry knives and/or
bats as major weapons. However, illegal gun ownership is also common among adult criminals, especially
those in organized criminal groups or gangs.

80
numbers of students have either used violence against others or have been victims of

violence on campus (Crime Investigation Bureau, 2008).

As mentioned earlier, the culture in Taiwan is highly influenced by the Confucian

ethos, which emphasizes educational success and respect for teachers and other authority

figures. Teachers and schools may be conferred a high level of authority to educate

students, which sometimes involves physical punishment; this can easily lead to

increased conflict between students and teachers. Moreover, as educational achievement

is emphasized in Taiwan, students who do not do well in classes are usually marginalized

in school, and many scholars argue that such marginalization is a major cause of school

violence and other forms of delinquency (Hou, 2003; Lee, 1998; Xu, 2007). Moreover, as

a result of the heavy focus on academic achievement, school curricula are mainly

designed to reach such goals; hence, other activities less related to academics (e.g.,

physical education, music education) are relatively underdeveloped. This makes school

education unattractive and sometimes burdensome, which could contribute further to an

increase in campus violence and delinquency (Lee, 1998; Zhou, 2001).

Youth gang problems, which have become more serious in recent years, are also

responsible for the increase in juvenile violent crime (Hou, 2003; Xu & Xu, 2000).

Although youth gangs are not new in Taiwan, since the middle-1990s, they have become

a social concern and a threat to many juveniles‘ safety. This problem is mainly due to

adult criminal organizations extending their influence into schools by recruiting students

from junior and senior high schools and marking their turf in these schools (Chai & Yang,

1999). A recent in-depth interview-based study indicated these gangs recruit students into
81
their groups because youth gang members are perceived to be ―low cost,‖ loyal, brave,

and are likely to be given more lenient punishment by the law (Xu & Xu, 2000; Yang,

2004). Students who join gangs are looking for protection, friendship, and fun, and

school drop-outs also consider gangs as a source of income (Xu & Xu, 2000). If youths

join gangs mainly for protection, this implicitly indicates that gang members not only are

seldom the target of criminal attacks, but also are ―immune‖ from retaliation. Yang (2004)

pointed out that the Confucian doctrines, which demand obedience and educational

attainment, clash with the modernized (Westernized) social perspective, and this creates

tensions between youth and social units (e.g., family, school). Students who struggle with

these tensions are more likely to be rejected by family and school (school failure is

commonly considered a shame to the family), so that often adolescents are left with few

options but to join gangs that provide support and acceptance (Yang, 2004).

Taiwan has a high population density, and students typically have long school

hours that include regular school hours (10 hours) and cram school hours (3 hours). Yi

and Wu (2004) described students as highly stressed and strained. With insufficient

outdoor recreational facilities, a result of Taiwan‘s high technology industries, and

because of the influence of Japan,13 several popular indoor entertainment businesses were

born—cyber cafés and KTV. These two popular indoor leisure activities have been

labeled as ―crime-prone‖ places by society, and research has found that over 75% and

13
Yi and Wu (2004) stated that Taiwanese youth admire the Japanese culture and this ―involves both
cultural dissemination and cultural assimilation‖ (p. 233).

82
54% of juvenile delinquents indicate they often go to cyber cafés and KTV, respectively,

for fun (Zhou, 2003).

In cyber cafés, which provide high-speed Internet access and well-equipped

computers, youth can enjoy online games and browse the Internet. According to a non-

profit organization‘s national survey on youth, 51% of Taiwanese youth have been to a

cyber cafe and 45% go to a cyber café once a week (Tosun Non-profit Organization,

2001). Because there are no adults monitoring how youths use the computers in cyber

cafés, these cafés can become blind spots where adolescents congregate (students or

drop-outs) and become involved in various deviant acts (Wu, 2001; Tosun Non-profit

Organization, 2001). Although the most common activity is playing online games, over

25% of the survey respondents reported committing deviant or delinquent acts, such as

prostitution,14 browsing erotic websites, or buying pirated goods (e.g., CDs, DVDs).

Besides these types of online deviant behaviors, adolescents who linger in cyber

cafés often become involved in related delinquencies, such as drug offenses. And,

because of the anonymous nature of online dating and chatting, these activities provide a

fantasyland for adolescents who are eager to make friends with opposite sex individuals,

whose teachers and parents commonly believe that cross-sex friendships will have a

negative effect on school performance. Moreover, because Taiwan is a collectivistic

culture, communications between the genders and with strangers is more restrained than

14
The problematic prostitution of young females is called Yuan-Zhu-Jiao-Ji in Chinese, which means
young girls sell their body for material goods. The activity is believed to have originated in Japan. In the
beginning, prostitution was arranged by phone, but now it mostly takes place through the Internet (e.g.,
instant message or discussion forum) and has become a ―new social issue.‖

83
in individualistic cultures. As a consequence, many students become addicted to such

non-realistic online activities, suffer harm (e.g., pregnancy), or become involved in risky

behaviors, such as one-night stands, sexual promiscuity, or substance use (Huang, 2003;

Wu, 2001). The high prevalence of Internet access, combined with the popularity of

cyber cafés, has caused some scholars to point out that such a phenomenon partly

accounts for increasing rates of ―offenses against sex autonomy‖ as well as violation of

copyrights (Xu, 2005).

Another popular indoor entertainment for Taiwanese youths is KTV (Karaoke TV)

(Ho et al., 20008). It is similar to American karaoke, except that Taiwanese youth enjoy

this activity in private rooms with friends. This type of entertainment is appealing, due to

its privacy, which fits into the collectivistic culture of not standing out. Besides providing

all the equipment, KTV facilities serve food and beverages. Hence, KTV has become a

popular feature of gathering places for Taiwanese teenagers for various occasions (e.g.,

birthday parties) and entertainment. While KTV is enjoyable for many Taiwanese

juveniles, it gradually has become a source of criminal and deviant activity. Due to its

privacy and lack of monitoring, it is an ideal place for youth to engage in underage

drinking and smoking, or even substance use/abuse and drug transactions (Yi & Wu,

2004). Several studies (Lee et al., 2009; Leung et al., 2008; Lee, Huang, Miao, 2000)

have shown that KTV is one of the most common forms of entertainment linked to

adolescents becoming involved in illegal substance use (e.g., MDMA, Ketamine). In

some cases, deviant sexual behavior (e.g., date rapes) and prostitution are conducted in

KTV establishments.

84
 Finally, juvenile substance use and abuse15 have become serious social problems

(Ho et al., 2008), although junior and senior high school students who were convicted of

violating substance regulation law numbered only 280 in 2007 (Zeng, 2008). However,

that number represents 95% of the total student violators in Taiwan. Two official

investigations also supported the belief that youth substance use and abuse are

problematic. Compared with 2006, youth substance offenses increased 18% in 2007

(Ministry of Justice), and urine screen tests, completed by the Ministry of Education from

2006-2008, revealed that positive results for MDMA and Ketamine increased yearly from

231 (2006) to 420 (2008) (Department of statistics, Ministry of Education, 2008).

Common substances used by the Taiwanese youth population are MDMA (―shaking

head‖ pill), Ketamine, Amphetamine, FM2, and marijuana (Lee et al., 2009). Although

the number of adolescents who use these illegal substances is increasing, the prevalence

rate is still low, around 1% to 1.4% (Zhou, 2000). The increased use of these drugs,

especially MDMA, is probably because it is cheap and provides excitement. High

educational stress and the prevalence of some entertainment options (e.g., KTV) where

adult control is low have been mentioned as causes of increased juvenile substance use

and abuse.

In sum, although juvenile delinquency rates remain stable, this review points out

that several important qualitative changes in juvenile delinquency have occurred in recent

years. Consequently, the juvenile offending rate is still low in Taiwan, compared with the
15
While this discusses only illegal substance use and abuse, alcohol and tobacco use are also common
among students. For example, Ma (2000) found that 46% of the adolescents reported they had never
smoked, and one culturally specific substance, the betel nut, is used by many adults, but its use is less
common in the adolescent population.

85
rates in other Western countries, but it still draws societal attention. The traditional

Chinese culture has gradually faded, due to economic growth as well as social and

political changes. This cultural change has inevitably contributed to increases in juvenile

delinquency. However, the traditional Chinese culture and Confucian philosophy have

not been eradicated totally, and as a result the juvenile delinquency rate in Taiwan is

lower than in most Western countries.

This review indicates that juvenile delinquency is somewhat different from that in

United States because of specific cultural and social settings, such as the prevalence of

motorcycles, cyber cafés, and KTV. This is consistent with Link (2008), who argued that

juvenile delinquency depends on cultural attitudes and perception of the social structure

and of differences in opportunity16. For example, because of the high population density

and lack of land in Taiwan, juveniles engage in many different kinds of indoor

entertainment, which encourages development of different kinds of delinquency than

youths in the U.S. In addition, because of popularity of motorcycles, youths in Taiwan

are involved in street racing and other kinds of delinquency (e.g., vandalizing properties)

in which motorcycles, not cars, are used. However, although these studies and reports are

informative, they are only descriptive and exploratory in nature. The next section will

present studies that have employed GST to explain juvenile delinquency in Taiwan and in

other non-Western countries, so as to provide a theoretical view of juvenile delinquency

in these countries.

16
One recent study on the victimization of drive-by street robbery also indicated that the uniqueness of this
phenomenon is due to Taiwan‘s special social and cultural settings (Kuo, Cuvelier, & Chang, 2009).

86
GST in Other Non-Western Countries and Taiwan

The cultural differences between Western (United States) and non-Western

countries (Taiwan) have already been discussed, and the possible effects of these

differences on the strain process and juvenile delinquency have also been presented.

However, studies that have examined GST in non-Western cultures are scarce. As

mentioned in Chapter 2, the majority of the published studies are based on U.S. samples

or samples from other Western countries (e.g., Canada). This hinders further

development of GST and delinquency theory in general (Kohn, 1987). Kim and

colleagues (1994) argued that cross-national studies would help revise a theory so as to

better accommodate cross-cultural differences, and studies that directly apply GST in

different nations have revealed the important role that culture plays in the GST/stress

process (Bao, Haas, & Pi, 2007; Botchkovar et al., 2009; Chun et al., 2006; Markus &

Kitayama, 1991; Tanzer et al., 1996). This section will briefly discuss the issues of cross-

national studies and provide a detailed review of the few studies that apply GST in non-

Western countries including Taiwan.

Cross-national study17

According to Kohn (1987), cross-national studies can be divided into different

types according to their different purposes; two of these types are related to the present

17
Although the present study uses cross-culture and cross-nation interchangeably, one must always keep in
mind that a nation can accommodate more than one culture (e.g., Native American culture in the U.S.). The
present study mainly considers the dominant culture in a nation as representative of that nation; hence,
Western culture to U.S. and Chinese culture to Taiwan.

87
study. The first type consists of studies of particular countries or cultures for their own

sakes. That is, the primary interest of the research is to know about, for example, Taiwan

and the United States. By contrast, the second type of cross-cultural study focuses on how

certain social construct/institutions influence individuals‘ behavior. In this perspective,

culture is treated as the context within which theoretical mechanisms are examined. The

present study intends to investigate the GST mechanism in two different countries, which

leads to adoption of approach two. Consequently, the cultures in Taiwan and the United

States are treated as the context that affects juveniles‘ perceptions of strain, the

consequent emotional responses, and coping behavior. However, before comparing the

GST process across nations, applying GST in Taiwan will provide insights into

understanding the theoretical process in that country (approach one).

Despite the two types of research that Kohn (1987) defined, cross-cultural/cross-

national research inevitably faces a related issue: the ―emic/etic‖ issue. The former

describes the study of a phenomenon within a particular culture (―idiographic‖ style).

The latter tries to apply a general theoretical model to all cultures, in an effort to find

universal behavior rules (―nomothetic‖ style). Using the emic approach helps ensure the

uniqueness of a particular culture is preserved, but the generalizability of the results is

limited, and there may be a risk of ethnocentricity. In contrast, applying the etic approach

may greatly enhance finding law-like theories, but the approach risks ignoring cultural

uniqueness. The present study, as a cross-national/cultural study, is a blend of emic and

etic approaches because it transports and examines a theory that has developed mainly in

the United States, within Taiwan, but at the same time the indigenous knowledge of

88
Taiwan is preserved because of the researcher‘s awareness. Also, the final results will be

compared between the United States and Taiwan, which will help verify, revise, and

extend the existing theory (Kim et al., 1994).

Applying GST in non-Western countries

So far, only six published studies have directly applied GST in three non-Western

countries: China (Bao et al., 2007), South Korea (Moon & Morash, 2004; Moon, Blurton,

& McCluskey, 2008; Moon, Morash, McCluskey, & Hwang, 2009; Morash & Moon,

2007), and the Philippines (Maxwell, 2001). In addition to these six studies, one other

study investigated GST and crime across three Western European countries (Greece,

Russia, and the Ukraine).

Maxwell (2001) used a convenience sample of sixth-grade students from one

urban area of the Philippines to study the impact of family strain (witnessing domestic

violence and parent-to-child violence) on antisocial behavior. Although the Philippino

society is influenced both by Spanish culture (e.g., over 85% of the population is Catholic)

and the culture of the United States (e.g., English is the medium of instruction), the

Philippines are similar to many Eastern countries in that family is a significant social

institution. The results of Maxwell‘s (2001) study indicate that witnessing domestic

violence is a strong predicator of students‘ self-reported antisocial behavior and teachers‘

predictions about students‘ antisocial behavior. However, direct parent-to-child violence

does not stand out as a significant cause of antisocial behavior. Maxwell argues that such

89
null findings are consistent with Agnew (1992), and cautions against ignoring cultural

difference in defining strain in cross-cultural research.

Bao, Haas, and Pi (2007), who applied GST in China, selected 615 students

(grade 8 to grade 11) from Guangdong Province. They examined both the strain-

delinquency relationship, as well as domain-specific and cross-domain buffering effects

of several conditioning variables on this relationship. The results revealed that strain from

family and school increases delinquency; however, negative relationships with peers do

not increase delinquency. Regarding the domain-specific buffering effects, they found

that school support dampens the effect of school strain on consequent delinquency. On

the cross-domain buffering effects, they indicated that whereas family support reduces

the effect of school strain on juvenile delinquency, school support protects youth from the

negative effect of family strain. Finally, they found that moral beliefs buffer the effects of

family, school, and peer strain on adolescents. This study also found that delinquent peers

increase delinquency among students who experience family strain and school strain. The

study also revealed gender differences in using social support. The authors argued that

girls are more likely to use social support in managing interpersonal strain, but boys are

more likely to be influenced by delinquent peers, which causes them to respond to

interpersonal strain with a higher frequency of delinquency. School support plays an

important role in modifying the strain-delinquency relationship in this study, which

echoes the importance of Confucian influence in China.

Four other studies that use GST to explain juvenile delinquency in Eastern

cultures were all conducted in South Korea. One study (Moon & Morash, 2004) recruited
90
385 11th graders from a large city in South Korea to examine the relationship between

culture-specific strains (exam-related strain and teacher strain) and juvenile delinquency.

This study revealed that teacher strain (emotional and physical abuse) did increase

delinquency, whereas exam-related strain did not. In addition, teacher strain and

delinquency relationships were conditioned by delinquent peer associations. The authors

suggested that GST is applicable to South Korea but, at the same time, they advocated

that researchers pay more attention to the specific types of strain unique to a particular

culture. The same authors (Morash & Moon, 2007) also tested the gendered strain

process that is delineated by GST (Broidy & Agnew, 1997). They found that boys are

more likely to experience teacher strain and that, in both genders, the combination of

teacher strain and delinquent peers is a strong predictor of violence. They also discovered

that girls are under the influence of different sorts of strains—parental strain, teacher

strain, and financial strain, than boys are. Girls who associated with delinquent peers

were twice as likely to respond to these strains with delinquency than girls who did not

associate with delinquent peers.

Unlike the two studies just mentioned, which investigated the basic GST process,

two recent studies conducted in South Korea focused on claims of the newly revised GST

(Agnew, 2001, 2006a). One of the studies (Moon, Blurton, & McCluskey, 2008)

examined how the recent perceived injustice of key strains (e.g., criminal victimization)

affect delinquency. Using a sample of 777 middle-school South Korean students, this

study indicated that some key strains—teachers‘ punishment and criminal

victimization—strongly predict juvenile delinquency regardless of whether such strains

91
happened recently (in the past six months) or long ago (more than a year). This somewhat

contradicts GST‘s prediction that says recent strain should have stronger effects on

delinquency. Also, this study found that chronic parental punishment and bullying

reduces delinquency. The perceived injustice of teachers‘ punishment and criminal

victimization did not stand out as a significant predictor of delinquency in the full model,

when other variables were included in the model (e.g., control variables). Furthermore,

these authors revealed that anger increases delinquency, but does not have a significant

mediating effect on the strain-delinquency relationship.

Using panel data based on South Korean middle-school students, Moon and

associates (2009) conducted a ―comprehensive‖ test of GST. They investigated the

relationships between key strains, trait-based and situational-based anger and depression,

several conditioning factors, and three different types of delinquency (violent, property,

and status). In general, they reported that most of their strain measures, as well as a

composite strain measure, had significant and positive effects on delinquency; however,

bullying, as a strain, did not increase delinquency. This study also revealed that both

situational and trait anger exert mediating effects on the strain-delinquency relationship,

especially violent delinquency, but the influence of trait anger was only minimal. On the

other hand, both trait and situational depression had slight mediating effects, even when

inner-directed delinquency (e.g., smoking) was an outcome variable, which was

inconsistent with GST‘s prediction as well as with results of previous research. Finally,

Korean youths who experience various strains (combined strain) were less likely to

commit violent crimes when they had a positive relationship with their parents, and such

92
adolescents were also less likely to become involved in property delinquency if they had

higher problem-solving skills. Although this study reported that a strained juvenile was

less likely to commit all forms of delinquency when this youth was associated with

delinquent peers, these authors argued that such results might be artificial, in that within

all levels of delinquent peer association, the strained students who had more delinquent

peers were more likely to be involved in delinquency than strained students who did not,

which was consistent with GST.

In sum, the above studies specifically applied GST to three Eastern countries and

found support for some of GST‘s propositions, which supported GST‘s generalizability.

However, as pointed out, some cultural influences were also apparent. For example, in

Eastern countries, with their more collectivistic cultures, family plays an important role in

youngsters‘ lives, which makes it a particular important source of strain. Also, since the

Confucian ethos prevails in many Asian countries, teacher importance is emphasized. As

Asian countries become more industrialized and modernized and, hence, more

individualistic (Hofstede, 2001; Triandis, 1995), the conflicts between traditional role

expectations and rules and modern individualized perceptions are expected to increase,

which will inevitably increase feelings of strain.

GST in Taiwan

As Taiwan has experienced economic growth and has become more modern,

juvenile delinquency as well as societal awareness of the well-being of adolescents have

also increased. Hence, many recent studies have been conducted to investigate the effect

93
of stress on students and juvenile delinquency. For example, Li and Chiang (2001)

reported that the most stressful life-events for youths (7th to 9th grades) are school-related

incidents (Wang, 2001) and that students are more adaptive to stress when they receive

material support (e.g., money, transportation). Xu and Huang (2004) also found that the

most influential life stressor for Taiwanese juveniles is school stress and that such

stressors are significantly related to various forms of delinquency, such as gang-related

delinquency (e.g., joining a gang) and pirating (e.g., cheating, illegal copying). In contrast,

Shi (2004) found that junior high school students who experience family-related, dating-

related, peer-relational, or future-related stressors are more likely to experience higher

levels of depression than those who experience school-related stressors.

Several studies have employed GST to explore the relationship between

stress/strain and juvenile delinquency. Chen (2000) found that stress related to school is

significantly correlated with delinquency and deviance; however, this stress significantly

predicts only delinquency (e.g., stealing), but not deviance (e.g., cheating). Peng (2002)

used a sample of junior- and senior-high school students from the southern part of

Taiwan to investigate the relationships among strain, negative emotions, and juvenile

delinquency. She found that negative life-events, daily hassles, and criminal victimization

all have significant effects on delinquency; however, daily hassles fail to exert a

significant effect on delinquency, when social control and self-control are in the model.

In addition, criminal victimization and daily hassles cause negative emotions, and the

combined negative emotions (e.g., anger, depression) along with victimization and daily

hassles increases delinquency. Tsai (2005) and Xu (2005) both found that strain

94
significantly increases various delinquent acts (e.g., vandalism, gambling) and deviant

acts (e.g., copying other students‘ homework) as well as some somatic symptoms (e.g.,

tiredness), after social support and demographic variables had been controlled for. Xu

(2005) found that strain not only directly affects all these negative life-outcomes—

delinquency, deviance, and somatic symptoms—but also indirectly affects them through

negative emotions.18 Tung (2007), who used a random sample of junior- high school

students (n = 1,540) to investigate relationships between life stressors and violent

behavior, found that negative life-events, relationship conflicts, and daily hassles all

significantly increased violent behavior, even after controlling for social support,

delinquent peers, and parental attitudes. In another report (Tung & Wu, 2008), the

original three types of strain- blockage of achieving goals, presentation of negative

stimuli, and removal of positive stimuli- was shown to be related to self-mutilation.

Although there have been a few studies that investigated the relationship between

strain and juvenile delinquency in Taiwan, these studies have only scratched the surface.

They focused on only the direct effects of various strains on delinquency, and only two

studies examined the role that negative emotions play in the strain-delinquency

relationship. Other important issues (e.g., conditioning effects) and the recently revised

GST propositions (e.g., magnitude) have not been empirically evaluated in Taiwan.

In sum, while studies have employed the GST approach to explain juvenile

delinquency in some non-Western countries (Maxwell, 2001; Morash & Moon, 2005;

18
The author does not specify this variable except to name it as ―negative emotion.‖ Upon close reading,
the items she uses to measure this variable are related to both anger and depression/anxiety.

95
Tung, 2003, 2007), these studies reflect certain limitations. First, studies applying GST to

non-Western countries did not have a similar U.S. sample for comparison. Hence, the

similarities and differences found in comparing the results to established evidence from

the United States are subject to various explanations, including cultural differences, and

sampling differences. Second, previous studies did not apply the full GST model, which

includes all major types of strain and negative emotions, to other countries (see Moon et

al., 2009, for an exception).

A more serious problem of many studies in the GST literature is the lack of a

systematic explanation of cultural differences and similarities. For example, the most

salient difference between Western (U.S.) and Chinese (Taiwan) cultures is the difference

between individualism and collectivism, which are the concepts that scholars have

identified as the major dimensions of cultural variability (Hofstede, 2001; Traiandis,

1995). How these two cultural variables interrelate with GST has not been explored. Also,

how other important aspects of cultural heritage (e.g., the Confucian ethos for the

Chinese) influence individuals‘ perception of strain, their emotional responses, and their

choice of coping strategies has also been ignored. For example, Chinese culture

emphasizes educational attainment and relational harmony, which may create or increase

strains emanating from these areas in an adolescent‘s life. Furthermore, the expression of

negative emotions will differ between Western cultures and the Chinese culture, because

expressing negative emotions is a sign of immaturity in Chinese culture.

96
 CHAPTER IV:

RESEARCH QUESTIONS

Drawing on the above discussion, there is a need to empirically evaluate the basic

GST model and conduct a more systematic cross-national comparison, especially

between Western countries and Eastern countries. This study uses a sample of U.S.

adolescents and a sample of Taiwanese juveniles to address this gap in the literature.

Specifically, the present study will address the following questions:

1. Is strain related positively to delinquency and aggression in the U.S. and Taiwan?

a. Is goal strain related positively to delinquency and aggression in the U.S. and

Taiwan?

b. Is unjust strain related positively to delinquency and aggression in the U.S. and

Taiwan?

c. Are negative life-events related positively to delinquency and aggression in the

U.S. and Taiwan?

d. Is victimization related positively to delinquency and aggression in the U.S. and

Taiwan?

2. Is strain related positively to anger and depression in the U.S. and Taiwan?

a. Is goal strain related positively to anger and depression?

b. Is unjust strain related positively to anger and depression?

c. Are negative life-events related positively to anger and depression?

97
 d. Is victimization related positively to anger and depression?

3. Does anger mediate the relationship between strain and delinquency/aggression?

4. Does depression mediate the relationship between strain and delinquency/aggression?

5. Is there any difference in the GST process in the U.S. and Taiwan?

a. Is the strain → delinquency process different in the U.S. and Taiwan?

b. Is the strain → aggression process different in the U.S. and Taiwan?

c. Is the strain → anger process different in the U.S. and Taiwan?

d. Is the strain → depression process different in the U.S. and Taiwan?

6. Is the GST mediating process different in the U.S. and Taiwan?

a. Is the strain → anger → delinquency process different in the U.S. and Taiwan?

b. Is the strain → depression → delinquency process different in the U.S. and

Taiwan?

c. Is the strain → anger → aggression process different in the U.S. and Taiwan?

d. Is the strain → depression → aggression process different in the U.S. and Taiwan?

98
 CHAPTER V:

METHODS

Sample

This study includes two samples: one from the U.S. and the other from Taiwan.

The U.S. sample is from an existing cross-sectional data set that was collected purposely

to examine juvenile delinquency among middle and high school students. All the subjects

from the U.S. sample were recruited from one public middle and one public high school

in Largo, Florida. New data for this study were collected from students in Taiwan who

were enrolled in a junior and a senior high school in one school district of Taiwan‘s

second largest city (Kaohsiung).

U.S. sample

The U.S. data were collected in Largo, Florida, in 1998. Largo is a metropolitan

area comprising 17.9 square miles and located about 23 miles west of Tampa. Its

population during the 1990s was around 69,000 people: 47% male, 92% white, and 16%

younger than 18 years (U.S. Census Bureau, 1990, 2000). About 6% of Largo‘s families

had incomes below the poverty level, and the city‘s 1998 median adjusted household

income was $42,000 (Largo Chamber of Commerce, 1998; U.S. Census Bureau, 1990,

99
2000). In 1998, the city‘s official crime rate (per 100,000) was 5,019 (Florida Department

of Law Enforcement, 1999).

The Largo middle school, one of the two area middle schools (grades 6-8),

enrolled 1,294 students during the 1998-1999 school year; the average class size was 25

students. Students from all Social Studies classes were invited to participate. Before the

actual survey, a passive parental consent form was distributed to students (see Verrill,

2008 for details). On the day of the survey, a researcher explained the purpose of the

study to all participants, reminded students that participation was voluntary, and

reassured them of the confidentiality of all the information they gave. The researcher then

remained on site to answer questions related to the survey. The final response rate was

81% (N = 1,049).

The Largo public high school, one of several high schools (grades 9-12) in this

area, enrolled about 1,848 students during the 1998-1999 school year; the average class

size was 33 students. As in the Largo middle school survey, a passive parental consent

procedure was used. Students from a random sample of 30 third-period school classes

were asked to participate. On the day of administration, a researcher described the

purpose of the study, explained that participation was voluntary and that the provided

information was confidential, and remained available to answer questions (Wareham,

Cochran, Sellers, & Dembo, 2005). The final response rate was 79% (n = 625).

100
Taiwanese sample

The additional data for this study were collected from a sample of junior and

senior high school students in Kaohsiung, which comprises about 59.3 square miles and

has a population of about 1.5 million: 49.4% male, 21% 18 or younger, and with a

marriage rate of about 4.78 per 1, 000 (Department of Budget, Accounting and Statistics,

Kaohsiung City Government, 2009). The city‘s average family income was $38,83219 in

2009, and about 1.5% of Kaohsiung‘s families were considered low income by the city

government (Department of Budget, Accounting and Statistics, Kaohsiung City

Government, 2009). The overall 2009 official crime rate (per 100,000) of Kaohsiung was

1,317 and the juvenile crime rate was 96720 (Department of Budget, Accounting and

Statistic, Kaohsiung City Government, 2009). The city has neither a remarkable

concentration of particular demographic groups nor a high crime area in any of its 11

districts.

The Zuo-Ying district, one of the 11 districts, has a rich historical background and

is an important military harbor of Taiwan. With a high speed rail station and rapid transit

system that were built in recent years, it has become an important business hub of the

northern part of Kaohsiung. The Zuo-Ying district comprises 7.48 square miles and has a

population of about 189,944: 49% male and about 25% age 18 or younger (Department

of Budget, Accounting and Statistics, Kaohsiung City Government, 2009).

19
This number is based on 1 (U.S. Dollar): 29 (NT Dollar) exchange rate.
20
The number here represents number of offenders per 100, 000 in Kaohsiung City.

101
 The selection of Taiwanese junior and senior high school students comparable to

those in the U.S. survey was not easy, because of educational and political system

differences. For example, Taiwan has 23 prefectures and two municipal cities with

populations over 1 million (Taipei and Kaohsiung city), and all of these prefectures and

municipal cities are under the central government‘s control. In contrast, the U.S. has 50

states, each of which has its own laws and independent political systems. In Taiwan, a

junior high school education is compulsory, but a senior high school education is not;

also, education systems offer three years of education. In the U.S., education in middle

(junior) school lasts for three years, but senior high school offers 4 years of education.

However, despite these differences, both Largo and the Zuo-Ying district are similar in

many ways. For example, both selected areas are near a coast and a metropolitan area

(Largo to Tampa; Zuo-Ying to Kaohsiung city), and the selected schools are similar in

size and geographic location (i.e., all in the same school district). Hence, the study

attempted in various ways to make the two samples comparable.

The junior high school selected for the present study from the Zuo-Ying district

was Zuo-Ying junior high school. There are five public junior high schools in the Zuo-

Ying district, and these junior high schools vary in size compared to the average number

of students in Kaohsiung City (Education Bureau, Kaohsiung City, 2010). Specifically,

three schools are relatively large with over 2,200 students each, one is in the middle

range with over 1,000 students and one is small, with only about 400 students. The Zuo-

Ying junior high school is one of the three large schools in this district, similar in size to

the Largo public middle school. The age range of junior high school students in Taiwan

102
as a whole is 12-15. The total number of students in the Zuo-Ying junior high school was

2,265 in 2010: male students constituted about 51% (n = 1,153), and the average class

size was 37 students (Education Bureau, Kaohsiung City, 2010).

In order to collect all subjects (junior and senior high school students) from the

same district, as the Largo study did, the senior high school sample was also selected

from the Zuo-Ying district. Of the approximately 2721 senior high schools in Kaohsiung

City, 4 are in the Zuo-Ying district; of these four, two are public high schools and two are

affiliated private senior high schools. Each of the two public senior high school has over

1,700 students, which is slightly more than the average for public senior high schools (n

= 1,546) (Education Bureau, Kaohsiung City, 2010). The Zuo-Ying senior high school

that was selected for the present study enrolled 1,789 students in 2010, with 48% male

students (n = 867) and an average class size of 35 (Education Bureau, Kaohsiung City,

2010). Compared to the Largo public high school, the Zuo-Ying senior high school has

slightly fewer students.

In Taiwan, a junior and senior high school student usually goes to his or her

homeroom every day where he or she will be with all those who will be classmates for

three school years. Almost all the major school subjects are taught by different teachers

in the homeroom, except some special subjects, such as music or physical education. In

addition students will need to decide their future track22 before starting their first year of

21
Among these 27 senior high schools, 5 are private senior high schools, and 5 of all public and 2 of all
private senior high schools are affiliated senior high schools.
22
The first track is geared toward language (e.g., Chinese, English), law, or business schools in universities.
The second track is geared toward medical school, social science (e.g., sociology, psychology), engineering
103
senior high school. The subjects taught will be the same for all students in a given track.

Hence, unlike the situation in the Largo high school, in which random samples of

students had been selected from 30 third-period high school classes, classes from each

track and grade were selected in order to have enough students23. Although the principals

of the selected schools provided support letters to the research team, they still showed

concern about letting ―outsiders‖ into school during regular school hours. A compromise

was reached whereby the principals agreed to supply teachers to help in administering the

survey by distributing the survey and then remaining on site to help answer questions24.

All students present on the day of the survey were given opportunity to participate, but

because of the voluntary nature of this survey, not all students present on the day

participated25. Before the day of survey administration, students received two letters from

the teacher: one letter explained the purpose of the study, indicated that participation was

voluntary, and the researchers would assure that no student names would be placed on

survey forms; the second letter, a passive parental consent letter, described the research

school (e.g., electronic engineering, mechanic engineering), and other schools in universities (e.g., biology,
veterinary medicine). The curricula are, therefore, different except for some fundamental courses (e.g.,
math, Chinese).
23
Enough students means that the study planned to collect a similar number of students as the Largo study
had. The Largo study had 625 high school students and 1,049 middle school students, thus, this study
decided to recruit about this same number.
24
Students might have been unwilling to respond to sensitive questions (e.g., delinquency) because
teachers were on site. Compared to other self-report studies in Taiwan where teachers were not on scene,
the proportion of students who reported delinquency in this study was not particularly low. For example,
about 12% and 27% of students had purposely damaged property and used alcohol in the present study. The
number was 17% and 20% respectively in Tung‘s research project in 2000. Similarly, the number was 9.5%
and 10.2% respectively in a government research report, which was based on a representative sample of
Taiwanese youths in 2003.
25
The participating rate is 91% (860 students out of 945 students) for the Zuo-Ying senior high school, and
the rate is 96% (960 students out of 999 students) for the Zuo-Ying junior high school.
104
and stated the exact day and time of the survey. In addition to these letters, the first page

of the survey again assured all students that the survey information would be kept

anonymous and that the information would be available only to the researchers.

The survey questionnaire was mainly developed in English and for research in the

U.S. Hence, some procedural problems must be addressed before further discussion of

the survey. Van de Vijver and Leung (1997, pp. 31-35) pointed out some pitfalls in the

cross-cultural research procedure. Briefly, these can be summarized as follows: the

researcher is unfamiliar with the target culture, the subjects in the target culture are

unfamiliar with the response procedures, and there are differences in the sample. The fact

that the present researcher is from the target culture reduces the first problem. The second

problem can be dealt with through a pretest (van de Vijver & Leung, 1997). The pretest

with a small sample of Taiwanese 7th graders indicated no problem with reading and

understanding. The response categories are familiar to Taiwanese adolescents, because

many research projects conducted in Taiwan have employed similar response choices

(e.g., Taiwan Youth Project) and because public polls also use these response options. In

addition, the present researcher and other Ph.D. students from Taiwan reviewed the entire

survey instrument. This procedure helped to identify items that did not make sense or

seemed awkward to Taiwanese readers, which could cause problems related to stimulus

characteristics. For example, although marijuana may be a common substance used by

students in America, Ketamine or MDMA are the substances that Taiwanese students

most often use.

105
 Besides the above mentioned pitfalls that might plague research procedures in

cross-cultural research, method bias can also occur. Some common method biases are:

response familiarity, differences in physical conditions during administration, and

communication bias (van de Vijver & Leung, 1997). The first and last biases were

reduced as discussed in the previous paragraph. Physical conditions were made as similar

as possible across the two cultural samples by having surveys administered during regular

class hours in the regular classroom and by using a paper-pencil format for both samples.

Survey preparation

The instrument used in the Taiwan survey is an adaptation of that used in the

Largo schools. ―Adaptation‖ refers to the literal translation of the original instrument

with some wording and content changes, in order to enhance the appropriateness of such

an instrument in a different culture (van de Vijver & Leung, 1997). One important issue

that most cross-cultural studies face is the equivalence of the instrument. The most

common remedy is back-translation (Brislin, 1986; Hofstede, 2001), which although

reducing errors of translation, does not guarantee a perfect or error-free translation

(Hofstede, 2001; Sanchez, Spector, & Cooper, 2005). Other steps can be employed to

reduce the ―variance‖ left unsolved by back-translation, including the use of translators

who not only understand both languages, but also have considerable deep experience

with both cultures (Hofstede, 2001; Sanchez et al., 2005). This can provide a more

accurate translation that carries the same meanings of the items into different cultures.

106
 The official language in Taiwan is Mandarin; hence, all the survey items adapted

from the U.S. study were translated into that language. Five steps were employed to

ensure the accuracy of the translation. First, all survey items in the English version were

translated into Mandarin by the author. Second, the same version was translated into

Mandarin by two doctoral students from Taiwan, who were studying at U.S. universities.

Third, the resolution of any differences between the two translated Mandarin versions

occurred through discussion between all these doctoral students and the author. Fourth,

the consensus Mandarin version was back-translated by a professional bilingual translator.

Finally, the differences between the original English version and the back-translated

English version were resolved through an in-depth discussion between the author and the

translator. The final version of the survey instrument resulted from this discussion.

After the language issue of the survey had been resolved, the final version of the

survey was converted to a scantron format, to reduce as much as possible errors that

occur while inputting raw data into a computer. Two incidents occurred during the time

period of collecting all the scantron sheets and performing final data inputs. First, a flood

during that time damaged about fifty of the answered scantron sheets; fortunately, the

flood damage was minor enough that all answer sheets were recovered. Second, many

students changed their answers, which caused a problem of data reading by the scanner.

With regard to the second problem, the researcher and the research assistants were able to

go through all the answer sheets to correct these errors, thus reducing to a minimum the

possible errors in the final data set.

107
Measurement of variables

The present study intends not only to apply GST to the study of juvenile

delinquency in Taiwan, but also to compare and contrast the GST model between the U.S.

and Taiwan. Consequently, the instruments used to measure all variables were identical

across the two samples, except for some slight changes made to fit Taiwan‘s social

conditions (e.g., marijuana is substituted by Ketamine or MDMA).

Both of the central components of GST are included: strain and negative emotions.

To capture each concept of strain and negative emotion, the survey items were subjected

to exploratory factor analysis (EFA) and, based on the results, a composite score was

created for each individual strain variable. Two sets of variables, delinquency (which

includes three different kinds of delinquent acts26) and physical aggression against

siblings, were used as the main endogenous variables. Finally, the present study also

incorporated two demographic variables (age and gender) as covariates. All these

variables were included in the final path analysis.

Delinquency27

Juvenile delinquency is defined broadly in the present study. It includes behavior

that is prohibited by law as well as acts carried out by youths who are within certain age

26
Three delinquent acts were chosen mainly because these three delinquent acts have enough variation and
are theoretically relevant. Detail is given in the delinquency section.
27
All but two items used in both the U.S. and Taiwanese surveys were worded identically. The two items
with different wording are about using marijuana and stealing a car or motorcycle. Marijuana is not
common for use by juveniles in Taiwan; the more ―popular‖ substances used are Ketamine and MDMA. In
Taiwan, the most common private means of transportation for juveniles are motorcycles and bicycles.
Consequently, the item ―stealing a car or motorcycle‖ was changed to ―stealing a motorcycle or a bicycle.‖
108
limits (e.g., 12-18) (Short & Hughes, 2008; Trojanowicz, Morash, & Schram, 2001). In

other words, delinquency refers to behavior that violates the law but is carried out by

minors (e.g., damaging property), and also to behavior that is prohibited to youth but not

to adults (e.g., alcohol use) (Stafford, 2004).

Self-reported delinquent behaviors included in the present study are fairly

commonly reported in the criminological literature (Elliott, Huizinga, & Ageton, 1985;

Piquero, Macintosh, & Hickman, 2002). The questionnaire asked students to report

whether they have done each of several delinquent acts. However, close examination of

these items revealed that for many items, less than 1% of subjects endorsed the acts in

both countries; this was especially true in Taiwan. Hence, these items were discarded

because of lack of variance to be explained. We finally chose 3 delinquent acts as our

outcome variables. These variables not only have relatively high percentages of students

reporting that they had done the acts in the past 12 months, but also were theoretically

relevant and representative of the general delinquency categories (e.g., property crime,

violent crime). For example, we included alcohol use, which is a substance offense and

which has been found to be related to strain (Aseltine & Gore, 2000).

The 3 selected delinquent acts are ―purposely damage property‖ (property

delinquent act), ―hit someone with intention to hurt them‖ (violent offense), and ―alcohol

use‖ (substance use). Individuals who report they have not committed a particular act

receive a score of 0, and students who report that they have received a score of 1 (See

Appendix A for detailed wording). Each of these three delinquent acts is examined

separately because they represent different domains of acts, and Agnew (2006a, 2006b)
109
has advocated separating various coping behaviors by type of act. This analysis helps to

identify strain-delinquent specific relationships.

Physical aggression

Card and colleagues (2008) refer to physical aggression as direct aggression (Xie, Farmer,

& Cairns, 2003), although direct aggression includes physical and verbal aggression. The

present study defines physical aggression as direct confrontation between perpetrator and

victims by any physical means (e.g., hitting or kicking) (Xie et al., 2003). Students were

asked to report how many times they had engaged in physically aggressive behavior

against a sibling in the past 12 months. Notice that one of the response categories is ―no

siblings.‖ Students who indicated that they had no siblings were dropped. Although this

might reduce sample size, the reduction was relatively small for both countries (7.8%, or

117 subjects, for the U.S.; 6.1%, or 106 subjects, for Taiwan). In addition, the analysis

for physical aggression was done separately from that for delinquency; hence, the

reduction of sample size did not affect other analyses (see Table 2 for the frequency

distribution for delinquency and aggression).

Strain

General strain theory (GST) identified strain as relationships in which the

―individual is not treated as he or she wants to be‖ (Agnew, 1992, p. 48). Later, Agnew

(2006a) defined strain as ―events or conditions that are disliked by individuals‖ (p.4). The

present study adopts the latter definition because it is broader and more inclusive.

110
 Agnew (1992) outlined three major types of strain: failure to achieve positively

valued goals, removal of positively valued stimuli, and presentation of noxious stimuli.

For the first kind of strain, the present study includes measures of (1) the discrepancy

between desired goals and actual outcomes and (2) experience with unjust outcomes.

Stressful life-events involving loss (3) were used as a measure of the second type of

strain. Finally, (4) victimization was used as a measure of the presentation of negative

stimuli (see Appendix A for detailed wording).

The purpose of the study was to investigate the effects of these four different

strains on subsequent delinquency. Consequently, a composite score was created for each

of the strain variables. To ensure that all items being used are reasonably able to ―hang

together‖ for both countries, we conducted exploratory factor analysis (EFA) for each

strain variable. The present study used principal axis factoring (PAF), which is the mostly

widely used method, instead of principal components analysis (PCA) to conduct EFA for

two reasons28.

First, EFA is aimed at explaining community (i.e., variance shared by items or

indicators) of the correlations among all items used to measure latent concepts, and thus

is highly consistent with the goal of determining whether the items used in the present

study could be accounted for by the underlying latent concept (e.g., anger). Many

scholars have suggested that when the goal of analysis is to identify a latent construct,

one should use EFA rather than PCA (Fabriga, Wegener, MacCallum, & Strahan, 1999;
28
The discussion that is followed focuses on the difference between EFA in general and PCA. The
intention is to clarify the confusions between EFA and PCA and the justification of using EFA. PAF is one
of many EFA methods, and the discussion can apply to PAF.

111
Gorsuch, 1983). In addition, it is necessary, though not sufficient, to demonstrate that the

latent concept (factor) accounts for the common relationship among a set of indicators,

before the validity of the construct, the model, and the indicator can be accepted

(Pedhazur & Schmelkin, 1991). In contrast, PCA extracts components that account for as

much variance as possible, and it does not differentiate between common variance and

unique variance. Moreover, the extracted components are not latent concepts. For these

reasons, Fabriga et al. (1999, p.275) stated that PCA ―is not a factor analysis at all.‖

Second, and related, the main diagonal of the matrix analyzed in EFA consists of

commonality, which usually is not equal to one. This is important because this does not

assume error-free measurement. Unlike EFA, PCA treats the main diagonal of the matrix

as error-free; that is, it puts unity at the main diagonal. This is counterintuitive, because

most social science research contains random error. Consequently, the present study uses

the principal axis factoring approach to EFA as a guide to create the conceptual variables.

One of the controversies surrounding EFA is with the regard to how many factors

should be retained. Although some statistical guidelines exist for this matter (e.g., Kaiser-

criterion), criticism of these rules persist. Pedhazur and Schmelkin (1991) argued that

decisions on EFA (e.g., number of factors to retain) should be ―made within a theoretical

context‖ (p.622). Following their admonition, this study used theory as the guideline to

decide on the number of factors to retain. In each EFA, we retained only one factor,

because the GST would predict that the items should reasonably represent a single

underlying theoretical construct. This decision later was supported by the statistical rules;

112
that is, the scree plot combined with the Kaiser-criterion agreed with the 1 factor solution

for each of the EFAs29 (the factor loadings in Appendix B).

After conducting EFA, the study created the composite score for each strain

variable. Two different scoring procedures were used. For goal strain and unjust strain,

all the items are on the ordinal level; hence, we divided each item by its own standard

deviation (SD) before summing all items together, rather than summing z-scores for each

item. By doing this, we avoided two problems. First, as when using z-scores, dividing the

raw score by its SD prevents assigning too much weight to items that have great variance;

there is no theoretical reason to believe that some items should have greater weight than

others. Second, dividing a raw score by its SD, rather than using a z-score which

subtracts the mean from the raw score before dividing by its SD, preserves the mean

differences across countries; this is important because one purpose of this study was to

discover whether there is a difference between the two countries. For the negative life-

event (10 items) and victimization scale (6 items), each item has only 2 categories; hence,

a z-score was created for each individual item and then the z-scores were summed to

create a composite score for each of these two strain variables.

29
The present study conducted six EFAs, four for strain variables and 2 for negative emotions. In each of
these analyses, each item loaded over .4 (Pedhazur & Schmelkin, 1991) on its respective factors. In some
cases the loadings were not over .4 but greater than .3; these items were still included because of theoretical
expectations. Moreover, the cut-off on loading is still much debated and subjective; some researchers
would accept .3 as a reasonable cut-off point (Child, 2006).

113
Failure to achieve positively valued goals−disjunction between desired and actual

outcome. Agnew (1992) criticized classic strain theories (Merton, 1938; Cohen, 1955;

Cloward & Ohlin, 1960) because they focus only on blockage of long-term monetary

goals. Later researchers argued that, although monetary goals are important, a juvenile

may have goals other than money, such as popularity and autonomy (Agnew, 1985;

Agnew, 2006a). The present study covers various goals that a youth might find important

in his or her current life (e.g., relationships, autonomy). Specifically, students were asked

to evaluate whether they strongly agree, agree, disagree, or strongly disagree with several

statements regarding receiving respect from parents and teachers (3 items), relationships

with others (2 items), and autonomy (2 items). Higher scores indicate that students did

not achieve or were unsatisfied with their situation regarding these goals and hence

experienced greater strain.

Failure to achieve positively valued goals−unjust outcomes. According to GST, strain

may be the result of a disjunction between a desired goal and an actual achievement, as

when an individual focuses on achieving but fails to attain a specific outcome (e.g.,

popularity). Another possible and related source of a strain is when an individual enters

into a relationship with the expectation that a certain rule of justice will be followed.

Agnew (1992) argued that a relationship is most stressful if the outcome/input ratio is not

equal, when individuals feel they have been under-rewarded (Hegtvedt, 1990). To capture

these feelings of unjust strain, seven items asked students whether they agree or disagree

with statements about unequal relationships in which they are involved. For example,

students can choose from strongly agree to strongly disagree in response to the statement,

114
―Many students don‘t study as hard as I do, but they still make better grades.‖ Although

these items do not specify the exact input/output ratio of all involved parties, all the

statements delineate clearly a situation in which the respondent does not ―get the best

deal‖ (Agnew, 1992). Higher scores indicated greater unjust strain (see Table B1 in

Appendix B for factor loadings for both variables).

Loss of positive stimuli−negative life-events. Strain can be a result of losing a positively

valued goal (Agnew, 1992). The most widely used instrument that captures this type of

strain and the presentation of negative stimuli, whether in the stress literature (Kaplan,

1983; Thoits, 1983) or in studies that examine GST (Eitle & Turner, 2003; Hoffmann &

Miller, 1997; Mazerolle & Piquero, 1997) is the negative life-event scale. Although many

different negative life-event scales exist, Turner and Wheaton (1995) argue that there is

no advantage in using one particular scale rather than another and that scales should be

tailored to fit the studied population (Herbert & Cohen, 1996). In addition, these authors

also suggest that a 1-year time frame should be used and unweighted indices are as useful

as any (Herbert & Cohen, 1996). The negative life-event scale (10 items) used in the

present study has all of the four recommended features, and all events are related to loss

of positive stimuli (e.g., death of a relative, loss of a friendship). Students reported

whether the event had happened to them (yes = 1) in the past 12 months or not (no = 0).

Higher scores indicate that a student experienced many stressful life-events and

consequently had a higher level of strain.

Presentation of noxious stimuli−victimization. Noxious stimuli by definition are those

events or incidents that are disliked by individuals. Criminal victimization is one of the
115
most severe noxious stimuli and types of strain. Here, victimization refers to any physical,

mental (emotional), and financial harm people suffer because of the criminal activities of

others (Karmen, 2001). Six different victimizations were used to tap into this strain, such

as being forced to give up money or possessions or being physically attacked by others.

Students were asked to indicate whether they have experienced an incident in the past 12

months (yes = 1) or have not (no = 0). Higher scores indicate a higher level of strain.

Negative emotion

Emotion refers to a person‘s response, after cognitive evaluation, to a stimulus

(e.g., an event, an object, or a person) related to his or her concerns or goals (Lazarus,

1991). Negative emotions, then, are present or felt when advancement of goals or

concerns is impeded. Simply put, a negative emotion reflects a gap between an ideal goal

and an actual goal (Larson & Asmussen, 1991; Lazarus, 1999; Solomon, 1976), and

matches well with GST.

The present study measures two important negative emotions: anger and

depression. The former, a central negative emotion in GST, is regarded as an outer-

directed emotion (Agnew, 1992; Jang & Johnson, 2003). The latter has gained increased

interest in research on GST because Agnew (2006a, 2006b) suggested that researchers

should pay more attention to other negative emotions, such as depression, an inner-

directed emotion (Ganem, 2008; Jang & Johnson, 2003).

Anger. Eight items were used to measure anger, five of which are adopted from

Spielberger‘s (1988) State-Trait Anger Expression Inventory (STAXI), which examines

116
anger as a personality trait that is situational (Wareham et al., 2005). One example of

these five items is ―When I get frustrated, I feel like hitting others.‖ The other three

items capture angry feelings or reactions that are also more situational. One example is

―It makes me mad when I don‘t get the respect from others that I deserve.‖ These eight

items appear to be primarily situational, which is in tandem with the suggestions of

Agnew and others (2006a; Baron, 2004; Mazerolle et al., 2003). Response categories

were coded so that a higher score indicates a higher level of anger (see Table B2 in

Appendix B for factor loadings for both anger and depression).

Depression. Four items are used to measure depression. In contrast to the anger measures,

which are more situational, these four items appear to be trait-like or symptoms of

depression. This may incur criticism because currently GST advocates use of measures

that tap into situational emotions. Three reasons may justify such use. First, most

measures of depression in psychology are similar to the present items, which capture

symptoms of depression, whether physical or behavioral (e.g., CES-D scale, Radloff,

1977). Second, studies that claim to measure situational distress/depression use similar

items/symptoms. The difference is that they ask respondents to answer questions based

on when they were experiencing strain. Third, in an even more direct measure of

situational depression by using vignettes, Ganem (2008) failed to capture pure situational

depression, in that respondents reported other emotions along with depression. Hence,

capturing situational depression may not be as easy as one would think, and without more

sophisticated strategies, one may measure negative emotions other than depression.

117
 The four depression items to be used in the present study are adopted from the

Beck Depression Inventory, second edition (BDI-II) (Beck, Steer, & Brown, 1996).

These items ask students to indicate how often the following statements describe them. (1)

I don‘t look forward to things as much as I used to, (2) I find it hard to keep my mind on

school work, (3) I sleep very well (reverse coded), and (4) I have lots of energy (reverse

coded). These four items include both a somatic component of depression (the last two

items) and an affective component of depression (the first two items) (Storch, Roberti, &

Roth, 2004). The response categories are identical to that of the anger measure (see

Appendix A for detailed wording). A higher score indicates a greater level of depression.

One caution must be presented here. The EFA results indicated that a 1 factor

solution is acceptable in the U.S. sample because the loading of these four items were all

over .4 except the first item, which had loading at .349. In contrast, these four items did

not load evenly in the Taiwanese sample, with the first two items having very low

loadings (see Table B2 in Appendix B for detail). However, the eigen value (sum of

squared loadings) was over 1 for the Taiwanese sample. Thus, this study still treated

depression as one factor in Taiwan so that the model construct was similar across the two

countries. The difference is due to the high correlation between the two somatic items in

the Taiwanese sample, but not in the U.S. sample. One possible explanation is that

Chinese people have the tendency to identify physical illness but report psychological

discomfort relatively seldom. This phenomenon, called somatization, refers to people

who emphasize physical symptoms of depression more than psychological ones (Heine,

2010). Moreover, in Chinese society, psychological illness incurs a great deal of stigma;

118
hence, the results obtained here would be expected in the Taiwan sample. However, to

make the model construct similar across the two countries, the present study still treated

depression as one factor for both samples (see Table 2 for the descriptive statistics for

strain, anger, and depression).

Demographic variables

The present study includes two important demographic variables, gender and age,

that have been shown to influence the strain-delinquency relationship. Studies have

shown that males and females may have different reactions and coping strategies under

strain/stress (Hoffmann & Su, 1997; Mirowsky & Ross, 1995; Perlin, 1989; Piquero &

Sealock, 2000; Sharp et al., 2005; Thoits, 1995; Van Gundy, 2002). Gender is therefore

included in the survey, with male coded 1 and female is coded 0. Age is also included

(students‘ age on the date of the survey), because as individuals grow older, they may

accumulate different experiences, and develop different responses to strain and different

coping skills (Hauser & Bowlds, 1990; Petersen, Kennedy, & Sullivan, 1991; Thoits,

1995; Turner, Wheaton, & Donald, 1995).

Race is not considered as a control variable because in Taiwan, there is essentially

no ―minority group,‖ except for some aboriginal individuals who constitute only 2% of

the total population (Directorate-General of Budget, Accounting, and Statistics,

Executive Yuan, Taiwan, 2010). One recent trend is an increase in numbers of so-called

―New Taiwanese Children,‖ children whose parents are not both Taiwanese. Usually the

mother is from another country (e.g., China, Vietnam), because the father is of relatively

119
low SES and thus has difficulty finding a mate in Taiwan. However, according to the

Ministry of Education (2009), in 2008, junior high school students from such families

comprised only 1% of the total junior high student population. Consequently, race is not

included as a covariate in the subsequent analysis for either sample (see Table 3 for the

frequency distribution for the two demographic variables).

Table 2 Descriptive Statistics of Strain and Negative Emotion Variables

U.S.(N = 1,516) Taiwan (N = 1,717)
Variable Mean¹ SD² Mean SD
Goal strain 15.03 3.70 14.57 4.13
Unjust strain 16.84 3.96 19.10 4.10
Life-event³ 0 4.97 0 5.25
Victimization 0 3.21 0 3.20
Anger 19.37 4.92 17.58 5.14
Depression 8.7 2.60 9.34 2.54
¹The reported mean is the average mean across the 5 datasets.
² The reported SD is the average SD across the 5 datasets.
³The means of Life-event and victimization scale are 0 because of standardization.

120
Table 3 Frequency Distribution of Demographic Variables, Aggression, and Delinquency
U.S. Taiwan
Gender n % Mean (SD) n % Mean (SD)
Male 730 48.2% 844 49.2%
.480 .490
Female 786 51.8% 873 50.8%
(.500) (.500)
Total 1,516 100% 1,717 100%
Age
11 185 12.2%
12 288 19% 93 5.4%
13 317 20.9% 452 26.3%
14 209 13.8% 306 17.8%
15 155 10.2% 13.780 383 22.3% 14.630
16 185 12.2% (1.984) 150 8.7% (1.604)
17 125 8.2% 297 17.3%
18 46 3% 31 1.8%
19 6 .4% 5 .3%
Total 1,516 100% 1,717 100%
Damage
property
Yes 206 13.6% 203 11.8%
.140 .120
No 1,310 86.4% 1,514 88.2%
(.343) (.323)
Total 1,516 100% 1,717 100%
Hit someone
Yes 357 24.5% 122 7.1%
.230 .070
No 1,159 75.5% 1,595 92.9%
(.424) (.257)
Total 1,516 100% 1,717 100%
Alcohol use
Yes 575 37.9% 474 27.6%
.380 .280
No 941 62.1% 1,243 72.4%
(.485) (.447)
Total 1,516 100% 1,717 100%
Aggression¹²
Never 524 37.6% 1,104 68.6%
1 205 14.7% 159 9.9%
2 140 10.1% 107 6.6%
2.000 2.030
3-5times 156 11.2% 96 6%
(1.819) (1.635)
6 or more
367 26.4% 144 8.9%
times
Total 1,392 100% 1,610 100%
¹The total number of aggression is not the same because of deleting those who have no
siblings.
²There is 5 missing cases in the U.S. sample and 1 in the Taiwan sample. These missing
cases were deleted.

121
Analytic strategy

Statistical method

The first purpose of the present study was to evaluate the GST model in the U.S.

and Taiwan. To achieve this, the study uses path analysis to test the basic GST model in

the U.S. and Taiwanese samples (see Figure 1)30. Using path analysis as a vehicle to

examine the research questions presented in Chapter IV has three advantages. First, path

analysis is commonly used to identify causal relationships and to test theoretical models

among manifest (observed) variables (Hatcher, 1994; Kline, 2005; Raykov &

Marcoulides, 2000).

Second, this statistical technique belongs to the family of structural equation

modeling (SEM)31; hence, many statistical estimations (e.g., ML, WLS) and methods

(e.g., multiple group comparison) may be applied to path analysis. Third, the GST model

also specifies mediating effects among theoretical variables, and James and Brett (1984)

suggest that researchers must use path analytic techniques to assess mediation. Baron and

Kenny (1986) likewise recommend path modeling to test mediation, noting that the

method allows simultaneous testing of all relevant paths.

30
The correlation between anger and depression was added because previous studies had found that
negative emotions were likely to co-occurred (Ganem, 2008; Sharp et al., 2001).
31
The SEM family includes several statistical models: path analysis models of observed variables,
confirmatory factor analysis models that examine the non-causal pattern of relationships among latent
constructs, structural regression models that specify causal relationships among latent constructs, and latent
growth models that examine effects over time (Kline, 2005; Raykov & Marcoulides, 2000).

122
 Goal Unjust Negative Victimization
blockage strain life-events

Anger Depression

Delinquency Aggression

Figure 1. The GST Path Model

Second, this statistical technique belongs to the family of structural equation

modeling (SEM)32; hence, many statistical estimations (e.g., ML, WLS) and methods

(e.g., multiple group comparison) may be applied to path analysis. Third, the GST model

also specifies mediating effects among theoretical variables, and James and Brett (1984)

suggest that researchers must use path analytic techniques to assess mediation. Baron and

Kenny (1986) likewise recommend path modeling to test mediation, noting that the

method allows simultaneous testing of all relevant paths.

32
The SEM family includes several statistical models: path analysis models of observed variables,
confirmatory factor analysis models that examine the non-causal pattern of relationships among latent
constructs, structural regression models that specify causal relationships among latent constructs, and latent
growth models that examine effects over time (Kline, 2005; Raykov & Marcoulides, 2000).

123
 To achieve the second purpose of the present study, that of comparing and

contrasting the GST model across countries, a SEM multiple-groups approach was used.

In conventional statistics, such as ANOVA, the group difference is estimated by a

categorical variable, which represents group membership (e.g., a dummy gender variable).

However, such a comparison is limited to mean difference; that is, delinquent boys, for

instance, have more delinquent friends than non-delinquent boys have. Other possible

differences, such as different processes, are ignored. Another conventional approach is to

fit the same model across different groups. That is, the same model was tested separately

in each group. Although this method provides a flexible approach to comparing groups, it

is also limited. For example, when a particular path is compared across groups, a

statistically significant result can indicate that a difference is present. However, this test

provides information on a difference rather than on a similarity. Hence, finding that

estimates are not significantly different, does not mean these parameters are the same

(Maruyama, 1997). For example, one might use Paternoster and colleagues‘ (1998)

equation to compare the path coefficients across male (b1) and female (b2) samples. The

z-score from this equation can provide information regarding whether b1 = b2. However,

like other significance tests, rejecting the null hypothesis does not mean one can accept

the alternative hypothesis because information is lacking on the type II error or power.

Moreover, the separate analyses each have lower statistical power. In contrast to these

conventional approaches, multiple-groups analysis in the SEM framework provides great

flexibility, offering direct comparisons across different groups on various parameters

(e.g., path coefficients, group means), and the comparisons reveal not only differences

124
but also similarities. For example, to test similarities, one can impose equality constraints

on paths across samples; if such constraint does not make the fit of the model worse, then

similarity is confirmed. In contrast, freeing one path provides the test for differences. In

addition, this approach utilizes all the subjects from all groups, which provides greater

statistical power. The sample size of both groups in the present study is over 1,500, which

is considered large by various statistical standards, and Maruyama (1997, p.259) argued

that with sufficiently large samples33, ―modeling groups as multiple populations is a

superior alternative to dummy coding.‖

Two points must be mentioned here before discussing missing data. First,

contrary to the requirements of regular SEM and path analysis models, which employ ML

(maximum likelihood estimation) and assume multivariate normality, the delinquent acts

in this study are all dichotomized variables, which violate the normality assumption.

Hence, regular ML estimation might be problematic. Although research has indicated that

ML is robust to minor departures from normality (Chou & Bentler, 1995), Brown (2006)

argued that when categorical data are encountered, ML should never be used. One

popular alternative estimator is weighted least squares (WLS), which adjusts the weight

matrix (W) by taking into account multivariate kurtosis in the variance/covariance matrix.

Brown (2006) argued that WLS does not perform well with small samples, and other

researchers also report that WLS does not provide proper estimations with the use of

categorical outcome variables (Flora & Curran, 2004). One proper alternative estimator

33
The argument here is to justify the use of multiple-groups analysis, not to devalue the MIMIC (multiple
indicators, multiple causes) approach. MIMIC is a valuable alternative to multiple-groups analysis. It has
several advantages: it is parsimonious and it requires fewer samples.

125
is weighted least square mean and variance (WLSMV), which is used to account for non-

normality. Flora and Curran (2004) have shown that WLSMV is an accurate estimator

when sample sizes vary from 100 to 1,000 with various degrees of non-normality and

model complexity. Each sample in the present study has more than 1,500 subjects; hence,

employing WLSMV as the estimator seems appropriate.

Second, several indirect effects are examined in the path model. Although the

traditional method is to use the Baron and Kenny (1986) approach or the Sobel test

(1982), simulation studies reveal that these two procedures perform poorly compared

with other modern methods (e.g., bootstrapping) (Fritz & MacKinnon, 2007; MacKinnon,

Lockwood, Hoffman, West, & Sheets, 2002). The problem with traditional methods is

that the significance of indirect effects is not distributed normally; hence, standard errors

and related significance tests are biased. To correct for this bias, bootstrapping can be

used. Bootstrapping is a resampling procedure in which the original data are randomly

drawn many times with replacement. These resampled samples provide a basis for

estimating the parameters of interest. Because of this random drawing with replacement,

no distribution assumption is involved. Brown (2006) called the sampling distribution

from bootstrapping ―concrete.‖ Hence, researchers often suggest that when the

assumption of normal-theory statistics is violated, bootstrapping is an attractive

alternative (Adèr, Mellenbergh, & Hand, 2008; Brown, 2006). Similarly, a growing

literature has advocated using bootstrapping when assessing indirect effects (Bollen &

Stine, 1990; Hayes, 2009; MacKinnon, Lockwood, & Williams, 2004).

126
Missing data

The present study employed two steps to handle missing data. First, students who

did not report either their gender or age were excluded from the analysis. For the U.S.

sample, about 27 students (1%) failed to report either gender or age, and for the Taiwan

sample, about 41 students (2%) did not report either gender or age. After these cases were

deleted, the U.S. sample had 1,647 subjects and the Taiwanese sample had 1,779 students.

However, further case deletion was done for ease of data imputation. Although multiple

imputation (MI) with the Markov Chain Monte Carlo (MCMC) approach can effectively

handle data with a high proportion of missing cases, it requires more iterations and

generates more complete sets of data, which is computationally intensive. To avoid this

problem, this study decided a priori that students who failed to report more than 2 items

on a scale were excluded from MI. For example, on the victimization scale (7 items),

three types of students were excluded from MI: the first type of students omitted all 7

items, the second type of students omitted 6 items, and the last type of students omitted 5

items. This criterion further reduced the U.S. sample to 1,516 and the Taiwanese sample

to 1,71734.

Second, after these cases were deleted, MI (Rubin, 1987) was used to replace the

missing values for all items. Although single imputation (e.g., mean replacement), in

which only a single value is used to replace the missing value, is easy to implement, two

serious disadvantages prevented the present study from employing this method. First,
34
For the U.S. sample, the negative life-event scale had most missing subjects (n = 73) who did not report
more than 2 items on the scale. On the other hand, for the Taiwan sample, the goal strain scale had the most
missing cases (n = 31). Other scales (e.g., anger, goal strain) together had 58 missing cases in the U.S. and
31 missing cases in Taiwan.

127
single imputation tends to underestimate standard errors, resulting in greater likelihood of

rejecting the null hypothesis. Second, it does not perform well even if the missing data

pattern is ignorable (MCAR or MAR). In contrast, MI uses several values to impute the

missing value, which results in multiple complete data sets. By using multiple sets of data

to estimate parameters of interest, a researcher adds variability into the estimation, which

can be used to adjust the standard error upward; that is, in turn it reduces Type I error

(McKnight, Mcknight, Sidani, & Figueredo, 2007). It has been shown that MI can be

successfully implemented on data missing not at random (MNAR) (Verbeke &

Molenberghs, 2000) and provide satisfactory results with minor departures from MAR

(Collins, Schafer, & Kam, 2001; Rubin, 1996)35. Moreover, this method has become the

most highly praised method for statistically dealing with missing cases (Allison, 2002;

Rubin, 1996; Shafer & Graham, 2002; McKnight et al., 2007), and is the dominant model

of handling missing data (Abraham & Russell, 2004). In addition, Rubin (1996) indicated

that MI not only provides generalizable estimates but also recovers variance for statistical

inference. On the basis of such literature support, using MI seemed to be justifiable and

appropriate.

MI with the MCMC approach is used in the present study. The advantage of using

the MCMC approach is that it can easily handle almost every kind of underlying

distribution; McKnight et al. (2007) have suggested MCMC as one approach to be used

when using MI with non-normal data, as in the present study, which has dichotomized

variables. Generally, the procedure involves two steps: the imputing (I-step) and the

35
In contrast this positive reference, Sinharay, Stern, and Russell (2001) indicated MI provided improved
estimation but it is still biased.
128
posterior (P-step). The I-step starts with an estimated mean and covariance matrix and

simulates a missing value for each observable data point. The P-step begins with the

complete data from the I-step and then generates a mean and covariance matrix based on

the posterior distribution36. The generated mean and covariance matrix from the P-step is,

then, used for the next I-step. The iteration between the I-step and P-step creates a

Markov chain with the goal of creating a distribution of missing data, from which

missing values are randomly drawn.

Before imputing missing values, both datasets were submitted to the Little‘s

(1988) MCAR chi-square test37, which helped to check whether the missingness of the

present data is completely at random (MCAR) or not. For the U.S. sample, the chi-square

test result was not significant at the .05 level, χ² (2,451, N = 1,516) = 2560, p = .06. This

indicated that the missing pattern in the U.S. sample is MCAR and MI is appropriate. In

contrast, the chi-square test result for the Taiwanese sample was significant, χ ² (980, N =

1,717) = 1109, p < .01. A series of comparisons across all the variables between missing

and non-missing cases revealed only a few significant differences, which should not be

necessarily viewed with caution. First, many comparisons were based on very few

missing cases. For example, when considering item4 on the unjust scale and item7 on the

anger scale, there are only 8 missing cases. With these few cases, the meaning of any

comparison is trivial. Second, the total number of pairs of comparisons is well over 100;

36
A posterior distribution is a distribution that is adjusted and updated based on information gained from
observing data.
37
This test compared observed means for each missing pattern with the expected population means, and
then, computed an overall weighted squared deviation. The test used the overall weighted squared deviation
and tested the null model (e.g., MCAR) by comparing it with the Chi-square distribution.

129
after the Bonferroni correction, one would need a p value far below .01 to obtain a

significant result. Many of the comparisons are significant at the .05 or .01 level, which

may be insignificant when applying the Bonferroni correction. Consequently, the

missingness might well be a matter of missing at random (MAR), which is also an

acceptable condition for using MI.

Besides the above concerns regarding conducting MI, some other decisions need

to be made. First, one must determine the number of iterations needed to achieve two

important conditions: (1) the algorithm has converged to the correct distribution and (2)

there is no statistical dependence between the observations in one generated data set and

another (Allison, 2003). Allison (2003, p.553) stated that ―[U]nfortunately, not much is

known about just how many iterations are needed to achieve these aims.‖ Thus, there is

no clear rule that can help researchers to make such decisions a priori.

For example, Allison (2002) argued that a small portion of missing data may be

estimated properly with a small number of iterations, but he did not provide specific

numbers. Similarly, Schafer (1997) presented different numbers of iterations, 50 to 1,000,

for conducting MI under different situations. Although there is no clear ―rule of thumb,‖

Allison (2002) did suggest that the number of iterations should be as large as it would be

with use of the EM (expectation maximization) algorithm. He later (Allison, 2003)

suggested that the default in the SAS program, 200 burn-in iterations and 100 iterations

to generate the first data set, is more than enough for the majority of missing data sets.

The present study uses 500 iterations to conduct MI, which is much greater than the

130
number of iterations when EM is used in SPSS38, and it is also greater than the default

mentioned by Allison (2003) in SAS.

The second related issue pertains to the number of complete data sets. The higher

the proportion of missing data, the more datasets are required. As mentioned before, the

missingness is not greater than 5% in both countries; hence, 2 or more complete datasets

should be enough. In fact, Rubin and Schenker (1986) suggested that 2 sets of imputation

are enough for missing at 10%. In a later Monte Carlo study, Schafer (1997) reported that

even with severe missingness (90% of the data missing), fewer than 20 imputations will

be required. The present study used the default, 5 complete data sets, which is usually

efficient (Allison, 2003). Schafer and Olsen (1998) have shown that 5 complete data sets

produce an efficient estimate even when the proportion of missing information is 50%. In

addition, Allison (2003, p.553) stated that 5 imputed data sets is ―widely regarded as

sufficient39 for a small to moderate amount of missing data.‖ As mentioned previously,

the fraction of missing cases in the present study is less than 5%, which is considered a

low amount of missing data. Consequently, 5 data sets are sufficient in the present

situation.

Although all these considerations have been taken into account, there is no

guarantee that the specified number of iterations would converge, given the large size of

the sample of the present study. Fortunately, there are some statistical methods that can

38
The default iteration in SPSS for EM is 25.
39
Efficient is defined ―[E]fficiency means that an estimator has a sampling variance that is at least as small
as that of any other estimator‖ (Allison, 2003, p.548, foot note 5).

131
be used to determine whether the problem of convergence is present. One simple

approach is to plot the parameter values (e.g., mean, SD) against the iteration history

(Schafer, 1997). For example, in the present study, the ideal imputation result is that the

mean and SD of the items for each scale are not varied across iteration, thus a stable

estimation of the mean and SD can be reached. Hence, plotting the mean and SD against

iteration history is checked. If there is no clear trend in the plot, then convergence has

been achieved. Allison (2002) suggested conducting this inspection on variables that have

the most missing values, because these are most likely the variables to be problematic.

For the U.S. sample, item 4 of the goal strain scale was selected because it has the largest

number of missing cases (n = 37); for the Taiwanese sample, item 7 of the unjust scale

was chosen (n = 19). Both plots showed no particular trend; in other words, they

appeared to be random (see Figure 2 for the U.S. sample and Figure 3 for the Taiwan

sample). This assured that the non-convergence problem was at a minimum.

132
 Panel A (Dataset number 1) Panel B (Dataset number 2)

Panel C (Dataset number 3) Panel D (Dataset number 4) Panel E (Dataset number 5)

Figure 2. The Iteration History Plot of Mean and Standard Deviation for the U.S. Sample

133
 Panel A (Dataset number 1) Panel B (Dataset number 2)

Panel C (Dataset number 3) Panel D (Dataset number 4) Panel E (Dataset number 5)

Figure 3. The Iteration History Plot of Mean and Standard Deviation for the Taiwanese Sample

134
 CHAPTER VI

RESULTS

The analyses that follow use the 5 complete datasets generated by the MI. The

first part of this chapter reports the results of applying the GST model in the U.S. sample.

The second part of this chapter presents the findings regarding to the Taiwanese sample.

The main purpose of these analyses is to examine the GST model, and to answer the

research questions of whether strain is related to delinquency and aggression, whether

strain is related to anger and depression, and whether strain affects delinquency and

aggression through anger and depression. The results also examine whether GST is

applicable to Taiwan.

The final part of this chapter provides the results from the multiple group analysis.

Although the first two parts examined the GST model, the analyses only focused on

whether GST is useful in explaining juvenile delinquency in each country and did not

examine the similarities and differences across two countries. With multiple group

comparison, the GST model can be further examined to see whether the same

relationships found in both countries are truly statistically similar (e.g., imposing

constraints). In addition, the multiple group analysis can also provide statistical

evaluation of the differences found in the separate analyses (e.g., freeing parameter).

135
 One caution must be made here before we turn to reported results. The sample

size of each country is well over 1,500 as such, statistically significant path coefficients

are likely to be detected. This is because sample size plays a role in null hypothesis

significance testing (NHST). The larger the sample size, the greater the ease of rejecting

the null hypothesis. As such, many scholars argue that one should not rely on NHST

alone but should also focus on substantively significant results40 (Cohen, 1994;

Nakagawa & Cuthill, 2007; Thompson, 1993). In other words, when statistically

significant results are almost certain, researchers should turn their attention to practically

significant results; that is, the result that has substantive meaning41. Applying this

thinking to the present case means that the path coefficients that are statistically

significant and are relatively large in magnitude are discussed more in the text than path

coefficients of relatively small magnitude. This is consistent with Levin (1993) who

argued that statistical significance should be built first and then the focus should turn to

practical significance.

GST in the U.S.

As previously mentioned, path analysis is adopted to examine the proposed causal

relationship among variables. Four outcome variables are incorporated in the analysis:

damaging property, hitting someone, alcohol use, and aggression against siblings. GST

would predict that students who experience strains should be more likely to commit

40
Substantive results refer to results that have a large effect size. In the present study, the substantive result
refers to the relatively large path coefficient, which can be seen as one kind of effect statistic.
41
The present study proposes the argument here is to justify the reporting practice in this study, but it does
not provide all the debates on the issue of effect size and NHST. Interested readers are referred to the whole
issue 4 of 1993 Journal of Experimental Education.
136
delinquent acts and aggression and experience negative emotions, such as anger and

depression. Table 4 has four models that give the results of the relationship between

strain and the four outcome variables. Browsing the results of Table 4, one finds that two

types of strain have statistically significant and sizeable effects on the three delinquent

acts. Negative life-events is related to damaging property, hitting someone, and alcohol

use. This result is consistent with previous reports that negative life-events is related to

increased alcohol use and other delinquency among youth (Broidy, 2001; Eitle, 2002;

Hoffmann, 2002; Hoffmann & Cerbone, 1999; Hoffmann et al., 2000). The more

consistent picture is that victimization has both statistically and substantively significant

effects on all four outcome variables. This suggests that victimization, as Agnew (2006a)

and previous studies have found (Hay & Evans, 2006; Lin, Cochran, & Mieczkowski,

2011), is a criminogenic strain. Besides these two strains, unjust strain has statistically

significant and relatively large effect on hitting someone and aggression against siblings.

Goal strain, however, did not have any direct effects on the four outcome variables. Some

previous studies have found similar null effects (Broidy, 2001; Moon et al., 2009; Sharp

et al., 2005).

Age and gender have different effects on these four outcome variables. For

example, older adolescents become involved in alcohol use more often than younger

adolescents, which is consistent with previous longitudinal analyses (Aseltine & Gore,

2000). On the other hand, older youth become less involved in aggression against siblings

than younger youth. Male students are more likely than female students to hit someone

137
and damage property. Surprisingly, female students report more aggressive behavior and

use more alcohol than male students.

Next, we turn to the examination of the relationship between strain and two

negative emotions: anger and depression. Figure 4 reveals that all four strain variables

have positive and significant effects on anger and depression. Among all the strain

variables, goal strain and unjust strain both have the largest statistically significant impact

on both anger and depression; that is, students who suffer from goal strain and unjust

strain are more likely to experience anger and depressive symptoms than their

counterparts who experience less of these two strains. These results provide support for

GST. However, Agnew (2006a) encouraged researchers to analyze the specific effect,

that is, which particular type of strain is related to which type of negative emotions. For

example, he suggested that the ―[r]esearcher should also examine whether particular

types of strains foster particular negative emotions…‖ (p.36). The present study, however,

did not find support for this statement in that all types of strains were related to anger and

depression. With regard to gender, consistent with previous studies, females experienced

higher depression than males, and there is no gender difference with regard to anger

(Broidy & Agnew, 1997; Broidy, 2001).

138
Table 4 The Relationship between all Strain Variables, Delinquency and Aggression in the U.S.¹²
Model1- Model2- Model3- Model4-
Variable Damaging Hitting Alcohol use Aggression
property someone
Goal strain .003(.014) .002(.008) .008(.012) .019(.014)
Unjust strain .010(.013) .026(.011)* .006(.011) .039(.013)**
Negative life-event .026(.009)** .017(.008)* .021(.007)** -.009(.009)
Victimization .068(.013)** .105(.011)** .075(.012)** .077(.014)**
Age .016(.025) -.034(.022) .289(.020)** -.099(.022)**
Male .161(.085) .317(.076)** -.188(.071)** -.446(.087)**

*p < .05. **p <. 01.

¹ Unstandardized coefficients are shown with standard errors in parentheses, and the
coefficients are the averaged estimation across the 5 data sets.
²The sample size is 1,516 for the three delinquent acts but is only 1,397 for aggression.

139
 GST argues that strain not only leads to delinquent reactions but also generates

negative feelings in the recipients, which in turn make the individual want to respond in

such a way as to correct the bad feeling. Although GST does propose that strain has

direct effects on delinquency, it is the indirect effect, through negative emotions, that

separates GST from other leading criminological theories (e.g., control theory). Table 5

presents the results of testing the full GST model, and Table 6 provides the results of total

indirect effects of a particular strain on the four outcome variables through anger and

depression.

In the full model, both anger and depression have significant influences on the

three delinquent acts and the effects are relatively large when compared to the effects of

all strain variables. However, only anger is significantly related to aggression against

siblings. As such, anger and depression are both potent candidates as mediators of strain

effects on the three delinquent acts but only anger could be a mediator when the outcome

is aggression (Baron & Kenny, 1986). The finding that anger is related to alcohol use is

surprising because previous studies suggest that anger is most likely to lead to outer-

directed delinquency, such as physical aggression (Aseltine et al., 2000; Jang & Johnson,

2005), rather than inner-directed coping strategies (e.g., substance use). In contrast,

depression was related to not only inner-directed delinquency (e.g., alcohol use) but also

outer-directed acts (e.g., hitting someone).

140
 Age
.298(.120)**
Goal strain .237(.178)**

Anger
.328(.264)**
Unjust strain .053(.053)*

.159(.242)** .262**

Negative life-event .305(.233)**

.122(.174)**

.212(.138)**
.028(.053)**
Victimization Depression
.051(.063)**

-.567(-.109)**
Male

Figure 4. The Path-Analytic Model of Strain and Negative Emotions in the U.S.
Unstandardized coefficients are shown with standardized coefficients in parentheses, and the
coefficients are the averaged estimation across the 5 data sets (insignificant paths are not shown ).
*p < .05. **p < .01.
141
Table 5 The Full GST Model in the U.S.¹²
Model for three delinquent acts(N = 1,516)
Damaging Hitting
Anger Depression Alcohol use
property someone
Goal strain .237(.032)** .122(.018)** -.021(-.014) -.017(.013) -.010(.012)
Unjust strain .328(.031)** .159(.018)** -.023(-.013) 0(.012) -.018(.011)
Negative life-event .053(.023)* .028(.012)* .020(.009)* .013(.008) .017(.007)*
Victimization .212(.036)** .051(.019)** .051(.012)** .091(.011)** .063(.012)**
Age .298(.065) .305(.030)** -.027(.024) -.068(.023)** .255(.020)**
Male .065(.230)** -.567(-.122)** .204(.083)* .352(.075)** -.148(.071)*
Anger .060(.009)** .047(.009)** .038(.008)**
Depression .082(.017)** .067(.017)** .075(.015)**
Model for aggression (N = 1,397 )
Anger Depression Aggression
Goal strain .245(.038)** .124(.020)** 0(.014)
Unjust strain .322(.036)** .159(.019)** .013(.013)
Negative life-event .050(.024)* .032(.013)* -.013(.009)
Victimization .209(.038)** .044(.020)* .061(.014)**
Age .283(.061)** .311(.032)** -.122(.022)**
Male .098(.240) -.565(.126)** -.45(.086)**
Anger .076(.010)**
Depression .006(.019)
*p < .05. **p < .01.
¹Unstandardized coefficients are shown with standard errors in parentheses, and the coefficients are the averaged estimation from the 5
complete datasets.
²All the models are estimated with anger being correlated with depression. The correlation is .262 when the outcome variables are the
three delinquent acts and is .260 when the outcome variable is aggression.

142
Although these findings may not echo those of previous studies in the GST literature,

results of studies from other areas have found similar results. For example, anger was

found to be related to the amount of alcohol consumption and substance use (Eftekhari,

Turner, & Larimer, 2004; Hussong & Chassin, 1994; Terrell, Miller, Foster, & Watkins,

2006), and even negative emotions closely related to anger, such as hostility (Hussong,

Hicks, Levy & Curran, 2001), increased alcohol use.

Similarly, depression also has strong and statistically significant effects on all

three delinquent acts. This finding is consistent with research by Beyers and Loeber

(2003), who found that depression was related to other-directed delinquent behaviors,

such as shoplifting and using force to get something. However, the present finding is

surprising because previous studies in the GST literature found that depression was

related to inner-directed deviant behavior, such as substance use (Jang. 2007; Jang &

Johnson, 2003) or purging (Sharp et al., 2005), not outer-directed delinquency.

The most pronounced result is that victimization has statistically significant and

relatively large effects on all the outcome variables and on both of the two negative

emotions. This suggests that students who experience more victimization incidents are

more likely to react to this stressor with delinquency and to display anger and depression.

Whereas the indirect effects from victimization to the four outcome variables through

anger are all significant, victimization has significant indirect effects on the three

delinquencies through depression only. The indirect relationship between victimization

and aggression through depression is not significant. While the indirect effects on the

three delinquent acts are all significant, these indirect effects account for only 13% to
143
25% of the total effects (see Table 7). These results again indicate that criminal

victimization, whether violent or property victimization, is a criminogenic strain (Agnew,

2006a), and most of its detrimental effect comes from the victimization itself.

Negative life-events is moderately and significantly related to property damage

and alcohol use but only marginally related to hitting someone. With regard to mediating

effects, negative life-events has indirect effects through anger on the four outcome

variables but has such effects through depression only on the three delinquent acts . The

most interesting difference between Table 4 and Table 5 is that of the negative life-event

→ hit someone relationship, which changes from significant to marginally significant

when both anger and depression were in the model. According to Baron and Kenny

(1986), this would indicate that a significant mediating effect exists. In Table 6, the

indirect effects of negative life-event to hitting someone through anger (.002) and

depression (.002) are all significant, although the effects are small. However, the

mediating effect is only partial because the direct effect still accounts for over 70% of

total effects of negative life-event to hitting someone (see Table 7).

In contrast to victimization and negative life-events, goal strain and unjust strain

are related to delinquency and aggression mostly through anger and depression. The total

indirect effect, for instance, of goal strain to property damage is greater than the direct

effect. For unjust strain, its impact on hitting someone, for example, is mediated through

anger (58%) and depression (38%).

144
Table 6 The Indirect Effects of Strains on Outcome Variables through Anger and Depression in the U.S.¹²³
Damaging
Variable Hitting someone Alcohol use Aggression
property³

Goal strain→Anger .014[.008, .021]² .011[.006, .017 ] .009[.005, .015] .019[.011, .027]
Unjust strain→Anger .020[.013, .028] .016[.010, .022] .012[.006, .018] .025[.016, .033]
Negative life-event→Anger .003[.007, .021] .002[0, .005] .002[0, .012] .004[0, .008]
Victimization→Anger .013[.008, .019] .010[.006, .015] .008[.004, .012] .016[.009, .023]
Goal strain→Depression .010[.006, .016] .008[.003, .014] .009[.005, .015] NSª
Unjust strain→Depression .013[.008, .019] .01[.005, .015] .012[.007, .018] NS
Negative life-event→Depression .003[0, .004] .002[0, .004 ] .002[0, .004] NS
Victimization→Depression .004[.001, .009] .004[.001, .007] .004[.001, .008] NS

¹The indirect effect reported here is the average from the 5 complete datasets.
² The total sample size is 1,516 for the three delinquent acts, and is 1,397 for aggression.
³ 95% confidence interval is in the bracket.
ªNon-significant indirect effect is not reported here.

145
Moreover, unjust strain has strong indirect effects on aggression against siblings through

anger, which accounts for 64% of total effects (see Table 7). This latter result not only

supports GST‘s contention regarding the mediating role of negative emotion in the GST

model (Agnew, 1992, 2001) but also undergirds recent empirical evidence of the

supremacy of state-like measures of negative emotion when more situational measures of

negative emotions are used (Capowich et al., 2001; Mazerolle et al., 2003).

In sum, two general conclusions can be reached. First, victimization has

statistically significant and relatively strong impact on students‘ life; that is, students who

experience criminal victimization have higher probability of involvement in delinquency

and aggression and these students are also more likely to experience anger and depression,

which in turn leads to greater involvement in delinquency, compared with students who

do not have such experience. Negative life-events, a commonly used measure of strain,

also has statistically significant and moderate effects on delinquency but not on

aggression. Goal strain and unjust strain, on the other hand, exert few or no direct effects

on delinquency and aggression; most of the effects these two strains have on the outcome

variables are through anger and depression.

Second, the proposed mediating effects of negative emotions on the strain-

delinquency relationship were present. In some cases, a total mediating effect was found.

For example, the unjust strain → hitting someone relationship was totally mediated by

anger and depression. This result not only supports GST‘s assertion that negative

emotion plays a causal role in crime and delinquency but also supports Agnew‘s (2006a,

2006b) argument that negative emotions other than anger should be included in the GST

146
model. On the other hand, both anger and depression are related to delinquency

regardless of whether it is inner-directed (e.g., alcohol use) or outer-directed delinquency

(e.g., hitting someone). This is unexpected, because GST would predict the domain

match relationship (e.g., outer-directed emotion (anger) → outer-directed coping (hitting

someone). However, studies from other areas provide some empirical support for the

mismatch results (Eftekhari et al., 2004; Beyers & Loeber, 2003; Terrell et al., 2006).

Table 7 The Direct, Indirect, and Total Effects in the U.S.

Paths Direct effect Indirect effect Total effect
Goal strain → damaging property -.021 .024 .003
Unjust strain → damaging property -.023 .033 .010
Negative life-event → damaging property .020 .006 .026
Victimization → damaging property .051 .017 .068
Goal strain → hitting someone -.017 .019 .002
Unjust strain → hitting someone 0 .026 .026
Negative life-event → hitting someone .013 .004 .017
Victimization → hitting someone .091 .014 .105
Goal strain → alcohol use -.010 .018 .008
Unjust strain → alcohol use -.018 .024 .006
Negative life-event → alcohol use .017 .004 .021
Victimization → alcohol use .063 .012 .075
Goal strain → aggression 0 .019 .019
Unjust strain → aggression .013 .026¹ .039
Negative life-event → aggression -.013 .004 -.009
Victimization → aggression .061 .016 .077
¹ The indirect effect is slightly greater than that in Table 5 because of the non-significant
indirect effect through depression.

147
GST in Taiwan

The above analyses show that the GST model is useful in explaining juvenile

delinquency and aggression in the U.S. This is understandable because the origins of

GST, the classic strain theory, and GST itself were developed in the U.S. As reviewed in

Chapter III, the cultural background of the U.S. is different from that of Taiwan. This

may raise the question of whether GST is extendable to non-western countries, such as

Taiwan. This section directly addresses this issue by using the data derived from the same

survey instrument, similar subjects, and identical statistical and theoretical models. In

addition, this can be considered as a replication of the U.S. study, which according to

Robinson and Levin (1997, p.26), is valuable in that it can not only ―confirm previous

findings but also extend those findings to new possibility.‖ This is especially true in the

present study because the replication is done by samples from different culture.

We first investigated whether the same four strain variables used in the U.S.

analysis have any effects on the three delinquent acts and on aggression. Table 8 provides

the results of this inquiry. Although goal strain and unjust strain have statistically

significant effects on delinquent acts, the effects are relatively small. In contrast, negative

life-events and victimization are statistically significant and strong predictors of

delinquency and of aggression. These results confirmed what was found in the U.S.

sample that victimization and negative life-events are crimnogenic.

Negative emotion plays an important role in GST; in fact, GST argues that strain

makes an individual feel bad, i.e., strain engenders negative emotions. The next step is to

test whether this argument holds in Taiwan. Figure 5 presents this test.
148
Table 8 The Relationship between all Strain Variables, Delinquency and Aggression in
Taiwan¹
Model1-
Model2-Hitting Model3- Model4-
Variable Damaging
someone Alcohol use Aggression
property²

Goal strain .024(.011)* .008(.012) .006(.009) .030(.009)**

Unjust strain .025(.011)* .031(.011)** .019(.009)* .007(.009)
Negative life-event .030(.007)** .039(.008)* .027(.006)** .016(.006)**
Victimization .031(.009)** .034(.010)** .013(.009) .032(.010)**
Age .004(.027) -.066(.032)* .061(.021)** -.144(.020)**
Male .445(.085)** .403(.100)** .039(.066) .071(.065)

*p < .05. **p <. 01.

¹ Unstandardized coefficients are shown with standard errors in parentheses, and the
coefficients are the averaged estimation across the 5 data sets.
²The sample size for the three delinquent acts is 1,717 but is only 1,610 for aggression.

As can be seen, the relationships between the four strain variables and anger and

depression are all significant. Among all the strain variables, goal strain and unjust strain

have strong and statistically significant effects on depression and anger. This is consistent

with the results found in the U.S. sample. One noticeable finding is that gender does not

have a significant and negative effect on depression. Previous studies have shown that

females usually experience and report higher levels of depression than do males

(Mirowsky & Ross, 1995; Sharp et al., 2005). However, as explained in a previous

chapter, Chinese people are more likely to identify somatic problems rather than

psychological ones. Consequently, a gender difference could be wiped out because of this

tendency.

149
 Age
.322(.101)**
Goal strain
.138(.111)**
Anger
.204(.162)**
Unjust strain .097(.099)**

.098(.157)** .186**

Negative life-event .133(.216)** .11(.069)**

.032(.067)*
*
Victimization .129(.080)** Depression
.

.651(.063)**
Male

Figure 5. The Path-Analytic Model of Strain and Negative Emotions in Taiwan:

Unstandardized coefficients are shown with standardized coefficients in parentheses, and the
coefficients are the averaged estimation across the 5 data sets (insignificant paths are not shown ).
*p < .05. **p < .01.

150
Also, male students report a higher level of anger than do female students. Although

Broidy and Agnew (1997) argue that females may experience a higher level of anger than

males, a study from Singapore showed that women from a Chinese culture, which is

patriarchal and places high priority on relational harmony, hesitated to reveal their anger

(Tanzer, Sim, & Spielberger, 1996). Consequently, female students may tend to

underreport their anger more than male students do.

The above two models show that students in Taiwan who experience strain are

more likely not only to experience anger and depression but also to be involved in

delinquency and aggression. However, these tests were performed only to build the

foundation for further examination of the full GST model, in which all variables are

incorporated in one model simultaneously. Table 9 provides the results of testing the full

GST model. Anger and depression mediate most of the relationships between goal strain,

unjust strain, and outcome variables (see Table 10).

More dramatically, anger and depression have totally mediated the goal strain-

hitting someone relationship; that is, the effects that goal strain have on hitting someone

is through anger and depression. Similarly, unjust strain also lost its influence on

damaging property and alcohol use. The indirect effect, for instance, from this strain

through anger to alcohol use is .004 and .006 through depression. The total indirect

effects account for about 53% of the total effects of unjust strain on alcohol use (see

Table 11). This indicates that most of the negative influence of unjust strain on youths is

through its instigation of anger and depression.

151
Table 9 The Full GST Model in Taiwan¹²
Model for three delinquent acts(N = 1,717)
Damaging Hitting
Anger Depression Alcohol use
property someone
Goal strain .138(.025)** .133(.014)** .014(.011) -.003(.012) -.005(.009)
Unjust strain .204(.027)** .098(.015)** .014(.011) .019(.011) .009(.009)
Negative life-event .097(.020)* .032(.012)** .026(.007)** .033(.008)** .023(.006)**
Victimization .129(.026)** .035(.017)* .026(.009)** .027(.010)** .008(.009)
Age .322(.080)** .110(.036)** -.010(.027) -.084(.032)** .047(.021)*
Male .652(.242)** -.143(-.118) .432(.085)** .381(.099)** .033(.066)
Anger .029(.008)** .043(.009)** .022(.006)**
Depression .046(.016)** .037(.017)* .059(.013)**
Model for aggression (N = 1,611 )
Anger Depression Aggression
Goal strain .121(.032)** .128(.016)** .024(.009)**
Unjust strain .201(.032)** .096(.016)** -.001(.009)
Negative life-event .099(.024)** .030(.011)* .013(.006)*
Victimization .140(.039)** .037(.019) .027(.01)**
Age .339(.077)** .119(.037)** -.157(.020)**
Male .654(.248)** -.148(.121) .052(.065)**
Anger .033(.007)**
Depression .015(.014)
*p < .05. **p <. 01.
¹Unstandardized coefficients are shown with standard errors in parentheses, and the coefficients are the averaged estimation from the 5
complete datasets.
²All the models are estimated with anger being correlated with depression. The correlation is .186 when the outcome variables are the
delinquent acts and is .177 when the outcome variable is aggression.

152
Table 10 The Indirect Effects of Strains on Outcome Variables through Anger and Depression in Taiwan¹
Damaging
Variable Hitting someone Alcohol use Aggression
property²

Goal strain→Anger .004[.001, .008]³ .006[.002, .011 ] .003[.001, .006] .004[.001, .008]
Unjust strain→Anger .006[.003, .009] .009[.005, .014] .004[.002, .008] .007[.003, .011]
Negative life-event→Anger .003[.001, .005] .004[.002, .007] .002[.001, .014] .003[.001, .006]
Victimization→Anger .004[.001, .008] .006[.001, .012] .003[0, .007] .005[.001, .010]
Goal strain→Depression .006[.001, .011] .005[0, .010]ª .008[.004, .013] NSª
Unjust strain→Depression .005[.001, .008] .003[0, .008]ª .006[.003, .009] NS
Negative life-event→Depression .001[0, .003] .002[0, .003 ] .002[.001, .003] NS
Victimization→Depression .001[0, .004] .001[0, .004] .002[0, .005] NS

¹The indirect effect reported here is the average from the 5 complete datasets.
² The total sample size is 1,717for the three delinquent acts, and is 1,611 for aggression.
³ 95% confidence interval is in the bracket.
ªNS: Non-significant at .05 and .1 level.

153
This indicates that most of the negative influence of unjust strain on youths is through its

instigation of anger and depression. This result reveals the importance of examining the

full GST model; leaving out negative emotions may provide only a partial picture of the

complex relationships between strain and delinquency in the adolescent year.

Negative life-events continually exert significant effects on all the outcome

variables. This is consistent with much previous research, whether the study was

conducted in western or eastern countries (Aseltine & Gore, 2000; Eitle & Turner, 2003;

Lin, in press; Lin & Mieczkowski, 2011). However, the indirect effects account only for

13% to 19% of the total effects of negative life-event on all the outcome variables.

Consequently, negative life-events are detrimental to youths primarily because of direct

effects.

Victimization, like negative life-events, has significant effects on delinquency and

aggression but not on alcohol use. As in the case of negative life-events, the most

negative influence comes from the victimization experience itself, because the indirect

effects account only for 14% to 20% of all the effects. The only exception is that anger

and depression have strong mediating effects on alcohol use. The indirect effect accounts

for about 38% of the total effect. Similar to the results with the U.S. sample,

victimization, measured in the present study, is detrimental to youths in Taiwan; and a

similar result was found in one previous study, which used a random sample of

Taiwanese adolescents (Lin & Mieczkowski, 2011).

154
Table 11 The Direct, Indirect, and Total Effects in Taiwan
Paths Direct effect Indirect effect Total effect
Goal strain → damaging property .014 .010 .024
Unjust strain → damaging property .014 .011 .025
Negative life-event → damaging property .026 .004 .030
Victimization → damaging property .026 .005 .031
Goal strain → hitting someone -.003 .011 .008
Unjust strain → hitting someone .019 .012 .031
Negative life-event → hitting someone .033 .006 .039
Victimization → hitting someone .027 .007 .034
Goal strain → alcohol use -.005 .011 .006
Unjust strain → alcohol use .009 .010 .019
Negative life-event → alcohol use .023 .004 .027
Victimization → alcohol use .008 .005 .013
Goal strain → aggression .024 .006¹ .030
Unjust strain → aggression -.001 .008¹ .007
Negative life-event → aggression .013 .003 .016
Victimization → aggression .027 .005 .032
¹ The indirect effect is slightly greater than that in Table 5 because of the non-significant
indirect effect through depression.

In sum, the results of the analysis in Taiwan generally support GST. First,

negative life-events and victimization have statistically significant and large direct effects

on delinquency and aggression. These two strains also have statistically significant

effects on anger and depression. As in the U.S., victimization and negative life-events are

detrimental to youths not only because of the strain itself but also through the negative

emotions—anger and depression—that ensue. However, the indirect effects account only

for 10 to 20 percent of total effects. Hence, the most negative influence of these two

strains on adolescents lies in the strain itself, a conclusion consistent with results of

previous studies that found that negative life-events and victimization both have effects

on delinquency and aggression.

155
 Second, although goal strain and unjust strain have various direct effects on

delinquency and aggression, in the full GST model, anger and depression mediate most

of their effects on the four outcome variables. The most significant change is that the

influences of goal strain on hitting someone and alcohol use are reduced to almost 0 but

the effects on hitting someone and alcohol use are taken up by anger and depression. This

same result is also found in the unjust strain-aggression relationship; in addition, anger

and depression totally mediated the strain-delinquency relationship.

Multiple group analysis

The final analysis provides the statistical basis for comparing the GST model

across the U.S. and Taiwan. As mentioned in the previous chapter, multiple group

analysis is preferable to other methods with regard to comparing path models across

different populations. Multiple group analysis is not only capable of discovering

similarities (e.g., imposing equality on a parameter) but also suitable for revealing

differences (e.g., freeing a parameter). The present study uses a step-up approach42 to

conduct the multiple group analysis; that is, the GST model is free to be estimated for

each group and then restrictions on a path are imposed one by one. Imposing a constraint

on a path frees a degree of freedom; if the chi-square difference is not over 3.84, the path

is said to be the same across countries. In contrast, if the chi-square difference is greater

than 3.84, the constrained path is said to be different43. A model with more degrees of

freedom is more parsimonious than a model with fewer degrees of freedom if both
42
In contrast to a step-down approach, a researcher starts with the most restricted model, and subsequent
models are evaluated by sequentially relaxing the constraints.
43
The chi-square is not directly useable for the chi-square difference testing because the estimator is not
ML but WLSMV. Fortunately, Mplus provides syntax (DIFFTEST) to fulfill this task.

156
models fit the data similarly. Hence, a parsimonious model is preferred to a complex

model with similar fit.

The restrictions that are enforced are based on the results of separated analyses

because the empirically found similarities and differences can be further tested. Before

any restrictions are imposed, the least restricted model, or the most complex model (e.g.,

all paths are free to be estimated) was tested. The results can be found in Table 12 and

Table 13. As can be seen, unjust strain and goal strain have no direct effects on

delinquency in both samples. Similarly, negative life-event and victimization continue to

exert statistically significant and large direct effects on delinquency and aggression in

both countries. The only exception is that in the U.S., negative life-events does not have a

significant effect on aggression, which is not the case in Taiwan. In addition, anger and

depression remain potent risk factors in delinquency in both countries, but only anger

increases the risk of involvement in aggression in both U.S. and Taiwan.

157
Table 12 The Full GST Model-Multiple Group Analysis for Delinquent Acts¹²³
Model for the U.S.(N = 1,516)
Damaging Hitting
Anger Depression Alcohol use
property someone
Goal strain .237(.032)** .122(.018)** -.021(.014) -.017(.013) -.010(.012)
Unjust strain .328(.031)** .159(.018)** -.023(.013) 0(.012) -.018(.011)
Negative life-event .053(.023)* .028(.012)* .020(.009)* .013(.008) .017(.007)*
Victimization .211(.036)** .051(.019)** .051(.012)** .091(.011)** .063(.012)**
Age .298(.080)** .305(.030)** -.027(.024) -.068(.023)** .255(.02)**
Gender .066(.242) -.567(.122)** .204(.083)* .352(.075)** -.148(.071)*
Anger .060(.009)** .047(.009)** .038(.008)**
Depression .082(.017)** .067(.017)** .075(.015)**
Model for Taiwan (N = 1,717)
Damaging Hitting
Anger Depression Alcohol use
property someone
Goal strain .138(.025)** .133(.014)** .014(.011) -.003(.012) -.005(.009)
Unjust strain .204(.027)** .098(.015)** .014(.011) .019(.011) .009(.009)
Negative life-event .097(.02)** .032(.012)* .026(.007)** .033(.008)** .023(.006)**
Victimization .129(.026)** .035(.017) * .026(.009)** .027(.010)** .008(.009)
Age .322(.080)** .110(.036)** -.010(.020) -.084(.032)** .047(.021)*
Gender .651(.242)** -.143(.118) .432(.085)** .381(.099)** .033(.066)
Anger .029(.008)** .043(.009)** .022(.006)**
Depression .046(.016)** .037(.017)* .059(.013)**
*p < .05. **p <. 01.
¹Unstandardized coefficients are shown, with standard errors in parentheses, and the coefficients are the averaged estimation from the
5 complete datasets.
²All path are free to be estimated.
³ Model is estimated with anger correlated with depression (r = .262-U.S.; r = .186-Taiwan).
158
Table 13 The Full GST Model-Multiple Group Analysis for Aggression¹²³
Model for the U.S. (N = 1,397)
Anger Depression Aggression
Goal strain .245(.038)** .124(.020)** 0(.014)
Unjust strain .322(.036)** .159(.019)** .013(.013)
Negative life-event .050(.024)* .032(.013)* -.013(.009)
Victimization .209(.038)** .044(.02)* .061(.014)**
Age .283(.061)** .311(.032)** -.123(.022)**
Gender .098(.240) -.565(.126)** -.045(.085)**
Anger .077(.010)**
Depression .004(.019)
Model for Taiwan (N = 1,611)
Anger Depression Aggression
Goal strain .121(.032)** .128(.016)** .024(.009)**
Unjust strain .201(.032)** .096(.016)** 0(.009)
Negative life-event .099(.024)** .030(.011)** .013(.006)*
Victimization .140(.039)** .037(.019) .027(.010)**
Age .339(.077)** .119(.037)** -.157(.020)**
Gender .654(.248)** -.148(.121) .053(.065)
Anger .033(.007)**
Depression .015(.014)
*p < .05. **p <. 01.
¹Unstandardized coefficients are shown, with standard errors in parentheses, and the
coefficients are the averaged estimation from the 5 complete datasets.
²All path are free to be estimated.
³Model is estimated with anger correlated with depression (r = .260-U.S.; r = .177-
Taiwan).

All these results indicate that if we stop here, the conclusion is that the GST

model is reasonably similar across cultural boundaries. This conclusion is important,

because the data were collected on the basis of almost identical survey items and because

the multiple group analysis allows direct comparison across populations. This conclusion

confirms that of previous studies that only indirectly compared results across countries

(Bao et al., 2004; Lin & Mieczkowski, 2011; Maxwell, 2001; Moon & Morash, 2004).

However, more can be gained if the differences of path coefficients are directly examined.

159
 The results from Table 12 and 13 provide the basis for imposing constraints on

paths. The most interesting differences44 related to damaging property are: goal strain →

damaging property, unjust strain → damaging property, and negative life-event →

damaging property. In addition, although the effects of strains on anger and depression

are all significant for both countries, the magnitude differs and studies have suggested

that cultural influences affect the expression of negative emotions. As such, imposing

constraints on the strain-negative emotions paths are also warranted.

Table 14 provides a summary of the results of the constraints imposed on paths as

outlined above. As can be seen, students who experienced negative life-events are more

likely to experience depression and anger in both the U.S. and Taiwan, and the magnitude

is the same for these strain-negative emotion relationships. On the other hand, students

who experience victimization are more likely to experience similar magnitude of

depression in both countries than students who do not have the experience. The same

applied to the goal strain-depression relationship. However, while unjust strain and goal

strain both increase students‘ anger, the magnitude is not the same across countries. The

magnitude is always stronger in the U.S. than is it in Taiwan.

44
Three criteria are used to impose constraints on a path: a path is significant in one country but not in
another country, the sign of the path is different across countries, or the difference of the magnitude of the
path is dramatically different across countries.

160
Table 14 The Full GST Model-Multiple Group Analysis with Constraints (Damaging Property)¹²³
Anger Depression Damaging property
U.S. Taiwan U.S. Taiwan U.S. Taiwan
Goal strain .230(.032)** .140(.025)** .129(.011)** .001(.005)
Unjust strain .328(.031)** .204(.027)** .159(.018)** .098(.015)** -.022(.013) .014(.011)
Negative life-event .077(.015)** .030(.009)** .023(.005)**
Victimization .211(.036)** .129(.011)** .042(.013)** .052(.013)** .025(.009)**
Age .298(.059)** .322(.080)** .305(.030)** .111(.036)** -.025(.024) -.011(.027)
Gender .066(.230) .651(.242)** -.567(.122)** -.143(.118) .202(.083)** .432(.085)**
Anger .058(.009)** .030(.008)**
Depression .079(.017)** .049(.016)**
*p < .05. **p <. 01.
¹Unstandardized coefficients are shown, with standard errors in parentheses, and the coefficients are the averaged estimation from the
5 complete datasets (N = 1,516-U.S.; N = 1,717-Taiwan).
² Model is estimated with anger correlated with depression (r = .268-U.S.; r = .185-Taiwan).
³ The model fits the data well: χ²(6) = 7.21, NS; CFI = .999; TLI = .992; RMSEA = .011

161
Table 15 The Full GST Model-Multiple Group Analysis with Constraints (Hitting Someone)¹²³
Anger Depression Hitting Someone
U.S. Taiwan U.S. Taiwan U.S. Taiwan
Goal strain .237(.033)** .138(.025)** .129(.011)** -.017(.013) -.003(.012)
Unjust strain .325(.031)** .207(.028)** .157(.018)** .098(.015)** .010(.008)
Negative life-event .077(.015)** .030(.008)** .023(.005)**
Victimization .211(.036)** .129(.026)** .042(.013)** .092(.011)** .027(.010)**
Age .298(.059)** .323(.080)** .305(.030)** .110(.036)** -.067(.023)** -.084(.032)**
Gender .064(.230) .651(.242)** -.567(.122)** -.143(.118) .351(.075)** .380(.099)**
Anger .045(.006)**
Depression .065(.017)** .039(.017)*
*p < .05. **p <. 01.
¹Unstandardized coefficients are shown, with standard errors in parentheses, and the coefficients are the averaged estimation from the
5 complete datasets (N = 1,516-U.S.; N = 1,717-Taiwan).
² Model is estimated with anger correlated with depression (r = .263-U.S.; r = .186-Taiwan).
³ The model fits the data well: χ²(7) = 8.37, NS; CFI =.999; TLI = .993; RMSEA = .011

162
 As reviewed in previous chapters, maintaining relational harmony is of high

priority in a collectivistic culture; in addition, the Confucian ethos regards the expression

of anger as immature. Therefore, the stronger effects may be attributed to the reluctance

of Chinese students to express anger, especially when the source of anger is goal-related

or relationship-related. When looking at results related to depression, the only cultural

difference is seen in the unjust strain-depression relationship, with this relationship being

stronger in the U.S. sample. The other three strain-depression relationships are similar

across the U.S. and Taiwan.

With regard to the three paths on which constraints were imposed, only one,

unjust strain → damaging property, was rejected (χ²(1) = 5.44, p<.05). Although the

magnitude and the sign of this path are different for the U.S. and Taiwan, the path has an

insignificant effect on damaging property in both countries. The other three imposed

constraints did not make the model fit worse; hence, goal strain does not have any

influence on damaging property in either country, whereas negative life-events exerts a

significant effect on damaging property in both countries, with the same magnitude.

Victimization, anger, and depression continue to have effects on damaging property in

both countries, although the magnitude of effects is different and is always stronger in the

U.S. than in Taiwan.

With regard to hitting someone, in addition to the imposed constraints on the

strain-negative emotion relationship, 4 other constraints were imposed: unjust strain →

hitting someone, negative life-events → hitting someone, victimization → hitting

someone, and anger → hitting someone. As shown in Table 15, the strain-negative
163
emotion relationships were all similar to those of Table 14. Of the four additional

imposed constraints, one was rejected (victimization → hitting someone) (χ²(1) = 21.78,

p<.01). Hence, although victimization is a risk factor for violent delinquency (e.g.,

hitting someone) in both countries, the effect of this stressor on hitting someone is

significantly stronger in the U.S. than in Taiwan.

Negative life-events is also an important risk factor for hitting someone in both

countries, but the influence is about the same. The most interesting similarity is that of

the anger-hitting someone relationship, because one would expect to see that this

relationship is stronger in the U.S. than in Taiwan. However, the imposed constraint does

not make model fit worse, which indicates that the relationship is the same across

cultures. This is surprising, because one would expect that Chinese students would have

lower levels of anger, which in turn leads to a lower incidence of violent acts. One

possible counter explanation is the recent surge of violent crime and campus violence in

Taiwan, as reviewed in Chapter III, which might make students more likely to vent anger

through violent acts. Moreover, what we discover here is the anger-hitting someone

relationship, which indicates that angered students are more likely to be involved in

violent delinquency but does not mean that students are equally angry or equally violent

in the two countries45.

45
The t-test showed that the U.S. students have significantly higher levels of anger than do Taiwanese
students (t = 10.1, p < .01). The proportion of students in Taiwan who report yes on hitting someone is only
7%, whereas the proportion for the U.S. is 24%.

164
 In Table 16, five paths deserve particular attention: goal strain → alcohol use,

unjust strain → alcohol use, victimization → alcohol use, anger → alcohol use, and

depression → alcohol use. The constraints on these four paths and those imposed on the

strain-negative emotions path are examined. The results (Table 16) are fairly similar to

those found in Table 14 and 15 insofar as strain-negative emotion is concerned. Of the

five listed paths, only one was rejected: victimization → alcohol use (χ²(1) = 14.93, p

< .01). Hence, victimization has dramatically different effects on alcohol use in both

countries. On the one hand, students in the U.S. drink alcohol to cope with victimization;

on the other hand, students in Taiwan do not employ such a coping strategy. This is

somewhat unexpected, but two explanations can be offered. In Taiwan, students are

under close supervision by family members and others because of the small land area and

crowding. Hence, opportunities for deviant behavior are greatly reduced. Another

possibility is that the drinking norms in the Chinese culture demand that individuals drink

with others or during feasts or meal time (Harrell, 1981). Students in Taiwan may be

aware of this norm, and may reject alcohol use as a coping strategy.

Other imposed paths do not make the model fit worse. As a result, the similarities

of the four paths are statistically confirmed. Hence, students in both countries who

experience anger and depression are more likely to use alcohol. However, it also shows

that goal strain and unjust strain have no effects on drinking alcohol in either the U.S. or

Taiwan. In contrast, negative life-events exerts a significant effect on alcohol use in both

countries, but the magnitude differs significantly.

165
Table 16 The Full GST Model-Multiple Group Analysis with Constraints (Alcohol Use)¹²³
Anger Depression Alcohol Use
U.S. Taiwan U.S. Taiwan U.S. Taiwan
Goal strain .237(.033)** .138(.025)** .129(.011)** -.007(.007)
Unjust strain .325(.031)** .207(.027)** .157(.018)** .099(.015)** -.001(.007)
Negative life-event .077(.015)** .030(.008)** .017(.007)* .023(.006)**
Victimization .212(.036)** .129(.026)** .042(.013)** .066(.012)** .007(.009)
Age .298(.059)** .322(.080)** .306(.030)** .110(.036)** .260(.020)** .044(.021)*
Gender .065(.230) .651(.242)** -.567(.122)** -.143(.118) -.152(.070)* .030(.066)
Anger .029(.005)**
Depression .067(.010)**
*p < .05. **p <. 01.
¹Unstandardized coefficients are shown, with standard errors in parentheses, and the coefficients are the averaged estimation from the
5 complete datasets (N = 1,516-U.S.; N = 1,717-Taiwan).
² Model is estimated with anger correlated with depression (r = .263-U.S.; r = .186-Taiwan).
³ The model fits the data well: χ²(8) = 9.31, NS; CFI =.999; TLI = .995; RMSEA = 0

166
 With regard to aggression (Table 17), constraints are imposed on four paths: goal

strain → aggression, unjust strain → aggression, negative life-events → aggression, and

victimization → aggression, in addition to the constraints imposed on strain-negative

emotions. Among all these imposed constraints, the latter two paths are rejected: negative

life-event → aggression (χ²(1) = 6.06, p<.05) and victimization → aggression (χ²(1) =

3.85, p<.05). Consequently, these two paths were seen to be different across countries.

The most dramatic change is the goal strain-aggression relationship, which becomes

significant in the U.S., although it was not significant before. Hence, goal strain and

victimization are related to aggression in both countries. Anger exerts significant effects

on aggression, which is consistent with GST‘s prediction; however, the magnitude is

different across countries, with a stronger effect in the U.S. sample. Similarly, depression

has no effects on aggression in either the U.S. or Taiwan

So far, the comparison has focused on the direct relationships between strain,

negative emotions, and outcome variables. Another important part of GST is

consideration of the indirect effect of strain on delinquency through negative emotions.

To examine whether a particular indirect effect is statistically different between the U.S.

and Taiwan, the Wald statistic is used46. The Wald statistic can be used to test the

specified parameter(s) all at once; that is, we can test all the indirect effects

simultaneously. However, it gives only the overall result (e.g., all the indirect effects are

the same).

46
The Wald statistic is compared against the Chi-square distribution. The number of the degrees of
freedom is based on the number of parameters tested in the hypothesis.
167
Table 17 The Full GST Model-Multiple Group Analysis with Constraints (Aggression)¹²³
Anger Depression Aggression
U.S. Taiwan U.S. Taiwan U.S. Taiwan
Goal strain .249(.037)** .121(.032)** .126(.012)** .017(.007)*
Unjust strain .319(.035)** .2(.032)** .158(.017)** .097(.015)** .003(.007)
Negative life-event .075(.017)** .031(.009)** -.013(.009) .013(.006)*
Victimization .175(.027)** .04(.014)** .059(.014)** .027(.010)**
Age .276(.061)** .33(.076)** .31(.032)** .119(.037)** -.118(.022)** .-156(.02)**
Gender .129(.239) .613(.247)* -.565(.125)** -.149(.12) -.47(.085)** .059(.065)
Anger .076(.010)** .033(.007)**
Depression .002(.018) .017(.013)
*p < .05. **p <. 01.
¹Unstandardized coefficients are shown, with standard errors in parentheses, and the coefficients are the averaged estimation from the
5 complete datasets (N = 1,397-U.S.; N = 1,611-Taiwan).
² Model is estimated with anger correlated with depression (r = .26-U.S.; r = .177-Taiwan).
³ The model fits the data well: χ²(7) = 5.51, NS; CFI =1; TLI = 1; RMSEA = 0

168
Hence, whether a particular indirect effect is statistically different between the U.S. and

Taiwan needs to be examined separately. The study used the Wald statistic to test all at

once indirect effects of the four strains, mediated by anger and depression, for each

outcome variable47. If the result is insignificant, no further individual test is performed. In

contrast, if the result turns out to be significant, an individual Wald test is conducted to

find out which of these indirect paths is significant.

Table 18 provides the results of the Wald statistic tests, including overall and

individual tests. As can be seen, the overall tests showed the consistent picture that the

indirect effects from strains through anger on delinquency and aggression were all

statistically different between the U.S. and the Taiwan. The indirect path through

depression on damaging property was the only one that differed across these two nations.

Close inspection of the significant overall tests provided a clear picture of the significant

group differences. For example, the indirect effect of negative life-events on aggression

through anger is not statistically different between the U.S. and Taiwan; however, the

indirect paths from the other three types of strain to aggression through anger are

statistically different.

Combining this particular result with those of Table 6 and Table 10, we

discovered that the goal strain → anger → aggression process was stronger in the U.S.,

for instance, than it was in Taiwan because the Wald test was significant and the indirect

effect was .019 in the U.S. and .004 in the Taiwan, a difference of over 4.5 times.

Another example from Table 18 is that the overall test for strains → depression →

47
As reported above, depression does not have a significant mediating effect on the strain-aggression
relationship. Hence, no Wald test is conducted for this indirect effect.
169
damaging property is significant but only unjust strain → depression → damaging

property is significant. When one looks at both Table 5 and Table 9, the indirect path

from unjust strain through depression to damage is seen to be twice as great in the U.S.

(.013) as in Taiwan (.005). Hence, students in both countries might experience depression

because of unjust strain but more students in the U.S. than in Taiwan cope with

depression by damaging property.

Table 18 The Wald test for Indirect Effect

Overall test χ² result
Strains→anger→damaging property χ²(4) = 18.01**
Strains→depression→damaging property χ²(4) = 10.67*
Strains→anger→hitting someone χ²(4) = 8.03†
Strains→depression→hitting someone χ²(4) = 7.2
Strains→anger→alcohol use χ²(4) = 9.91*
Strains→depression→alcohol use χ²(4) = 4.94
Strains→anger→aggression χ²(4) = 25.61**
Individual test χ² result
Goal strain→anger→damaging property χ²(1) = 10.62**
Unjust strain→anger→damaging property χ²(1) = 12.52**
Victimization→anger→damaging property χ²(1) = 7.8**
Unjust strain→depression→damaging property χ²(1) = 5.79*
Goal strain→anger→hitting someone χ²(1) = 2.89†
Unjust strain→anger→hitting someone χ²(1) = 2.93†
Goal strain→anger→alcohol use χ²(1) = 5.63**
Unjust strain→anger→alcohol use χ²(1) = 6.06*
Victimization→anger→alcohol use χ²(1) = 4.5*
Goal strain→anger→aggression χ²(1) = 13.71**
Unjust strain→anger→aggression χ²(1) = 16.08**
Victimization→anger→aggression χ²(1) = 8.45**
†p < .1, *p < .05, **p < .01

In sum, the multiple group analysis provided a deeper understanding of the GST

model in both countries. Some cultural differences were discovered, whether in the direct

or indirect effects. Generally, the strains used in the present study engendered depression

170
in students in both countries; in contrast, students in the U.S. were more likely than

students in Taiwan to respond to strain with anger. On the other hand, students in both

countries responded to negative life-events with similar levels of anger.

With regard to delinquency and aggression, victimization seemed to increase

students‘ involvement in delinquency and aggression in both countries, but the effects

were different, being stronger in the U.S. than in Taiwan. Goal strain and unjust strain

had similar non-significant effects on delinquency and aggression in both countries in

most cases. Negative life-events had similar and significant effects on hitting someone

and damaging property, but this same strain exerted different effects on alcohol use and

aggression in either country.

The final tests of indirect effects provided even closer examination of the GST

process. The general pattern was that students in the U.S. were more likely than

Taiwanese students to commit delinquent acts and aggression because of strain and anger.

On the other hand, the indirect effects of strains on outcome variables through depression

were very similar across countries. The only exception was that students in the U.S. who

experienced unjust strain, which caused depressive feelings, were more likely than their

counterparts in Taiwan to cope with the emotion by damaging property.

171
 CHAPTER VII

CONCLUSION AND DISCUSSION

The present study set out to exam the GST model in both the U.S. and Taiwan.

The major purposes were: whether the GST model was useful in explaining juvenile

delinquency in both countries and whether the GST processes were different in the two

countries. It is worth mentioning that this study directly compares the GST model in two

countries that have very different cultural backgrounds. In addition, this study used

stringent statistical methods to test the mediating effects of negative emotions in the GST.

Thus, the results of this study provide great insight into the issue of applying

criminological theory in a cultural setting other than that of Western countries. As

outlined in Chapter IV, this study attempted to answer six research questions. The first

and second questions were to determine whether the basic GST model could be utilized

in the U.S. and Taiwan, that is, whether strain affected delinquency and negative

emotions in both countries. The third and fourth questions asked whether anger and

depression, the two negative emotions examined in this study, mediated the strain-

delinquency relationship in both countries. The final two research questions were mainly

concerned with the similarities and differences of the GST processes in the U.S. and

Taiwanese samples. In order to answer these research questions, two sets of data

collected in the two countries by use of identical survey instruments, were examined by

172
path analysis, a statistical tool particularly useful in testing theoretical causal models and

indirect effects across different populations. This chapter provides a summary of the

major findings of the statistical analysis and discusses these findings.

Summary of findings

The findings of the present study can be divided into three parts: the GST model

in the U.S. sample, the GST model in the Taiwanese sample, and the multiple group

analysis of the GST model.

First, the study found that the basic GST model was useful in explaining juvenile

delinquency and aggression in both the U.S. and Taiwan. The results show that negative

life-events and victimization are detrimental to youngsters in both countries, not only

because they increase youthful delinquency involvement but also because they cause high

levels of depression and anger in students. In addition, victimization has positive effects

on aggression in both countries (see Table 19). These findings are consistent with those

of previous studies that found that victimization and negative life-events are criminogenic

to youth (Carson et al., 2010; Eitle & Turner 2002; Harrell 2007; Hay & Evans 2006; Lin

et al., 2011). However, victimization failed to have an effect on alcohol use in the

Taiwanese sample.

In contrast to these supportive results, goal strain, as measured in the present

study, did not have a statistically significant effect on any of the outcome variables in the

U.S. (see Table 19). However, goal strain did have impacts on Taiwanese students‘

delinquent behavior, although the magnitude of the effect is not large when compared to

173
victimization and negative life-events. Whereas unjust strain had significant effects on

hitting someone and on aggression in the U.S. sample, the same strain increased students‘

involvement in damaging property and alcohol use in the Taiwanese sample. One reason

for this different cultural response pattern might be due to the collectivistic nature of

Chinese culture, which as mentioned earlier, emphasizes relational harmony. As such,

Taiwanese students might prefer less violent coping behavior (e.g., drinking alcohol) to

more violent coping strategies (e.g., hitting someone). One might argue that damaging

property is also a ―violent‖ behavior. However, damaging property was measured in a

sense more closely related to vandalism, where the owner of the property is not present.

Hence, Taiwanese students may be willing to engage in this form of destructive behavior.

A consistent finding from the basic model analysis was that all the strains had

positive and significant effects on depression and anger in both the U.S. and Taiwan (see

Table 18). However, only victimization and negative life-events have both statistically

and practically significant effects on delinquency and aggression. Consequently, although

the results may be consistent with previous studies in the U.S. (Broidy 2001; De Coster &

Kort-Butler, 2006; Olweus 1994; Vaux & Ruggiero, 1983) or in Asia (Moon et al., 2008;

Lin, in press; Lee & Larson, 2000), the practical meaning of the findings may be limited.

Second, to answer the question of whether anger and depression mediate the

strain-delinquency relationship, the study examined the full GST model, which

incorporated strain, anger, depression, and outcome variables simultaneously in a path

model, in the U.S. and Taiwan.

174
Table 19 Summary for the Basic GST model in the U.S. and Taiwan
Negative life- Victimization
Variable Goal strain Unjust strain
event

U.S. Taiwan U.S. Taiwan U.S. Taiwan U.S. Taiwan

Damaging
+NS +* +NS +* +** +** +** +**
property

Hitting +** +**

+NS +** +* +NS +* +*
someone

Alcohol use +NS +NS +NS +* +** +** +** +NS

Aggression +NS +** +* +NS -NS +** +** +**

Anger +** +** +** +** +** +** +** +**

Depression +** +** +** +** +** +* +** +NS

+ = positive effect. - = negative effect.

NS = non-significant. * = significant at .05 level. ** = significant at .01 level.

In addition, bootstrapping was employed to investigate properly the indirect effects.

Table 20, a summary for the full GST model, shows that, in the full model, goal strain

and unjust strain had only minimal effects on delinquency and aggression. In contrast,

negative life-events and victimization were criminogenic to youths in both countries.

That is, these two strains continually have statistically significant and large effects on

delinquency and aggression. Although the results seemed to duplicate what we found

with the basic GST model, the magnitude of relationships between strain, negative

emotions, and delinquency was changed; in some cases, a significant relationship

disappeared. According to the classic work of Baron and Kenny (1986), mediating effects

are at work.

175
 Table 21 provides a summary of the mediating effects. Looking across this table,

one finds that anger and depression significantly mediate all of the strain-delinquency

relationships, with one exception: depression did not significantly mediate the strain-

aggression relationship. Although most of these mediating effects were only partial and

moderate to small, in some instances full mediation was found48. For example, unjust

strain had a significant effect on hitting someone, but this relationship was reduced to

almost zero in the full model. Anger and depression totally mediated the unjust strain-

delinquency relationship. Moreover, consistent with GST‘s proposition that outer-

directed emotion (e.g., anger) is related to outer-directed behavior (e.g., hitting someone),

the total mediating effect comes mostly through anger. Besides the full mediating effect

found in the U.S. sample, some fully mediating effects were found in the Taiwanese

sample as well. For example, the goal strain-damaging property relationship was reduced

to non-significant levels when anger and depression are included in the model. Table 21

shows that the full mediating effect was mostly through depression for goal strain but

through anger for unjust strain.

Although the analysis of mediation seemed to support GST‘s proposition that

negative emotions mediate the strain-delinquency relationship, most of the mediating

effects were small; only in the fully mediated situation were strong mediating effects

found. Consequently, the result suggested that strain had detrimental impacts on youths

48
In some cases, the direct effects (e.g., goal strain → hitting someone) become negative or larger when
anger and depression are incorporated in the model. This phenomenon is called suppression or inconsistent
mediation (Davis, 1985). Little and colleagues (2007) grouped suppression under the rubric of ―partial
mediated relationship;‖ hence, this study also designates the suppression effect as partial mediation. For
more information on suppression, consult Mackinnon, Krull, and Lockwood (2000).

176
mostly because of strain itself. In addition, the relatively small path coefficients that were

found in this study seem to suggest that goal strain and unjust strain may have only

limited utility in understanding juvenile delinquency.

Finally, with regard to the last two research questions, which addressed the

similarities and differences between the U.S. and Taiwan, Table 22 provides a summary

of results obtained with the multiple group analysis. As can be seen, most of the tested

paths were similar in the two countries and significant in both. Hence, this study found

that the strain-delinquency/aggression relationships were similar in magnitude in the U.S.

and Taiwan. However, some differences were also evident. Victimization was

significantly related to aggression in both countries, but the magnitude of the relationship

differed between the two. The sharpest differences were found in the victimization-

alcohol use and the negative life-event-alcohol use relationship. The former was

significant only in the U.S. sample but the latter was significant only in the Taiwanese

sample. With regard to the strain-negative emotion relationship, strain-anger relationships

were different between the U.S. and Taiwan, with negative life-event → anger as the sole

exception. The result was that most students in the U.S. are more likely than Taiwanese

students to react to the four strains with anger. In contrast, strain-depression relationships

were similar in the two samples, with exception of unjust strain → depression. Strains

were significantly related to depressive feelings for students in both countries, but the

depressive feelings associated with unjust strain were stronger in the U.S. students than in

the Taiwanese students.

177
Table 20 Summary for the Full GST model in the U.S. and Taiwan
Variable Anger Depression Damage property Hit someone Alcohol use Aggression
U.S. Taiwan U.S. Taiwan U.S. Taiwan U.S. Taiwan U.S. Taiwan U.S. Taiwan
Goal strain +** +** +** +** -NS +NS -NS -NS -NS -NS +NS +**
Unjust strain +** +** +** +** -NS +NS +NS +NS -NS +NS +NS -NS
Negative life-event +* +* +* +** +* +** +NS +** +* +** -NS +*
Victimization +** +** +** +** +** +** +** +** +** +NS +** +**
Anger +** +** +** +** +** +** +** +**
Depression +** +** +** +* +** +** +NS +NS
+ = positive effect. - = negative effect.
NS = non-significant. * = significant at .05 level. ** = significant at .01 level.

Table 21 Summary for the Indirect Effect of Anger and Depression in the U.S. and Taiwan
Path Damage property Hit someone Alcohol use Aggression
U.S. Taiwan U.S. Taiwan U.S. Taiwan U.S. Taiwan
Goal strain → anger/depression *(p) (a) *(t) (d) *(p) (a) *(p) (a) *(p) (s) *(p) (d) *(p) (a) *(p) (a)
Unjust strain → anger/depression *(p) (a) *(t) (a) *(t) (a) *(p) (a) *(p) (s) *(t) (d) *(p) (a) *(p) (a)
Negative life-event →
*(p) (s) *(p) (a) *(p) (s) *(p) (a) *(p) (s) *(p) (d) *(p) (a) *(p) (a)
anger/depression
Victimization
*(p) (a) *(p) (a) *(p) (a) *(p) (a) *(p) (a) *(p) (a) *(p) (a) *(p) (a)
→anger/depression
*Indirect effect is significant at .05 level.
p = partial mediation. t = total mediation.
a = most of the mediating effect is from anger. d = most of mediating effect is from depression. s = effects are similar.

178
 The bottom part of Table22 gives the results of testing for similarities for the

indirect effects. The summary shows that the indirect effects of goal strain and unjust

strain through anger on all four outcome variables were different between the U.S. and

Taiwan, and the differences were mostly due to stronger indirect effects in the U.S. In

addition, differences were also found in the indirect effects of victimization on aggression

and damaging property through anger. Thus, all the differences with regard to indirect

effects of victimization were related to anger.

Discussion of the findings

Strain and its characteristics

Agnew (2006a) recently identified some strains that are most likely to be related

to delinquency. Victimization, which is on the list, was found in this study to be

criminogenic to youths in both the U.S. and Taiwan. In contrast, negative life-events,

although not on the list, was also found to be strongly related to youthful delinquency in

previous studies (Aseltine & Gore, 2000; Hoffmann & Cerbone, 1999; Hoffmann et al,

2000) as well as in the present study, a result that only partially supports the recent

revision of GST; however, the usefulness of GST in explaining juvenile delinquency and

its generality are confirmed. Although this strain, as well as victimization, and the

relationship of both of these two strains to negative emotions and delinquency, are

different in some respects, the negative impacts they have on youths in both countries are

unquestioned. These findings are especially important because Cohen (1994) and others

(Robinson & Lavin, 1997) argued that external replications, studies that investigate the

179
same research questions but with different subjects, is the only way to provide

generalizability and is invaluable for accumulating knowledge in a given domain.

Agnew (2006a) argued that strains are criminogenic if they are seen as ―high in

magnitude‖ and ―unjust,‖ are ―associated with low social control,‖ and create ―some

pressure or incentive for criminal coping‖ (Agnew, 2006a, pp. 58-68). After providing

these characteristics of strain, Agnew gave a list of strain that possessed these

characteristics, hence, criminogenic. Although giving the characteristics of strain (e.g.,

magnitude) and the list of criminogenic strain enhance GST on the theoretical ground,

close scrutiny may raise some challenges. One challenge is that the purpose of providing

the four characteristics was to counter the ―unfalsifiable‖ accusation (Jensen, 1995). The

listing of criminogenic strains may increase confusion rather than clarification. For

example, one would question whether direct measurement of the characteristics of strain

is necessary because the listed strains, according to Agnew, already include these

characteristics. As in this study, the characteristics of strain were not measured directly,

but the criminogenic strain—victimization—was included, which contains all the listed

characteristics. To stick with the list of criminogenic strain will once again lead

researchers back to the previous state; that is, GST is ―unfalsifiable‖ because one may

always find a strain that is related to delinquency but is not on the list. For example, the

list did not include negative life-events, but this strain has been found to be related to

delinquency in many previous studies as well as in the present study.

180
Table 22 Summary for the Tested Similarities and Differences in the Full GST Model
between the U.S. and Taiwan
Path (strain and negative emotions to delinquency and
Result
aggression )
Goal strain → damage property +NS
Unjust strain → damage property -NS
Negative life-event → damage property +**
Unjust strain → hit someone +NS
Negative life-event → hit someone +**
Victimization → hit someone -**
Anger → hit someone +**
Goal strain → alcohol use +NS
Unjust strain → alcohol use +NS
Victimization → alcohol use -**(U.S. only)
Anger → alcohol use +**
Depression → alcohol use +**
Goal strain → aggression +*
Unjust strain → aggression +NS
Negative life-event → aggression -*(Taiwan only)
Victimization → aggression -**
Path (strain to negative emotions)
Goal strain → anger -**
Unjust strain → anger -**
Negative life-event → anger +**
Victimization → anger -**
Goal strain → depression +**
Unjust strain → depression -**
Negative life-event → depression +**
Victimization → depression +**
Path (indirect effect)
Goal strain→anger→damage property -*
Unjust strain→anger→damage property -*
Victimization→anger→damage property -*
Unjust strain→depression→damage property -*
Goal strain→anger→hit someone -*
Unjust strain→anger→hit someone -*
Goal strain→anger→alcohol use -*
Unjust strain→anger→alcohol use -*
Victimization→anger→alcohol use -*
Goal strain→anger→aggression -*
Unjust strain→anger→aggression -*
Victimization→anger→aggression -*
+ = path is similar. - = path is different
* = the path is significant at .05 level. ** = the path is significant at .01 level.
181
 Another issue related to the above argument is that unjust strain, as measured in

the present study, has relatively little effect on delinquency and aggression. This raises

yet another challenge to GST because one of the characteristics of criminogenic strain is

unjust. Comparing the measurements used in this study and in that of Moon et al. (2008)

with Agnew‘s (2006a) argument on unjust strain, many similarities were found. For

example, Agnew suggested that a strain is more likely to be seen as unjust when ―victims

believe the strain they experienced is undeserved‖ or ―the strain strongly violates strongly

held social norms or values‖ (pp. 63-64, emphasis in origin). Moon and colleagues (2008)

took the former approach by asking students to rate whether they deserve the strain or not;

and the present study took the latter approach by presenting statements that violates

norms or rules (e.g., imbalance of input/gain). Both studies failed to find that unjust strain

was criminogenic, as GST would predict. With regard to the findings of Moon et al., the

explanation might be that victimization itself is so negative to students that it leads to

delinquency whether it is unjust or not. Hence, measuring the characteristics of such a

strain may be redundant49. With regard to the present finding, the explanation is that

many of the unjust strain statements are related to situations that are least likely to cause

crime (e.g., unpopular with peers, demands associated with conventional pursuit) (Agnew,

2006a, pp. 75-77). Incidents related to conventional pursuits that are in themselves

unlikely to cause crime may be less criminogenic to youths even if they are seen as unjust.

Together, these results may suggest that researchers should probably focus on directly

49
The redundant criticism is also found in Slocum et al. (2005) but with a somewhat different purpose;
theirs was to evaluate various dimensions of one of the four characteristics, magnitude.

182
measuring the characteristics of strain that Agnew gave, but at the same time we should

use the listed crimnogenic strains as a guide to guard against redundancy.

Notwithstanding Agnew‘s revision and above arguments, another possible

extension of GST is to consider the hierarchical order of strain. Maslow (1970) argued

long ago that human needs constitute a hierarchy; that is, one first satisfies the lower

ranking needs and then moves up. The basic needs, physical needs, must be fulfilled

before one shows concern about his or her safety needs and other higher ranking

needs/success (e.g., love and belonging, esteem). Victimization is criminogenic because

such an incident threatens one of the lower ranking needs, safety. One of the focal strains

in classic strain theory, strain related to monetary needs, which earlier research had found

to be criminogenic (Agnew, 1994; Agnew, Mathews, Bucher, Welcher & Keyes, 2008;

Baron, 2004), can be seen as a threat to the very bottom of the hierarchy of human needs,

physical needs (e.g., food). This may better account for many cross cultural similarities

because strains that threaten the lowest ranked needs would be universally stressful,

which in turn would lead to antisocial behavior because individuals want to satisfy such

need in an expedient way, which is usually criminal. Therefore, one of the characteristics

of strain might be its rank on the hierarchy of needs; the lower the rank the more

criminogenic it might be.

Negative emotions

The measure of anger in the present study is situational. Recent arguments from

the GST literature have suggested that a situational or state-like measure of negative

183
emotion is better than a trait-like measure (Agnew, 2006b; Capowich et al., 2001;

Mazerolle et al., 2003). Mazerolle et al. (2003, p.131) stated that ―the relationship

between anger and deviant outcomes is attenuated when trait-based measures of anger are

used.‖ The same may be true of other negative emotions as well. On this score, Agnew

(2006a, 2006b) suggested that the proper test of GST requires researchers to incorporate

emotional states. Consistent with these suggestions, anger, measured in the present study,

not only affected delinquency and aggression but also mediated the strain-

delinquency/aggression relationship. On the other hand, depression, in the present study,

as measured by the depressive symptom check-list, which is considered a trait-like

measure, also exerts strong effects on delinquency as well as mediating the strain-

delinquency relationship. However, in most cases, anger had a stronger mediating effect

than depression had. This may support the argument that a state-like measure of negative

emotions is more important and appropriate than a trait-like measure.

Although the results of this study seem to advocate acceptance of Agnew‘s and

others‘ arguments on the situational measure of negative emotions, several challenges

arise. First, while the above results seem to support the distinction between state-like and

trait-like measures, the differences between the effects of anger and depression on

delinquency and between their mediating effects are not large. Most of the state vs. trait

argument derives from one negative emotion, anger. Whether such an argument can be

extended to other negative emotions is an open question. At present, this may not apply

to depression, because this study did not find a large difference between the direct and

indirect effects that anger and depression have on delinquency. For example, depression

184
and anger both have statistically significant effects on all delinquent acts, except

aggression, which is only related to anger. In addition, the results show that anger and

depression exert similar mediating effects on strain-delinquency relationships. Previous

studies using similar measures of depression also found that depression is related to

delinquency (Beyers & Loeber, 2003; Lin, in press) and that it mitigates the strain-

delinquency relationship (Carson et al., 2009; Hoffmann & Su, 1998; Lin et al., 2011;

Walls et al., 2007). Hence, some negative emotions can perhaps be distinguished on the

basis of being situational or dispositional; other negative emotions may not be so clearly

differentiated on this issue, especially those negative emotions related to clinical

diagnosis.

A second, and related challenge arose from a study that was conducted by Ganem

(2008), who used several different scenarios to capture several negative emotional states

and used these measures to examine the role of negative emotions in GST. One of her

conclusions was that human emotions often occurred together; that is, even with a

properly delineated scenario, subjects reported different negative emotions other than the

sole emotion that the scenario was supposed to induce. The problem of co-occurrence

was found in previous studies (Sharp et al., 2005; Sharp et al., 2001; Sigfusdottir et al.,

2004) and in the present study, within which anger and depression were correlated.

Hence, there might be a more complex relationship between strain, negative emotion, and

delinquency than GST has offered (see Sharp et al., 2001).

Besides the co-occurrence of negative emotions, which may hinder the proper

measurement of negative emotions, research from other areas (e.g., psychology) has
185
found that mood is not static but dynamic during adolescent years. For example,

Schneiders and colleagues (2007) used the Experience Sampling Method to ―collect data

from participants at selected moments during their daily activities‖ (pp.703-704). They

found that mood changed across location and social context. As such, the above

measurement issue of negative emotion might not be simply a state vs. trait dichotomy.

A third, and more fundamental challenge, is explaining that the role of negative

emotions in GST. Many of the strain-delinquency relationships could be explained from

many different viewpoints (Agnew, 2001). For example, the victimization-delinquency

relationship can be explained by low self-control theory (Schreck, 1999; Schreck, Stewart,

& Fisher, 2006). Schreck (1999) has argued that individuals with low self-control are

more likely to commit crimes but also more likely to become victims because they are

highly likely to be in situations where they will be victimized. One way to distinguish

between equally valid theoretical explanations is the inclusion of negative emotion. This

inclusion clearly delineates the mechanism linking independent variable (strain) to the

outcome (delinquency). In contrast to most mainstream criminological theories (e.g.,

self-control, social learning), GST is the only theory that takes into account negative

emotions and the linkage of strain, negative emotions and delinquency. Hence, for

example, victimization leads to negative emotion, which in turn leads to delinquency; this

separates GST from other theories.

Although negative emotion provides clear mechanism linking strain to

delinquency, vagueness surrounds the role of negative emotion. From the earlier version

of GST (Agnew, 1992) to the recent revision (Agnew, 2001, 2006a), GST has never
186
clearly stated whether one should expect a full mediation or only a partial mediation from

negative emotions, although the recent revision seems to suggest a full mediation effect

(Agnew, 2006a, p.19, Figure 1.1). If this is the theoretical argument of GST, the present

study would reject GST‘s proposition, as would other studies (Aseltine, et al., 2000;

Brezina, 2001; Hay, 2003; Hay & Evans, 2006; Mazerolle & Maahs, 2000; Perez,

Jennings, & Gover, 2008). However, on some occasions, Agnew has argued that the same

coping strategies are used to deal with both strain and negative emotions. Hence, strain

remains a potent risk factor for delinquency. The accumulated evidence (Broidy, 2001;

Gibson et al., 2001; Jang & Johnson, 2003) and the results from this study seem to

support this theoretical argument.

The final challenge is the domain matching or specificity argument. Agnew

(2006a, 2006b) argued that researchers should explore how a specific strain may be

related to specific negative emotions, which in turn lead to specific forms of delinquency.

The first part of this argument refers to the strain-negative emotion relationship and the

second part refers to the negative emotion-delinquency relationship. Agnew (1992, p.60)

provided a clear delineation of the second part. Outer-directed negative emotions (e.g.,

anger) were most likely to be related to outer-directed delinquency (e.g., aggression), and

inner-directed negative emotions (e.g., depression) were more likely to be related to

inner-directed deviance (e.g., substance use). The present study provided only limited

support for this assertion. On the one hand, anger was found to be related to both outer-

(e.g., hitting someone) and inner-directed delinquency (e.g., alcohol use), and depression

was found to be a predictor of both types of delinquency. On the other hand, only anger

187
was related to aggression. However, as explained in the previous chapter, this mismatch

has been found by other researchers (Beyers & Loeber, 2003; Terrell et al., 2006). In

contrast, Jang and Johnson (2003) and Ganem (2008) directly tested this domain match

hypothesis and found support for it. For example, Jang and Johnson (2003) found that

strained individuals were more likely to feel anger, which had stronger effects on fighting

than on drug use, whereas depressive mood had the opposite effects on fighting and drug

use.

The results from this study and studies of others suggest that the domain match

argument might need to be further investigated or revised, so that researchers can

anticipate which negative emotions should be related to which type of delinquency.

Ganem‘s conclusion provided a direction for this endeavor. She stated that ―certain

crimes are positively predicted by certain negative emotions, some are negatively

influenced, and some are not influenced at all…‖ (p.74). In addition, Agnew (2006a)

proffered another way to deal with these mixed results. He argued that the objects in a

strain situation might engender different forms of delinquency. For example, if people are

the cause of one‘s anger, this anger may increase aggression; in contrast, if objects cause

one‘s anger, damaging property or stealing may be the ―ideal‖ way to cope with anger.

With regard to the first part of the specificity argument, studies from psychology

have found that the scripts most likely to induce anger include someone or something

interfering with one‘s plan, someone making a demand offensive to the recipient, an

individual feeling that others are trying to harm him or her in some way, or the perception

of the disadvantage of unfairness (e.g., procedural justice) (Lazarus, 1999; Shaver,

188
Schwartz, Krison, & O‘Connor, 1987; Bies & Tripp, 2001). The strain measured in the

present study contains some of these characteristics; hence, the significant relationship

found between strain and anger was expected. On the other hand, the scripts of ―sadness‖

or ―depressive mood‖ include the experience of an undesirable outcome, with the

perception that one is either unable to change or one is hopeless to correct the situation,

or the experience is perceived as creating irrevocable harm (Lazarus, 1999; Shaver et al.,

1987). Negative life-events and victimization fit this description and are related to

depression, whereas the other two strains may not include these characteristics.

Consequently, the present study seems to reject the specificity argument; in contrast to

this conclusion, Ganem‘s (2008) finding that ―certain emotions are more likely than

others to occur under certain types of strain‖ (p.73) seems to support this argument.

The above argument indicates lack of consistency with regard to the results of

studies of domain matching assertion. Clearly, future research is needed to clarify the

incongruent findings from the present study and previous studies as well as GST‘s

assertion. This may enhance the theoretical development of GST and its usefulness in

explaining crime and delinquency.

Similarities and differences in GST across cultures

The multiple group analysis revealed some differences between the samples from

two countries. One interesting finding is that the strain-anger relationship differs in most

cases between the two countries, in that the magnitude is always higher in the U.S.

sample than in the Taiwanese sample. This difference is consistent with Heine (2008,

189
p.352), who concluded that ―[l]ooking at emotional experience, there is more evidence

for cultural diversity.‖ Markus and Kitayama (1994) stated that in an individualistic

culture, negative emotion that is related to self (e.g., goal, self identity) is more likely to

be felt and expressed. Goal strain, unjust strain, and victimization threaten an individual

in various ways (e.g., blocking goals, harming oneself), which increases anger. In

addition, Chinese students often attribute their failures to themselves but their successes

to the group (Heine, 2008; Heine et al., 2001; Yang, 1986; Yu, 1996); this self-attribution

might make one attribute the strain experience to oneself, which in turn could lead to

lower anger, because anger is more likely when one has external attributions. One study

has documented that many Chinese immigrants have maintained the traditional Chinese

culture and parenting practices (Wu, 1996), so the cultural differences in the strain-anger

relationship may suggest that applying GST to these and other collectivistic cultures

needs to be done with caution.

Another explanation of the difference may be that the Chinese are socialized at an

early age to control affective display (Wu, 1996). As reviewed earlier, in a collectivistic

culture, great effort is made to maintain social harmony; hence, expressing negative

emotions, especially anger, may be prohibited because of its potential to damage

interpersonal relationships. Consequently, Taiwanese students might be just as likely to

experience anger as their counterparts in the U.S. but may prefer not to disclose it. The

reluctance to express negative emotions is even more so when these emotions incur great

social stigma, such as depression (Russell & Yik, 1996). Moreover, the unwillingness in

the Chinese culture to express negative emotions is also evident in the mediating effects

190
that anger and depression have on strain-delinquency relationships; in most cases, if not

all, the mediating effects are larger in the U.S. than in the Taiwanese sample.

Another interesting difference that emerged in the multiple group analysis is that

the victimization-alcohol use relationship is significant only in the U.S. As explained

previously, this might be due to the drinking norms that regulate alcohol use in Taiwan,

combined with a living environment that makes excessive use difficult. The difference

found for this particular case indicates that it may be useful to include macro social

factors (e.g., cultural norms) in the GST (Agnew, 2006a).

The negative life-events-aggression relationship was significant only in the

Taiwanese sample; this is interesting, because one would expect Chinese students to have

lower aggression levels than students in the U.S. An early study (Ho, 1984) showed that

Chinese parents instill impulse control in their children and are less tolerant of aggressive

behavior than U.S. parents. One possibility is that the control that Chinese parents

exercise over their children is intended to limit their aggression against out-group

individuals, but not in-group individuals, because of the fear of retaliation. The

aggression measured in the present study was restricted to physical aggression against

one‘s siblings, and aggression against one‘s own siblings might not be thought of as

serious. Moreover, a phenomenon in Chinese culture is that parents emphasize school

achievement during their children‘s adolescent years and tend to be lenient about other

behavior, which may lead to greater physical aggression against siblings in the Taiwan.

191
 Aside from the differences noted in this study, some similarities were also found.

Students who have experienced more negative life-events and victimization incidents are

more likely to hit someone and damage property, regardless of which country they live in.

These findings support the generality of GTS and confirm that both negative life-events

and victimization have negative impacts on youths. The recent surge of school violence

and bullying in Taiwan might be explained by studies based on GST.

Anger, despite the already discussed great cultural differences, has similar effects

across countries on alcohol use and on hitting someone. This may be contradictory to the

discussion of effects of cultural differences on negative emotion expression. One

explanation is that the Chinese, although indeed reluctant to express anger or depression

in general, do, however, express these emotions in the same ways as their counterparts in

other countries.

Similarities may become the case in situations in which the cultural differences

are gradually fading. Scholars have documented that as a country moves greater

development, individuals become more individualistic (Trandi, 1995; Hofsted, 2001). In

Taiwan, for example, as it advanced to becoming a developed country and was greatly

influenced by western culture, the traditional ethos has gradually faded (Smith & Hung,

2005). Lin and Mieczkowski (2011) argued that globalization and free trade have

accented the impact of Western culture on the traditional culture; thus, the collectivistic

and Confucian mentality has diminished notably. This, in turn, leads Taiwanese

adolescents to be more westernized and thus be vulnerable to a similar strain-delinquency

mechanism. However, the influence of traditional culture on students has not lost its grip
192
altogether. As such, students in Taiwan have become more westernized but still retain

traditional thought.

Some studies have found that individuals in Taiwan hold both individualistic and

collectivistic self or bi-cultural self images (Lu & Yang, 2006; Lu, 2008; Lu, Kao, Chang,

Wu, & Jin, 2008). Consequently, students in modern Taiwan not only consider similar

strains as stressful but also replicate the behavior of their counterparts in the U. S. with

regard to the ways in which they cope with strains and negative emotions. The bi-cultural

explanation is only tentative, future studies may want to measure both individualistic and

collectivistic self of Taiwanese students, and comparing the GST process between

students who are high on individualistic self (westernized) but low on collectivistic self,

students who are high on collectivistic self (traditional) but low on collectivistic self, and

students who are high on both. This may empirically evaluate the bi-culture self and

westernization arguments.

Another possible factor that contributes to the development of the bi-cultural self

is the low birth rate in Taiwan and the one-child policy in Mainland China. With low

birth rate, many couples may have only one child, which may make them spoil their one

and only child (Wu, 1996). As a result, the child may develop a high level of

individualism early, and thus his or her reactions to strain and negative emotion may

become similar to those of children in the U.S. Moreover, as students progress in their

schooling, they become more and more individualistic (Greenfield, 1997). Hence, the

bicultural phenomenon favors the application of western-developed theories in Taiwan

193
and other eastern developed countries. Finding similarities and cross-cultural supportive

results for the theories may become more common in the future.

Limitations of the present study and future study

Although this study is valuable and contributes to the literature on GST, several

limitations need to be addressed. First, the characteristics of strain (e.g., magnitude,

related to low social control) were not measured directly, and although some researchers

suggested that inclusion of the characteristics has limited utility (Botchkovar et al., 2009;

Lin & Mieczkowski, 2011), others have found that these characteristics provided the

prediction of delinquency (Moon et al., 2008; Slocum et al., 2005).

Second, the depression measure in this study is not ideal. Depression is measured

through four questions commonly used in depression symptom check-lists. Whether this

should be seen as a situational or dispositional measure is not clear. For example,

physicians often use symptoms to diagnose one‘s health condition, which may render the

symptoms as indictors of one‘s current state of health. On the other hand, these symptoms

usually last for some time, so that they can be qualified as an indicator of some illness.

Hence, symptoms might include characteristics of both state-like and trait-like measures

of negative emotions. A recent study argued that, because depression should be treated as

a clinical disorder from the health perspective, distinguishing between state- and trait-

depression is problematic (Manasse & Ganem, 2009). Hence, although the measure is not

ideal, it was still found to have strong effects on delinquency and in mediating the strain-

delinquency relationships.

194
 Third, this study measures only anger and depression; other negative emotions,

such as fear, anxiety or shame, need to be included. This is especially important given

that different cultures may have different focal negative emotions. For example, shame

may be the most commonly felt negative emotion for Taiwanese students (Fung & Chen,

2001), but anger may be the regular emotional reaction to strain in the U.S.

Fourth, another important component in the GST consists of the conditioning

factors. Agnew (2006a) has argued that whether individuals cope with strain in a

delinquent manner depends on these conditioning factors, but the literature contains

mixed results in this regard (Baron, 2004, 2007; Baron & Hartnagel, 2002; Mazerolle,

Burton, Cullen, Evans, & Payne, 2000; Eitle & Turner, 2002, 2003). In addition, it is

statistically difficult to find conditioning effects with the use of survey data (McClelland

& Judd, 1993) and to implement them in the SEM approach50.

Finally, the data are only cross-sectional; hence, this study cannot firmly establish

the causal relationship between variables. However, the path analysis used in this study is

meant to evaluate causal patterns; thus, the results can at least indicate where to look for

causal relationships among variables. Consistent with the causal relationship found in this

study, Agnew and White (1992), who used longitudinal data, also found similar results.

Moreover, the measurement of variables other than delinquency has no time limit,

whereas questions about delinquent behaviors asked subjects about their involvement in

these delinquent acts during the past one year only. This provides some control for the

50
In Mplus, a random coefficient has been used to model the conditioning effect when the interaction is
between latent variable. However, there is not much research on this issue so far.

195
temporal order problem. Some scholars have argued that the relationships between strain,

negative affect, and reactions are fairly simultaneous (Agnew 1992; Piquero & Sealock

2000, 2004). Notwithstanding the limitation, longitudinal data are still needed to

replicate the present path model in order to confirm the findings found in this study.

One related limitation is that the U.S. sample was collected more than a decade

ago. Hence, it might be unwise to compare those data to the recent data from Taiwan.

However, this may be a desired feature rather than a limitation. The U.S. was more

developed than Taiwan was a decade ago51 but the difference today is smaller.

Consequently, comparing these two datasets should not cause too great a problem.

Future research can build upon the present study in several ways. First,

characteristics of strain may need to be measured directly. This may help to clarify the

challenges raised by results of this study and other studies. Second, a recent study argued

that depression measured by means of a clinical symptom checklist should be regarded as

trait-depression, and that consequently, the relationship between strain and trait-like

depression is moderating rather than mediating (Manasse & Ganem, 2009). Future

studies may need to examine this argument when the measure of negative emotion is

trait-like. Such studies could greatly advance GST by clarifying the role of negative

emotions in GST. Also, future studies should consider the co-occurrence of negative

emotions and to model this complex strain-negative emotion-delinquency relationship

51
The comparison was based on annual personal income. In the U.S., the annual personal income was
about $28,000 in 1999 (U.S. Census Bureau, 2011), when the Largo survey was conducted, and was
$39,000 in 2009. The same number was $16,000 in Taiwan in 2010 (Directorate-General of Budget,
Accounting, and Statistics, Executive Yuan, R.O.C., 2011).

196
directly. Finally, direct comparison across cultures provides a valuable way to extend

criminological theories in general and GST in particular. Without such efforts, any

generality as well as possible differences in the theoretical mechanisms may not be

discovered.

In conclusion, although previous studies may have come to the same conclusion

that the present study provides, the lack of direct comparison prevents these studies from

giving firm conclusion. This research contributes to the current GST literature by directly

comparing and contrasting the GST model in the U.S. and Taiwan. The results show that

GST is useful in explaining juvenile delinquency in both the U.S., a more individualistic

culture, and Taiwan, a more collectivistic culture. In addition, this study found that

victimization and negative life-events are criminogenic to youths; hence, future studies

that do not include these two strains may risk of model misspecification. Anger and

depression are found to be detrimental to youths in both countries not only because they

affect adolescents‘ wellbeing but also because they lead to delinquent coping strategies.

Consequently, these two negative emotions should be incorporated into the GST model.

197
 REFERENCES

Abraham, T., & Russell, D. W. (2004). Missing data: A review of current methods and
applications in epidemiological research. Current Opinion in Psychiatry, 17, 315-
321.

Adèr, H. J., Mellenbergh, G. J., & Hand, D. J. (2008). Advising on research method: A
consultant’s companion. Huizen, Ntherlands: Johannes van Kessel.

Adler, F. (1996). Our American Society of Criminology, the world, and the state of the
art- The American Society of Criminology 1995 presidential address.
Criminology, 34, 1-9.

Agnew, R. (1984). Goal achievement and delinquency. Sociology and social research, 68,
435-451.

______ (1985a). A revised strain theory of delinquency. Social Forces, 64, 151-167.

______ (1985b). Neutralizing the impact of crime. Criminal Justice and Behavior, 12,
221-239.

______ (1987). On ―testing structural strain theories.‖ Journal of Research in Crime and
Delinquency, 24, 281-286.

______ (1989). A longitudinal test of the revised strain theory. Journal of Quantitative
Criminology, 5, 373-387.

______ (1991). Strain and subcultural crime theory. In J. Sheley (Ed.), Criminology: A
contemporary handbook. Belmont, CA: Wadsworth.

______ (1992). Foundation for a general strain theory of crime and delinquency.
Criminology, 30, 47-87.

______ (1993). Why do they do it? An examination of the intervening mechanisms

between social control variables and delinquency. Journal of Research in Crime
and Delinquency, 30, 245-266.

______ (1994). Delinquency and the desire for money. Justice Quarterly, 11, 411-427.
198
______ (1997). Stability and change in crime over the life course: A strain theory
explanation. In T. P. Thornberry (Vol. Ed.), Developmental theories of crime and
delinquency: Vol. 7. Advances in criminological theory (pp.101-132). New
Brunswick, NJ: Transaction.

______ (2001). Building on the foundation of general strain theory: Specifying the types
of strain most likely to lead to crime and delinquency. Journal of Research in
Crime and Delinquency, 38, 319-361.

______ (2002). Experienced, vicarious, and anticipated strain: An exploratory study on

physical victimization and delinquency. Justice Quarterly, 19, 603-632.

______ (2003). An integrated theory of adolescent peak in offending. Youth & Society,
34, 263-299.

_______ (2005). Juvenile delinquency: Causes and control (2nd ed.). Los Angeles, CA:
Roxbury.

______ (2006a). Pressured into crime: An overview of general strain theory. LA:
Roxbury Publishing Co.

______ (2006b). General strain theory: Current status and direction. In F. T. Cullen, J. P.
Wright, & K. R. Blevins (Eds.), Taking stock: The status of criminological theory:
Vol. 15. Advances in criminological theory (pp.101-123). New Brunswick:
Transaction.

Agnew, R., & Brezina, T. (1997). Relational problems with peers, gender, and
delinquency. Youth & Society, 29, 84-111.

Agnew, R., Cullen, F. T., Burton, V. S., Jr., Evans, T. D., & Dunaway, R. G. (1996). A
new test of classic strain theory. Justice Quarterly, 13, 681-704.

Agnew, R., Matthews, S. K., Bucher, J., Welcher, A. N., & Keyes, C. (2008).
Socioeconomic status, economic problems, and delinquency. Youth & Society, 40,
159-181.

Agnew, R., & White, H. R. (1992). An empirical test of general strain theory.
Criminology, 30, 475-499.

Akers, R. L. (2000). Criminological theories: Introduction, evaluation, and application

(3rd ed.). Los Angeles, CA: Roxbury.

199
Akers, R. L., & Cochran, J. K. (1985). Adolescent marijuana use: A test of three theories
of deviant behavior. Deviant Behavior, 6, 323-346.

Anderson, E. (1999). Code of the street. New York, NY: W.W. Norton & Company.
Aseltine, R. H., Jr., & Gore, S. L. (2000). The variable effects of stress on alcohol use
from adolescence to early adult. Substance & Misuse, 35, 643-668.

Aseltine, R. H., Jr., Gore, S. L., & Gordon, J. (2000). Life stress, anger, anxiety, and
delinquency: An empirical test of general strain theory. Journal of Health and
Social Behavior, 41, 256-275.

Allison, P. D. (2002). Missing data. Sage University Paper Series on Quantitative

Applications in the Social Sciences, 07-136. Thousand Oaks, CA: Sage.
______ (2003). Missing data techniques for structural equation modeling. Journal of
Abnormal Psychology, 112, 545-557.

Bao, Wang-Ning, Haas, A., & Pi, Y. (2007). Life strain, coping, and delinquency in the
People‘s Republic of China: An empirical test of general strain theory from a
matching perspective in social support. International Journal of Offender Therapy
and Comparative Criminology, 51, 9-24.

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in

social psychological research: Conceptual, strategic, and statistical considerations.
Journal of Personality and Social Psychology, 51, 1173-1182.

Baron, S. W. (2004). General strain, street youth and crime: A test of Agnew's revised
theory. Criminology, 42, 457-483.

______ (2006). Street youth, strain theory, and crime. Journal of Criminal Justice, 34,
209-223.

______ (2007). Stress youth, gender, financial strain, and crime: Exploring Broidy and
Agnew‘s extension to general strain theory. Deviant Behavior, 28, 273-302.

______ (2009). Street youths' violent responses to violent personal, vicarious, and
anticipated strain. Journal of Criminal Justice, 37, 442-451.

Baron, S. W., & Hartnagel, T. F. (1997). Attributions, affect, and crime: Street youths‘
reactions to unemployment. Criminology, 30, 519-533.

______ (2002). Street youth and labor market strain. Journal of Criminal Justice, 30,
519-533.

200
Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Manual for the Beck depression
inventory- II. San Antonio, TX: Psychological Corporation.

Bellah, R. N., Madsen, R., Sullivan, W. M., Swidler, A., & Tipton, S. M. (1996). Habits
of the heart: Individualism and commitment in American life. Los Angeles, CA:
University of California Press.

Bernard, T. J. (1984). Control criticisms of strain theories: An assessment of theoretical

and empirical adequacy. Journal of Research in Crime and Delinquency, 21, 353-
372.

______ (1987). Replay to Agnew. Journal of Research in Crime and Delinquency, 24,
287-290.

Beyers, J. M., & Loeber, R. (2003). Untangling developmental relations between

depressed mood and delinquency in male adolescents. Journal of Abnormal Child
Psychology, 31, 247-266.

Bies, R. J., & Tripp, T. M. (2001). A passion for justice: The rationality and morality of
revenge. In R. Cropanzano (Ed.), Justice in the workplace: From theory to
practice (pp.197-208). Mahwah, NJ: Lawrence Erlbaum.

Bond, M. H., & Hwang, K. K. (1986). The social psychology of Chinese people. In M. H.
Bond (Ed), The Psychology of the Chinese People (pp.107-170). New York, NY:
Oxford University Press.

Bond, M. H., Wan, K. C., Leung, K., & Giacalone, R. A. (1985). How are responses to
verbal insult related to cultural collectivism and power distance? Journal of
Cross-Cultural Psychology, 16, 111-127.

Bollen, K. A., & Stine, R. (1990). Direct and indirect effects: classical and bootstrap
estimates of variability. Sociological Methodology, 20, 115-140.

Botchkovar, E. V., Tittle, C. R., & Antonaccio, O. (2009). General strain theory:
Additional evidence using cross-cultural data. Criminology, 47, 131-176.

Brandt, D. (2006). Delinquency, development, and social policy: Current perspective in

psychology. New Haven, CT: Yale University Press.

Brezina, T. F. (1998). Adolescent maltreatment and delinquency: A question of

intervening processes. Journal of Research in Crime and Delinquency, 35, 71-99.

______ (1999). Teenage violence toward parents as an adaptation of family strain. Youth
& Society, 30, 416-444.
201
Broidly, L. M. (2001). A test of general strain theory. Criminology, 39, 9-35.

Broidly, L. M., & Agnew, R. (1997). Gender and crime: A general strain theory
perspective. Journal of Research in Crime and Delinquency, 34, 275-306.

Brown, T. A. (2006). Confirmatory Factor Analysis: For Applied Research. New York,
NY: The Guilford Press.

Brislin, R. W. (1986). The wording and translation of research instruments. In W. J.

Lonner, & J. W. Berry (eds.), Field methods in cross-cultural research (pp.137-
164). Beverly Hills, CA: Sage.

Burton, V. S., Jr. (1991). Explaining adult criminality: Testing strain, differential
association and control theories. Unpublished doctoral dissertation, University of
Cincinnati.

Burton, V. S., Jr., & Cullen, F. T. (1992). The empirical status of strain theory. Journal of
crime and Justice, 15, 1-30.

Campus Security Report Center. (2006). 教育部94學年度各級學校校園事件統計分析

報告。 [The statistical report of school incidents of school year 2005]. Available
at https://csrc.edu.tw/FileManage.mvc/FrontDetail/184

Cao, L. (2007). Return to normality: Anomie and crime in China. International Journal
of Offender Therapy and Comparative Criminology, 51, 40-51.

Capowich, G. E., Mazerolle, P., & Piquero, A. (2001). General strain theory, situational
anger, and social networks: An assessment of conditioning influences. Journal of
Criminal Justice, 29, 445-461.

Card, N. A., Stucky, B. D., Sawalani, G. M., & Little, .T. D. (2008). Direct and indirect
aggression during childhood and adolescence: A meta-analytic review of gender
differences, intercorrelations, and relations to maladjustment. Child Development,
79, 1185-1229.

Carson, D., Sullivan C., Cochran, J. K., & Lersh, K. (2009). General strain theory and the
relationship between early victimization and drug use. Deviant Behavior, 30, 54-
88.

Chai, D. H., & Yang, S. L. (1999). Gangs in schools: Problems and resolutions. Journal
of Student Guidance Bimonthly, 65, 8-17.

202
Chang, E. C. (2001). A look at the coping strategies and style of Asian Americans:
Similar and different? In C. R. Snyder (Ed.), Coping with stress: Effective people
and processes (pp.222-239). London, UK: Oxford University Press.

Chen, C. W., & Chen, J. Q. (2000). 青少年犯罪成因及預防策略。[Causes of juvenile

delinquency and its prevention]. Criminal Justice Policy & Crime Research, 3,
113-128.

Chen, J. C. (2000). 青少年生活壓力、家庭氣氛與偏差行為之關係研究。

[A study on the relationship between juvenile life-stress, family atmosphere and
delinquency]. Unpublished master thesis, National Changhua University of
Education.

Chou, C. P., & Bentler, P. M. (1995). Estimates and tests in structural equation modeling.
In R. H. Hoyle (Ed.), Structural equation modeling: Concepts, issues, and
applications (pp.37-55). Thousand Oaks, CA: Sage.

Chun, Chi-Ah, Moons, R. H., & Cronkite, R. C. (2005). Culture: A fundamental context
for the stress and coping paradigm. In P. T. P. Wang & L. C. J. Wong (Eds.),
Handbook of multicultural perspectives on stress and coping (pp.29-53). New
York, NY: Springer.

Chu, M. M. (2000). Taiwan. In G. Barak (Ed.), Crime and crime control: A global view
(pp.199-217). Westport, CT: Greenwood Press.
Chung, T., Langenbucher, J., Labouvie, E., Panadina, R. J., & Moos, R. H. (2001).
Changes in alcoholic patients‘ coping responses predict 12-month treatment
outcomes. Journal of Consulting and Clinical Psychology, 69, 92-100.

Child, D. (2006). The essential of factor analysis (3rd ed.). New York, NY: Continuum
International.

Clinard, M. B. (1964). Anomie and deviant behavior: A discussion and critique. New
York: The Free Press.

Cloward, R., & Ohlin, L. (1960). Delinquency and opportunity. Golencoe, IL: Free Press.
Cohen, A. K. (1955). Delinquent boys. Golencoe, IL: Free Press.

Cohen, J. (1994). The earth is round (p < .05). American Psychologist, 49, 997-1003.

______ (1965). The sociology of the deviant act: Anomie theory and beyond. American
Sociological Review, 30, 5-14.

Collins, L. M., Schafer, J. L., & Kam, C. M. (2001). A comparison of inclusive and
restrictive strategies in modern missing-data procedures. Psychological Methods,
203
 6, 330-351.

Colten, M. E., & Gore, S. (1991). Adolescent Stress: Causes and Consequences. New
York: Aldine De Gruyter.

Compas, B. E., & Wanger, B. M. (1991). Psychosocial stress during adolescence:

Intrapersonal and interpersonal processes. In M. E. Colten & S. Gore (Eds.),
Adolescent stress: Causes and Consequences (pp. 67-92). New York: Aldine De
Gruyter.

Copper, J. F. (2003). Taiwan: Nation-state or province? Cambridge, MA: Westview.

Crime Investigation Bureau. (2008). 犯罪預防。[Crime prevention]. Available at

http://www.cib.gov.tw/crime/Crime_Book.aspx
Cross, S. E. (1990). The role of the self-concept in cross-cultural adaptation.
Unpublished doctoral dissertation, University of Michigan.

Cullen, F. T. (1984). Rethinking crime and deviance theory: The emergence of a

structuring tradition. Totowa, NJ: Rowman & Allanheld.

Cullen, F. T., Wright, J. P., & Blevins, K. R. (2006). Taking Stock: The Status of
Criminological Theory. New Brunswick, NJ: Transaction Publishers.

Dai, Y. (2005). The development of E-Scooter in Taiwan. E-Scooter Industry Periodical,

7, 2-4.

Davis, J. A. (1985). The logic of causal order. Sage University Paper Series on
Quantitative Applications in the Social Sciences, 07-055. Thousand Oaks, CA:
Sage.

De Coster, S., & Kort-Butler, L. (2006). How general is general strain theory? Assessing
determinacy and indeterminacy across life domains. Journal of Research in Crime
and Delinquency, 43, 297-325.

Department of Budget, Accounting and Statistic, Kaohsiung City Government. (2009). 高

雄市統計年報。 [Kaohsiung city annual statistic report]. Available at
http://kcgdg.kcg.gov.tw/book/yearbook/index.htm

Department of Statistics, Ministry of Education. (2008). 教育部教育統計。[Education

statistics] Available at
http://140.111.34.54./statistics/content.aspx?site_content_sn=8869

______ (2009). 主要統計表。[Major statistics] Available at

http://www.edu.tw/statistics/content.aspx?site_content_sn=8869
204
Directorate-General of Budget, Accounting, and Statistics, Executive Yuan, R.O.C.
(2011). 人口靜態統計。[Statistics of Population]. Available at
http://www.dgbas.gov.tw/ct.asp?xItem=15408&CtNode=4594
______ (2011). 國民所得統計。[Statistics of Personal Income] Available at
http://www.stat.gov.tw/ct.asp?xItem=28862&ctNode=3565

Drapela, L. A. (2006). The effect of negative emotion on licit and illicit drug use among
high school dropouts: An empirical test of general strain theory. Journal of Youth
and Adolescence, 35, 755-770.

DuRant, R H., Getts, A., Cadenhead, C., Emans, S. J., & Woods, E. R. (1995). Exposure
to violence and victimization and depression, hopelessness, and purpose in life
among adolescents living in and around public housing. Developmental and
Behavioral Pediatrics, 16, 233-237.

Durkheim, E. (1947). Division of labor in society (G. Simpson, Trans.) Glencoe, IL: Free
Press. (Original work published 1893).

______ (2006). On suicide (R. Buss, Trans.). New York: Penguin Books. (Original work
published 1897).

Education Bureau, Kaohsiung City Government. (2010). 教育現況統計。[Current

Educational Statistics]. Available at
http://163.32.250.1:8000/Members/grp07/new_page_3.htm

Eftekhari, A., Turner, A. P., & Larimer, M. E. (2004). Anger expression, coping, and
substance use in adolescent offenders. Addictive Behaviors, 29, 1001-1008.

Eitle, D. J. (2002). Exploring a source of deviance-producing strain for females:

Perceived discrimination and general strain theory. Journal of Criminal Justice,
30, 95-111.

Eitle, D. J., & Turner, R. J. (2002). Exposure to community violence and young adult
crime: The effects of witnessing violence, traumatic victimization, and other
stressful life events. Journal of research in Crime and Delinquency, 39, 214-237.

______ (2003). Stress exposure, race, and young adult crime. Sociological Quarterly, 44,
243-269.
Ekman, P. (1999). Basic emotions. In T. Dalgleish & M. J. Power (Eds.), Handbook of
cogniyion and emotion (pp.45-60). UK: John Wiley & Sons.

205
Elliott, D. S. (1994). Serious violent offenders: Onset, developmental course, and
termination- The American Society of Criminology 1993 Presidential Address.
Criminology, 32, 1-21.

Elliott, D. S., Ageton, S. S., & Canter, R. J. (1979). An integrated theoretical perspective
on delinquent behavior. Journal of Research in Crime and Delinquency, 16, 3-27.
Elliott, D. S., Huizinga, D., & Ageton, S. S. (1985). Explaining delinquency and drug use.
Thousand Oaks, CA: Sage.

Elliott, D. S., & Voss, H. L. (1974). Delinquent and dropout. Lexington, MA: D. C.
Health.

Ellwanger, S. J. (2007). Strain, attribution, and traffic delinquency among young drivers.
Crime & delinquency, 53, 523-551.

Eve, R. A. (1978). A study of efficacy and interactions of several theories for explaining
rebelliousness among high school students. Journal of Criminal Law and
Criminology, 69, 115-125.

Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. J. (1999). Evaluating
the use of exploratory factor analysis in psychological research. Psychological
Methods, 4, 272-299.

Farnworth, M., & Leiber, M. J. (1989). Strain theory revisited: Economic goals,
educational means and delinquency. American Sociological Review, 54, 263-274.

Farrington, D. P. (1989). Early predictors of adolescent aggression and adult violence.

Violence and Victims, 22, 79-100.

Florida Department of Law Enforcement. (1999). Crimes and crime rates

statistics [Data file]. Available at http://fl.rand.org/cgi-bin/annual.cgi
Flora, D. B., & Curran, P. J. (2004). An empirical evaluation of alternative methods of
estimation for confirmatory factor analysis with ordinal data. Psychological
Methods, 9, 466-491.

Ford, J. A., & Schroeder, R. D. (2009). Academic strain and non-medical use of
prescription stimulants among college students. Deviant Behavior, 30, 26-53.

Fritz, M. S., & MacKinnon, D. P. (2007). Required sample size to detect the mediated
effect. Psychological Science, 18, 233-239.

Froggio, G. (2007). Strain and juvenile delinquency: A critical review of Agnew‘s

general strain theory. Journal of Loss and Trauma, 12, 383-418.

206
Froggio, G., & Agnew, R. (2007). The relationship between crime and "objective" versus
"subjective" strains. Journal of Criminal Justice, 35, 81-87.

Fung, H., & Chen, E. C. H. (2001). Across time and beyond skin: Self and transgression
in the everyday socialization of shame among Taiwanese preschool children.
Social Development, 10, 420-437.

Gabrenya, W. K., & Wang, Y. (1983). Cultural differences in self schemata. Paper
presented at the Southeast Psychological Association Convention, Atlanta, 1983.

Gallois C., Giles, H., Ota, H., Pierson, H. D., Ng, S. H., Lim, T. S., et al. (1996).
Intergenerational communication across the Pacific Rim: The impact of filial
piety. In H. Grad, A. Blanco, & J. Georgas (Eds.), Key issues in cross-cultural
psychology (pp.192-211). Lisse, Netherlands: Swets & Zeitlinger.

Gates, H. (1987). Chinese working-class lives: Getting by in Taiwan. Ithaca, NY: Cornell
University Press.

Genem, N. M. (2008). The role of negative emotion in general strain theory: An

empirical study. La Vergne, TN: Lightning Source Inc.

Gibson, C. L., Swatt, M., & Jolicoeur, J. R. (2001). Assessing the generality of general
strain theory: The relationship among occupational stress experienced by male
police officers and domestic forms of violence. Journal of Crime and Justice, 24,
29-57.

Gorsuch, R. L. (1983). Factor analysis (2nd ed.). Hillsdale, NJ: Erlbaum.

Gottfredson, M. R., & Hirschi, T. (1990). A General Theory of Crime. Stanford,

California: Stanford University Press.

Greenfield, P. M. (1997). You can‘t take with you: Why ability assessments don‘t cross
cultures. American Psychologist, 52, 1115-1124.

Harnish, J. D., Aseltine, R. H., Jr., & Gore, S. (2000). Resolution of stressful experience
as an indicator of coping effectiveness in young adult: An event history analysis.
Journal of Health and Social Behavior, 41, 121-136.

Harrell, E. (2007). Adolescent victimization and delinquency behavior. New York, NY:
LFB.

Harrell, S. (1981). Normal and deviant drinking in rural Taiwan. In A. Kleinman & L.
Tsung-Yi (Eds.), Normal and abnormal behavior in Chinese culture (pp.49-59).
Boston, MA: D. Reidel.
207
Hay, C. (2003). Family strain, gender, and delinquency. Sociological Perspectives, 46,
107-135.

Hay, C., & Evans, M. M. (2006). Violent victimization and involvement in delinquency:
Examining predictions from general strain theory.‖ Journal of Criminal Justice,
34, 261-274.

Hatcher, L. (1994). A step-by-step approach to using SAS system for factor analysis and
structural equation modeling. Cary, NC: SAS Institute.

Hayes, A. F. (2009). Beyond Baron and Kenny: Statistical mediation analysis in the new
millennium. Communication Monographs, 76, 408-420.

Hauser, S. T., & Bowlds, M. K. (1990). Stress, coping, and adaption. In S. S. Feldman &
G. R. Elliott (Eds), At the threshold: The developing adolescent (pp.388-413).
Cambridge, MA: Harvard University Press.

Hegtvedt, K. A. (1990). The effects of relationship structure on emotional response to

inequity. Social Psychology Quarterly, 53, 214-228.

Heine, S. J. (2008). Cultural psychology. New York, NY: W. W. Norton.

Heine, S. J., Kitayama, S., Lehman, D. R., Takata, T., Ide, E., Leung, C., & Matsumoto,
H. (2001). Divergent consequences of success and failure in Japan and North
America: An investigation of self-improving motivations and malleable selves.
Journal of Personality and Social Psychology, 81, 599-615.

Heine, S. J., & Lehman, D. R. (1995). Social desirability among Canadian and Japanese
students. Journal of Social Psychology, 135, 777-779.

Herbert, T. B., & Cohen, S. (1996). Measurement issues in research on psychological

stress. In H. B. Kaplan (Ed.), Psychological stress: perspectives on structure,
theory, life-course, and methods (pp.295-332). New York: Academic Press.

Hirschi, T. (1969). Causes of delinquency. Los Angeles, CA: University of California

Press.

Ho, H. Z., Chen, W. W., & Kung, H. Y. (2008). Taiwan. In I. Epstein & J. Pattnaik (Eds.),
Asia and Oceania: Vol. 1. The Greenwood encyclopedia of children’s issues
worldwide (pp.439-464). Westport, CT: Greenwood Press.

Ho, D. Y. F. (1986). Chinese patterns of socialization: A critical review. In M. H. Bond

(Ed), The Psychology of the Chinese People (pp.1-37). New York, NY:
208
 Oxford University Press.

_______ (1996). Cognitive socialization in Confucian heritage cultures. In P. M.

Greenfield & R. R. Cocking (Eds.), Cross-cultural roots of minority child
development (pp.285-313). Hillsdale, NJ: Lawrence Erlbaum.

_______ (1996). Filial piety and its psychological consequences. In M. H. Bond (Ed.),
The handbook of Chinese psychology (pp. 155-165). New York, NY:
Oxford University Press.

Ho, D. Y. F., & Chiu, C-Y. (1994). Component ideas of individualism, collectivism, and
social organization: An application in the study of Chinese culture. In U. Kim, H.
C. Triandis, C. Kagitcibasi, S-C. Choi, & G. Yoon (Eds.), Individualism and
collectivism: Theory, method, and applications (pp.137-156). Thousand Oaks, CA:
Sage.

Hoffmann, J. P. (2002). A contextual analysis of differential association, social control

and strain theories of delinquency. Social Forces, 81, 753-785.

Hoffmann, J. P., & Cerbone, F. G. (1999). Stressful life events and delinquency
escalation in early adolescence. Criminology, 37, 343-373.

Hoffmann, J. P., Cerbone, F.G., & Su, S. S. (2000). A growth curve analysis of stress and
adolescent drug use. Substance Use & Misuse, 35, 687-716.

Hoffmann, J. P., & Miller, A. S. (1998). A latent variable analysis of general strain theory.
Journal of Quantitative Criminology, 14, 83-110.

Hoffmann, J. P., & Su, S. S. (1997). The conditional effects of stress on delinquency and
drug use: A strain theory assessment of sex differences. Journal of Research in
Crime and Delinquency, 34, 46-78.

_______ (1998). Stressful life events and adolescent substance sue and depression:
Conditional and gender differentiated effects. Substance Use & Misuse, 33, 2219-
2262.

Hofstede, G. (2001). Culture’s Consequences (2nd ed.).Thousand Oaks: Sage Publications.

Hofstede, G., & Bond, M. H. (2001). The Confucius Connection: From cultural roots to
economic growth. In M. J. Gannon (Ed.), Cultural metaphors: Readings research
translations, and commentary (pp.31-50). Thousand Oaks, CA: Sage.

209
Hollist, D. R., Hughes, L. A., & Schaible, L. M. (2009). Adolescent maltreatment,
negative emotion, and delinquency: An assessment of general strain theory and
family-based strain. Journal of Criminal Justice, 37, 379-387.

Hou, C. (2003). 青少年犯罪問題與政策現況。[Current state of juvenile delinquency

and policy]. Criminal Justice Policy & Crime Research, 6, 131-148.

Hsu, F. L. K. (1983). Rugged individualism reconsidered: Essay in psychological

anthropology. Knoxville, TN: University of Tennessee Press.

Huang, D. H. (2003). Characteristics and problems for the development of adolescents in

Taiwan. In F. Pajares & T. Urdan (Eds.), International Perspectives on
Adolescence (pp.303-318). Greenwich: Information Age Publishing.

Hui, C. H., & Villareal, M. J. (1989). Individualism –collectivism and psychological

needs: Their relationships in two cultures. Journal of Cross-Cultural Psychology,
20, 310-323.

Hussong, A. M., & Chassin, L. (1994). The stress-negative affect model of adolescent
alcohol use: Disaggregating negative affect. Journal of Studies on Alcohol, 55,
707-718.

Hussong, A. M., Hicks, R. E., Levy, S. A., & Curran, P. J. (2001). Specifying the
relationship between affect and heavy alcohol use among young adults. Journal of
Abnormal Psychology, 110, 449-461.

Introduction of Juvenile Guidance Sections. (1991). Taipei, Taiwan (in Mandarin).

James, L. R., & Brett, J. M. (1984). Mediators, moderators, and tests for mediation.
Journal of Applied Psychology, 69, 307-321.

Jang, S. J. (2007). Gender differences in strain, negative emotions, and coping behaviors:
A general strain theory approach. Justice Quarterly, 24, 523-553.

Jang, S. J., & Lyons, J. A. (2006). Strain, social support, and retreatism among African
Americans. Journal of Black Studies, 37, 251-274.

Jang, S. J., & Johnson, B. R. (2003). Strain, negative emotions, and deviant coping
among African Americans: A test of general strain theory. Journal of Quantitative
Criminology, 19, 79-105.

______ (2005). Gender, religiosity, and reactions to strain among African Americans.
Sociological Quarterly, 46, 323-357.

210
Jensen, G. F. (1995). Salvaging structure through strain: A theoretical and empirical
critique. In F. Adler & W. S. Laufer (Eds.), The legacy of anomie theory: Vol. 6.
Advances in criminological theory (pp.139-158). New Brunswick, NJ:
Transaction.

Johnson, M. C., & Morris, R. G. (2008). The moderating effects of religiosity on the
relationship between stressful life events and delinquent behavior, Journal of
Criminal Justice, 36, 486-493.

Jones, R. A. (1986). Emile Durkheim: An introduction to four major works. Beverly Hill,
CA: Sage.

Juvenile Law. (1981). Taipei, Taiwan (in Mandarin).

Kaplan, H. B. (1983). Psychosocial stress: Trends in theory and research. New York,
NY: Academic Press.

Karmen, A. (2001). Crime victims: An introduction to victimology (4th ed.). Los Angeles,
CA: Wadsworth.

Karstedt, S. (2001). Comparing cultures, comparing crime: Challenges, prospects, and

problems for a global criminology. Crime. Law, and Social Change, 36, 285-308.

Kaufman, J. M. (2009). Gendered responses to serious strain: The argument for a general
strain theory of deviance. Justice Quarterly, 26, 410-444.

Kohn, M. L. (1987). Cross-national research as an analytic strategy: American

Sociological Association, 1987 Presidential Address. American Sociological
Review, 52, 713-731.

Kim, U., & Choi, S. H. (1994). Individualism, Collectivism, and child development: A
Korean perspective. In P. M. Greenfield & R. R. Cocking (Eds.), Cross-cultural
roots of minority child development (pp.227-257). Hillsdale, NJ: Lawrence
Erlbaum.

Kim, U., Triandis, H. C., Kagitcibasi, C., Choi, Sang-Chin, & Yoon, G. (1994).
Introduction. In U. Kim, H. C. Triandis, C. Kagitcibasi, Sang-Chin, Choi, & G.
Yoon (Eds.), Individualism and Collectivism; Theory, Method, and Applications
(pp.1-11). Thousand Oaks: Sage Publications.

Kennedy, L. W., & Baron. S. W. (1993). Routine activities and a subculture of violence:
A study of violence on the street. Journal of Research in Crime and Delinquency,
30, 88-112.

211
Kilpatrick, D. G., Saunders, B. E., Veronen, L. J., Best, C. L., & Von, J. M. (1987).
Criminal victimization: Lifetime prevalence, reporting to police, and
psychological impact. Crime & Delinquency, 33, 479-498.

King, A. Y. C. (1981). Individual and group in Confucianism: A relational perspective.

Paper presented at the Conference on Individualism and Wholism, York, Maine.

King, A. Y. C., & Bond, M. H. (1985). The Confucian paradigm of man: A sociological
view. In W. S. Tseng & D. Y. H. Wu (Eds.), Chinese culture and mental health
(pp.29-45). Orlando, FL: Academic Press.

Kirk, R. E. (1996). Practical significance: A concept whose time has come. Educational
and Psychological Measurement, 56, 746-759.

Kitsuse, J. I., & Dietrick, D. C. (1959). Delinquent boys: A critique. American

Sociological Review, 24, 208-215.

Kleinman, A. (1986). Social origins in distress and disease. New Haven, NY: Yale
University Press.

Kline, R. B. (2005). Principles and practices of structural equation modeling. New York,
NY: Guilford Press.

Kornhauser, R. R. (1978). Social sources of delinquency: An appraisal of analytic models.

Chicago, IL: University of Chicago Press.

Kroeber, A. L., & Kluckhohn, C. (1952). Culture: A critical review of concepts and
definitions (Papers of the Peabody Museum of Archaeology and Ethnology).
Cambridge, MA: The Museum.

Kroeber, A. L., & Parsons, T. (1958). The concepts of culture and of social system.
American Sociological Review, 23, 582-583.

Kuo, J. H., & Wu, H. L. (2003). The placement guidance for delinquent children and
youth. Criminal Justice Policy & Crime Research, 6, 149-168.

Kuo, S. Y., Cuvelier, S. J., & Chang, K. M. (2009). Explaining criminal victimization in
Taiwan: A lifestyle approach. Journal of Criminal Justice, 37, 461-467.

Lam, C. M. (2007). Not growing up forever: A Chinese conception of adolescent

development. New York, NY: Nova Science.

Largo Chamber of Commerce. (1998). Largo demographics [Data file]. Available

at http://www.largochamber.com
212
Larson, R., & Asmussen, L. (1991). Anger, worry, and hurt in early adolescence: An
enlarging world of negative emotions. In M. E. Colton & S. Gore (Eds.),
Adolescent Stress: Causes and Consequences (pp. 21-42). New York, NY: Aldine
de Gruyter.

Laungani, P. (2004). Asian perspective in counseling and psychotherapy. New York, NY:
Brunner-Rutledge.

Lauritsen, J. L., Sampson, R. J., & Laub, J. H. (1991). The Link between Offending and
Victimization among Adolescents. Criminology, 29, 265-292.

Lazarus, R. S. (1991). Emotion and adaptation. New York, NY: Oxford University Press.

_______ (1999). Stress and emotion: A new synthesis. New York, NY: Springer.

Lee, J. M., Huang, H. L., & Miao, D. F. (2000). 青少年物質使用之社會學習及社會連

結因素研究---已在學生為例。[Applying social learning and social bonding on
juvenile substance use by using a student sample]. 健康促進暨衛生教育雜誌, 20,
17-34.

Lee, M., & Larson, R. (2000). The Korean ‗examination hell‘: Long hours of studying,
distress, and depression. Journal of Youth and Adolescence, 29, 249-271.

Lee, S. X., Lin, G. N., Yang, H. R., Fu, L. A., Liu, X. W., & Lee, S. Q. (2009). A
qualitative study of knowledge, attitude, practice, and coping regarding drugs
amongst adolescent drug offenders. Journal of Research in Delinquency and
Prevention, 1, 1-28.

Lee, Y. N. (1998). 校園暴力問題防治方法之探討。[Discussion on the problem of

campus violence and its prevention]. Criminal Justice Policy & Crime Research,
1, 217-236.

Leung, K., Au, Y. F., Fernandez-Dols, J. M., & Iwawaki, S. (1992). Preference for
methods of conflict processing in two collectivist cultures. International Journal
of Psychology, 27, 195-209.

Leung, K., & Bond, M. H. (1982). How Chinese and Americans reward task-related
contributions: A preliminary study. Psychologia, 25, 32-39.
Leung, K., & Li, W. K. (1990). Psychological mechanisms of process control effects.
Journal of Applied psychology, 75, 613-620.

213
Leung, K. S., Li, J. H., Tsay, W. I., Callahan, C., Liu, S. F., Hsu, J., Hoffer, L., & Cottler,
L. B. (2008). Donosaur girls, candy girls, and trinity: Voices of Taiwanese club
drug users. Journal of Ethnicity in Substance Abuse, 7, 237-257.

Levin, J. R. (1993). Statistical significance testing from three perspectives. Journal of

Experimental Education, 61, 378-382.

Li, X. Y., & Chiang, I C. (2001). The study of the subject life stress, social support,
coping behaviors and psycho-physical health situation of the junior high school
students in Taoyuan. 衛生教育學報, 15, 115-132.

Lin, C., & Liu, W. T. (1999). Intergenerational relationships among Chinese immigrant
families from Taiwan. In H. P. McAdoo (Ed.), Family ethnicity: Strength in
diversity (pp.271-286). Newbury Park, CA: Sage.

Lin, W. H. (in press). General strain theory in Taiwan: A latent growth curve modeling
approach. Asia Journal of Criminology.

Lin, W. H., Cochran, J. K., & Mieczkowski, T. (2011). Direct and Vicarious violent
victimization and juvenile delinquency: An application of general strain theory.
Sociological Inquiry.

Lin, W. H., & Mieczkowski, T. (2011). Subjective strains, conditioning factors, and
juvenile delinquency: General strain theory in Taiwan. Asian Journal of
Criminology, 6, 69-87.

Link, T. C. (2008) Adolescent substance use in Germany and the United States: A cross-
cultural test if the applicability and generalizability of theoretical indicators.
European Journal of Criminology, 5, 453-480.

Liska, A. E. (1971). Aspirations, expectations, and delinquency: Stress, and additive

models. Sociological Quarterly, 12, 99-107.
Little, T. D, Card, N. A., & Bovaird, J. A. (2007). Structural equation modeling of
mediation and moderation with contextual factors. In T. D. Little, J. A. Bovaird,
& N. A. Card (Eds.), Modeling Contextual Effects in Longitudinal Studies (pp.
207-230). Mahwah, NJ: Lawrence Erlbaum.

Lonner, W. J. (1994). Culture and human diversity. In E. J. Trickett, R. J. Watts, & D.

Brirman (Eds.), Human diversity: Perspectives on people in context (pp.230-243).
San Francisco, CA: Jossey-Bass.

Lu, L. (2008). The individual-oriented and Social-oriented Chinese bicultural self:

Testing the theory. The Journal of Social Psychology, 148, 347-373.

214
Lu, L., Kao, S. F., Chang, T-T, Wu, H-P, & Jin, Z. (2008). The individual- and social-
oriented Chinese bicultural self: A subcultural analysis contrasting mainland
Chinese and Taiwanese. Social Behavior and Personality, 36, 337-346.

Lu, L., & Yang, K. S. (2006). Emergence and composition of the traditional-modern
bicultural self of people in contemporary Taiwanese societies. Asian Journal of
Social Psychology, 9, 167-175.

Ma, A. L. (2000). The research into the cognition, obeying willingness and behavior of
adolescents for Tobacco Hazards Act in Taiwan. Doctoral dissertation, National
Taiwan Normal University.

MacKinnon, D. P., Krull, J. L., & Lockwood, C. M. (2000). Equivalence of mediation,

confounding and suppression effect. Prevention Science, 1, 173-181.

MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002).
A comparison of methods to test mediation and other intervening variable effects.
Psychological Methods, 71, 83-104.

MacKinnon, D. P., Lockwood, C. M., & Williams, J. (2004). Confidence limits for the
indirect effect: Distribution of the product and resampling methods. Multivariate
Behavioral Research, 39, 99-128.
Manasse, M. E., & Ganem, N. (2009). Victimization as a cause of delinquency: The role
of depression and gender. Journal of Criminal Justice, 37, 371-378.

Markus, H. R., & Kitayama, S. (1991). Culture and the self: Implications for cognition,
emotion, and motivation. Psychological Review, 20, 568-579.

_______ (1994). The cultural construction of self and emotion: Implications for social
behavior. In S. Kitayama, & H. R. Markus (eds.), Emotion and cultural:
Empirical studies of mutual influence (pp.89-130). Washington, DC: American
Psychology Association.

Maruyama, G. M. (1997). Basics of Structural equation modeling. Thousand Oaks, CA:

Sage.

Maslow, A. H. (1970). Motivation and personality. New York: Harper & Row.

Maxwell, S. (2001). A focus on familial strain: Antisocial behavior and delinquency in

Filipino society. Sociological Inquiry, 71, 265-292.

Mazerolle, P. (1998). Gender, general strain, and delinquency: An empirical examination.

Justice Quarterly, 15, 65-91.

215
Mazerolle, P., Burton, V. S., Cullen, F. T., Evans, D. T., & Payne, G. L. (2000). Strain,
anger, and delinquent adoptions: Specifying general strain theory. Journal of
Criminal Justice, 17, 753-778.

Mazerolle, P., & Maahs, J. (2000). General strain and delinquency: An alternative
examination of conditioning influences. Justice Quarterly, 17, 753-778.

Mazerolle, P., & Piquero, A. (1997). Violent responses to strain: An examination of

conditioning influences. Violence and Victim, 12, 323-343.

_______ (1998). Linking exposure to strain with anger: An investigation of deviant

adaptations. Journal of Criminal Justice, 26, 195-211.
Mazerolle, P., Piquero, A., & Capowich, G. E. (2003). Examining the links between
strain, situational and dispositional anger, and crime. Youth & Society, 35, 131-
157.

McClelland, G. H., & Judd, C. M. (1993). Statistical difficulties of detecting interactions

and moderator effects. Psychological Bulletin, 114, 376-390.

Mcknight, P. E., Mcknight, K. M., Sidani, S., & Figueredo, A. J. (2007). Missing data: A
gentle introduction. New York, NY: Guiford.

Merton, R. K. (1938). Social structure and anomie. American Sociological Review, 54,
597-611.

_______ (1959). Social structure and anomie: Revisions and extensions. In R. N. Anshen
(Ed.), The family: Its function and destiny: Vol. V. Science of Culture Series
(pp.226-257). New York: Harper & Brothers.

_______ (1964). Anomie, anomia, and social interaction: Contexts of deviant behavior.
In M. B. Clinard (Ed.), Anomie and deviant behavior (pp.213-242). New York,
NY: The Free Press.
_______ (1968). Social structure and anomie (Enlarged ed.). New York: The Free Press.
Mesquita, B., & Frijda, N. H. (1992). Cultural variations in emotions: A review.
Psychological Bulletin, 112, 179-204.

Messner, S. F. (1988). Merton‘s ―social structure and anomie‖: The road not taken.
Deviant Behavior, 9, 33-53.

Ministry of Justice. (2007). 法務部統計指標 。[Statistical index] Available at

http://www. moj.gov.tw/ct.asp?xItem=126021&CtNode=2345&map=001

216
_______ (2008). 97年少年兒童犯罪概況及其分析。[The current state of juvenile and
child delinquency and its analysis]. Available at
http://www.criminalresearch.moj.gov.tw/public/Attachment/910719591765.pdf

Mirowsky, J., & Ross, C. E. (1995). Sex difference in distress: Real or artifact? American
Sociological Review, 60, 449-468.

Moffitt, T. E. (1993). Adolescence-Limited and Life-Course-Persistent antisocial

behavior: A developmental taxonomy. Psychological Review, 100(4), 674-701.

Moon, B., Blurton, D., & McCluskey, J. D. (2008). General strain theory and
delinquency: Focusing on the influences of key strain characteristics on
delinquency. Crime & Delinquency, 54, 582-613.

Moon, B., Hays, K., & Blurton, D. (2009). General strain theory, key strains, and
deviance. Journal of Criminal Justice, 37, 98-106.

Moon, B., & Morash, M. (2004). Adaptation of theory for alternative cultural contexts:
Agnew's general strain theory in South Korea. International Journal of
Comparative and Applied Criminal Justice, 28, 77-103.

Moon, B., Morash, M., McCluskey, C. P., & Hwang, H. W. (2009). A comprehensive test
of general strain theory: Key strains, situational- and trait-based negative
emotions, conditioning factors, and delinquency. Journal of Research in Crime
and Delinquency, 46, 182-212.

Moore, C. A. (1967). Introduction: The humanistic Chinese mind. In C. A. Moore (Ed.),

The Chinese mind: Essential of Chinese philosophy and culture (pp.1-10).
Honolulu, HI: University of Hawaii Press.

Morash, M., & Moon, B. (2007). Gender differences in the effects of strain on the
delinquency of South Korean youth. Youth & Society, 38, 300-321.

Mu, R. H. (1991). Stress, coping and physical and mental health in two cultures: Taiwan
and the United States. Unpublished doctoral dissertation, Kent State University.

Nagin, D. S., & Paternoster, R. (1991). On the relationship of past to future participation
in delinquency. Criminology, 29, 163-189.

Nakagawa, S., & Cuthill, I. C. (2007). Effect size, confidence interval and statistical
significance: A practical guide for biologists. Biological Review, 82,591-605.

217
Nation Police Agent, Ministry of the Interior. (2002). 2000 年世界各國刑案統計比較提
要分析。 [2000 worldwide criminal cases comparison and analysis]. Available at
http://www.npa.gov.tw/NPAGip/wSite/ct?xItem=26978&ctNode=11398&mp=1

_______ (2009).刑事案件處理嫌疑犯人數－年齡別 。[Criminal cases and suspects-

age distribution]. Available at
http://www.npa.gov.tw/NPAGip/wSite/lp?ctNode=11395&CtUnit=1740&BaseD
SD=7&mp=1

Nation Youth Commission, Executive Yuan. (2005). 青少年政策白皮書。 [Youth

policy]. Available at http://www.nyc.gov.tw/

Offer, D., Ostrov, E., Howard, K. I., & Atkinson, R. (1988). The teenage world:
Adolescent‘s self-image in 10 countries. New York, NY: Plenum Medical.

Oishi, S. (2003). Goals as cornerstones of subjective well-being: Linking individuals and

cultures. In E. Diener & E. M. Smith (Eds.), Culture and subjective well-being
(pp.87-112). Cambridge, MA: M.I.T. Press.

Olweus D. (1994). Bullying problems among school children: Long-term outcomes for
the victims and an effective school-based intervention program. In L. R.
Huesmann (Ed.), Aggressive behavior: Current perspectives (pp.411-448).
Mahwah, NJ: Lawrence Erlbaum.

Ostrowsky, M. K., & Messner, S. F. (2005). Explaining crime for a young adult
population: An application of general strain theory. Journal of Criminal Justice,
463-476.

Oyserman, D., Coon, H. M., & Kemmelmeier, M. (2002). Rethinking individualism and
collectivism: Evaluation of theoretical assumptions and meta-analyses.
Psychological Bulletin, 128, 3-72.

Parsons, T. (1951). The social system. Golencoe, IL: Free Press.

Paternoster, R., & Bachman, R. (2001). Explaining criminals and crime: Essays in
contemporary criminological theory. Los Angeles, CA: Roxbury.

Paternoster, R., Brame, R., Mazerolle, P., & Piquero, A. (1998). Using the correct
statistical test for the equality of regression coefficients. Criminology, 36, 859-
866.

Paternoster, R., & Mazerolle, P. (1994). General strain theory and delinquency: A
replication and extension. Journal of Research in Crime and Delinquency, 31,
235-263.
218
Pearlin, L. I. (1989). The sociological study of stress. Journal of Health and Social
Behavior, 30, 241-256.

Pedhazur, E. J., & Schmelkin, L. P. (1991). Measurement, design, and analysis: An

integrated approach. Hillsdale, NJ: Lawrence Erlbaum.

Peng, Y. F. (2002). 緊張、負面情緒與青少年偏差行為之研究−以台南地區為例。[A

study on strain, negative emotions and juvenile delinquency: A case of Tainan
area]. Unpublished master thesis, National Cheng Kung University.

Perez, D. M., Jennings, W. G., & Gover, A. R. (2008). Specifying general strain theory:
An ethnically relevant approach. Deviant Behavior, 29, 544-578.

Petersen, A. C., Kennedy, R. E., & Sullivan, P. (1991). Coping with adolescence. In M. E.
Colton & S. Gore (Eds.), Adolescent stress: Causes and consequences (pp.21-42).
New York, NY: Aldine de Gruyter.

Piquero, A., Macintosh, R., & Hickman, M. (2002). A validity of a self-reported

delinquency scale. Sociological Methods & Research, 30, 492-529.

Piquero, N. L., & Sealock, M. D. (2000). Gender and general strain theory: A preliminary
test of Broidy and Agnew‘s gender/GST hypotheses. Justice Quarterly, 21, 125-
158.

_______ (2004). Gender and general strain theory: A preliminary test of Broidy and
Agnew‘s gender/GST hypotheses. Justice Quarterly, 21, 125-158.

Plutchik, R. (1980). Emotion: A psychoevolutionary synthesis. New York, NY: Harper &
Row.

Pridemore, W. A., Chamlin, M. B., & Cochran, J. K. (2007). An interrupted time-series

analysis of Durkheim‘s social deregulation thesis: The case of the Russian
federation. Justice Quarterly, 24, 271-290.

Quicker, J. C. (1974). The effect of goal discrepancy on delinquency. Social Problem, 22,
76-86.

Radloff, L. S. (1977). A self-report depression scale for research in the general

population. Applied Psychological Measurement, 1, 385-401.

Raykow, T., & Marcoulides, G. A. (2000). A first course in structural equation modeling.
Mahwah, NJ: Lawrence Erlbaum.

219
Research, Development, and Evaluation Commission, Executive Yuan. (2001). 民眾對
目前生活及社會問題的看法。[Public opinion on current life situation and
social problem]. Available at
http://www.rdec.gov.tw/ct.asp?xItem=4023323&ctNode=12142&mp=100

Robbers, M. L. P. (2004). Revisiting the moderating effect of social support on strain: A

gendered test. Sociological Inquiry, 74, 546-569.

Robinson, D. H., & Levin, J. R. (1997). Reflections on statistical and substantive

significance, with a slice of replication. Educational Research, 26, 21-26.

Rubin, D. B. (1987). Multiple imputation for nonresponse in surveys. New York, NY:
Wiley.

_______ (1996). Multiple imputation after 18+ years. Journal of the American Statitsical
Association, 91, 473-489.

Rubin, D. B., & Schenker, N. (1986). Multiple imputation for interval estimation from
simple random samples with ignorable nonresponse. Journal of the American
Statistical Association, 81, 366-374.

Russell, J. A., & Yik, M. S. M. (1996). Emotions among the Chinese. In M. H. Bond
(Ed.), The handbook of Chinese psychology (pp.166-188). New York, NY:
Oxford University Press.

Sampson, R. J., & Laub, J. H. (1993). Crime in the making: Pathways and Turning
Points through Life. Cambridge, MA: Harvard University Press.

Sanchez, J. I., Spector, P. E., & Copper, C. L. (2005). Frequently ignored methodological
issues in cross-cultural stress research. In P. T. P. Wang & L. C. J. Wong (Eds.),
Handbook of Multicultural Perspectives on Stress and Coping (pp.29-53).
Springer.

Schafer, J. L. (1997). Analysis of incomplete multivariate data. London: Chapman & Hall.

Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state of the art.
Psychological Methods, 7,147-177.

Schafer, J. L., & Olsen M. K. (1998). Multiple imputation for multivariate missing-data
problems: A data analyst‘s perspective. Multivariate Behavioral Research, 33,
545-571.

220
Schneiders, J., Nicolson, N. A., Berkhof, J., Feron, F. J., deVries, M. W., & Os, J. V.
(2007). Mood in daily contexts: Relationship with risk in early adolescence.
Journal of Research on Adolescence, 17, 697-722.
Schreck, C. J. (1999). Victimization and low self-control: An extension and test of a
general theory of crime. Journal of Quantitative Criminology, 16, 633-654.

Schreck, C. J., Stewart, E. A., & Fisher, B. S. (2006). Self-control, victimization, and
their influence on risky lifestyle: A longitudinal analysis using panel data. Journal
of Quantitative Criminology, 22, 319-340.

Schwartz, S. H. (1990). Individualism-collectivism critique and proposed refinements.

Seiffge-Krenke, I. (2000). Causal links between stressful events, coping style, and
adolescent symptomatology. Journal of Adolescence, 22, 675-691.

Selya, R. M. (1995). Taipei. New York, NY: John Wiley & Sons.

Sharp, S. F., Brewster, D., & Love, S. R. (2005). Disentangling strain, personal attributes,
affective response and deviance: A gendered analysis. Deviant Behavior, 26, 133-
157.

Sharp, S. F., Terling-Watt, T. T., Atkins, L. A., Gilliam, J. T., & Sanders, A. (2001).
Purging behavior in a sample of college females: A research note on general
strain theory and female deviance. Deviant Behavior, 22, 171-188.

Shaver, P., Schwartz, J., Kirson, D., & O‘Connor, C. (1987). Emotion knowledge:
Further exploration of a prototype approach. Journal of Personality of Social
Psychology, 52, 1061-1086.

Shek, D. T. L., & Lee, T. Y. (2007). Parental control and Chinese adolescents in the
context of positive youth development programs in Hong Kong. In J. Merrick &
A. Omar (Eds.), Adolescent behavior research: International perspectives (pp.3-
17). New York, NY: Nova Science.

Shi, Y. W. (2004). 國中生生活壓力、負向情緒調適、社會支持與憂鬱情緒之關聯。[

The relationships among life stress of junior high school students, negative mood
regulation, social support, and depressive mood]. Unpublished master thesis,
National Cheng Kung University.
Short, J. F., & Hughes, L. A. (2008). Juvenile delinquency and delinquents: The nexus
of social change. New Jersey: Pearson Education LTD.

Sigfusdottir, I.-D., Farkas, G., & Silver, E. (2004). The role of depressed mood and anger
in the relationship between family conflict and delinquent behavior. Journal of
Youth and Adolescence, 33, 509-522.
221
Simons, R. L., & Chen, Y. F., Stewart, E. A., & Brody, G. H. (2003). Incidents of
discrimination and risk for delinquency: A longitudinal test of strain theory with
an African American sample. Justice Quarterly, 20, 827-854.

Sinharay, S., Stern, H. S., & Russell, D. (2001). The use of multiple imputation for the
analysis of missing data. Psychological Methods, 6, 317-329.

Slocum, L. A., Simpson, S. S., & Smith, D. A. (2005). Strained lives and crime:
Examining intra-individual variation in strain and offending in a sample of
incarcerated women. Criminology, 43, 1067-1110.

Smith, C. S., & Hung, L-C. (2005). Juvenile delinquency in Taiwan: Sign of times? Paper
presented at the annual meeting of American Society of Criminology, Toronto,
Canada.

Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural

equation models. Sociological Methodology, 13, 290-312.

Sollenberger, R. T. (1968). Chinese-American child-rearing practices and juvenile

delinquency. Journal of Social Psychology, 74, 13-23.

Solomon, R. C. (1976). The passions. Garden City, NY: Doubleday.

Spielberger, C. D. (1988). Manual for the sate-trait anger expression inventory (Research
edition). Odessa, FL: Psychological Assessment Resource Inc.

Stafford, M. C. (2004). Juvenile delinquency. In G. Ritzer (Eds.), Handbook of social

problems: A comparative international perspective (pp.480-494). Thousand
Oaks, CA: Sage.

Steinberg, L., & Cauffman, E. (1996). Maturity of judgment in adolescence: Psychosocial

factors in adolescent decision making. Law and Human Behavior, 20, 249-272.

Steinberg, L., Dornbush, S., & Brown, B. B. (1993). Ethnic differences in adolescent
achievement: An ecological perspective. American Psychologist, 47, 723-729.

Storch, E. A., Roberti, J. W., & Roth, D. A. (2004). Factor structure, concurrent validity,
and internal consistency of the Beck Depression Inventory Second-Edition in a
sample of college students. Depression and Anxiety, 19, 187-189.

Tanzer, N. K., Sim, C. Q. E., & Speilberger, C. D. (1996). Experience, expression, and
control of anger in a Chinese society: The case of Singapore. Stress and Emotion:
Anxiety, Anger, and Curiosity, 16, 51-65.
222
Terrell, F., Miller, A. R., Foster, K., & Watkins, C. E. Jr. (2006). Racial discrimination-
induced anger and alcohol use among black adolescents. Adolescence, 41, 485-
492.

Thoits, P. (1983). Dimensions of life events that influence psychological distress: An

evaluation and synthesis of the literature. In H. B. Kaplan (Eds.), Psychological
stress: Trends in theory and research (pp.33-101). New York: Academic Press.

______ (1995). Stress, coping, and social support process: Where are we? What next?
Journal of Health and Social Behavior, Extra Issue, 53-79.

Thompson, B. (1993). AERA editorial policies regarding statistical significance testing:

Three suggested reform. Education Research, 25, 26-30.

Tolan, P. H., & Thomas, P. (1995). The implications of age of onset for delinquency risk
II: Logitudinal evidence. Journal of Abnormal Child Psychology, 23, 157-182.

Tosun Non-profit Organization. (2001). E 世代青少年網咖經驗調查報告。 [The report

of E generation adolescent cyber café experience]. Available at
http://www.tosun.org.tw/database_detail.asp?main_id=00002

Tsai, M. H. (2005). 生活壓力、制握信念、社會支持與青少年偏差行為之關係。[The

relationships among life stress, locus of control, social support, and juvenile
delinquency]. Unpublished master thesis, National Cheng Kung University.

Trandis, H. C. (1989). The self and social behavior in differing cultural contexts.
Psychological View, 96, 506-520.

______ (1994). Culture and social behavior. New York, NY: McGraw-Hill.

______ (1995). Individualism & Collectivism. Bulder: Westview Press.

Trojanowicz, R.C., Morash, M., & Schram, P. J. (2001). Juvenile delinquency: Concepts
and control (6th ed.). NJ: Prentice Hall.

Tung, Y. Y. (2003). 一般化緊張理論的實證性檢驗。[An empirical test of general

strain theory]. 犯罪學期刊, 6, 103-128.

______ (2007). 社會支持對生活壓迫性因素與年暴力行為間的關聯之影響。 [The

effects of social support on strain-juvenile violence relationship]. Criminal Justice
Policy & Crime Research, 10, 229-247.

223
Tung, Y. Y., & Wu, Y. L. (2008). 青少年生活緊張與自我傷害之探討。[Discussion on
the relationship among juvenile life strain and self-mutilation]. Available at
http://www.moj.gov.tw/ct.asp?xItem=98681&ctNode=28261&mp=001

Turner, R. J., & Wheaton, B. (1995). Checklist measurement of Stressful life event. In S.
Cohen, R. C. Kessler, & L. U. Gordon (Eds.), Measuring stress: A guide for
health and social scientists (pp.29-58). New York: Oxford University Press.

Turner, R. J., Wheaton, B., & Lloyd, D. (1995). The epidemiology of social stress.
American Sociological Review, 60, 104-125.

U.S. Census Bureau. (1990). United States census 1990 [Data file]. Washington,
DC: Available at http://www.census.gov

______ (2000). United States census 2000 [Data file]. Washington,

DC: Available at http://www.census.gov

______ (2011). Statistic Abstract. Available at

http://www.census.gov/compendia/statab/cats/income_expenditures_poverty_we
alth.html

Van de Vijver, F., & Leung, K. (1997). Methods and data analysis for cross-cultural
research. Thousand Oaks, CA: Sage.

Van Gundy, K. (2002). Gender, the assertion of autonomy, and the stress process in
young adulthood. Social Psychology Quarterly, 65, 346-363.

Vaux, A., & Ruggiero, M. (1983). Stressful life change and delinquent behavior.
American Journal of Community Psychology, 11, 169-183.

Verbeke, G., & Molenberghs, G. (2000). Linear mixed models for longitudinal data. New
York, NY: Springer-Verlag.

Verrill, S. W. (2008). Social Structure and Social Learning: Mediating or moderating?

El Paso, TX: LFB Scholarly.

Wan, Z. Z. (2001). 台北縣國中生之壓力源、因應方式與生活適應之先關研究。[A

study on Taipei county junior high school students‘ stresses and their coping
behaviors]. 生活科學學報, 7, 61-86.

Wareham, J., Cochran, J. K., Dembo, R., & Sellers, C. S. (2005). Community, strain,
and delinquency: A test of a multi-level model of general strain theory. Western
Criminological Review, 6, 117-133.

224
Walls, M. L., Chapple, C. L., & Johnson, K. D. (2007). Strain, emotion, and suicide
among American Indian youth. Deviant Behavior, 28, 219-246.

Wilson, R. W. (1970). Learning to be Chinese: The political socialization of children n

Taiwan. Cambridge, MA: M.I.T. Press.

Wu, C. (2001). 青少年網路應用問題之探討。[Discussion on the problem of juvenile

internet use]. Criminal Justice Policy & Crime Research, 4, 127-140.

Wu, D. Y. H. (1996). Chinese childhood socialization. In M. H. Bond (Ed.), The

handbook of Chinese psychology (pp.143-154). New York, NY: Oxford
University Press.

Xie, H., Farmer, T. W., & Cairns, B. D. (2003). Different forms of aggression among
inner-city African-American children: Gender, configurations, and school social
network. Journal of School Psychology, 41, 355-375.

Xu, C. J., & Xu, C. Z. (2000). 青少年不良幫派成行過程及相關因素之研究。 [A study

on the correlates of juvenile gang emerging]. Criminal Justice Policy &
Crime Research, 3, 77-111.

Xie, J. Q. (1998) 少年飆車行為之分析與防制。 [Analysis and prevention of juvenile

joyriding]. Criminal Justice Policy & Crime Research, 1, 185-216.

Xu, F. S. (2005). 台灣地區少年非行狀況與防制策略之探討。[Discussion on Taiwan‘s

juvenile delinquent situation and prevention strategies]. Criminal Justice Policy &
Crime Research, 8, 265-288.

_______ (2007). 校園暴行之成因與防制策略。 [The causes of school violence and its

prevention strategies]. Available at
http://www.tosun.org.tw/database_detail.asp?main_id=00038

Xu, F. S., & Hwang, F. M. (2004). 青少年痛苦指數與偏差行為之研究。[Empirical

study on the relationship between juvenile negative life index and delinquency].
Criminal Justice Policy & Crime Research, 7, 133-158.

Xu, Z. Y. (2005). 緊張因素、負面情緒、制握信念與男女性青少年偏差行為之關聯

性研究。[The relationships among strain factors, negative emotions, locus of
control and adolescent deviance]. Unpublished master thesis, National Cheng
Kung University.

Yamaguchi, S., Kuhlman, D. M., & Sugimori, S. (1995). Personality correlates of

allocentric tendencies in individualist and collectivist cultures. Journal of Cross-
225
 Cultural Psychology, 25, 146-158.

Yang, J. (2004). Taiwan. In M. P. Duffy & S. E. Gillig (Eds.), Teen gangs: A global view
(pp.179-193). Westport, CT: Greenwood Press.

Yang, K. S. (1986). Chinese personality and its change. In M. H. Bond (Ed), The
Psychology of the Chinese People (pp.107-170). New York, NY: Oxford
University Press.

_______ (1995). Chinese social orientation: An integrative analysis. In T. Y. Lin, W. S.

Tseng, & E. K. Yeh (Eds.), Chinese societies and mental health (pp.19-39). New
York, NY: Oxford University Press.

Yang, K. S., & Lu, L. (2005). Social- and individual-oriented self-actualizer: Conceptual
analysis and empirical assessment of their psychological characteristics.
Indigenous Psychological Research in Chinese Societies, 23, 71-143.

Yi, C. C., & Wu, C. L. (2004). Taiwan. In J. J. Slater (Ed.), Teen life in Asia (pp.223-241).
Westport, CT: Greenwood Press.

Yoshihama, M. (2002). Battered women‘s coping strategies and psychological distress:

Differences by immigration status. American Journal of Community Psychology,
30, 429-452.

Yu, A. B. (1996). Ultimate life concerns, self, and Chinese achievement motivation. In M.
H. Bond (Ed.), The handbook of Chinese psychology (pp.227-246). New York,
NY: Oxford University Press.

Zeng, H. Q. (2008). 青少年藥物濫用問題分析。[Analysis of juvenile drug abuse

problem] (National Policy Foundation, Culture and Education Analysis 097-011).
Available at http://www.npf.org.tw/post/3/4293

Zhang, W. B. (2003). Taiwan’s modernization: Americanization, modernizing Confucian

manifestations. River Edge, NJ: World Scientific.

Zhou, B. S. (2000). 台灣地區在校青少年藥物使用流行病學調查研究。

[Epidemiological study on illicit drug abuse by school adolescents in Taiwan].
Taipei, Taiwan: Department of Health, Executive Yuan, Taiwan.

Zhou, S. X. (2001). 愛上學的孩子，不會變壞嗎？論學校因素與青少年偏差行為之

關係。[Children who love to go to school are not going to be bad? Discussion on
the relationship between school factors and juvenile delinquency]. 應用心理學
研究, 11, 93-115.

226
APPENDICES

227
 Appendix A: Survey questionnaire

Delinquency (both U.S. and Taiwan version)

Have you ever…in the past 12 months?
1. Purposely damaged or destroyed property that did not belong to you.
2. Hit someone with the idea of hurting them.
3. Used alcohol.
Response: 0 = No, never; 1 = Yes
Aggression (both U.S. and Taiwan version)
How many times in the past 12 months have you…?
1. Used physical actions (such as slapping, kicking, or hitting hard) against a brother or
sister.
Response: 1 = Never; 2 = Once; 3 = Twice; 4 = 3-5times; 5 = 6 or more times;
6 = no brother/sister
Goal strain scale (both U.S. and Taiwan version)
How much do you agree or disagree with the following statements?
1. The teachers here embarrass you when you don‘t know the right answers (respect).
2. My teachers don‘t respect my opinions as much as I would like (respect).
3. My parents don‘t respect my opinions as much as I would like (respect).
4. My classmates do not like me (relationship with others).
5. People my age tend to push me around (relationship with others).
6. My parents don‘t give me a say in what the rules should be (autonomy).
7. People my age treat me like I‘m still a kid (autonomy).
Response: 1 = Strongly agree; 2 = Agree; 3 = Disagree; 4 = Strongly disagree
Unjust strain scale (both U.S. and Taiwan version)
How much do you agree or disagree with the following statements?
1. Many students don‘t study as hard as I do, but they still make better grades.
2. I don‘t have as much money as other students in this school.
3. Even though I try hard, my grades are never good enough.
4. Other students get special favors from the teachers here that I don‘t get.
5. Even though I work hard, I never seem to have enough money.
6. Compared to the rules my friends have to abide by, the rules my parents set for me
are unfairly strict.
7. No matter how responsible I try to be, my parents don‘t trust me to do things on my
own.
Response: 1 = Strongly agree; 2 = Agree; 3 = Disagree; 4 = Strongly disagree
Negative Life-event scale (both U.S. and Taiwan version)
Which of the following things happened to you in the past 12 months?
1. Change schools.
2. Parents divorced .
3. Parent moved out or away.
4. Broke up with boyfriend or girlfriend.
228
5. Moved to new neighborhood.
6. Death of a relative.
7. Lost a friendship.
8. Pet died or disappeared.
9. Dropped from or quit athletic team or school activities.
10. Parents lost job for more than two months.
Response: 0 = No; 1 = Yes
Victimization scale (both U.S. and Taiwan version)
Have the following things ever been done to you personally?
1. Been picked on or teased by other kids who were being mean.
2. Had your backpack, lunch money, or other personal things stolen from you.
3. Had a bicycle or motorcycle stolen.
4. Been hit by someone trying to hurt you.
5. Had someone use a weapon or force to get money or things from you.
6. Been attacked by someone with a weapon or by someone trying to serious hurt or kill
you.
Response: 0 = No; 1 = Yes
Anger scale (both U.S. and Taiwan version)
How often do you think these statements describe you?
1. I feel annoyed when people don‘t notice that I‘ve done good work.
2. When I get mad, I say nasty things.
3. It makes me very mad when I am criticized in front of others.
4. When I get frustrated, I feel like hitting others.
5. I feel furious when I work hard but get a poor grade.
6. It makes me mad that others are able to spend more money than I can.
7. It makes me mad when I don‘t get the respect from others that I deserve.
8. If things upset other people, it‘s their problem, not mine.
Response: 1 = Almost never; 2 =Sometimes; 3 = Often; 4 = Almost always
Depression scale (both U.S. and Taiwan version)
How often do you think these statements describe you?
1. I find it hard to keep my mind on school work.
2. I don‘t look forward to things as much as I used to.
3. I sleep very well.
4. I have lots of energy.
Response: 1 = Almost never; 2 =Sometimes; 3 = Often; 4 = Almost always

229
 Appendix B: EFA tables

Table B1 Principal Axis Factor Analysis for Goal Strain and Unjust Strain
Factor Factor Loadings
U.S. Taiwan
Goal strain
Goal strain1-teacher does not respect me .452 .669
Goal strain2-parents do not respect me .623 .725
Goal strain3-teachers embarrass you .422 .569
Goal strain4-parents do not give me a say .533 .686
Goal strain5-people treat me like a kid .576 .503
Goal strain6-my classmates do not like me* .443 .530
Goal strain7-people push me around .441 .537
Sum of squared loadings 1.775 2.591
Unjust strain
Unjust strain1-students do not study hard but make better grade .323 .318
Unjust strain2-I do not have as much money as other students .361 .452
Unjust strain3-my grades are never good enough even I try hard .539 .423
Unjust strain4-other students get special favors from teachers .365 .443
Unjust strain5-I never seem to have enough money .585 .596
Unjust strain6-my parents set unfairly strict rules .496 .550
Unjust strain7-my parents do not trust me to do things on my own .532 .623
Sum of square loadings 1.529 1.726
*Reverse coded

230
 Table B2 Principal Axis Factor Analysis for Anger and Depression
Factor Factor Loadings
U.S. Taiwan
Anger
Anger1-I feel annoyed if people do not notice my good work .507 .550
Anger2-when I get mad, I say nasty things .485 .509
Anger3-I get very mad when I am criticized in front of others .616 .671
Anger4-When I get frustrated, I feel like hitting something .470 .607
Anger5-I feel furious when I work hard but get a poor grade .557 .517
Anger6-I feel mad when people do not let me make my .555 .642
Anger7-I feel mad that others have more money than I do .539 .461
Anger8-I feel mad when I do not get respect that I deserve .614 .624
Sum of squared loadings 2.378 2.661
Depression
Depression1-I find it hard to keep my mind on school work .349 .106
Depression2-I do not look forward to things .475 .149
Depression3-I sleep very well* .575 .793
Depression4-I have lots of energy* .484 .784
Sum of square loadings .913 1.277
*Reverse coded

231