5 - Banche Mercati Territoriali 2010

Banks, Local Credit
Markets and Credit Supply

Seminari e convegni
Workshops and Conferences

n
u
m
b
e
r
5
A
u
g
u
s
t

2
0
1
0
2
0
0
9

-

2
M
e
z
z
o
g
i
o
r
n
o

e

p
o
l
i
t
i
c
h
e

r
e
g
i
o
n
a
l
i
Banks, Local Credit
Markets and Credit Supply
Seminari e convegni
Workshops and Conferences
Earlier versions of the papers in this volume were presented at the Conference on Banks, Local Credit
Markets and Credit Supply, held in Milan, Universit Bocconi, on 24 March 2010. The workshop
was organized by Banca dItalia - Milan Branch, jointly with the Paolo Baff Centre of Universit
Bocconi. Donato Masciandaro (Universit Bocconi) and Paola Rossi were in charge of the scientifc
organization of the Conference. The papers were selected by Michele Benvenuti, Luigi Cannari,
Giorgio Gobbi, Paolo Emilio Mistrulli, Sauro Mocetti, Marcello Pagnini and Paola Rossi. The volume
has been edited by Luigi Cannari, Marcello Pagnini and Paola Rossi, with the editorial assistance of
Paolo Natile. These papers are part of a research project conducted by Banca dItalia on Banking
organization and local credit markets, coordinated by Luigi Cannari, Marcello Pagnini and Paola
Rossi. We wish to thank all the banks and fnancial institutions for kindly participating in the survey on
lending organization that was at the core of the project.
The proceedings may be requested from:
Banca dItalia
Biblioteca
Via Nazionale, 91
00184 Roma
Italy
They are also available on the Internet at: www.bancaditalia.it
The views expressed in this volume are those of the authors of the individual papers and do not necessarily
refect those of Banca dItalia.
Printed by the Printing Offce of Banca dItalia, Rome, August 2010
2010 Banca dItalia

CONTENTS

Introduction

Luigi Cannari, Marcello Pagnini and Paola Rossi .. 5

Session 1
NEW TECHNOLOGIES AND BANKING ORGANIZATION
1. The organization of lending and the use of credit scoring techniques in Italian
banks: Results of a sample survey

Giorgio Albareto, Michele Benvenuti, Sauro Mocetti, Marcello Pagnini and
Paola Rossi ... 11
2. Information technology and banking organization
Sauro Mocetti, Marcello Pagnini and Enrico Sette .. 49
3. Bank acquisitions and decentralization choices
Enrico Beretta and Silvia Del Prete .. 77
Discussion
Elena Beccalli ....... 103
Barbara Chizzolini .... 107

Session 2
BANKING ORGANIZATION AND SMALL BUSINESS LENDING
4. The retail activity of foreign banks: Effects on credit supply to households and
firms
Luigi Infante and Paola Rossi ....... 113
5. Debt restructuring and the role of lending technologies
Giacinto Micucci and Paola Rossi ... 143
6. Loan officer authority and small business lending. Evidence from a survey
Michele Benvenuti, Luca Casolaro, Silvia Del Prete and Paolo Emilio Mistrulli .. 175
Discussion
Angelo Baglioni ........ 193
Paola Bongini ........ 197

CONTRIBUTORS

Angelo Baglioni (Catholic University of Milan)
Alberto Barcella (Confindustria Lombardia)
Elena Beccalli (Catholic University of Milan and London School of Economics)
Michele Benvenuti (Bank of Italy)
Enrico Beretta (Bank of Italy)
Paola Bongini (University of Milan-Bicocca)
Franco Bruni (Bocconi University)
Luigi Cannari (Bank of Italy)
Luca Casolaro (Bank of Italy)
Barbara Chizzolini (Bocconi University)
Gregorio De Felice (Intesa Sanpaolo)
Silvia Del Prete (Bank of Italy)
Francesco Giordano (Unicredit Group)
Luigi Infante (Bank of Italy)
Donato Masciandaro (Bocconi University)
Salvatore Messina (Bank of Italy)
Giacinto Micucci (Bank of Italy)
Paolo E. Mistrulli (Bank of Italy)
Sauro Mocetti (Bank of Italy)
Marcello Pagnini (Bank of Italy)
Angelo Porta (Bocconi University)
Paola Rossi (Bank of Italy)
Salvatore Rossi (Bank of Italy)
Enrico Sette (Bank of Italy)
INTRODUCTION
Luigi Cannari
-
, Marcello Pagnini
**
and Paola Rossi
***

In the past fifteen years the Italian banking system changed dramatically. Regulatory
evolution, increased internationalization and the increased market concentration following M&As
in the sector (including cross-border mergers) altered the characteristics of the banking industry as
well as the supply of banking services. Partly in response to the spread of ICT (information and
communication technologies), banks reorganized their lending activity, by changing their mode of
interaction with customers and adapting the procedures they use to assess borrowers
creditworthiness and to monitor loan performance.
Internal organization is crucial to banks activity. It has been examined in depth including
for supervisory purposes to assess the adequacy of operational risk safeguards and, more recently,
corporate governance and control. The papers presented at this conference focus on the interaction
between branch managers and small and medium-sized enterprises (SMEs). This choice reflects the
idea that lending relations are built through repeated interaction over time between the branch
manager and the borrower. Such interaction is shaped by the type of organization the bank has
adopted: the degree of decision-making centralization, modes of communication between
hierarchical levels, and the adoption of statistical techniques for risk evaluation can significantly
affect the behaviour of branch managers; this, in turn, may be reflected in the allocation of credit to
SMEs.
The papers rely on a special survey, conducted in 2007 on more than 300 Italian banks
(Albareto, Benvenuti, Mocetti, Pagnini and Rossi). The survey gathered information on factors like
the distance geographical and hierarchical between the headquarters of the bank (i.e. the holder
of the formal authority within the bank) and its branch network, the decision-making powers
delegated to branch managers (proxied by the size of the loans they can grant on their own
authority to small firms), their incentives and tenure in office, the adoption of credit scoring
techniques, the type of data used to generate the scores, and the actual use of these procedures in
the loan process (decision to grant, decision on size, pricing, etc.).
While these organizational features differ markedly from bank to bank, this heterogeneity is
only partially explained by the differences in bank size. The survey also highlighted the fact that
between 2003 and 2006 the autonomy of branch managers increased; this trend went hand in hand
with the enhanced mobility of these managers as well as a greater use of compensation schemes
linking their salaries to the performance of the local branch they managed.
During the three years considered, credit scoring and internal rating techniques, while still
absent at many smaller banks, were almost universally adopted by larger ones. Although scoring
techniques play a central role in larger banks decision to grant credit, they are not frequently used
to determine the terms and conditions of loans.
The papers presented at the conference address two main issues: how bank concentration and
the introduction of ICT affect some organizational variables that are at least potentially relevant in

-
Bank of Italy, Economic Research and International Relations Area.
**
Bank of Italy, Bologna.
***
Bank of Italy, Milan.
6 Luigi Cannari, Marcello Pagnini and Paola Rossi
explaining relations between banks and SMEs; and the relation between structural changes in
supply and organizational variables on the one hand and lending to the business sector on the other.
Among the papers dealing with the first issue, Mocetti, Pagnini and Sette analyse whether
the increasing use of ICT altered the delegation of credit power to branch managers. The new
technologies can either reduce or increase the degree of decision-making centralization within a
bank, depending on whether the greater capacity of the centre to monitor the branches following
the introduction of these technologies, a force pushing toward decentralization, would prevail over
the lower cost of communication between branch and head office, an effect that would lead to more
centralization. They find that the banks relying most heavily on ICT have generally delegated more
power to branch managers.
Beretta and Del Prete examine the changes in the distribution of lending power together with
the turnover of branch managers subsequent to acquisitions. After a takeover, the size of the loan
that branch managers or CEOs can grant on their own authority tends to increase, while the average
term of office of branch managers diminishes. The increase in decision-making independence in
the target bank is more pronounced when the takeover involves banks located in the same local
areas as the branch network and head office.
Within the group of papers dealing with the effects of organizational variables on credit
supply, Infante and Rossi investigate how the increasing presence of foreign banks in Italy affects
competition among different types of borrowers (households vs. firms) and geographical areas.
Competition from foreign intermediaries is stronger when products are standardized and
information asymmetries less important: an increase in the activity of foreign banks leads to greater
customer mobility, both for households and firms; it reduces interest rates and the collateral
required from households, while no impact on interest rates and only a slight reduction in collateral
are detected for firms.
Benvenuti, Casolaro, Del Prete and Mistrulli show how the branch loan officers authority
plays a key role in explaining bank specialization in small-business lending: banks that delegate
more decision-making power to their branch managers, assign greater importance to soft
information and have lower manager turnover are more willing to lend to small firms than other
banks. As bank size increases, the propensity to lend to SMEs tends to diminish; this effect is partly
offset by organizational measures heightening local loan officers independent decision-making
powers.
Banking organization also influences the decision to intervene in favour of troubled firms.
Micucci and Rossi find that about one third of the firms in the sample succeed in getting their debt
restructured a maturity extension or further lending by at least one of their lending banks. The
probability of debt restructuring is higher when the bank is geographically closer to borrowing
firms, relies more on soft than hard information, and adopts a decentralized structure with more
power allocated to local managers. However there are important complementarities among
organizational variables: the adoption of credit scoring increases the likelihood of restructuring if
banks also use these techniques systematically in the monitoring process and if they adopt more
decentralized structures.
The data collected in the survey as well as the entire research project referred to a time
period that preceded the financial crisis and the subsequent recession that has affected the world
and the Italian economy in recent years. Specific inquiries were conducted to determine what
features of bank-firm relations and what organizational arrangements helped to mitigate the
contraction of credit during the crisis. In terms of statistical data-gathering, new surveys of credit
Introduction 7
demand and supply at regional level were conducted.
1
Further surveys have commenced recently.
The additional data will make it possible to gauge the extent to which the tendency towards more
decentralized decision-making may have been modified by the heightening of credit risk due to the
crisis and by the need to prevent deterioration in loan asset quality. The data will also help to
investigate the relations between the organization of branches and the granting of loans to SMEs
during the crisis. The operation of rating systems can also be analysed in the light of the impact of
recession. Some preliminary studies indicate that as the crisis worsened, automatic credit scoring
methods became increasingly important in the lending decisions of major banking groups.
However, the accounting data that constitute the input for automatic scoring systems suffer a
structural lag of several months, and rarely include qualitative factors helpful in assessing a firms
future prospects.
The main message to emerge from the studies presented today is that mergers and
acquisitions in banking and more intensive use of ICT do not imply, per se, neglect of the needs of
SMEs. These analyses indicate that the vocation for local banking does not necessarily depend
on the banks size but does require suitable organizational arrangements. In particular, banks with
more decentralized decision-making and with longer tenure for branch managers appear to have a
greater propensity to lend to SMEs and to intervene in debt restructuring on behalf of firms in
financial distress.
The use of rating and credit scoring techniques is increasingly common. Gathering and
processing data is essential to the evaluation of creditworthiness. Implementing methodologies that
analyse the available statistical data within a consistent framework is not incompatible with the
initiative and autonomy of the local branch manager. Indeed, within some types of organization
these methodologies reinforce the role of local branch managers, enabling headquarters to grant
them greater authority. But especially during a period of crisis, it is essential to combine codified
data (hard information) with knowledge of the customer (soft information). In the words of
Governor Draghi in his address to the Italian Banking Association on 8 July 2009, It is equally
important that in deciding on loans banks use all the information available, that they supplement
statistical scoring methods which lose some of their predictive power at times of crisis with
direct knowledge of customers and their real potential for growth and profitability in the long
term.

1
Economie regionali 22, February 2010.

SESSION 1
NEW TECHNOLOGIES AND BANKING ORGANIZATION

THE ORGANIZATION OF LENDING AND THE USE OF CREDIT SCORING
TECHNIQUES IN ITALIAN BANKS: RESULTS OF A SAMPLE SURVEY
Giorgio Albareto
-
, Michele Benvenuti
**
, Sauro Mocetti
***
,
Marcello Pagnini
***
and Paola Rossi
****

1. Introduction and main results
During the 1990s two major factors affected the Italian banking industry: liberalization and
an intensive wave of technological innovation originating in the ITC sector. As a result, the
banking system underwent a consolidation process, banks expanded and entered new markets, and
internal decision-making processes for granting loans were completely overhauled. The ways in
which households and firms accessed credit changed. In the wake of these transformations, banks
grew in size and organizational complexity, and now found themselves having to manage their
presence in a number of different geographical and product markets. Large banks were not the only
ones affected by this trend, as small and medium-sized banks frequently joined bigger banking
groups or expanded internally, in both cases determining a leap in size that sometimes led to
organizational discontinuity. The rapid advances in ICT technologies had a profound effect on the
output of the entire banking industry. These transformations imply that we need to have an updated
and deeper knowledge about how banks organize the several aspects of their lending activities
(customer screening, the terms and conditions of lending, monitoring of the borrowers conduct
etc.)
The literature on bank-firm relationships generally treats banks as unitary entities and
neglects the characteristics of their internal structure. Recently, however, the literature on
organization has spawned a number of papers that emphasize the importance of the strategic
interaction among managers in charge of various functions within the banking organization, who
have different information and who are bearers of interests that do not necessary coincide. It has
been shown both theoretically and empirically that the ways in which this interaction occurs
can affect the effectiveness of credit allocation, especially in the case of SMEs.
One of the main consequences technological change had for credit markets was the
introduction of credit scoring
1
techniques based on standardized data. Despite their increasing
importance, including in the Italian market, there are few studies on the diffusion and use of these
procedures.
To collect data useful for understanding these changes, in 2007 the Bank of Italy conducted a
survey on over 300 banks, representing the universe of intermediaries of a certain minimum size
and organizational complexity. This report presents the results of the survey.

-
**
Bank of Italy, Florence.
***
****
The authors wish to thank Guglielmo Barone, Enrico Beretta, Luigi Cannari, Xavier Freixas, Giorgio Gobbi, Giacinto Micucci and
Paolo Emilio Mistrulli for their comments. We are especially indebted to Guglielmo Barone for his help in formulating the
questionnaire that was distributed to the banks. A special thank you also goes to Rino Ferlita and Angela Romagnoli whose help in
the various phases of the survey proved indispensable, and to Paolo Natile for his assistance in the preparation of several programs.
The survey was made possible by the kind cooperation of the Bank of Italys branch network.
1
We use this term to denote all the automated techniques for assessing creditworthiness, described later in the paper.
12 Giorgio Albareto, Michele Benvenuti, Sauro Mocetti, Marcello Pagnini and Paola Rossi

The analysis of banks internal organization revealed profound differences among the Italian
intermediaries from four standpoints: the geographic distance between a banks headquarters and
its branch network; the decision-making autonomy of the branch manager, proxied by the amount
of small business credit he or she is authorized to extend in proportion to that which can be
approved by the CEO; the length of branch managers tenure; the use of incentives for their
remuneration. Part of this heterogeneity is accounted for by the size and institutional differences of
the banks surveyed. However, some heterogeneity still exists among homogenous groups of
intermediaries. The results confirm how the internal structure of lending activities adapts to specific
circumstances and forms a crucial component of banks competitive strategies. An implication of
these findings is that an analysis of these phenomena must employ a broader and richer taxonomy
than the traditionally one, based on banks size.
For most of the participating banks, between 2003 and 2006 the distance between their
headquarters and branch network increased. Bank managers enjoyed greater mobility and
autonomy in decision-making, and economic incentives were more frequently used for their
remuneration. The results do not support the thesis that the advent of new technologies greatly
diminished the role of bank managers, with negative repercussions for banks direct interaction
with SMEs. On the contrary, it is possible that lower communication costs favoured the greater
autonomy of local managers in the periphery (branches or local decision-making centres).
Increased mobility could be the result of partially exogenous events to the banks strategy (i.e.
mergers and acquisitions, tough competition in the local credit markets). It could also be the result
of an active policy by banks to reduce the costs of opportunistic practices by local branch
managers.
The survey has shown that credit scoring have spread among Italian intermediaries, with the
sharp acceleration in recent years probably related to the entry into effect of the new Capital
Adequacy Accords (Basel II). The diffusion process was more pronounced among large banks, in a
position to exploit economies of scale. Credit scoring techniques mostly process balance sheet data,
historically the most frequently used element for evaluating creditworthiness. Larger banks
generally use internally developed models, and give great emphasis on the qualitative data, such as
borrowers corporate governance and organization and the project to be financed (for large firms
especially).
Although scoring techniques play a central role in the decision to grant credit, they are not
frequently used to determine the terms and conditions of loans. The scores generated by the
application of these techniques appear more stringent for large banks than for smaller ones.
Overall, the results suggest that the new techniques, which banks are still in the process of adapting
to, flank but do not entirely replace the previous evaluation processes and data sources.
The report is organized as follows: section 2 surveys the literature on the internal
organization of banks and the use of credit scoring; section 3 contains information on the sample of
banks interviewed; section 4 analyzes the relationship between the centre and branch network in
banks internal organization and the role of bank managers; section 5 examines the diffusion and
importance of credit scoring in business lending. The appendix reproduces the questionnaire
distributed to the participating banks.

2. The organization of lending and the use of credit scoring: the main issues
Usually banks rely on a mix of data sources to assess firms creditworthiness. Some
informations are easy to classify and transmit from a distance (hard information), while others are
acquired through personal contacts and comprise qualitative elements that are difficult to
communicate to the persons other than those who collected them (soft information). It is generally
The organization of lending and the use of credit scoring techniques in Italian banks: Results of a sample survey 13

thought that qualitative data play a greater role in the evaluation of start-ups and small businesses,
which are prone to having more opaque information (for example owing to fewer years of
experience), less reliable processed data, or in any event are subject to less strict information
requirements compared with large corporations (such as accounting data).
The collection of quantitative but especially qualitative information on small firms is done
though the branch manager. It is usually at this level of the bank organization that the first contact
is made with the small firm, the assessment of creditworthiness activated, and the relevant
information transmitted for evaluation at the higher levels of the banking structure. In some
instances the decision whether or not to grant a loan and its terms and conditions is taken in
complete autonomy by the branch manager.
The literature on corporate organization acknowledges that it is possible for the objectives of
branch managers to diverge from those of the ultimate control holders. It also emphasizes how
specialization in the gathering of data and information asymmetries between the headquarters and
the branch network within a complex organization can generate the need for a two-way transfer of
information along the hierarchy. The literature on banking shows how the many distinctive
elements featuring internal organization (the organization chart, the extent to which decision-
making is centralized or decentralized, internal control systems, procedures for communicating
between the various organizational levels, etc.) have a decisive influence on the strategies of the
branch managers and, through them, on the allocation of credit to small firms.
The central role played, de facto, by branch managers in lending to SMEs has not been given
sufficient attention by the literature, in part owing to a lack of data. Some recent contributions have
begun to fill this gap. Liberti (2005) and Liberti and Mian (2006) show that the higher up one goes
in the hierarchy of the bank organization, and the more distant the customer is from the decision-
making centre, the less qualitative elements will weigh on loan decisions. This shows how
qualitative information is in fact collected and stored at the lower hierarchical levels and how its
transmission costs increase with geographical and organizational distance (defined as the number
of levels involved in the decision-making process).
Stein (2002), and Berger et al. (2005) broaden the debate to include the effects that different
organizational models can have on the incentives for banks branch managers. In particular, Stein
uses a theoretical model to show how a large bank can discourage the acquisition of soft
information by a branch manager where data must be communicated along multiple hierarchical
layers and its transmission becomes extremely costly. This effect does not occur in the case of
small banks, where there is markedly less physical and hierarchical distance between the
headquarters and the branch network. As a result, the branch managers of major banks may have a
greater incentive to collect hard data, which are more easily provided by large firms, while smaller
intermediaries can specialize in the acquisition of qualitative information and in small business
lending.
For Japan, Uchida, Udell and Yamori (2006) show how branch managers traits are not
important for the purposes of accumulating soft information and explain this result with reference
to the likelihood that the strong social cohesion of Japanese society lowers the costs of transmitting
qualitative information.
In an empirical analysis of Italian data, Ferri (1997) shows the positive correlation between
branch manager mobility and bank size, and explains this result by reference to the greater
difficulties large banks face in limiting the moral hazard stemming from the potential for collusion
between branch managers and bank customers. The large physical and organizational distance
between the headquarters and branches of major banks increases the costs of monitoring, driving
these intermediaries to use the mobility of banking managers as a tool for limiting their
opportunities to reap private benefits. By contrast, the geographical and organizational proximity

typical of small banks encourages them to maximize the benefits of the stability of local managers,
while at the same time maintaining monitoring costs at reasonably low levels, given that the top
managers may belong to the same local community as the branch manager. Moreover, if a large
bank is specialized in lending to medium-sized and large firms capable of providing hard data
(Stein, 2002), it follows that these banks will be less motivated to keep the local managers in the
same branch for a long time. In these cases, mobility can represent a way of furthering local
managers careers and of warding off excessive inertia in the administration of local branches.
Hertzberg, Liberti and Paravisini (2007) use data on the turnover of the local bank managers of an
Argentinean bank and show that the bank utilizes mobility to persuade these managers to report
information on the creditworthiness of borrowers accurately. Scott (2006) shows that for a sample
of small U.S. firms the turnover of local managers increases the likelihood of credit rationing.
The literature just surveyed brings out four key themes relating to the characteristics of
branch managers activities:
A1) Hierarchical and geographical distance between bank headquarters and branch
managers. As we have seen, distance can affect the costs for transmitting qualitative informations
collected at local level and accordingly the incentive to acquire them. It can also determine the cost
of monitoring branches at central level.
A2) Decision-making autonomy of the branch manager. This variable undoubtedly enhances
the incentives for a branch manager to acquire soft information, but at the same time it increases
the costs of control and coordination at central level.
A3) Tenure of the branch manager. Also in this case a trade-off can be generated similar to
the one described earlier: the more stable a branch managers position is, the more incentives there
may be to acquire soft information, but the costs of control can also increase (due, for example, to
moral hazard).
A4) Incentives. Economic incentives for branch managers can help reduce the problem of
moral hazard by aligning the objectives of peripheral agents with the bank central management,
with the danger, however, of transferring excessive risk to the branch manager.
The role played by technological innovation in the development of the procedures for
granting loans adopted by banks from the 1990s onwards was recalled earlier. One of the most
important consequences of ITC advances in the banking industry was the sharp reduction in data
processing costs, in other words in the use of data for administrative purposes at the various
organizational levels. The new regulations on minimum capital requirements also provided strong
incentives for the adoption of statistical techniques for measuring credit risk: the methods vary but
their distinctive feature is the ability to group customers within a finite number of categories,
associating to each one a synthetic indicator that expresses the probability of default and
accordingly the degree of risk. The introduction of these techniques can influence the role of
branch managers in the allocation of credit in various ways. Indeed, credit scoring can represent an
alternative means of assessing creditworthiness to the decision-making processes that emphasize
qualitative information and the close interaction of branch managers with customers. At the same
time, the adoption of scoring techniques allows, at least partly, the transformation of soft
information into processed data and facilitates the control of branches. These issues, which are
closely interrelated, cannot be dealt with without further exploring the specific nature of credit
scoring, including its relatively recent introduction into the Italian banking system. From this
perspective the analysis of credit scoring complements that of organizational variables and central-
peripheral relations.
Quantitative methods include both credit scoring models, which distinguish cases of
expected default from non-default using statistical techniques such as discriminant analysis or logit

and probit, and internal rating systems, which map more or less automatically individual borrowers
(or, in the most sophisticated cases, the different lines of credit of each borrower) on a scale of
judgments (Allen, DeLong and Saunders, 2004). These technologies can allow, even before the
formulation of any final judgment, discretionary interventions by one or more persons to assess
qualitative elements not explicitly considered in the model. The degree of flexibility in the use of
credit scoring therefore varies, depending both on the characteristics of the procedures adopted and
on their importance in lending decisions and customer relationships. In this report the term scoring
refers to all these instruments indiscriminately. What follows is a brief survey of the main questions
related to the adoption and use of credit scoring techniques, which will be described more fully in
section 5.
B1) Banks characteristics and the adoption of credit scoring techniques. In Europe the
introduction of credit scoring was later and more gradual than in the United States (Degryse and
Ongena, 2004). Recently, however, credit scoring has been widely adopted by Italian banks. The
literature concurs in emphasizing how the adoption of credit scoring techniques is influenced both
by the size of banks and by their organization. Size acts in the usual ways: bigger banks have more
resources to invest in new techniques and the cost of investments is then distributed among a
broader loan portfolio. Moreover, large banks report greater diseconomies owing to distance:
difficulties in transmitting soft infromation internally, in selecting and monitoring loans, and in
designing the right incentives for local managers. The adoption of credit scoring techniques lowers
the costs of screening and monitoring firm activity and of controlling branch managers, mitigating
the problems of monitoring from a distance (Berger, Frame and Miller, 2005).
B2) The characteristics of the scoring techniques. The scoring techniques can differ both in
their origin (developed internally or acquired externally) and in the data set processed. The use of
methods developed internally by banks implies greater control and more flexibility, in the sense
that they are easier to modify if the bank no longer considers them adequate. Moreover, broad
recourse by banks to externally developed techniques could lead to greater homogeneity in the
criteria for assessing creditworthiness by the various intermediaries.
These techniques mostly rely on quantitative data inputs to formulate scores. As mentioned
earlier, there is also concern over the possible decline in the importance of soft information in
lending decisions, with the result that start-ups and smaller firms, relying more on bank credit,
could be adversely affected by an intense and rigid use of these methods (Cowan and Cowan,
2006). However, the empirical evidence to date, above all concerning the United States, does not
appear to justify these fears (Berger, Frame and Miller, 2005).
B3) The importance of scoring techniques and how they are used. Even though scoring
techniques have by now been widely adopted by banks, their importance in assessing customers
creditworthiness is not a foregone conclusion. This issue is by no means secondary, since their
impact on the extension of credit varies depending on whether banks use them as the main
instrument for assessing creditworthiness or as a supplementary instrument along with other
evaluation techniques (Berger, Frame and Miller, 2005). Based on the results of a recent survey
conducted on a sample of U.S. banks, the scores for small firms are considered less important than
the traditional indicators of creditworthiness, such as cash flow and available collateral (Cowan and
Cowan, 2006).
In section 3 the distinction between intermediaries is based on size and institutional set-up.
Given that information asymmetries, agency problems, data transmission costs and the economies
of scale linked to the use of IT become more complex when we move from individual banks to
groups, the survey also distinguishes between stand-alone banks and members of a group.


3. The survey
Our data are taken from a qualitative questionnaire, reproduced in the appendix, aimed at
gathering information on the organizational aspects and scoring techniques used in the lending
process.
The questions capture banks organizational structure, meaning both their specialization by
customer segment and the number of hierarchical layers involved in the decision to lend to a firm.
For each hierarchical level, the survey establishes the degree of autonomy granted to approve loan
applications and, possibly, to establish their terms and conditions. Finally, the existence and type of
economic incentives for branch managers and the length of their tenure are also considered. In the
second part of the questionnaire, the questions explore the adoption of statistical-quantitative
techniques for evaluating firms, their use in setting terms and conditions of loans, as well as in the
monitoring of the loans. Next, the characteristics of the models are surveyed, in particular the data
used to calculate the scores and the role of quantitative and qualitative information in evaluating
new loan applicants.
The questionnaire was submitted to intermediaries during 2007 through the Bank of Italys
branch network. The selection of the banks was aimed at ensuring adequate coverage, both
geographically and by type of bank (medium-sized and large banks, small stand-alone banks and
others belonging to groups, mutual banks). The sample design was based on that used for a data
base including interest rates (TAXIA), which until the end of 2006 surveyed 215 reporting
institutions, selected by size (measured by total lending to customers), geographical location and
share of loans reported to the Central Credit Register. The original 215 banks sample was modified
in tow ways. First, intermediaries specialized in activities such as factoring, leasing, household
mortgages and credit recovery were excluded, given that the questionnaire focused on lending to
firms in traditional technical forms. Second, it was left up to local data collectors to submit the
questionnaire to banks excluded from the TAXIA sample, provided they did not belong to the
minor size category.
The final sample included 333 banks and 322 responses. The missing answers all refer to
banks whose business lending was marginal or nil. The composition of the sample by geographical
area and size is shown in Table 1. The sample accounts for 82.8 per cent of the total amount of
outstanding loans to non financial firms in 2007 (Figure 1). Coverage is high throughout the
country, ranging from 81.0 per cent in the regions of the Northwest to 90.4 per cent in the South;
among the size/institutional classes this percentage contracts for small stand-alone banks (34.7 per
cent), a category that includes several branches of foreign banks which were not surveyed.
Table 2 shows the mean and median size of the four classes used in the report. It appears
significant that the average size of the stand-alone banks is around half that of those belonging to
the same class size but part of a group.
The questionnaire contains mostly qualitative questions, and accordingly the data have been
aggregated in two separate ways: either as simple frequencies of responses or as weighted
frequencies where the weights are equal to loans SMEs or large firms, depending on which type of
borrower the question refers to. The significance assigned in the first case is that of the diffusion of
the phenomena (for example, what share of the banks use credit scoring), and in the second case
that of the likelihood of the borrowers encountering a given phenomenon (using the same
example, the share of lending to SMEs or large firms granted by banks using scoring techniques).

Figure 1
Coverage of the sample
(percentages of total lending to firms)
0 10 20 30 40 50 60 70 80 90 100
Italy
South
Centre
Northeast
Northwest
Mutual banks
Small stand-alone banks
Small group-member banks
Medium-sized and large banks

4. Banks organization, SMEs lending, and the role of the branch manager
4.1 Distance from headquarters and banking presence in local credit markets
Some recent literature has focused on the effects that the distance between bank and firm can
have on access to credit and its terms and conditions. This literature often treats the bank as a single
entity, but if instead, as noted above, we recognize that the bank has an articulated internal
structure, the question can be approached from two different standpoints, one referring to the
distance between the borrower firm and the unit within the bank charged of lending decisions, the
other to the distance between that unit and the head office, where the persons who exercise control
powers are located. This is illustrated below.
This relationship between centre and branch may be afflicted by agency problems, just as
severe as those traditionally considered in bank-firm relations. Apart from the costs of monitoring
the branch managers activity, distance can also increase the costs of transmitting qualitative
information, lessen the incentives for the collection of this information, as well as enlarge the effort
required to pass on best practices.
2
When instead of geographical distance there is hierarchical
distance measured by the number of organizational levels along the chain of command involved
in the decision the effects are similar.

2
Berger and De Young (2001) have shown that a bank holding companys ability to transfer its own efficiency to the other banks in a
group diminishes as the distance between the former and latter increases. Deng and Elyasiani (2005) have found that the risk of the
holding company increases with the distance from its subsidiaries and interpret this in terms of greater ability of the local manager
to engage in opportunistic behaviour when he operates in far-away markets.

In what follows we describe several indicators of the distance between centre and periphery
for the banks in our sample, the number of local markets in which they do business, and how these
factors evolved between 2000 and 2006.
To compute distance, first we must define the relevant geographical units and devise a
suitable gauge. To this end, we define the location of the persons who ultimately control the bank
as the city in which the banks legal head office is established and that of the branch manager as the
locality of the branch, itself defined as the main city within each of the local labour market area
mapped by Istat in 2001. This procedure successfully balances the need for a geographical
classification detailed enough to allow for precise measurement of the distance between head office
and branch with the need to capture the differences between the market areas where the head office
and the branches are located.
3

Once the local labour market area where the banks legal office is located is determined, we
calculate the distance between that and all the other local labour markets where the bank has at
least one branch.
4
The distances so obtained are weighted according to the local labour markets
share in the banks total lending. Weighting is necessary to prevent branches in far-off areas but not
accounting for a significant portion of the banks assets from having a disproportionate effect on
the average distance.
The results are given in Table 3. The mean for this indicator in 2006 is 47 kilometres, the
median 21 kilometres. The larger banks have a more extensive branch network and accordingly

3
Some of the foregoing observations on the relationship between head office and local branches and the related agency problems
might suggest adopting a continuous space and defining distance simply as the distance in kilometres between the head office and
the branch. However, other aspects of the relationship are better defined by reference to the distance between the local markets of
the two units. The first method would give a better gauge of the physical distance; the second highlights the distance, hence the
differences, between the broader areas within which the two units are located.
4
The local labour systems are identified by the geographical coordinates of their main municipality; the distance is calculated by the
great circle formula, assuming the earth is spherical. Another assumption concerns the distance to branches located within the
same local labour system as the head office. This could be set equal to zero or else calculated on the basis of the land area of that
system. The results do not change when the distance is set to zero.
Legal head office of the bank
Branch
manager
Borrower
firm

higher average distances than medium-sized, small and mutual banks. Statistical tests on these
averages indicate that the differences between classes of bank are significantly different from zero.
There are also significant inter-category differences in the number of local labour systems in which
the bank has at least one branch (Table 4).
The average distance is highly variable. The inter-quartile difference is equal to 132 per cent
of the median for all banks and is also high even within each of the size categories.
Between 2000 and 2006 there was a broad increase in the distance between the local labour
system of the head office and that of the branch. The median rose for all categories of bank (except
for mutual banks, for which it held constant), but the mean diminished slightly among large banks,
while increasing for small and mutual banks. Statistical tests on the averages indicate that the
differences are significantly different from zero for small and mutual banks but not for the large
banks. The average number of local labour market areas in which banks had branches also
increased over the period. In general the increase in distance mainly involved medium-sized banks,
especially for banks belonging to groups, and also, though to a lesser extent, small non-group banks
and mutual banks. In part, this is because some large banks involved in mergers reduced the
number of branches and of local labour systems in which they were present, shortening their range
of operations. Even excluding these banks, however, the increase of the average distance for the
large banks is small and less pronounced than that recorded for the other size categories.
It is also interesting to examine the evolution of headquarters-branch distances not for single
banks but for banking groups, considering all the local labour market areas with at least one branch
and computing the distance from the parent banks head office. For this we exclude groups that
have only one bank and any branches or subsidiaries abroad. For the 33 banking groups so
identified, the mean distance rises from 180 to 217 kilometres, the median from 131 to 241
kilometres, and the average number of local labour markets from 107 to 142.
The foregoing suggests several considerations. First, the variability in banks size is reflected
in huge differences in the extent of branch networks and in distance between centre and peripheral
branches. But even within the various size and institutional classes the differences remain, which
means that the problem of the cost of transmission of qualitative information, and agency problems
in general, between branch managers and head offices differ in intensity from bank to bank, as do
the organizational structures adopted to deal with them (see below).
Second, between 2000 and 2006 there was a general lengthening of distances, owing in part
to the introduction of information and communications technology, which greatly reduced the cost
of acquiring, processing and transmitting information (Berger, 2003).
5

Third, the increase in distance was greater among small banks belonging to groups and to
small non-group banks, and also to some extent among mutual banks. Large banks held the average
distance of branches from the head office roughly unchanged; their overall range of action was
increased via their group structure. This jibes with some recent works showing that small and
minor banks have grown more than large banks (Bonaccorsi di Patti, Eramo and Gobbi, 2005).

4.2 Organizational structure and layers of hierarchy
The previous section considered the distance between head office and branches in
geographical terms. Now we consider it in terms of organizational structure, measuring the
distance between the branch manager and the top management of the bank by the organizational

5
Felici and Pagnini (2008) show that the banks with larger ICT endowments, other things (including profitability) being equal,
increase their capacity to move into markets far away from their existing localities.

model adopted and the number of layers of management between them. The model may have a
divisional structure, in which customers are segmented by size, or a single structure that performs
all lending activities. In a divisional structure, the middle management and/or the type of peripheral
unit varies with the size of the borrower firm; along with the modification of organizational roles,
the delegation of powers and the way in which lending is planned and carried out also change. The
number of layers of hierarchy is used to gauge the depth of the organizational structure.
The model of organization. The responses to the questionnaire indicate that over 70 per
cent of the large banks are organized by divisions, with customers segmented by size, typically
divided into SMEs and large firms. For small banks belonging to groups this percentage falls to 33
per cent, and for small stand-alone banks to 24 per cent; for mutual banks it is 10 per cent. As a
rule the variable used for segmenting firms is sales, although in a significant number of cases small
banks use loan size. The modal sales threshold dividing small from large firms is 2.5 million.
Among small banks, organization by divisions is a very recent phenomenon; almost 50 per cent of
the small and mutual banks with this model adopted it in the last three years, compared with only
about 10 per cent of the large banks.
Typically, large banks have greater differentiation of products, markets, and customers.
Accordingly, the divisional organizational can exploit specialization, by assigning a single manager
responsible for a given product or geographical area. Division structure also makes it easier to
tailor adaptation to the industrial or local situation. And the large banks, thanks to potential scale
economies, can more readily sustain the costs of divisional organization, which entails more staff
and structures engaged in similar activities. Among other potential costs there is the need for closer
coordination of relatively autonomous, diversified units.
Layers of hierarchy. The length of the chain of command is given by the number of ranks
between the branch manager and the CEO.
6
These ranks are hierarchically ordered, and each has
specific powers in terms of maximum loan amounts. In what follows we consider the positions
involved in lending to SMEs.
7
The average number of layers varies significantly by institutional
category: 5 for large banks, 4 for small group-member banks, 3.6 for small stand-alone banks and
2.8 for mutual banks. Figure 2 shows the distribution of layers of hierarchy according to this
classification.
As bank size increases, not only does the average number of hierarchical layers increase but
so does the variance of their distribution. The large banks are those with the greatest diversity in the
length of the chain of command. One of five has a simplified organization (with fewer than three
layers), while one in three has a highly complex structure (more than six layers). For mutual banks,
more than 40 per cent have an elementary structure with just two levels: branch manager and CEO.
However, even in this class there are more complex organizations; one in five has four or more
layers.
The number of layers of hierarchy has a significant impact on a banks operation. On the one
hand more layers may mean higher costs for the transmission of information from one level to
another and longer decision times. On the other, a flatter organizational chart, for a given size of
staff, implies larger units, hence a larger area to control in other words, a larger number of
subordinates under each supervisor.

6
We consider only the ranks that correspond to a significant rise in hierarchical level. This excludes, for instance, deputies and
auxiliary staff but includes all the grades assigned to run an organizational unit.
7
We do not comment on the data on layers of hierarchy involved in lending to large firms or the degree of decision-making
autonomy enjoyed by branch managers in such lending. This decision follows from the limited role of branch managers in this
segment. Moreover, only a minority of banks have different organizational structures for lending to SMEs and to large firms.

Figure 2
Distribution of number of layers in hierarchy (1)
(numbers and percentages)
0
5
10
15
20
25
30
35
40
45
50
2 3 4 5 6 7 8 10
Medium-sized/large banks Small group-member banks
Small stand-alone banks Mutual banks

Source: Sample survey of 322 banks.
(1) The number of layers is the number of hierarchically ranked positions from branch manager through chief executive officer. The
horizontal axis gives the number of layers, the vertical axis the percentage shares of the number of layers for each class of institution.

4.3 Branch managers decision power
The branch manager obviously has a privileged position for acquiring the information
relevant to loan decisions. However, the formal authority the person holding the right of control
belongs to the top management of the bank. The organizational problems posed by the lack of
coincidence of these two figures can be coped with either by transferring the information to the
person endowed with the formal authority (central decision-making) or by assigning the power to
decide to the person who has obtained the information (decentralized decision-making).
8
One
solution creates problems of information transmission, the other problems of control.
9
Between
these two extremes there is a continuum of degrees of decentralization to resolve the trade-off.
The literature on corporate organization has dealt with the problem of measuring the degree
of decentralization of the power of decision from various standpoints.
10
In this survey the banks

8
See Christie et al. (2003). The distinction between real and formal authority is from Aghion and Tirole (1997).
9
If the power of decision were totally centralized, in the hands of top management, this could lead to organizational failures as a
result of the branch managers lack of incentive to acquire information. Also, the transmission of information from lower to higher
levels may entail a loss of information, or at any rate a lag between information acquisition and decision. Finally, if agents are
rationally constrained la Simon, the banks top management might not be capable of handling a large information flow.
Decentralization, on the other hand, while allowing the power of decision to be in the hands of the person with the information,
means a loss of top management control over that persons choices. The costs of delegating formal authority are defined as agency
costs and depend on the fact that the aims pursued by the bank as such do not necessarily coincide with the personal objectives of
the staff. Agency problems typically involve collusion between the branch manager and the borrower firm or the manipulation of
the data she has gathered.
10
Christie et al. (2003) and Acemoglu et al. (2006) identify autonomy of decision with the presence of profit centres within the
corporation. This indicator reflects the observation that a cost centre controls either revenue or costs but not both, while a profit
(continues)

were asked to indicate, for every hierarchical level, the maximum amount of credit that it could
grant on its own authority. This information was used to construct an indicator of the degree of
decentralization. This section too refers to lending to SMEs.
11

The amount of credit that the branch manager can grant on his own power increases with the
size of the bank (Table 5). The mean is 550,000 (and the median 250,000) for large banks,
200,000 for small banks belonging to groups, 90,000 for small stand-alone banks and 50,000
for mutual banks. However, these means conceal some variability within the subgroups, as is
shown by the interquartile difference as a ratio to the median. The greatest variability is found
among mutual banks; in one fifth of them the value is 0, and in one fifth it is more than 100,000.
The amount of credit that the branch manager can extend autonomously is larger for mortgage
loans (124,000; Table 6) and smaller for uncollateralized overdraft facilities (62,000) or
unsecured loans (54,000). These differences reflect the role of collateral.
Comparing the loan authorization power of the branch manager with that of the top
management, we can construct an indicator to measure the degree of decentralization. The branch
manager (or head of the local unit of the bank) and the CEO are the two figures that appear in
virtually all the organization charts. Comparing the powers delegated to them, we can build an
index of the branch managers autonomy with respect to the powers of the CEO. This index equals
5 per cent in large banks, 11 per cent in small banks belonging to groups, 14 per cent in small
stand-alone banks and nearly 20 per cent in mutual banks (Table 5). The index is negatively
correlated with bank size because as the bank becomes larger the powers of the CEO increase more
than proportionately with respect to those of the branch manager. The CEO himself performs
different functions depending on type of bank. In mutual banks, for example, the most important
decisions on loans to SMEs are taken directly by the banks board or council, whereas in large
banks they are taken lower down in the chain of command. Rather than compare such drastically
different types of bank, again it is more meaningful to observe the variability within each category.
And again the ratio of the interquartile difference to the median shows great variability.
The evidence presented so far refers to the year 2006. The questionnaire also asked about the
trend in the past three years. Half the banks reported a tendency to greater decentralization, while
just 4 per cent declared that they had centralized their decision-making powers (Table 7). This
tendency characterized all the banks but was most pronounced for the large ones (about three
quarters of which reported greater decentralization).
Several conclusions follow. First, the size of the bank is a major determinant of the
delegation of powers, in absolute value, to the branch manager. Some large banks, which tend to be
farther away from the local credit markets and the borrower firms, could endow their branches with
considerable autonomy with a view to creating streamlined structures, relatively immune to the
inertia and lack of dynamism typical of large corporations, and closer to the local community. The
amount of lending authority may also be affected by the type of customer served. Banks with larger
customers tend to have higher ceilings on their local units lending powers.
Second, decentralization is more pronounced in small than in large banks. Greater
geographical proximity of local to central offices and less complex lending business may foster the
decentralization of powers of decision, thanks among other things to greater ease of control of top
management over local managers. Yet by itself bank size does not explain the variability in the
degree of decentralization. Within each of our size classes, the observed variance would appear to
indicate significant variety in organizational arrangements.

centre takes decisions involving both costs and revenue. The degree of decentralization is thus 0 when the level immediately below
the CEO is a cost centre and 1 when it is a profit centre.
11
The questionnaire also had a question on the autonomy of branch managers to price loans, but owing to the number of non-
responses we have elected not to comment on these data.

The responses further reveal a general tendency towards decentralization. Together with the
recent adoption of divisional models, this highlights a certain organizational dynamism,
presumably in connection with increasing geographical distance between head office and branches
and with the diffusion of ICT. Finally, the greater decentralization among large banks may be a
response to the competition of smaller institutions in lending to SMEs in local credit markets.

4.4 Branch managers tenure
The tenure of branch managers faces the bank with a trade-off. Greater stability in the
position facilitates contacts and familiarity with the local market, hence the acquisition of soft
information. But it also heightens the informational asymmetry between branch manager and head
office, possibly enabling the former to reap private benefits (by collusion with local borrowers, say,
or manipulation of the information that he transmits up the chain of command).
The survey found that the mean time for which branch managers held their position was
nearly 4 years, against a median of 38 months and a mode of 36 months (Table 8). These figures
are similar to those found in a similar survey by Ferri (1997). The mode (3 years) could depend on
corporate routines and widespread organizational models shared by many banks. The term of office
is shorter in the larger banks, longer in small stand-alone banks and mutual banks. The standard
statistical tests show that all these differences are significantly different from zero. Small banks
belonging to banking groups have values similar to large banks, suggesting that group
organizational strategies extend to smaller intermediaries. The degree of mobility of branch
managers shows considerable variability not only for the entire sample but also within each
category of banks.
As to trends in tenure, nearly 40 per cent of the sample banks reported that it had shortened
in the last three years, while 14 per cent reported a lengthening. The tendency towards greater
mobility was broadly uniform also within bank classes. The heightened mobility of branch
managers may be related to the introduction of ICT, which in practice may have reduced the rents
deriving from close local bank-firm relations; to the increase in head-branch distance, which in turn
presumably increases the costs of monitoring local managers activities; to mergers and corporate
restructuring, which have involved a large number of banks since the 1990s; and to heightened
competition in credit markets, leading in turn to stiffer competition in the local bank manager
markets.

4.5 Branch managers incentives
The potential costs connected with the distance between head office and local units and with
the misalignment between the personal objectives of local agents and those of the bank as such can
be mitigated by incentive systems linking agents compensation to results (Milgrom and Roberts,
1992). However, incentive systems can also entail excessive transfer of risk to the local agent.
Incentives are most common in the bigger banks (Figure 3), with 83 per cent of the large
banks in our sample stating that the use of incentives in connection with lending for branch
managers compensation is very important compared with an overall sample mean of 57 per cent.
Mutual banks make least use of incentives (46 per cent). This may be explained by the higher
incidence of agency costs in larger banks. That is, the greater the geographical and organizational
distance between centre and periphery, the less ability there is to monitor local agents activity. So
incentives should help to align branch managers goals with the banks objectives.
As to the factors on which incentives are based, the most common response was the overall
profitability of the local unit (e.g., gross income). This factor was especially important for large

banks and small group-member banks (Table 9). Practically nine-tenths of all the banks using
incentives said overall profitability of the branch was a very significant factor. By contrast, small
stand-alone banks and mutual banks are much more sensitive to the bad loan ratio or variations in
bad loans. To simplify, larger banks and those belonging to groups tend to link local managers
incentives to the profitability of the branch and of its loan portfolio, while other banks, including
mutual banks, tend to stress the containment of bad loans and limitation of credit risk.

Figure 3
Use of economic incentives for branch managers (1)
0
10
20
30
40
50
60
70
80
90
Medium-sized/lar ge
banks
Small gr oup-member
banks
Small st and-alone banks Mut ual banks

(1) Percentage of banks that describe as very important at least one of the factors shown in Table 9 for the compensation of branch
managers. The horizontal line represents the overall sample mean.

In the last three years the banks have made greater use of incentive schemes for branch
managers compensation (Table 10). The factors whose relative importance has increased have
been gains in the overall profitability of the branch and the containment of bad loans. The share
reporting an increase in the relevance attached to these factors was more than 40 per cent, against
fewer than 5 per cent reporting a reduction in their importnace. Again, the mutual banks were those
paying the closest attention to the incidence and variation of bad loans and substandard loan assets.

5. The importance of credit scoring techniques in assessing creditworthiness
5.1 The diffusion of credit scoring techniques
The plunging cost of data processing in recent years has fostered banks to introduce
statistical techniques for measuring credit risk, supplementing their external and internal sources of
information. Our survey of the Italian banking system shows the diffusion of credit scoring
techniques for business lending. Consistent with our hypotheses in the introductory section, and as

other studies have shown (Bofondi and Lotti, 2005), these techniques have been mainly adopted by
the larger banks with extensive branch networks, which can exploit economies of scale.

Figure 4
Diffusion of credit scoring
(percentages; unweighted frequencies)

46,9
83,8
42,2
59,4
64,1
97,3
56,5
0 20 40 60 80 100 120
Mutual banks
Medium-sized and large
banks
SMEs
Large corporates


At the end of 2006, 57 per cent of the sample banks had scoring techniques in place to assess
the creditworthiness of firms, whether large or small (Table 12). But the distribution was not
uniform by type of intermediary: diffusion reached 97 per cent for medium-sized and large banks,
66 per cent for small banks belonging to groups, 59 per cent for small stand-alone banks, and just
over 40 per cent for mutual banks (Figure 4). Moreover, scoring techniques were systematically
more common for lending to SMEs. For large banks the difference was more than 13 percentage
points (Table 12 and Figure 4). That is, the prevailing tendency is to use these techniques to reduce
selection costs for smaller firms, even where the bank is less specialized in this segment, thus
freeing resources for screening in the core business segment.
Italian banks began to adopt the scoring techniques substantially in 2003, and their
introduction has been by degrees, accelerating sharply in recent years. In 2000 under 10 per cent of
the banks had such techniques in place, in 2003 almost 25 per cent, and in 2006 more than half
(Figure 5 and Table 12).

Presumably the spread of scoring techniques in recent years is related to the new capital
adequacy accords (Basel II), which links capital requirements more directly to customer risk, with
incentives for more accurate evaluation of the quality of the loan portfolio.
12

Figure 5
Introduction of credit scoring
(percentages; unweighted frequencies)

0
10
20
30
40
50
60
70
80
90
100
2000 2001 2002 2003 2004 2005 2006
Mutual banks
SMEs
0
10
20
30
40
50
60
70
80
90
100
2000 2001 2002 2003 2004 2005 2006
Large corporates


The credit scoring techniques adopted by Italian banks differ both in origin
(internal/external) and in the data used. At the end of 2006 more than 50 per cent of the banks that
had introduced a credit scoring methodology had participated actively in its development, either
alone or in cooperation with other institutions or consortia (Table 13). The degree of participation
is correlated with bank size. Nearly all the larger banks had contributed actively to the development
of the scoring method, small banks less commonly.

5.2 The data used for credit scoring
One of benefits deriving from credit scoring involves the management of the data available
to banks. Banks can now fully exploit this information, integrating and combining data for
systematic, replicable use. But accurate data extending over a suitably long period are
indispensable to the reliability of the models forecasts. The new techniques also impose a
standardization of the documentation required for loan applications, which among other things
facilitates subsequent securitization.
Adapting internal information systems originally designed for different purposes is one of
the most serious problems and has slowed the introduction of the new techniques.
13
Consequently
the models focus on the factors that have traditionally been used to assess creditworthiness (firms

12
After classifying their retail customers by risk under the internal ratings approach, banks must estimate the main risk components
for each class and then calculate, by the Basel Committees methodology, specific capital requirements for each. Thus the
introduction of credit scoring may be highly advantageous to the banks, insofar as it can lower capital charges (Jankowitsch et al.,
2007). So far very few banks, and only large ones, have begun the procedures towards the adoption of these methodologies for
calculating capital requirements. To do so, in fact, banks must meet stringent qualitative and quantitative requirements, subject to a
complex process of supervisory validation. But the possibility that the internal models may be validated and recognized for
supervisory purposes has nevertheless fostered their diffusion (Bank for International Settlements, 2005) both by stimulating studies
on the methodologies by specialized companies and by creating incentives for initiatives by consortiums of institutions.
13
The numerous interbank mergers of recent years have also complicated the integration of data from different sources.

financial statements and credit history), while other data both from external sources and available
within the banking group are used less frequently (Table 14). The question on this matter was
phrased in ordinal terms, asking respondents to rank the various sources of information by
importance (see the appendix). Figure 6 reconstructs the ranking based on the frequency of the
answers very important and decisive.
There are significant differences in the weights of the factors used in credit scoring for
SMEs. For mutual banks (and for the small banks), the most important factor is the financial
statement, followed by the credit history with the bank and with the rest of the banking system. The
larger banks, by contrast, assign greater importance to the firms past credit performance than to its
accounting data (Figure 6).
Less importance attaches to the firms economic sector and geographical area, which in fact
are not even considered in many cases (about a third of the models, accounting for 18 per cent of
loans). Other external data sources, including the interbank register of bad cheques and payment
cards and the Chamber of Commerce database, are of little importance and are often not used at all.
Large banks and mutual banks are the ones that most commonly ignore them (31 and 26 per cent
respectively, corresponding to 38 and 19 per cent of loans). Qualitative information is generally
included in the estimation models, although with relatively modest weight in the banks overall
assessment. Finally, large banks also consider any relations between the firm and other members of
the banks group; even so, about 40 per cent of the large banks do not include this information in
their models.
Figure 6
Information sources included in scoring systems
(ranking of information sources by importance)

Medium-sized and large banks Mutual banks
1
st
State of credit relations with banking system Income statement and balance sheet
2
nd
State of relations with bank State of credit relations with banking system
3
rd
Income statement and balance sheet State of relations with bank
4
th
Other outside sources Other outside sources
5
th
Qualitative information Qualitative information
6
th
Relations with banking group Area and sector of activity


In evaluating the credit worthiness of large firms, the importance of the various sources of
information used by the larger banks is similar to that used for small firms, but there are some
differences in their relative ranking. In the former case, greater importance attaches to the company
accounts and especially to qualitative information, generally relating to organizational structure, the
stature of the management, and the quality of the investment project to be financed (Table 14).
Overall, the models appear more flexible, as they allow for more judgmental components. With
respect to large firms, more attention is paid to the state of the borrowers existing relations with

the other members of the banks group, although a substantial 35 per cent of large banks do not
consider this factor.

5.3 The importance of credit scoring techniques in assessing creditworthiness
Statistical scoring techniques have gained considerable importance in the lending process, in
particular in the decision whether or not to grant a loan. In most cases the score is decisive or very
important (Table 15 and Figure 7). There are significant differences between banking groups, and
the relative importance of quantitative techniques is definitely greater among the larger banks and
decreases with the size of the institutions. Generally, the rating/scoring influences the size of the
loan, and for the smaller firms it also affects the amount of collateral requested.

Figure 7
Importance of scoring in lending to SMEs
(frequency of decisive and very important answers, weighted by lending to SMEs)

0
10
20
30
40
50
60
70
80
90
100
Decision Amount Pricing Maturity Collateral Monitoring
Mutual banks


Though scoring techniques are widely used, they are still rarely employed to determine
interest rates and loan maturities. This lack of an immediate impact of the borrowers credit score
on the interest rate the bank charges would appear to indicate that the effect found by the literature
in other countries, namely an expansion of the volume of lending to marginal customers at higher
interest rates, has not (yet) emerged in Italy. But, consistently with developments in risk
management and control, these methodologies have been widely used to monitor the situation of
firms and the state of loans and accounts. The differences in practices concerning small and large

firms do not appear pronounced, though there is a somewhat greater tendency to use ratings to
determine the pricing of loans to the larger firms (Table 15).
The banks flexibility in using scoring techniques is highly variable. It depends, on the one
side, on the characteristics of the procedures chosen, which may provide for adapting scores to take
account of information elements not included into the model, and, on the other side, on their actual
importance in lending decisions and management, according to whether they are the main
evaluation tool or a supplement to another method of assessment. The models degree of flexibility
was gauged with reference to decisive as one of the possible responses on the importance of
scoring methods. Decisive was counted separately from very important precisely in order to
capture the possibility for loan officers to derogate from the automatic scores; the answer
decisive was interpreted to mean practically no such flexibility. In all cases, the scores were more
binding for large than for small banks.
For decisions on whether or not to lend to SMEs, credit scoring tools, while important, are
decisive for only 18 per cent of the entire sample of banks and for a third of the large banks;
weighted by volume of loans, the frequencies are higher (Figure 8), confirming that the
discretionary power of loan officers tends to diminish as the size of the bank (and of the borrower
firm) increases. Further, greater specialization in lending to SMEs corresponds to lower weight
assigned to scoring in loan decisions.
Figure 8
Flexibility in use of scoring techniques
(percentage share of answers decisive and very important in loan approvals to SMEs,
weighted)

0
10
20
30
40
50
60
70
80
90
100
D
e
c
is
io
n
A
m
o
u
n
t
M
o
n
it
o
r
in
g
D
e
c
is
io
n
A
m
o
u
n
t
M
o
n
it
o
r
in
g
D
e
c
is
io
n
A
m
o
u
n
t
M
o
n
it
o
r
in
g
D
e
c
is
io
n
A
m
o
u
n
t
M
o
n
it
o
r
in
g
very important
decisive
Medium-sized
and large banks
Small group-member
banks
Mutual banks
Small stand-alone
banks


As we have seen, the likelihood of a banks developing its own scoring system internally, at
least in part, increases with bank size. The purchase of a statistical procedure from outside could

reduce the banks exercise of control over its own instrument for selecting borrowers, fostering the
perception of the system as a black box, both initially in relation to the algorithm and database
and subsequently at times of revision (Frame, Srinivasan and Woosley, 2001). Our survey shows
that in each size class of bank, credit scoring techniques for SMEs are more frequently decisive or
very important in lending decisions when they are developed by the bank itself (Table 16).

5.4 The information used in deciding on loan applications
Technological change has affected the phase of selection of borrowers, making it easier to
build quantitative models that can sum up the data on a potential borrower.
Credit scoring techniques can be used to enhance the information from other sources that
banks ordinarily use in screening borrowers, or even replace them and become the main means of
evaluation. One of our questions was what importance the bank assigned to the various sources of
information it used to decide whether or not to grant a loan to a first-time applicant; the output of
the statistical model is just one such source, and not always the most important.
The results (Table 17) are in line with expectations. For loans to SMEs, scoring methods are
assigned high importance more frequently by larger banks, less frequently by smaller ones, while
qualitative information is emphasized more commonly by mutual banks. In selecting large
corporate borrowers, not SMEs, the statistical models are less important and qualitative information
more so. Finally, small and mutual banks tend to assign considerable weight to the loan applicants
credit relations with the entire system, a tendency that is accentuated when the potential borrower is
a larger firm.

TABLES

Table 1
Sample composition
(number of banks)

Class of bank Northwest Northeast Centre South/Islands Total

Medium-sized/large 17 9 7 4 37
Small group-member 27 21 24 20 92
Small stand-alone 10 9 10 3 32
Mutual 54 57 40 10 161
Total 108 96 81 37 322

Table 2
Size of sample banks
(millions of euros, number of employees and branches)

Volume of business Employees Branches
Class of bank
Mean Median Mean Median Mean Median

Medium-sized/large 39,380 20,125 5,647 3,342 519 347
Small group-member 3,168 2,350 571 422 67 51
Small stand-alone 1,491 1,460 287 178 32 22
Mutual 668 518 116 96 14 12


Table 3
Distance head office-branch (1)
(kilometers, percentages)

2000 2006
Class of bank
Mean Median Difference (2) Mean Median Difference (2)
Medium-sized/large 152 116 126 151 117 136
Small group-member 50 30 99 66 39 100
Small stand-alone 26 21 67 31 22 69
Mutual 15 14 64 16 14 68
Total 42 19 121 47 21 132

(1) Banks present in sample both in 2000 and in 2006. Distance is mean distance in kilometers of head office from the local labour
systems in which the bank is present, weighted by the banks lending in that local labour system. (2) Interquartile difference over
median, in percent.

Table 4
Geographical diffusion (1)
(numbers, percentages)
2000 2006
Class of bank
Mean Median Difference (2) Mean Median Difference(2)
Medium-sized/large 96 60 147 111 70 117
Small group-member 12 10 100 19 15 97
Small stand-alone 9 5 80 11 7 100
Mutual 3 2 50 3 3 67
Total 17 4 225 21 5 280

(1) Banks present in sample both in 2000 and in 2006. Number of local labour systems in which the bank has at least one branch.
(2) Interquartile difference over median, in percent.


Table 5
Delegation of powers
(thousands of euros and percent)
Autonomous decision-making power of
branch manager, thousands of euros (1)
Index of relative power delegated (2)
Class of bank
Mean Median
Difference
(3)
Mean Median
Difference
(3)
Medium-sized/large 546 250 146 5.3 3.1 159
Small group-member 202 125 118 11.0 8.6 119
Small stand-alone 92 80 125 13.7 14.3 130
Mutual 53 30 217 19.2 16.7 122
Total 154 75 173 14.7 10.8 177
(1) Banks were asked: Consider the granting of loans to non-financial companies applying to your bank for the first time and which,
based on the information available, have no solvency problems. What is the maximum amount of credit (in thousands of euros) that can
be autonomously granted by .... The figure represents the power delegated to the branch manager or head of local unit. Lending to
SMEs. (2) The relative index is the amount of power delegated to the branch manager normalized with respect to that of the CEO.
(3) Interquartile difference over median, in percent.

Table 6
Delegation of powers by technical form (1)
(thousands of euros and percentage index)
Autonomous decision-making
power of branch manager (1)
Index of relative power delegated (2)
Matter for decision
Mean Median Mean Median

Current account overdraft (3) 62 25 16.8 10.0
Self-liquidating credit (3) 107 50 17.4 11.8
Unsecured loans 54 25 16.7 10.0
Mortgage loans 124 60 18.3 10.0
Total credit 154 75 14.7 10.8

(1) The figure represents the decision-making power delegated to the branch manager or head of local unit, in thousands of
euros;.lending to SMEs. (2) The relative index is the amount of power delegated to the branch manager normalized with respect to
that of the CEO. (3) For loans not secured by collateral.


Table 7
Decentralization of powers of decision (1)
(percentages)
Tendency in decentralization
Class of bank
Decrease No change Increase
Medium-sized/large 5.4 21.6 73.0
Small group-member 7.5 38.7 53.8
Small stand-alone 2.7 45.9 51.4
Mutual 2.6 47.7 49.7
Total 4.2 42.1 53.7

Source: Sample survey of 322 banks..
(1) Respondents were asked: In the last three years has the degree of decentralization of powers of decision on lending in your bank
increased, decreased or remained basically the same?. Lending to SMEs.

Table 8
Average tenure of branch managers
(absolute values and percent)
Tenure in months (1) Trend in the last three years (1)
Class of bank
Mean Median Difference (2) Shortened Unchanged Lengthened
Medium-sized/large 32 32 33 39.4 45.5 15.1
Small group-member 38 36 50 36.5 54.0 9.5
Small stand-alone 49 38 63 26.7 60.0 13.3
Mutual 50 48 50 41.8 42.5 15.7
Total 45 38 71 38.5 47.7 13.8

(1) Respondents were asked: Indicate the average length of tenure of branch managers, in months (even an estimate). In the last three
years has it lengthened, shortened or remained unchanged?. (2) Interquartile difference over median, in percent.


Table 9
Factors considered in determining incentives for branch managers (1)
(percentages)

Medium-
sized/large
banks
Small
group-
member
banks
Small
stand-
alone
banks
Mutual
banks
Total
banks

Growth in lending 28.6 19.7 33.3 36.1 29.6
Bad debt and/or impaired loan rate 5.7 16.4 37.5 53.0 32.0
Change in bad debts and/or impaired loans 8.6 21.3 58.3 49.4 35.0
Net earnings on loan portfolio 25.7 14.7 29.2 19.3 20.2
Overall profitability of unit
(e.g., gross income)
88.6 90.2 62.5 60.2 74.4
Average potential riskiness of loan
portfolio
11.4 18.0 20.8 32.5 23.2

(1) Percentage of banks that consider each factor very important for determining branch managers compensation. The choices
were very important, fairly important, not very important and not important at all. Sample limited to banks using incentives
linked to the factors specified.

Table 10
Trend in use of incentives for branch managers (1)
(percentages)

Medium-
sized/large
banks
Small
group-
member
banks
Small
stand-
alone
banks
Mutual
banks
Total
banks

Growth in lending 28.1 25.0 13.6 5.2 15.9
Bad debt and/or impaired loan rate 34.6 33.3 31.6 48.6 40.1
Change in bad debts and/or impaired
loans
39.3 42.6 36.8 48.0 43.7
Net earnings on loan portfolio 23.1 19.2 35.0 20.9 22.5
Overall profitability of unit
(e.g., gross income)
26.5 48.3 36.4 41.0 40.1
Average potential riskiness of loan
portfolio
48.3 33.3 33.3 43.6 40.3

(1) Percentage of banks reporting an increase in the last three years in the importance of the factors indicated in branch managers
compensation. The possible answers were increased, essentially unchanged, decreased and not relevant. Sample limited to
banks indicating a tendency.

Table 11
Organizational variables by region (1)
(number of banks and percentages)

Number of
layers of
hierarchy (2)
Power of
branch
manager (3)
Index of
delegation of
powers (4)
Tendency in
decentraliza-
tion (5)
Tenure in
months
(6)
Incentives for
branch
managers (7)
Northwest 3.28 89 16.6 50.0 40 53.1
Northeast 3.29 125 12.3 54.0 57 58.6
Centre 3.20 90 18.7 51.5 45 47.9
South and
Islands
3.33 94 18.2 45.4 42 51.5
(1) Only small and mutual banks. (2) Mean number of hierarchical levels from branch manager through CEO. (3) Mean loan
decision-making power of branch manager (thousands of euros). (4) The relative index is the amount of power delegated to the
branch manager normalized with respect to that of the CEO. (5) Percentage of banks reporting an increase in decentralization in the
last three years. (6) Median tenure of branch manager at the branch. (7) Percentage of banks that consider at least one of the
factors in Table 6 as very important to branch managers compensation.


Table 12
Diffusion of credit scoring models in assessing firms creditworthiness (1)
(percentages)

Class and location of bank 2000 2003 2006
SMEs Large firms
Class of bank
Banks using statistical/quantitative methodologies
(percentage frequency of responses)
Medium-sized/large 8.1 29.7 97.3 97.3 83.8
Small group-member
6.5 20.7 66.3 64.1 56.5
Small stand-alone 9.4 28.1 59.4 59.4 46.9
Mutual 11.8 24.8 42.2 42.2
Total 9.6 24.5 57.1 56.5
Location of head office
(small and mutual banks only)

Northwest 5.5 19.8 51.7 51.7
Northeast 8.1 19.5 42.5 42.5
Centre 14.9 31.1 58.1 56.8
South 15.2 30.3 63.6 60.6
Total small and mutual banks 9.8 23.9 51.9 51.2
Class of bank
Banks using statistical/quantitative methodologies (2)
(response frequencies weighted by volume of lending to firms)
Small group-member
6.2 20.7 68.5 62.2 63.1
Mutual 11.4 21.6 39.0 40.0
Total 9.9 31.4 87.8 81.4
Location of head office
(small and mutual banks only)

Northwest 4.4 20.9 51.6 53.0
Northeast 9.3 24.3 49.8 51.8
Centre 6.2 20.4 79.2 63.2
South 16.4 29.4 74.9 60.4
Total small and mutual banks 8.0 22.9 61.4 56.2

(1) Respondents were asked: In assessing creditworthiness, do you use automatic scores generated by statistical/quantitative
methodologies (credit scoring and internal ratings)? Please indicate whether these methods are used for lending to SMEs and to large
firms. (2) The percentage frequencies are weighted by share of total lending to firms for the data on 2003 and 2006, by lending to
SMEs and large firms, respectively, for the last two columns.


Table 13
How scoring models were developed (1)
(percentages)
Bank class and firm size Internal
In collaboration
with other
institutions
Purchase
from group
company
Purchase
from outside
company
Other

Lending to SMEs
Small group-member 24.4 24.6 22.5 26.0 0.8
Mutual 11.9 40.5 18.3 24.6 4.7

Lending to large firms
Small group-member 18.6 21.7 19.6 25.7 14.4

(1) Percentage frequency of responses, weighted by volume of lending to SMEs and to large firms, respectively.


Table 14
Information used in scoring models (1)
(percentages)
SMEs Large firms
Information used (2)
Medium-
sized/larg
e banks
Small
group-
member
banks
Small
stand-
alone
banks
Mutual
banks
Medium-
sized/large
banks
Small
group-
member
banks
Small
stand-
alone
banks

Financial statement data 46.5 71.0 56.1 82.6 86.7 77.5 29.9
Geographical area and
economic sector 2.1 1.3 1.7 2.4 0.0 1.2 5.8
Credit relations with entire
system (3) 85.1 35.6 75.4 50.0 35.0 47.3 76.1
Other outside sources of
data (4) 11.8 9.8 9.8 9.8 3.1 7.3 16.0
Relations with bank 50.4 63.9 54.0 46.9 27.3 50.0 47.3
Relations with banks group 3.2 6.9 0.0 0.0 8.7 9.3 0.0
Qualitative information (5) 4.7 13.4 5.5 3.8 33.7 8.8 22.6

(1) Sum of frequencies of responses indicating each source as one of the two most important, weighted by volume of lending to SMEs
and large firms, respectively. Data for 2006. (2) Banks were asked: If you use statistical-quantitative methodologies for assessing
firms creditworthiness, please rank by decreasing order of importance the data considered in your calculation engine in assigning
the overall score: 1 for the most important, 2 for the next most important, and so on. If you do not use any particular factor, answer
NA. (3) Source: Central Credit Register or other credit bureau. (4) Interbank register of bad cheques and payment cards, Protest
bulletin, etc. (5) Codifiable data, as through special questionnaires, on organization of the firm, characteristics of the project to fund,
and so on.

Table 15
Importance of scoring models in lending decisions (1)
(percentages)
Bank class and firm size Lend/not Amount Pricing Maturity Collateral Monitoring

Lending to SMEs
Medium-sized/large 91.8 57.0 14.7 15.8 42.9 70.0
Small group-member 67.0 39.3 15.5 6.7 34.1 73.7
Small stand-alone 54.1 31.4 21.4 13.3 34.0 71.0
Mutual 47.5 31.8 16.5 17.5 31.9 48.5

Lending to large firms
Medium-sized/large 88.0 70.2 20.3 29.0 35.6 82.6
Small group-member 67.6 34.8 17.3 9.0 27.9 61.4
Small stand-alone 50.1 33.2 32.1 14.3 32.1 81.9

(1) Banks were asked to: Rank from 1 to 5, in decreasing order of importance. 1=decisive, 2=very important, 3=fairly important,
4=not very important, 5=not important at all. NA=not applicable. The table gives the sum of the frequencies of answers 1 and 2
(decisive or very important), sample limited to banks that use statistical-quantitative methods. Data for tghe end of 2006. Frequencies
weighted by volume of lending to SMEs and large firms respectively.

Table 16
Importance of credit scoring models in loan approval (1)
(percentages)
Bank class and size of
firm
Developed internally
or in cooperation with
other institutions
Purchased
from outside
Developed internally
or in cooperation with
other institutions
Purchased
from outside
Simple frequencies Weighted frequencies (2)
Lending to SMEs
Medium-sized/large 93.5 20.0 95.2 36.6
Small group-member 73.9 60.0 72.2 56.5
Small stand-alone 40.0 35.7 30.1 64.7
Mutual 54.5 51.4 52.8 43.9

(1) Banks were asked to: Rank from 1 to 5, in decreasing order of importance, the factor in the decision to lend. 1=decisive, 2=very
important, 3=fairly important, 4=not very important, 5=not important at all. NA=not applicable. The table gives the sum of the
frequencies of answers 1 and 2 (decisive or very important), sample limited to banks that use statistical-quantitative methods. Data for
the end of 2006. (2) Frequencies weighted by volume of lending to SMEs.

Table 17
Importance of factors in assessing creditworthiness of new loan applicant (1)
(percentages)
SMEs Large firms
Information used (2)
Medium-
sized/larg
e banks
Small
group-
member
banks
Small
stand-
alone
banks
Mutual
banks
Medium-
sized/large
banks
Small
group-
member
banks
Small
stand-
alone
banks

Statistical-quantitative
methods 70.2 27.6 18.9 8.9 59.6 32.9 0.0
Financial statement data 95.2 85.7 95.2 96.5 100.0 95.2 98.2
Credit relations with entire
system (3) 82.6 86.7 97.2 89.5 72.0 92.5 98.2
Availability of guarantees 28.3 51.8 45.7 42.0 3.9 24.4 33.5
Qualitative information (4) 35.8 33.5 33.4 49.9 69.2 48.8 61.8
First-hand information 16.3 15.9 9.6 15.0 3.9 4.7 2.9

(1) Sum of frequencies of responses indicating each source as one of the two most important, weighted by volume of lending to SMEs
and large firms, respectively. Data for the end of 2006. (2) The banks were asked: For the granting of loans to non-financial firms
that apply to you for the first time, please rank in decreasing order of importance the factors used in deciding whether or not to grant
the loan. 1 for the most important, 2 for the next most important, and so on. Two different factors cannot be given the same rank. If
you do not use any particular factor, answer NA. (3) Source: Central Credit Register or other credit bureau. Interbank register of
bad cheques and payment cards, Protest bulletin, etc. (4) Codifiable data, as through special questionnaires, on organization of the
firm, characteristics of the project to fund, and so on.


APPENDIX

BANKS LENDING: ORGANIZATIONAL AND TECHNICAL FACTORS IN ASSESSMENT
Unless otherwise specified, answers as at 31 December 2006

Bank: ___ ______________________________________ ABI code_ ________
Banking group: __ _______________________________
SECTION A: THE ORGANIZATION OF LENDING BUSINESS
A1 With reference to lending to non-financial firms, please indicate whether your bank:
a) is organized by divisions, distinguishing between large
and small borrower firms (e.g. retail and corporate) and, if
so, since what year

Yes |__| No |__|

yr. |______|
SMEs share of your banks
total lending to non-financial
firms
|______|%
b) within its group (only for group members) specializes in
one customer size segment; if so, since what year

Yes |__| No |__|
yr. |______|
A2 Still on lending to non-financial firms, please indicate the segments into which you divide customers by size class,
specifying variable and threshold value (1).

Name of segment
Variable (unit of
measurement)
Threshold (2)
Sales equivalent of
threshold value (3)
1) SMEs

2) Large firms

(1) If the bank is organized by divisions, specify the segments to which the divisions correspond. If your bank is specialized within a
group, indicate your customer segment, the measurement variable and threshold value. If you do not segment customers, indicate
only the most important size class of firms for your business. (2) For instance: up to 100 should be written <100; between 100
and 200 by 100-200; more than 100 by >100. (3) Only if you do not segment customers by volume of sales.

In the rest of this questionnaire, SMEs and large firms mean, respectively, the size classes identified in the
first and the second lines of question A2. If your business is exclusively or predominantly with one of the two
segments, please fill out only the parts of the questionnaire relevant to that segment.
A3 Consider the granting of loans to first-time applicant non-financial firms. The process may involve a number of
positions within the bank (1): please indicate, for each of the customer segments specified above, these positions in
increasing order of importance, starting from the initial manager of the lending relationship (e.g., branch manager,
corporate loan officer), the geographical area and organization unit to which he belongs (e.g. branch, head office). If your
bank belongs to a group, indicate the positions with the parent bank that are involved (N.B.: if necessary, add additional
lines to the table).

SMEs Large firms
Position
Geographical area
(2)
Organizational
unit
Position Geographical area (2)
Organizational
unit
1

2

3

4

5

6

(1) Specify all the positions that may play a role in loan approval. For instance: branch loan officer, branch manager, coordinating
branch manager, credit area director, CEO, Executive Committee, Board or Directors and, for group members, parent bank credit
area director, and parent bank CEO, Executive Committee and Board of Directors. (2) 1 = sub-provincial, 2 = province, 3 =
multi-province, 4 = region, 5 = multi-region, NA = not applicable.

A4.1 Consider loans to first-time applicant non-financial firms (1). Please specify the maximum amount of credit
(in thousands of euros) that can be autonomously granted by each of the positions named in question A3, distinguishing
according to customer segment and technical form of credit (N.B.: if necessary, add additional lines to the table).
SMEs Large firms
Current
account
facility (2)
(3)
Self-
liquidating
loans (2) (3)
Unsecured
loans (2)
Mortgage
loans
(2)
Total credit
(3)
Current
account
facility
(2) (3)
Self-
liquidating
loans (2) (3)
Unsecured
loans (2)
Mortgage
loans
(2)
Total
credit (3)
1

2

3

4

5

6

(1) Answer for a firm which according to available information has no problems of solvency. (2) If the category may include more
than one case (as when the amount varies with type of branch), give the value representative of the largest number of loans. If the
position responsible has no degree of autonomous power of decision, put 0. If the bank does not make this type of loan, put NA.
If a position has no lending ceiling save that set by law, put max. (3) Consider loans not backed by collateral.
A4.2 With reference to lending both to new borrowers and to past customers, your bank probably also has a grid of
standard terms concerning interest rates. What is the deviation, in basis points, from the standard terms (e.g. 25 basis
points) that can be decided autonomously by the first three positions specified above (1), distinguishing by firm size
and technical form (2).
SMEs Large firms
Current
account
facility (2) (3)
Self-
liquidating
loans (2) (3)
Unsecured
loans (2)
Mortgage
loans (2)
Total
credit
(3)
Current
account
facility (2) (3)
Self-
liquidating
loans (2) (3)
Unsecured
loans (2)
Mortgage
loans (2)
Total
credit (3)
1

2

3

(1) Refer to the positions specified in lines 1, 2 and 3 of question A3. (2) If the category may include more than one case (as when
the amount varies with type of branch), give the value representative of the largest number of loans. If the position responsible has
no degree of autonomous power of decision, put 0. If the bank does not make this type of loan, put NA. (3) Consider loans not
backed by collateral.
A4.3 In the last three years has the decree of decentralization of powers of decision on lending to non-financial firms
in your bank increased, decreased or remained basically the same?
SMEs: increased|__| no change|__| decreased|__| NA |__| Large firms: increased|__| no change|__| decreased|__| NA |__|
A5 What is the weight of the following factors in determining incentives for branch managers compensation? In the
last three years has the importance of each factor increased, decreased or remained essentially unchanged?
Order of
importance (1)
Tendency (2)
Growth in lending

Bad debt and/or impaired loan ratio

Change in bad debt and/or impaired loan ratio

Net earnings on loan portfolio

Overall profitability of unit (e.g. gross income)

Average potential riskiness of loan portfolio

Growth of direct fund-raising

Growth of indirect fund-raising

Other (please specify)

(1) 1 = very important, 2 = fairly important, 3 = not very important, 4 = not important at all; NA = not applicable.
(2) 1 = increased; 2 = essentially unchanged; 3 = decreased; NA = not applicable.

A6 Please indicate the average length of tenure of branch managers, in months (even an estimate). In the last three
years has it lengthened, shortened or remained unchanged?
Mean tenure: |_____| (months) Tendency in last three years: lengthened|__| unchanged|__| shortened|__| NA |__|
SECTION B: ASSESSMENT OF CREDITWORTHINESS
B1 In assessing creditworthiness, do you use automatic scores generated by statistical/quantitative methodologies
(credit scoring and internal ratings)? Please indicate whether these methods are used for the types of lending listed below,
the year they were introduced, whether they were developed internally or purchased from outside, and their importance in
the decision whether or not to lend, amount, pricing, maturity, collateral and monitoring.
Importance of the method in decisions on:

Yes /
No
Year of
introductio
n (1)
Internal /
external
(2)
Loan
approval
(3)
Amou
nt (3)
Pricing
(3)
Maturity
(3)
Collat
eral
(3)
Monitorin
g (3)
Loans to households

Consumer credit

Loans to SMEs

Loans to large firms

(1) Year when first introduced. (2) Specify whether: 1 = the methodology was developed exclusively within the bank; 2 = it was
developed in cooperation with other institutions or consortia; 3 = it was purchased from an outside company belonging to your
group; 4 = it was purchased from an outside company not belonging to your group; 5 = other. (3) Rank from 1 to 5 in decreasing
order of importance: 1 = decisive, 2 = very important, 3 = fairly important, 4 = not very important, 5 = not important at all, NA =
not applicable.

B2 If you use statistical-quantitative methodologies for assessing firms creditworthiness, please rank by decreasing
order of importance the data considered in your calculation engine in assigning the overall score: 1 for the most
important, 2 for the next most important, and so on. No two factors can be given the same rank. If you do not use the
factor, answer NA.

Financial
statement
data
Geographical
area and
economic
sector
Relations
with banks
(1)
Other outside
data sources
(2)
Relations
with your
bank (3)
Relations with
your banking
group (3)
Qualitative
informatio
n (4)
Other
(5)
___
SMEs

Large firms

(1) Central Credit Register and/or other credit bureaus. (2) Interbank register of bad cheques and payment cards, Chambers of
Commerce, specialized companies, etc. (3) E.g., loans and deposits of firm with your bank. (4) Firms organizational structure,
project to finance, etc.. (5) Specify.


B3 For the granting of loans to non-financial firms that apply to you for the first time, please rank in decreasing order of
importance the factors used in deciding whether or not to grant the loan. 1 for the most important, 2 for the next most
important, and so on. No two factors can be given the same rank. If you do not use the factor, answer NA.

SMEs Large firms
Statistical-quantitative methods exclusively
Financial statement data (1)
Credit relations with entire system (data from Central Credit Register and/or other credit bureaus or
public sources, i.e. Interbank register of bad cheques and payment cards, Bulletin of protests, etc.
(1)

Availability of personal guarantees or collateral
Qualitative information (organizational structure of firm, characteristics of project, etc.) (1)
Other assessments based on first-hand information
Other (specify)
(1) With respect to the statistical-quantitative methodologies referred to in question B2, please answer as regards the use of each
source of information outside the algorithms.


REFERENCES
Acemoglu, D., P. Aghion, C. Lelarge, J. Van Reenen and F. Zilibotti (2006), Technology,
information and the decentralization of the firm, mimeo.
Aghion, P. and J. Tirole (1997), Formal and real authority in organizations, Journal of
Political Economy, 105: 1-29.
Allen, L., G. DeLong and A. Saunders (2004), Issues in the credit risk modelling of retail
markets, Journal of Banking & Finance, 28: 727-752.
Bank for International Settlements (2005), Studies on the validation of internal rating
systems, Basel Committee on Banking Supervision, Working Paper, 14, May 2005.
Berger, A.N. (2003), The economic effects of technological progress: Evidence from the
banking industry, Journal of Money Credit and Banking, 35: 141-176.
Berger, A.N. and R. DeYoung (2001), The effects of geographic expansion on bank
efficiency, Journal of Financial Services Research, 19: 163-84.
Berger, A.N., W.S. Frame and N.H. Miller (2005), Credit scoring and the availability, price,
and risk of small business credit, Journal of Money, Credit and Banking, 191-222.
Berger, A.N., N.H. Miller, M.A. Petersen, R.G. Rajan and J.C. Stein (2005), Does function
follow organizational form? Evidence from the lending practices of large and small banks,
Journal of Financial Economics, 76: 237-269.
Bofondi, M. and F. Lotti (2005), Innovation in the retail banking industry: The diffusion of
credit scoring, Review of Industrial Organization.
Bonaccorsi di Patti, E., G. Eramo and G. Gobbi (2005), Piccole e grandi banche nel mercato
del credito in Italia, Banca Impresa Societ, XXIV(1): 3-34.
Christie, A., M.P. Joyeb and R.L. Watts (2003), Decentralization of the firm: Theory and
evidence, Journal of Corporate Finance, 9: 336.
Cowan, C.D. and A.M. Cowan (2006), A survey based approach of financial institution use
of credit scoring for small business lending, Working Paper, Office of Advocacy, United States
Small Business Administration, SBAH-04-Q-0021.
Degryse, H. and S. Ongena (2004), The impact of technology and regulation on the
geographical scope of banking, Oxford Review of Economic Policy, 20: 571-590.
Deng, S. and E. Elyasiani (2005), Geographic diversification, distance, BHC return and
risk, mimeo.
Felici, R. and M. Pagnini (2008), Distance, bank heterogeneity and entry in local banking
markets, Journal of Industrial Economics, 56: 500-534.
Ferri, G. (1997), Mobilit dei dirigenti ed efficienza allocativa: banche locali e nazionali,
Quaderno di Moneta e Credito, 245-265.
Frame, W.S., A. Srinivasan and L. Woosley (2001), The effect of credit scoring on small
business lending, Journal of Money Credit and Banking, 33: 813-825.
Hertzberg, A., J.M. Liberti and D. Paravisini (2007), Information and incentives inside the
firm: evidence from loan officer rotation, mimeo.
Jankowitsch, R., S. Pichler and W.S.A. Schwaiger (2007), Modelling the economic value of
credit rating systems, Journal of Banking and Finance, 181-198.

Liberti, J.M. (2005), How does organizational form matter? Distance, communication and
soft information, mimeo.
Liberti, J.M. and A. Mian (2006), Estimating the effect of hierarchies on information use,
mimeo.
Milgrom, P. and J. Roberts (1992), Economics, organization and management, Prentice
Hall, NJ.
Scott, J. (2006), Loan officer turnover and credit availability for small firms, Journal of
Small Business Management, 44(4): 544562.
Stein, J.C. (2002), Information production and capital allocation: Decentralized versus
hierarchical firms, Journal of Finance, 57: 1891-1921.
Uchida, H., G.F. Udell and N. Yamori (2006), Loan officers and relationship
lending, mimeo.

INFORMATION TECHNOLOGY AND BANKING ORGANIZATION
Sauro Mocetti
-
, Marcello Pagnini
-
and Enrico Sette
**

1. Introduction
Most banks organize their lending activity through a network of geographically dispersed
branches, operating in different local credit markets. Those units are normally run by a local branch
manager (henceforth LBM) who plays a critical role in collecting information about potential
borrowers, in particular small and medium size firms. Through face to face interactions with the
firms managers and their local community, the LBM gains access to relevant information about
firms creditworthiness, builds relationships with the most promising potential borrowers in the
market, screens applicants. Since the efficiency of the decision-making process requires the person
responsible for decisions to have access to the knowledge which is valuable to those decisions, and
since information transmission can be costly, some decision rights should be delegated to the
LBMs, especially concerning small business lending. However, as stressed by principal-agent
models, delegation raises agency costs since the LBM might pursue private benefits that are
potentially conflicting with the maximization of banks profits. Quite naturally, the optimal degree
of delegation is the result of the trade-off between agency costs and the costs of transferring
information within the organization.
Information and communication technologies (henceforth ICT) that deeply shocked credit
markets during the last decades affect this trade-off along several dimensions. On the one hand,
they brought about a substantial improvement in the CEOs ability to measure and observe LBMs
output thereby reducing agency costs and favouring decentralization. On the other hand, they also
contributed to reducing bottom-up communication costs and to tackling CEOs information
overload, widening the possibilities to move decisions up in the hierarchy.
Moving from these opposing effects, the aim of this paper is to examine whether and how
the introduction of ICT affects the allocation of decision rights inside the bank hierarchy. The issue
is particularly relevant since it is usually held that an autonomous and active LBM may facilitate
credit access for small firms that are more opaque and difficult to evaluate.
We derive a simple model of organizational design where the contrasting effects of ICT on
the delegation of authority are represented. The model confirms the ambiguous effects of ICT on
decentralization and also yields an equation for the optimal level of delegation which is estimated
on a unique dataset reporting information on internal organization of a representative sample of
Italian banks.
1
The survey concerns the role of LBMs in small business lending. The focus on
lending to small and medium sized enterprises is motivated by the fact that typically LBMs have
some authority on lending to small business, while decision powers on loans to large corporations
are usually granted to managers at higher hierarchical level.

-
**
We thank Guglielmo Barone, Michele Benvenuti, Luigi Cannari, Moshe Kim, Paolo Mistrulli, Hamid Mehran, Antonello Scorcu,
and other seminar participants at the Bank of Italy for useful comments. An earlier version of this paper was previously circulated
under the title ICT and power delegation in small business lending and was presented at the Third Italian Congress of
Econometrics and Empirical Economics (Ancona).

1
Data was gathered through a specially designed survey conducted by the Bank of Italy in 2007 reporting, among other variables, the
distribution of power delegation across hierarchical levels and the adoption of credit scoring for small business lending. The sample
include all medium and large banks and a representative sample of small banks and credit cooperative banks. See Albareto et al.
(2008) for an illustration of the survey and for some evidences based on it.
50 Sauro Mocetti, Marcello Pagnini and Enrico Sette

Our main econometric findings indicate that banks equipped with more ICT capital and
resorting to credit scoring delegate more decision-making power to their LBMs in small business
lending activity. These results are robust to several additional controls and robustness checks,
including instrumental variable estimation. Moreover, we find that the impact on decentralization is
strengthened for banks with a greater orientation towards small business lending and for those that
jointly hold higher ICT capital endowments and adopt credit scoring.
In light of our model, the interpretation of this evidence is that the impact of new
technologies on the agency costs more than offset that on the costs of transmitting information
(from local branches to headquarter) and on the improvement in central managers decision ability.
Information technology helps banks to improve the internal monitoring activity, thanks to timely
information about lending practices and capability of controlling LBMs activity and performance
of local branches. Finally, credit scoring models, intranet infrastructures and more cost-effective
voice and internet communication have made it easier to analyze credit applications at the branch
level, to share documentation and to combine information stored in different database. All these
factors lessen agency problems and favour delegation toward the peripheral units within the
banking organization.
2

The present paper is at the crossing of three strands of literature. One is related to the LBMs
role in the banking industry. Despite its primacy in small business lending, research in this field is
relatively scant. Previous contributions focus on the impact of organizational form
decentralization or hierarchy on information production and transmission inside an organization
(Stein, 2002), and on how a bank actually disseminates hard and soft information across different
hierarchical layers (Liberti, 2005).
3
Though interesting from our point of view, most of the
empirical findings in this literature rely on clinical studies of a single organization and are therefore
difficult to generalize to the entire banking sector. The second strand of literature concerns the
effects of new technologies in the banking industry. Berger (2003) assesses the impact of
technological progress on productivity growth and on the structure of the banking industry.
4
Berger
et al. (2005) examine the effect of adoption of credit scoring on credit availability and price
conditions for small business lending. Felici and Pagnini (2008) investigate how the ability of
banks to open branches in distant markets changed following the advent of ICT. None of these
papers analyzed the impact of the new technologies on banks organizational structure. This topic
is at the core of a third strand of literature. In particular, Brynjolfsson and Hitt (1998) and
Bresnahan et al. (2002), using firm-level data, find evidence of complementarities between
information technology and workplace organization.
5
Bloom et al. (2009) distinguish between
information and communication technologies and find that the former are associated with more
autonomy and a wider span of control; the latter, by contrast, decrease autonomy for both workers
and plant managers. These studies employ data from firms belonging to diverse industries, whereas

2
Obviously, there can be other channels through which new technologies are enabling greater decentralization. For example, they can
facilitate the information transmission from the centre to the periphery, coordination of local branches and/or reduce the costs of
acquiring information locally. Following Autor et al. (2003), new technologies can also substitute the LBM in carrying out a limited
and well-defined set of standardized and manual activities say routine tasks increasing the efficiency of cognitive and
interactive tasks. These further channels are omitted in our model for sake of simplicity. However, their effects are discussed in the
following section.
3
Stein (2002) argue that decentralization is more attractive when information is soft whereas large hierarchical firms with multiple
layers of management have a comparative advantage when information is hard and can be costlessly passed along within the
hierarchy. Liberti (2005) and Liberti and Mian (2009) empirically show that loans that go to higher levels (that are more distant
from the source of information) rely more on hard information and less on soft information. Uchida et al. (2008) analyze whether
loan officer attributes affect the production of soft information. The role of loan officer has been investigated also in Udell (1989),
Ferri (1997), Scott (2006), and Hertzberg et al. (2009). For a more general analysis of the importance of bank organization on
lending activity see Berger and Udell (2002) and Berger et al. (2005).
4
See also Casolaro and Gobbi (2007) for evidence on Italian banking industry.
5
See also Gurbaxani and Whang (1991), Christie et al. (2003), Colombo and Delmastro (2004) and Rajan and Wulf (2006).
Information technology and banking organization 51

our paper is the first to examine the impact of the new technologies on the internal organization of
banks. Focusing on a single industry, with relatively homogenous inputs and outputs, may help
mitigate problems of combining data from heterogeneous industries. Besides, factors like
asymmetries of information and agency costs underlined by the theories on firm internal
organization are likely to have a special relevance in the banking sector.
The rest of the paper is organized as follows. Section 2 outlines a theoretical framework
describing the trade-off between centralization and decentralization and investigates the impact of
ICT on its terms. Sections 3 and 4 illustrate the empirical strategy and the data used in the paper,
respectively. Section 5 presents the results and a set of robustness checks. Section 6 concludes.

2. The impact of ICT on banking internal organization
As explained in the introduction, the expected impact of ICT on decentralization is
theoretically ambiguous. In order to clarify this intuition and to get some guidance in setting-up an
estimable equation we derive a simple model of decision making in banking, trying to capture the
main trade-offs highlighted by the theoretical literature on delegation.
6
First, delegation provides
incentives to the agent (the LBM); second, delegation determines a loss of control for the principal
(the CEO) and may lead to bad decisions being made if the incentives of the agent are not fully
aligned to those of the principal; third, centralization entails additional costs for the principal
related to collecting information and making a decision; fourth, the principal may wish to monitor
the agent when she is delegated power, and this generates some extra costs, as well.
7
The new
technologies affect the costs for transmitting and processing information as well as those for
monitoring LBMs activity, thus influencing the choice between centralization and
decentralization.
8

Our model is as follows: two agents operate in a bank, the LBM and the CEO (or
headquarter). The latter decides how much authority to delegate to the LBM. Loans of different
size randomly arrive to the bank. Loan size L is distributed according to an uniform distribution
with support | | L , 0 . The LBM exerts unobservable effort to improve the quality of loan origination.
She collects information on the local credit market and on potential borrowers, and she builds
relations with firms which may apply for a loan. The more effort she exerts in this loan generation
activity, the higher the probability that potential borrowers are of good quality, and therefore that
loans generate a high payoff. We assume that neither the loan generating activity, nor the quality of
borrowers is verifiable so that contracts to incentive the LBM cannot be written. The LBM derives

6
Our model blends elements from the literature on delegation as a response to agency problems with elements from the literature on
delegation as a way to minimize the costs of transmitting information within the organization.
7
Delegation of authority to the LBM may increase her initiative and participation in the organization and her propensity to acquire
and use soft information (Aghion and Tirole, 1997; Stein, 2002; Zabojinik, 2002). Moreover, by delegating power to the agent who
directly collects information, banks can fully exploit economies arising from local capabilities and task specialization (Geanakoplos
and Milgrom, 1991), thereby saving on the costs of transferring information within the organization. However, principal agent
models point to the fact that decentralization implies a loss of control, i.e. the agent is freer to pursue private benefits that are
potentially conflicting with principal or CEOs interests. Typical agency costs in the banking industry consist of LBM incentives to
grant loans to poor quality borrowers in order to increase the funds allocated to the branch, or because of a personal friendship with
the owner, the prospect of a future job offer from the borrowing firm or illegal kickbacks. These agency costs may be alleviated
through monitoring or incentive mechanisms that are however, in the presence of incomplete information, imperfect and costly to
implement. For an analysis of organizational failures due to leaks and delays in information transmission between the top and the
bottom of the hierarchy and vice versa, see also Radner (1993) and van Zandt (1999).
8
ICT is likely to improve the monitoring ability of the central management, raising the chances to detect misbehaviour from the
LBM. This, in turn, would increase the scope for decentralization by reducing agency costs (Hubbard, 2000). On the other hand,
ICT adoption lowers the costs of bottom-up communications and it improves CEOs computation abilities thereby reducing her
information overload (Bresnahan et al., 2002). The result would be, in this case, greater centralization.

a private benefit if she is delegated power and the loan is granted. If decision powers are delegated,
the CEO monitors the LBMs activity, learns the quality of the firm, and may decide to over-rule
LBMs decision.
In order to better illustrate our model, we sketch its timing:
1. The CEO designs the organization, by choosing the threshold loan size
*
L above which
decision are centralized.
2. LBM exerts effort in the loan generation activity.
3. The bank receives a loan application of size L. The decision to grant the loan is
centralized or decentralized according to whether L is above or below the threshold
*
L .
4. The person in charge of the decision observes the quality of the firm and decides about
granting the loan. CEO monitors the LBM if the decision is decentralized, and possibly
overturns it.
5. Payoffs are realized.
If the CEO centralizes loans above a threshold
*
L , her expected payoff is

D
CEO
C
CEO
L
L
L
L L
t t
* *
+

where
C
CEO
t is the payoff of the CEO from centralization while
D
CEO
t is the payoff from
delegation. Denote as x the probability the decision is delegated (that corresponds to L L , since
loans are distributed uniformly).
We assume that the effort exerted by the LBM in the loan generation activity raises the
probability that the quality of the loan is high. Let e denote this probability and assume that it
coincides with the effort exerted by the LBM. Our interpretation is that the more the LBM exerts
effort in collecting soft information and building relationship with prospective borrowers, the
higher the average quality of firms applying for credit.
The LBMs payoff is:

( )
2
2
1
1 e x x
C
LBM
D
LBM LBM
H + H = H

where:

( ) ( ) | | 0 1 0 1 + + = H e eB m m
D
LBM

0 = H
C
LBM

B is the private benefit accruing to the LBM if the loan is granted (granting more loans
increases the funds allocated to the branch, the connections with the debtor, etc.). Since the LBM
derives the private benefit B when loans are granted, she decides to offer the loan independently of
firm quality. By monitoring the LBMs decision, the CEO learns the quality of the firm. In case of
delegation, if monitoring does not succeed (with probability 1-m) we assume it is optimal not to
grant the loan. If monitoring succeeds (with probability m) the loan is granted only if the borrower

is of high quality (it happens with probability e).
9
In case of centralization, payoff of the LBM is
normalized to 0.
The LBM chooses effort optimally to maximize her expected payoff. Hence optimal effort
is:

xmB e =

which shows that effort increases provided the LBM is delegated more power with higher
probability (same intuition as in Aghion and Tirole, 1997), monitoring is more likely to succeed,
and private benefit is larger.
10

CEOs payoff is:

( )
C
CEO
D
CEO CEO
x x H + H = H 1

where:

( ) ( ) | |
mon D
CEO
c e eH m m + + = H 0 1 0 1

( )
dec C
CEO
c c e eH + = H
inf
0 1

In case of delegation, if monitoring does not succeed (with probability 1-m) the loan is not
granted and CEOs payoff is normalized to 0. If monitoring succeeds (with probability m) the loan
is granted only if it is of high quality (with probability e); if the quality of firms applying for credit
is good the payoff is high and will be denoted by H. If the loan is bad instead, the CEO manages to
over-rule the decision by the LBM, the loan is not granted and this yields a low payoff normalized
to 0. Monitoring is beneficial, since it allows the CEO to provide incentives to the LBM.
11

Delegating decision power also entails a monitoring activity that is costly (
mon
c ).
In case of centralization, the screening activity and lending decisions by the CEO also
benefit from the loan generating activity performed by the LBM: if the quality of firms applying for
credit is good (with probability e) then loan is granted and the CEOs expected payoff is H; if firms
is not good (with probability 1-e) the loan is not granted, and payoff is normalized to zero.
Centralizing decision power also entails costs of transmitting information from the LBM (
inf
c )
and of processing information and making a decision (
dec
c ).
12

The optimization problem of the CEO is

( )
C
CEO
D
CEO
x x
x
H + H 1
max

or

9
We could assume that when monitoring fails the loan is granted, and results would be essentially unchanged.
10
Optimal effort must be such that e1, so that e is a well defined probability.
11
An alternative formulation posits that when monitoring succeeds and the firm is of bad quality, the CEO is able to fetch a positive
payoff by modifying the decision of the LBM. What matters for our results is the assumption that successful monitoring yields a
larger payoff to the CEO than if monitoring fails.
12
With no loss of generality, we are normalizing the cost of making a decision about the loan to zero for the LBM. This follows since
we can assume that she has better access to information about borrowers and also that given her incentives, she does not exert much
effort to evaluate the quality of borrowers, since she always prefer to grant a loan.

( ) ( )
dec mon
c c eH x c meH x
x
+
inf
) 1 (
max

where xmB e = and 1 0 s s x .
It can be verified that this function is concave in x and optimal delegation is:

( )
( ) | |
mon dec
c c c mHB
BH m m
x + +
=
inf *
1 2
1

which shows that the CEO delegates authority more, the higher the costs of transmitting
information and of making a decision, and the lower the cost of monitoring.
13

The effect of ICT endowment on delegation is ambiguous, since it affects the cost of
transmitting information, the cost of deciding about the loan, and the cost of monitoring. Both
according to theory, and on the basis of availability of data, we assume that the cost of transmitting
information within the bank depends upon the ICT endowments, the distance of the branch from
the headquarter and the size of the bank, as follows:

size dist ICT c
3 2 1
inf
+ + =

where we assume that 0
1
< , as ICT reduces the cost of transmitting information within the
organization, while it is assumed that 0
2
> , 0
3
> since these costs are likely to increase in the
case of wider geographical branch networks implying longer centre-periphery distances and of
larger and more complex organizations. Finally, we assume that
dec
c and
mon
c depend on the same
variables as those determining
inf
c :

size dist ICT c
dec
3 2 1
u u u + + =

size dist ICT c
mon
3 2 1
+ + =

Moreover the effects of those variables are also assumed to be the same as before implying
that 0
1
< u , 0
2
> u , 0
3
> u and 0
1
< , 0
2
> , 0
3
> . By plugging the expressions for costs
into the equation for optimal delegation, we obtain:
( ) ( ) ( ) ( )
size
BH m m
dist
BH m m
ICT
BH m m BH m m
mBH
x
+
+
+
+
+ +
+
=
1 2
) (
1 2
) (
1 2
) (
1 2
3 3 3 2 2 2 1 1 1 *
u u u

and this yields an estimable equation for the optimal delegation by bank i:

i
i
i
size dist ICT
L
L
x c | | | | + + + + = =
3 2 1 0
*
(1)
This simple estimable model suggests that the effect of ICT on delegation of authority is
ambiguous, since it depends upon the sign of ) (
1 1 1
u + , which is to say, that it depends upon
whether ICT impacts more strongly on the cost of transmitting information, on the cost of decision
making, or on the cost of monitoring.

13
Notice that the non-negativity constraint may be binding, and x*=0 if the numerator is negative. Similarly, if the optimal x*>1, the
constraint x1 will be binding.

There are a few assumptions of the model which are critical to obtain a simple linear
estimable equation. First, the distribution of loans is assumed to be uniform. In a more general
model, the dependent variable could be a complex function of the absolute loan size which is
delegated to the LBM. Hence, as a robustness check, we also estimate the model using absolute
delegation as a dependent variable, since this is just the maximal size of loans on which decision
powers are granted to the LBM. Second, we are assuming that optimal effort e does not depend on
ICT. Removing this assumption would generate a non linear relationship between optimal
delegation and ICT. Then, the first order condition for optimal delegation could still be brought to
data and estimated by non-linear least squares, or it could be approximated by a linear equation.
More generally, we could have written a more complex and richer model to capture the different
channels through which ICT affect delegation of authority.
14
We believe our model captures the
most relevant channels through which ICT affects delegation. Adding further channels would not
change the main prediction that ICT has an ambiguous effect on delegation at the cost of yielding a
complex non linear model to estimate. Third, we are modelling the LBMs activity as one of loan
generation. The LBM collects information and builds relationships with the most promising
potential borrowers, in this way raising the average quality of firms that actually apply for a loan.
However, LBM, also collects information about borrowers when they actually apply for a loan (in
our model, after a loan of size L arrives), in order to support the decision to grant the loan. This
occurs both if the decision is delegated and if it is centralized. Hence, we can easily extend our
model to include this further activity performed by the LBM, allowing for ICT to impact on that,
and little would change.

3. Empirical strategy
In order to take equation (1) to data, it is necessary to think about the process through which
banks invest in ICT, especially since organizational form and ICT adoption may be jointly
determined. The scenario we have in mind starts from the recent waves of innovation in the
telecommunication and computer sectors. These developments brought about a dramatic fall in the
price of hardware and software and induce banks as well as firms in other industries to speed up
ICT adoption. However, the intensity and speed of reaction to this largely exogenous price shock
differ across banks depending on the quality of the management, the lending practices existing
before the introduction of the new technologies and other organizational variables.
In particular, we assume that the introduction of the new technologies will take time to exert
its effects because of the existence of learning costs related to their use and the uncertainty
surrounding their returns. Moreover, after some time, ICT will affect those components of the
organizational structure that are more flexible and present lower adjustment costs such as our proxy
for the LBMs decision power (more on this below). This justifies our econometric strategy based
on the idea that ICT adoption influences the degree of decision-making centralization in small
business lending. In the long run it is likely that ICT adoption and organizational factors will be
jointly determined as underlined by the literature (Breshnan et al., 2002). But in the short and
medium term, it is reasonable to assume that the casual link runs from ICT to organizational
variables.
The empirical analysis is carried out on a cross section of about 300 Italian banks and the
empirical specification follows from equation (1):

14
Just to make a few examples, new technologies can affect the decision costs of the LBM when she is in charge of the decision; they
affect the LBMs ability to monitor the borrower after the loan is granted; they affect the cost of transmitting information from the
headquarter to local branches (besides bottom-up communications) and diminish the coordination costs of peripheral units.

i i i i i
X SCORING CAP ICT PD o | | o + + + + =
2 1
_

where PD
i
represents the empirical counterparts of x*, the optimal delegation identified in
equation (1). More specifically, PD
i
equals the ratio between the loan size that the LBM of bank i
can autonomously grant to a small firm applying for credit and the corresponding amount that can
be extended by banks CEO. In our data, the power delegation of the CEO mirrors the upperbound
of the loan distribution that arrives to the bank. The implicit assumption is that the maximum loan
amount applied by a firm corresponds to the maximum sum of money that can be granted by the
vertex of the bank. Therefore the ratio between LBMs and CEOs power delegation identifies the
threshold that discriminates between delegation and centralization. Among the explanatory
variables, new technologies are represented through ICT_CAP and SCORING. ICT_CAP
i
is the log
of ICT capital stock per employee in bank i considered with a three year lag; SCORING
i
is a
discrete variable that distinguishes the following situations: no adoption of credit scoring for small
business lending, adoption in the last 2 years and adoption since at least 3 years. As explained
above, the use of lagged values for these variables is consistent with a short run causal relationship
in which power delegation is determined by new technologies adoption. Moreover, the use of lags
for all the regressors may help solving simultaneity problems that might affect our econometric
findings.
The use of ICT_CAP or some of its related proxies is relatively common in empirical studies
on organization. It picks up the effects that the investment in computers and software might have
on the relationship between the vertex of the bank organization and the peripheral managers. The
role of SCORING is relatively new and it is clearly specific to lending markets. These techniques
provide a standardized screening procedure that allows banks to reduce costs of recording and
transmitting information and to speed up the loan approval process. Implementation of such a
lending technology is intimately related to the use of the new technologies and consequently is
considered here as part of that process. Credit scoring models, intranet infrastructures and
communication technologies allow timely and frequent information about lending practices and
local branches performances. Other improvements include efficiency gains in accessing data,
sharing documentation, in combining information stored in different database, and in
communicating between headquarter and peripheral units. Finally, there can be positive spillovers
between credit scoring and ICT endowments since scoring models affect how information is
recorded and stored while software and other applications affect how information is processed. We
also include a set of covariates (size, distance, etc.) varying across banks and that are also taken
with a three year lag.
15

1
and
2
are the key parameters to be estimated.

4. Data
The dataset is obtained by combining information on banking internal organization taken
from a specially designed survey and additional data derived from the Supervisory Reports (SR) of
the Bank of Italy. The survey, conducted by the Bank of Italy in 2007, included questions about the
organizational structure of the lending activity, the distribution of power delegation across
hierarchical levels and the adoption of credit scoring for small business lending. The targeted
sample consisted of more than 300 banks that include the universe of medium and large banks and
a representative sample of small and cooperatives banks.
16
We excluded those intermediaries that

15
In what follows, we will use lagged values for the explanatory variables whenever it was possible. For some regressors, we observe
data for the year 2006 only (see Table A1 in the appendix).
16
The accuracy of the questionnaire was guaranteed through several checks before and after the realization of the survey. First, the
questionnaire was tested with preliminary interviews with bank managers to detect potential ambiguities in the phrasing and
(continues)

do not lend to small firms (e.g. those specialized in lending to large firms) and foreign bank
branches. The final sample consists of 297 observations. The SR contains balance sheet data, loan
portfolio characteristics and other information disclosed by banks complying with prudential
regulation requirements.
In the following subsections we will describe our key variables whereas Table A1 in the
Appendix reports a complete list and description of the covariates included in vector X
i
and that
will be used in the empirical analysis.

4.1 Power delegation
One of the goals of the survey was to get a measure of the degree of delegation in the loan
origination process. To this end, banks were asked to list all the figures within the organization
involved in lending activity, ordered hierarchically from the local loan officer up to the board of
directors (see Table A2 in the Appendix). In addition, banks were asked to indicate the maximum
loan amount that each figure can grant in autonomy to a small firm applying for a banks credit for
the first time.
17
This choice was motivated by the need to avoid that past characteristics of a banks
loan portfolio had a too strong impact on the decision to decentralize authority. At last, levels of
delegation had to be referred to applicants exhibiting a risk level that a bank judge a priori as
normal. In other words, we wanted to observe bank organizational strategies for the pool of
applicants that are neither exceptionally good or bad because these borrowers are normally rare and
at the same time they may induce an abnormally low or high level of delegation toward the LBM.
The distribution of power delegation across the hierarchical layers mirrors the allocation of
formal authority within the organization. Following Aghion and Tirole (1997), formal authority
refers to the right to make decisions, where the attribution of authority is ratified by a contractual
relationship. In our framework, this kind of power is transferred from the CEO to the subordinates
by means of an act of delegation. When the CEO keeps formal authority, she is in charge of the
decision process concerning the loan approval. If the CEO delegates decision-making power to the
LBM, the latter is able to act autonomously within the limits set by the act of delegation. Starting
from the distribution of power delegation across the hierarchical layers, we built a measure of
decentralization for each bank, as suggested by our structural model. This is obtained by
normalizing the credit amount offered by the LBM with respect to that of the CEO.
18

On average the LBM has a power delegation that is 15 percent of that of the CEO. This
percentage is smaller for large banks (about 5 percent) and higher for cooperative banks (nearly 20
percent). Power delegation is zero for about 10 percent of the banks whereas the maximum is 68
percent. The distribution of the variable is right-skewed with the mass concentrated around values
that are not very distant from zero. However there is quite a lot of variance across banks: the 25 and
75 percentiles are respectively equal to 3 and to 22 percent.
19

LBMs autonomy is clearly a multifaceted concept and therefore is very difficult to represent
through a unique index. Under this respect, our proposed measure may clearly have some
limitations. First, it does not pick up other aspects of the loan contract that may characterize LBMs
authority like the power to set interest rates and to ask for collaterals. Second, being based on a

improper question formulations. Second, several rounds of telephone interviews were conducted to avoid missing data and to correct
inconsistent responses in the questionnaire. The response rate was nearly 100 percent.
17
This means that a specific applicant was not a customer of that bank at time t-1.
18
In our sample there is a great variance across banks with respect to organizational structure and hierarchical layers. The LBM and
the CEO are two figures featuring almost all the banking organizations.
19
The distribution of LBMs absolute power delegation across banks is reported in Figure A1 in the Appendix.

concept of formal authority, our index could misrepresent LBMs real authority that pertains to the
effective control over decisions and that derives from the knowledge of crucial information
(Aghion and Tirole, 1997). For instance, an high level of PD could not necessarily signal LBMs
real authority for those banks that do not allow a LBM to remain in the same branch for a sufficient
time span as to accumulate qualitative knowledge of the local credit applicants.
The first objection can be addressed by observing that the power to grant a loan is crucial
within a bank-firm relationship and that power delegation is arguably correlated with other aspects
that characterize LBMs authority in the lending process. The second objection will also be
addressed by introducing a set of robustness checks aimed at controlling for other organizational
variables that affect the LBMs capability of acquiring information.
Finally, our index can also be evaluated through a comparison with alternative proxies for
the decentralization of decision-making power used by the literature on organization. Bresnahan et
al. (2002) propose an indicator based on the importance of self-managing teams among the
production workers. Colombo and Delmastro (2004) rank modes of allocating decision-making
centralization, partial delegation and full delegation depending on the autonomy of plant
managers in taking some strategic decisions related to plants activity. Christie et al. (2003) and
Acemoglu et al. (2007) define an organizational unit of the firm as decentralized when it is
designed as a profit centre. The basic assumption is that a profit-centre unit has a broader set of
decision rights than a cost-centre unit.
20
Compared to the other indexes, our indicator is based on
objective data and not on a subjective assessment about the importance of the peripheral manager.
Second, it can be directly interpretable as a measure of how much decision rights are allocated to
the local branch level. Third, it is continuous and therefore it is not subject to the loss of
information connected to a discrete measure.

4.2 ICT capital stock and credit scoring
As stated above, new technologies are represented through ICT capital stock and credit
scoring adoption. The former has been computed using the perpetual inventory method. Formally:
( ) | |

+ = =

j
t j i t j i j
j
t j i t i
I K K CAP ICT
, , 1 , , , , ,
1 _ o

where K
i,j,t
is the capital stock of bank i for a particular asset type j at time t,
j
is the constant
rate of depreciation for asset j, and I
i,j,t
is the real investment by bank i for asset j at time t. Types of
assets include hardware, software and premises for computing equipment. Bank nominal
investment flows are deflated using the hedonic price indexes developed by the Bureau of
Economic Analysis and adjusted for the variation in the EU/USD exchange rate.
21
Then, capital
stock is obtained as a weighted sum of the past real investments, with weights given by the relative
efficiency of capital goods. Depreciation rates are assumed to be constant over time and different
across types of assets. Following Jorgenson and Stiroh (2000), software and hardware are assumed
to depreciate at a yearly rate of 31.5 percent, premises for computing equipment at a rate of 11.5
percent.
22

20
In Christie et al. (2003) the decentralization measure is obtained from a questionnaire, which asks the firms management to identify
the second level below the CEO and tell whether it is a profit center, cost center or a mixture of the two.
21
We rely on hedonic prices to supplement traditional price index methods that do not properly account for the rapid technological
progress in ICT industry. Just to give an example, a new computer might offer twice the memory of its predecessor with no change
in its production cost. Therefore the data on ICT investments in current prices need to be deflated to arrive at constant-quality prices.
22
Data on ICT investments are available starting from 1984. For 1983 the capital stock is set equal to zero. Although not fully
satisfactory, this assumption does not affect our results because of the high capital stock depreciation rate.

The ICT capital had been steadily increasing during the last decade: in 2006 its real value
was almost three times as large as in 1995 (Figure 1). Investments in hardware augmented by more
than 400 percent, the software component by 70 percent. This trend was mostly due to the
exponential decline in the price/performance ratio of computers and related technology, reflected in
the BEA price deflator.

Figure 1
Real ICT investments and capital stock
(1995=100)

100
200
300
400
500
600
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
100
120
140
160
180
200
220
240
260
280
300
Hardware investments Software investments ICT capital stock (right axis)

Source: Supervisory Reports of the Bank of Italy, Istat and BEA.

Credit scoring is the process of assigning a quantitative measure the score to a potential
borrower (Feldman, 1997). The score represents an estimate of the borrowers future loan
performance and is obtained combining information from firms balance sheet data, credit history
of the loan applicant and, sometimes, qualitative information. Our survey documents the diffusion
of this financial innovation across Italian banks (Albareto et al., 2008). Banks were asked to report
whether they adopted credit scoring for small business lending and (if so) the number of years since
the adoption. In 2006 about 60 percent of the banks in the sample adopted those techniques. The
adoption rate accelerated in recent years, probably because of the New Basel Capital Accord
(Figure 2). Almost all the medium and large banks adopted credit scoring whereas about 65 percent
of small banks and 45 percent of credit cooperative banks followed the same strategy.
23

23
See also Akhavien et al. (2005).

Figure 2
Adoption of small business credit scoring by year

0
20
40
60
80
100
120
140
160
180
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006

Source: Survey on Banking internal organization.
Number of banks that adopted credit scoring per year. The sample consists of 297 observations.

5. Results
5.1 Main findings
Correlations between the explanatory variables reported in Table A3 show that most of them
are strongly correlated with bank size (SIZE). To cope with the potential collinearity problem, we
start with a very parsimonious specification in which regressors include only our ICT variables and
SIZE (see Table 1). SIZE is included to get rid of the differences across banks attributable to the
heterogeneity in terms of organizational complexity and institutional nature. Moreover our key
explanatory variables, ICT_CAP and SCORING, are included separately in columns I and II and
simultaneously in column III. Columns IV through VII add stepwise further controls to the basic
specification.
SIZE has a negative and significant effect on the dependent variable. This finding is
consistent with the idea that large banks face higher agency costs and therefore delegate less to the
peripheral units within the organization. The other covariates are correlated with SIZE therefore it
might not be possible to correctly and separately identify their effects. Since we are not interested
in a structural interpretation of their parameters, these regressors are added to the specification as
they can potentially affect LBMs decisional power and their omission could distort our
econometric findings. We start adding a set of dummies for CEOs location. These regional
dummies pick up all the influences that local environment might have on CEOs attitudes and
strategies including its propensity to delegate; they are particularly relevant in the case of small
banks since they control also for the characteristics of the local markets where these banks operate.
DISTANCE, which is equal to the log of the average distance between the local branches and the
bank headquarter, is negatively and significantly associated to the degree of decentralization. This
evidence is consistent with the fact that, all else being equal, agency costs increase with the CEO-

LBM physical distance thereby augmenting the need to centralize decision-making within the
organization in response to a loss of control.
24
LOAN SIZE (the average size of loans that each bank
offers to non financial firms) is introduced since LBMs role is enhanced when lending activity is
focused on small business. However, LOAN SIZE does not significantly affect the degree of
decentralization, probably because of the collinearity problems mentioned before. Last, we include
BRANCH SIZE to control for the fact that a LBM running a larger branch could occupy a higher
position within the bank hierarchy and because of that might also be given a larger autonomy. The
estimated parameter for this variable is positive, consistently with our expectations.

Table 1
Baseline results

I II III IV V VI VII
ICT_CAP 0.013*** 0.012** 0.012*** 0.013*** 0.013*** 0.012***

(0.004) (0.005) (0.004) (0.004) (0.004) (0.004)
SCORING 0-2 -0.012 -0.018 -0.024 -0.020 -0.020 -0.022

(0.017) (0.017) (0.018) (0.017) (0.018) (0.018)
SCORING 3+ 0.069*** 0.060** 0.063** 0.071*** 0.069*** 0.070***

(0.024) (0.024) (0.024) (0.025) (0.025) (0.024)
SIZE
-0.035*** -0.031*** -0.033*** -0.032*** -0.013 -0.018* -0.028***

(0.005) (0.005) (0.005) (0.005) (0.009) (0.009) (0.010)
DISTANCE
-0.040*** -0.034** -0.030*

(0.015) (0.016) (0.016)
LOAN SIZE
0.026 0.009

(0.021) (0.023)
BRANCH SIZE
0.070**

(0.030)
FE
- - - YES YES YES YES
Observations
291 296 291 291 291 289 289
R-squared
0.13 0.16 0.18 0.26 0.27 0.28 0.29

OLS estimates. The dependent variable is PD that is the ratio between the loan size that a LBM can autonomously grant to a small
firm applying for credit and the corresponding amount that can be extended by the CEO. The key explanatory variables are ICT (log
of ICT capital stock per employee) and SCORING (dummies equal to 1 if the bank adopted small business credit scoring in the last 2
years or since at least 3 years, respectively). Robust standard errors are reported in brackets; * significant at 10%; ** significant at
5%; *** significant at 1%.

As far as our key explanatory variables are concerned, our findings clearly indicate that
banks holding more ICT capital and having adopted credit scoring techniques since at least 3 years
increase LBMs decision power over small business lending. Estimated parameter for ICT_CAP
and SCORING 3+ are both positive and significantly different from zero. On the contrary, the
impact of SCORING 0-2 is not significant. It is likely that learning costs associated to the
introduction of scoring techniques and uncertainty surrounding their screening capabilities do not
enable them to have an immediate and/or a short-term impact on power delegation. Notably, these
results are confirmed even when ICT_CAP and SCORING 3+ are included in the same regression,

24
Other papers analyzing the effects of center-periphery distance on bank strategies include Felici and Pagnini (2008) and Berger and
De Young (2001).

showing that their effects can be separately identified. Moreover, the magnitude of the coefficients
and their statistical significance is substantially similar in all the specifications proposed.
Bresnahan et al. (2002) and Colombo and Delmastro (2004) find similar results on the relationship
between ICT adoption and decentralization using firm level data.
The role of ICT variables in shaping internal organization is also quantitatively relevant.
According to our estimate, the adoption of credit scoring since at least 3 years increases our index
of power delegation by 0.07, nearly one-half the mean value. Moving from the first to the third
quartile in the distribution of the ICT stock per employee, would increase the index of power
delegation by about 10 percent. To compare the magnitudes of the effects associated to the two
variables, we treat SCORING as it were continuous. It turns out, that a one standard deviation
augment for ICT_CAP (SCORING 3+) would lead to an increase by 14 (20) percent of the standard
deviation for PD, ,thereby showing that the intensity of the two effects is not dissimilar.
Our findings suggest that the new technologies complement rather than substitute the role of
the LBM in small business lending. By increasing its investments in computers and software and
by adopting credit scoring, a bank improves the quality and the frequency of reporting on the
performance of the different local branches. Improvements in the internal monitoring, in turn,
favour more delegation exploiting LBMs local capabilities and specialization in business lending.
In the parlour of our model in section 2, these savings on agency costs brought about by new
technologies could overcome those generated through the fall in communication costs and the
improvement in CEOs decisional process. In turn, these effects would modify the balance between
centralization and decentralization favouring the latter.
A potential criticism that can be raised against our interpretation is the following. Suppose
that banks adopting credit scoring organize their loan origination process in a hierarchical way.
Namely, in the first stage of that process they screen loan applicants through the scoring system. In
the second stage, they attribute loan power origination to a LBM if and only if that applicant
obtained a sufficiently high score in the first stage of the evaluation process. Hence, under that
organization mode, the positive correlation between PD and scoring would merely reflect the fact
that a LBM could originate loans only for those applicants passing the test carried out in the first
stage screening procedure. This argument however is not fully convincing in the context of our
empirical analysis. In fact, the banks participating the survey were asked to indicate delegation
levels referred to relatively good quality borrowers, i.e. presumably those credit applicants passing
the test in the first stage screening procedure described above.
25

5.2 Alternative measure of power delegation
In this subsection we control the robustness of our results using a different definition of
power delegation, namely the (absolute) maximum loan amount that a LBM can grant in autonomy
to a small firm. We start again with a very parsimonious specification and then we add stepwise the
main covariates (Table 2). The absolute amount of power delegation is positively and significantly
associated with SIZE. Again, most of the bank variables are strongly correlated with size and
therefore it might not be possible to correctly and separately identify their effects. For example, the
coefficients on SIZE is reduced when we add LOAN SIZE. Larger banks typically focus their
lending on larger firms which, in turn, borrow greater amount of money. Therefore, the positive

25
Furthermore, our findings continue to be interpretable as a positive relationship between scoring adoption and decentralization of
decision rights, since PD measures the amount of lending that can be granted at the branch level (i.e., without relying on the
approval of higher hierarchical layers), independently from the chosen organizational mode.

association between the loan officer delegation and the size of the bank partly reflects the fact that
larger banks need to delegate more in order to keep lending decisions at the local branch level.
Table 2
Baseline results with absolute level of delegation

ICT_CAP
0.135** 0.119** 0.107* 0.109* 0.119* 0.110**
(0.057) (0.060) (0.061) (0.061) (0.058) (0.056)
SCORING 0-2 0.336 0.302 0.339 0.347 0.361* 0.324
(0.213) (0.216) (0.219) (0.218) (0.214) (0.218)
SCORING 3+ 0.412* 0.352 0.398* 0.414* 0.390* 0.393*
(0.212) (0.215) (0.219) (0.222) (0.222) (0.221)
SIZE 0.646*** 0.631*** 0.620*** 0.558*** 0.601*** 0.490*** 0.427***
(0.057) (0.060) (0.061) (0.064) (0.119) (0.105) (0.118)
DISTANCE -0.092 0.035 0.083
(0.198) (0.171) (0.173)
LOAN SIZE 0.544** 0.430*
(0.248) (0.260)
BRANCH SIZE 0.401
(0.387)
FE
- - - YES YES YES YES
Observations
291 296 291 291 291 291 289
R-squared
0.33 0.32 0.34 0.43 0.43 0.45 0.45

OLS estimates. The dependent variable is the loan size that a LBM can autonomously grant to a small firm applying for credit. The
key explanatory variables are ICT (log of ICT capital stock per employee) and SCORING (dummies equal to 1 if the bank adopted
small business credit scoring in the last 2 years or since at least 3 years, respectively). Robust standard errors are reported in brackets;
* significant at 10%; ** significant at 5%; *** significant at 1%.

As far as ICT_CAP and SCORING 3+ are concerned, our results are fairly consistent with
those presented in Table 1. The impact of the new technologies on the level of LBMs power
delegation is positive and significant in almost all the specifications. In particular, increasing the
ICT capital stock per employee by 10 percent leads to an augment by 1 percent in the maximum
amount of the loan that a LBM can autonomously grant to a small firm applying for credit. Having
adopted credit scoring since at least 3 years increases the absolute level of power delegation by 10
percent.
Although our empirical findings are qualitatively confirmed when we use the absolute level
of LBMs power delegation as dependent variable, we will use the relative measure in the rest of
the empirical analysis. Apart from the fact that the relative index is directly suggested by our
structural model (see above), it is also much closer to what we are really interested in since
establishing the degree of decentralization in an organization necessarily means measuring the
autonomy of the local manager with respect to the pinnacle of the hierarchy. Moreover, a relative
index enables us to get rid of the huge differences existing across Italian banks in terms of size,
business focus and institutional nature.

5.3 Robustness checks
In this subsection we provide several robustness checks.
26

Table 3
Controlling for other banks characteristics and organizational variables

ICT_CAP 0.012*** 0.012*** 0.014*** 0.011*** 0.014*** 0.014*** 0.014***

(0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004)
SCORING 0-2 -0.020 -0.020 -0.015 -0.017 -0.034* -0.034* -0.026

(0.018) (0.019) (0.017) (0.018) (0.020) (0.020) (0.021)
SCORING 3+ 0.071*** 0.072*** 0.072*** 0.081*** 0.070*** 0.070*** 0.083***

(0.024) (0.025) (0.024) (0.025) (0.026) (0.026) (0.026)
SIZE
-0.029*** -0.025** -0.020* -0.030*** -0.036*** -0.035*** -0.030**

(0.011) (0.011) (0.011) (0.011) (0.012) (0.012) (0.015)
DISTANCE
-0.031* -0.031* -0.030* -0.031* -0.021 -0.023 -0.028

(0.017) (0.018) (0.016) (0.017) (0.018) (0.019) (0.021)
LOAN SIZE
0.005 -0.013 0.013 0.005 0.028 0.029 -0.006

(0.024) (0.027) (0.023) (0.029) (0.027) (0.027) (0.034)
BRANCH SIZE
0.074** 0.070** 0.066** 0.090*** 0.077** 0.076** 0.112***

(0.030) (0.030) (0.029) (0.033) (0.032) (0.032) (0.037)
ROA
0.224 0.837

(0.848) (2.737)
BAD LOANS
0.116 0.008

(0.192) (0.331)
NUMLEV
-0.067** -0.045

(0.029) (0.035)
TURNOVER
0.017 -0.015

(0.023) (0.027)
CEO EDU
-0.014 -0.011

(0.017) (0.020)
CEO AGE
-0.001 0.000

(0.002) (0.002)
FE
YES YES YES YES YES YES YES
Observations
287 285 289 268 254 254 231
R-squared
0.30 0.29 0.31 0.32 0.31 0.31 0.34

In Table 3 we add further controls to our preferred empirical specifications. We start by
adding bank profitability, approximated by return on assets (ROA) that can be considered as a
proxy for CEOs managerial talent.
27
A talented CEO could have a stronger propensity to adopt

26
In unreported evidence we also control the robustness of our results using different empirical specifications. Namely, we use a
TOBIT model instead of OLS in order to take account of the fact that our measure of decentralization, PD, varies between 0 and 1.
We also consider an alternative estimation method (Statas glm procedure) enhanced to deal with fractional response data (see
Papke and Wooldridge, 1996). The empirical findings are substantially unchanged.
27
This variable could also signal bank market power.

new technologies and could be also better able to select good quality LBMs and therefore to
delegate more power to them. In column II we add BAD LOANS (the ratio of bad loans to total
lending) that should capture the effects of the riskiness of the loan portfolio on delegation. In
particular, a riskier loan portfolio could induce a lower degree of decentralization. Based on our
findings, none of these variables play a significant role.
Columns III through VII add variables referring to the internal bank organization and to
CEOs characteristics. NUMLEV is the log of the number of hierarchical levels between the LBM
and the CEO and it should capture how the depth of the hierarchical structure affects the degree of
decentralization. NUMLEV is negatively and significantly associated to the degree of power
delegation. This evidence is consistent with the fact that in a deeper organization decision rights are
distributed on a higher number of intermediate positions and therefore peripheral managers are
likely to have less decision-making power. TURNOVER measures the average LBMs permanence
within the same local branch. The introduction of TURNOVER is aimed at controlling for further
variables that might affect real LBMs authority in small business lending. The latter might vary
with the banks internal policy concerning LBMs turnover across local branches since a higher
permanence within the same local branch can be associated to better knowledge of the local credit
market.
Finally, we control for CEOs individual characteristics since they may affect both the
adoption of new technologies and the banking internal organization. For instance, a younger and/or
a more educated CEO may be more willing to adopt innovative organizational strategies and to
introduce new technologies. Therefore we include CEO EDU a dummy variable that is equal to 1
if the CEO is graduated and CEO AGE the age of the CEO of the bank. Most of the coefficients
on these additional controls are not significantly different from zero. On the contrary, the
coefficients on ICT_CAP and SCORING 3+ are unaffected by all these robustness checks, i.e. they
remain stable and highly significant in all the specifications adopted.
Table 4
Sample splits

By size: By small business specialization:
Small banks Large banks Small fraction of
small business
Large fraction of
small business
ICT_CAP 0.017*** 0.006 0.004 0.017**

(0.006) (0.004) (0.005) (0.007)
SCORING 0-2 -0.031 -0.000 0.003 -0.031

(0.029) (0.018) (0.026) (0.023)
SCORING 3+ 0.082** 0.053** 0.059* 0.075**

(0.040) (0.026) (0.031) (0.036)
SIZE
-0.015 -0.033*** -0.031*** -0.027**

(0.026) (0.006) (0.008) (0.011)
Observations
145 146 146 145
R-squared
0.09 0.21 0.17 0.13


Banks of our sample are very heterogeneous in terms of their size that, in turn, is strongly
correlated with other variables approximating lending practices and internal organization.
Therefore, one might want to examine whether the empirical findings hold when considering
samples of banks more homogenous in terms of size. We define small (large) banks those with total
assets below (above) the median and we run separate regressions for these subsamples (see Table
4). It is worth noticing that the results should be interpreted with caution given the limited number
of observations available. Our key variables (ICT_CAP and SCORING 3+) have the expected
positive sign in both subsamples, even though the impact seems to be somewhat stronger (and with
higher level of significance) for small banks. This effect might be driven by higher specialization of
these banks in small business lending, as emphasized by the second sample split. In fact, the
magnitude and the statistical significance of the coefficients of our key variables are higher for
banks having a larger fraction of small business in their portfolio. By exploiting the predictions of
our model, a possible explanation for this evidence is that the reduction of monitoring costs due to
ICT adoption was larger for banks with smaller and more opaque customers. Alternatively (and
perhaps more reasonably), new technologies have reduced the cost of transmitting and processing
information (and thus pushed towards more centralization) especially for banks relying extensively
on hard information.
Table 5
Complementarities between ICT and credit scoring

Baseline Interaction
ICT_CAP 0.012*** 0.009***

(0.004) (0.004)
SCORING 0-2 -0.022 -0.019

(0.018) (0.018)
SCORING 3+ 0.070*** 0.176**

(0.024) (0.068)
ICT SCORING 3+ 0.020*

(0.011)
SIZE
-0.028*** -0.028***

(0.010) (0.010)
DISTANCE
-0.030* -0.031**

(0.016) (0.016)
LOAN SIZE
0.009 0.007

(0.023) (0.023)
BRANCH SIZE
0.070** 0.064**

(0.030) (0.030)
FE
YES YES
Observations
289 289
R-squared
0.29 0.30



Finally, we look for the existence of possible spillovers between ICT variables and the
adoption of credit scoring techniques. In Table 5 we report our baseline specification and an
additional column with the interaction term between ICT and credit scoring. The coefficients on
ICT_CAP and SCORING 3+ remain significant and with the expected sign. In addition, the
coefficient on the interaction term is positively and significantly associated to the degree of
decentralization. This means that the effects of ICT capital on decentralization are stronger for
those banks adopting credit scoring; on the other way round the impact of credit scoring is
enhanced when intermediaries have large ICT capital endowments. Therefore, multiple
complementarities between decentralization strategy, ICT endowments and scoring techniques
arise.

5.4 IV estimation
So far, our results suggest that there is a strong correlation between ICT variables and the
degree of power delegation to the LBM. Lagged values were used to avoid potential endogeneity of
some explanatory variables. However, there might still be some omitted determinants of power
delegation that are also correlated with the adoption of new technologies. Moreover, the direction
of causality could be reversed with respect to the one suggested by our empirical specification. To
address these endogeneity issues, we instrument ICT_CAP and SCORING. Specifically, we use the
lagged value of ICT_CAP as instrument for the current ICT endowment and the adoption of credit
scoring for either consumption credit or mortgages as instruments for the adoption of credit scoring
for small business lending. Lets discuss them in turn.
Our first instrument is ICT_ 95, the log of the ICT capital stock per employee in 1995. Using
lags of the explanatory variables as instruments is frequently seen as controversial. In particular,
instruments cannot be sufficiently correlated with the endogenous variable or there might be time
invariant omitted variables driving both the dependent variable and the lagged values of the
instruments. In our empirical framework, the first problem does not arise since our instrument is
highly correlated with the current endowments of ICT capital. As for the existence of time invariant
omitted factors, we are confident that this is not an issue in our case because of the length of the
time span and the peculiarity of the period considered. The Italian banking system underwent a
deep restructuring in the 90s including M&A that induced wide changes in the management of
each bank. Also the external conditions deeply changed and banks nowadays face an environment
largely unpredictable in the 90s.
28
Moreover, the ICT capital stock within a specific time span is
strongly related to ICT investments in that period because of the high depreciation rate of this type
of assets. ICT investments, in turn, are volatile and lumpy i.e. they concentrate in a relatively
short period of time and they are plausibly unrelated to the organizational structure observed ten
years later. As additional instrument we also introduce ICT_00, the log of the ICT capital stock per
employee for the year 2000. We exploit the fact that many banks made significant investments in
new software and hardware in that period because of the widespread concern that the information
system would cease operating between December 31, 1999 and January 1, 2000 the so-called
millennium bug. Once again, the determinants of these investments are plausibly unrelated with the
current allocation of decision-making power.

28
In the decade in question, there has been an increase in the size and in the geographical reach of the banks, the competition in the
local credit markets is augmented and the skill composition of the workforce is changed, thus largely modifying the environment in
which banks operate.

Table 6
IV estimates

I II III IV
ICT_CAP 0.029** 0.029** 0.011*** 0.011***

(0.012) (0.012) (0.004) (0.004)
SCORING 0-2 -0.026 -0.026 -0.022 -0.023

(0.025) (0.024) (0.024) (0.023)
SCORING 3+ 0.066* 0.064* 0.061 0.060*

(0.039) (0.033) (0.039) (0.033)
SIZE
-0.028** -0.028** -0.030*** -0.030***

(0.012) (0.012) (0.011) (0.011)
DISTANCE
-0.035* -0.034* -0.027 -0.027

(0.019) (0.018) (0.018) (0.018)
LOAN SIZE
0.001 0.001 0.010 0.010

(0.028) (0.028) (0.030) (0.030)
BRANCH SIZE
0.060* 0.061* 0.066** 0.066**

(0.033) (0.033) (0.029) (0.029)
FE
YES YES YES YES
Instrumental variables
ICT_ 95
SCORE_CC
ICT_ 95
SCORE_MO
ICT_ 00
SCORE_CC
ICT_ 00
SCORE_MO
Observations
263 263 278 278
R-squared
0.09 0.21 0.17 0.13

IV estimates. The dependent variable is PD that is the ratio between the loan size that a LBM can autonomously grant to a small firm
applying for credit and the corresponding amount that can be extended by the CEO. The key explanatory variables are ICT (log of
ICT capital stock per employee) and SCORING (dummies equal to 1 if the bank adopted small business credit scoring in the last 2
years or since at least 3 years, respectively). The lagged values of ICT (ICT_95 and ICT_00) are used as instrument for the actual
ICT endowment; the adoption of credit scoring for either credit consumption or mortgages (SCORE_CC and SCORE_MO,
respectively) are used as instrument for the adoption of credit scoring for small business lending. * significant at 10%; ** significant
at 5%; *** significant at 1%.

SCORE_CC is the instrument for the adoption of credit scoring for small business lending.
This is variable is built in the same way as SCORING with the further condition that the adoption
of credit scoring for consumption credit must have occurred in the same year or earlier that for
small business lending. The correlation between SCORE_CC and SCORING is due to the fact that
adoption of a scoring system for consumer credit historically preceded and hence facilitated the
introduction of these lending technologies in small business lending.
29
Data seem to support this
hypothesis. Apart from the indirect effect through SCORING, SCORE_CC has no direct effect on
PD. This assumption is based on the recognition that consumers and small firms have historically
been approached with two different lending technologies. In consumption credit the
creditworthiness is quickly assessed through standardized procedures and in an impersonal way. In
contrast, in small business lending it is usually emphasized the role of soft information and the
primacy of the loan officer who personally interacts with the firm. This is why credit scoring has
become first used for consumer credit decisions (automobile, credit card, and mortgage loans) and

29
The experience accumulated with credit scoring in the local branches may favor the extension of these techniques to other activities.
Prior adopters are also likely to have a more codified database that are necessary to implement a credit scoring mechanism.

it has taken longer to be adopted for business loans.
30
In sum, we believe that the motivations
driving the adoption of credit scoring for consumer credits are plausibly uncorrelated with the
determinants of power delegation to the LBM in business lending. Using a similar argument, we
use SCORE_MO adoption of credit scoring for mortgage as a further instrumental variable.
The IV estimates are reported in Table 6. As mentioned before, the instruments proposed are
strongly correlated with our potentially endogenous explanatory variables. The first stage F-
statistics are above the rule-of-thumb of 10 traditionally used to assess the strength of instruments.
The results clearly confirm previous findings that ICT adoption gets along with a larger
decentralization in small business lending. Estimated parameter for SCORING is only marginally
affected by the use of instrumental variables. As far as the estimated coefficient for ICT_CAP is
concerned, it is (if any) upward revised and it remains highly significant.

6. Conclusions
Over the past decades, advances in information technology have transformed the production
process and the workplace organization. As far as we know, this paper represents the first attempt
at examining the impact of the new technologies on the internal organization in the banking
industry.
The issue is relevant for two main reasons. First, the allocation of decision-making power
across hierarchical layers is one of the distinctive feature of an organization. Specifically, the
degree of autonomy of the LBM is the key variable as far as small firms credit access is
concerned. Second, the banking industry is particularly interesting since it is nowadays one of the
most intensive adopters of ICT. Moreover, factors like asymmetries of information and agency
costs underlined by the theories on firm internal organization are likely to have a special relevance
in the banking sector.
We derive a simple model of organization design that yields an estimable equation about the
relation between technology adoption and degree of delegation in lending activity. This relation has
been empirically investigated using a unique and specifically tailored dataset including about 300
Italian banks. We find that banks equipped with more ICT capital and resorting to credit scoring
delegate more decision power to their LBMs. These results are robust to many additional controls,
including instrumental variable estimation.
Hence, worries about the introduction of the new technologies and the negative effects on
small business credit access seem to be exaggerated. Indeed, ICT adoption has favoured delegation
of the decision rights to the LBMs. Our interpretation is that the lessening of agency problems
(leading to more delegation) due to the adoption of new technologies more than offset the
efficiency gains in transmitting and processing information (that, in turn, would push towards more
centralization). It also likely that computers and related technologies has substituted the LBM in
carrying out standardized and manual activities, leaving her free to focus on border lending
practices. Decentralization may imply that, de facto, a larger number of lending practices are
decided at the branch level, where relevant information resides. Granting more responsibilities to
the LBMs can promote their initiative and effort in collecting and act on soft information. All these
facts might have relevant consequences in terms of credit access for small firms, especially those

30
In the U.S., the first country where credit scoring was adopted during the fifties, these techniques concerned mainly mortgage loans
and credits of small size. Only in the nineties these procedures have become applied also to small business lending activity. The
adoption of credit scoring in Italy started much later. However, also in Italy the adoption concerned first mortgage loans and
consumer credit and only after small business lending even though the time gap is much more restricted. See Bofondi and Lotti
(2005) for an analysis of the diffusion of credit scoring in Italy. See Albareto et al. (2008) for more recent evidences.

with a limited operating history or that are more opaque. It is also reasonable to expect, in the near
future, an increase of the degree of decentralization due to the wide adoption of credit scoring in
recent years and the continuing progress in ICT industry. Quite evidently, all these considerations
hold true in those circumstances in which the LBM has actually better access to proprietary
information about borrowers quality with respect to CEO and she is also better able to act on that
information.
As a task for the future research agenda, it could be interesting to elaborate an empirical
strategy aimed at identifying the different channels through which ICT may affect decentralization.
Moreover, it could also be of interest an attempt at analysing the heterogeneity of the strategies
followed by the banks in the adoption of credit scoring methods and how these reflect on credit
allocation.

APPENDIX

Table A1
Descriptive statistics

Variable Description (1) Mean St. dev.
PD
Ratio between the loan size that a LBM can autonomously grant to
small firm applying for credit and the corresponding amount that can be
extended by the CEO (2006)
0.15 (0.144)
PD (absolute
level)
Log of the loan size that a LBM can autonomously grant to small firm
applying for credit (2006)
3.90 (1.775)
ICT_CAP
Log of ICT capital stock per employee -5.54 (1.616)
SCORING 0-2
Dummy equal to 1 if the bank adopted small business credit scoring in
the last 2 years
0.35 (0.477)
SCORING 3+
Dummy equal to 1 if the bank adopted small business credit scoring
since at least 3 years
0.23 (0.419)
ICT_ 95
Log of ICT capital stock per employee (1995) -5.67 (1.309)
ICT_ 00
Log of ICT capital stock per employee (2000) -5.43 (1.435)
SCORE CC 0-2
Dummy equal to 1 if the bank adopted credit scoring for credit
consumption in the last 2 years and before or in the same year of the
adoption of credit scoring for small business lending
0.15 (0.356)
SCORE CC 3+
Dummy equal to 1 if the bank adopted credit scoring for credit
consumption since at least 3 years and before or in the same year of the
adoption of credit scoring for small business lending
0.22 (0.412)
SCORE MO 0-2
Dummy equal to 1 if the bank adopted credit scoring for mortgages in
the last 2 years and before or in the same year of the adoption of credit
scoring for small business lending
0.18 (0.381)
SCORE MO 3+
Dummy equal to 1 if the bank adopted credit scoring for mortgages
since at least 3 years and before or in the same year of the adoption of
credit scoring for small business lending
0.23 (0.421)
SIZE
Log of total assets 20.69 (1.476)
DISTANCE
Log of average distance (in kilometers) between the headquarter of the
bank and the local markets where the bank has at least one branch. The
distance is weighted by the amount of loans borrowed in the market
where the local branches are situated
3.10 (0.895)
LOAN SIZE
Log of the average loan size in the banks portfolio 11.57 (0.548)
BRANCH SIZE
Log of the number of employees per branch 1.66 (0.350)
ROA
Returns on assets 0.01 (0.008)
BAD LOANS
Ratio of bad loans on total lending 0.04 (0.044)
NUMLEV
(Log of the) number of hierarchical levels between the LBM and the
CEO (2006)
1.19 (0.355)
TURNOVER
(Log of the) average permanence (in months) of the LBM within the
same branch (2006)
3.71 (0.430)
CEO EDU
Dummy equal to 1 if the CEO has a university degree 0.38 (0.487)
CEO AGE
Age of the CEO 53.48 (6.123)

Source: Supervisory Report of the Bank of Italy, Survey on Banking internal organization.
(1) Data refer to 2003 if not otherwise specified.


Table A2
Survey on Banking internal organization

Question 1: Hierarchical levels
Consider the lending activity to small business. This activity
may involve several figures of the banks internal organization.
Please list all these figures starting from the one at the bottom of
the hierarchy (e.g. loan officer) up to the board of the directors. If
the bank belongs to a group, list also the figures of the group who
may participate at the lending process.

Description of the hierarchical levels:
1) ____________________________________________
2) ____________________________________________
3) ____________________________________________
4) ____________________________________________
5) ____________________________________________
6) ____________________________________________
7) ____________________________________________
8) ____________________________________________
9) ____________________________________________
10) ____________________________________________
11) ____________________________________________
12) ____________________________________________

Question 2: Power delegation
Consider the lending practice for a firm applying for credit
for the first time. Please write down the maximum amount of
loan that can be granted in autonomy by each of the figure
indicated. Levels of delegation had to be referred to applicants
exhibiting a risk level that a bank judge a priori as normal.

Maximum amount (in thousands of euros):
________________
________________
________________
________________
________________
________________
________________
________________
________________
________________
________________
________________

Source: Survey on Banking internal organization.

Figure A1
The distribution of absolute power delegation of LBM

0,0
5,0
10,0
15,0
20,0
25,0
0 1-20 21-50 51-100 100-200 201-500 >500

The x-axis reports classes of absolute power delegation of the LBM (in thousands of euro); the y-axis the frequencies.
Source: Survey on Banking internal organization


Table A3
Matrix of correlation between the explanatory variables

ICT_
CAP
SCOR
ING
3+
SIZE
DIST
ANCE
LOAN
SIZE
BRANC
H SIZE
ROA
BAD
LOAN
S
NUML
EV
TUR
NOV
ER
ICT_CAP

SCORING
3+
0.118
(0.044)
SIZE 0.159 -0.018
(0.007) (0.754)
DISTANCE 0.146 0.055 0.787
(0.013) (0.349) (0.000)
LOAN SIZE -0.054 -0.032 0.208 -0.084
(0.363) (0.582) (0.000) (0.152)
BRANCH
SIZE
0.171 -0.050 0.595 0.407 0.268
(0.004) (0.392) (0.000) (0.000) (0.000)
ROA -0.037 0.020 0.235 -0.025 0.301 0.272
(0.531) (0.729) (0.000) (0.665) (0.000) (0.000)
BAD LOANS 0.097 0.012 0.016 0.169 0.301 0.049 -0.114
(0.102) (0.843) (0.788) (0.004) (0.000) (0.407) (0.054)
NUMLEV 0.170 -0.021 0.622 0.503 0.086 0.360 0.194 -0.024
(0.004) (0.725) (0.000) (0.000) (0.137) (0.000) (0.001) (0.684)
TURNOVER 0.032 0.058 -0.293 -0.318 0.004 -0.143 0.150 0.073 -0.336
(0.598) (0.337) (0.000) (0.000) (0.950) (0.019) (0.013) (0.231) (0.000)
The p-values are reported in brackets.

REFERENCES
Acemoglu, D., P. Aghion, C. Lelarge, J. Van Reenen and F. Zilibotti (2007), Technology,
information and the decentralization of the firm, Quarterly Journal of Economics, 122:
1759-1799.
Aghion, P. and J. Tirole (1997), Formal and real authority in organizations, Journal of Political
Economy, 105: 1-29.
Akhavien, J., W.S. Frame and L.J. White (2005), The diffusion of financial innovations: An
examination of the adoption of small business credit scoring by large banking
organizations, Journal of Business, 78: 577-596.
Albareto, G., M. Benvenuti, S. Mocetti, M. Pagnini and P. Rossi (2008), Lorganizzazione
dellattivit creditizia e lutilizzo di tecniche di scoring nel sistema bancario italiano: risultati
di unindagine campionaria, Questioni di Economia e Finanza, 12, Bank of Italy.
Autor, D.H., F. Levy and R.J. Murnane (2003), The skill content of recent technological change:
An empirical exploration, Quarterly Journal of Economics, 118: 1279-1333.
Berger, A.N. (2003), The economic effects of technological progress: Evidence from the banking
industry, Journal of Money Credit and Banking, 35: 141-176.
Berger, A.N. and G.F. Udell (2002), Small business credit availability and relationship lending:
The importance of bank organizational structure, Economic Journal, 112: F32F53.
Berger, A.N. and R. DeYoung (2001), The effects of geographic expansion on bank efficiency,
Journal of Financial Services Research, 19: 163-184.
Berger, A.N., N.H. Miller, M.A. Petersen, R.G. Rajan and J.C. Stein (2005), Does functions
Berger, A.N., W.S. Frame and N.H. Miller (2005), Credit scoring and the availability, price, and
risk of small business credit, Journal of Money, Credit and Banking, 37: 191-223.
Bloom, N., L. Garicano, R. Sadun and J. Van Reenen (2009), The distinct effects of Information
Technology and Communication Technology on firm organization, NBER Working Paper,
14975.
Bofondi, M. and F. Lotti (2005), Innovation in the retail banking industry: The diffusion of credit
scoring, Review of Industrial Organization, 28: 343-358.
Bresnahan, T.F., E. Brynjolfsson and L.M. Hitt (2002), Information technology, workplace
organization, and the demand for skilled labor: Firm-level evidence, Quarterly Journal of
Economics, 117: 399-376.
Brynjolfsson, E. and L.M. Hitt (1998), Information technology and organizational design:
Evidence from micro data, mimeo.
Casolaro, L. and G. Gobbi (2007), Information technology and productivity changes in the
banking industry, Economic Notes, 36: 43-76.
Christie, A.A., M.P. Joye and R.L. Watts (2003), Decentralization of the firm: Theory and
evidence, Journal of Corporate Finance, 9: 3-36.
Colombo, M.G. and M. Delmastro (2004), Delegation of authority in business organizations: An
empirical test, Journal of Industrial Economics, 52: 53-80.

Feldman, R. (1997), Small banks and a big change in technology called credit scoring, Federal
Reserve Bank of Minneapolis, Region, 19-25.
Felici, R. and M. Pagnini (2008), Distance, bank heterogeneity and entry in local banking
markets, Journal of Industrial Economics, 56: 500-534.
Quaderno di Moneta e Credito, 245-265.
Geanakoplos, J. and P. Milgrom (1991), A theory of hierarchies based on limited managerial
attention, Journal of the Japanese and International Economies, 5: 205-225.
Gurbaxani, V. and S. Whang (1991), The impact of information systems on organizations and
markets, Communications of the ACM, 34: 59-73.
Hertzberg, A., J.M. Liberti and D. Paravisini (2009), Information and incentives inside the firm:
evidence from loan officer rotation, mimeo.
Hubbard, T.N. (2000), The demand for monitoring technologies: The case of trucking, Quarterly
Journal of Economics, 115: 533-560.
Jensen, M.C. and W.H. Meckling (1992), Knowledge, control and organizational structure Parts I
and II, in L. Werin and H. Hijkander (eds.) Contract Economics, Cambridge, Basil
Blackwell.
Jorgenson, D.W. and K.J. Stiroh (2000), Raising the speed limit: U.S. economic growth in the
information age, Brookings Papers on Economic Activity, 1: 125-235.
Liberti, J.M. (2005), How does organizational form matter? Distance, communication and soft
information, mimeo.
Liberti, J.M. and A. Mian (2009), Estimating the effect of hierarchies on information use, Review
of financial studies, forthcoming.
Papke, L.E. and J. Wooldridge (1996), Econometric methods for fractional response variables with
an application to 401(k) plan participation rates, Journal of Applied Econometrics, 11: 619-632.
Radner, R. (1993), The organization of decentralized information processing, Econometrica, 62:
1109-1146.
Rajan, R.G. and J. Wulf (2006), The flattening firm: Evidence from panel data on the changing
nature of corporate hierarchies, Review of Economics and Statistics, 88: 759-773.
Scott, J.A. (2006), Loan officer turnover and credit availability for small firms, Journal of Small
Business Management, 44: 544-562.
Uchida, H., G.F. Udell and N. Yamori (2008), Loan officer and relationship lending to SMEs,
Working Paper 2008-17, Federal Reserve Bank of San Francisco.
Udell, G.F. (1989), Loan quality, commercial loan review and loan officer contracting, Journal
of Banking and Finance, 13: 367-382.
Van Zandt, T. (1999), Real-time decentralized information processing as a model of organizations
with boundedly rational agents, Review of Economic Studies, 66: 633-658.
Zbojnk, J. (2002), Centralized and decentralized decision making in organizations, Journal of
Labor Economics, 20: 1-22.

BANK ACQUISITIONS AND DECENTRALIZATION CHOICES
Enrico Beretta
*
and Silvia Del Prete
**

1. Introduction
Banking consolidation has strongly intensified since the mid-1990s involving an increasing
number of intermediaries. The number of Italian banks decreased from 970 in 1995 to 806 in 2007;
over the same period, mergers and acquisitions (M&As) accounted for more than 70 per cent of
banks total assets. Furthermore, at the end of 2007 the top five Italian holding banks represented
more than 50 per cent of the banking systems total assets. One of the most common concerns is
that the consolidation process creates larger and more complex banks; therefore, it could favour
hierarchy versus decentralization, increasing the functional distance between lender and borrower
and reducing relationship lending (Alessandrini et al., 2005). Nevertheless, studies focusing on
M&As do not agree that the effect of consolidation on credit supply or post-dealing portfolio
diversification is unambiguous. Empirical analysis usually indicates that M&As among small banks
enhance small business lending, whereas the opposite effect emerges when large banks are
involved. The effect of consolidation may also diverge across geographical areas and types of deal.
Merged banks usually reduce the share of credit granted to SMEs while acquired banks show a
higher orientation towards small business lending, especially if they have a local geographical
reach (Beretta and Del Prete, 2007).
How could this heterogeneity in the consequences of M&As be explained? We believe that it
could be originated by changes in the internal organization of a bank following the consolidation
process, especially in terms of centralization versus decentralization choices. This intermediate
effect on organization has often been neglected by the empirical studies on bank M&As, mainly
due to lack of specific data: this is why bank size has usually been employed as a simple proxy of
bank complexity (Berger and Udell, 2002; Cerqueiro et al., 2007).
The link between consolidation and organizational re-shaping is crucial. According to the
theory, merged or acquired banks could suffer from scope diseconomies in retail services (such as
SME finance), discouraging decentralization because of increased agency costs (Aghion and
Tirole, 1997; Stein, 2002).
1
Decentralization implies higher agency costs in order to face loss of
control over local officers and to limit moral hazard behaviour. The trade-off between
decentralization and control defines the organizational design: during the restructuring phase that
follows a merger or an acquisition, it is likely that the allocation of lending decision rights will
change across hierarchical levels, in order to create a new equilibrium within the banks
organization.

*
Bank of Italy, Genoa.
**
We are grateful for useful suggestions to Elena Beccalli, Michele Benvenuti, Luigi Cannari, Barbara Chizzolini, Sauro Mocetti,
Marcello Pagnini, Matteo Piazza, Paola Rossi, and participants at seminars held at the Bank of Italy. Although this paper is the
result of all the coathours joint work, Enrico Beretta mainly contributed to Sections 3, 6 and 8, and Silvia Del Prete to Sections 1,
2, 4, 5 and 7.
1
In Steins model decentralized banks, whose local loan officers have more decision-making power, improve research incentives to
collect and use proprietary information on opaque borrowers.
78 Enrico Beretta and Silvia Del Prete

The purpose of our paper is to verify possible structural changes in the decentralization
versus centralization choice following banking acquisitions.
2
These changes may especially affect
small business lending, for which it is more difficult to transfer information across a hierarchy.
Using a unique data set on banking organization, we focus on the internal loan approval mechanism
concerning SME financing. We then investigate whether Italian bank acquisitions affect power
delegation across hierarchical levels and personnel turnover. We also investigate how the degree of
dissimilarity among bidder and target banks can hamper the integration process and favour
organizational complexity versus decentralization. This can lead to pinpointing the mechanism at
work to shed light on the mixed results of banking consolidation on relationship lending as well.
3

To the best of our knowledge there are no empirical studies examining whether banks
involved in M&As experience significant changes in their internal structures and small business
lending strategies: our paper is a first attempt to go in that direction.
Organizational data are drawn from a special survey conducted at the end of 2006 by the
Bank of Italy on a sample of around 300 banks. The sample represents almost the entire universe of
domestic banks, excluding intermediaries specialized in consumer credit, leasing and factoring, and
Italian branches of foreign banks. We combine such data with census and balance sheet indicators.
Our main results are the following. Acquisitions positively affect the amount of loans the
branch manager of the target bank can autonomously grant to SMEs. The power delegated to the
Chief Executive Officer (CEO) shows a greater increase. These interventions tend to smooth the
differences between the dealing partners. Within acquired bank organizations, these adjustments
engender a systematically different distribution of decision rights across hierarchical levels. At the
same time, we find that branch officers turnover is higher for banks involved in acquisitions. Since
the local managers decisional power is positively affected by the deal, we cannot support the
common concern that the consolidation process reduces decentralization. We argue that the higher
turnover of local officers is mainly related to monitoring the costs of larger organizations
associated with the increasing decision-making power delegated to local branch managers.
The paper is organized as follows. Section 2 presents a review of the literature concerning
the main effects of bank consolidation on relationship lending and related organizational
implications. Section 3 describes the data and variables employed in the econometric exercise
while Section 4 sets out the empirical methodology. Section 5 presents the baseline results;
Sections 6 and 7, respectively, report more detailed estimates and robustness checks, and Section 8
concludes.

2. Related literature and research questions
Our paper is related to two strands of the literature: on the one hand, those studies analysing
the empirical consequences of M&As; on the other, papers examining agency problems in
delegating power across hierarchical levels.

2
We focus on acquisitions rather than mergers because we still have the chance to distinguish between target and bidder,
characterized by a specific organizational framework. Our principal interest is, moreover, to study the changes of power delegation
in target banks; this would not be possible in the case of a merger, because the target disappears after the deal.
3
We do not investigate the effect of organizational changes on credit supply, because this further link has already been examined by a
large strand of literature, summarized in the following section. Concerning the relation between more decentralized banking
organizations and the greater extent of relationship banking, see Benvenuti et al. (2009). The relation between organizational
framework, business specialization and the economic performance of a bank is empirically analysed for the Italian banks, in Bongini
et al. (2007).
Bank acquisitions and decentralization choices 79

With respect to the first, many researchers have studied the impact of M&As on banking
performance and credit availability. Some studies have documented a reduction in credit supply to
small borrowers, due to the growing organizational complexity of banks (Berger et al., 1995; Peek
and Rosengren, 1996; Udell, 1998; Scott and Dunkelberg, 2003). For Italian banks, Focarelli et al.
(2002) show that a merger usually causes a greater share of credit to be granted to large companies,
persistent in the long-run, while this effect is not always found after an acquisition. Credit rationing
has been better quantified by considering the geographical extent of bank activity or by
distinguishing results by bank size (Keeton, 1996; Peek and Rosengren, 1998; Avery and Samolyk,
2000; De Vincenzo and Iannotti, 2002; Sapienza, 2002; Beretta and Del Prete, 2007). Other
empirical findings have supported the idea of the dynamic effects of M&As which, in the long-run,
stimulate the role of local and new banks and only lead to a temporary reduction in credit supply to
SMEs
4
(Berger et al., 1998; Berger et al., 2001; Berger et al., 2000; Keeton, 2000; Bonaccorsi di
Patti and Gobbi, 2001; Bonaccorsi di Patti and Gobbi, 2007).
However, these studies do not provide evidence on the root cause of the changes in
banking credit policy after M&As. Moreover, mixed results of M&As between large and small
banks underline the existence of a crucial link between banking organization, lending technology
and specialization in SME finance. Since in this literature organizational changes are latent
variables, size and geographical reach are widely employed as proxies of banking organization,
even if these proxies might not be fully satisfactory.
The second strand of the literature can shed further light on this topic. When firm size
enlarges, agency and information costs tend to develop more than proportionally (Berger and
Udell, 2002). Larger intermediaries, stemming from consolidation, may experience diseconomies
la Williamson (1988),
5
because of higher coordination costs in lending during the transition period
(especially if information is innately soft, as for SMEs); as a consequence, organizational
diseconomies can shrink the scope of large banks (Cerqueiro et al., 2007).
Agency problems could be important in explaining the different results of M&As. When two
or more very dissimilar institutions consolidate, they face higher integration costs due to
differences between dealing partners in terms of corporate culture, performance, information
technology, lending practices and workplace environments. In such cases, the cost of control could
increase and cause some difficulties for target banks in sustaining previous credit relations or
performance. Consequently, parent banks have to cope with an organizational trade-off between
delegation and control over the target intermediary.
Agency models highlight how organizational shocks, mainly caused by M&As, could affect
in different ways incentives for local loan managers to set up relationship lending and gather soft
information. Following Steins framework, within larger banks information has to go through
several hierarchical layers (organizational complexity); thus the transmission of proprietary (soft)
information upward could be difficult, because it cannot be directly verified by anyone other than
the agent who produces it. This means that in more complex organizations, CEOs are less efficient
in creating incentives and in distributing funding across branches to finance more opaque firms.
There are many reasons for this: more levels of bureaucracy, lower motivation for line managers,
greater agency or moral hazard problems (which arise when the decision-making power is
separated from expertise on information). Stein (2002) argues that, when soft information is

4
If loans that are adversely affected have a positive net present value, the consolidation process may generate business opportunities
for rival banks and modify their lending organization. This may also happen because local banks benefit from their knowledge of
the economic environment and from their comparative advantage in gathering soft information.
5
Studies on banking performance after M&As (Focarelli et al., 1999 and Del Prete, 2002) suggest diversification revenues following
a merger and portfolio restructuring after an acquisition; they confirm that there are few cost savings, due to rigidity in personnel
and organizational structures, both for large and mutual banks.

crucial, decentralized banks are more efficient in financing smaller firms. This model, based on
agent incentives and soft information, could explain why M&As involving small banks generally
have a lower impact on small business lending.
Consolidation can also generate greater organizational complexity and hence higher costs for
monitoring local managers; therefore the local managers tenure at the same branch could be lower
(Ferri, 1997), providing fewer incentives to establish face-to-face relationships with SMEs.
Organizational shocks after a deal could be influenced by technological innovation, too. The
implementation of innovative internal rating systems should reduce the disadvantage of large and
centralized banks in the treatment of SME information (Akhavein et al., 2005; Berger et al., 2005a;
Berger and Udell, 2006). These techniques may also mitigate agency problems, since their adoption
in the evaluation of creditworthiness can also facilitate the remote control of lending decisions by
local managers. Evidence suggests that large and consolidated banks show a higher share of credit
granted to SMEs if they have adopted rating or scoring systems (Frame et al., 2001) or a closer
correlation between interest rates and the risk of default (Panetta et al., 2004).
These theoretical predictions have been tested using the link between size and hierarchical
complexity (Berger et al., 2005b) and by case studies. Liberti (2003 and 2005), using data on a
large banks loan approvals, argues that transmission of, and reliance on, soft information is higher
under decentralized than centralized institutions. This is consistent with the findings of Liberti and
Mian (2006), who show that soft information depreciates across hierarchical levels and that
whoever is given more authority puts more effort into collecting soft information.
To our knowledge, no empirical evidence is at present available about the effect of M&As
on banks organizational features. In this paper, we take up the following question: Do bank
acquisitions influence organization design (centralization/decentralization) and potentially affect
local loan officers incentives?. We also try to test whether Steins hypothesis of a trade-off
between delegation and control is true. To this end, we analyse banks entering a group following an
acquisition, so as to corroborate and better explain the mixed effects of consolidation.

3. Data and variables
3.1 Dependent variables: organizational features
Organizational data are drawn from a survey conducted by the Bank of Italy on a non-
random sample of about 300 banks, depicting lending practices at the end of 2006 and representing
around 80 per cent of the Italian banking systems total loans to firms. Intermediaries specialized in
consumer credit, leasing and factoring financing were excluded, as well as all Italian branches of
foreign banks. The survey covers almost the whole spectrum of banks above a minimum threshold
in terms of size and organizational complexity, including virtually all large, medium and small
banks, and excluding very small intermediaries. The accuracy of the data collected is reasonably
high.
6

In order to compare two homogenous groups of intermediaries, we have excluded from the
sample those banks that underwent a merger between 2000 and 2005. Therefore, the econometric
analysis compares banks that have experienced an acquisition, either as bidder or target, and banks
that did not take part in any deal during the same period. We have also excluded outlier
observations and banks that only finance long-term investment projects.

6
Preliminary interviews with bankers helped in organizing the questions properly, avoiding potential ambivalences; ex post
interviews helped to fill missing information and to fix erratic responses. For more details, see Albareto et al. (2008).

The survey focuses on the organizational design of business lending and, secondly, on the
implementation of rating systems. The information about bank organizational structure is very
analytical. Banks were asked to give information about: the number of hierarchical layers involved
in the decision to grant loans to SMEs, the amount of finance up to which each hierarchical level
may lend autonomously (from the branch loan officer up to the board of directors), the type of
information required (soft versus hard information), the frequency of loan officer turnover, the
importance of credit scoring and internal rating methodologies in the assessment of credit.
Our delegation indexes are equal to the amount of loans the local manager and the CEO are
respectively allowed to grant to SMEs (logarithms).
7
The first indicator provides information about
the power of the periphery to grant credit autonomously, the latter informs about the authority of
the top management. We have chosen the General Manager to embody the CEO because this
hierarchical level is always present. The branch loan officer and the top management can
respectively represent the agent and the principal indicated in the theoretical literature (Stein,
2002).
8

The second step involved drawing information from the survey sample on loan officers
turnover to assess whether the length of tenure managing a branch (in terms of the number of
months) is affected by involvement in M&As. On the one hand, establishing close and long-lasting
relations with opaque borrowers should be directly linked to the stability of local managers. On the
other hand, the longer the time a manager works in the same branch, the higher the likelihood of
moral hazard behaviour. The loan officers turnover is intrinsically related to relationship lending,
but it is also a device for the top management to reduce monitoring costs, which increase when
power must be delegated. This is the case, for instance, in financing opaque SMEs, for which
information is not easily transferable upwards.
Descriptive statistics on the sample employed for the baseline estimates are presented in
Table 1, with variable definitions. On average, a branch manager can grant loans up to 136,000
autonomously; the median is considerably lower (70,000), since the upper bound is lower for
small and mutual banks. The power delegated to the CEO is about 2.4 million; again the median
value is significantly lower (0.5 million). Local officers stay in charge for an average of 46
months; the distribution is relatively narrow: the first quartile is 35 months, the third is 60.
Table 2 reports descriptive statistics on organizational features across the banks involved
(either as bidder or target) or not involved in acquisitions. By and large, acquiring banks are large
intermediaries; their delegation indexes are higher than the sample mean. Their local officers show
a faster turnover, most likely because of the extent of their branch network. Target banks in
acquisition deals are larger than sample mean, too. Their officers have considerable delegated
power (albeit, in median, less than the power of officers in bidder banks) and a fast turnover. Banks
not involved in the consolidation process are largely small and local: this contributes to explain the
lower degree of power delegated to their local officers and their slower turnover, too.

3.2 Dummies accounting for bank acquisitions
In order to capture systematic differences in lending strategy following an acquisition, we
build a dummy for each kind of dealing partner (bidder or target banks); it is equal to 1 for each

7
The threshold under which a borrower can be classified as an SME varies from bank to bank; the most frequent classification term
is the total of net sales.
8
In some unreported estimates, we have also experimented an index of internal decisional decentralization, calculated for each bank
i, as the ratio between the maximum amount of finance up to which a branch loan officer may lend autonomously and the CEO can
lend (internal decentralization).

bank involved in a deal, and zero otherwise. Information on acquisitions is drawn from the Bank of
Italys Register of Banking Groups.
9

We try to overcome the difficulties of trying to capture a dynamic phenomenon in a
cross-section estimate by building two sets of dummies: these distinguish recent acquisitions
(between 2003 and 2005, new) from older ones (between 2000 and 2002, old). Disentangling
operations over time allows us to verify whether there are temporary effects or not: we expect
different restructuring results depending on the time these deals took place.
Another interesting feature to investigate is the different effect of deals among
geographically contiguous banks from those involving distant intermediaries. With this aim, we
interact the dummy identifying a target bank with dummy variables measuring the geographical
reach between dealing partners. The first (Overlap) refers to the overlapping in branch networks of
the target and its bidder: it is equal to 1 if more than 50 per cent of the branches of the target is
located in provinces in which the bidder has also branches,
10
and zero otherwise; at the same time
NoOverlap is its orthogonal dummy in order to split target banks into two homogenous groups. A
second set of dummies, Inmkt and its orthogonal Outmkt, refers to the legal headquarters of dealing
partners: dividing Italy into two areas (North and Centre; South), the dummy Inmkt is equal to 1 if
the target is headquartered in the same area of the bidder, and zero otherwise. It is the opposite for
the dummy Outmkt.
Finally, to control if the effects of an acquisition are different for large and small banks, we
interact the dummy target bank with a dummy Small (Large) bank, which is equal to 1 if at the
end of 2006 the total assets of the bank were less (more) than 1.3 billion, and zero otherwise.

3.3 Other controls
In our econometric exercise, we control for other bank-level features (average loan, portfolio
risk, profitability, banks belonging to banking groups, geographical location of the headquarters,
presence in industrial districts), potentially able to affect bank organization.
The log of total assets is our size variable; as already seen, banks involved in acquisitions
(either as bidder or target) are larger than others (Table 2). Our specialization index in small
business lending is measured as the average size of loans to firms and households; for this feature,
there does not seem to be a strong variability between the different groups of banks. The ratio
between profits before taxes and total assets (ROA) informs us about bank profitability: as
expected, the profitability of acquiring banks is higher than both target and independent banks. The
share between non-performing loans and total loans is our proxy for the bank portfolio credit risk;
also in this respect, bidders perform better than targets, and rivals are in the middle. Data used to
build previous bank-level characteristics are drawn from the Bank of Italy Supervisory Reports.
Our control for local environment is the district index, constructed for each bank as the
share of branches located in municipalities classified as industrial districts according to Istat
taxonomy. Our hypothesis is that banks with wide activity in district areas will be encouraged to set
up close relationships with local SMEs; this may enhance the importance of soft information and
promote decentralized organizations. The index is on average 36 per cent, and it is higher for stand-
alone banks.

9
Information on merged banks, instead, is drawn from the Census of Banks (SIOTEC) and it is employed to cut them out of the
sample.
10
The distribution of this ratio is polarized between very small and very large values; the threshold of 50 per cent seems to represent a
satisfactory cut-off to split the sample into the two groups.

A control for the banks geographical location is provided by the dummy South, that is equal
to 1 if the intermediarys headquarters are located in the South of Italy. With respect to governance,
the dummy Group is equal to 1 if, at the beginning and at the end of the sample period, the bank
was part of a bank group and it has not been acquired in the meantime.
Other variables contribute to define bank organizational design in lending to SMEs. First of
all, organizational complexity depends on the ability to process information. For this purpose, we
distinguish between banks that have adopted a credit scoring system and intermediaries which have
not.
11
Moreover, using the pecking order on sources of information about borrowers (financial
statements, credit registers, qualitative data, etc.), we operate a distinction between intermediaries
giving great importance to soft information and banks for which this kind of information is less
important.
12

4. Empirical methodology
An acquisition can generate changes in the degree of hierarchy and in the decentralization of
the decision-making process of the target bank, due to the need to merge the different cultures,
information systems and know-how of the target and the bidder. The latter may experience some
organizational change, too: for instance, some managers could be temporarily employed in the
acquired banks structure.
13
If this is the case, by comparing banks that have experienced
acquisitions in a six-year period before 2006 with those which were not involved in any deal, we
should be able to capture different organizational strategies.
Therefore, the purpose of our empirical analysis is to verify to what extent the organization
of SME loan approval of bank i is explained by its involvement in consolidation. To this end, we
employ two sets of regressions to assess if bank acquisitions affect the decision-making rights of
both the loan officer and the CEO. A third econometric exercise investigates the effect on the loan
officers turnover, in order to capture both branch managers incentives to finance small firms and
headquarters monitoring strategy. In fact, while the delegation system sheds lights on formal
authority, turnover conditioning local officers real decision-making power is likely to represent
a better proxy for the agents incentives in relationship lending (Scott, 2006) and for monitoring the
costs of large institutions (Ferri, 1997).
The equations to be estimated have the following form:
LO Delegation
i
= f (Dummy Acq
Bidder
, Dummy Acq
Target
, X
i
) (1)
CEO Delegation
i
= f (Dummy Acq
Bidder
, Dummy Acq
Target
, X
i
) (2)
LO Tenure
i
= f (Dummy Acq
Bidder
, Dummy Acq
Target
, X
i
) (3)
We capture acquisition events by means of the dummy variables described in the previous
section (Dummy Acq); moreover, we introduce controls for individual characteristics (vector Xi),
such as bank size, average loan size, credit portfolio risk, profitability, governance (banks
belonging to a group), geographical localization of the headquarters, and presence in district areas.
All bank-level variables aim to control for differences across intermediaries, potentially causing
heterogeneity in organizational complexity and in lending technologies.

11
With respect to target banks, we examine if the scoring was introduced after the acquisition.
12
The indicator referring to the importance of soft information is based on the ranking of two factors: qualitative information and
personal knowledge of the customer.
13
Albareto et al. (2008) show first descriptive evidence from the survey conducted by the Bank of Italy (see Section 3). This study
emphasizes the difference, both in branch officers delegated power and turnover, between local banks belonging to groups and
other stand-alone local banks.

The analysis is carried out with an OLS estimate on a cross-section of 235 banks that
provided information on their organization in lending to large and small firms. Our main findings
are checked by using splitting samples and matching methods. Each dependent variable is
measured at the end of 2006, while regressors are one-period lagged, or taken as the mean over the
whole six-year period, to tackle endogeneity.

5. Results
5.1 Delegation of power across hierarchical levels
In the first step of our econometric exercise, we aim to investigate the distribution of power
delegation across hierarchical layers among banks experiencing acquisitions and those not involved
in any deal. To this end, our dependent variable is defined as the maximum amount that a local
branch officer can autonomously grant to SMEs. At the same time, we also check the
decision-making power of the CEO, in order to clearly disentangle the different impacts of the
reorganization both at the top management and at the branch level within a target bank.
Furthermore, the simultaneous analysis of these two measures of (formal) authority for each
hierarchical level shows which variable drives bank centralization-decentralization choices
concerning small business lending in the post-acquisition period. In order to implement robustness
checks on our results, we start with a very parsimonious model and then, following a stepwise
approach, we insert other bank-level controls in the estimation. Findings from the different models
and our baseline specification are shown in Table 3.
We find that acquisitions positively affect the powers delegated both to the loan officer and
to the CEO for loan approval; the effect on the latter is higher than on the former.
14
Consequently,
being acquired in the period 2000-2005 improves formal decision-making power at both
headquarters and branch level within a target bank. This result is very significant, especially in our
baseline specification (model IV), which controls for many bank characteristics. The coefficient of
the dummy target suggests that the power delegated is more than double for acquired banks with
respect to those not involved in any deal in the same period. Therefore, having recently become
part of a group causes empowerment along hierarchical layers in order to move towards the higher
levels of the bidder. This evidence is consistent with the strand of the empirical literature
supporting the idea that target banks specialize more on small business loans after becoming part of
a group (Beretta and Del Prete, 2007).
15

The increasing delegation of power across all the organizational levels of the target bank
may have different causes. One possibility is that the CEO of the parent bank will appoint a new
top management with higher powers to enhance its supervision. Even if no removal strategy is
followed, another hypothesis is that the power delegated to the CEO of the target bank may be
increased in order to match the higher level of the bidders CEO, and to strategically merge
different loan approval mechanisms. The real determinant may depend upon on the main

14
This is why in an unreported regression, using the ratio between LO and CEO powers as a dependent variable, we obtain a negative
correlation with the dummy identifying a target bank, suggesting a lower internal decentralization degree. Consequently, it does not
seem to provide evidence of a decline in the decision-making power of banks newly belonging to banking groups, but only of a
major functional distance, in terms of decision-making power, between the vertex and the periphery in the lending process of
target banks.
15
Our results also seem to suggest that acquiring banks show a lower delegation of power in lending to SMEs for both top
management and loan officers. Even if this evidence is not statistically significant, it is likely that the increasing complexity after the
deal may reduce their comparative advantage in treating soft information, by raising coordination costs across branches and
headquarters. All these changes within a banking group may favour higher specialization on the part of acquiring banks in lending to
large and medium-sized firms and a prominent expertise of their targets in SME finance.

motivation of the deal: when acquisition is driven by the desire to restructure the target, the former
hypothesis is more likely, otherwise the latter is more plausible.
Among the other controls, bank size plays an important role in explaining the different
degrees of delegation across hierarchical levels. As expected, bank size is positively correlated with
the amount of credit that CEOs and branch officers can autonomously grant to small firms; in fact,
delegation to these two management layers is set higher by large and complex banks. Similar
effects are determined by the governance scheme: a bank belonging to a group, even if not acquired
recently, shows higher decision-making rights at both hierarchical levels, in order to foster size
growth.
Power delegation seems not to be affected by bank risk and profitability. Instead, it is lower
if the banks headquarters are located in a Southern region, a geographical area where firms are
riskier due to higher local diseconomies and negative externalities.
16
Finally, we find that local loan
officers decision-making power is positively affected by bank specialization in Italian district
areas (i.e. agglomerations of SMEs): the greater the share of district branches, the greater the power
assigned to local managers. This result is in line with the evidence that district areas create
information spillovers and increase opportunities for local managers to establish face-to-face
relationships with small firms to collect soft information more easily.

5.2 Branch loan officers turnover
In the second step, we measure to what extent the local loan officers length of tenure at the
same branch (number of months, in logarithms) is influenced by the banks involvement in
acquisition. The results of this set of estimates are presented in Table 3, which shows a stepwise
approach as in the previous section.
As pointed out above, this indicator is affected by the trade-off between the relationship
lending approach and the moral hazard problem. On the one hand, the longer the time a local loan
officer manages the same branch, the greater the personal incentive to develop close relationships
with opaque borrowers and to collect soft information (Uchida et al., 2006).
17
We find that bank
profitability lengthens the time spent by loan officers at the same branch, while portfolio risk
increases their turnover. This evidence supports the fact that bank profitability and risk could affect
loan officers incentives to develop relationship lending; in fact, bank performance is often adopted
as a scale to define local managers compensation schemes.
On the other hand, more complex organizations cost more to control: reducing the length of
tenure of the branch officer is an effective strategy to reduce agency problems, even if this is in
conflict with relationship banking. Our results show that size is negatively correlated with loan
officer length of tenure, as is belonging to a banking group. These findings are consistent with
previous evidence on Italian banks, supporting a greater turnover of loan officers in large banks
with respect to small ones, especially mutual banks, because of growing monitoring costs (Ferri,
1997).
Moreover, banks that grow in size following acquisitions face higher integration and agency
costs. In line with our suppositions, the turnover of local loan officers is greater in acquired banks.

16
The effect of the dummy South is statistically significant relatively to CEOs decision-making right, while for loan officer
delegation we get the expected sign but the result is not robust: this evidence is probably due to the fact that the monitoring is firstly
on top management behaviour, mainly if the acquisition is driven by restructuring the target.
17
These findings are consistent with Scott and Dunkelberg (1999), who emphasize the primacy of the loan officers role in the
relationship paradigm. In fact, by examining the effect of US banking industry consolidation on SME finance, they argue that a loan
officers turnover is negatively correlated with credit availability and positively correlated with the SME searching for a new bank.

In particular, from our baseline specification, we can argue that these officers stay in the same
branch for a period lasting around 21 per cent less than others not involved in any deal. This
evidence could offer insight into the fact that the greater turnover is actually a method of reducing
agency costs by monitoring loan officers moral hazard risks within more complex banks, above all
if delegated powers increase. Our results are consistent with the findings provided by Udell (1989),
who argues that those banks delegating more responsibility to their loan officers, have to invest
more in monitoring their outcomes.
18

6. Further specifications
6.1 Disentangling different lending techniques
As already seen in the literature review, the preferred way of processing information and,
more specifically, soft information may influence the organizational design, and in particular
variables such as loan officers decision-making power and turnover. In the spite of better
qualifying our baseline results on the effect of acquisitions, we look at two other organizational
characteristics: the adoption of scoring systems in the assessment of customers creditworthiness
and the importance given by the bank to soft information.
With respect to the use of credit scoring, we add to the baseline model a dummy Score
signalling the implementation after the acquisition, in the case of target banks of an internal
rating system concerning SME finance. In order to capture special effects due to reorganization
after the deal, we also interact the dummy Score with that identifying an acquired bank. The
results, for each of the three dependent variables examined in the previous section, are shown in
Table 4.1: the dummy Score is positive but not significant, while the parameter of its interaction is
negative and statistically significant in explaining the loan officers decision-making power. This
suggests that for banks that widely employ credit scoring models, the post-acquisition growth of
the local managers autonomy is less important. This is an expected result, because the use of these
hard information lending techniques should be a way for a parent bank to strictly monitor local
officers outcomes within the newly acquired intermediary.
In order to analyse the importance a bank gives to soft information, we employ a dummy
Soft identifying the bank for which this feature is important, and its interaction with the dummy
Target (Table 4.2). In this case the interaction term has no significant effects: no differential effects
are found for the delegation schemes of acquired banks. We also find that the higher the
importance of the soft information the lower the local officers turnover (branch officers stay in
charge for a longer period). This also seems to make sense: a fast turnover is a better way to control
officers, but it is likely to engender the disruption of personal knowledge and relationships.

6.2 Recent and old acquisitions
In order to mitigate the difficulty of measuring a dynamic phenomenon like an acquisition in
a cross-section estimation, we try to distinguish the deals performed in the three years before the
reference-year of the survey (2006) from the older ones. The economic literature about M&As
often identifies a transition period, generally lasting three years, in which some changes in the
organizational structure progressively emerge to find a new equilibrium, followed by a completion

18
Further evidence, emerging from our exercise, is that, ceteris paribus, turnover grows after an acquisition for bidder banks as well.
This is probably due to the willingness of the acquiring intermediary to place its managers in a key role within some of the most
crucial branches of the target bank.

period, when the consequences of the M&A should have consolidated. To make this distinction,
as seen in Section 4, we build two sets of dummy target banks to distinguish recent events (new)
from older ones (old), respectively between 2003 and 2005 and between 2000 and 2002.
Disentangling acquisitions over time allows us to verify whether or not there are long-term effects:
we expect different restructuring results depending on the time deals took place.
The results of this exercise are presented in Table 5.1. The post-acquisition growth of the
autonomy of the local officer and of the CEO is significant in the transition period, but its effect is
reinforced after a three-year period. This seems to be consistent with the progressive
implementation of a new delegation scheme by acquiring banks. The effects on turnover are similar
in the two sub-periods, even if the parameter of the older acquisitions is more significant.
Consequently, these findings generally suggest that the integration of different organizations and
cultures takes time.

6.3 Small versus large banks
Another interesting question is whether the centralization-decentralization choice after an
acquisition differs according to the size of the target. A priori, we can suppose that the smaller the
target, the harder and the faster the impact of the acquisition on its organization. To test this
hypothesis, we split the dummy target bank between small and large intermediaries with the
same approach employed in previous sections.
The results are shown in Table 5.2. As expected, the effects of acquisitions on powers
delegated to both the local officer and the CEO are stronger for smaller targets. In fact, the
parameters of small banks are greater than those of the larger intermediaries. The effects on
turnover are stronger for small banks, but in this case the difference is small.

6.4 Geographical proximity between bidder and target
The effect of an acquisition on the internal organization of the target bank could be
influenced by the geographical proximity to its bidder. Proximity can make dealing partners more
similar and reduce integration costs in the transition period. We can suppose that the bidder will be
more worried about moral hazard behaviour if its target operates far away, in a less known area:
this could lead, at the same time, to less delegation of power and a faster turnover of local officers.
To testing this hypothesis, as seen in Section 4, we employ different interactions with the dummy
target bank: the first is related to the overlapping of the branch networks of bidders and targets,
the second investigates if the headquarters of bidders and targets are (or are not) located in the
same area.
Results are presented in Table 5.3. As expected, if the provinces covered by the branch
network of the target are largely new provinces for the bidder (NoOverlap), the post-acquisition
growth of local managers and top managements decision-making powers is weaker. This suggests
that the bidder is willing to strictly monitor the officers of its target, because a lower level of
knowledge of the new local markets can generate higher risks. However, we do not find any
significant difference in turnover. On the one hand, in the case of low overlapping, the need to limit
moral hazard would suggest a faster turnover; on the other hand, this would cause a loss of specific
local information, that could be useful at banking group level.
The second exercise (see also Alessandrini et al., 2008) better qualifies previous findings.
Results (see Table 5.4) suggest that the post-acquisition growth of a local officers power emerges
only in the case of an in-market deal, that is if the headquarters of the dealing partners (bidder
and target) are located in the same area (Inmkt). The growth of the CEOs power holds in both

hypotheses, but if bidder and target are headquartered in different areas, the improvement of the top
managements decision-making power within the acquired bank is smaller and less statistically
significant. Since in our sample the acquiring banks are all located in Northern and Central Italy,
these results suggest that when the target is a Southern intermediary, the bidders organizational
behaviour in terms of power delegated is more careful, due to higher credit risk in the Southern
area.
The increase in turnover is stronger following an out-market deal. This may indicate the
willingness of the bidder to place its own managers in control of some of the targets branches,
perceived as more critical.

7. Robustness checks: matching methodologies
It is important to note that acquired banks might not be a random sample from the
population. To the extent that the acquisition targets differ systematically from other banks, a
problem of simultaneity between ownership status and other performance and organizational
variables could arise and bias the estimates. In order to address this endogeneity problem, we check
previous results by running a propensity score matching. This procedure controls for the selection
bias by restricting the comparison to differences within carefully selected pairs of banks.
19

The aim of the analysis is to estimate the causal effect of the concentration process on a
banks organizational features, in terms of power delegated and turnover. So, we can write the
following difference:
)] ( ) ( [ )] ( ) ( [ ) (
, 0
0
, 1
0
, 0
0
, 1
1
, 1
0 1
X T X T X T X T X T
Y E Y E Y E Y E Y Y E
= = = = =
=
(4)
The second term of the first difference (Y
0
) is, however, an unobserved counterfactual; the
matching technique is a way of constructing this missing counterfactual by making comparisons
conditional on a vector X of observable bank characteristics. Moreover, the underlying assumption
is that, conditional on vector X relevant to the acquisition decision, a target and a stand-alone bank
would exhibit similar organizational features under the same circumstances. So, the second
difference in squared brackets in the equation (4) represents the selection bias, which is assumed to
be zero conditional on X. Since conditioning on the propensity score is equivalent to conditioning
on all variables in the treatment model, we follow this methodology.
The propensity score (p-score) is the predicted probability of treatment which, in our case, is
the probability of a bank being acquired. Then, we employ propensity score matching and we build
homogenous groups to compare the organizational features of banks within the pairs of
observations matched on their p-score value (Becker and Ichino, 2004). Hence, if our matching
process is successful, we can give a causal interpretation of the average difference in power
delegation and turnover indicators between treated and control banks.
Following this approach, as a first step, we use a probit regression to model the binary
outcome of a bank acquired in the period 2000-06 as a function of bank-specific characteristics. In
the selection of the relevant variables for the acquisition decision, we follow Focarelli et al. (1999)

19
Each pair consists of an acquired and a stand-alone bank with similar observable characteristics in a two-year period preceding the
acquisition. In order to maximize the matching probability, for this exercise, we use the whole sample of acquired banks for which
we have organizational data, without dropping banks that have undergone a merger in the period under investigation. These
matching methodologies have had their original application in labour economics, but they have become increasingly popular in
other causal economic analyses. They are often combined with a difference-in-difference approach, in order to capture differences in
the path of performance before and after the treatment event. Unfortunately, we are not able to follow this approach because our
organizational variables are observed only at the end of 2006.

and we control for bank size, portfolio risk, profitability, assets and liabilities composition and cost
efficiency; for target banks all these indicators are taken as a mean on a two-year period preceding
the deal, in order to avoid endogeneity. Results of the probit estimation (not reported) suggest as
expected that acquired banks are more problematic in terms of risk and efficiency and that their
acquisition would be motivated by the restructuring strategies of the holding bank.
20

The predicted probabilities (p-score) are used to assign to each acquired bank the closest
bank not involved in acquisition. To measure how well the propensity score performs in our
exercise, we test the balancing hypothesis, by implementing the procedure suggested by Becker
and Ichino (2004). This procedure calculates the difference between the treated and the control
group in terms of each of the above variables and runs simple t-tests on the differences within
5 bands of the propensity score.
21
Using both the predicted probabilities and the five balancing
blocks, for all treated (acquired) banks in our sample we perform on each organizational dependent
variable (delegation measures and loan officers turnover) an Average Treatment Test (ATT) to
compute the average effect on the treated. We compare this result with the coefficient of the
dummy target estimated in our OLS baseline specification (Table 3). To calculate the ATT value,
we employ the Kernel Matching method; therefore, all treated observations are matched with a
weighted average of all controls, with weights that are inversely proportional to the distance
between the propensity scores of those treated and the controls.
Results are presented in Table 6 and are in line with alternative evidence (not reported),
obtained using the Neighbour Matching method. Findings suggest the robustness of previous OLS
estimations. Consequently, we can conclude that when a bank is included in a banking group, the
new top management of the target benefits from an enlargement of its decision-making power,
because of the alignment to the higher level of the holding bank. At the same time, the loan
officers power increases, even if to a lesser extent. Moreover, the local officers tenure at the same
branch seems to become lower relative to the other banks not involved in acquisitions.

8. Concluding remarks
In this paper we investigate the causal links between the organizational features of SME
lending policies and the banking consolidation process, mainly with respect to bank acquisitions.
This process has been particularly intense since the mid-1990s; consequently, most of the evidence
documented by studies on M&As could depend on the relevant impact of the reorganization of the
banking industry.
The availability of a new data set, built using a special survey on Italian banks, allowed us to
analyse the effect of M&As on organizational design for small business lending. In particular, in
this paper we focus on power delegated across hierarchical levels and on the branch officer
turnover. We take into account other bank-level characteristics, too, which are able to affect
lending techniques and organizational design.
According to our results, the decision-making power at various hierarchical levels within an
acquired bank gains from the deal: after controlling for many characteristics, the maximum amount
that a local officer can autonomously lend increases after an acquisition, as well do the decision-

20
In the probit model we have also inserted among the regressors dummy variables for the event that each bank will take part in a
merger in the subsequent period and other bank-level characteristics, such as: presence in district areas, specialization in small
business lending, etc. All these variables are not significant in the probit estimation for the event that a bank will be acquired in the
period 2000-2005.
21
All the differences for the five groups (the default) are found to be small and statistically insignificant (the so-called balancing
hypothesis). As a result, we are confident that we can group together relatively homogeneous banks.

making rights of the CEO of the target bank. This evidence can be explained by the need to move
towards the higher autonomy of the acquiring bank, in order to better merge different lending
policies after the deal.
Conversely, entering a banking group may worsen agency problems and increase monitoring
costs, because of the higher complexity during the integration phase, especially if the strategy of
the acquiring bank is to delegate more power. In this case, we find that loan officer turnover
increases for target intermediaries, most likely as a mechanism to contain monitoring costs.
In the case of acquisition, decentralization strategy and monitoring policy seems to be
favoured if bidder and target banks are similar, thereby reducing integration costs. Consequently,
the gains in terms of decision-making power are greater if the dealing partners have a more similar
geographical reach, both in terms of branch network and the location of their headquarters.
Our study recognizes the links between the consolidation process and bank organization. Our
findings seem to support a post-acquisition trade-off between decentralization choices and agency
problems, exacerbated by the growing complexity and by the need to mitigate moral hazard
behaviour within expanding banking groups. However, further research could provide many
interesting insights. In particular, it would be useful to analyse in greater depth the timing of a
banks organizational set-up, moving from a simple cross-section to a more informative panel data
analysis.

APPENDIX
Table 1
Descriptive statistics on the estimation sample

Variable Description Mean
Std.
dev.
Median Q25 Q75

Organizational indicators (year 2006)
Loan
officers
delegation
Amount of finance up to which a local loan
officer can autonomously grant to a small firm
(here in thousand of euros; in the estimates in
log)
136.31 314.44 70.00 15.00 150.00
CEOs
delegation
Amount of finance up to which the CEO can
autonomously grant to a small firm (here in
thousand of euros; in the estimates in log)
2,428.50 6,988.91 500.00 150.00 1,600.00
Loan
officers
tenure at
the same
branch
Average period of tenure of a local loan officer
at the head of the same branch (here in number
of months; in the estimates in log)
46.06 20.84 40.00 35.00 60.00
Bank-level characteristics

Average
loan size
Average individual size of loans granted to
firms and households (log of euros)
11.06 0.54 11.04 10.75 11.32
Bank size Total assets (log of euros) at the end of 2005 20.72 1.31 20.40 19.77 21.39
Roa
Ratio between profits before taxes and total
assets (percentage); average 2000-05.
0.85 0.61 0.90 0.71 1.10
Portfolio
risk
Ratio between non-performing loans and total
loans (percentage); average 2000-05
3.70 4.21 2.56 1.69 4.13
District
branches
Share of branches located in municipalities
classified as industrial districts (percentage)
36.21 36.03 21.01 0.00 66.67
The estimation sample is the one employed for the baseline specification (Table 3, Model IV).


Table 2
Descriptive statistics: differences across clusters of banks

Variables and cluster of
banks
Obs. Mean Std. dev. Median Q25 Q75

Acquiring banks (bidders)
Loan officers delegation
11 242.55 190.84 250.00 100.00 400.00
CEOs delegation
11 7,204.55 6,487.85 4,500.00 3,500.00 15,000.00
Loan officers tenure at the same
branch
10 31.20 7.02 32.00 26.00 36.00
Acquired banks (targets)
Loan officers delegation
32 267.19 537.48 150.00 101.50 257.50
CEOs delegation
32 5,451.13 8,264.99 2,750.00 1,300.00 6,200.00
branch
30 34.07 10.61 35.50 25.00 36.00
Other banks
Loan officers delegation 190 108.12 260.19 50.00 10.00 100.00
CEOs delegation 191 1,647.03 6,562.41 310.00 120.00 900.00
branch
177 48.93 21.60 48.00 36.00 60.00
The estimation sample is the one employed for the baseline specification (Table 3, Model IV).


Table 3
Baseline results

Loan officers delegation (log level) CEOs delegation (log level)
I II III IV I II III IV
DBidder -0.6905 -0.4205 -0.5266 -0.5264 -0.2370 -0.0370 -0.1453 -0.1453
[0.6479] [0.6383] [0.6451] [0.6434] [0.3776] [0.4204] [0.4251] [0.4259]
DTarget 0.9173*** 1.1482*** 1.1741*** 1.1833*** 1.1018*** 1.2729*** 1.2993*** 1.2996***
[0.2597] [0.2635] [0.2678] [0.2632] [0.2105] [0.2124] [0.2109] [0.2113]
Average loan size -0.4204* -0.3428 -0.3925* -0.3732* -0.2506 -0.1931 -0.2439 -0.2433
[0.2341] [0.2277] [0.2349] [0.2225] [0.2307] [0.2262] [0.2255] [0.2232]
Bank size 0.7183*** 0.6150*** 0.6399*** 0.6578*** 1.0151*** 0.9386*** 0.9641*** 0.9646***
[0.0890] [0.0949] [0.0965] [0.0971] [0.0668] [0.0702] [0.0710] [0.0725]
Roa 0.2009 0.1949 0.1736 0.1480 -0.0228 -0.0272 -0.0490 -0.0498
[0.2262] [0.2279] [0.2332] [0.2247] [0.0903] [0.0888] [0.0929] [0.0937]
Risk -0.0024 -0.0030 0.0156 0.0238 -0.0236 -0.0240 -0.0050 -0.0047
[0.0262] [0.0232] [0.0285] [0.0284] [0.0211] [0.0177] [0.0188] [0.0193]
DGroup 0.8850*** 0.8390*** 0.8532*** 0.6558*** 0.6088*** 0.6092***
[0.2538] [0.2566] [0.2593] [0.2168] [0.2183] [0.2178]
DSouth -0.5341 -0.4056 -0.5456** -0.5416**
[0.4314] [0.4374] [0.2446] [0.2392]
District branches 0.0062* 0.0002
[0.0032] [0.0027]
Constant -6.8520** -5.7206* -5.6748* -6.5069** -12.0770*** -11.2387*** -11.1919*** -11.2178***
[2.9820] [3.0248] [3.0069] [2.9510] [2.6889] [2.7048] [2.6973] [2.6426]
Observations 235 235 235 235 235 235 235 235
R-squared 0.29 0.31 0.31 0.32 0.59 0.61 0.61 0.61
OLS estimation with robust standard errors [in brackets]; * significant at 10%; ** significant at 5%; *** significant at 1%.


Table 3 (cont.)
Baseline results

LO's tenure at the same branch
I II III IV
DBidder -0.1931* -0.2435** -0.2179* -0.2180*
[0.1051] [0.1213] [0.1231] [0.1226]
DTarget -0.1848*** -0.2325*** -0.2389*** -0.2387***
[0.0667] [0.0697] [0.0703] [0.0704]
Average loan size -0.0043 -0.0132 -0.0009 0.0011
[0.0678] [0.0705] [0.0727] [0.0734]
Bank size -0.0729*** -0.0527** -0.0585** -0.0611**
[0.0224] [0.0245] [0.0249] [0.0253]
Roa 0.2171*** 0.2206*** 0.2273*** 0.2303***
[0.0671] [0.0662] [0.0665] [0.0667]
Risk -0.0077 -0.0072 -0.0119* -0.0129**
[0.0049] [0.0054] [0.0062] [0.0065]
DGroup -0.1686** -0.1581** -0.1592**
[0.0710] [0.0711] [0.0705]
DSouth 0.1324 0.1166
[0.0966] [0.0981]
District branches -0.0008
[0.0008]
Constant 5.1748*** 4.8813*** 4.8614*** 4.9275***
[0.7926] [0.8106] [0.8123] [0.8195]
Observations 223 223 223 223
R-squared 0.22 0.23 0.24 0.24


Table 4
Comparing different lending techniques

1. Use of credit scoring 2. Relevance of soft information
LO deleg. CEO deleg. LO tenure LO deleg. CEO deleg. LO tenure
DBidder -0.6528 -0.1425 -0.2469** -0.4178 -0.1612 -0.2514**
[0.6402] [0.4478] [0.1215] [0.6472] [0.4268] [0.1187]
DTarget 1.9887*** 1.3269*** -0.1251 1.1252*** 1.3300*** -0.2037**
[0.3159] [0.2501] [0.1392] [0.3282] [0.2867] [0.0952]
Average loan size -0.4339* -0.2654 -0.0434 -0.3948* -0.2408 0.0003
[0.2280] [0.2300] [0.0677] [0.2232] [0.2270] [0.0717]
Bank size 0.6708*** 0.9738*** -0.0550** 0.6479*** 0.9654*** -0.0609**
[0.1006] [0.0775] [0.0266] [0.0977] [0.0723] [0.0254]
Roa 0.1069 -0.0727 0.2498*** 0.1478 -0.0476 0.2498***
[0.2225] [0.0986] [0.0641] [0.2159] [0.0979] [0.0642]
Risk 0.0177 0.0029 -0.0153** 0.0220 -0.0045 -0.0125**
[0.0300] [0.0193] [0.0070] [0.0281] [0.0192] [0.0059]
DGroup 0.8137*** 0.6147*** -0.1765** 0.8493*** 0.6067*** -0.1576**
[0.2662] [0.2266] [0.0705] [0.2571] [0.2194] [0.0679]
DSouth -0.5810 -0.7409*** 0.1166 -0.3990 -0.5431** 0.1255
[0.4976] [0.2532] [0.1012] [0.4276] [0.2414] [0.0930]
District branches 0.0061* -0.0002 -0.0010 0.0064* 0.0001 -0.0009
[0.0033] [0.0027] [0.0008] [0.0033] [0.0027] [0.0008]
DScore 0.2468 -0.1285 0.0001
[0.2553] [0.1813] [0.0619]
DTarget*DScore -1.1766*** -0.0178 -0.1298
[0.4125] [0.3544] [0.1557]
DSoft -0.3251 0.0424 0.1289**
[0.2409] [0.1864] [0.0588]
DTarget*DSoft 0.0715 -0.0677 -0.0953
[0.4663] [0.4014] [0.1324]
Constant -6.1618** -11.0742*** 5.2958*** -5.8856* -11.2822*** 4.8451***
[3.0350] [2.6109] [0.7840] [3.0113] [2.6936] [0.8309]
Observations 227 227 215 234 234 221
R-squared 0.33 0.61 0.26 0.33 0.61 0.27


Table 5
Focus on time of deal and on bank size effect - I

1. Old vs. New 2. Small vs. Large
DBidder -0.5679 -0.1938 -0.2242* -0.6309 -0.3078 -0.2080
[0.6465] [0.4232] [0.1236] [0.6539] [0.4367] [0.1274]
DTarget*Old 1.1527*** 1.3235*** -0.2394***
[0.2875] [0.2562] [0.0749]
DTarget*New 0.8255* 0.8114** -0.2393**
[0.4832] [0.3502] [0.1115]
DTarget*Small 1.5555*** 1.8785*** -0.2808***
[0.3627] [0.2611] [0.0701]
DTarget*Large 0.9249*** 0.8978*** -0.2158**
[0.3174] [0.2551] [0.0973]
Average loan size -0.3572 -0.2288 -0.0039 -0.4050* -0.2927 0.0033
[0.2301] [0.2263] [0.0738] [0.2276] [0.2237] [0.0741]
Bank size 0.6687*** 0.9776*** -0.0593** 0.6937*** 1.0205*** -0.0646**
[0.0985] [0.0729] [0.0256] [0.1040] [0.0789] [0.0276]
Roa 0.1271 -0.0788 0.2291*** 0.1692 -0.0168 0.2310***
[0.2108] [0.0884] [0.0660] [0.2311] [0.1064] [0.0661]
Risk 0.0263 -0.0026 -0.0139** 0.0234 -0.0053 -0.0128*
[0.0291] [0.0204] [0.0065] [0.0287] [0.0193] [0.0065]
DGroup 0.8277*** 0.5817*** -0.1625** 0.7977*** 0.5230** -0.1544**
[0.2593] [0.2165] [0.0711] [0.2673] [0.2216] [0.0723]
DSouth -0.4580 -0.6047** 0.1290 -0.3917 -0.5199** 0.1161
[0.4481] [0.2618] [0.0966] [0.4398] [0.2366] [0.0987]
District branches 0.0062* 0.0001 -0.0008 0.0061* 0.0000 -0.0008
[0.0032] [0.0027] [0.0008] [0.0032] [0.0027] [0.0008]
Constant -6.8765** -11.6005*** 4.9508*** -6.8966** -11.8239*** 4.9717***
[3.0457] [2.6549] [0.8138] [2.9809] [2.6344] [0.8195]
Observations 235 235 223 235 235 223
R-squared 0.32 0.61 0.24 0.33 0.62 0.24


Table 5 (cont.)
Focus on target and bidder geographical reach - II

3. Network overlapping 4. In-market Vs. Out-market
DBidder -0.5915 -0.1828 -0.2050* -0.5164 -0.1279 -0.2137*
[0.6501] [0.4163] [0.1234] [0.6457] [0.4276] [0.1225]
DTarget*NoOverlap 0.9530*** 1.1303*** -0.2384***
[0.3233] [0.2397] [0.0863]
DTarget*Overlap 1.3706*** 1.6043*** -0.2459***
[0.3335] [0.2787] [0.0915]
DTarget*Inmkt 1.2465*** 1.4094*** -0.2180***
[0.2866] [0.2296] [0.0760]
DTarget*Outmkt 0.9149 0.8335* -0.3227**
[0.6050] [0.4382] [0.1588]
Average loan size -0.3301 -0.1916 -0.0103 -0.3776* -0.2509 -0.0007
[0.2363] [0.2246] [0.0758] [0.2212] [0.2227] [0.0740]
Bank size 0.6826*** 0.9833*** -0.0662** 0.6564*** 0.9622*** -0.0619**
[0.1013] [0.0709] [0.0264] [0.0973] [0.0723] [0.0253]
Roa 0.1414 -0.0510 0.2372*** 0.1489 -0.0481 0.2297***
[0.2172] [0.0891] [0.0633] [0.2271] [0.0968] [0.0669]
Risk 0.0252 -0.0029 -0.0133** 0.0222 -0.0074 -0.0135**
[0.0283] [0.0190] [0.0066] [0.0289] [0.0185] [0.0067]
DGroup 0.8146*** 0.5897*** -0.1537** 0.8538*** 0.6103*** -0.1583**
[0.2614] [0.2130] [0.0714] [0.2597] [0.2175] [0.0705]
DSouth -0.3459 -0.4808** 0.1198 -0.3133 -0.3812 0.1458
[0.4495] [0.2384] [0.0997] [0.5135] [0.2580] [0.1082]
District branches 0.0065** 0.0005 -0.0008 0.0062* 0.0001 -0.0008
[0.0033] [0.0027] [0.0008] [0.0032] [0.0027] [0.0008]
Constant -7.4882** -12.1876*** 5.1507*** -6.4327** -11.0889*** 4.9639***
[3.2549] [2.6053] [0.8716] [2.9614] [2.6548] [0.8298]
Observations 235 235 223 235 235 223
R-squared 0.32 0.61 0.24 0.32 0.61 0.24


Table 6
Robustness checks: matching methodology

Dependent
Variables
N. obs. treated N. Controls
Average
treatment effect
Bootstrapped
stand. errors
Significance
level

Loan officers
delegation
36 120 0.681 0.357 **
CEOs delegation 36 120 1.184 0.401 ***
Loan officer's tenure
at the same branch
39 190 -0.117 0.073 *
The Average Treatment Test is calculated using the Kernel methodology; the common support option has been applied; * significant at
10%; ** significant at 5%; *** significant at 1%.


REFERENCES
Economy, 105(1): 1-29, February.
Akhavein, J., W.S. Frame and L.J. White (2005), The diffusion of financial innovations: Adoption
of small business credit scoring by large banking organizations, Journal of Business, 78(2):
577-596.
dellattivit creditizia e lutilizzo delle tecniche di scoring nel sistema bancario italiano:
risultati di unindagine campionaria, Questioni di Economia e Finanza, 12, Banca dItalia,
March.
Alessandrini, P., M. Croci and A. Zazzaro (2005), The geography of banking power: The role of
functional distance, BNL Quarterly Review, LVIII, 235: 129-167, December.
Alessandrini, P., G. Calcagnini and A. Zazzaro (2008), Asset restructuring strategies in bank
acquisitions: Does distance between dealing partners matter?, Journal of Banking and
Finance, 32: 699-713.
Avery, R.B. and K.A. Samolyk (2000), Bank consolidation and the provision of banking services:
The case of small commercial loans, Federal Deposit Insurance Corporation, Working
Papers, 00-1.
Becker, S.O. and A. Ichino (2004), Estimation of average treatment effects based on propensity
scores, mimeo.
Benvenuti, M., L. Casolaro, S. Del Prete and P. E. Mistrulli (2009), Loan officer authority and
small business lending. Evidence from a survey, mimeo, Bank of Italy.
Beretta, E. and S. Del Prete (2007), Aggregazioni bancarie e specializzazione nel credito alle PMI:
peculiarit per area geografica, Temi di discussione, 644, Banca dItalia, November.
Berger, A.N., S.D. Bonime, L.G. Goldberg and L.J. White (2000), The dynamics of market entry:
The effects of mergers and acquisitions on the de novo entry and small business lending in
the banking industry, Board of Governors of the Federal Reserve System, Working Paper.
Berger, A.N., W.S. Frame and N.H. Miller (2005a), Credit scoring and the availability, price, and
risk of small business credit, Journal of Money, Credit and Banking, 37( 2): 191-222.
Berger, A.N., L.G. Goldberg and L.J. White (2001), The effects of dynamic changes in bank
competition on the supply of small business credit, European Finance Review, 5: 115-139.
Berger, A.N., A.K. Kashyap and J.M. Scalise (1995), The transformation of the U.S. banking
industry: What a long, strange trip its been, Brookings Papers of Economic Activity, 2: 55-218.
Berger, A.N., N.H. Miller, M.A. Petersen, R.G. Rajan and J.C. Stein (2005b), Does function
Berger, A.N., A. Saunders, J.M. Scalise and G.F. Udell (1998), The effects of bank mergers and
acquisitions on small business lending, Journal of Financial Economics, 50: 187-229.
Berger, A.N. and G.F. Udell (2002), Small business credit availability and relationship lending:
The importance of bank organisational structure, Economic Journal, 112: 32-53.
Berger, A.N. and G.F. Udell (2006), A more complete conceptual framework for SME finance,
Journal of Banking and Finance, 30(11): 2945-2966.

Bonaccorsi di Patti, E. and G. Gobbi (2001), The effects of bank consolidation and market entry
on small business lending, Temi di discussione, 404, Banca dItalia.
Bonaccorsi di Patti, E. and G. Gobbi (2007), Winners or losers? The effects of banking
consolidation on corporate borrowers, Journal of Finance, 62(2): 669-695.
Bongini, P., M.L. Di Battista and L. Nieri (2007), Assetti organizzativi e performance dei gruppi
bancari italiani, in AA.VV., Banche italiane: unindustria al bivio. Mercati, consumatori,
governance, XII Rapporto, Fondazione Rosselli, Rome, Bancaria Editrice.
Cerqueiro, G., H. Degryse and S. Ongena (2007), Distance, bank organizational structure and
credit, mimeo.
Del Prete, S. (2002), I processi aggregativi dimpresa. Unanalisi delle peculiarit delle
concentrazioni bancarie, Milano, Giuffr.
De Vincenzo, A. and S. Iannotti (2002), Il credito alle imprese, le relazioni di clientela e le
operazioni di concentrazione, mimeo, Banca dItalia.
Quaderni di Moneta e Credito, 197: 245-265.
Focarelli, D., F. Panetta and C. Salleo (1999), Why do banks merge?, Temi di discussione, 361,
Banca dItalia, December.
Focarelli, D., F. Panetta and C. Salleo (2002), Why do banks merge?, Journal of Money, Credit
and Banking, 34(4): 1047-1066.
Frame, W.S., A. Srinivasan and L. Woosley (2001), The effect of credit scoring on small-business
lending, Journal of Money, Credit, and Banking, 33(3): 813-825.
Keeton, W.R. (1996), Do bank mergers reduce lending to businesses and farmers? New evidence
from tenth district states, Economic Review, Federal Reserve Bank of Kansas City, 81(3): 63-76.
Keeton, W.R. (2000), Are mergers responsible for the surge in new bank charters?, Federal
Reserve Bank of Kansas City Economic Review, 85: 21-41.
Liberti, J.M. (2003), Initiative, incentives and soft information: how does delegation impact the
role of bank relationship managers?, London Business School of Economics.
Liberti, J.M. (2005), How does organizational form matter? Communication, distance and soft
information, London Business School of Economics.
Liberti, J.M. and A. Mian (2006), Estimating the effect of hierarchies on information use,
London Business School, London.
Panetta, F., F. Schivardi and M. Shum (2004), Do mergers improve information? Evidence from
the loan market, Temi di discussione, 521, Banca dItalia.
Peek, J. and E.S. Rosengren (1996), Small business credit availability: How important is the size
of the lender?, in Saunders A. and I. Walter, Universal Banking: Financial System Design
Reconsidered, Irwin Burr Ridge, Illinois, 628-655.
Peek, J. and E.S. Rosengren (1998), Bank consolidation and small business lending: Its not just
bank size that matters, Journal of Banking and Finance, 22: 799-819.
Sapienza, P. (2002), The effects of banking mergers on loan contracts, Journal of Finance, LVII
(1): 329-367.
Scott, J.A. (2006), Loan officer turnover and credit availability for small firms, Journal of Small
Business Management, 44(4): 544-562.

Scott, J.A. and W.C. Dunkelberg (1999), Bank consolidation and small business lending: Asmall
firms perspective, in Blanton J.L., A. Williams and S.L.W. Rhine (eds.), Business access
to capital and credit, A Federal Reserve System Research Conference, 328-361.
Scott, J.A. and W.C. Dunkelberg (2003), Bank mergers and small firm financing, Journal of
Money, Credit and Banking, 35: 999-1017.
Uchida, H., G.F. Udell and N. Yamori (2006), Loan officers and relationship lending, mimeo.
Udell, G.F. (1989), Loan quality, commercial loan review and loan officer contracting, Journal
of Banking and Finance, 13: 367-382.
Udell, G.F. (1998), The consolidation of the banking industry and small business lending, in
Amihud Y. and G. Miller, Bank Mergers and Acquisitions, Boston, Kluwer Academic
Publisher.
Williamson, O. (1988), Corporate finance and corporate governance, Journal of Finance, 43: 567-591.

DISCUSSION
Elena Beccalli
*

The paper Information technology and banking organization by Mocetti, Pagnini and Sette
is an interesting contribution to the debate on information technology in banking. It inquires into
how ICT impacts on the local loan officers autonomy in small business lending, finding that
decentralization increases when more ICT is employed and when credit scoring techniques are
adopted.
The contribution of the paper is strong in several respects. First, the literature has covered the
effects of technology on banks performance but not on organizational structures using bank-level
data. This study complements the existing literature with an analysis of the impact of information
technology and credit scoring on the autonomy of branch managers in lending to SMEs, and it
exploits a unique dataset including ICT and decentralization data, for a sample of 300 Italian banks
in 2006.
The paper might have explored the link between technology and organizational structure in
more open fashion. Its empirical model assumes a one-direction causality in the relationship, and
thus presumes the validity of the technological imperative: technology determines organization.
Yet the organizational and management literature has concluded that the technological imperative
has a long history, but the empirical evidence has generated contradictory findings. [...] Information
systems both centralized and decentralized authority (Markus and Robey, 1988, p. 585). It seems
that since the eighties a new imperative has been emerging: the interaction between technology and
organization. Two relevant dimensions characterize the interaction. First, in presence of
uncertainty, the interaction is affected by flexibility, that is the ability of a bank to develop an
internal organizational system (such as an organizational structure with high modularity and high
organicness) that is able to adapt rapidly to modified market conditions (Fusconi, 1996). Second,
the interaction takes place not only in the long run, but also in the short and medium term
(Bresnahan et al., 2002). The existence of the interaction between technology and organization is
something that might be fruitfully explored by a test of causality, such as the Granger causality
Wald test based on a VAR (Granger and Newbold, 1986), in addition to the test using instrumental
variables.
The paper considers two types of ICT investment: hardware and software. While these are
surely major components of ICT spending, one might also include ICT services, since hardware
and software account for only about 50 per cent of the total ICT investment by EU banks (Beccalli,
2007). This is especially true for small and medium-sized banks, which mainly acquire ICT
services via outsourcing (just consider the services provided in outsourcing by the central
associations and consortiums). Given that services represent a substantial and growing proportion
of total ICT investment, this outsourcing expenditure should be included in the calculation.
Additionally, alternative measures of the ICT index might have been employed in robustness
checks: not only ICT capital stock per employee but also the ratio of ICT capital stock to total
costs.
The ICT explanatory variable is included with a lag of three years. This attempt to consider a
lagged effect of ICT on organization is highly appropriate. While the ICT literature sets the time
required to observe the impact of ICT on performance at five years (Brynjolfsson and Hitt, 2000),

*
Catholic University of Milan, Department of Economics and Business Administration.
104 Elena Beccalli

we are not aware of any public evidence showing that three years are required to observe the effect
of ICT on the organization. The paper would benefit from a further analysis of the length of the lag.
A more general issue concerns the test of the hypothesis underlying the overall analysis, i.e.
the fact that greater delegation of decision-making power increases lending to SMEs. One wonders
whether it would be possible to devise an empirical test of the effect of decentralization on the
quantity of loans granted by a specific bank. The paper might also make a greater effort to
distinguish the delegation effects according to the technology intensity of the bank (as by
comparing banks in the first and the last quartiles in ICT investment). Are there differences in
decentralization correlated with greater/less investment in ICT? Finally, the paper could well be
more ambitious in investigating the joint effect of ICT and decentralization on productivity. Do
more ICT and greater decentralization lead to improvements in productivity? This question is at the
heart of the debate (Solow, 1987; Brynjolfssojn and Hitt, 1998; Carr, 2003), and it would be very
interesting to pursue it.
The paper Bank acquisitions and decentralization choices by Beretta and Del Prete
investigates the impact of acquisitions on decentralization. The focus is on the changes in the
decision-making powers of CEOs and branch managers of the Italian banks involved in
acquisitions before 2006. The main findings are: i) an increase in the powers delegated to the target
bank after the acquisition, occurring immediately and lasting for three years; and ii) an increase in
the branch managers turnover for banks involved in acquisitions.
The analysis is based on a unique, high-quality, detailed dataset with bank-level information
and it examines an interesting and relevant topic not previously analysed in the extensive literature
on bank mergers and acquisitions (perhaps precisely because of the lack of data at the bank level).
And it uses a methodology (propensity score matching) that overcomes the limitations of the OLS
methodology.
Decentralization, the papers dependent variable, is crucial but particularly complicated to
measure. The proxy that the authors adopt is the delegation of power in lending decisions, but this
is just one of the various dimensions of decentralization identified by the organizational literature
(Pugh et al. 1968; Hall, 1991). First, decentralization can be defined not only in terms of decision-
making autonomy of the periphery, but also in terms of the scope and effectiveness of control
systems. From this perspective, credit scoring can be interpreted as a tool to monitor the branch, or
rather to compensate for the presence of greater delegation. Second, decentralization depends on
the manner in which exceptions are handled, i.e. the flexibility in the use of the scoring system and
the possibility of derogation to the ratings resulting from it. Another issue is that decentralization
represents just one dimension of the organizational structure. Consequently, it could be appropriate
to consider other structural dimensions (e.g., hierarchical levels).
Some other suggestions could be addressed to the sample and the model. As far as the
sample is concerned, the number of targets and bidders under investigation is very limited (11
bidders and 32 targets). There is an asymmetry with respect the control group (190 observations).
Further, the ratio between the number of discriminatory observations (11 and 32) and the number
of explanatory variables (6 or 9) could raise concerns of model over-fitting, which could be
overcome by using a panel dataset. With reference to the model, the paper employs an appropriate
methodology (propensity score matching) as a robustness check; since this approach overcomes
many of the limitations of the OLS methodology, it might be better to present propensity score
matching as the reference methodology. Also, the model assumes the instantaneous impact of the
acquisition on the delegation of powers. As there is no previous evidence of such immediacy,
further tests might well have been proposed, particularly with the introduction of lagged variables.
As regards the variables, the results could be strengthened by considering industry-adjusted
values (that is, bidder minus control sample, and target minus control sample). The paper uses
Session 1: New technologies and banking organization - Comments 105

unadjusted values for all variables, but the values of bidders and targets are affected by
developments in the industry unrelated to acquisitions, and this could be fruitfully taken into
account. The paper would also benefit from a test on the collinearity problems potentially affecting
the set of explanatory variables.
Finally, the papers contribution could be further enhanced in two dimensions. First, the
assumption (not tested) is that the root cause of the variation in credit policies after acquisitions is
the organizational change. But is the organizational change necessarily the actual root cause?
Second, an interesting consequence to be tested is whether the increase in the delegation of powers
to the target and the increased turnover of local managers lead to changes in the amount of credit
granted to SMEs. Specifically, it could be interesting to test the causality of the relationship
between organizational change and lending volume.
106 Elena Beccalli
REFERENCES
Beccalli, E. (2007), Does IT investment improve bank performance? Evidence from Europe,
Journal of Banking & Finance, 31: 2205-2230.
Brynjolfsson, E. and L.M. Hitt (2000), Information technology, organization transformation, and
business performance, Journal of Economic Perspective, 14: 23-48.
Brynjolfsson, E. and L.M. Hitt (1998), Beyond the productivity paradox, Communications of the
ACM 41, 49-55.
Carr, N.G (2003), IT doesnt matter, Harvard Business Review, 81: 41-50.
Fusconi, A. (1996), La variabile flessibilit nel modello strutturale, organizzativo e gestionale
della banca, Atti XVII Convegno Cefor Flessibilit, struttura e organizzazione nelle
Banche, Verona.
Granger, C.W.J. and P. Newbold (1986), Forecasting economic time series, Academic Press Inc.,
California.
Hall, R.H. (1991), Organizations: Structures, processes, and outcomes, Prentice-Hall.
Markus, M.L. and D. Robey (1988), Information technology and organization change: Causal
structure in theory and research, Management Science, 34: 583-598.
Pugh, D.S., D.J. Hickson and C.R. Hinings (1968), Dimensions of organization structure,
Administrative Science Quarterly, 13: 65-91.
Solow, R. (1987), Wed better watch out, New York Times Book Review, July 12.

DISCUSSION
Barbara Chizzolini
*

1. The papers by Enrico Beretta, Silvia Del Prete and by Sauro Mocetti, Marcello Pagnini,
Enrico Sette deal with the causes and effects of changes in banking organization after
liberalization, the increased concentration in banking markets and the onset of widespread use of
information and communication technologies. Together they offer a clear picture of the
organization and strategy of the Italian banks operating in local banking markets before the global
financial crisis of 2007-08, which I suspect is still substantially unchanged.
Both papers are mainly concerned with the delegation of decision power on lending approval
from bank headquarters to the periphery. Both exploit an interesting and unique dataset concerning
the internal organization of more than 300 Italian banks drawn from a special survey. Their results
may be summarised as follows:
- Information and communication technologies tend to increase the delegation of decision-
making power to local branch managers.
- The power to grant credit to SMEs increases for both CEOs and local loan officers of the
acquired banks immediately after the merger; that is, decision-making delegation increases.
- The mobility (turnover) of branch managers also increases.

2. Theory points to a tradeoff in decentralization, between the benefits from greater
knowledge of local markets and the costs of monitoring and of potential moral hazard in the
behaviour of local branch officers. The optimal amount of delegation, as measured by a threshold
value for loan size relative to maximum loan size below which the power of decision is transferred
to the branch manager, is derived in Mocetti, Pagnini and Settes Information technology and
bank organization from a theoretical model
1
as a function of the expected benefits for both the
branch manager and the central CEO and of information transmission, decision-making and
monitoring costs.
The theory is tested by regressing PD, a measure of the maximal loan size a branch manager
may approve relative to that which the CEO can grant, on IT investment and adoption, on measures
of credit scoring procedure adoption and on controls. The study finds that ICT and credit scoring
affect delegation/decentralization strongly and positively.
I have three main comments:
i. The benefits to the branch manager. Theory indicates that the amount of delegation
depends on the profitability/quality of loans approved and on managers private benefits.
In particular, the latter affect the effort that branch managers make in securing high-quality
loans. There is no measure of these benefits in the empirical model. Why not? Would it be possible

*
Bocconi University, Milano, Department of Economics E. Bocconi.
1
This theoretical model, however, could be rewritten in a much simpler way.
108 Barbara Chizzolini

to find an observable variable to control for them, maybe using the answers to the survey questions
about incentives to local bank managers?
The ICT and scoring coefficients might be biased upwards because of the omission of this
potentially relevant variable.
ii. Risk and information technology. The authors control for the possibility that risk may bias
their estimates by including in the definition of PD only loans that are judged normal in terms of
risk. But this leaves the reader with a desire to know something more about whether and how much
ICT and credit scoring affect risk-taking at local level. This is likely to be the subject of future
research.
iii. Endogeneity issues. Most of these issues have been taken into account in several
robustness checks, but to my mind the structure of the demand side of the market has not been
considered sufficiently: the location (hence the characteristics of customers) of the branching
network (not only of the banks central headquarters) might be determined jointly with the centrally
decided amount of delegation and investment in technology.
2
Which comes first in the causal series
of events is not entirely clear to me. Is it possible to at least introduce a control for the main
geographical area of operations of banks in the sample?
Again, the ICT and scoring coefficients might be biased.

3. To test whether and how bank acquisitions (merged banks are excluded from the sample)
affect delegation/decentralization choices, Enrico Beretta and Silvia Del Prete in Bank
acquisitions and decentralization choices estimate the following three equations:
LO Delegation = f(Dummy acquired, dummy acquiring, X)
CEO Delegation = f(Dummy acquired, dummy acquiring, X)
LO Tenure = f(Dummy acquired, dummy acquiring, X)
Where LO (CEO) Delegation is the maximum amount that a local Loan Officer (CEO) can
autonomously grant to SMEs and LO Tenure is the number of months of tenure of a local Loan
officer.
Controlling for a number of factors, such as bank size, group membership, profitability and
risk, geographical proximity between bidder and target, the authors find that Acquisitions
positively affect the powers delegated both to the loan officer and to the CEO for loan approval; the
effect of the latter is higher than on the former, and that the turnover of local loan officers is
greater in acquired banks.
I have one main question about this paper, very much related to the endogeneity issue I
raised for the previous paper:
Is the maximum amount of the LOs loan approval authority on average greater because the
bank has been party to an acquisition, or has the bank been acquired because its loan portfolio in
local markets is characterized by relatively large loans?
The treatment analysis and all the robustness checks suggest that the first causal chain
applies. Still, I would be more comfortable if the dependent variables in the delegation equations
were measured in relative terms, either with respect to the average loan size of the bank or, if
possible, of the market where the bank operates. As it is, taken together with the finding of very
high branch manager mobility (LO turnover) in acquired banks, they may actually measure the

2
See Cerasi et al. (2000), Bresnahan (2002).
Session 1: New technologies and banking organization - Comments 109

banks specialization in the loan market, with acquired banks specializing in lending to medium-
sized customers while smaller and mutual banks specialize in lending to very small firms.
3

4. I found both papers interesting and convincing in their results and, taken together, they
indicate that liberalization and concentration in banking markets have given the larger banks that
were the first to invest in ICT and in more flexible, decentralized organization a competitive
advantage in lending to medium-sized firms in local markets through fairly standardized and
efficient procedures.

3
See Berger (1998), for the bimodal distribution of banks by size in the loan markets after deregulation and the waves of M&A deals
in banking.
110 Barbara Chizzolini

REFERENCES
Berger, A.N., A. Saunders, J.M. Scalise and G.F. Udell, (1998), The effects of bank mergers and
acquisitions on small business lending, Journal of Financial Economics, 50: 187-229.
Bresnahan, T.F., E. Brynjolesson and L.M. Hitt, (2002), Information technology, workplace
organization, and the demand for skilled labor: Firm-level evidence, The Quarterly Journal
of Economics, February.
Cerasi, V., B. Chizzolini and M. Ivaldi, (2000), Branching and competitiveness across regions in
the italian banking industry, in Polo M. (ed.), Industria Bancaria e Concorrenza, Il Mulino.

Session 2
BANKING ORGANIZATION AND SMALL BUSINESS LENDING

THE RETAIL ACTIVITY OF FOREIGN BANKS:
EFFECTS ON CREDIT SUPPLY TO HOUSEHOLDS AND FIRMS
Luigi Infante
*
and Paola Rossi
**
1. Introduction
In the last decade, the increased presence of foreign intermediaries in many countries has
heated the debate about the effect of the new entrants on the banking sectors of the host countries.
While the empirical evidence has confirmed that multinational banks may improve competition in
credit markets, the theoretical literature has emphasized how the existence of information
asymmetries give a significant advantage to national banks making it difficult for new entrants to
compete. Following Sengupta (2007), foreign competitors are expected to focus on those segments
of the market that are less opaque where products and conditions may be more easily standardized.
According to DellAriccia and Marquez (2004), new entrants will act more aggressively towards
clients for whom information problems are less important, while incumbent banks will then
maintain a large market share in segments with substantial entry barriers, where they can extract
higher profits from relationship lending (flight to captivity).
Although the impact of foreign players should be different depending upon the importance
of information asymmetries, most of the empirical literature considers country aggregates, mainly
because of data constraints, without differentiating according to the class of customer or the type of
loan. This paper tries to overcome this shortcoming. Analysing the effects of the increasing activity
of foreign banks in Italy, we are especially concerned with distinguishing between lending towards
households and businesses and taking into account different loan contracts.
In this respect, Italy is an interesting testing ground to analyse the effect of multinational
banking. Since the 1990s, the Italian banking sector has undergone profound changes, following
significant regulatory developments: barriers to branch expansion as well as restrictions to banking
activities were removed; obstacles to foreign bank entry were overcome, thanks to the principles of
mutual recognition and home country control introduced by the European Second Banking
Directive. In response to the ongoing process of liberalization, the banking sector has undertaken a
demanding process of restructuring and reorganization, with a substantial recovery in efficiency.
While there is widespread evidence of increasing competition within the Italian banking sector,
1

the role played in this process by the growing presence of foreign intermediaries has been partially
overlooked.
The share of foreign banks in Italy has now risen to about 19 per cent of total assets (Bank of
Italy, 2007); it was just above 5 per cent in the mid-1990s. Furthermore, market shares are
extremely diversified across Italian provinces and they are generally higher when calculated on
credit granted to households than to businesses.
The debt burden of Italian households while increasing rapidly in recent years remains
limited in an international comparison; most of the debt consists of mortgages for housing

*
**
The authors wish to thank A. Accetturo, G. Albareto, P. Alessandrini, M. Bofondi, L. Cannari, R. De Bonis, G. Gobbi, P. E.
Mistrulli, A. Mori, S. Ongena, A. Presbitero, A. Zazzaro, the seminar participants at the Bank of Italy for their helpful comments
and suggestions following a previous version of the paper. All remaining mistakes are the responsibility of the authors alone.
1
Angelini and Cetorelli (2003), Gobbi and Lotti, (2004), Bofondi, Bonaccorsi di Patti and Gobbi, 2006. On the effect of M&As, see
Focarelli and Panetta (2003).
114 Luigi Infante and Paola Rossi

purchases. In the last decade, foreign banks have sharply increased their lending to households,
while foreign intermediaries have specialized in mortgage loans or consumer credit (Casolaro,
Gambacorta and Guiso, 2005).
Conversely, the development of business lending by foreign banks has been limited, partly
because of the peculiar structure of the Italian productive sector. In comparison with other
economies at a similar level of industrialization, Italy is characterized by the prevalence of small
and medium-sized enterprises, for which bank debt is nearly the only source of external financing
(if we exclude commercial debt): there are very few listed companies or companies with access to
international financial markets.
Besides increasing market share, foreign banks in Italy display a fairly differentiated
degree of business expansion across Italian provinces; we exploit this geographical variability to
measure how the presence of foreign banks affects the competitive pressure within local credit
markets. We test it with reference to: i) a customers mobility index;
2
ii) the level of interest rates
and the spreads between lending and deposit rates applied on average to the different classes of
customers; and iii) the collateral required. Our dataset covers the period 1997-2006 and draws on
information at the provincial level. In the estimates, we control for local credit market
characteristics (concentration and size of the market, number of new branches, credit risk, and
financial development), to identify the distinctive role currently played by foreign intermediaries.
The choice of the province where a foreign bank decides to operate may give rise to some kind of
endogeneity; we tackle this problem by using a two-step estimator, according to the methodology
proposed by Robinson (1989) and we end up with stable results.
We show that foreign intermediaries exert significant competitive pressure in credit
markets. Their presence increases the mobility of banks market share, both for loans to households
and to businesses. For the latter, however, the displacement is decidedly more limited and it
appears mainly in short-term lending.
Then we test whether this greater competitive pressure from foreign intermediaries has an
impact on average lending rates either directly, through more aggressive credit policies, or
indirectly, inducing other banks to react by reducing their rates to maintain their positions. In our
regressions, lending rates to households and to businesses are measured by looking at the
difference from the average three-month interbank rate, which approximates the cost of banks
internal funds.
3

As far as mortgage loans are concerned, increased competition by foreign banks comes
with a reduction in interest rates: a rise in the foreign banks share from the 25th to the 75th
percentile of the distribution implies a fall in interest rates of 15 basis points (18 points between
2002 and 2006). In recent years, we also detect a reduction in collateral posted by households, an
outcome that is consistent with the gradual introduction of high loan-to-value mortgages by foreign
banks.
On the other hand, there is no evidence of a significant impact on lending rates applied to
the business sector: neither short-term rates, nor longer maturity rates seem affected. In our view,
this result is rooted in the information asymmetries in (small) business lending. Furthermore,
products supplied and conditions applied to firms cannot be easily standardized. Still, we note a

2
We adjust market shares to account for mergers and acquisitions among banks in the sample period.
3
We rerun the estimates considering the spread between loan and deposit rates in the same area. This variable is commonly used to
measure the mark-up, and it is a proxy for bank market power (Levine, 2004; Martinez Peria and Mody, 2004). Deposit rates,
however, are available only for sight deposits, a source of funds not usually related to mortgages. Therefore, we use these estimates
as robustness checks for our previous results.
The retail activity of foreign banks: Effects on credit supply to households and firms 115

reduction in real collateral posted on long-term loans. The effect, initially weak, strengthens over
time.
Our findings are consistent with recent theoretical literature. The increased mobility of
customers without any significant effect on lending rates may partially support the hypothesis put
forward in DellAriccia and Marquez (2004). They model the propensity of local banks to
specialize in opaque market segments (flight to captivity). On the contrary, a greater mobility of
households, accompanied by a reduction in mortgage rates, confirms that the competitive pressure
is more effective when products are standardized and information asymmetries less important
(Sengupta, 2007; Leaner and Schnitzer, 2006). The evidence tends to reject the hypothesis that
collateral is used to select low-risk customers (Sengupta, 2007). However, this result may be
peculiar to our data, because of the well-known inefficiencies of the Italian judicial system.
The work is organized as follows. Section 2 briefly reviews the theoretical and empirical
literature about the determinants and effects of foreign bank entry. Section 3 accounts for the
development of foreign intermediaries in Italy in the last decade. Empirical analysis follows in
Section 4 where we consider the impact of the presence of foreign banks on the mobility of market
share, on lending rates and collateral required. Section 5 performs a series of robustness checks,
conditioning on the likelihood of operating in a given province. The last section concludes.

2. Related literature
The decision of a bank to enter a foreign country has been driven historically by the
internationalization of borrowers: the growing interdependence of the world economy has
transformed many banks into multinational corporations. This trend reflects the need to provide
credit and financial services to customers going abroad (following the client).
4
Over time, however,
the need to expand into new markets (local market opportunities) as well as to diversify assets
internationally has gained importance.
5

According to industrial organization theory, the entry (or the threat of entry) of a new
company into the market will lead to greater competition and lower market power. However, the
banking industry shows significant peculiarities because of the pervasive presence of information
asymmetries between lenders and borrowers. This market failure affects the decision to enter and
modifies its effects: the advantage of incumbents resulting from information gathering through
repeated interactions with entrepreneurs limits the ability of new entrants to provide better
conditions for customers, because of the risk of attracting bad borrowers (the winner's curse).
Local banks have more information about their customers, which is crucial to small business
lending (the home field advantage). Barriers to entry may be even greater if a bank arrives from
abroad: foreign banks generally belong to large holding groups, whose headquarters may be
geographically remote and also culturally and institutionally distant. The empirical evidence
confirms that foreign banks experience greater difficulties in establishing relationships with small
and more opaque enterprises (Berger, Klapper and Udell, 2001). According to Mian (2006),

4
Among others, see Grubel (1977), Aliber (1984), Williams (1997). Empirical studies on the determinants of entry and activity of
foreign banks have largely confirmed the importance of economic integration among counties (Goldberg and Saunders, 1981;
Goldberg and Johnson, 1990; Grosse and Goldberg, 1991; Brealey and Kaplanis, 1996; Yamori, 1998; Sagari, 1992); an exception
is Williams (1996).
5
Local market opportunities are emphasized by the eclectic theory (Dunning, 1977). Among localization factors, the literature has
stressed the role played by market features of host countries, such as inefficiencies, regulatory restrains or limits to banking
activities that offer profit opportunities to new entrants. By contrast, the agglomeration economies offered by financial centres may
be an important factor of attraction (Davis, 1990). For empirical evidence, see Claessens, Demirg-Kunt and Huizinga (2001),
Focarelli and Pozzolo (2001), Magri, Mori and Rossi (2005).


distance constraints are mainly driven by information-gathering and agency costs, rather than
greater enforcement problems.
On the other hand, intermediaries that operate internationally may be more efficient, because
they may have better screening techniques, skilled management, lower capital costs, and they can
exploit economies of scale or scope (Berger, DeYoung, Genay and Udell, 2000; Claeys and Hainz,
2006; Lehner and Schnitzer, 2006). These factors could partially offset the information
disadvantage (the global advantage hypothesis). Indeed, according to Claessens, Demirg-Kunt
and Huizinga (2001), foreign competitors tend to be more efficient and profitable (Mahajan,
Rangan and Zardkoohi, 1996).
In Sengupta (2007), the ability of the foreign entrant to attract good borrowers is limited by
the need to offer lower lending rates (pooling equilibrium): either the new entrant is more efficient
in some respect, or the market will be dominated by the incumbent bank, which will have both an
information and a cost advantage. The more opaque the borrowers, the higher the efficiency
required to overcome the initial information disadvantage. Alternatively, the incoming bank can
ask for collateral as a discriminating device against high-risk entrepreneurs (separating
equilibrium). Collateral pledging is costly and improves efficiency only in those countries with a
well-functioning judicial system and adequate legal protection for creditors. In Claessens and Van
Horen (2008), the institutional framework plays a crucial role in determining the likelihood of
entry of foreign intermediaries: an environment that is similar both in the home and the destination
county may give rise to a competitive advantage.
According to Dell'Ariccia and Marquez (2004), having to compete with efficient new
entrants, the incumbent bank will try to exploit its superior information by specializing in those
market segments where adverse selection problems are more severe (flight to captivity). Incumbent
banks will keep large market shares in segments with substantial entry barriers, where they can
extract higher profits from relationship lending. To clients for whom information problems are less
important, new entrants may act more aggressively and interest rate reductions should be more
pronounced. Yet, the effect on small, opaque firms may still be positive. According to Giannetti
and Ongena (2009), an increase in foreign bank presence is correlated with a higher probability
that previously unbanked firms will establish bank relationships mainly with domestic banks; a
firms access to bank lending is possibly favoured by increased foreign competition.
Again, in Lehner and Schnitzer (2006) this increased competitive pressure is expected to be
sharper in markets with low product differentiation. According to Claeys and Hainz (2006), in the
case of entry by acquisition, the local bank's information advantage is coupled with the higher
efficiency of the foreign bank. Given the pre-eminent position of the new intermediary, the overall
effect on rates will be less than in the case of entry via the on-site opening of a branch.
By and large, new entrants from abroad are expected to be welfare improving, but there are
possible drawbacks. In the case of a negative correlation between the degree of imperfect
information and the average quality of entrepreneurs, the specialization in captive segments (more
opaque and then more risky) by local banks could lead to a deterioration in the average quality of
assets (DellAriccia and Marquez, 2004). In developing countries, the competition from foreign
players may bring about the so-called cream skimming effect: smaller and more opaque firms are
separated from larger companies and pooled with riskier firms, to which higher interest rates are
applied (Detragiache, Tressel and Gupta, 2006). Concern has also been expressed by Berger et al
(2005): if large foreign banks crowd out smaller domestic ones, the supply of loans to information-
opaque small businesses, lending could be negatively affected.
There is widespread empirical evidence that increased competition from foreign banks has
positive effects. According to Claessens, Demirg-Kunt and Huizinga (2001), on the basis of data
on almost 8,000 banks from 80 countries, the presence of foreign banks improves efficiency and

reduces profit margins. Uiboupin (2004) reports similar results for the countries of Central and
Eastern Europe. Moreover, restricting the entry of foreign intermediaries would raise interest
margins (Levine, 2004) and the market power of incumbents (Claessens and Laeven, 2003). The
effects on competition are also detected in the Italian case, albeit indirectly, where the decision by
foreign operators to enter the market is determined by the profit opportunities offered (Magri, Mori
and Rossi, 2005). Martinez Peria and Mody (2004) find that interest rate spreads are lower for
foreign banks operating in Latin America, while their presence tends to reduce the operating costs
of domestic banks. De Haas and Van Lelyveld (2003, 2008) finds that foreign banks have a
stabilizing effect on credit supply in Central and Eastern Europe. The access to the internal capital
market can alleviate the negative effects of a liquidity shock (because the holding company can
provide support). On the other hand, multinational banks may amplify pro-cyclicality, switching
their activity towards booming economies at the expenses of those counties that are experiencing a
downturn. Furthermore, according to Detragiache Tressel and Gupta (2006), the more foreign
intermediaries there are in developing countries the lower the credit supply - a result that is not
replicated in advanced economies.
Sector analyses are limited. According to Clarke, Cribs Peri and Martinez (2001), financial
constraints are mitigated by the presence of foreign intermediaries. Giannetti and Ongena (2005)
find that the entry of foreign banks in Eastern European countries has a positive impact on credit
availability, and this is especially true for large companies. Mian (2006), using data on more than
80,000 loans to firms located in Pakistan, finds that foreign banks tend to avoid lending to
information-opaque firms, which would require more relationship banking.
Overall, both theoretical analysis and empirical evidence seem to indicate that the impact of
foreign competitors will differ depending on the level of information asymmetry, the degree of
standardization of products and the importance of customer relationships. We expect a partial
reallocation of customers among incoming banks and local intermediaries: the former will focus on
more transactional and competitive markets, where the impact on interest rates should be stronger;
the latter will specialize in segments that are more opaque. Collateral could either increase, if
employed as a selection mechanism, or decrease, if it is used as a competitive device.

3. Foreign banks in Italy: some stylized facts
At the end of 2007, there were 98 foreign banks in Italy, 74 branches and 24 subsidiaries.
The two categories are based on the location of the controlling shareholders of the bank.
6
Most of
them (72 out of 98) were headquartered in Milan, the main financial centre of the country.
7
The
total number of incoming banks has been steadily increasing over the last two decades (Chart 1).
Since the second part of the 1990s, entry by means of subsidiaries and acquisitions has been
gaining importance with respect to branches, consistently with the increased activity in retail
markets.
8

6
We only analyse foreign branches and subsidiaries, thus excluding minority shareholdings. By subsidiary we mean a bank that is
legally autonomous but controlled (through the majority of votes) by a holding company located abroad, while a branch is part of a
foreign bank and has no legal autonomy. On the presence of foreign banks with the form of minority shareholdings, see Focarelli
and Pozzolo (2001). In the estimates, we did not consider acquisitions by foreign banks made in 2006 (the last year of observation)
to avoid discontinuity in the data.
7
A first wave of entries took place between 1977 and 1983, together with the integration of the world economy and the improvement
of Italys international trade position. In this period, the regulatory restrictions to Italian banks offered profit opportunities for
foreign operators to be exploited in loco. A new wave of arrivals occurred with the launch of the European single market (Magri,
Mori, Rossi, 2005).
8
Sunk costs affect the decision whether and how to enter a foreign market (Bugamelli and Infante, 2003). The choice is associated
with the activity the bank proposes to carry out (Claeys and Hainz, 2006, Van Tassel and Vishwasrao, 2005). Acquisitions and
(continues)

Figure 1
FOREIGN BANKS IN ITALY
-10
10
30
50
70
90
1
9
8
0
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
2
0
0
5
2
0
0
6
net inflows from abroad
N. of foreign banks
N. of foreign branches

Source: Banks Register.

Figure 2
DISTRIBUTION OF FOREIGN BANKS MARKET SHARE IN THE ITALIAN PROVINCES
0
2
4
6
8
10
12
14
16
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Inter-quartile range
Median
5th and 95th percentile
Italian average
Households
0
2
4
6
8
10
12
14
16
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Inter-quartile range
Median
5th and 95th percentile
Italian average
Firms

Source: Bank Supervisory Report. Data refer to branches and subsidiaries of foreign banks; takeovers of Italian banks taking place in
2006 are not included.

subsidiaries give a network of branches to the new entrant. This choice is consistent with the aim of operating in the retail sector,
where information asymmetries tend to favour banks with a widespread presence in the territory.

Between 1995 and 2006, the market share of foreign intermediaries in Italy rose from 5.2 to
13.7 per cent of total assets (Table A1); in early 2007 it was 19 per cent. Initially, they used to
operate mainly in wholesale markets and with large multinational companies. In 1995, lending
activities were limited, while off-balance sheet operations accounted for more than 42 per cent of
the system, especially for trading in securities and derivatives. Since then, activity has gradually
changed; market shares in lending activity have grown strongly and deposit-taking has increased
too, albeit at a slower pace. Simultaneously, off-balance sheet activities have been greatly reduced.
However, this evolution has not been uniform nationwide. Figure 2 charts the main
statistical indicators of the distribution of foreign banks market share across the Italian provinces.
It shows the high variability of their presence. For businesses, the national average is always above
the third quartile, pointing to very concentrated activity in a limited number of zones. The findings
confirm there has been a higher increase in positions for households than for businesses. In the
following sections, this geographical variability and its dynamics over time is used to measure the
effect of foreign intermediaries on competitive conditions within credit markets.

4. The effect on market share, interest rates and collateral
4.1 Customer mobility
The first test we perform is to verify whether and to what extent the entry of foreign
players improves customer mobility.
Several studies, focusing on the degree of rivalry within a market, use market share changes
as a measure of competitive pressure stemming from either incumbents or new competitors (Caves
and Porter, 1978; Davies and Geroski, 1997). Relative stability in the ranking of firms in the
market could suggest collusive behaviour (Heggestad and Rhoades, 1976). Likewise, from a
theoretical point of view, few movements in market share do not necessarily mean there is a lack of
rivalry, i.e. in a perfect competition framework, market shares remain stable. However, this is very
unlikely (Shepherd, 1970): even in markets that are not highly concentrated, entrepreneurs will use
various strategies to differentiate their products, which leads to movements in market share.
In the banking industry, tight entry regulation reduces market share instability. In Italy,
during the last two decades, constraints to geographical expansion and to various activities have
been removed. In the same period, several mergers and acquisitions have taken place.
Notwithstanding the increased concentration of the market, big banks have suffered from market
share losses, while small and foreign banks have improved their positions (Bonaccorsi di Patti,
Eramo, Gobbi, 2005). Most likely, reorganizations following M&As have loosened customer
relationships (Sapienza, 2002).
In order to isolate the changes not induced by competition, in this section all the indicators
have been adjusted to account for mergers and acquisitions among Italian banks in the sample
period, by re-building pro-forma consolidated data for banks resulting from the aggregations.
We therefore computed the following (absolute) mobility index (MI) by province and market
segment:
[1] MI
k, p, t+1
=
j=1, n
| MS
j, k, p, t+1
MS
j, k, p, t
|
where MS is the market share of the j-th bank in the p-th province at time t (as said adjusted
for mergers and acquisitions); n is the overall number of banks operating in the p-th province; k are
the market segments considered (loans, short-term loans and mortgages to households or

businesses).
9
This index has the advantage of exploiting the variability of the distribution as a
whole.
10

The MI index has been regressed as in the following specification:
[2] MI
k, p, t+1
= o + (Foreign Bank market share
k
)
p, t-1
+ (Controls
k
)
p, t-1
+ d
t
+ u
p
+ c
p, t

where d
t
is a time dummy, and u
p
is unobserved provincial heterogeneity. The estimates take into
account fixed effects at provincial level. The time span runs from 1997 up to 2006; data are yearly
averages for 103 provinces. All regressors are lagged by one period, to reduce endogeneity
problems. Since the dependent variable is equal to the change in market shares between time t and
t+1, there is a two-year lag with respect to right-hand side variables. Table A2 reports some
descriptive statistics of the variables.
The regressor of interest is the foreign banks market share. This variable should give a more
precise measure of the impact of foreign intermediaries than other indicators such as the number of
foreign banks or branches in the market (Claessens, Demirg-Kunt and Huizinga, 2001). In fact,
as we will discuss in Section 5, foreign banks are involved in retail activities remotely as well
(without on-site presence) and other measures could understate the true effect of these
intermediaries.
As a control variable, we introduced the Herfindahl concentration index, a common measure
for market structure.
11
We further considered the number of new branches (per 10 thousand
inhabitants), in order to control both the entry of new banks and the raising of the level of rivalry
among incumbents as a result of the opening of new branches.
A measure of indebtedness of households and businesses is also included, in the form of
mortgage per capita or total loans over value-added, respectively. The share of small business
lending over total loans in the province is considered to capture the dimensional structure of firms
in local credit markets and, thereby, the importance of relationship lending.
Credit risk is proxied by the ratio between non-performing loans over total loans to
households and firms; markets with lower credit risk should attract more intermediaries and could
be associated with a higher degree of rivalry. As a further control, we used the size of the local
credit market, which also proxies the degree of financial development; this is computed as the sum
of loans and deposits over the population (Kato and Honjo, 2006; Woodrow Eckard, 1987).
Table A3 shows the results in reference to households. Foreign bank presence is positively
related to the instability index for mortgages (column [5]), while it is not statistically significant for
short-term operations or total loans to households (columns [1] and [3], respectively).
As far as mortgages are concerned, moving from the first to the third quartile in the
distribution of foreign banks market shares enhances the mobility by 1 percentage point. Since the
average mobility index is around 13 per cent, this would account for a 7 per cent increase in the
mobility of customers.

9
As far as households are concerned, we do not consider consumer credit, due to the lack of a geographical breakdown before 2002.
10
There is a heated debate on the best indicator to measure the instability of market share, i.e. absolute versus relative or computed on
the top 3 or 5 largest firms, and so on. See Kato and Honjo (2006) for a recent contribution. For Italy, market share instability in the
banking sector is analysed in De Bonis and Ferrando (2000), to verify whether the joint presence of the same banks in different
provinces leads to collusive behaviour.
11
Other structural indicators, such as the share of the 3 largest banks in the market or the number of banks on the population, produced
nearly identical results.

We verified the possible different impacts of foreign intermediaries over time. To this end,
the foreign bank variable is split into two sub-periods, the first between 1997 and 2001; the second
between 2002 and 2006 (Table A3, columns [2], [4] and [6]).
In column [6], the coefficients are statistically similar (with a P-value equal to 0.83 and the
F-test equal to 0.05). However, foreign banks recently expanded their activity unevenly across the
Italian provinces. As a consequence, the impact between 2002 and 2006 is higher: moving from the
25th to 75th percentile of the distribution of foreign banks market share, mobility rises by 1.1
percentage points, almost two and a half times higher than before, corresponding to 9 per cent of
the average mobility of the second sub-period.
Among the control variables, greater market concentration (with a higher Herfindahl index)
implies higher customer mobility. Other concentration indexes pointed to the same results. This
finding is consistent with other evidence on this issue. According to Sakakibara and Porter (2001),
when non-price competition is crucial, rivalry in a concentrated industry becomes harder. For Italy,
De Bonis and Ferrando (2000) read the positive effect of the Herfindahl index as the consequence
of the entry of larger banks into areas where small intermediaries, with high market power, used to
prevail.
Developed credit markets offer greater profit opportunities, implying greater customer
mobility in all the regressions. The degree of indebtedness has a negative sign; most likely this
variable also captures part of the potential risk. The other variables - credit risk or opening new
branches - are not significant.
Moving on to look at firms (Table A4), foreign banks seem to exert some competitive
pressure on both total and short-term loans (columns [1] and [3], respectively), while there is not
any influence longer maturity (column [5]).
For short-term loans, an increase in the market share of foreign banks from the 25th to the
75th percentile amplifies the mobility of firms by 0.4 percentage points, corresponding to 2.8 per
cent of total mobility. The effect although statistically significant appears small from the
economic point of view. This result is consistent with the theory emphasizing the difficulties for
new entrants to evaluate firms creditworthiness without well-established customer relationships.
Our estimates suggest that market shares are basically reallocated in short-term lending, while no
effect is detected for loans with longer maturities, usually related to collateralized loans. This
further confirms other evidence (see Gormley, 2010) in which a large reduction in business lending
to firms with fewer tangible assets is observed following the entry of foreign banks.
The variable measuring small business lending in the provinces has a negative and
significant impact, signalling lower mobility in those markets where the presence of small firms is
higher and, consequently more relationship lending is needed. Other variables results are in line
with what has been already discussed for households.
The intermediation of foreign banks is narrow in some provinces; this could affect the
significance of some coefficients of the previous estimates. Therefore, we verified the robustness
of our results through weighted regressions, where the weights consist of the foreign banks market
shares in all provinces. Our main findings are confirmed (Table A5).

4.2 Interest rates
As a further step, our analysis focuses on establishing to what extent, and for what products,
foreign banks influence the average interest rate charged to customers.

Our dependent variable is defined as the difference between lending rates and the three-
month interbank rates, the latter approximating the cost of internal funds for the bank. Ceteris
paribus, a decrease in this differential, associated with an increase in the presence of foreign banks,
would point to an improvement in market competition. As a robustness check, we consider the
differential between the lending and deposit rates applied in each province. This variable has the
advantage of measuring the spread in each province; on the other hand, interest rates on bank
liabilities are reported only for sight deposits, a source of funding not usually correlated to medium
and long-term lending.
12

For households, we take the interest rates applied to mortgages for housing purchases. As far
as business lending is concerned, interest rates are defined as follows: i) the interest rate on
overdraft facilities generally used by bank as the reference rate on short-term uncollateralized
loans; ii) the interest rate on medium and long-term (usually collateralized) loans. The interbank
rate is equal to the yearly average of the three-month Euribor rate.
We want to test whether competition from foreign players has an impact on provincial rates
either directly, through more aggressive credit policies, or indirectly, forcing other banks to reduce
their prices to retain customers. However, our database allows us to detect only the spillover effect
on the average rates applied in the province. Data on interest rates come from a survey that, until
2004, was limited to a sample of about 70 intermediaries, none of which was a foreign bank. Since
2004 onwards, the sample has been enlarged to about 250 banks (Bank of Italy, 2003), but very
few foreign intermediaries were included. At the end of 2005, foreign banks accounted for only 2.2
per cent of mortgages to households reported in the survey; for businesses, the coverage was even
lower, equal to 1.6 per cent for long-term loans and 0.3 per cent for short-term loans. Therefore, we
are not able to verify if these intermediaries follow more aggressive credit policies and apply lower
interest rates; we can only assess whether their presence induces incumbents to react to greater
competitive pressure which leads, on average, to lower interest rates.
To test this hypothesis, we used a simple specification where lending rates in difference to
either Euribor [Eq. 3] or deposit rates [Eq. 4] are a function of foreign banks market share (in
the province and market segment analysed) and other control variables, according to the following
equation:
[3] (Interest rate on loans
k
-Euribor)
p,t
= o + (Foreign banks market shares
k
)
p, t
+
(Controls
k
)
p, t
+ d
t
+ u
p
+ c
p,t

[4] (Interest rate on loans
k
Interest rate on deposits)
p,t
= o + (Foreign banks market shares
k
)
p,t
+
(Controls
k
)
p,t
+ d
t
+ u
p
+ c
p,t

where d
t
is a set of annual dummies, u
p
is the unobserved heterogeneity at the provincial
level. The subscript k refers to the separated regressions related to: mortgages to households,
overdraft facilities and long-term loans to firms. Data cover the period from 1997 to 2006.
Equations [3] and [4] contain a complete set of time dummies to take into account cyclical
effects not caught by other variables and to which the banks usually react with changes in the

12
However, Kok Sorensen and Lichtenberger (2007) use deposit rates as a proxy of the cost of funding when explaining the dispersion
in interest rates on mortgages across countries in the Euro area.

margins. As before, further differences across Italian geographical areas, for which we do not
control adequately, are captured using fixed effects at the provincial level.
13

We use the same regressors already discussed in the previous section. Again, all the right-
hand side variables are lagged by one period, to limit endogeneity problems. As previously
discussed, since data on interest rates on lending do not include foreign banks (at least up to 2004),
this further reduces the simultaneity issue that could arise between market shares held by foreign
intermediaries and interest rates.
We expect that the effect on interest rates stemming from the presence of foreign banks to
vary according to the market segments where these players compete: if foreign banks compete on
the same line of business with domestic banks, interest rates should be systematically reduced
(Martinez Peria and Mody, 2004). On the other hand, greater competitive pressure could drive
domestic banks towards opaque markets, where they have an information advantage and charge
higher interest rates. The result could be an increase in the average interest rate applied by
incumbents, due to a change in the composition of their customers (DellAriccia and Marquez,
2004). If this is the case, even an interest rate reduction to a good borrower would be difficult to
identify in aggregated data and the foreign presence could be associated to a positive effect on
average interest rates charged in the province.
Tables A6 and A7 report the results.
As for mortgages, a greater presence of foreign banks reduces the average spread (Table A6,
columns [1] and [3]). When the market share of foreign banks moves from the 25th to the 75th
percentile of the distribution, the spread between the lending rate and the Euribor decreases by 15
basis points. The results are very similar when the dependent is defined in terms of the difference
from deposit rates (about 13 basis points). Furthermore, although the coefficients of foreign banks
in the two sub-sample periods are very similar, the impact of their presence is stronger between
2002 and 2006 (column [2]): moving from the 25th to the 75th percentile, the spread decreases by
18 basis points, a reduction that is nearly double that detected in the first period.
This result is consistent with the arrival of foreign banks specialized in housing finance. In
the meantime, the growth of the Italian real estate market has boosted households indebtedness.
Competition in the mortgage market has intensified: innovative contracts have been introduced
especially by foreign banks and the range of available products has widened. The use of credit
scoring techniques could have contributed to these developments (Albareto et al. 2008); these
techniques have turned mortgage contracts into a relatively standardized product, making the entry
of foreign intermediaries easier.
Among the other regressors, the Herfindahl index is always statistically significant and
positively correlated to interest rates.
14
New branches boost competition and reduce interest rates
on mortgages. Household indebtedness (mortgages per capita in the province) does not affect the
spread between lending rates and the interbank rate, but it leads to an increase in the spread with
deposit rates (Table A6, col. [1] and [3]), in line with other empirical evidence (Kok Sorensen and
Lichtenberger, 2007). Financial market development and credit risk (non-performing loans over
total loans) do not have significant effects on interest rates.

13
Data on value added, employment rate, and so on, at a provincial level, are available with a high lag. Notwithstanding the loss of
observations, using these variables we do not detect any significant differences in the estimated coefficients.
14
The expected sign of the Herfindahl index is ambiguous and it leads to a different interpretation (see Berger, 1995, for a review).
Empirical evidence for Italy is mixed, too; Focarelli and Panetta (2003) find that concentration is significant, reducing deposit rates;
on the contrary, it does not affect interest rates for Bofondi, Bonaccorsi di Patti and Gobbi (2006), while it is significant for Fatica,
Fiori and Piersante (2006).

For business lending, the activity of foreign banks does not exert any statistically significant
effect on the interest rates applied to firms (Table A7, col. [1]-[8]).
On the other hand, well-developed financial markets reduce interest rates on short-term
lending, while the coefficient is not significant on long-term lending.
15
The Herfindahl
concentration index is not statistically significant, while opening up new branches reduces the
interest rates applied to medium and long-term loans to firms. Furthermore, the greater the share of
loans granted to small firms in the province, the higher the interest rates applied, due to a high
degree of opaqueness and the consequent difficulty of the market to evaluate creditworthiness.
Credit risk increases the short-term interest rate while it is not significant on the long-term rate.
The latter is affected by the degree of firms indebtedness, a result that could catch part of the
potential long-term default risk, which is not well controlled by non-performing loans. The degree
of firms indebtedness in the province is negatively correlated with short-term interest rates, most
likely behaving as a proxy for the development of the local financial market.
In our data, lending rates are reported only for loans larger than 75,000, giving rise to a
censoring of small lending which could bias our estimates. To address this issue, we added, to the
previous equations [3] and [4], a proxy for the coverage of the reported data in the survey, built as
the ratio between the amount of loans reported in the survey and the actual stock of loans (of the
same type and in the same area) reported by Bank Supervisory statements. The results, both in
direction and in size, do not change.
Our conclusions are confirmed also by weighted regressions (Table A8), where the weights
are computed as the foreign banks market shares at the provincial level.

4.3 The collateral required
Foreign banks could compete with domestic banks on other features of the loans rather than
in terms of interest rates, for instance by adopting a different policy for the collateral required. As a
result, the collateral could be reduced as a reaction to the increased competition, or it could be
raised as a device to screen good customers to whom lower interest rates could be offered
(Sengupta, 2007).
To verify these two opposite hypotheses, in the following regression we used the share of
collateralized loans over total loans as the dependent variable:
[5] (Collateralized loans/Total loans
k
)
p, t
= + (Foreign banks market shares
k
)
p, t
+
(Controls
k
)
p, t
+ d
t
+ u
p
+
p, t

We split real and personal guarantees. In the estimates (not reported) on short-term
collateralized loans, we do not detect any effect for households or firms. On the other hand,
significant results have emerged for medium and long-term collateralized loans (Table A9). For
households, an increase in the activity of foreign banks is associated with a reduction in real
collateral, partially balanced by an increase in the personal guarantees posted (Table A9, col. [1],
[2] and [3]). These results appear consistent with the growing diffusion of high loan-to-value
mortgages (where the share of real collateral falls), commonly backed by extra personal
guarantees.
This result holds for firms, too: the greater the presence of foreign intermediaries, the smaller
the share of collateralized loans (columns [5] and [6]). The other variables point to an increase in

15
It is significant when the spread is measured in relation to deposit rates. For these loans, the use of the interest rate on sight deposits
as proxies of the cost of long-term funding is particularly problematic and could capture a spurious effect.

the competitive pressure: the collateral posted is negatively correlated to the degree of financial
development and to the number of new branches opened; on the contrary, it increases along with
the greater presence of smaller firms.
In the two sub-periods analysed, the results are confirmed, both for households and firms.
As before, we verify the robustness of the estimates running weighted regressions. The main
findings are confirmed (Table A10).

5. Choosing where to operate: robustness checks with a two-step procedure
Our findings could be affected by the ex ante choice of the provinces where foreign banks
decide to operate significantly. To minimize their initial information disadvantage, these
intermediaries could only enter those areas where credit risk is lower and customer screening less
important; consequently, lending rates are lower in these areas.
16

This decision could give rise to a selection problem, due to the particular features of the area
where foreign banks operate, which is difficult to assess by means of instrumental variables
(Garen, 1988). Therefore, to address this issue we used a two-stage estimator, similar to that
proposed by Heckman (1978, 1979), with the aim of controlling for any remaining selection bias
driven by unobserved factors changing over time and in a different way across provinces.
However, despite the limited number of provinces where foreign banks decided to open a
branch, they grant loans to customers located in all provinces. Still, the level of activity is not
uniformly distributed over the different areas, with some provinces showing a level of foreign bank
intermediation close to zero.
In the first stage, we computed the probability that the level of activity of foreign banks in a
province is above a certain threshold, identified according to the distribution of foreign banks
market share. We considered different thresholds, equal to the 10th and to the 20th percentile of the
distribution;
17
corresponding respectively to market shares equal to about 1 and 1.5 per cent. The
selection choice is defined as follows:
[6] Selection equation: y
p, t
= + (X)
p, t
+ d
t
+
p, t
where y
p, t
=1 if foreign banks grant total loans greater than the 10th [or 20th] percentile
of their market shares distribution.
As a further test, we estimated a different selection equation, where the dependent variable is
equal to 1 if foreign intermediaries have at least one branch in that area. The decision to open a
branch indicates the aim of increasing the penetration in the area, but hinges on the need to balance
the related sunk costs with the expected profits. Therefore, we deem that the features of local
financial markets, the expected income and potential risks of the lending activity play a key role in
the entry decision (Gobbi and Lotti, 2004).
As already mentioned, even without any branches, the bank can operate by remote control.
The selection, differently from Heckman (1978, 1979), does not generate censoring of the

16
Such a critique should be particularly important for firms, for which lending by foreign banks is more concentrated in the northern
areas. Nonetheless, this is exactly the case in which we did not find any significant results. On the contrary, lending to households is
decidedly widespread nationwide and we detect a significant effect of foreign intermediaries.
17
It has also considered a threshold corresponding to the fifth percentile, which gives results very similar to those for the 10th
percentile.

observations in the structural equation. Lee (1978), dealing with a similar case, uses a selection
equation in the first stage (as in equation [6]) and estimates in the second step two structural
relations:
[7a] T
0
p,t
=
0
+
0
1
Z
0
p, t
+
0
2
| (.)
p, t
/(1-u (.))
p, t
+ q
0
p,t
if y
p,t
=0
[7b] T
1
p,t
=
1
+
1
1
Z
1
p, t
+
1
2
| (.)/u (.)
p, t
+ q
1
p,t
if y
p,t
=1
where, in our case, T represents the mobility index, interest rates or the share of
collateralized loans and Z indicates the control variables. The two equations are clearly augmented
with a further control, respectively | (.)/(1-u (.)) and | (.)/(u (.)), or the inverse Mills ratio.
Robinson (1989) estimates the two equations jointly, obtaining a more efficient estimator in
the following way:
[8] T
p,t
=
0
+ (
1
-
0
)y
p,t
+
1
Z
p, t
+
0
2
p, t
+ (
1
2
-

0
2
)
p, t
* y
p,t
+ d
t
+ u
p
+ q
p,t

where
p, t
is equal to | (.)
p, t
/(1-u (.))
p, t
if y
p,t
= 0 while it is equal to | (.)/u (.)
p, t
if y
p,t
= 1
Therefore, we estimate the equations reported in the previous section, conditioned to the
likelihood that foreign banks operate in the province with at least one branch or their activity
exceeds the selected threshold (equation [8]).
The various specifications of the selection equation [6] include, as regressors, the degree of
financial development of the area (total loans and deposits pro capite); credit risk (non-perfoming
loans over total loans in the province); and the Herfindahl concentration index. As before, d
t
is a
set of annual dummies.
We consider three identification variables. First, the population living in the province (in
logarithms), which captures the potential size of the market,
18
while it is not correlated to credit
variables in the second step. Second, we add the average distance of the province p from the
headquarters of foreign banks in Italy. The distance affects the knowledge of the area: the greater
the distance, the more difficult it is to screen the customers and monitor the relationship (Mian,
2006). Therefore, we expect it will reduce the activity level as well as the probability of opening a
branch in the area. Nevertheless, we cannot exclude that the lack of knowledge induces a foreign
intermediary to open a branch to fill the information gap. As a further identification variable, we
used the number of crimes against property carried out by unknown persons to the total number of
crimes against property in all provinces. This variable is intended as a proxy for social capital in
the province. The variable should capture the degree of opaqueness of the market, measuring the
share of crimes not dealt with by the judiciary system because the culprit could not be ascertained.
In Table A11, we report the different specification of equation [6] with probit models. The
results are in line with our expectations. Foreign banks tend to open a branch in more developed,
low concentrated and low risk credit markets.
The distance reduces the probability of operating and opening a branch. Population is always
significant and affects positively both the entry decision and the activity level in the province. The
number of crimes against property significantly reduces the probability of operating in the province
or opening a new branch.
Conditioning our previous estimates for probability estimated in equation [6], the results are
confirmed. For the sake of simplicity, in Table A12 we report only the main coefficients referring
to foreign banks market shares and the inverse Mills ratio, along with the correlations between the

18
Alternatively, the value added has been considered. Specifications that replace population with this variable lead to similar results.

errors in the first and second stage regressions (Rho). The significance of the inverse Mills ratio is
poor and mainly affects interest rates and collateral, showing that a selection process is at work in
the estimates. Nonetheless, the results are robust to the new specifications and previous
interpretations are confirmed.

6. Conclusions
In this paper we have measured the impact of foreign banks activity on Italian credit
markets. To capture the level of activity of foreign intermediaries, we used their market shares in
the different geographical areas and with reference to the various lines of business. We studied this
variable in relation to: i) customers mobility index at the provincial level; ii) interest rates applied
on average to the different customers (households and firms); and iii) the collateral required.
In the estimates, we controlled for other factors that could affect our dependent variables,
such as the degree of concentration of the market, the number of newly opened branches, credit
risk, the development of financial markets, and the degree of indebtedness of households and
firms. A two-step procedure controlled for possible endogeneity problems related to the provinces
where foreign banks mainly operate or have their establishments.
Our results show that foreign intermediaries exert significant competitive pressure in the
different lines of business analysed. They increase the mobility of customers, both households and
businesses. For the former, such a pressure comes with a reduction in the interest rates on mortgage
loans. The reduction is economically significant, equal to 15 basis points over the decade, about 18
basis points in the last five years. Recently, we have also observed a reduction in real collateral
posted; this finding is consistent with the steady diffusion of contracts whose loan-to-value is
higher than in the past.
There no evidence of a significant impact on interest rates applied to firms, though we
observe a slight reduction in the share of collateralized loans associated with the greater presence
of foreign intermediaries.
Consistently with the banking theory, in our view the different effect we detect for
households and businesses depends upon both the degree of standardization of the products
(greater for mortgages to households) and the exposure to asymmetric information (higher for
credit to firms). Firms heterogeneity remains to be analysed; conditions applied to business
lending are not very standardized. They depend on a firms characteristics and the bargaining
process between the bank and its customer. Most likely, in this field, competition from new
entrants is complex and more time is needed for its effects to be felt.

APPENDIX
Table A1
Market Share of Foreign Banks in Italy
(percentage values; annual averages)

1995 1998 2001 2003 2006
Total assets
5.2 9.1 6.8 6.7 13.7
Total loans
2.7 5.8 6.9 8.4 16.0
Interbank loans
15.7 22.6 10.8 10.2 17.8
Securities
6.3 11.5 11.0 6.2 14.1
Funding
2.0 3.3 2.9 3.9 12.0
Interbank deposits
16.3 27.2 20.9 21.3 28.3
Endorsement loans
5.4 9.7 16.8 11.4 16.2
Guarantees and commitments
42.6 67.5 11.0 6.1 12.3
Securities custody
5.6 6.6 10.3 10.4 23.5

Source: Bank Supervisory Report. Data refers to branches and subsidiaries.

Table A2
Descriptive Statistics

Average
Std.
Dev.
50 perc. 25 perc. 75 perc. Source

Market share instability:

Total credits to households
0.124 0.054
0.113
0.094 0.140 Supervisory Report, Bank of Italy
Mortgages to households
0.134 0.059
0.123
Short-term credits to households
0.182 0.077
0.168
Total credits to firms
0.145 0.068
0.129
Long-term credits to firms
0.179 0.080
0.162
Short-term credits to firms
0.163 0.083
0.144
Current account to firms
0.165 0.088
0.146
Interest rates differentials between:

Long-term rate to households and Euribor
1.880 0.953
1.727
1.227 2.409 Bank of Italys survey
Long-term rate to households
and funding rate
3.890 0.960
3.727
Overdraft rate to firms and Euribor
5.970 1.482 3.801 3.286 4.456 Bank of Italys survey
Overdraft rate to firms and fundig rate
7.980 1.406 7.994 7.013 8.984 Bank of Italys survey
Long-term rate to firms and Euribor
1.990 1.353
1.813
Long-term rate to firms and funding rate
4.001 1.322
3.801
Market share of credits to households:

Long-term credits with personal collateral
0.072 0.038
0.063
Long-term credits with real collateral
0.722 0.090
0.734
Market share of credits to firms:

Long-term credits with personal collateral
0.183 0.063
0.174
Long-term credits with real collateral
0.584 0.098
0.600
k

Table A2 (continued)
Descriptive Statistics

Average
Std.
Dev.
50 perc. 25 perc. 75 perc. Source
Foreign banks market share related
to:

Households credits
0.047 0.034
0.039
0.031 0.063
Supervisory Report, Bank of Italy
Mortgages to households
0.029 0.027
0.022
0.011 0.037
Short-term loans to households
0.090 0.067
0.071
0.035 0.132
Loans to firms
0.020 0.025
0.012
0.007 0.023
Long-term loans to firms
0.009 0.018
0.003
0.001 0.008
Short-term loans to firms
0.008 0.017
0.003
0.001 0.007
Current account to firms
0.007 0.015
0.002
0.001 0.006
Herfindahl index over households loans
0.118 0.052
0.106
0.082 0.140
Herfindahl index over households
mortgages
0.136 0.062
0.119
0.094 0.160
Herfindahl index over firms loans
0.105 0.040
0.095
0.077 0.122
Financial development
(loans+deposits per capita)
20.173 10.568
19.227
11.780 26.075
Supervisory Report and Istat
Loans to households per capita
2.804.4 1.474.4
2.406.1
1.664.8 3.784.4
Mortgages to households per capita
1.953.5 1.178.5
1.612.1
1.039.3 2.639.5
Short term loans to households per capita
850.9 351.9
798.9
577.6 1.060.9
Credits to firms over the Value Added
0.38 0.16
0.36
0.25 0.50
Long-term loans to firms over the Value
Added
0.13 0.05
0.12
0.09 0.16
Short-term loans to firms over the Value
added
0.13 0.06
0.12
0.08 0.17
Share of loans to small firms
0.204 0.107
0.182
0.122 0.274
Bad loans over total loans to household
0.087 0.074
0.062
0.036 0.106
Bad loans over total loans to firms
0.178 0.132
0.084
0.046 0.190
New branches (per 10,000 inhabitants)
0.143 0.307
0.092
0.033 0.183
Bank of Italy and Istat


Table A3
Market shares instability at t+1 in the households sector
(within estimator with provincial fixed effects)

Total loans Short-term loans Mortgages
[1] [2] [3] [4] [5] [6]

Foreign banks market
shares
t-1

0.119
(0.130)
___ -0.081
(0.101)
___ 0.403***
(0.137)
___
shares
t-1 1997-2001

___ -0.207
(0.252)
___ -0.441**
(0.206)
___ 0.448*
(0.253)
Foreign banksmarket
shares
t-1 2002-2006

___ 0.060
(0.135)
___ -0.117
(0.103)
___ 0.409***
(0.140)
Herfindahl
t-1

0.562***
(0.063)
0.554***
(0.063)
0.493***
(0.089)
0.475***
(0.090)
0.473***
(0.050)
0.473***
(0.050)
t-1

1.472**
(0.684)
1.543**
(0.685)
2.289**
(0.899)
2.397***
(0.899)
1.281*
(0.711)
1.273*
(0.712)
Log (Loans per capita
t-1
)
-0.090***
(0.020)
-0.094***
(0.020)
___ ___ ___

___

Log (Mortgages per capita
t-
1
)
___ ___ ___ ___ -0.047**
(0.019)
-0.046**
(0.019)
Log (Short-term loans per
capita
t-1
)
___ ___ -0.096***
(0.021)
-0.103***
(0.022)
___ ___
Bad loans
t-1

0.081
(0.058)
0.082
(0.058)
0.017
(0.083)
0.012
(0.082)
0.081
(0.062)
0.081
(0.062)
New branches
t-1

-0.002
(0.005)
-0.001
(0.005)
-0.000
(0.007)
-0.001
(0.007)
0.002
(0.005)
0.002
(0.005)
Number of observations 1.022

1.022

1.022

1.022

1.022

1.022

R
2
within 0.26 0.26 0.24 0.24 0.28 0.28

All the estimates include annual dummies and a constant term. Standard errors in parenthesis. * level of significance between 5 and
10 per cent, ** level of significance between 1 and 5 per cent, *** level of significance at 1 per cent.


Table A4
Market shares instability at t+1 in the firms sector

[1] [2] [3] [4] [5] [6]

Foreign banks market shares
t-1

0.347**
(0.141)
___ 0.693**
(0.267)
___ 0.360
(0.230)
___
shares
t-1

1997-2001

___ 0.365*
(0.215)
___ 0.476*
(0.255)
___ 0.905**
(0.375)
shares
t-1

2002-2006

___ 0.343**
(0.146)
___ 0.750***
(0.262)
___ 0.245
(0.239)
t-1

-0.209**
(0.086)
-0.208**
(0.086)
-0.256**
(0.107)
-0.261**
(0.107)
-0.269**
(0.105)
-0.294**
(0.106)
Herfindahl
t-1

0.230*
(0.128)
0.230*
(0.128)
0.619***
(0.150)
0.621***
(0.150)
0.076
(0.147)
0.055
(0.147)
t-1

1.494
(1.006)
1.512
(1.019)
0.885
(1.216)
0.757
(1.220)
-1.314
(1.108)
-1.177
(1.109)
Log(Loans over value added
t-2
)
-0.012
(0.019)
-0.013
(0.019)
___ ___ ___

___

Log(Mortgages over
value added
t-2
)
___ ___ ___ ___ -0.029*
(0.017)
-0.032*
(0.018)
Log (Short-term loans over
value added
t-2
)
___ ___ -0.023
(0.019)
-0.020
(0.019)
___ ___
Bad loans
t-1

0.012
(0.034)
0.012
(0.034)
-0.007
(0.042)
-0.012
(0.043)
0.057
(0.041)
0.060
(0.041)
New branches
t-1

0.004
(0.006)
0.004
(0.006)
0.009
(0.008)
0.009
(0.008)
0.000
(0.008)
0.001
(0.008)
Number of observations
R
2
within
919
0.06
919
0.06
919
0.05
919
0.06
919
0.10
919
0.10
All the estimates include annual dummies and a constant term. Standard errors in parenthesis. , * level of significance between 5 and


Table A5
Market shares instability at t+1
(within estimator weighted by total loans of foreign banks in all provinces)

[1] [2] [3]

Households
t-1
0.193
(0.187)
0.242*
(0.143)
0.376*
(0.198)
Firms
t-1
0.496**
(0.206)
0.887***
(0.338)
0.282
(0.338)

Table A6
Medium and long-term interest rates applied to households

Differential with respect to
Euribor
Differential with respect to funding
rate

[1] [2] [3] [4]

t-1

-6.255***
(1.527)

-5.475***
(1.737)

(Foreign banks market shares
t-1
)
1997-2001

-6.081***
(2.225)

-4.778*
(2.530)
(Foreign banks market shares
t-1
)
2002-2006

-6.244***
(1.531)

-5.432***
(1.741)
Herfindahl
t-1

1.474**
(0.571)
1.478**
(0.572)
1.960***
(0.649)
1.974***
(0.650)
t-1

-1.127
(7.559)
-1.206
(7.599)
4.048
(8.594)
3.731
(8.639)
t-1
)
0.291
(0.214)
0.293
(0.214)
0.796***
(0.243)
0.804***
(0.244)
Bad loans
t-1

-0.015
(0.611)
-0.016
(0.612)
-0.057
(0.695)
-0.063
(0.696)
New branches
t-1

-0.171***
(0.057)
-0.172***
(0.057)
-0.141**
(0.065)
-0.142**
(0.065)
Number of observations 1.030 1.030 1.030 1.030
R
2
within
0.73 0.73 0.60 0.60

Table A7
Interest rates applied to firms

Differential with respect to Euribor Differential with respect to funding rate

Current account Mortgages Current account Mortgages
[1] [2] [3] [4] [5] [6] [7] [8]

shares
t-1

-1.147 -2.667 -1.956 -2.058

(2.369) (3.666) (2.628) (3.772)
shares
t-1 1997-2001

-2.496 -3.873 -3.472 -2.960

(2.535) (5.535) (2.813) (5.693)
shares
t-1

2002-2006

1.492 -2.231 1.008 -1.799

(2.959 (3.856) (3.283) (3.967)
Share of loans to small
firms
t-1

2.576*** 2.497*** -0.035 0.009 3.413*** 3.323*** 0.666 0.700
(0.855) (0.856) (1.655) (1.663) (0.949) (0.950) (1.702) (1.711)
Herfindahl
t-1
0.521 0.477 -0.528 -0.478 0.841 0.792 -0.104 -0.067

(1.235) (1.234) (2.379) (2.387) (1.370) (1.369) (2.447) (2.455)
t-
1

-41.001*** -42.882*** 27.147 26.934 -30.455*** -32.568*** 40.821** 40.662**

(9.696) (9.771) (19.179) (19.204) (10.758) (10.840) (19.729) (19.755
Log (Loans over value
added
t-2
)
-0.452*** -0.422*** 0.709* 0.712* -0.302* -0.268 0.742** 0.744**

(0.153) (0.154) (0.372) (0.372) (0.169) (0.171) (0.382) (0.383)
Bad loans
t-1
0.657** 0.617* -0.025 -0.030 0.928** 0.884** 0.242 0.238

(0.337) (0.338) (0.648) (0.648) (0.374) (0.375) (0.666) (0.667)
New branches
t-1
-0.009 -0.010 -0.263** -0.263** 0.027 0.026 -0.227* -0.228*
(0.064) (0.064) (0.123) ((0.123) (0.071) (0.071) (0.127) (0.127)
Number of obs.
919 919 919 919 919 919 919 919
R
2
within 0.76 0.76 0.38 0.38 0.37 0.38 0.24 0.24
All the estimates include annual dummies and a constant term. Standard errors in parenthesis.
* level of significance between 5 and 10 per cent, ** level of significance between 1 and 5 per cent ,*** level of significance at 1 per
cent.

Table A8
Medium and long-term interest rates


Differential with respect to
Euribor
Differential with respect to funding rate

Households
t-1
-6.337**
(2.687)
-7.758***
(2.346)
Firms
Current account Mortgages Current account Mortgages
t-1
-1.428
(3.451)
-4.379
(5.154)
-1.625
(3.028)
-3.917
(5.021)


Table A9
Medium and long-term collateral

Households Firms
Personal Real Personal Real

[1] [2] [3] [4] [5] [6]

t-1
0.199*** -0.677*** 0.201 -0.568***
(0.048) (0.137) (0.161) (0.181)
shares
t-1

1997-2001

-0.569*** -0.474*
(0.199) (0.273)
shares
t-1 2002-2006

-0.670*** -0.595***
(0.137) (0.190)
t-1
0.090 0.158** 0.155*
(0.073) (0.082) (0.082)
Herfindahl
t-1
-0.013 -0.080 -0.079 -0.150** 0.059 0.057
(0.018) (0.051) (0.051) (0.066) (0.074) (0.074)
t-1
0.420* 0.352 0.303 1.365 -2.947*** -2.932***
(0.241) (0.676) (0.679) (0.835) (0.937) (0.938)
t-1
) -0.010 -0.033* -0.032*
(0.007) (0.019) (0.019)
Log (Loans over value added
t-2
) 0.014 -0.025 -0.025
(0.017) (0.019) (0.019)
Bad loans
t-1
0.044** -0.030 -0.030 -0.027 -0.034 -0.034
(0.019) (0.055) (0.055) (0.028) (0.031) (0.031)
New branches
t-1
-0.000 -0.008 -0.008 0.006 -0.014** -0.013**
(0.002) (0.005) (0.005) (0.005) (0.006) (0.006)
R
2
within
1.030
0.44
1.030
0.29
1.030
0.29
919
0.08
919
0.19
919
0.19
All the estimates include annual dummies and a constant term. Standard errors in parenthesis. * level of significance between 5 and 10
per cent, ** level of significance between 1 and 5 per cent, *** level of significance at 1 per cent.

Table A10
Medium and long-term collateral

Personal Real

Households
Foreign bank market sharest-1 0.346***
(0.069)
-0.826***
(0.190)
Firms
Foreign bank market sharest-1 0.319*
(0.186)
-0.975***
(0.246)

Table A11
Probability of operating or entering a province with an establishment
(probit estimates)

dummy 10
percentile
dummy 20
percentile
dummy presence
of
establishments

[a] [b] [c]

t-1

-14.946
(24.081)
21.837
(15.255)
28.984***
(10.301)
Herfindahl
t-1

-7.995***
(1.882)
-11.339***
(1.613)
-4.413***
(1.415)
Bad loans
t-1

-7.420***
(1.640)
-2.479**
(1.162)
-1.923**
(0.865)
Log of population
t-1

3.813***
(0.386)
2.664***
(0.228)
1.761***
(0.127)
Distance
-0.004***
(0.001)
-0.003***
(0.000)
-0.001***
(0.000)
Crimes against properties by person(s) unknown
t-1

-9.318***
(2.386)
-2.023
(1.601)
-3.735***
(1.104)
1030 1030 1030
Pseudo R
2

0.685 0.546 0.418


Table A12
Corrections for the operativity and presence of a foreign bank establishment

[a] [b] [c]
Dependent
variable
Foreign
banks
market
shares
t-1

Inverse
Mills
Ratio (1)
Rho
Foreign
banks
market
shares
t-1

Inverse
Mills
Ratio (1)
Rho
Foreign
banks
market
shares
t-1

Inverse
Mills
Ratio (1)
Rho

Market share
instability in
households
mortgages
0.424 ***
(0.142)
-0.003
(0.009)
-0.05 0.410 ***
(0.143)
0.008
(0.007)
0.14 0.439***
(0.143)
0.016
(0.013)
0.28
Market share
instability in short-
term loans to firms
0.691**
(0.267)
-0.020
(0.014)
-0.20 0.701***
(0.267)
0.007
(0.011)
0.07 0.681**
(0.266)
-0.039*
(0.022)
-0.44
Market share
instability in long-
term loans to firms
0.357
(0.231)
0.024*
(0.014)
0.24 0.387*
(0.231)
0.025
(0.010)
0.30 0.373
(0.232)
0.012
(0.021)
0.16
Medium and long-
term interest rate
to households
-6.284***
(1.533)
-0.058
(0.095)
-0.09 -6.152***
(1.535)
0.051
(0.074)
0.08 -6.205***
(1.532)
0.193
(0.138)
0.29
Interest rates on
overdraft to firms
-1.004
(2.351)
0.259**
(0.114)
0.22 -1.178
(2.359)
-0.008
(0.088)
-0.01 -0.911
(2.349)
0.343**
(0.172)
0.36
Medium and long-
term interest rates
to firms
-2.756
(3.672)
-0.040
(0.217)
-0.03 -2.851
(3.679)
0.070
(0.167)
0.05 -2.908
(3.690)
0.001
(0.334)
0.00
Real collateral to
mortgages to
households
-0.671***
(0.137)
0.004
(0.008)
0.05 -0.673***
(0.137)
0.005
(0.007)
0.06 -0.653***
(0.137)
0.021*
(0.012)
0.25
Real collateral to
medium and long-
term loans to firms
-0.581***
(0.181)
0.026**
(0.011)
0.23 -0.597***
(0.181)
0.009
(0.008)
0.11 -0.620***
(0.180)
-0.005
(0.016)
-0.05
(1) Computed with estimates reported in the respective columns of table A11.

REFERENCES
Albereto, G., M. Benvenuti, S. Moretti, M. Pagnini and P. Rossi (2008), Lorganizzazione
di unindagine campionaria, Banca dItalia, Questioni di Economia e Finanza, 12.
Aliber, R.Z. (1984), International banking: a survey, Journal of Money, Credit and Banking, 16:
661-678.
Angelini, P. and N. Cetorelli (2003), The effects of regulatory reform on competition in the
banking industry, Journal of Money, Credit and Banking, 35(5): 663-684.
Banca dItalia (2007), Relazione Annuale anno 2006, Rome, 31 May.
Banca dItalia (2003), Larmonizzazione delle statistiche europee sui tassi dinteresse bancari e le
scelte metodologiche italiane, Supplementi al Bollettino Statistico Note metodologiche e
informazioni statistiche, 57, 29 October.
Banca dItalia (2007), Leconomia della Lombardia nellanno 2006, Milan, May.
Berger, A.N. (1995), The profit-structure relationship in banking-tests of market power and
efficient structure hypotheses, Journal of Money, Credit and Banking, 27: 404-456.
Berger, A.N., R. DeYoung, H. Genay and G.F. Udell (2000), Globalisation of financial
institutions: Evidence from cross-border banking performance, Brookings-Wharton papers
on financial services, 23-158.
Berger, A.N., L.F. Klapper and G.F. Udell, (2001), The ability of banks to lend to informationally
opaque small businesses, Journal of Banking and Finance, 25: 2127-2167.
Berger, A.N., N.H. Miller, M.A. Petersen, R.R. Rajan and J.C. Stein (2005), Does function follow
organizational form? Evidence from the lending practices of large and small banks, Journal
of Financial Economics, 76: 237-269.
Bonaccorsi di Patti, E., G. Eramo and G. Gobbi (2005), Piccole e grandi banche nel mercato del
credito in Italia, Banca Impresa Societ, 1: 3-34, April.
Bofondi, M., E. Bonaccorsi di Patti and G. Gobbi (2006), Gli effetti dellingresso di nuove banche
nei mercati locali del credito, Rivista Italiana degli Economisti, 1: 27-50.
Brealey, R.A. and E.C. Kaplanis (1996), The determination of foreign banking location, Journal
of International Money and Finance, 15(4): 577-597.
Bugamelli, M. and L. Infante (2003), Sunk costs of exports, Banca dItalia, Temi di Discussione,. 469.
Casolaro, L., L. Gambacorta and L. Guiso (2005), Regulation, formal and informal enforcement
and the development of the household loan market. Lesson from Italy, Banca dItalia, Temi
di Discussione, 560.
Caves, R.E. and M.E. Porter (1978), Market structure, oligopoly, and stability of market shares,
The Journal of Industrial Economics, 26 (4): 289-313.
Cesarini, F. (1981), Le filiali di banche estere in Italia, in F. Capriglione and V. Mezzacapo
(eds), Il sistema valutario italiano, Milan, Giuffr.
Claessens, S., A. Demirg-Kunt and H. Huizinga (2001), How does foreign entry affect domestic
banking markets?, Journal of Banking & Finance, 25(5): 891-911, May.
Claessens, S. and L. Laeven (2003), What drives bank competition? Some international
evidence, World Bank Policy Research Working Paper, 3113.

Claessens, S. and N. Van Horen (2008), Location decisions of foreign banks and institutional
competitive advantage, DNB Working Paper,.172/2008.
Clarke, G.R.G., R. Cull and M.S. Martinez Peria (2001), Does foreign bank penetration reduce
access to credit in developing countries?, World Bank Policy Research Working Paper, 2716.
Claeys, S. and C. Hainz (2006), Acquisition versus greenfield: the impact of the mode of foreign
bank entry on information and bank lending rates, ECB Working Paper Series, 653, July
2006.
Davies S.W. and P.A. Geroski (1997), Changes in concentration, turbulence, and the dynamics of
market shares, The Review of Economics and Statistics, 79(3): 383-391.
Davis, E.P. (1990), International financial centres: An industrial analysis, Bank of England,
Discussion Paper, 51.
De Bonis, R and A. Ferrando (2000), The multimarket contacts theory: An application to Italian
banks, Banca dItalia, Temi di Discussione, 387.
Dell'Ariccia, G. and R. Marquez (2004), Information and bank credit allocation, Journal of
Financial Economics, 72(1): 185-214.
Detragiache, E., T. Tressel and P. Gupta (2006), Foreign banks in poor countries: Theory and
evidence, IMF Working Paper, 18.
De Haas, R. and I. Van Lelyveld (2003), Foreign banks and credit stability in central and eastern
Europe. A Panel Data Analysis, DNB Working Paper, 109/2003.
Dunning, J.H. (1977), Trade, location of economic activity and the MNE: a search for an eclectic
approach, in B. Ohlin, P.O. Hesselborn P.O. e P.K. Wijkman (eds.), The international
allocation of economic activity, London, Macmillan.
Fatica, S., R. Fiori and F. Piersante (2006), Competition in the Italian banking system: some new
empirical evidence, Banca dItalia, mimeo.
Focarelli, D. and F. Panetta (2003), Are mergers beneficial to consumers? Evidence from the
market for bank deposits, American Economic Review, 93(4): 1152-1172.
Focarelli, D. and A.F. Pozzolo (2001),The patterns of cross-border bank mergers and
shareholdings in OECD countries, Journal of Banking & Finance, 25: 2305-2337.
Garen, J. (1988), Compensating wage differentials and the endogeneity of job riskiness, Review
of Economics and Statistics, 70(1): 3-16.
Giannetti, M. and S. Ongena (2005), Financial integration and entrepreneurial activity evidence
from foreign bank entry in emerging markets, European Central Bank, Working Paper, 498.
Giannetti, M. and S. Ongena (2009), Lending by example: Direct and indirect effects of foreign
banks in emerging markets, 1 September, mimeo.
Gobbi, G. and F. Lotti (2004), Entry decisions and adverse selection: An empirical analysis of
local credit markets, Journal of Financial Services Research, 26: 225-244.
Goldberg, L.G. and D. Johnson (1990), The determinants of U.S. banking activity abroad,
Journal of International Money and Finance, 9(2): 123-137.
Goldberg, L.G. and A. Saunders (1981), The determinants of foreign banking activity in the
United States, Journal of Banking and Finance, 1: 17-32.
Gormley, T.A. (2010), The impact of foreign bank entry in emerging markets: Evidence from
India, Journal of Financial Intermediation, 19: 26-51.

Grosse, R.E. and L.G. Goldberg (1991), Foreign bank activity in the United States: An analysis by
country of origin, Journal of Banking and Finance, 15: 1093-1112.
Grubel, H. (1977), A theory of multinational banking, Banca Nazionale del Lavoro, Quarterly
Review, 349-363.
Heckman, J.J. (1978), Dummy endogenous variables in a simultaneous equation system,
Econometrica, 46(4): 931-959.
Heckman, J.J. (1979), Sample selection bias as specification error, Econometrica, 47(1): 153-161.
Heggestad, A.A. and S.A. Rhoades (1976), Concentration and firm stability in commercial
banking, The Review of Economics and Statistics, 45, November.
Kato, M. and Y. Honjo (2006), Market share instability and the dynamics of competition: a panel
data analysis of Japanese manufacturing industries, Review of Industrial Organization, 28:
165-182.
Kok Sorensen, C. and J.-D. Lichtenberger (2007), Mortgage interest rate dispersion in the euro
area, ECB Working Paper Series, 733.
Lee, L.F. (1978), Unionism and wage rates: A simultaneous equations model with qualitative and
limited dependent variables, International Economic Review, 19(2): 415-433.
Lehner, M. and M. Schnitzer (2006), Entry of foreign banks and their impact on host countries,
Munich Economics - Discussion Paper 2006-29, September.
Levine, R. (2003), Denying foreign bank entry: Implications for bank interest margins, Central
Bank of Chile, Working Papers, 222,August.
Magri, S., A. Mori and P. Rossi (2005), The entry and the activity level of foreign banks in Italy:
An analysis of the determinants, Journal of Banking & Finance, 29 (5): 1295-1310.
Mahajan A., N. Rangan and A. Zardkoohi (1996), Cost structures in multinational and domestic
banking, Journal of Banking & Finance, 20: 283-306.
Martinez Peria, M.S. and A. Mody (2004), How foreign participation and market concentration
impact bank spreads: Evidence from Latin America, Journal of Money, Credit and Banking,
Ohio State University Press, 36 (3): 511-537, June.
Mian, A. (2006), Distance constraints: The limits of foreign lending in poor economies, Journal
of Finance, LXI, (3): 1465-1505.
Robinson, C. (1989), The joint determination of union status and union wage effects: Some tests
of alternative models, Journal of Political Economy, 97(3): 639-667.
Sagari, S.B. (1992), United States foreign direct investment in the banking industry,
Transnational Corporations, 1(3): 92-123.
Sakakibara, M. and M.E. Porter (2001), Competing at home to win abroad: Evidence from
Japanese industry, Review of Economics and Statistics, 83: 310-322.
Sapienza, P. (2002), The effect of banking mergers on loan contracts, Journal of Finance, 57(1):
329-367.
Sengupta, R. (2007), Foreign entry and bank competition, Journal of Financial Economics,
84:502-528.
Shepherd, W.G. (1970), Market power and economic welfare, New York, Random House Inc.

Uiboupin, J. (2004), Effects of foreign banks entry on bank performance in the CEE countries,
University of Tartu, Working Paper.
Van Tassel, E. and S. Vishwasrao (2005), Asymmetric information and the mode of entry in
foreign credit markets, Department of Economics, Florida Atlantic University, Working
Paper.
Williams, B. (1996), Determinants of the performance of Japanese financial institutions Australian
1987-1992, Applied Economics, 28: 1153-1165.
Williams, B. (1997), Positive theories of multinational banking: Eclectic theory versus
internalisation theory, Journal of Economic Surveys, 11(1): 71-100.
Williams, B. (1998), Factors affecting the performance of foreign-owned banks in Australia: A
cross-sectional study, Journal of Banking and Finance, 22: 197-219.
Woodrow Eckard, E (1987), Competition and market shares instability, Journal of Business, 60
(4): 539-552.
Yamori, N. (1998), A note on the location choice of multinational banks: The case of Japanese
financial institutions, Journal of Banking and Finance, 22: 109-120.

DEBT RESTRUCTURING AND
THE ROLE OF LENDING TECHNOLOGIES
Giacinto Micucci
-
and Paola Rossi
**

1. Introduction
Debt restructuring may preserve the ongoing concern of firms facing financial distress but
which still have profitable investment projects. It is a complex bargaining process which involves
the firm and its lenders. When a borrower faces financial distress, the lender bank has to decide
whether to take part in the workout process or not. We study this decision, taking into account
banks heterogeneity in terms of organization and lending technologies.
Most of the existing literature on this argument considers bank debt easier to renegotiate than
public debt, because banks are better able to produce and process information on firms economic
prospects, especially in the case of small and medium-sized enterprises (SMEs). Banks may both
mitigate information asymmetries and facilitate coordination among lenders (Bolton and
Scharfstein, 1996; Bris and Welch, 2005). They decide whether to liquidate or to reorganize the
firm by providing the extra funds needed to ensure that the firm stays in operation. However, there
is wide cross-country evidence that firms maintain relationships with many banks.
1
In a multiple
banking framework, banks may behave differently as regards workout decisions, i.e. some banks
decide to restructure their loans to financially distressed firms, while others do not. To the best of
our knowledge, so far heterogeneity in restructuring has received little attention in the literature.
Recent theoretical and empirical literature have highlighted the role of different organization
and lending technology adopted by the bank in shaping the type of relationship developed with the
borrower (Stein, 2002; Berger and Udell, 2002; Berger et al., 2005a; Berger, Frame and Miller,
2005b). We expect these factors to be especially important in the renegotiation process, when
coordination problems arise and information asymmetries become even more relevant. Using a
recent survey about Italian banks, we are able to link the choice to restructure the debt of a
financially distressed firm to a series of variables that pin down organizational factors within the
lending banks. Thus, this paper connects the literature on bank organization with the literature on
debt restructuring.
In order to evaluate the restructuring decision, we perform an econometric analysis on Italian
firms and their lending banks. The Italian case is particularly interesting to analyse: multiple
banking relationships are widespread among SMEs (Detragiache, Garella and Guiso, 2000), while
a harsh bankruptcy regime and the inefficiencies of the judicial system imply that reorganizations
are better done privately. We use loan-level data, with the addition of information on both
borrowing firms and lending banks. Starting from firms financial statements (Cerved), we select
those firms that faced financial distress (that is, whose cash flows fell below their net financing
costs) at least once between 2002 and 2004. Balance-sheet data have been combined with

-
Bank of Italy, Ancona.
**
The authors wish to thank for helpful comments two anonymous referees of the Temi di Discussione, Antonio Accetturo, Michele
Benvenuti, Paola Bongini, Giorgio Calcagnini, Luigi Cannari, Hans Degryse, Moshe Kim, Pasqualino Montanaro, Andrea Nobili,
Massimo Omiccioli, Steven Ongena, Marcello Pagnini, Gregory F. Udell, participants at seminars held at the Bank of Italy, at the
Barcelona Banking Summer School, at the Ancona Conference on The Changing Geography of Money, Banking and Finance in a
post-crisis world.
1
According to Ongena and Smith (2000), large firms located in 22 countries have an average of 5.6 banking relationships.
144 Giacinto Micucci and Paola Rossi

information concerning relations with the banking system (Central Credit Register) and with a
special survey carried out in 2007 by the Bank of Italy on the organization and technology used in
lending activities by Italian banks (Bank organization survey). The survey investigates the degree
of bank decentralization and the comparative relevance of soft and hard information (credit
scoring) in assessing credit risk. Thus we have a sample of about 30,000 loan-level observations
(bank-firm relations), relating to more than 9,000 distressed Italian firms and 300 Italian banks.
These data allow us to evaluate the divergent strategies followed by banks when a borrowing
firm becomes financially distressed. We define the restructuring decision either as the rescheduling
of bank loan maturity or the extension of new loans to a firm in distress. These interventions loosen
the borrowers financing constraints. They are economically significant: in our sample, among the
firms that were able to renegotiate outstanding debt with at least one of the lending banks, only 8
per cent leaves the market, against 30 per cent among those firms that did not renegotiate their
debt. Therefore, even if we do not address the default issue, we are confident that studying banks
decisions has an important economic motivation.
When we study this decision, we have to take into account the sample selection issue. While
we have a representative sample of firms in distress, we only observe the banks which actually lent
to those firms: firms choose the banks from which they borrow and banks decide whether to lend or
not. This selection process, clearly endogenous, may bias the results. It is related to firm
characteristics, such as size, risk and location, and, more importantly, to the credit policy of the
bank. These elements are likely to influence, ex ante, the characteristics of lending banks in terms
of size, technology, etc. and, ex post, the restructuring decision. We address this issue by
estimating a probit model with sample selection. In the selection equation, we consider as potential
lenders all those intermediaries that have at least one branch in the province where the firm is
located. In this way, we study the restructuring decision which all the banks operating in the local
credit market could have taken with respect to the firms in distress.
We find that bank heterogeneity accounts for the decision to restructure credit to SMEs in
distress. While one third of the firms can restructure outstanding debt with at least one bank, it
never happens that all the lending banks decide to restructure simultaneously. The probability of
restructuring is higher when a lending bank relies more on soft information (relationship lending)
rather than on hard information (transactional lending) and adopts a decentralized structure with
more power allocated to local managers. The banks restructuring decision also depends negatively
on its distance from borrowing firms.
However there are significant complementarities among organizational variables: the
adoption of credit scoring increases the likelihood of restructuring if banks also use these
techniques systematically in the monitoring process and if they adopt more decentralized
structures.
From a methodological point of view, our results emphasize the importance of taking bank
heterogeneity into account when analysing the restructuring process. From a policy point of view,
we argue that even if bank consolidation and the increasing use of credit scoring may raise
concerns for SMEs in distress, the enhanced implementation of statistical techniques and effective
decentralization of decision-making can offset their effects.
The paper is organised as follows. Section II contains a brief review of the theoretical
background. Section III describes data sources and introduces some descriptive statistics. The
empirical results are reported in Sections IV and V. Section VI concludes.

Debt restructuring and the role of lending technologies 145

2. Literature review
This paper is rooted in two different strands of literature. The first analyses the bank-debt
restructuring in the event of financial distress. The second studies the role of different bank
organizations and lending technologies in company financing. To our knowledge, so far these two
strands have followed different tracks. In recent years research assuming banks heterogeneity has
blossomed at least when analysing variables such as credit availability, its price and risk, and
market entry decisions while the empirical literature dealing with financial restructuring has
continued to assume that bank behaviour is homogenous.
According to Jensen (1989), as long as the going-concern value of the distressed firm
exceeds the liquidation value, its debt will be renegotiated to assure survival. Moving from this
statement, a well-developed body of literature has focused on the borrower-lender bargaining
process in the design of debt contracts (Hart and Moore, 1998). With complete contracts, ex ante
bargaining makes renegotiation in cases of distress unnecessary. On the contrary, when contracts
are incomplete there is room for ex post renegotiation. Nonetheless, renegotiation is hampered by
information problems, causing both coordination failures and difficulties in evaluating company
prospects. More in general, the literature focuses on three aspects that may affect debt
restructuring: i) the coordination between lenders; ii) the evaluation of firms prospects after the
occurrence of the financial distress (information asymmetries); and iii) the lenders incentives to
restructure.
As regards coordination, the presence of many lenders may lead to liquidation even if this is
economically inefficient (White, 1989). Most of the theory assumes that coordination failures arise
when debt is public or, at least, highly dispersed, while privately-held debt mostly bank loans is
easier to renegotiate (Brown, James and Mooradian, 1993; Bolton and Scharfstein, 1996; Bolton
and Freixas, 2000). If creditors do not coordinate their responses to a renegotiation offer or to a
default threat, a single creditor will prefer to free-ride on the debt restructuring offer.
2
An increase
in the number of lenders reduces the incentive of the borrower to repudiate debt payments
(strategic default incentive), since contracts are more difficult to renegotiate (Bergloef and von
Thadden, 1994; Bolton and Scharfstein, 1996; Manove, Padilla and Pagano, 2001). In a recent
paper, however, Bris and Welch (2005) reach the opposite conclusion on the optimal number of
creditors: since creditors recover less when they are dispersed, a firm that opts for multiple
creditors ex ante has a better bargaining position in the case of financial distress ex post. Therefore,
a good quality firm may choose just one house bank to signal its confidence it will not go bankrupt,
while a risky firm will tend to increase the number of lenders. Similarly, Carletti, Cerasi and
Daltung (2007) predict a greater use of multiple-bank lending when firms are less profitable and
monitoring costs are high. Coordination may be achieved by means of a formal contract among the
creditors, such as bank pools (Brunner and Krahnen, 2008). However, large pools (i.e. pools with
many member banks) have a negative impact on the likelihood of a successful turnaround.
3
Again,
the evidence is against a simple trade-off between the number of banks and the borrowers ability
to obtain a debt renegotiation.
It is particularly difficult to evaluate firm prospects when financial distress occurs, because
of lower reliance on hard information about a firms records (Giammarino, 1989; Boot, 2000;
Davydenko and Franks, 2008). Relationship lending is assumed to help to restructure the

2
Small lenders have an incentive to hold out, even if restructuring would be in the general interest (Bergman and Callen, 1991;
Gertner and Scharfstein, 1991; Rajan, 1992).
3
Using Italian data, Carmignani and Omiccioli (2007) evaluate the conflicting effects of the concentration of creditors, showing that
the overall effect of a high concentration of bank credit is a higher probability of financial distress but a lower probability of
liquidation.

outstanding debt of firms facing financial distress (Longhofer and Santos, 2000; Hoshi, Kashyap
and Scharfstein, 1990). It is worth mentioning that in this situation creditors could prefer to
postpone the liquidation decision, leaving the debt high and waiting for more information about the
firms viability. This is what Kahl (2002) calls controlled liquidation. Liquidation is assumed to
be a dynamic process, in which postponing the decision could be economically efficient for the
bank. In fact, it may limit the downside risk; if assets lose value, creditors may proceed with an
informed liquidation decision later on. This strategy comes at the cost of inducing less efficient
investment decisions on the part of distressed firms. Therefore, it is an optimal choice only when
there is a fair degree of uncertainty about a firms recovery prospects.
In the literature on financial restructuring, creditors within the same category or seniority are
commonly assumed to behave uniformly. While dispersed publicly-owned debt is considered
separately from more concentrated bank debt (Bolton and Freixas, 2000), it is frequently assumed
that banks will follow similar strategies, with few exceptions.
4
Detragiache, Garella and Guiso
(2000) also consider the possibility that different banks may behave in different ways according to
their liquidity problems. Moreover, according to Manove, Padilla and Pagano (2001) well secured
banks may under invest in project screening activities (the lazy bank hypothesis). Kordana and
Posner (1999) suggest that the secured creditors incentives are skewed towards liquidation over
reorganization.
Banks have usually been thought of as single agents, thus neglecting the complexity of their
organizational forms and the potentially diverging objectives of the agents inside them.
5
However,
a more recent strand of literature, deriving from firm theory, has underlined the importance of
banks heterogeneity also in an organizational perspective (Stein, 2002). The adoption of different
lending technologies is intrinsically related to the design of the proper incentives within the bank.
According to Aghion and Tirole (1997), the amount of communication within an organization
depends on the allocation of formal authority. Increasing the decision-making power at the lower
level of the hierarchy fosters information gathering and initiative, the cost being a loss of control
over local managers, who may pursue their own private interests. We believe that organizational
schemes and lending technologies are also important when the bank has to decide whether to
restructure its loans to a distressed firm.
Information on a firms prospects is clearly related to lending technologies. The adoption of
remote banking and credit scoring tends to enhance the use of impersonal methods of contact:
transaction-based lending may substitute relationship lending and this could affect bank lending
policy when it comes to restructuring decisions, since soft information may be less important. Mian
(2006) finds that in Pakistan foreign banks are less prone to renegotiate bilaterally in the case of
default than domestic banks, suggesting that relationship lending makes financial restructuring
easier than transactional lending. On the other hand, the effect of credit scoring techniques is not
obvious: improved accuracy in evaluating creditworthiness may allow the bank to expand its
lending towards informationally opaque firms (Berger et al., 2005a); it may also reduce moral
hazard, a key factor in cases of distress.
Relationship lending is based on proprietary information collected, in particular, by the loan
officer through repeated interaction over time with the firm (Berger and Udell, 2002).
Decentralization increases the incentives to collect soft information, which is not wasted along
hierarchical lines (Stein, 2002). It significantly alters the interaction between the banks

4
In Bulow and Shoven (1978), three classes of claimants are analysed: bondholders, bank lenders and equity holders. Bondholders
are assumed to be a non-cohesive group of lenders; therefore, to avoid bankruptcy, the main bank must provide the extra funds
necessary to pay other creditors. Small banks are not included in the model, while their loans are placed with the bondholder claims.
Therefore, bank heterogeneity is not analysed in the model.
5
Rajan (1992) assumes that bank debt is easily renegotiated because the bank is a monolithic, readily accessible creditor.

headquarters and its local branches and thus the decision-making process in the financial
restructuring phase. Power delegation may convey stronger incentives towards debt restructuring;
among the drawbacks, it may expose the loan officer to the risk of capture by the borrowing firm,
delaying liquidation even when this is the most efficient solution. For many reasons, it may be in
the loan officers interest to conceal a deteriorating borrowers conditions, a problem that is
exacerbated by delegating more authority to loan officers (Berger and Udell, 2002).
Agency costs to monitor local managers tend to increase with distance. The top management
of the bank will be less familiar with the local economic and social environment. Greater physical
and cultural distance makes it difficult to collect and communicate soft information and to monitor
local managers (Degryse and Ongena, 2004; Alessandrini, Presbitero and Zazzaro, 2009). This
could explain why the allocation of decision-making power to these officers tends to decrease with
distance (Mocetti, Pagnini and Sette, 2010) and why foreign banks tend to be weaker at relational
functions such as renegotiation and recovery of bad loans, functions that require strong information
and control mechanisms (Mian, 2006).
More detailed measures of bank-organization heterogeneity have gradually begun to appear
in the literature, recognizing the importance of the organizational schemes and technological
progress (Berger, 2003; Berger et al., 2005a, 2005b; Degryse et al., 2009; Frame et al., 2001;
Petersen and Rajan, 2002; Scott, 2006). However, studies adopting a bank heterogeneity
perspective have focused on aspects such as credit availability, its price and risk, and market entry.
To the best of our knowledge, so far the literature on debt restructuring has instead downplayed the
role of bank organization and lending technology.

3. Data description and stylized facts
To perform the econometric analysis, we build a unique data set at the loan level, using three
sources of information, respectively regarding firms, their lending banks, and the characteristics of
bank-firm relations.
First we use a dataset from Cerved

consisting of company accounts. Cerved is a reliable
source of information on Italian companies. Information is drawn from official data recorded at the
Italian Registry of Companies and from financial statements filed annually at the Italian Chambers
of Commerce on a compulsory basis. Cerved provides information nearly on the entire universe of
Italian joint stock and public and private limited liability companies. The information provided
includes company profiles and summary financial statements (balance sheets, income statements
and financial ratios). We restrict the analysis to manufacturing firms or those offering private non-
financial services, whose turnover before the occurrence of financial distress was between 1 and 50
million euros. Then we select those firms whose cash flows (earnings before provisions, interest
and taxes) fell below their net financing costs at least once within the 2002-04 period (i.e. their
coverage ratio became less than one).
6
In this situation, the firm is not able to meet its contractual
repayment obligations. This is what we will call financial distress from here on. We consider
only firms presenting this condition for the first time, in order to select those at the onset of the
crises. We identify the distress event as the year in which this situation occurs for the fist time. We
close the sample in 2004, in order to have three years after the distress event (up to 2007) to assess
the result of the workout.
7
We do not include listed companies in the sample, nor companies with
access to financial markets; for firms in the sample, bank debt is mostly the only source of external
finance, if we exclude trade credit.

6
We also tried using different thresholds without significantly altering the results.
7
That is, if financial distress occurs in 2004, the three-year window after the distress event is 2005-07.

This first set of information has been matched with Central Credit Register data (Centrale
dei rischi; CR hereafter). The CR reports data on credit lines granted by every bank to each of the
selected firms.
8
The CR also provides the number of banking relations and the quality of loans. For
any distressed firm, data are taken at time t (the year of the distress event) and at time t+3 (in order
to evaluate the evolution of bank-firm relations and the result of the workout). We drop from the
sample those firms that were classified as bad borrowers at the moment of the distress event by
most of the lending banks, while we keep those that became non-performing in the following
years.
9

Then, we add additional information on lending banks, by using a Bank of Italy special
survey conducted in 2007, which collected unique information on organization and technology
used in lending activities by more than 300 Italian banks involved in the survey (Albareto et al.,
2008). The survey focuses on two aspects. First, it explores the role of local branches in small
business lending, reporting the limits assigned to local loan officers and CEOs in extending credit
to small businesses autonomously, the average turnover of local managers and the use of economic
incentives to reward their activity. Second, it describes the diffusion of statistical models to manage
credit risk: the year of introduction; their effective use in lending, pricing and monitoring the loan;
and the importance the bank gives to other sources of information, such as personal knowledge. We
also include other information from the Banking Supervision Reports.
Once all the sources of information have been matched, the sample numbers 9,150 firms
facing financial distress and more than 300 banks. Our unit of observation is defined by a unique
bank-firm pair. Since we observe about 3.5 banks for each borrowing firm, we have about 30,000
bank-firm observations. The econometric analysis has been conducted at this loan (bank-firm
relationship) level.
Between 2002 and 2004, the average share of firms in financial distress in the total
population was 5 per cent (Table 1). The share was higher for the service sector than for
manufacturing; inside manufacturing, it was higher in traditional industries, such as textiles,
clothing and footwear.
At time t+3, about one fifth of the distressed firms has exited the market, while almost 80
per cent still survive, in most of the cases no longer in financial distress (53 per cent; Table 2).
As far as the restructuring process is concerned, slightly more than one third of the firms is
allowed to restructure outstanding debt by at least one bank (Table 3). Among the firms that have
restructured their debt, exits are less than 10 per cent, against 30 per cent for the others. By and
large, only some of the lending banks agree to restructure, while others do not. Thus, when
computed at bank-firm level, the incidence of debt renegotiation goes down to 16 per cent. More
specifically, in 23.5 per cent of cases only one bank restructures; two or more banks are found to
restructure in the remaining cases (13 per cent; Table 3). Apart from one-to-one banking
relationships, it never happens that all the lending banks decide to restructure simultaneously: in a
nutshell, banks show heterogeneous lending behaviour.
Table 4 gives some descriptive statistics on the evolution of the main bank aggregates over
the three-year window following the financial distress. From t to t+3, the credit used (considering
only firms that have survived) increased at a yearly rate of about 4 per cent, while a higher growth
rate was recorded for the economy overall. The increase in short-term lending was lower than in

8
The CR records the exposures of banks for which the amount of credit granted or drawn or the guarantee provided exceeds the
threshold of 75,000 euros.
9
Since non-performing loans are assimilated to defaults in Italian supervisory regulation, these cases are similar to cross-defaults.

longer maturity exposure, whose incidence thus increased, from 45 to 48 per cent. The average
number of lending banks and the credit concentration index remained substantially stable over the
sample period. The incidence of bad debts on the sum of performing and non-performing loans
(considering both exits and survivors) consistently increased, from 0 to 23 per cent.

4. The probability of bank debt restructuring
4.1 The model
When a borrower faces financial distress, the bank has to decide whether to take part in
workout activities (occurrence of workout). Workout activities consist of rescheduling loan
maturity and/or extending new credit; it may also encompass the arrangement of a lender syndicate,
the provision of consultancy services, debt equity swaps and other kinds of involvement of the
bank in restructuring (Elsas and Krahnen, 2002). Some time after this decision, we observe if the
firm still survives (success of workout) or has been liquidated. The bulk of the existing empirical
literature is concerned with the success of workout.
10
In our analysis we depart from the prevailing
literature as we observe every single bankfirm relationship (loan-level data) and report bank
features such as organization and lending technologies, while previous studies use mainly
information on firms and on the type of relationship with the banking system.
We study how a single bank contributes to the workout, adopting credit decisions such as
rescheduling a maturity or agreeing a new credit extension. Following the taxonomy introduced by
Brunner and Krahnen (2008), bank debt is restructured if one of the following two conditions are
present in the three years after the crisis:
a) long term debt has increased while total loans have stayed the same or decreased only
slightly (no more than 10 per cent);
b) total loans have increased.
Therefore, our dependent variable (RESTR) is a dummy that is equal to 1 if the bank
provides a maturity extension (option a) or an increase in total credit during the three years after
the crisis (option b), 0 otherwise. With these interventions, the bank is loosening the borrowers
financing constraints, most likely because it is willing to continue the relationship with the firm.
11

Consistently with our purpose, we also consider as non-restructured (RESTR=0) those firms that
run into voluntary or forced liquidation, or are classified as bad borrowers by their lenders.
12
As a
robustness check, in Section 5 we explore different definitions of the dependent variables. First, we
consider the two previous conditions separately. We control for the cases in which restructuring is
approved before time t+3 (i.e. t+1 or t+2), but this decision is reversed afterwards.
13
Lastly, we
define the restructuring operation in terms of the amount of new credit extended.
We estimate the probability that a bank restructures loans to a firm in distress as a function
of balance-sheet data, relationship lending information and data on the organization and technology
used by lending banks.

10
At least in the short run. As usual in the related empirical literature, we follow the firms history for 3 years after the occurrence of
financial distress.
11
We do not consider other types of restructuring, such as debt equity-swaps, lender syndicates or others, which are very uncommon
for SMEs in distress.
12
Due to inefficiencies in the Italian judicial system, the length of bankruptcy procedures often exceeds that of the analysed period (3
years after the occurrence of financial distress). Therefore the bad loan classification is usually a preliminary phase leading to
liquidation.
13
The previous definition classified these cases as non-restructured firms.

As a first step, we estimate a simple probit model of the type:
(1)
) 0 ( 1 > + = c | x
y
,
where 1(.) is equal to 1 if the relation inside the brackets is true.
The log likelihood is:
{ } { } ) ( 1 ln ) ( ln
0 1
| | x x LL
y y
u + u =

= =
,
where denotes the standard cumulative normal distribution.
Still, this method does not address the primary concern which may bias our results: firms
choose the banks from which they borrow and banks decide whether or not to lend (Berger et al.,
2005a; Guiso and Minetti, 2007). As a consequence, while we have a representative sample of
firms in distress, we observe only those banks which actually lent to those firms, i.e. only the
selected banks.
This selection process is clearly endogenous: it is related to firm characteristics, such as size,
risk and location, and, more important to our purposes, to the credit policy followed by the bank.
This process is of prime importance in the case of a firm which is small and risky. In fact, firms
may decide to have relations with those banks they believe more willing to restructure outstanding
debt in case of distress. On the contrary, they may just want to conceal the situation, moving to
banks they think have more difficulties in detecting the real prospects of investment projects.
Furthermore, according to Bris and Welch (2005), the entrepreneur will try to enlarge the number
of creditors to gain a better bargaining position in case of distress. All these elements are likely to
influence, ex ante, the characteristics of lending banks in terms of size, technology, etc. and, ex
post, the restructuring decision.
To address this issue, we estimate a probit model with sample selection (Heckman, 1979) by
means of maximum-likelihood estimation method, of the type:
(2) Sample selection:
) 0 ( 1
1 1 1
1
> + = c | x
y

Restructuring equation:
) 0 ( 1
2 2 2
2
> + = c | x
y
observed if Y
1
=1

1
,
2

N(0, 1) , corr(
1
,
2
)=
where 1(.) is equal to 1 if the relation inside the brackets is true. The main variable is
observed only if the first equation is verified. The log likelihood is then:
{ } { } { } ) ( 1 ln ) , , ( ln ) , , ( ln
1 1
0 1
2 2 1 1 2
0 2 ; 1 1
2 2 1 1 2
1 2 ; 1 1
| | | | | x x x x x LL
y y y y y
u + u + u =

= = = = =

In the sample selection equation, we consider as potential lending banks all those
intermediaries that have at least one branch in the province in which the firm is located. By

controlling for the selection of the lending banks, our aim is to study the restructuring decision
which all the banks operating in the local credit market could have taken with respect to the firm
facing distress.
The number of observations increases to more than 340,000 in the selection equation.
Obviously, at this stage we do not introduce any information regarding bank-firm relations (which
only exist for those banks actually selected). Our identification variable is the market share of the
bank, computed on the number of branches. This variable has a univariate correlation of 0.33 with
the selection variable (Table 6) against a correlation of 0.04 with the restructuring variable. The
choice of this variable is based on the idea that if the bank has a widespread presence in the area,
most likely the firm will select it. If the bank is not selected, this might be because the firm did
apply for a loan but the bank rejected it, or because the firm did not apply for a loan in the first
place, expecting the bank to be tough i.e. not willing to help in case of distress. We tried to
capture the latter hypothesis by means of the second identification variable, which is the average
collateral over total loans the bank requires in the same area from its business clients. Even if the
correlation of this variable with the selection variable is statistically weak, we think it has important
economic significance.
* * *
To control for the anticipated going-concern value of the firm, we introduced various
balance-sheet indices, such as the ratio of financial liabilities on total liabilities (leverage), return
on assets (ROA) and net interest payments over total assets (financing costs over assets); we also
introduce the firm size (log of total assets), as well as sector (two-digit Ateco code) and regional
dummies. The balance-sheet variables are calculated as an average of the three years before the
crisis.
To capture the concentration of lending across banks, we consider two variables: the first is a
dummy (Single), equal to 1 if the firm has just one lending relationship; the second is the share of
the bank over the total outstanding debt. We follow previous studies (Detragiache, Garella and
Guiso, 2000; Guiso and Minetti, 2004 and 2007), in which multiple lending is a two-step decision:
first, the firm decides whether to borrow from one or more banks; then it decides how concentrated
its bank debt should be. The share of collateral over total credit extended by each bank is then
introduced to control for the guarantee posted by the firm to the lending bank.
Our key variables are those capturing the heterogeneity among banks. All these regressors
come from the special survey carried out by the Bank of Italy.
We first consider the lending technology used by the bank. We follow the conceptual
framework described by Berger and Udell (2006). We introduce a dummy variable indicating
whether a bank has adopted credit scoring techniques in small business lending (Scoring).
According to Berger and Udell (2006), credit scoring is a transaction technology based primarily
on hard information. Moreover, in the survey, banks are asked to evaluate how important they
regard soft information in the lending process; therefore, we are able to add a dummy if the bank
has stated that qualitative information is an important element in assessing the credit worthiness of
SMEs (Soft information). We also use two dummy variables to capture the differences between
banks that use credit scoring only as a decision-making rule and those also using the scores in the
monitoring process.
We deem the degree of decentralization of the lending decision to be very important on bank
behaviour (Stein, 2002). Delegation improves incentives to gather soft information and encourages
initiative on the part of local managers. Using data from the survey, we computed a delegation
index as the amount of credit the local branch manager of the bank can extend autonomously to
small businesses, with respect to the amount the CEO himself can grant without referring to the

banks board (Delegation). To verify the soundness of our results, we also use the index in absolute
terms, as the amount (in logs) of credit that a local branch manager can autonomously extend to a
small firm.
Decentralized banks may partially offset the different objectives of local and central
managers by designing appropriate incentives. To ensure local managers apply the appropriate
effort in obtaining and processing information, their remuneration is usually linked to the economic
performance of the branch they manage: we introduce a dummy for the existence of economic
incentives to remunerate local managers (Incentives). On the other hand, to avoid excessive
involvement in local economies, branch managers are commonly moved on a regular basis. In the
estimates, we also add the average length of tenure (in months) of these managers (Turnover).
Finally, the size of the bank and its distance from the borrower may be important factors in
influencing the decision whether to support a firm in distress. Therefore, we introduce dummies for
the type and dimension of the bank and we measured the physical distance between bank
headquarters and the location of the firm (Distance). A growing number of contributions address
this issue.
14
Distance is related to the type of relationship with the firm. Larger banks tend to use
transactional technology in lending to firms located far away (Berger, Frame and Miller, 2005b).
Monitoring costs tend to increase with borrower-lender distance. Furthermore, intermediaries
lending at a greater distance are less involved in local economies than smaller banks, which are part
of the community and have fewer ways of diversifying their assets. Mian (2006) finds that
geographical and cultural distance is an important attribute in explaining the lending, recovery and
renegotiation differences between domestic and foreign banks.
Table 5 reports some descriptive statistics about the main variables used in the estimates,
including balance-sheet indices and bank organization variables. At this descriptive stage, we just
note that in the case of a bank-firm restructured relationship (observations for which RESTR=1),
the average value is higher for Roa and Size, and lower for Scoring and Distance, that is
restructured firms are larger, more profitable and the lending banks are geographically closer and
make less use of credit scoring. Finally, Table 6 shows the cross-correlations between variables.

4.2 The results
Table 7 reports the results of the econometric analysis. The first two columns report the
restructuring equation and the selection equation from our baseline specification; we compare these
results with a restructuring equation estimated by means of a simple probit (Col. 3). Then, we
introduce some extensions for credit scoring techniques, by controlling if they are used in the
lending decision and monitoring procedures (Col. 4). Additional specifications are reported in the
following columns of Table 7.
As far as the Heckman procedure is concerned, in the selection equation (Col. 1) the two
identification variables are strongly significant and they show the expected signs: banks with a
higher market share in the province are more likely to be picked up; the same is true when they ask
for a lower level of collateral. The selection process is an important issue to address, as suggested
by the significance of the estimated . Furthermore, most of the effects in the restructuring equation
(Col. 2) are stronger when estimated by the Heckman procedure than when a simple probit is used
(Col. 3), confirming that we need to correct for sample selection.

14
On the effects of the physical and functional distance of a lending bank from a borrowing firm, see Alessandrini, Presbitero and
Zazzaro (2009) and Degryse and Ongena (2004) for a review.

In the restructuring process (Cols. 2-8), our results related to balance-sheet variables seem to
point to consistent outcomes. The more favourable the financial position of the firm (lower interest
payments and leverage), the higher the likelihood of restructuring. This is also true for firm
profitability (measured by ROA), which increases the probability of restructuring outstanding debt.
On the other hand, debt restructuring is more frequent for smaller firms.
Moving to variables related to bank concentration, the econometric analysis provides new
insights on the recent theoretical and empirical literature on credit concentration. Our results show
that a non-linearity does exist, which in turn may support the theory put forward in Detragiache,
Garella and Guiso (2000) and Guiso and Minetti (2004 and 2007), in which multiple lending is
modelled as a two-step decision: first, the firm decides whether to borrow from one or more banks;
then it decides how concentrated its bank debt should be. In our results, single banking helps to
restructure outstanding debt: the estimated coefficient for Single is positive and significant. On the
other hand, once multiple banking is chosen, it is rational for a risky firm to spread its debt: the
probability of obtaining debt restructuring in the case of distress decreases with the size of the
lending share of each bank.
15
An alternative specification (not reported) in which we use the
Herfindahl concentration index points in the same direction.
We introduce the share of the bank over the total outstanding debt separately for the leading
bank, i.e. the bank which has extended the higher share (Share of the main bank), and the other
intermediaries (Col. 6). The effect of lending shares is lower for the main bank than for the other
intermediaries the test that the two variables have the same coefficient is rejected at 1 per cent
but it is still detectable in the data.
16
Thus, banks tend to limit further increases in their exposure
towards risky business when it is already high. As described in more detail in the next section, this
interpretation is further improved by comparing the results in relation to the type of restructuring: it
becomes more and more difficult to get further credit as the share of the bank increases, while it
has no effect on the likelihood of obtaining a maturity extension for outstanding debt.
This evidence can add a further rationale to the findings of Ongena, Tmer-Alkan and von
Westernhagen (2007), who show that firms with a higher probability of default are expected to
have a larger number of banks and a significantly lower degree of credit concentration. It is also
shown that the usual imitative behaviour of small banks with respect to the main bank applies to a
lower extent in the case of a firm in distress, which borrows more from other banks rather than
from the main bank. Our result is also consistent with the theory stated in Detragiache, Garella and
Guiso (2000), since the risk of fund withdrawal because of a bank liquidity shock is more difficult
to balance for a firm in distress. Furthermore, according to Bris and Welch (2005), a firm that opts
for multiple creditors ex ante assumes a better bargaining position in the case of financial distress
ex post (Guiso and Minetti, 2004 and 2007; Ongena, Tmer-Alkan and von Westernhagen, 2007).
The coefficient estimated for Collateral is negative and significant, suggesting that banks
restructure less when bank credit is better secured. This finding appears to be consistent with some
explanations proposed in the theoretical literature. According to the lazy bank hypothesis (Manove,
Padilla and Pagano, 2001), in the case of distress, well secured banks may free-ride on screening
activities and other initiatives aiming at preserving a firms operations. In addition, it should be
more difficult to reconcile divergent borrowers preferences when borrowers are endowed with
different seniorities (Bolton and Scharfsten, 1996). This finding is also consistent with the asset-
based lending theory, where collateral is the primary source of repayment (Berger and Udell,
2006).

15
Since the variables Single and Bank share are correlated (Table 6), we also introduced them separately. The above described results
are confirmed.
16
Nearly the same result is obtained in a specification (not reported) which excludes the dummy Single.

With regard to the variables capturing bank heterogeneity, the dummies for the type of bank
maintain some explanatory power. It is worth noting that small cooperative banks tend to
restructure more than large and medium banks (our benchmark), while other small banks do not.
Therefore, we believe that cooperative banks do follow different strategies when dealing with small
firms in distress, even after controlling for their lending technology and organization. Most likely,
this result is related to their nature as cooperatives rather than to their size: the other small banks in
the sample do not show any detectable difference in restructuring propensity.
While controlling for bank size, the type of organization and lending technology still explain
the restructuring decision. The estimated sign of the variable Scoring is negative and very
significant, i.e. the introduction of credit scoring techniques reduces the probability that a bank will
restructure credit to firms in distress (Table 7, Col. 1).
Since this issue is of great interest, we have explored in detail the role of credit scoring
techniques, exploiting more comprehensive information on how credit scoring is actually
implemented by banks.
Banks were asked to assess whether credit scoring techniques were important both in the
loan decision and in the monitoring process. In the restructuring equation, we introduced two
dummies, to control for the use of scores in credit-extension decisions (Score_Decision) and in the
monitoring process (Score_Monitoring), respectively.
17
The estimates in Column 4 report our
preferred specification. It turns out that the estimated coefficient is negative for Score_Decision; on
the contrary, it assumes a positive sign for Score_Monitoring. Note that banks use credit scoring
techniques more frequently in the decision-making process to extend credit, rather than in the
monitoring process, thus explaining the overall negative effect of Scoring. In our view, the opposite
signs estimated for the two dummies suggest that the use of credit scoring is not by itself an
obstacle to restructuring, while the way it is implemented matters. In the monitoring stage, the use
of statistical techniques signals a continuous process of gathering and analysing information over
the entire life cycle of the firm-bank relationship, rather than an episodic application of automatic
techniques. Most likely, the use of these methods reduces monitoring costs, thus helping debt
renegotiation for firms in a very crucial phase of their life.
We further investigate the use of credit scoring by defining this variable differently. Since
we know the year of adoption of these techniques by banks, we can exploit this information. The
credit scoring dummy is now equal to 1 if credit scoring was adopted before the restructuring
decision but after the distress event; 0 otherwise. In doing so, we are now sure the variable is
completely exogenous with respect to the selection process. Our results are qualitatively and
quantitatively confirmed (Col. 7).
The decentralization-degree variable (Delegation) positively affects the probability of
restructuring. This result is also confirmed when we introduce the variable in absolute terms
(column 5 of table 7). Being closer to SMEs and having more power in the decision-making
process, local managers may use personal relations to collect soft information. This helps the
restructuring process. Of course, we cannot rule out that local managers have the incentive to hide a
deteriorating borrowers condition or that they are looking for private benefits (Aghion and Tirole,
1997; Berger and Udell, 2002). We will come back to this point later on. It is worth mentioning
that when Scoring and Delegation are interacted, the resulting interaction effect assumes a positive
value, indicating that the use of credit scoring increases the probability of restructuring if banks
also adopt more decentralized structures (Col. 8). The mean of the interaction effect is equal to

17
We did not change the selection equation in the first step, since we believe that is difficult for the borrower to know the way these
techniques are implemented by the lending bank. However, the results are very similar considering these variables in the selection
equation.

0.176, with a standard error of 0.0649; for the sake of simplicity, it is calculated with a simple
probit estimation, using the inteff procedure proposed by Norton, Wang and Ai (2004). This
result is consistent with the hypothesis that even when accomplished by larger banks,
decentralization may counterbalance the effects of a rigid implementation of automatic techniques.
The Turnover variable is never significant. This may attenuate the possible concern about the
estimated (positive) coefficient of the Delegation variable. In fact, more delegation may entail
better incentives and information for local loan officers, thus allowing a proper selection of better
firms; on the other hand, it may introduce the risk that credit is allocated to unworthy borrowers,
because of agency problems between the loan officer and his bank. In any case, this risk should
increase with the length of tenure of the local officer. To disentangle the hypothesis of more
accurate information from that of capture, we therefore interact Turnover and Delegation. Since the
interaction variable turns out to be always insignificant, this gives some further support to the
improved information explanation.
18

The bank-firm distance is a measure for both the type of lending relationship and
information about the overall economic conditions of the area where the firm is located. However,
this variable is obviously linked both to bank size and the type of bank organization.
19
Therefore
we add it only at the end, in order to analyse whether it affects the results or not. We find that
lending at a greater distance significantly decreases the probability of restructuring. In the
meantime, our main results remain stable. It is also worth mentioning that the coefficient for
Cooperative banks decreases, suggesting that part of its effect depends on a very close proximity to
distressed firms; at the same time, the coefficient for Cooperative banks remains highly significant
while we do not detect any effect on restructuring likelihood as far as other small banks are
concerned, supporting the hypothesis that the effect comes from the cooperative form rather than
size.

5. Robustness check
5.1 Sample splits and different definitions of the dependent variable
In this section, starting from our preferred specification (Spec. 4 of Table 7) we perform
some robustness checks. The results are reported in Table 8.
In the first place, we split the sample by considering separately: small firms with total sales
lower than 10 million euros (Tab. 8, Col. 1); and mediumsized businesses with sales of 10 to 50
million euros (Col. 2). Our results remain generally unchanged. In line with the theory, soft
information and economic incentives are (only slightly) significant for small firms, while they are
insignificant for medium-sized businesses; this is true both in the selection and the restructuring
equations. In the latter, the distance seems to play a role only for smaller firms, while it loses
significance as firm size increases. As far as credit scoring in lending and monitoring is concerned,
the two variables have the same sign but the marginal effect for medium-sized firms is more than
twice that of smaller firms.
Then we explore different definitions of the dependent variables (in the table we report only
the second stage, since the selection equation remains unchanged).
First, we define the restructuring in terms of credit granted instead of credit actually used
(Col. 3). Then we control for the cases in which the concessions rescheduling or new finance

18
At the same time, we think that this issue is sufficiently important to merit further analysis, in future research work.
19
For the sake of simplicity, we do not report the various steps in which we progressively add the variables.

are approved before time t+3 (that is t+1 or t+2), but this decision is reversed afterwards in the
three-year window (Col. 4).
20
Next, we split the variable, considering separately the cases in which
only the maturity is rescheduled (point a in the previous definition; Col. 5), from cases in which the
firm obtains new finance (point b in previous definition; Col. 6). In the next paragraph we will also
define the restructuring in terms of the quantity of new credit extended.
However, through these various definitions of the dependent variable, the estimated results
for the regressors of interest remain mostly unchanged. As further evidence, we signal that banks
are more prone to reschedule maturity rather than to extend new credit towards highly leveraged
firms. As a matter of fact, the marginal effects of balance-sheet variables are much higher when
banks decide to allow new credit lines rather than to extend loan maturity. This evidence is
consistent with the idea that banks tend to postpone the liquidation decision by rescheduling their
debt claims if there is increasing uncertainty about a firm's prospects (Kahl, 2002).
Disentangling the type of restructuring also improves our understanding of the credit
concentration effect: as a banks lending share increases, it tends to avoid the extension of new
loans, while its share has no effect on the likelihood of maturity rescheduling.
Finally, we focus our attention on the eventuality that the restructuring process is an
inefficient economic outcome. To ensure economic efficiency, only the debt of firms with positive
investment opportunities should be renegotiated, while non-viable firms are better liquidated.
In the previous estimates, we controlled for the expected net present value of distressed firms
by introducing the average balance-sheet indexes computed before the crisis. This is a backward
looking strategy, since we do not have the proper variable to split viable from non-viable firms
according to their future prospects; nevertheless we can presume that the most likely candidates to
exit the market in the near future are those firms still in distress at t+3 (about one-fourth of the
sample; Table 2). Thus, we conduct an econometric analysis considering only those firms which
are no longer in financial distress at t+3. When we replicate the econometric exercise with
reference to these firms (Col. 7 of Table 6), our previous results are confirmed. This evidence
provides some support to the hypothesis of more efficient restructuring of distressed firms, but the
issue would require further analysis and is left for further research.

5.2 More on the endogeneity of organizational variables
Indeed, most of the organizational variables are likely to be endogenous. If incentive
schemes, decentralization and use of technology are properly designed and implemented, one
would expect them to lead to the selection of better firms. Thus, the bank should be keener on
renegotiating the debt of these firms since they are good clients that have been hit by bad luck and
the bank wishes to maintain the relationship. In other words, proper organizational schemes could
promote an ex ante selection (at the very first credit decision) of better firms. If this is the case,
firms will obtain debt renegotiation irrespective of the actual importance of the organization
variables.
Note that we have already tackled this issue, by controlling for the matching of the bank-firm
relationship in the first stage, when our key variables are used in the selection equation.
Furthermore, we introduced the financing position and profitability of the firm (leverage, ROA,
financing costs on assets) before the occurrence of the distress. These controls allow us to

20
The previous definition classified these cases as not restructured firms.

distinguish between ex ante better firms (those suffering distress only after a period of good
profitability) and riskier firms (for which distress occurs in unfavourable economic conditions).
Notwithstanding these controls, a sceptic could still argue that residual borrower-level
variation is spuriously correlated with firms quality. To further address this concern, following
Mian (2006), we add firm-level fixed effects to the estimated equation, thus controlling for all
possible borrower-specific variations. For the sake of simplicity, we use a simple linear probability
model, but we obtain similar results when a conditional logit model is used. Obviously we drop all
firms with just one lending relation (Single=1). This methodology emphasizes the heterogeneous
behaviour of different lending banks towards the same firm in distress.
The estimated coefficients (see Table 9) regarding our main organizational variables
(distance, scoring, soft information, incentive, delegation) remain mainly unaffected, suggesting
that our results are not driven by unobserved borrower characteristics since they are robust to the
inclusion of borrower fixed effects. Thus, we are more confident of the relevance of lending
technologies and banks organizational schemes for the financial-restructuring process.
Lastly, we introduced fixed effect by banks. In this way we can test whether there is some
time-discontinuity in the behaviour of the same bank towards different firms before and after
introducing credit scoring techniques. We know the year of adoption of these techniques and we
can exploit this information.
21
Thus, we redefined the credit scoring dummy: it is now equal to 1 if
it was adopted before the restructuring decision; 0 if it was adopted afterwards. We can see how the
implementation of credit scoring changes the probability after the introduction. Again the result
points to a harsher regime following the adoption of credit scoring, in line with our previous
results.

5.3 The intensity of bank debt restructuring
So far, the results discussed in the previous section are robust to different specifications of
the dependent variable and to different estimation methods.
We further check the robustness of these estimates using an alternative definition of the
restructuring process. First, we try to capture the intensity of bank involvement towards the
distressed firm. More in detail, we assess the extent by which a bank makes further concessions, by
using the log of new credit extended in the three years following the crises. We use this information
as a new definition of our dependent variable. Obviously, we should again address a selection
problem. As a matter of fact, we have to take into account that: i) the observed bank-firm
relationship is the result of a self selection process; ii) only the subgroup of restructured firms had
the chance to obtain new finance. It is worthwhile recalling here that one condition we used to
define debt restructuring is actually an increase in credit extended in the case of distress (option b
in our definition). These aspects call for an appropriate technique to solve this double-selection
problem. To this purpose, we adopt a standard bivariate decision problem as proposed by Meng
and Schmidt (1985) and Tunali (1986).
22
This method allows us to control for the fact that a firm
may agree on a restructuring plan and obtain new credit only after a two-stage participation
decision, i.e. the firms decision to participate in the credit market and the banks decision to
extend credit to that firm. Therefore in the first stage we estimate these two equations by a bivariate
probit with partial observability:
(3) y
1ij
= x
1ij
b
1
+ e
1ij
(firm i applies for credit to bank j if y
1ij
> 0)

21
From the survey, we know the year of adoption only for credit scoring, not for other variables.
22
For an application to the labour market analysis, see Mohanty (2005).

(4) y
2ij
= x
2ij
b
2
+ e
2ij
(bank j decides to restructure the debt of firm i if y
2ij
> 0)
We include in x
1,ij
and x
2,ij
a set of variables aimed at capturing all the economic and financial
determinants of the credit decision. We also include the two identification variables already
discussed in the previous section: the bank market share in the province, measured by the number
of branches, and the average collateral requested by the bank from firms in the same province.
In the second step of the analysis, we estimate a restructuring equation for the borrowing
firms only, where the dependent variable, Y
i
m
, is the new credit extended; we use OLS and add the
two corrections for the self-selection bias:
(5)
u c c b x
y
i i i i
m
i
+ + + =
2 13 1 13 3 3

Where:
Y
i
m
=log loans
t+3
- log loans
t
,
i i i i
c c u
2 23 1 13
c =
,
( )
( )
, ,
1
2 2 1 1
2
1 1 2 2
1 1
1
b x b x F
b x b x
b x
i i
i i
i
i
|
|
.
|
\
|
u
=

( )
( )
, ,
1
2 2 1 1
2
2 2 1 1
2 2
2
b x b x F
b x b x
b x
i i
i i
i
i
|
|
.
|
\
|
u
=

| and represent, respectively, the density and the cumulative function of a standard normal
distribution, while is the correlation of the error terms in the bivariate probit.
Table 10 reports the estimated results using the method discussed above (Col. 2), compared
with those obtained by an estimation without sample-selection correction (Col. 1). By applying this
method, our results appear to be unaffected or, in some cases, even strengthened.
As an alternative estimation technique, we considered a Tobit model,
23
whose dependent
variable is zero in the case where no restructuring is recorded (if RESTR=0) and it is equal to the
growth of credit in the three years after the crises (log loans
t+3
- log loans
t
) in the case of
restructuring (if RESTR=1). The (log) likelihood function is:
(
|
.
|
\
|
+ |
.
|
\
|

= u

= >
o
|
o
|
|
o
i i i
x x y
LL
i i
'
ln
1
ln 1
0 y 0 y
,
The results, reported in the last column of Table 10, confirm our previous evidence.

23
The hypothesis underlying the Tobit model was that the explanatory variables would influence in the same way both the probability
of restructuring and the quantity of credit extended. This hypothesis may be too restrictive, in that the explanatory variables may
affect the two phenomena in different ways or may be relevant in only one of the decisions.

6. Concluding remarks
In this paper we have studied the bank-debt restructuring process of SMEs in financial
distress, taking bank heterogeneity into account thanks to a survey conducted by the Bank of Italy
on bank organization and lending technology. Our results show that banks follow heterogeneous
strategies when facing financially distressed firms, in that they may behave very differently
according to their organization and lending technologies. While bank size seems far less important,
in line with the literature on lending technologies we show that more impersonal bank-firm
relations, lending at a great distance and relying on credit scoring techniques (transactional lending)
may decrease the likelihood of bank debt restructuring, compared with cases where there is
relationship banking.
However, we show that technological and organizational arrangements may balance the
effects of transactional lending. Decentralization of decision-making and a more systematic use of
statistical techniques, especially in the monitoring process, can facilitate financial restructuring. In
particular, the use of credit scoring is associated with a higher probability of debt restructuring
when a bank also adopts a more decentralized structure and also uses these methods to monitor the
lending relationship.
From a policy perspective, better understanding of the mechanisms behind banks decisions
towards firms in distress is economically significant. The restructuring decision is a crucial step in
preserving the continuation of a firm in distress but still with profitable investment opportunities.
While consolidation and the increasing use of credit scoring may raise concerns for SMEs in
distress, our results suggest that enhanced implementation of statistical techniques and effective
decentralization of decision-making may offset these effects. At the same time, the growing
emphasis of regulators on bank organization appears to be consistent with our results. Our work
also suggests that the common assumption of uniform bank behaviour in the case of distress should
be reconsidered; at the same time, that bank size per se cannot fully explain this heterogeneity
among banks, as individual choices relating to strategies, organizational forms, and lending
technologies may have important consequences on banks behaviour.


APPENDIX
Definition of variables
Dependent variables (Source: Central Credit Register):
Restr: binary variable equal to 1 if at least one of these two conditions has occurred: i) long-
term debt has increased while total loans stayed the same or decreased only slightly (no more than
10 per cent); ii) total loans have increased. Firms in voluntary or forced liquidation, and those
classified as bad borrowers are considered as non-restructured.
Firm variables (Source: Cerved):
Financing cost on assets: ratio of net interest expenses to total assets;
Leverage: Ratio of financial debts to the sum of financial debts and net equity;
Size: (Log of ) total assets;
ROA: Return on assets;
Coverage: ratio of earnings before interests, taxes and depreciation to net interest expenses.
[This variable has been used to select firms facing financial distress].
Bank-firm relationship variables (Source: Central Credit Register):
Distance: distance (Log of KMs) between the banks headquarters and the location of the
borrowing firm;
Bank share: share of the bank over total outstanding debt;
Share of the main bank: share of the main bank over total outstanding debt, when the bank is
the principal one (with the larger position); 0 otherwise.
Share of other banks: share of the bank over total outstanding debt, when the bank is not the
principal one; 0 otherwise;
Collateral: ratio of collateralized loans to total loans;
Single: binary variable equal to 1 if there is just one lending bank;
Bank organization variables (Source: Bank of Italy, Survey on bank organization);
Scoring: binary variable equal to 1 if a bank uses scores in 2003 as a rule in the decision-
making process, instead of adopting a more flexible approach;
Score_decision: binary variable equal to 1 if a bank uses credit scoring techniques in 2003
and these are considered as an important instrument when deciding new credit extension;
Score_monitoring: binary variable equal to 1 if a bank uses credit scoring techniques in 2003
and these are considered as an important instrument in the monitoring process;

Delegation: index of delegation in favour of local managers, computed as the amount of
loans to SMEs that they can extend autonomously with respect to the amount the CEO of the bank
can extend;
Soft information: binary variable equal to 1 if a bank has stated that qualitative information is
important in credit decisions;
Incentive: binary variable equal to 1 if local managers are compensated by means of a
performance-related wage;
Turnover: average length of tenure of the local loan officer (in months);
Time: time (in years) since the adoption of credit-scoring techniques in 2003.
Other bank variables (Source: Bank of Italy, Banking Supervision Reports:
Small banks in groups: binary variable equal to 1 if a bank is both small in size and a
member of a banking group;
Small banks: binary variable equal to 1 if a bank is small and it is not member of a banking
group;
Cooperative banks: binary variable equal to 1 if a bank is both small and a cooperative
(Banca di Credito Cooperativo, according to the Italian banking law).


Table 1
Firms in distress
(% of firms)

Firms in distress for the first time in
the year considered
(% of firms)
2002 2003 2004 Average
Total 3.7 6.4 5.0 5.0
By sector
Manufacturing 3.2 5.8 4.4 4.5
Services 4.2 7.1 5.5 5.6
By area
North 3.5 6.4 4.7 4.8
Centre 3.7 6.6 5.4 5.2
South 4.8 6.6 5.5 5.6
By size
1-10 million euros 3.5 6.7 5.1 5.1
10-50 million euros 5.0 5.1 4.2 4.8
Source: Cerved. Simple frequencies on the number of firms not in distress at the beginning of the period.

Table 2
Situation at t+3 of firms in distress in t
(% of firms)

By sector By area
By size
(million euros)
Situation at t+3
(% of firms)
Manufac-
turing
Services North Centre South 1-10 10-50
Total
Exits
19.8 22.4 20.5 20.9 24.4 21.7 18.3 21.1
Survivors
80.2 77.6 79.5 79.1 75.6 78.3 81.7 78.9
Still in distress
22.3 28.5 25.2 26.5 25.3 24.2 31.7 25.5
Out of distress
57.9 49.1 54.2 52.6 50.3 54.1 50.0 53.4
Total
100 100 100 100 100 100 100 100

Source: Cerved, Central Credit Register. Simple frequencies on the number of firms in distress in t.


Table 3
Financial restructuring of firms in distress
(% of firms)

By sector By area
By size
(million euros)
Situation at t+3
(% of firms)
Manuf. Serv. North Centre South 1-10 10-50
Total
Exits
19.8 22.4 20.5 20.9 24.4 21.7 18.3 21.1
Survivors
80.2 77.6 79.5 79.1 75.6 78.3 81.7 78.9
At least one bank restructures
(a+b)
40.4 33.7 37.1 39.8 32.0 34.7 48.2 36.4
Only one bank
24.9 22.7 23.5 25.3 22.2 23.9 22.6 23.5
2 banks
9.3 7.3 8.6 8.2 6.5 7.4 12.5 8.0
3 banks
3.8 2.4 3.0 3.8 2.4 2.4 6.3 3.0
4 or more banks
2.5 1.4 2.02 2.4 0.9 1.0 6.7 1.9
All the banks restructures (a)
2.0 4.1 2.8 3.3 4.4 3.4 1.4 3.1
Some banks restructures, others
do not (b)
38.4 29.6 34.3 36.5 27.6 31.3 46.8 33.3
Less than half of banks
33.8 25.1 29.7 31.4 24.0 27.0 40.8 28.7
More than half
4.6 4.5 4.6 5.1 3.6 4.3 6.0 4.6
Financially restructured (1)
16.1 16.6 16.3 17.1 15.5 15.9 17.9 16.4

Source: Cerved, Central Credit Register. All the statistics refer to the simple frequencies of firms on the total of distressed firms,
except (1), which refers to the number of bank-firm relationships.


Table 4
Main credit variables

Variables
t
(year of financial distress)
t+3
Bank credit (1)
13.3
Number of banks
4.7 4.7
Herfindhal index, computed on bank credit
0.47 0.48
Long-term bank credit (%share of bank credit)
44.8 47.8
Collaterals (%share of bank credit)
17.6 18.7
Bad loans (% share of bank credit) (2)
0.0 23.0

Source: Cerved, Central Credit Register. All the statistics refer to the survivor firms, except (2), which refers to all distressed firms in
the sample.
(1) Growth rate of bank credit between t and t+3.


Table 5
Sample statistics
(bank-firm relationships)

All the sample of distressed firms
(RESTR=0;1)
Only the non
restructured
firms
(RESTR=0)
Only the
restructured
firms
(RESTR=1)
Variables
Mean
Std.
Dev.
Median
1
st
quartile
3
rd
quartile
Mean Mean
Dependent variable
Restructuring 0.16 0.37 0.00 0.00 0.00 0.00 1.00
Firm variables

Leverage 0.74 2.05 0.79 0.60 0.91 0.74 0.74
Size (log of total assets) 8.35 1.10 8.30 7.53 9.10 8.30 8.40
ROA -0.05 0.13 -0.02 -0.07 0.00 -0.05 -0.03
Financing cost on assets 0.03 0.02 0.02 0.01 0.04 0.03 0.03
Bank relationship variables

Herfindhal 0.58 0.29 0.52 0.34 0.90 0.59 0.53
Distance 1.60 1.92 0.89 0.24 2.26 1.60 1.50
Bank organization variables

Scoring 0.25 0.43 0.00 0.00 1.00 0.26 0.21
Score_decision 0.23 0.42 0.00 0.00 0.00 0.23 0.19
Score_monitoring 0.16 0.37 0.00 0.00 0.00 0.16 0.16
Soft Information 0.54 0.50 1.00 0.00 1.00 0.54 0.52
Delegation 0.07 0.09 0.03 0.01 0.09 0.07 0.08
Incentive 0.84 0.37 1.00 1.00 1.00 0.83 0.87
Turnover 31.7 11.5 30.0 24.0 36.0 31.8 31.0

Source: Cerved, Central Credit Register, Bank of Italy Survey on bank organization. All the statistics refer to the bank-firm
relationships.


Table 6
Correlations between variables

RESTR Selection
Finan.
costs
/assets
ROA Leverage Size
Collateral
(bank-
firm
level)
Single
Bank
share
RESTR 1
Selection . 1
Finan. costs /assets -0.027 0.062 1
ROA 0.042 0.001 -0.038 1
Leverage -0.014 0.058 0.143 -0.016 1
Size
(log of total assets) -0.050 0.116 -0.134 -0.035 0.062 1
Collateral
(bank-firm level) -0.087 0.289 0.023 -0.002 0.033 0.046 1
Single -0.011 0.158 -0.030 0.005 -0.023 -0.025 0.075 1
Bank share -0.051 . -0.120 -0.010 -0.128 -0.356 0.228 0.635 1
Scoring -0.019 -0.013 0.002 0.003 -0.002 -0.007 -0.009 -0.002 -0.016
Score_decision -0.019 0.015 0.002 0.003 -0.002 -0.006 -0.006 0.006 -0.012
Score_monitoring 0.018 -0.002 0.004 0.003 0.001 -0.005 -0.007 -0.004 -0.040
Delegation 0.025 -0.066 0.001 -0.001 0.001 -0.001 0.002 -0.021 -0.005
Soft information 0.000 -0.018 -0.001 0.000 0.003 0.002 0.001 -0.006 0.009
Incentive 0.010 0.048 0.011 0.007 0.001 -0.005 0.007 0.012 -0.002
Turnover 0.013 -0.090 0.004 0.004 0.009 0.000 -0.014 -0.023 0.013
Distance -0.037 -0.043 -0.004 0.008 -0.021 -0.023 -0.003 -0.011 -0.047
Share of bank
branches in the prov. 0.042 0.328 0.016 0.010 0.004 -0.008 0.084 0.119 0.169
Avg. collateral,
by bank in the prov. -0.011 -0.010 -0.005 0.005 -0.001 -0.016 0.056 0.003 0.115
continues

Scoring
Score
decision
Score
monitor
Delega-
tion
Soft
informa-
tion
Incentive Turnover Distance
Share
(bank
branches)
Scoring 1
Score_decision 0.840 1
Score_monitoring 0.733 0.608 1
Delegation 0.100 0.075 0.139 1
Soft information 0.070 0.036 -0.037 0.100 1
Incentive -0.079 -0.082 -0.139 -0.156 -0.091 1
Turnover 0.023 0.014 0.031 0.260 0.023 -0.254 1
Distance 0.084 0.132 0.087 -0.074 -0.071 0.046 -0.178 1
Share of bank
branches in the prov. 0.035 0.085 -0.035 -0.154 -0.030 0.097 -0.121 -0.126 1
Avg. collateral,
by bank in the prov. -0.037 -0.074 -0.089 0.164 -0.024 -0.112 0.194 -0.086 0.050

Source: Cerved, Central Credit Register, Bank of Italy Survey on bank organization.

Table 7
Probability of restrucuturing outstanding bank debt
(Heckman and simple probit regressions; marginal effects for the restructuring equations)

1 2 3 4 5 6 7 8
Baseline
heckprob
Probit of
baseline
Preferred
heckprob
Ddelegation
(absolute)
Share of the
main bank
Scoring
exogenous
Scoring *
delegation
Variables
Selection Restruct. Restruct. Restruct. Restruct. Restruct. Restruct. Restruct.
Financing costs/assets 7.492*** -1.311*** -0.881*** -1.341*** -1.366*** -1.348*** -1.395*** -1.299***
[19.73] [5.88] [5.99] [5.99] [6.10] [6.02] [5.83] [5.941]
ROA 0.152 0.399*** 0.361*** 0.401*** 0.397*** 0.393*** 0.400*** 0.389***
[1.58] [4.41] [5.18] [4.41] [4.38] [4.35] [4.43] [4.289]
Leverage 0.173*** -0.027* -0.019** -0.028** -0.029** -0.029** -0.026** -0.028*
[22.9] [1.86] [2.02] [1.95] [2.01] [1.97] [1.84] [1.940]
Size (log of tot. assets) 0.230*** -0.040*** -0.027*** -0.042*** -0.043*** -0.044*** -0.038*** -0.039***
[36.04] [8.82] [9.03] [9.31] [9.55] [9.56] [8.51] [9.032]
Collateral (single firm) --- -0.189*** -0.172*** -0.193*** -0.182*** -0.195*** -0.192*** -0.190***
[9.89] [9.97] [10.05] [9.33] [10.14] [10.05] [10.011]
Single (1 / 0) --- 0.070*** 0.058*** 0.071*** 0.073*** 0.063*** 0.068*** 0.070***
[3.42] [3.17] [3.48] [3.56] [3.08] [3.31] [3.417]
Bank share --- -0.190*** -0.157*** -0.190*** -0.194*** --- -0.189*** -0.188***
[10.48] [10.02] [10.49] [10.67] [10.44] [10.463]
Share of the main bank --- --- --- --- --- -0.192*** --- ---
[10.59]
Share of other banks --- --- --- --- --- -0.277*** --- ---
[7.19]
Scoring (1 / 0) -0.163*** -0.029*** -0.027*** --- --- --- -0.046***
[22.51] [3.69] [3.76] [4.459]
Score_decision (1 / 0) --- --- --- -0.087*** -0.070*** -0.087*** --- ---
[8.28] [6.43] [8.25]
Score_monitoring (1/0) --- --- --- 0.115*** 0.099*** 0.116*** --- ---
[8.48] [7.34] [8.46]
Score exogenous --- -0.027*** ---
[3.61]
Delegation (relative) 0.409*** 0.258*** 0.209*** 0.122*** --- 0.121*** 0.208*** 0.153***
[10.49] [6.12] [5.57] [2.72] [2.71] [5.09] [2.798]
Delegation (absolute) --- --- --- --- 0.018*** --- --- ---
[7.15]
Scoring*delegation --- --- 0.204***
[2.725]
Soft information (1 / 0) 0.015** 0.011 0.003 0.018*** 0.013* 0.019*** -0.002 0.011*
[2.59] [1.59] [0.56] [2.75] [1.95] [2.75] [0.34] [1.672]
Incentive (1 / 0) 0.009 0.025** 0.023** 0.040*** 0.031** 0.040*** 0.029** 0.018*
[0.90] [2.46] [2.49] [3.81] [2.87] [3.78] [2.77] [1.749]
Turnover -0.007*** 0.000 0.000 0.000 0.000 -.0001 0.000 0.000
[23.44] [0.64] [0.32] [0.37] [1.37] [0.40] [0.33] [0.214]
Distance -0.054*** -0.009*** -0.015*** -0.009*** -0.010*** -0.009*** -0.011*** -0.008***
[25.13] [3.65] [7.95] [3.88] [4.02] [3.91] [4.80] [3.381]
Small banks in groups -0.332*** -0.003 -0.023*** -0.001 0.018* -0.001 -0.014 -0.007
[38.86] [0.73] [2.80] [0.12] [1.79] [0.12] [1.49] [0.754]
Small banks -0.372*** 0.015 -0.016 0.012 0.037* 0.012 -0.002 0.020
[20.17] [0.71] [0.90] [0.54] [1.69] [0.56] [0.12] [0.951]
Cooperative banks -0.543*** 0.068*** 0.018 0.087*** 0.139*** 0.086*** 0.057*** 0.072***
[35.45] [3.88] [1.29] [4.93] [8.20] [4.90] [3.21] [4.187]
Bank share in the prov. 7.173*** --- --- --- ---
[163.23]
Avg. collateral -0.122*** --- --- --- ---
[6.94]
Rho -0.151*** -0.165*** -0.169*** -0.166*** -0.143
Uncensored
and total obs.
29,491
340,736 29,491
29,491
340,736
29,515
340,960
29,491
340,736
29,491
340,736
29,491
340,736
The regressions also include the constant as well as regional, sector and year dummies. Robust z statistics are in brackets. Standard
errors are adjusted for clustering in firms. * significant at 10%; ** significant at 5%; *** significant at 1%.


Table 8
Restructuring and financial distress: robustness check
(Heckman selection estimation, restructuring equation. Marginal effects)

1 2 3 4 5 6 7
Variables
Sample up
to 10
million
Sample 10
to 50
million
Restructuring
on credit
granted
Restructuring
before t+3
Restructuring
only maturity
extension
Restructuring
with new
finance
Only firms out
of distress at
t+3

Financing costs on assets -1.417** -1.253** -1.637*** -1.573*** -0.601*** -1.465*** -0.402
[5.96] [2.26] [7.34] [6.98] [3.13] [6.48] [-1.31]
ROA 0.350*** 0.509* 0.448*** 0.433*** 0.280*** 0.379*** 0.475***
[3.75] [1.74] [4.59] [4.25] [3.31] [4.17] [4.01]
Leverage -0.043*** -0.023 -0.009 -0.039** 0.017* -0.041*** -0.009
[2.67] [1.29] [0.70] [ 2.33] [1.84] [2.76] [-0.70]
Size (log of total assets) -0.049*** -0.077*** -0.025*** -0.049*** -0.007* -0.044*** -0.033***
[8.15] [5.32] [5.61] [-10.81] [1.90] [9.54] [-5.63]
Collateral (bank-firm level) -0.185*** -0.188*** -0.203*** -0.188*** -0.106*** -0.160*** -0.227***
[8.66] [4.36] [10.95] [ 9.16] [6.19] [8.31] [-8.41]
Single (1 / 0) 0.068*** 0.001 0.004 0.066*** -0.030* 0.124*** 0.068**
[3.14] [0.02] [0.23] [3.43] [1.84] [5.90] [2.42]
Bank share -0.178*** -0.245*** -0.057*** -0.209*** 0.012 -0.261*** -0.223***
[9.13] [5.24] [3.20] [11.76] [0.85] [13.93] [-8.96]
Score_decision (1 / 0) -0.070*** -0.160*** -0.060*** -0.108*** -0.082*** -0.080*** -0.088***
[5.97] [6.82] [5.67] [ 10.19] [8.90] [7.65] [-6.07]
Score_monitoring (1 / 0) 0.092*** 0.204*** 0.083*** 0.134*** 0.099*** 0.111*** 0.118***
[5.88] [7.46] [6.04] [10.48] [6.93] [7.94] [6.40]
Delegation (relative) 0.089* 0.230** 0.209*** 0.082* 0.189*** 0.094** 0.070
[1.76] [2.41] [4.71] [1.86] [4.71] [2.13] [1.15]
Soft information (1 / 0) 0.024*** -0.001 0.019** 0.006 0.011* 0.018*** 0.023**
[3.04] [0.08] [2.92] [0.88] [1.85] [2.71] [2.48]
Incentive (1 / 0) 0.053*** -0.006 0.039*** 0.036*** 0.053*** 0.034*** 0.042***
[4.37] [0.29] [3.71] [3.49] [5.71] [3.29] [2.81]
Turnover 0.000 -0.001 -0.001** 0.000 -0.001*** 0.000 0.000
[0.18] [1.23] [2.17] [ 0.30] [3.36] [0.18] [-0.79]
Distance -0.012*** -0.006 -0.008*** -0.010*** -0.008*** -0.008*** -0.011***
[4.09] [1.21] [3.51] [ 4.08] [3.74] [3.30] [-3.15]
Small banks in groups -0.007 0.019 -0.009 -0.008 -0.030*** 0.004 -0.011
[0.66] [0.94] [0.95] [ 0.88] [3.62] [0.39] [-0.79]
Small banks 0.014 0.009 -0.027 0.009 -0.015 0.012 -0.026
[0.60] [0.18] [1.29] [0.42] [0.83] [0.59] [-0.86]
Cooperative banks 0.083*** 0.101** 0.050*** 0.056*** 0.023 0.091*** 0.077***
[4.21] [2.61] [2.92] [3.35] [1.39] [5.12] [3.25]

Rho -0.137*** -0.281*** -0.180*** -0.153*** -0.170*** -0.161*** -0.1461***

Uncensored
and total obs.
22,285
285,496
7,241
55,240
29,491
340,736
29,491
340,736
29,491
340,736
29,491
340,736
14,857
172,605

The regressions also include the constant, as well as regional, sector and year dummies. Robust z statistics are in brackets. Standard


Table 9
Restructuring and financial distress: fixed effects for firms or banks
(linear probability models)

Fixed effects for firms Fixed effects for banks
Collateral (bank-firm level) -0.130***

Financing cost on assets -0.933***
[6.11] [6.46]
Bank share -0.128***

ROA 0.294***
[6.37] [4.93]
Scoring (1 / 0) -0.033***

Leverage -0.019**
[3.49]

[2.33]
Delegation (relative) 0.135***

Size ( log of total assets ) -0.027***
[2.81]

[9.56]
Soft information (1 / 0) 0.013*

Collateral (bank-firm level) -0.114***
[1.71]

[7.40]
Incentive (1 / 0) 0.022*

Single (1 / 0) 0.065***
[1.78]

[3.83]
Turnover -0.000

Bank share -0.172***
[0.80]

[11.46]
Distance -0.014***

Scoring in the year of restructuring -0.0533***
[5.52]

[2.96]
Small banks in groups -0.016

Distance -0.019***
[1.53]

[8.85]
Small banks -0.011

[0.45]

Cooperative banks 0.044**

[2.43]

n. observation 22,250

n. observations 29,491
n. of groups 7,702

n. of groups 270
The regressions also include the constant.
The regressions also include the constant, as well as regional,
sector and year dummies.


Table 10
New credit to firms in distress
(bivariate probit and Tobit estimation)

1 2 3 4
Variables
New credit
without
sample
selection
correction
New credit
with sample
selection -
baseline
specification
New credit
with sample
selection -
preferred
specification
Tobit
estimates
(marginal
effects)
Financing cost on assets -9.162*** -31.579*** -29.153*** -2.584***
[8.93] [5.75] [5.53] [11.33]
ROA 0.387 7.306*** 6.430*** 0.522***
[0.94] [3.95] [3.87] [4.78]
Leverage -0.522*** -0.991*** -0.946*** -0.129***
[3.26] [5.04] [4.92] [7.93]
Size (log of total assets) -0.123*** -0.812*** -0.763*** -0.059***
[6.98] [4.64] [4.59] [12.66]
Collateral (bank-firm level) 0.112* -3.347*** -2.947*** -0.210***
[1.64] [3.84] [3.71] [7.45]
Single (1 / 0) 1.898*** 3.119*** 2.982*** 0.384***
[18.37] [9.40] [9.69] [13.33]
Bank share -2.561*** -5.858*** -5.434*** -0.628***
[25.98] [6.97] [7.16] [23.97]
Scoring (1 / 0) -0.046 -0.554***
--- ---
[1.26] [4.07]
Score_decision (1 / 0) --- --- -1.482*** -0.124***
[4.18] [7.93]
Score_monitoring (1 / 0) --- --- 1.861*** 0.144***
[4.08] [7.52]
Delegation (relative) 0.097 4.465*** 1.762*** 0.133**
[0.50] [4.02] [3.52] [2.16]
Soft information (1 / 0) 0.026 0.212*** 0.315*** 0.019**
[0.80] [3.69] [3.91] [1.98]
Incentive (1 / 0) 0.086 0.529*** 0.704*** 0.054***
[1.96] [4.49] [4.31] [3.55]
Turnover 0.000 0.003** -0.002* -0.000
[0.01] [2.01] [1.35] [0.99]
Distance 0.006 -0.142*** -0.132*** -0.017***
[0.60] [3.62] [3.48] [5.89]
Small banks in groups 0.020 -0.038 -0.001 -0.026**
[0.46] [0.80] [0.03] [2.04]
Small banks -0.071 0.200* 0.076 -0.043
[0.83] [1.80] [0.76] [1.49]
Cooperative banks -0.083 1.056*** 1.188*** 0.024
[1.31] [3.56] [3.56] [1.09]
Blamda1
--- -1.019*** -0.977*** ---

[3.96] [3.87]
Blamda2
--- 9.329*** 8.153*** ---
[3.98] [3.86]
Observations 9,117 9,117 9,117 29,491
R-squared 0.125 0.127 0.127 0.026

The regressions also include the constant as well as regional, sector and year dummies. Robust z statistics are in brackets. Standard


REFERENCES
Economy, 105: 1-29.
di unindagine campionaria, Questioni e Economia e Finanza, 12, Banca dItalia.
Alessandrini, P., A.F. Presbitero and A. Zazzaro (2009) Banks, Distances and Firms Financing
Constraints, Review of Finance, 13: 261-307.
Berger, A.N. (2003), The economic effects of technological progress: Evidence from the banking
industry, Journal of Money, Credit and Banking, 35: 141-176.
Berger, A.N., N.H. Miller, M.A. Petersen, R.R. Rajan and J.C. Stein (2005a), Does function
Berger, A.N., W.S. Frame and N.H. Miller (2005b), Credit scoring and the availability, price, and
risk of small business credit, Journal of Money, Credit and Banking, 191-221.
Berger, A.N. and G.F. Udell (2002), Small Business Credit Availability and Relationship
Lending: The Importance of Bank Organisational Structure, The Economic Journal, 112
(477): F32-F53.
Journal of Banking and Finance, 30: 2945-2966.
Bergloef, R. and E.L. von Thadden (1994), Short-term versus long-term interests: Capital
structure with multiple investor, Quarterly Journal of Economics, 109(4):.1055-1084.
Bergman, Y.Z. and J.L. Callen (1991), Opportunistic underinvestment in debt renegotiation and
capital structure, Journal of Financial Economics, 29:.137-171.
Bolton, P. and X. Freixas (2000), Equity, bonds, and bank debt: Capital structure and financial
market equilibrium under asymmetric information, Journal of Political Economy, 108: 324-351.
Bolton, P. and D.S. Scharfstein (1996), Optimal debt structure and the number of creditors,
Journal of Political Economy, 104(1): 1-25.
Boot, A. (2000), Relationship banking: what do we know?, Journal of Financial Intermediation,
9: 7-25.
Bris, A. and L. Welch (2005), The optimal concentration of creditors, Journal of Finance, 60(5):
2193-2212.
Brown, D.T., C.M. James and R.M. Mooradian (1993), The information content of distressed
restructuring involving public and private debt claims, Journal of Financial Economics, 33:
93-118.
Brunner, A. and J.P. Krahnen (2008), Multiple lenders and corporate distress: Evidence on debt
restructuring, Review of Economic Studies, 75: 415-442.
Bulow, J.I. and J.B. Shoven, (1978), The bankruptcy decision, The Bell Journal of Economics,
9(2): 437-456.
Carletti, E., V. Cerasi and S. Daltung (2007), Multiple-bank lending: Diversification and free-
riding in monitoring, Journal of Financial Intermediation, 16: 425-451.

Carmignani, A. and M. Omiccioli (2007), Costs and benefits of creditor concentration: An
empirical approach, Temi di Discussione, 645, Banca dItalia.
Davydenko, S.A. and J.R. Franks (2008), Do bankruptcy codes matter? A study of defaults in
France, Germany, and the U.K, Journal of Finance, LXIII: 565-608.
Degryse, H., L. Laeven and S. Ongena (2009), The impact of organizational structure and lending
technology on banking competition, Review of Finance, 13: 225-259.
Degryse, H. and S. Ongena (2004), The impact of technology and regulation on the geographical
scope of banking, Oxford Review of Economic Policy, 20(4): 571-590.
Detragiache, E., P. Garella and L. Guiso (2000), Multiple versus single banking relationships:
Theory and evidence, Journal of Finance, 55: 1133-1161.
Elsas, R. and J.P. Krahnen (2002), Collateral, relationship lending and financial distress: An
empirical study on financial contracting, CFS Working Paper, 17.
Frame, W.S., A. Srinivasan and L. Woosley (2001), The effect of credit scoring on small business
lending, Journal of Money, Credit and Banking, 33: 813-825.
Gertner, R. and D. Scharfstein (1991), A theory of workouts and the effects of renegotiation law,
Journal of Finance, 46 (4): 1189-1222.
Giammarino, R.M (1989), The resolution of financial distress, Review of Financial Studies, 2: 25-47.
Guiso, L. and R. Minetti (2004), Multiple creditors and information rights: Theory and evidence
from US firms, CEPR Discussion Paper, 4278.
Guiso, L. and R. Minetti (2007), The structure of multiple credit relationship: Evidence from US
Firms, EUI Working Papers, ECO 46, European University Institute.
Hart, O. and J. Moore (1998), Default and renegotiation: A dynamic model of debt, Quarterly
Journal of Economics, 113: 1-41.
Heckman, J.J. (1979), Sample selection bias as a specification error, Econometrica, 47: 153-161.
Hoshi, T., A. Kashyap and D. Scharfstein (1990), The role of banks in reducing the costs of
financial distress in Japan, Journal of Financial Economics, 27: 67-88.
Jensen, M.C. (1989), Active investors, LBOs, and the privatization of bankruptcy, Journal of
Applied Corporate Finance, 2(1): 35-44.
Kahl, M. (2002), Economic distress, financial distress, and dynamic liquidation, Journal of
Finance, 57(1): 135-168.
Kordana, K.A. and Posner E.A. (1999), A positive theory of chapter 11, New York University
Law Review, 74: 161-234.
Longhofer, S.D. and J.A.C. Santos (2000), The importance of bank seniority for relationship
lending, Journal of Financial Intermediation, 9(1).
Manove, M., A.J. Padilla and M. Pagano (2001), Collateral vs. project screening: A model of lazy
banks, Rand Journal of Economics, 32(4): 726-744.
Meng, C. and P. Schmidt (1985), On cost of partial observability in bivariate probit model,
International Economic Review, 26: 71-85.
Mian, A. (2006), Distance constraints: The limits of foreign lending in poor economies, Journal
of Finance, LXI(3): 1465-1505.

Mocetti, S., M. Pagnini and E. Sette (2010), Information technology and banking organization,
mimeo, Bank of Italy.
Mohanty, M.S. (2005), An alternative method of estimating the workers reservation wage,
International Economic Journal, 19: 501-522.
Norton, E.C., H. Wang and C. Ai, (2004), Computing interaction effects and standard errors in
logit and probit models, The Stata Journal, 4(2): 154-167.
Ongena, S., and D.C. Smith, (2000), What determines the number of bank relationships? Cross-
country evidence, Journal of Financial Intermediation 9: 26-56.
Ongena, S., G. Tmer-Alkan and N. von Westernhagen (2007), Creditor concentration: An
empirical investigation, Deutsche Bundesbank, Discussion Paper, Series 2: Banking and
Financial Studies, 15.
Petersen, M.A. and R.G. Rajan(2002) Does distance still matter? The information revolution in
small business lending, Journal of Finance, 57: 2533-2570.
Rajan, R.G. (1992), Insiders and outsiders: The choice between informed and arms-length debt,
Journal of Finance, 47: 1367-1399.
Scott, J. (2006), Loan officer turnover and credit availability for small firms, Journal of Small
Tunali, I. (1986), A general structure for models of double selection and an application to a joint
migration/earning process with remigration, Research in Labor Economics, 8: 235-282.
White, M.J. (1989), The corporate bankruptcy decision, Journal of Economic Perspective, 3:
129-153.

LOAN OFFICER AUTHORITY AND SMALL BUSINESS LENDING.
EVIDENCE FROM A SURVEY
Michele Benvenuti
*
, Luca Casolaro
**
, Silvia Del Prete
*
and Paolo Emilio Mistrulli
***

1. Introduction
A large body of the literature has highlighted how large banks are less prone to lend to small
firms (e.g. Berger et al., 2005). An interpretation of the reluctance of large banks to lend to small
firms is that these borrowers are opaque and can only be screened and monitored on the basis of
soft information (Stein, 2002).
Soft information has two main characteristics that would make large banks less able to lend
to opaque borrowers than small intermediaries. Firstly, soft information is gathered by a face-to-
face interaction between lenders and borrowers which requires physical proximity. In principle, the
proximity to borrowers may be achieved by both large and small banks by means of their branch
networks. Banks would only need to delegate the acquisition of soft information to their branch
managers. However, the second key characteristic of soft information i.e. poor verifiability
gives rise to principal-agent problems when the production of information is delegated. According
to Stein (2002), agency problems are particularly severe in complex organizations where the
information has to be passed on through many hierarchical layers. By assuming that the size of the
bank and the complexity of its organization are closely linked, Steins model also predicts that
large banks are less able than small banks to produce soft information and thus to lend to opaque
borrowers.
The model proposed by Stein (2002) is based on the intuition that the ability of branch
managers to credibly transfer soft information i.e. the efficiency of the communication
technology depends on the organizational complexity of the bank. Therefore the efficiency of the
communication technology is crucial for the functioning of the so-called internal capital market, by
means of which internal financial resources are allocated by bank headquarters among branches.
Loan officers collect information, both soft and hard, about potential borrowers and then compete
for funds on the internal capital market. Naturally, bank headquarters can assess the quality of the
projects that branch officers want to finance only on the base of verifiable information. This
implies that, insofar as branch managers are unable to harden it, soft information is useless for
capital budgeting purposes. Thus, branch manager incentives to exert effort in the production of
soft information are positively related to their ability to show their headquarters that the projects
they want to finance are creditworthy. The efficiency of the communication technology, by means
of which information is passed on from branches to higher organizational layers, is then strictly
related to the level of the loan officers authority, that is the degree of control over lending
decisions that branch loan managers have.

*
**
Bank of Italy, Livorno.
***
Bank of Italy, Potenza.
We would like to thank for their very helpful suggestions Piero Alessandrini, Emilia Bonaccorsi di Patti, Paola Bongini, Luigi
Cannari, Hans Degryse, Giovanni Ferri, Giorgio Gobbi, Steven Ongena, Paola Rossi, Gregory Udell, Alberto Zazzaro and the
participants at seminars held at the Bank of Italy, and at the Ancona Conference on The Changing Geography of Money, Banking
and Finance in a Post-Crisis World.
176 Michele Benvenuti, Luca Casolaro, Silvia Del Prete and Paolo Emilio Mistrulli

Most of the extant empirical literature, investigating bank specialization in small business
lending, is based on the hypothesis that organizational complexity is directly related to bank size.
In particular, these contributions
1
assume that large banks are more hierarchical (i.e. the head office
is less prone to delegate) and branch loan officers have less authority compared with small banks
loan officers. However, Albareto et al. (2008), by exploiting a very unique dataset on the
organizational characteristics of Italian banks recently surveyed by the Bank of Italy, showed that
there is some heterogeneity in bank organization even within banks of similar size. As a result,
even if size is probably an important driver affecting bank organizational design, it is not the only
influence and it is also important to look at the nature of the loan officers authority.
In this paper, we test whether the branch loan officers authority affects bank specialization
in small business lending. Differently from previous contributions, we do not proxy the branch loan
officers authority with the size of the bank. Indeed, we exploit the recent Bank of Italy survey
which has provided us with detailed information about the number of hierarchical layers involved
in small business lending approvals, loan officer turnover, the use of credit-scoring techniques, the
amount of money up to which branch loan officers are entitled to lend, and the type of loan officer
compensation schemes. Furthermore, the availability of detailed and new information on the
organizational structure of a very large sample of Italian banks gives us the opportunity to shed
light on how the various characteristics of bank organization interact among themselves.
While the relation between the loan officers authority and small business lending
specialization is clearly stated in Steins model (2002), it is not an easy task to test that link
empirically. Firstly, following Aghion and Tirole (1997), one has to be aware that there is a
distinction between formal authority (the right to decide) and real authority (the effective control
over decisions), and econometricians may, at best, reasonably observe the formal authority of a
loan officer while, if they want to test Steins predictions, they need to observe the real authority.
Secondly, loan authority and effort depend on many factors, such as: the maximum amount loan
officers are entitled to lend on their own, their discretion in setting rates, the kind of loan officer
compensation scheme, and so on. On the one hand, these factors interact among themselves in
shaping the organizational design of the loan approval process within a bank; on the other hand,
formal organizational features may impinge on the real authority of a loan officer in quite a
complex way. For these reasons, instead of focusing on a specific organizational characteristic, we
adopt an holistic approach and we include in the estimation different organizational variables
which are presumably related to the real authority of the loan officer in order to investigate bank
size being equal how different organizational solutions in modelling local officers incentives can
affect bank portfolio choices. Thus, we contribute to the literature investigating the role of
organization for bank lending specialization in two main ways. Firstly, by using a unique dataset
including information about bank organizational features and lending techniques, we are able to go
beyond the comparison between large and small banks. Secondly, we show that small business
lending specialization is positively correlated with many bank organizational characteristics linked
to a loan officers authority; moreover, we support the idea that the link between formal and real
authority may be quite complex.
2
In other terms, by using the survey on bank organization we are
able to shed light on some features of the organizational black box. However, we are aware that
this is only the first step to understanding the way that box may actually work.

1
A comprehensive review on banking organizational issues is summarized in Cerqueiro, Degryse and Ongena (2007).
2
Our econometric specification is not able to avoid endogeneity problems and our results cannot therefore be interpreted as causal
effects.
Loan officer authority and small business lending. Evidence from a survey 177

Another insight of the paper is that we verify how the distance between the borrower and
bank headquarters (functional distance)
3
interacts with the organizational scheme, especially with
the loan officers authority, and then affects bank willingness to lend to opaque borrowers.
Following DellAriccia and Marquez (2004) and Hauswald and Marquez (2006), we argue that the
ability of a branch loan officer to transfer soft information to higher hierarchical layers which is
closely related to the loan officers authority depends not only on the organizational complexity
of the bank, as put by Stein (2002), but also on functional distance.
4

Finally, we exploit geographical and sector heterogeneity across banks to compute our
measure of specialization in small business lending for each bank at the province/sector level. This
allows us to disentangle supply (bank organization characteristics) and demand factors (firm
characteristics) and to control also for some lending relationship features, like the functional
distance between bank headquarters and borrowers.
The main results of the analysis are the following. Firstly, our analysis confirms that large
banks are generally less prone to lend to small firms than other banks. This is in line with previous
empirical contributions, supporting the view that small banks are better able to lend to opaque
borrowers compared with larger intermediaries.
Secondly, we show that bank size is only part of the whole story since, even after having
controlled for it, bank organizational features help to explain differences across banks in small
business lending. In particular, our analysis suggests that loan officer authority is positively related
to small business lending activity. Indeed, we find the more loan officers are involved in loan
approval decisions and/or in setting loan pricing, the more the bank is specialized in small business
lending.
Conversely, the amount of money up to which loan officer may lend autonomously is
negatively correlated with bank specialization in small business finance. This result is due to the
fact that the amount of money loan officers are entitled to lend does not necessarily entail a greater
authority for loan officers. Indeed, this result is consistent with the hypothesis that loan officers
gain private benefits by expanding the overall amount of money they lend (the so-called empire-
building hypothesis). Consequently, they can more easily expand their lending portfolio when they
are entitled to grant a larger amount of money per each contract. In this way, they obtain the same
overall branch target by lending to a smaller number of borrowers. Furthermore, by assuming that
bank debt is a good proxy for firm size and opaqueness they can save on soft information-gathering
efforts.
We also find that, banks which give greater importance to soft information in lending
decisions are more willing to lend to small businesses, consistently with the view that small firms
are opaque borrowers and that their creditworthiness assessment is largely based on proprietary
information. Small business lending specialization is also negatively affected by loan officers
turnover. This result suggests that loan officers efforts to gather soft information are less if they
stay at a given branch for a shorter time.
In the same direction, our econometric analysis indicates that the distance between bank
headquarters and borrowers (functional distance) affects the availability of lending to small
businesses: the shorter the distance between dealing partners, the higher the bank specialization in
small business finance. This is consistent with the hypothesis that functional distance lowers the

3
To the best of our knowledge, Alessandrini et al. (2005 and 2008) were the first to use this definition, in opposition to the physical
proximity between borrowers and bank branches (the so-colled operational distance).
4
Mistrulli and Casolaro (2008) provide some evidence concerning functional distance and loan rate setting, supporting the view that
distance affects the ability of a loan officer to harden soft information. Jimenez, Salas-Fumas and Saurina (2008) investigate the
relationhip between functional distance and the use of collateral.

capability of branch loan officers to credibly pass on soft information to higher hierarchical layers
of the bank, reducing their ability to compete for internal resources and, as a consequence, the
specialization of banks in lending to opaque borrowers too.
The rest of the paper proceeds as follows. Section 2 describes the data and the econometric
methodology. Section 3 presents the empirical findings. Section 4 concludes.

2. Data and methodology
2.1 Sources of data
In 2007 the Bank of Italy surveyed the lending practices in use by Italian banks at the end of
2006. More than 300 banks participated in the survey, accounting for 83 per cent of the overall
Italian banking systems lending to firms. Leasing and factoring companies, branches of foreign
banks, and consumer credit banks were not included in the survey since these intermediaries are
highly specialized in specific segments of the credit market. The sample covers almost all Italian
banks, except for the very smallest ones. The accuracy of the data collected is high. Firstly,
preliminary interviews with some bankers allowed the questionnaire to be designed so as to avoid
potential misunderstandings; secondly, ex post interviews helped fill missing information and
clarify erratic answers.
Banks participating in the survey were questioned about lending to small enterprises: the
number of hierarchical layers involved in the decision to grant a loan, the kind of information
required in order to make a lending decision (soft versus hard information), and information about
branch loan officers, such as the amount of money up to which they are allowed to lend
autonomously, the speed of their turnover, and their compensation schemes (see Albareto et al.,
2008).
5

Initial evidence for bank organization suggests that loan officer power delegation and
turnover generally increase with bank size (Table 1), but a certain variability in these
organizational features persists within homogenous groups of banks sharing a similar governance
scheme. Furthermore, the adoption of different lending techniques (e.g soft information or rating
systems) also shows mixed solutions across groups of banks (Table 2).
We then match the information obtained from the survey with bank Supervisory Reports
allowing us to compute a measure of specialization in lending to small businesses at both sector (23
sectors) and geographical level (103 provinces) for each bank participating in the survey. We also
use bank balance sheet data to compute control variables to take into account bank-level
characteristics affecting portfolio composition. After a cleaning procedure to detect outliers, 239
banks remain.
Since the main obstacle of the econometric exercise is to proxy the local officers real
authority, because this is not directly observable, we adopt a pragmatic approach. In particular, it is
reasonable to assume that both loan officers decisional power and their efforts depend crucially on
organizational design (in term of decentralization and control). Thus, we use a set of bank
organizational controls to capture local officers authority.

5
For a similar survey concerning US banks, see Udell (1989).

2.2 The research question and the econometric set-up
The main aim of the econometric analysis is to verify whether banks which delegate more
real authority to their loan officers (that we approximate with various bank organizational features)
are more oriented to small business lending, as suggested by Stein (2002).
Our baseline regression is a cross-section of the following type:

( )
z j i iz i i i ijz
Geography Sector Bank Disth ing CreditScor ation SoftInform on Organizati f SHSE , , , , , , =
The dependent variable (SHSE), which ranges between 0 and 1, is the outstanding amount of
loans, granted by bank i to small businesses (with less than 20 employees) of sector j and
headquartered in province z, as a share of the overall amount of lending to non-financial firms of
sector j headquartered in province z, granted by bank i.
We have computed bank specialization in small business finance separately for each
province/sector pair to which a bank lends. In this way we are able to disentangle loan demand
factors which vary at both sector and geographical level and supply factors and in particular
bank organizational characteristics which vary at the bank level. To avoid eventual bias due to
minimal amounts of lending, we drop all bank/sector/province observations accounting for less
than 0.002 per cent of bank-level total loans.
6
Since the bank organizational characteristics are
observed at the end of 2006, our measure of bank specialization in small business finance has been
computed as the average of the four 2007 quarters. In this way we mitigate both endogeneity and
seasonality problems. The econometric exercise is carried out by an OLS estimation and the results
are checked with a tobit regression since our dependent variable is bounded between 0 and 1.
We investigate the impact of different factors on bank specialization in small firm finance.
Firstly, Organization
i
is our key set of variables, controlling for bank organization
characteristics related to a loan officers real decision-making power (see Table 3 for more details).
In particular, we include in the estimation variables focusing on:
- the scope of the power delegated (i.e. the amount of finance up to which loan officers may
autonomously lend, their discretion in setting interest rates or asking for collateral);
- the turnover policy adopted by banks;
7

- whether loan officers play a relevant role in the loan approval process or not;
- the kind of reward scheme: some bank headquarters may place a greater emphasis on
branch profitability, while others may be more oriented to pursue a loan volume target or risk
mitigation.
Bank lending specialization could also be affected by the type of lending technologies
adopted (Berger and Udell, 1995 and 2002). In order to control for their impact we use two dummy
variables. On the one hand, SoftInformation
i
is a dummy that is equal to 1 if, according to the
survey, bank i considers soft information as crucial, very important or important in lending
decisions. Since this kind of information is difficult to verify, if banks require loan officers to base

6
The value corresponds at the 10
th
percentile of the bank level total loan distribution. Alternative thresholds do not affect our main
results.
7
In some unreported regressions we have divided the level observed of a loan officers delegation and turnover for bank i by the
maximum computed over all the banks lending to small firms headquartered in province z, since we approximate the boundaries of
local credit markets the market relevant to small firms with those of the 103 Italian provinces. In this way, following Degryse,
Laeven and Ongena (2007), we aim to capture how bank is organization differs from that of its competitors (other bank lending to
firms in province z). Our main findings are robust to the employment of these average measures of delegation and turnover.

their lending decisions on qualitative information or on personal knowledge, this implies that loan
officers are largely autonomous. On the other hand, CreditScoring
i
is a dummy that is equal to 1 if
a bank has introduced a rating system mainly to monitor and control borrowers creditworthiness.
Moreover, we also take into account that the distance between borrowers and bank
headquarters can affect the ability of loan officers to pass information along the hierarchy, which
then lowers the ability of loan officers to compete in the internal capital market. To this aim we
compute a dummy variable (DISTh) that is equal to 1 if the province
8
where borrowers are
headquartered is the same province where the lending bank i has its headquarters. We have adopted
a discrete measure for functional distance since most of lending relationships are concentrated
within the province where both the lender and the borrower are located.
Finally, Bank
i
is a vector of variables taking into account other bank-level characteristics,
such as size, risk and cost efficiency, potentially correlated with portfolio composition, while
Sector
j
and Geography
z
are respectively firm sector and province dummies, controlling for loan
demand factors.
Descriptive statistics are reported in Table 3, which shows further information on the
definition of variables employed in the econometric analysis.

3. Results
3.1 Basic regressions
Results are reported in Table 4. In the first model (model a) we have regressed our measure
of bank specialization in small business finance over some bank and firm characteristics. In
particular, after having controlled for loan demand factors, i.e. for the sector of activity and the
province where the firm is headquartered (sector and province dummies), we find evidence
supporting the view that bank size (Size) affects lending portfolio choices. In line with many other
papers (e.g. Berger et al., 2005), our results show that large banks are less prone than small banks
to lend to small firms. In particular, moving from the 25
th
percentile of the bank size distribution to
the 75
th
percentile, the share of lending to small firms decreases by about 9 percentage points. In
the first model, we also control for the number of branches a bank has in the province where the
borrowers are headquartered (Branches). We find that, other things being equal, banks are more
prone to lend to opaque borrowers in local markets where their network of branches is larger. Our
interpretation is that, since lending to small firms is mostly based on soft information and this
information may be gathered only at a local level, a large branch network helps banks to collect soft
information, thus making easier to screen and monitor opaque firms.
We also control for the functional distance between the borrowers and the lenders
headquarters. In particular, Disth is a dummy which equals 1 if the firm is headquartered in the
same province where the bank has its headquarters, 0 otherwise. We find that bank specialization in
small business lending is higher when the lender and the borrower are functionally close.
The following equations add controls for loan officer authority. The main aim is to verify
whether or not bank size fully captures all relevant organizational characteristics of banks affecting
loan officer authority and then banks lending choices. Briefly, we investigate whether, for any
given bank size, additional controls for loan officer authority have some explanatory power for
bank specialization in small business lending. If this is the case, that would imply that there is some

8
Italy is divided into 103 provinces, 20 regions and 5 areas (North West, North East, Centre, South, Islands).

heterogeneity in terms of organizational structure among banks of the same size and that bank size
does not capture all the significant characteristics of bank organization.
First of all we control for the scope of loan officer delegation (model b). In particular,
delegation_approval is a dummy variable which is equal to 1 if banks participating in the survey
reported that the role of branch officers in loan approval is important,
9
and 0 otherwise. We find
that compared with banks which delegate less power to their loan officers, banks reporting that loan
officer power is large are more prone to lend to small businesses; in particular, their share is 2.9
percentage points greater. We consider this variable as crucial to capture the scope of the power
delegated to loan officers that, within reason, goes beyond the strict approval decision. Indeed, it is
reasonable to think about loan approval delegation as a pre-requisite for delegating the decision
about specific loan contract covenants. For example some banks may place more emphasis than
others on the discretion loan officers may have to set loan interest rates, while others may stress
more the quantity of money loan officers are entitled to lend on their own initiative. Model c) then
adds a measure of loan officers discretion in setting interest rates. The dummy delegation_price
equals 1 if price discretion is important,
10
0 otherwise. We then interact this variable with
delegation_approval. Results suggest that, conditional on having being delegated the power to
approve loans, branch officers incentives in gathering soft information and then in lending to small
business are even greater when they have some discretion to set interest rates. In particular, the
share of small business lending is on average 4.2 per cent greater for banks which delegate
significant decision-making power to loan officers both in approving loan requests and setting
interest rates.
Another component of loan officer authority might be the amount of finance up to which
he/she is entitled to lend on his/her own initiative to each borrower. We then plug into the equation
(model d) another control variable which is defined as the logarithm of the maximum amount of
money loan officers are entitled to lend to any single borrower (delegation_loan). We get a
negative sign for the coefficient of delegation_loan, which means that the specialization of banks
in small business lending is decreasing in terms of the amount of money loan officers are entitled to
lend on their initiative. To interpret this result one has to take into account three aspects of lending.
First, there is a strict positive correlation between firm size and the size of the loan. Second, firm
opaqueness correlates positively with the firm size. Third, the effort loan officers have to exert in
order to screen and monitor firms is positively correlated with firm opaqueness. Thus, when loan
officers are able to choose among a wider set of loan contracts because their headquarters have
defined a higher threshold for the amount loan officers may lend on their own initiative they tend
to finance a set of larger firms which therefore allows them to lend the same amount of money
overall while exerting less effort in screening and monitoring borrowers compared with the case in
which they lend to a set of smaller firms.
11
The negative coefficient for delegation_loan indicates
that loan officers tend to switch to less opaque borrowers when the amount they can lend to each of
them increases.

3.2 Loan officer tenure and lending technologies
In this section, we test whether banks specialization in small business lending depends on
loan officer turnover and the lending technologies adopted. We argue that both sets of variables
may affect loan officer incentives: the tenure of loan officers at a given branch may affect their

9
See Table 3 for many details on variable definition.
10
See previous footnote.
11
Note that, even if the variable delegation_loan refers to SMEs, there is a positive and strong correlation between the maximum
amount of money loan officers may lend to SMEs and the amount they may lend to larger firms.

ability to benefit from efforts to gather soft information, which requires a prolonged interaction
with borrowers. Lending technologies may differ in the use of soft or hard information. When a
bank puts more emphasis on hard information (e.g. by adopting credit scoring techniques), this may
entail a reduction in the capability of loan officers to have control over lending decisions (i.e. loan
officers real authority) thus deterring soft information acquisition.
Model a) in Table 5 includes a control for the time spent by loan officers at a branch (Loan
Officer Tenure). The coefficient for this regressor is positive and statistically different from zero,
showing that banks which are more prone to lend to small firms allow a longer interaction of loan
officers with the local credit market.
12
This result is consistent with the hypothesis that the
investment in soft information requires a longer time to yield some return. As a consequence, loan
officers have greater incentives to invest in soft information if their stay at the same branch is
expected to be longer, otherwise their efforts will mostly benefit incoming loan officers.
Another aspect which may affect loan officer authority is the kind of information considered
more important within the bank. Equation b) includes a dummy variable (Soft Information), which
is equal to 1 if banks make greater use of soft information than hard information and 0 otherwise.
We find that the coefficient for the prevalence of soft information is positive and significant too
(the relevance of soft information increases the share of small business lending by more than 1 per
cent), reinforcing the interpretation that small business lending requires relationship lending
technology. Similarly, by adding a control for the use of credit scoring techniques (credit scoring),
we get a negative coefficient (the impact on small business lending ranges between 1 and 3 per
cent) for this variable (model c), indicating that the use of hard information is negatively related to
the specialization of banks in small business lending.
Finally, model d) includes only one measure of loan officer power delegation (delegation)
while it keeps the control for loan officer tenure and the use of soft information. This more
parsimonious specification is used in the rest of the paper as a benchmark (baseline) equation.

3.3 Loan officer compensation schemes
In this section we take into account the characteristics of the loan officer compensation
schemes (Table 6). In particular, we add some variables which are related to the variable
component of loan officer pay-off. Since in general loan officers get a bonus on the base of short-
run portfolio performance, one has to keep in mind when interpreting the relative coefficients that
these variables are short-run constraints. First, we plug a dummy variable (loan portfolio size) that
equals 1 if the size of the loan portfolio held by loan officers is a stringent constraint, and 0
otherwise. In other words, we compare banks which put great emphasis on lending portfolio size
targets (loan portfolio size =1) with other banks which do not. Model a) shows that if loan
portfolio size is a stringent constraint then loan officers are more prone to lend to small firms. For
those banks, the share of business lending is 1.3 per cent greater. This result is consistent with the
view that when loan officers are rewarded for the expansion of a loan portfolio per se, they are less
worried about lending to opaque borrowers whose ex post profitability may depend on information
which has been observed by them but that could not be transferred in a credible way to
headquarters.
Model b) includes another control for the loan officer compensation scheme. In particular,
the dummy loan portfolio profitability equals 1 if the profitability of the overall loan portfolio held
by the branch manager is a stringent constraint, 0 otherwise. We find that the coefficient for

12
Scott (2006) found similar results for U.S. Banks. Ferri (1997) investigated a similar issue for Italian banks.

profitability targets is negative (the impact on small business specialization is more than 3 per cent)
indicating that lending to opaque firms requires an investment in soft information which yields
some return in the medium-long run.
By interacting loan portfolio profitability and loan portfolio size (model c) we may also
check whether the effect of loan portfolio profitability differs among banks which put a higher
emphasis on the size of loan portfolio than other banks. The coefficient for this interaction term
(model c) is positive, signalling that the effect of profitability constraints is smoothed by the
quantity constraint. In other terms, the profitability constraint is less stringent when banks also
emphasize quantity targets. Loan officers are therefore less concerned with generating returns in
the short-run since they are also rewarded better if they expand their lending overall.

3.4 Robustness checks
In order to check the robustness of our main findings, we run some alternative econometric
exercises on our baseline specification (Table 5, model d).
First, since our measure of bank specialization in small business lending is computed as a
share (defined between 0 and 1), it can be considered as censored variable. Under this assumption,
it would be more feasible to estimate our parameters using a tobit methodology. Secondly, we
verify how our results are robust to a more severe measure of bank specialization in small business
lending (e.g. firms with less than 5 employees). Table 7 (respectively, models a and b) shows that
our baseline results are mostly confirmed and that findings are not affected by a different definition
of our dependent variable or by different estimation techniques. In particular, the more important
the role of the loan officer in loan approval process, the higher the bank specialization in small
business finance. Moreover, banks which put much emphasis on the use of soft information in
lending to opaque borrowers or which invest in loan officer tenure show a higher orientation
towards small business finance.
13

Even within banks of comparable size we observe wide heterogeneity in terms of
organizational characteristics. Looking at Tables 1 and 2, it is possible to note some meaningful
differences in organization and lending techniques between large and small banks, especially for
banks not belonging to groups. In particular, large banks exhibit on average higher loan officer
turnover and delegation in lending to small firms; furthermore, large intermediaries usually assign
less importance to soft information and make greater use of credit scoring systems in loan
approvals. We then investigate whether our results are driven by some non-linear effect related to
bank size. In order to confirm that our results, once we have controlled for size, are capturing this
within group heterogeneity, we run our baseline regression only for the sub-sample of small
banks. Table 7 (model c) shows that our results are robust also in a more homogenous group of
banks by size; this evidence allows us to be confident that our main results are not just driven by a
non linear-size effect.

13
It is important to point out that when we consider bank orientation towards very small firms (those with less than 5 employees), the
distance between lenders and borrowers negatively correlates with a banks specialization in lending to this kind of firm. At the
same time, organizational variables related to loan officers authority seem to be less important in supporting small business
lending. This may be due to the fact that these firms are so opaque that it is so hard for loan officers to channel information through
the banks higher hierarchical levels. As a consequence, organizational solutions may be of little help to induce loan officers efforts
to gather soft information.

4. Conclusions
In this paper we investigate whether bank organization characteristics affect bank
specialization in lending to small firms. By using a unique dataset based on a survey of Italian
banks, we are able to go beyond the traditional large versus small banks comparison. We also
exploit detailed data on lending activity at province level to investigate how the functional
distance between bank headquarters and borrowers affects bank specialization.
Our main findings are the following. In line with previous results, we support the idea that
bank size negatively affects bank specialization in small firm finance due to small banks
comparative advantage in investing in soft information gathering and assessing opaque borrowers
creditworthiness. Moreover, we argue that, apart from size, a banks organizational features (loan
officer-specific ones) play an important role in explaining its orientation to small business lending.
In particular, we show that the higher the real involvement of the branch officer in the loan
approval process, as well as in setting interest rates, the higher the bank specialization in financing
small and opaque firms. In the same direction, we find that the longer the local officer stays at the
same branch, the more banks are focused on small business finance; experience gained over time
can improve a loan officers ability to establish long-lasting and informative relationships with
opaque borrowers. We also find that the amount of money up to which loan officer may lend
autonomously is negatively correlated with bank specialization in small business finance. This
result is consistent with the hypothesis that loan officers gain private benefits by expanding the
overall amount of money they lend (the so-called empire building hypothesis). In this way, they
obtain the same overall branch target by lending to a smaller number of borrowers, thus saving soft
information gathering effort.
Bank specialization in small business lending depends crucially on lending technologies, too.
Banks more involved in soft information gathering exhibit a greater share of credit granted to small
firms, while the improvement of internal rating systems (mainly to monitor borrowers
creditworthiness) negatively affects small business lending orientation, giving a prominent role to
hard information.
Finally, we show that banks are less prone to finance small firms in those provinces that are
far from their headquarters. This is in line with Hauswald and Marquezs study (2006) and
Mistrulli and Casolaros findings (2008), showing that functional distance is a proxy for lender-
borrower asymmetric information problems.
All in all, our results suggest that, consistently with Steins model (2002), an increase in a
loan officers authority is positively correlated with bank specialization in small business lending.
However, in line with Aghion and Tirole (1997), we are aware of the distinction between loan
officers formal authority (the right to decide) and real authority (the effective control over
decisions). Indeed, we control for a rich set of formal rules (loan officers tenure, the power
delegated to loan officers, both in terms of the size of the loan and the interest rate charged, the
structure of loan officers compensation scheme, etc.) that all contribute to shape a loan officers
real authority.


APPENDIX
Table 1
Bank size, delegation and loan officer (LO) turnover (1)
(thousand of euros and months)

Maximum amount of money LOs
are allowed to lend autonomously
Months LO stays in a branch
mean p25 p50 p75 mean p25 p50 p75

Large and medium-sized banks 458 108 200 380 32 26 32 36
Small banks in banking groups 211 80 125 250 40 30 36 48
Stand-alone small banks 112 44 90 150 48 36 40 60
Cooperative banks 114 10 30 100 49 36 48 60

Total 176 18 71 150 45 33 38 60

(1) According to the size classification provided by the Bank of Italy, small banks have total assets of less than 7 billion euros.

Table 2
Bank size, loan pricing and lending technology (1)
(units)

Loan officers
allowed to lower
interest rate by
more than 25 b.p.
Importance of soft
information (2)
Credit scoring
crucial in lending
decision (3)
YES NO YES NO YES NO

Large and medium-sized banks 2 12 20 17 30 6
Small banks in banking groups 3 38 38 36 32 19
Stand-alone small banks 0 14 10 10 8 9
Cooperative banks 12 91 108 76 45 41

Total 17 155 176 139 115 75

(1) According to the size classification of the Bank of Italy, small banks have total assets of less than 7 billion euros. (2) This
classification is based on a dummy variable that is equal to 1 (YES) for those banks that ranked face-to-face relationships with the
entrepreneur or the usage of non-traceable information as crucial or very important. (3) In this case YES is used for banks using
credit scoring and/or internal rating systems for SME finance, whose score index is crucial for the assessment of credit.


Table 3
Descriptive statistics on the estimation sample

Variables Description Mean Median Std. Dev.
Expected
sign
Dependent variable
Specialization in
small business
lending
Share of credit granted by each bank i to sole-
proprietorships and other non-financial
enterprises with less than 20 employees
operating in province z (103 provinces) and
sector j (23 economic activities).
0.34 0.18 0.37
Bank characteristics
Size Log of total assets. 22.89 23.03 1.74 -
Risk Bad loans on total loans (percentage). 3.31 2.64 2.57 -
Cost-to-income ratio Operational costs divided by gross income
(percentage).
61.42 60.17 11.41 +/-
Branches Number of branches of each bank i in each
province z.
8.32 2.00 20.05 +
Bank organization
Soft information The dummy is equal to 1 if bank i ranked face-
to-face relationships with the entrepreneur or
the usage of non-traceable information in the
first three places (crucial, very important
or important).
0.53 1 0.50 +
Delegation_approval The dummy is equal to 1 if the LO has a
crucial or very important role in loan
approvals.
0.89 1 0.31 +
Delegation_price The dummy is equal to 1 if the LO has a
crucial or very important interest rate
discretion.
0.60 1 0.49 +
Delegation_loan This index is the log value of the maximum
loan that a local officer can autonomously
grant to SMEs.
5.19 5.11 1.47 +/-
Loan Officers length
of tenure
The index is calculated as the logarithm of
months the LO of bank i stays in the same
branch.
3.53 3.58 0.37 +
Credit Scoring The dummy is equal to 1 if for a given bank
the internal rating system is crucial or very
important in monitoring SMEs
creditworthiness.
0.56 1 0.50 -
Loan portfolio
profitability
The dummy is equal to 1 if the LOs pay-off is
related in a significant way (crucial or very
important) to the loan portfolio profitability.
0.74 1 0.44 -
Loan portfolio size The dummy is equal to 1 if the LOs pay-off is
related in a significant way (crucial or very
important) to the size of the loan portfolio
(volume targets).
0.87 1 0.34 +
Functional distance
Disth This dummy is equal to 1 if SMEs are
headquartered in the same province where
bank i has its headquarters.
0.08 0 0.28 +


Table 4
Loan Officer authority and small business lending: Basic regressions

Model a Model b Model c Model d
Bank Characteristics
Size -0.0326*** -0.0325*** -0.0320*** -0.0214***
[0.0009] [0.0009] [0.0009] [0.0011]

Risk -0.0037*** -0.0038*** -0.0032*** -0.0026***
[0.0006] [0.0006] [0.0006] [0.0006]

Cost-to-income ratio 0.0018*** 0.0020*** 0.0017*** 0.0014***
[0.0001] [0.0001] [0.0001] [0.0001]

Branches 0.0012*** 0.0011*** 0.0011*** 0.0010***
[0.0001] [0.0001] [0.0001] [0.0001]

Lender-borrower distance
Disth 0.0088* 0.0119** 0.0135*** 0.0109**
[0.0049] [0.0049] [0.0049] [0.0049]

Loan Officer authority
Delegation_approval 0.0296*** 0.0039 0.0111**
[0.0047] [0.0052] [0.0051]

Delegation_approval*Delegation_ price 0.0363*** 0.0317***
[0.0030] [0.0030]

Delegation_loan -0.0229***
[0.0012]

Constant 0.7440*** 0.7030*** 0.7098*** 0.6034***
[0.0289] [0.0301] [0.0301] [0.0302]

Observations 62,763 62,763 62,763 62,763
R
2
0.18 0.18 0.18 0.19

The dependent variable, ranging between 0 and 1, is the outstanding amount of loans granted by bank i to small businesses (firms with
less than 20 employees) of sector j and headquartered in province z, as a ratio of the overall amount of lending to non-financial firms
of sector j headquartered in province z, and granted by bank i. Sector and province dummies are always included. Robust standard
errors are reported in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%.


Table 5
Loan Officer authority and small business lending:
Loan Officer turnover and lending technologies
Model a Model b Model c Model d
Size -0.0193*** -0.0195*** -0.0194*** -0.0261***
[0.0011] [0.0011] [0.0011] [0.0010]

Cost-to-income ratio 0.0014*** 0.0015*** 0.0015*** 0.0020***
[0.0001] [0.0001] [0.0001] [0.0001]

Risk -0.0029*** -0.0030*** -0.0031*** -0.0044***
[0.0006] [0.0006] [0.0006] [0.0006]

Branches 0.0010*** 0.0010*** 0.0010*** 0.0011***
[0.0001] [0.0001] [0.0001] [0.0001]

Disth 0.0106** 0.0107** 0.0112** 0.0121**
[0.0048] [0.0048] [0.0048] [0.0049]

Delegation_approval 0.0166*** 0.0200*** 0.0184*** 0.0369***
[0.0052] [0.0053] [0.0053] [0.0048]

Delegation_approval*Delegation_ price 0.0303*** 0.0316*** 0.0312***
[0.0030] [0.0030] [0.0030]

Delegation_loan -0.0215*** -0.0211*** -0.0192***
[0.0012] [0.0012] [0.0013]

Loan Officers tenure 0.0318*** 0.0308*** 0.0296*** 0.0413***
[0.0045] [0.0045] [0.0045] [0.0045]

Lending technologies
Soft information 0.0153*** 0.0131*** 0.0100***
[0.0028] [0.0028] [0.0028]

Credit scoring -0.0135*** -0.0325***
[0.0032] [0.0030]

Constant 0.4294*** 0.4211*** 0.4229*** 0.4209***
[0.0384] [0.0384] [0.0384] [0.0385]

Observations 62,763 62,763 62,763 62,763
R
2
0.19 0.19 0.19 0.19

of sector j headquartered in province z and granted by bank i. Sector and province dummies are always included. Robust standard

Table 6
Loan Officer authority and small business lending:
Loan Officer compensating schemes

Model a Model b Model c
Size -0.0200*** -0.0200*** -0.0196***
[0.0012] [0.0013] [0.0013]
Cost-to-income ratio 0.0023*** 0.0000 0.0002
[0.0002] [0.0002] [0.0002]
Risk -0.0057*** -0.0022** -0.0036***
[0.0007] [0.0010] [0.0011]
Branches 0.0011*** 0.0010*** 0.0010***
[0.0001] [0.0001] [0.0001]
Disth 0.0055 0.0174*** 0.0170***
[0.0057] [0.0063] [0.0063]

Delegation_approval 0.0634*** 0.0373*** 0.0374***
[0.0062] [0.0070] [0.0070]

Loan Officers compensation schemes
Loan portfolio size 0.0132*** 0.0382*** 0.0185**
[0.0048] [0.0053] [0.0078]
Loan portfolio profitability -0.0383*** -0.0690***
[0.0039] [0.0102]
Loan portfolio size* Loan portfolio profitability 0.0384***
[0.0117]
Loan Officers tenure 0.0440*** 0.0128** 0.0102*
[0.0051] [0.0057] [0.0058]

Lending technologies
Soft information 0.0078** 0.0236*** 0.0257***
[0.0032] [0.0037] [0.0038]
Credit scoring -0.0307*** -0.0313*** -0.0359***
[0.0034] [0.0038] [0.0041]

Constant 0.4173*** 0.4891*** 0.4987***
[0.0435] [0.0486] [0.0488]
Observations 49,074 39,768 39,768
R
2
0.19 0.19 0.19

of sector j headquartered in province z, and granted by bank i. Sector and province dummies are always included. Robust standard


Table 7
Loan Officer authority and small business lending: Robustness checks

Model a)
Tobit estimation:
whole sample
(marginal effects)
Model b)
OLS estimation:
very small
firms
Model c)
OLS estimation:
small banks
Size -0.0105*** -0.0177*** -0.0395***
[0.0006] [0.0009] [0.0027]
Cost-to-income ratio 0.0014*** 0.0011*** 0.0008***
[0.0001] [0.0001] [0.0002]
Risk -0.0029*** -0.0019*** -0.0013
[0.0004] [0.0005] [0.0009]
Branches 0.0009*** 0.0006*** 0.0004**
[0.0001] [0.0000] [0.0002]
Disth 0.0187*** -0.0171*** 0.0146**
[0.0030] [0.0038] [0.0070]
Delegation_approval 0.0203*** 0.0259*** 0.0138**
[0.0034] [0.0040] [0.0070]
Loan Officers tenure 0.0269*** 0.0243*** 0.0589***
[0.0027] [0.0037] [0.0061]
Lending techniques
Credit scoring -0.0166*** -0.0066*** -0.0107**
[0.0018] [0.0025] [0.0045]
Soft information 0.0057*** 0.0040* 0.0345***
[0.0018] [0.0024] [0.0044]

Constant - 0.2493*** 0.7147***
- [0.0308] [0.0761]
Observations 62,763 62,763 31,193
R
2
0.10 0.17 0.16
In the tobit estimation (model a) the dependent variable, ranging between 0 and 1, is the outstanding amount of loans granted by
bank i to small businesses (firms with less than 20 employees) of sector j and headquartered in province z , as a ratio of the overall
amount of lending to non-financial firms of sector j headquartered in province z, and granted by bank i. In the OLS estimation in
model b the dependent variable, ranging between 0 and 1, is the outstanding amount of loans granted by bank i to very small
businesses (firms with less than 5 employees) of sector j and headquartered in province z, as a ratio of the overall amount of lending
to non-financial firms of sector j headquartered in province z, and granted by bank i. In the OLS estimation in model c the dependent
variable, ranging between 0 and 1, is the outstanding amount of loans granted by small bank i to small businesses (firms with less
than 20 employees) of sector j and headquartered in province z, as a ratio of the overall amount of lending to non-financial firms of
sector j headquartered in province z, and granted by bank i. Sector and province dummies are always included. Robust standard


REFERENCES
Economy, 1, 1-29.
di unindagine campionaria, Questioni di Economia e Finanza, 12, Banca dItalia.
Alessandrini, P., M. Croci and A. Zazzaro (2005), The geography of banking power: The role of
functional distance, BNL Quarterly Review, 235: 129-167.
Alessandrini, P., F. Presbitero and A. Zazzaro (2008), Banks, distances and financing constraints
for firms, Review of Finance, forthcoming.
Berger, A. and G. Udell (1995), Relationship lending and lines of credit in small firm finance,
Journal of Business, July.
Berger, A. and G. Udell (2002), Small business credit availability and relationship lending: The
importance of bank organizational structure, The Economic Journal, 112: 32-53, February.
Berger, A., N. Miller, M. Petersen, R. Rajan and J. Stein (2005), Does function follow
organization form? Evidence from the lending practices of large and small banks, Journal
Cerquiero, G., H. Degryse and S. Ongena (2007), Distance, bank organizational structure and
credit, mimeo, Tilburg University.
Degryse, H., L. Laeven and S. Ongena (2007), The impact of organizational structure and lending
technology on banking competition, CentER-Tilburg University, Tilburg.
DellAriccia, G. and R. Marquez (2004), Information and bank credit allocation, Journal of
Financial Economics, 72: 185-214.
Ferri, G. (1997), Branch manager turnover and lending efficiency: Local vs. national banks, BNL
Quarterly Review, Special issue, March.
Hauswald, R. and R. Marquez (2006), Competition and strategic information acquisition in credit
markets, Review of Financial Studies, 3: 967-1000.
Jimenez, G., V. Salas-Fumas and J. Saurina (2008), Organizational distance and use of collateral
for business loans, Documento de Trabajo, 0816, Banco de Espaa.
Mistrulli, P.E. and L. Casolaro (2008), Distance, lending technologies and interest rates, mimeo,
Bank of Italy.
Scott, J. (2006), Loan officer turnover and credit availability for small firms, Journal of Small
Stein, J. (2002), Information production and capital allocation: Decentralized versus hierarchical
firms, Journal of Finance, 5: 1891-1921.
Udell, G. (1989), Loan quality, commercial loan review and loan officer contracting, Journal of
Banking and Finance, December.

DISCUSSION
Angelo Baglioni
-
The paper by Infante and Rossi focuses on the role played by foreign banks in Italy. The
authors document the increasing presence of foreign banks over the last fifteen years, with a
transition from wholesale to retail business. To summarise the main findings, foreign banks have
had a significant impact on competition only as far as households are concerned, particularly for
mortgage loans; this impact has been reflected in customer mobility, interest rates and collateral
requirements. The impact on competition has been much less significant as far as firms are
concerned. The authors interpretation is that foreign banks have focused on the sectors where
informational asymmetries play a minor role and credit services are more standardised, while
domestic banks have focused on sectors where opacity and customer relationships enable them to
enjoy some market power.
I find the analysis of the paper interesting and convincing. It is consistent with the theory of
banking, in particular some contributions of the early nineties, showing the competitive advantage
of the incumbent banks over new entrants, which is due to essentially two factors that limit the
contestability of the market for business lending: (i) the entrant necessarily has to screen a sample
of potential borrowers, drawn from a population of firms that has already been screened by the
incumbent; (ii) an incumbent bank has typically established customer relationships with some
firms, giving it an informational advantage.
The presence of foreign banks in Italy may be traced back to the Second EU Directive on
Banking, which introduced cross-border liberalisation from 1992. Previously, many scholars and
practitioners in the banking sector had forecast a rapid increase of foreign banks presence in Italy
from that date, but this did not happen. For a number of years the business of foreign banks was
limited and basically restricted to the wholesale segment. Only in recent years have we witnessed a
significant increase in their retail business. This suggests substantial difficulties for foreign banks
interested in entering the Italian banking market. I believe that the reasons for this difficulty differ
between firms and households.
While in the case of firms the problem may related to informational asymmetries, this factor
is not so relevant for households, since business in this segment consists mainly of collateralised
loans (mortgages) and deposit-taking. The relevant explanation, for households, is the switching
costs incurred by depositors owing to the location of bank branches and the bundling of several
financial products, such as asset management and payment services. Italian banks have been quite
active in supplying such bundles of services tied to deposits, in order to increase switching costs
and so develop market power vis--vis households. This may be seen as a reaction to the fall in the
profitability of the traditional deposit-taking and lending business following liberalisation. In this
regard, the analysis of the competitive impact of foreign banks in the collection of deposits and in
provision of asset management services seems to be a promising line of future research (I am afraid
that the impact has been rather weak so far).

-
Catholic University of Milan, Institute of Economics and Finance.
194 Angelo Baglioni

Actually Italian banks have tried to limit their customers mobility in the mortgage loan
sector as well, where the competitive challenge from the foreign banks has been stiffest. For
instance, in May 2008 they signed an agreement on the renegotiation of variable rate mortgages,
inducing borrowers to ask for a loan renegotiation with their present bank rather than look for a
better alternative at another bank. Moreover, the Italian anti-trust authority has found some
irregular behaviour in the application of the mortgage loan portability rule.
Finally, a more technical comment on the use of the Herfindahl index is in order. This
measure of concentration produces ambiguous results: greater concentration in the mortgage loan
market is correlated both with greater customer mobility (suggesting more competition) and with
higher interest rates (suggesting less competition). These unsurprising findings reveal the weakness
of the structure-conduct-performance approach as applied to the banking industry. The approach
works well in industries with only a few producers, but the number of banks currently operating in
Italy is quite large, especially in Milan where most of the foreign banks are present. The limits to
competition, rather than in the concentration of the market structure, are to be found in the market
segmentations mentioned due to informational asymmetries and switching costs and to
collusion, either explicit (through agreements like those related to payment services) or implicit
(leading to a coordination in the provision of some services).
Session 2: Banking organization and small business lending - Comments 195

REFERENCES
Broecker, T. (1990), Credit-worthiness tests and interbank competition, Econometrica, 58: 429-
452.
Rajan, R. (1992). Insiders and outsiders: The choice between informed and arm's length debt,
Journal of Finance, 47: 1367-1400.
Sharpe, S. (1990), Asymmetric information, bank lending and implicit contracts: A stylized model
of customer relationships, Journal of Finance, 45: 1069-1087.

DISCUSSION
Paola Bongini
*
The two papers Loan officer authority and small business lending. Evidence from a survey
by Benvenuti, Casolaro, Del Prete and Mistrulli and Debt restructuring and the role of lending
technologies by Micucci and Rossi take a unique perspective on the role of organizational forms
and lending technologies in explaining differences in bank behavior that cannot be fully captured
by bank size.
Since the authors have done a good job of explaining: a) the rich and unique dataset they rely
on; b) the advanced econometric techniques implemented; and c) the basic results, which reveal a
deep knowledge of both the theoretical background and banking practice, I wont review their
specific empirical findings. Instead, I want to discuss what we can learn from the results and what
questions still remain open.
Both studies deal with a topic that has become relevant in recent academic research, that is
banks internal organization and the use of lending technologies. Although the papers focus on two
different moments of the lending activity the former on the decision to lend to SMEs; the latter on
the decision to restructure claims on distressed SMEs they both take the same approach of
considering heterogeneity in banks behavior. Such heterogeneity goes beyond the traditional
dualism between large and small banks or bank debt versus public debt: instead, heterogeneous
behavior is to be found within homogeneous groups of banks. Thus not every large bank is
necessarily little inclined to lend to SMEs, nor should every bank be willing to restructure its loans
to financially distressed firms just because bank debt is easier to renegotiate than market debt.
What is more relevant, as the authors state, is the way in which banks shape their internal
organization and which lending technologies they use to screen and monitor loans to SMEs.
Let us analyze the first issue, i.e. banks organization. A recent strand in the literature, both
theoretical and empirical, has highlighted the role of organization in shaping bank-firm
relationships, in particular with SMEs. Relationship lending is bound up with the specific role and
authority of the local branch manager, in terms of : a) his/her autonomy in extending credit to
SMEs ; b) the number of hierarchical layers involved in small business loan approval; c) the banks
policy on bank manager tenure and turnover; d) the compensation scheme used to reward bank
manager performance.
The study by Benvenuti et al. suggests that banks are more willing to lend to small business
when:
a) the loan officer is more involved in the approval decision and/or in setting loan pricing.
Conversely the amount of money up to which loan officers may lend autonomously is
negatively correlated with bank specialization in small business finance;
b) the loan officer tenure at the local branch lengthens;
c) the loan officers compensation is linked to the size of the lending portfolio

*
University of Milan-Bicocca, Department of Business and Finance.
198 Paola Bongini

In the study by Micucci and Rossi, banks are more prone to restructure outstanding debt in
case of distress when:
a) the loan officer is delegated more power, in terms of the amount of loans to SMEs that he
can extend autonomously with respect to the amount the CEO of the bank can extend;
b) the loan officer receives some performance-related compensation.
On these findings I have two main comments.
First, the internal organization of a bank represents, for both studies, a key variable in
defining the type of bank-firm relationship that the bank intends to institute with SMEs
(transactional or relationship lending). However, the two studies deliver different results with
respect to the specific features that such organization should have. Two questions are of special
interest. First, why are these results not aligned, given the use of similar data-sets? Second, how do
the authors justify the fact that banks that are more willing to lend to SMEs are the very same
banks that are less prone to restructure small business debt?
Second, a flourishing new stream of theoretical literature (Stein, 2002; Berger et al, 2005)
has highlighted how the hierarchical distance or number of layers between the banks
headquarter and the local branch could be one the main causes keeping large and organizationally
complex banks from lending to small, opaque borrowers. However, this thesis has been subjected
to little empirical testing due to the difficulty of measuring the hierarchical layers in a banks
organization by means of publicly available information. One exception, for the US system, is
Rajan and Wulf (2006): using a detailed database of managerial job descriptions in over 300 large
U.S. firms, they find that firm hierarchies are becoming flatter, with increasing delegation of
authority to divisional managers. Similar data was available to the authors of the papers discussed
here, in that the two studies exploit a Bank of Italy survey on banks organizational aspects that
also provides detailed information on the number of hierarchical layers involved in loan approval.
Unfortunately neither one uses this variable in its econometric tests, which would have offered
important support to the thesis that large banks are more inclined to interact with SMEs when they
have a flat organization.
The second factor that can crucially differentiate banks behavior is the role of lending
technologies. The term lending technologies was introduced by Berger and Udell in their
extensive work on relationship banking. According to Berger and Udell (2006), each lending
technology is a unique combination of the primary information (hard or soft), screening and
underwriting policies/procedures, structure of the loan contracts, and monitoring strategies and
mechanisms. Thanks to these combinations the traditional pairs relationship banking-opaque
firms and transactional banking- transparent firms can be superseded.
Both studies proxy the concept of lending technologies by means of a dummy variable
indicating:
a) whether a bank has adopted credit scoring techniques in small business lending;
b) whether a bank considers qualitative information an important factor in assessing the
creditworthiness of SMEs.
They find similar results: banks that assign greater importance to soft information in lending
decisions even when credit scoring techniques are implemented are more willing to lend to
SMEs and to restructure their debt.
Such results are not surprising given the importance of qualitative (soft) information in
screening and monitoring opaque borrowers, which most SMEs are; they are also consistent with
the relevance of the loan officers authority in the lending process: the acquisition of soft
information needs to be delegated to the branch managers, who have an incentive to do so properly
Session 2: Banking organization and small business lending - Comments 199

only if they can use such information in their lending decisions. In other terms, soft information is
closely related to the extent of the loan officers authority.
Rewarding the acquisition and use of soft information requires face-to-face interaction
between the lender and the borrower, which was initially conceived in terms of physical proximity
of the branch to the borrowing firm. More recently, geographical distance has been supplanted by a
new concept, namely the functional distance between the banks headquarters and the borrowing
firms, which also considers social and cultural factors that help determine the distance between the
bank and its customers. In line with this new notion of distance, both the studies discussed here
analyze the importance of functional distance in explaining banks specialization in lending to
SMEs or their propensity for debt restructuring. Functional distance is proxied by :
a) a dummy variable which takes the value 1 if the borrower and bank have their
headquarters in the same province (Benvenuti et al.);
b) the distance (log of kilometers) between the banks headquarters and the location of the
borrowing firm (Micucci and Rossi).
Both proxies are statistically significant and positively correlated with the banks decision to
support SMEs. For the study by Benvenuti et al. this result is somewhat troubling, in that their
proxy implies that a firm with headquarters in the province of Milan is as just as distant,
functionally, from, say, the UBI group (headquartered in Bergamo) as a firm headquartered in a
region far removed from Lombardy, where both the group and the firm are located. The clear
implication of the statistical significance of this proxy for distance is that geographical proximity is
a binding constraint for successful small business lending. This would offer additional support for
the thesis that only small banks, especially mutual banks, have a propensity to lend to small
borrowers.
How to define and measure a banks proximity to its borrowers is not a trivial question. It is
an important concern for the top management of large banks and the subject of what has become a
substantial body of academic research.
Research in the eighties addressed the issue of optimal bank size, producing policy
implications that corresponded to the substantial concentration of the banking sector over the past
fifteen years. In the nineties, mainstream research focused on value creation: the answer was
business and revenue diversification, which also meant, especially for large banks, a shrinkage of
traditional lending business. In the first decade of the new century, the main issue was how to
combine being global with acting locally. The answer was to separate production from
distribution as the issue was analyzed mainly with respect to banking activity in the securities
industry. Today, following the global financial crisis, the main issue in banking is how to define a
banks proximity to its customers, given that such proximity seems to represent the key variable in
a successful bank-firm relationship. The answers appear to lie in the specific design of the banks
organization and its lending techniques. The studies by Micucci and Rossi and Benvenuti et al. are
good examples of timely empirical research in this field.

200 Paola Bongini

REFERENCES
Berger, A.N., N.H. Miller, M.A. Petersen, R.G. Rajan and J.C. Stein (2005), Does function follow
organizational form? Evidence from the lending practices of large and small banks, Journal
Journal of Banking and Finance, 30: 2945-2966.
Rajan, R.G. and J. Wulf (2006), The flattening firm: Evidence from panel data on the changing
nature of corporate hierarchies, The Review of Economics and Statistics, 88(4): 759-773.

5 - Banche Mercati Territoriali 2010

Uploaded by

Copyright:

Available Formats

5 - Banche Mercati Territoriali 2010

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

5 - Banche Mercati Territoriali 2010

Uploaded by

Copyright:

Available Formats

Banks, Local Credit

Markets and Credit Supply

You might also like