1

SME financing and the choice of lending technology in Italy: complementarity or

substitutability?

Francesca Bartoli, Giovanni Ferri

**, Pierluigi Murro

***, Zeno Rotondi

****

Abstract

This paper investigates SME financing in Italy. The literature distinguishes

between two main different lending technologies (LTs) for SMEs: transactional

and relationship LTs. We find that banks lend to SMEs by using both LTs

together, independently of the size and proximity of borrowers. Moreover, we

show that the use of soft information decreases the probability of firms being

credit rationed. Finally, we find that more soft information is produced when

the bank uses relationship LT as primary technology individually or coupled

with transactional LT. Our results support the view that LTs can be

complementary, but reject the hypothesis that substitutability among LTs is

somehow possible for outsiders by means of hardening of soft information.

Keywords: Lending technologies, bank-firm relationship, soft information,

hard information, small business finance

JEL Classifications: G21, G30, O16

1. Introduction

Among academics and policymakers there is a clear perception that small and medium-

sized enterprises (SMEs) lack adequate financing and need to receive special assistance

(see, e.g., Vos et al., 2007). Recent research shows that also large banks provide large

amounts of funding and other services to small firms (e.g., De la Torre et al., 2010).

Nevertheless, uunder the current paradigm in SMEs lending research, large banks are

believed to specialize towards relatively large, informationally transparent firms using

hard information, while it is held that small banks have advantages in lending to smaller,

less transparent firms using soft information. Hence large banks would tend to specialize

in the transaction-based lending technology, while small bank in the relationship lending

technology. Theories based on incomplete contracting suggest that small organizations

have a comparative advantage in activities that make extensive use of soft information.

The model of Stein (2002) predicts that large banks will tend to shift away from small-

business lending, because this is an activity that relies more heavily on the production of

soft information, which cannot be verifiably documented in a report that the loan officer

can pass on to his superiors. Hence, the loan officer’s incentives to produce high-quality

information are weak when she works in a large bank. On the contrary, in the case of

larger firm lending can be based more heavily on verifiable information, such as firm’s

financial statements, the balance sheet, etc.. In this case the model suggests that a large

bank will have no problem at providing incentives for hard information production.

UniCredit; e-mail: francesca.bartoli@unicredit.eu;

Lumsa University; e-mail: g.ferri@lumsa.it;

Luiss University; e-mail: pmurro@luiss.it;

UniCredit; e-mail: zeno.rotondi@unicredit.eu;

(corresponding author). The views put forward in the paper belong exclusively to the authors and do not

involve in any way the institutions of affiliation. Pierluigi Murro gratefully acknowledges financial support

from UniCredit and University of Bologna, Research Grant on Retail Banking and Finance.

2

The above paradigm has found some support in the empirical literature (see, e.g., Berger

et al., 2005). There have been some important refinements, but the conclusion on the

dichotomy between large versus small banks in the choice of lending technologies still

holds. In turn, Berger and Udell (2006) argue that a common oversimplification is the

treatment of transactions technologies as homogeneously unsuitable for lending to

informationally opaque SMEs. Moreover, Berger and Black (2011) suggest that large

banks do not have equal advantages in all of these transactions technologies and such

advantages are not all increasing monotonically in firm size. They also analyze lines of

credit without fixed-asset collateral in order to focus on relationship lending and confirm

that small banks have a comparative advantage in relationship lending. But their

comparative advantage appears to be strongest for lending to larger firms.

Hence, the current paradigm on SME lending emphasizes the dichotomy between large

and small banks by predicting that relationship lending will be used to the exclusion of

transaction-based lending technologies and vice versa, i.e. suggesting that these

alternative types of lending technologies tend to be mutually exclusive due to the

presence of incentives in specializing in one of them. In other words, the possibility of

complementarity is not neglected but is considered a less likely outcome in practice and

therefore has so far gathered very little attention in the literature.1

The aim of the present paper is to fill this gap in the literature and investigate the

possibility for banks of combining lending technologies for financing SMEs,

independently of their size. In particular, we argue that the paradigm that suggests that

large financial intermediaries are disadvantaged in relationship-based lending to opaque

SMEs is misleading. We show that complementarity among transactions and relationship

lending technologies is indeed a prevailing phenomenon, compared to specialization in

one primary lending technology, and that complementarity is higher for large banks

compared to small local banks.

To address these issue, we use a novel component of survey micro-data allowing us to

learn the lending technology used by the firm’s main bank. The data refer to the end of

2006 and come from the tenth wave of the Survey of Italian Manufacturing Firms (SIMF)

run by UniCredit banking group.2 This survey constitutes an ideal testing ground for two

main reasons. First, the data set provides unusually detailed information on the

relationship between the firm and its main bank, based directly on firms’ responses to

survey questions. Second, the small and medium size of the businesses in our sample and

the central role of banks in the external financing of investment renders Italy an ideal

environment to study the firm-main bank relationship. In fact, in Italy stock and bond

markets are relatively underdeveloped so that SMEs that are denied loans by banks are

usually forced to scale down their investment plans.

In the first part of our empirical analysis we study the specific features and the

deployment of lending technologies that appear to be more widespread toward SMEs. In

1 To our knowledge of the literature, an exception is represented by Uchida et al. (2008, 2006) who have

found the existence of complementarity among lending technologies for the case of Japan, by using an

approach for identifying lending technologies similar to that followed in the present paper, although the

analysis developed by them is rather different. They argue that further research based on other countries

should be developed in order to assess whether their findings should be interpreted as being inconsistent

with the prior literature or rather better interpreted as a reflection of an idiosyncratic situation prevailing in

Japan. 2 Formerly the survey was run by Mediocredito Centrale and Capitalia banking group.

3

particular, we provide empirical evidence on the existence and relevance of

complementarity between transaction-based and relationship-based lending technologies.

A possible explanation of the novelty of our result compared to the existing literature can

be found in the methodology. In fact, the identification strategies used in the literature for

the two alternative lending technologies imply by construction that they are mutually

exclusive (see, for instance, Berger and Black, 2011). On the contrary, our identification

strategy does avoid this problem by defining lending technologies in terms of their

specific features allowing by construction the possibility that the bank is using at the

same time more than a single lending technology when financing a given firm.

The second part of the analysis addresses the role of lending technologies in the

production of soft information under both hypotheses of a single primary lending

technology or complementarity among lending technologies. Previous literature has

associated soft information with relationship lending only. Various papers suggest that

more hierarchical banks, such as large and foreign banks, are relatively less capable of

processing and quantifying soft information and transmitting it through the channels of

large and complex organizations (Berger et al., 2001; Stein, 2002). However, Petersen

(2004) conjectures that transactional lenders might be able to “harden” soft information

to boost their local competitive stance and allow them to compete more aggressively

outside core markets.

The approach used in the present analysis for identifying lending technologies allows us

to shed new light providing a better understanding of the above issues. In order to tackle

these issues, the second part of our analysis is divided into two steps. Firstly, we measure

the production of soft information by the main bank and investigate its impact on the

probability that a firm is credit-rationed. The results show that soft information lowers the

probability that SMEs are credit rationed. Secondly, we try to understand the impact of

transaction-based and relationship lending technologies in the production of soft

information by the main bank. We find that the production of soft information increases

when either the relationship lending technology is used alone or together with the

transactions lending technology. By contrast, when the transactions lending technology is

used alone it seems to be ineffective in producing soft information. The implications of

these findings are twofold. First, the way soft information becomes embodied in the

lending decision might still differ between relational and transactional technologies.

Second, substitutability between lending technologies for outsiders by means of

hardening of soft information might be rather unfeasible.

The paper is structured as follows. Section 2 briefly discusses the literature on lending

technologies and the role of soft information for financial intermediaries. Section 3

presents the dataset, as well as the methodology we use to construct the variables

employed. Sections 4 presents the empirical evidence on lending technologies and the

role of soft information. Section 5 concludes.

2. Related literature

2.1 Lending technologies

Banks lend to SMEs by means of a variety of technologies. Berger and Udell (2006)

define a lending technology as a unique combination of primary information source,

4

screening and underwriting policies/procedures, loan contract structure, and monitoring

strategies/mechanisms. Different banks use different lending technologies (Rajan, 1992).

Thus, the choice of the main bank is a key component for the strategy of any firm, in

particular for SMEs that usually depend on bank financing as a source of external

funding. Among the various lending technologies used to finance firms, the literature has

thus far focused mostly on two classes: transaction-based lending technologies and

relationship lending technologies. According to the prevailing paradigm large banks hold

a comparative advantage in transactional lending, while the smaller-sized or local banks

have an edge in relationship lending (Stein, 2002).

There is a growing literature on the lending technologies that banks use to finance SMEs.

The empirical research has tried to test the results derived from the theoretical models. In

particular, several papers have analyzed the impact of relationship lending on the

financing of SMEs. For the United States, Cole (1998) finds that a lender is less likely to

grant credit to a firm if the customer relationship has lasted for one year or less, or if the

firm deals with other financial counterparts. On data for Italy, Angelini et al. (1998) show

that the intensity of relationship banking reduces the probability that borrowing firms will

be rationed, even though the lending rates charged by the banks tend to increase as the

bank-firm relationship lengthens. For Belgian enterprises, Degryse and Van Cayseele

(2000) detect the role of relationship banking along two different dimensions: borrowing

rates increase as the bank-firm relationship lengthens, while borrowing rates decrease

when the scope of the bank-firm relationship – defined as the purchase of additional

information intensive services (other than the loan) – increases.

Recently, both the theoretical and the empirical strands of the literature analyze also the

transaction-based lending technologies. Often, the literature has used the transaction

lending label for any type of loan based on information that is easily verifiable by

outsiders. By contrast, some authors underline that transaction lending is not a single

homogeneous lending technology but should be separated into a number of distinct

transaction technologies used by financial institutions. Berger and Udell (2006) suggest

that transactions technologies include financial statement lending, small business credit

scoring, asset-based lending, factoring, fixed-asset lending, and leasing. They define and

describe each of these lending technologies, highlight its distinguishing features, and

show how the technology addresses the opacity problem. Also the empirical literature

tries to explain the transaction-based lending technologies. A number of studies focus on

each individual technology separately. For example, Berger and Frame (2007) study

credit scoring and Udell (2004) asset-based lending.

2.2 Hard and soft information

The two classes of lending technologies can be primarily distinguished by means of the

type of information a bank uses in granting and monitoring the loan. Transaction-based

lending technologies are typically based (primarily) on hard quantitative information,

such as information derived from the borrowers’ balance sheets and/or the collateral

guarantees they offer (Berger and Udell, 2006). Instead, relationship lending technologies

assign a key role to soft information, qualitative information obtained via personal

interaction/acquaintance and that is difficult to codify (Rajan, 1992). For these reasons,

the literature suggests that a transaction lending technology is more desirable for more

5

transparent firms, while considering the relationship lending technology to be more

appropriate for opaque firms (suffering more intense asymmetries of information).

Although the literature generally distinguishes the lending technologies on the basis of

the type of information which is exchanged between the firm and the bank, only few

papers try to study in detail what is meant by hard and soft information. According to

Petersen (2004) hard information is quantitative, easy to store and transmit in impersonal

ways, and its content is independent of the collection process. Instead, soft information is

qualitative, often communicated in text, and so not easy to store. Also, soft information

contents depend on the collector of the information. In fact, soft information is gathered

personally and the decision maker is the same person as the information collector. Thus,

according to Stein (2002), less hierarchical banks are better able to use soft information in

their decisions.

However, Uchida et al. (2012) suggest that loan officers appear to be capable of

producing as much soft information at large banks as they do at small banks. Their

findings are related to a strand of literature dealing with the effects of organizational

variables on soft information production. Liberti and Mian (2009) show that delegating

decision-making at a lower level in a large multi-national bank’s hierarchy is likely to be

more soft information-intensive. Similarly, Agarwal and Hauswald (2010) find that

branches in a large bank that are more independent produce more soft information.

Related empirical evidence, going in the same direction, is provided by Mocetti et al.

(2010), who examine the interaction between information technology and banking

organization. In particular, they show that banks equipped with more information and

communication technology capital and resorting to credit scoring delegate credit

decisions relatively more to local branch managers in small business lending activities. In

turn, Benvenuti et al. (2010) show that the propensity to lend to SMEs tends to diminish

as bank size increases. This effect is partly offset by organizational measures

strengthening local loan officers’ independent decision-making power, by affecting

branch manager’s degree of delegation of credit power, incentives schemes and turnover.

The distinction among lending technologies derives from the idea that there are two types

of production functions using distinct inputs: hard and soft information. However, the

nature of information is not exogenously fixed. In fact, the lenders practices show us that

it may be possible to change the nature of information. For example, Frame et al. (2001)

find that credit scoring is associated with an increase in the portfolio share of U.S. small-

business loans, reducing information costs between borrowers and lenders. Moreover,

Berger et al. (2005) show results consistent with the hypothesis that the use of credit

scoring increases SME credit availability, in particular for relatively risky credits.

Albareto et al. (2008), reporting the results of an Italian survey conducted by the Bank of

Italy in 2007, illustrate that medium and large banks use soft information in their credit

scoring models. Their empirical evidence for Italy supports a trend of convergence in

recent years in the structures of lending organization between local and large banks, with

a greater degree of decentralization of credit decisions. Finally, another example of a

change in the nature of information is group lending, such as that involving Mutual

Guarantee Institutions (MGIs). Columba et al. (2010) and Bartoli et al. (2013) suggest

that banks, especially large ones, appreciate this kind of lending technology in order to

lend to SMEs. They find that MGIs, through peer monitoring and joint liability, help

banks to mitigate SMEs’ asymmetric information problems.

6

3. Data and empirical model

3.1 Methodology and data description

We model the complementarity among lending technologies as:

.111 iiii uxzy (1)

where yi is the lending technology used to finance firm i, zi is the vector of control

variables, xi accounts for the presence of relationship lending and ui is the vector of

heteroskedastic-robust standard errors. We estimate these equations using an OLS

model.3

Our main data source is the Tenth wave of the Survey on Italian Manufacturing Firms

(SIMF), run by the UniCredit banking group in 2007. Every three years this survey

gathers data on a sample of Italian manufacturing firms having more than 10 employees.

The 2007 wave consists of 5,137 enterprises. All the firms with more than 500 employees

are included, while those having a number of employees in the range 11 to 500 are

sampled according to a stratified selection procedure based on their size, sector, and

geographic localization. The main strength of this database is the very detailed

information it collects on individual firms. In particular, the 2007 wave features

information regarding the firm’s: a) ownership structure; b) number and skill degree of

employees; c) attitude to invest in R&D and whether it has made innovations; d) extent of

internationalization and exports; e) quality of the financial management and relationships

with the banking system. This information refers to the three years previous to the survey

year, in our case 2004-2006.4 The firms in the sample cover approximately 9% of the

reference universe in terms of employees and about 10% in terms of value added. Thus,

the sample is highly representative of the economic structure of Italian manufacturing.

Italy provides an ideal testing ground for examining the characteristics of lending

technologies that banks use to finance SMEs. The industrial structure is, in fact,

characterized by the large presence of small and medium-sized firms for which banks are

the main source of external funding. In 2009, the total value of stocks traded (% of GDP)

was 64.2%, compared with 258.8% in the United States. In this framework, SMEs are

financially vulnerable due to their almost exclusive dependence on bank financing as a

source of external finance.

Table 2 presents the descriptive statistics. At the average, the firms have been in business

for 22 years; beyond 60% of them have fewer than 50 employees (below 4% of the firms

have more than 500 employees); 70% of them are localized in the North. Only 1% are

listed on the Stock Exchange, while 37% have their balance sheet certified by external

auditors. Using Pavitt’s taxonomy (Pavitt, 1984), the distribution among sectors shows

that almost half of the enterprises belong to traditional sectors, while only 5% have their

business in the high tech sectors. Moving on to their financial set up, the average length

of the relationship with the main bank is 17 years; 49% of the firms have a national bank

3 As a robustness check, we estimate model (1) using a Tobit model. The results, available upon request,

are qualitatively similar. 4 The survey is public and available upon request to UniCredit banking group.

7

as their main banking counterpart, 10% entrust a larger-sized cooperative bank, 7%

feature a savings bank as their main bank, 5% entrust a smaller-sized cooperative mutual

bank, while 28% of the firms have another type of bank as their main bank. Finally,

multiple banking is extensive: on average firms do business with five banks and the share

of loans obtained from the main bank is 32% of the total banking loans received.5

Particularly relevant for our analysis, the 2007 wave of the survey features a peculiarity

with respect to the previous waves. Specifically, an entirely new set of questions is

specifically tailored to investigate in depth the relationship between the firm and its main

bank. In this paper we focus particularly on two questions where the firm is asked to state

which of the characteristics – choosing from a given list – have been important in the

firm’s selection of its main bank, as well as stating which characteristics, in the firm’s

view, best describe the way its main bank grants credit. Unsurprisingly, given the fact

that answering this section of the survey was relatively more time-consuming, only one

third of the surveyed enterprises (exactly 1,541 firms) answered these questions. We

cannot rule out self-selection. In other words, it is possible that the choice by a firm to

answer this part of the questionnaire was not random. However, we think that this self-

selection in the respondents would bias the results only if it were systematically related to

one or more of the explanatory variables.

To complement the survey, we also use data from other sources (see Table 1 for details

on the variables). We employ data from the Bank of Italy on the presence of banks in

local markets. We use data provided by the Italian National Statistics Office (ISTAT) on

civil suits and population per judicial district, as well as on the value added and

population of provinces. Finally, we employ data on social capital by Guiso et al. (2004).

3.2 Lending technology indices

We consider four indicators of lending technology similar to those in Uchida et al.

(2006). We capture the characteristics of different lending technologies using the

question “In your view, which criteria does your bank follow in granting loans to you?”.

In answering this question the firm was required to give a weight (going, in descending

order, from 1, very much, to 4, nil) to fifteen factors (see the Appendix). Most of these

factors are related to one of the lending technologies. We then link the factors that we

believe to be most closely associated with each lending technology based on the Berger

and Udell (2006) classification scheme. For reason of comparability of our results with

those in Uchida et al. (2008, 2006), we focus only on four lending technologies from this

classification.6

First, financial statement lending, is a transactions technology based primarily on the

strength of a borrower’s financial statements. Berger and Udell (2006) argue that banks

underwrite commercial loans using the financial statement lending technology for firms

5 Employing data on self-reporting firms might raise concerns about firms overstating or understating their

characteristics. We do not believe this has occurred for two main reasons. On the one hand, the Italian law

(l. 675/1996) on the treatment of personal data forbids using them for objectives other than that mentioned

in the survey, namely the elaboration of statistical tables. Hence, firms should have no incentive to lie.

Second, the personnel employed in the survey are highly qualified, firms’ answers pass through several

filters and double checks by this personnel. 6 Berger and Udell (2006) consider six different transaction-based lending technologies: (i) financial

statement lending, (ii) small business credit scoring, (iii) asset-based lending, (iv) factoring, (v) fixed-asset

lending, and (vi) leasing, together with relationship lending.

8

with a strong financial condition based on an assessment of verified (i.e., audited)

financial statements. From the list of fifteen criteria shown in the Appendix, we use the

initial four. These factors (financial solidity, profitability, growth of sales and ability of

the firm to repay its debt) represent qualities that are best assessed by an analysis of the

firms’ audited financial statements. From these four factors we created the financial

statement lending index, LT_FS, by calculating the average of the four dummy variables

which take a value of one if the firm answered 1 (very much) to the four relevant lending

factors, respectively. The virtue of using an average index is that it can be directly

compared with the other (averaged) indices, as we explain below, since all the indices are

constructed from dummy variables and thus take a value in the [0,1] range.7 Next, we

focus on fixed-asset lending. Fixed-asset lending technologies involve lending against

assets that are long-lived and are not sold in the normal course of business (e.g.,

equipment, motor vehicles, or real estate). The factors that are related to fixed-asset

lending are items 5, 6 and 8. We make a clear difference between real estate lending and

other fixed-asset lending, and so we construct two indices. The first, LT_RE, is a dummy

variable that takes the value one if the firm answered 1 (very much) to lending factor no.

5. Second, LT_OF, is an average of the two dummy variables which take a value of one

if the firm answered 1 (very much) to lending factors no. 6 and 8, respectively.8 Indeed,

as a robustness check, we aggregate the three transactional lending technologies in a

single index (we label it LT_TRANS).9 We take the three transactional lending

technologies as well as the aggregated index as the endogenous variable.

Finally, as the key explanatory variable, we consider the relationship lending technology.

Under relationship lending, the financial institution relies primarily on soft information

gathered through contact over time with the SME, its owner and the local community to

address the opacity problem. We construct the relationship lending index, LT_RL, using

the factors that seem most related to soft information accumulation by banks through

close relationships. The index is an average of six dummy variables which take a value of

one if the firm answered 1 (very much) to lending factors, no. 9, 10, 11, 13, 14 and 15,

respectively.

These indices are not likely to be perfect proxies for the use of different lending

technologies, since they are based on the firms’ perception of the lending factors used by

the bank in underwriting its loans, and thus may not be precisely capturing the banks’

screening process. However, constructing these indices using the firms' perspectives has

some advantages. In fact, previous researches on SME finance suffer from the problem

that the lending technologies are usually not identified (Kano et al., 2011). Our data allow

us to perceive the actual features of the bank at the time the firm is asked. Thus, we can

distinguish between lending technologies. Such information was not available in the prior

literature.

3.3 Credit rationing and soft information

7 We also conducted preliminary analysis using the first principal component of the principal component

analysis over the dummy variables. The results, available upon request, are qualitatively similar. 8 Note that the basic technology used in real estate lending and other fixed-asset lending is the same, and

the distinction is solely based on the type of collateral. 9 This index is an average of seven dummy variables which take a value of one if the firm answered 1 (very

much) to lending factors, no. 1, 2, 3, 4, 5, 6 and 8, respectively. In Table 3 we report the partial correlation

between LT_TRANS index and LT_RL index.

9

In the second part of the empirical analysis we address the role of soft information on the

probability that a firm is credit-rationed. To define our indicator of credit rationing, we

use the firms’ answers to three questions of the survey that are not necessarily related to

the main bank. The questions are:

1. In 2006 would your firm have wished a larger amount of loans at the prevailing

interest rate agreed with the bank?

2. In 2006, did the firm demand more credit than it actually obtained?

3. To obtain more credit, were you willing to pay a higher interest rate?

The variable of credit rationing is a dummy variable taking value one if the firm answers

yes to the first question and to at least one of the other two, zero otherwise. In order to

construct a proxy variable for the production of soft information we consider a

methodology similar to that used in Scott (2004) and Uchida et al. (2012). We use the

question of the survey: “Which characteristics are key in selecting your main bank?”. In

answering this question the firm was required to give a value, with descending order of

importance, from 1 to 4 to fourteen factors (see the Appendix for the details on this

question). We focus on the following characteristics:

a. The bank knows you and your business.

b. Frequent contacts with the credit officer at the bank.

The variable Soft is a dummy variable that takes value one if the firm chose the highest

value for both the above characteristics a and b, zero otherwise.

3.4 Control variables

In this section, we discuss the other variables included in the regressions. We first

classify banks into two types: large banks and local banks. The variable for large banks,

is a dummy that takes value one if the main bank is a national bank or a foreign bank,

whereas the variable for local banks, is a dummy taking value one if the main bank is

smaller-sized mutual bank, larger-sized cooperative bank, a saving bank or “other type of

bank”. We use different variables to represent firm characteristics as controls: the age of

the firm; the logarithm of the number of employees, as a proxy for size; a dummy

variable that takes value one if the firm is a corporation, zero otherwise; and the degree of

financial leverage, given by the ratio of total loans to the sum of the total loans and the

firm’s assets. We control also for the firm’s geographic localization, defining two

dummies for whether a firm is located in the Center or in the South of Italy, and its sector

based on a two-digit ATECO sectors. Moreover, we insert some variables describing the

characteristics of the local economic environment: the GDP per capita in the province in

2004, the length of the first-degree trial by the courts located in the province in 2004 and

the provincial level of social capital as measured in Guiso et al. (2004). Finally, we insert

some variables to control for the structural characteristics of the banking sector, such as

the average value of the Herfindhal-Hirschman index of concentration on bank loans in

the province during 1991-2004 period and the average number of branches per thousands

inhabitants in the province during 1991-2004 period in the province.

10

4. Findings

4.1 Complementarity between lending technologies

Comparing across the various lending technology indices, we first ascertain which one is

more widespread. We analyze the degree of complementarity between the various

technologies, by looking at the correlation among the indices as well as via multivariate

regressions of the indices. First, we consider the relative importance of each lending

technology individually, by directly comparing the magnitude of each index. Table 2

shows the summary statistics of these variables. The lending factors related to financial

statement technology are relatively more frequently emphasized, in fact the index of

financial statement technology is the largest among the four indices. This result shows

that financial statement is the most widespread lending technology. Moreover, this

finding is robust also whether we check for bank type. The level of LT_FS is 0.193 for

the firms whose main bank is a national bank, 0.092 for those that interact with a local

bank. The relationship lending index is the second largest index (0.053 for local banks

and 0.087 for national banks), followed by the real estate lending index and the other

fixed-asset lending. However, the presence of a ranking in the use of lending technologies

does not rule out the possibility of complementarity. In fact, it can be reasonably argued

that different technologies require screening and monitoring processes that are similar in

nature and intensity, so that these may be used simultaneously instead of being strictly

distinct from each other.

In order to examine in a descriptive way the issue of complementarity, we construct two

dummy variables equal to one if the main bank uses as primary lending technology either

the relationship or the transactional lending technology, zero otherwise. We construct a

third dummy variable equal to one if the bank exploits both lending technologies as

primary lending technology.10

The summary statistics confirm that complementarity

among lending technologies is more frequent than specialization in one specific

technology. This result is robust also when we distinguish between local and national

banks, although complementarity is relatively more frequent for national banks compared

to local banks.

We study more deeply the phenomenon of complementarity by estimating regressions

among the indices. In particular, we examine four different specifications, considering as

dependent variable the three transaction-based lending technologies and an aggregate

index. Table 4 reports the results. The main explanatory variable is the bank’s use of

relationship lending. In all the specifications, the coefficient for relationship lending is

positive and significant at less than the 1% level. Specifically, taking into account the

magnitude of the coefficients, relationship lending results more closely tied with financial

statement lending (0.72) than with the two fixed-asset lending technologies (real estate

and other fixed-asset, both with a coefficient of about 0.59). In column 4, we consider the

10

In particular, the dummy variable for the use of relationship lending as primary lending technology is

equal to one if the index LT_RL is larger than the 75% percentile of the distribution, while the index

LT_TRANS is lower than 75%. The dummy variable for transactional lending as primary lending

technology is defined with similar thresholds. Finally, the dummy variable for multiple primary lending

technologies is equal to one if both LT_RL and LT_TRANS are larger than 75% percentile. As a

robustness check, we use also other thresholds for the construction of these indexes. The results are similar.

11

transaction-based lending technologies in aggregate. The value of the coefficient of this

index is 0.66. The estimates show complementarity between relationship lending and

transaction-based lending considered in aggregate and in each single component. In other

words, the findings suggest that different technologies could complement each other and

loans could contain characteristics of different technologies at the same time. These

results also provide information about what combination of lending technologies is the

most important. According to the magnitude of the estimated coefficients, financial

statement lending is relatively more correlated with relationship lending. Furthermore,

taking into account the mutual interrelationship among the four lending technologies, the

two fixed-asset lending technology and the relationship lending technology are less used

at the same time.

Finally, in the analysis of the other determinants of the lending technology choice, we

find that local banks use less the transactional lending technologies. In fact, if the main

bank is local the probability of using the transactional technologies decreases and the

effect is strongest for the financial statement technology.

Regarding the other control variables considered, we find that the judicial inefficiency

decreases the probability of using real estate technology and the other-fixed asset

technology. The rationale for this result is that the judicial inefficiency, reducing the

expected return from the fixed assets pledged as collateral, decreases the incentive of

using these technologies. By contrast, a higher level of social capital is associated with a

larger use of financial statement technology.

4.2 Robustness checks

To check that our findings are robust, we report the results from running the regressions

on sub-samples of observations. First, we split the sample based on the type of the main

bank of the firm. Our aim is to investigate whether the contemporaneous use of

transaction-based lending and relationship lending is a phenomenon that characterizes

smaller-sized/territorial banks more than large banks, as suggested by the literature (see,

e.g., Berger and Black, 2011). For the sake of simplicity we consider only the

complementarity between transaction-based lending technologies taken in aggregate and

relationship lending. Results are reported in Table 5 (column 1-2). The estimated coefficients are positive and significant at less than 1% level, either when the firm’s main

bank is a large bank (coefficient equal to 0.64), or when it is a local one (coefficient equal

to 0.70).

In columns 3-6 we report the estimates for the regression on the determinants of

transactional lending technology for several sub-samples based on firms’ characteristics.

First, we distinguish between small and large firms, based on the number of employees

(columns 3 and 4). The impact of relationship lending on transactional lending is

significant for both firms with less and more than 30 employees (the median number of

employees in our sample). The coefficients are qualitatively similar. These results remain

virtually unchanged if we split the sample at the median value of sales. Finally, when we

split the sample according to the age of the firms (columns 5 and 6), we find that

relationship lending is significant in explaining transactional lending technologies for all

the firms independently of their age. These findings suggest that the importance of the

complementarity of lending technologies is not affected by firm’s characteristics (size

and age) and type of main bank (local versus large).

12

4.3 The role of soft information

In the previous sections, we find that there is pervasive complementarity among lending

technologies. An interesting issue raised by our finding lies in studying how soft

information enters in this picture. For this reason, in this section we investigate the

impact of soft information on the probability that a firm is credit rationed and how the

lending technology used by the firm’s main bank impacts on the soft information

production.11

Table 6 reports the results. In particular, columns 1 and 2 show regression

results and marginal effects for the probability of being credit rationed.12

As we expected,

the findings show that soft information has a negative and significant impact on credit

rationing in all regressions. The marginal effect of soft information is -0.19, it is

significant at the 5% level and economically sizable. This finding shows that the

production of soft information can reduce credit rationing. This is consistent with results

in the previous literature that show the positive effects of soft information for borrowers

(e.g. Uchida et al., 2012).

In columns 3 and 4, we consider the cases of individual and multiple primary lending

technologies - i.e. specialization in one lending technology versus complementarity with

specialization on both lending technologies - and examine their impacts on the production

of soft information. In order to perform this analysis we consider again the dummy

variables used in the descriptive analysis reported in section 4.1.

Following the literature (see, e.g., Petersen, 2004), if hardening of soft information is

feasible we would expect that transactional lending technology has a statistically

significant impact on the production of soft information. Our results show that

relationship lending increases the production of soft information. Moreover, the

complementarity among lending technologies seems to be also useful for the bank in the

production of soft information. These findings imply that more soft information is

produced when the main bank uses a relationship lending technology as primary lending

technology individually or in combination with transactional lending technology also

taken as primary lending technology. By contrast, transactional lending as primary

technology individually is not significant. This finding implies that substitutability

between relationship and transactional lending technologies may not be feasible by means

of hardening of soft information.

5. Conclusions

In this paper we have investigated the firm-main bank relationship by using a large

sample of Italian manufacturing firms, featuring a pervasive presence of small and

11

As explained in Section 3.3, the measures of credit rationing are not necessarily related to the main bank

(i.e., it could be that firms perceived that they are credit rationed by other banks). Therefore, the results

need to be interpreted with some caveats. Unfortunately, the low number of observation used in this

regression does not give us the possibility to better investigate this relation. 12

In order to tackle for potential problems of small sample bias we performed Montecarlo bootstrapping.

Results showed to be stable with 100 replications. In addition to the control variables used in the first part

of the analysis, we consider the length of the firm-main bank relationship and the firm’s profitability,

measured by the average value of the firm’s return on assets in the 2004-2006 period.

13

medium-sized enterprises. We start considering a recent strand of literature stressing that

banks want to serve SMEs and find this segment profitable, especially as margins in other

banking markets narrow due to intensified competition (De la Torre et al., 2010). This

literature finds that, partly thanks to the enormous progress in information and

communication technologies, even large and foreign banks (normally arm’s-length

lenders) may now be capable of lending to SMEs referring also to firm’s soft information.

This could imply that substitutability between relationship and transactional lending

technologies may to some extent be possible for outsiders by means of hardening of soft

information. Another possibility, not explored in the literature, could be that the

alternative lending technologies can indeed be complementary, i.e. used in combination

for the same firm. In this paper we tried to address these issues.

Our results highlight that a given firm may receive credit via different lending

technologies, hence supporting the hypothesis of complementarity among lending

technologies. We show that this form of complementarity is found not only at large banks

but also at smaller-sized ones. Moreover, after showing that soft information lowers the

probability of credit rationing, we find that the production of soft information depends

crucially on whether the firm’s main bank uses mainly relationship lending technology

alone or combined with transaction-based lending technology, taken both as a primary

lending technologies. On the contrary, when the main bank uses mainly transaction-based

lending technology there is no production of soft information.

Thus, it appears that the way soft information becomes embodied in the lending decision

might still differ between relational and transactional banks/technologies. These findings

suggest that complementarity among lending technologies, pursued by organizational

measures aimed at increasing the degree of delegation and lowering the turnover of

branch manager, might be more effective for the loan decision process rather than new

soft information communication techniques.

14

References

Agarwal, S., Hauswald, R., (2010), “Authority and information”. Discussion paper,

SSRN #1566062.

Albareto, G., Benvenuti, M., Mocetti, S., Pagnini, M., Rossi, P. (2008),

“L’organizzazione dell’attività creditizia e l’utilizzo di tecniche di scoring nel

sistema bancario italiano: risultati di un’indagine campionaria”, Occasional Papers,

No.12, Bank of Italy.

Angelini, P., Di Salvo, R., Ferri, G. (1998), “Availability and cost of credit for small

businesses: customer relationships and credit cooperatives”, Journal of Banking and

Finance, Vol. 22, 925-54.

Bartoli, F., Ferri, G., Murro, P., Rotondi, Z. (2013), “Bank-firm relations and the role of

Mutual Guarantee Institutions at the peak of the crisis”, Journal of Financial

Stability, Vol. 9, 90-104.

Benvenuti M., Casolaro L., Del Prete S., Mistrulli P.E. (2010), Loan officer authority and

small business lending: evidence from a survey. In: Cannari L., Pagnini M., Rossi

P. (eds), Banks, local credit markets and credit supply, Bank of Italy.

Berger, A. N., Black L.K., (2011) “Bank size, lending technologies, and small business

finance”, Journal of Banking and Finance, Vol. 35, 724–735.

Berger, A. N., Frame, W.S., Miller, N.H., (2005), “Credit Scoring and the Availability,

Price, and Risk of Small Business Credit”, Journal of Money, Credit and Banking,

Vol. 37, 191-222.

Berger, A.N., Frame W.S., (2007), “Small business credit scoring and credit availability”,

Journal of Small Business Management, Vol. 46, 5-22.

Berger, A.N., Klapper, L., Udell, G., (2001), “The ability of banks to lend to

informationally opaque small businesses”, Journal of Banking and Finance, Vol.

25, 2127–2167.

Berger, A.N., Miller, N.H., Petersen, M.A., Rajan, R.G., Stein, J.C., (2005), “Does

function follow organizational form? Evidence from the lending practices of large

and small banks”, Journal of Financial Economics, Vol. 76, 237–269.

Berger, A.N., Udell, G.F. (2006),“A more complete conceptual framework for SME

finance”, Journal of Banking and Finance, Vol. 30, 2945–2966.

Cole, R.A. (1998), “The Importance of Relationship to the Availability of Credit”,

Journal of Banking and Finance, Vol. 22, 959-977.

Columba, F., Gambacorta, L., Mistrulli, P.E. (2010), “Mutual guarantee institutions and

small business finance”, Journal of Financial Stability, Vol. 6, 45–54.

Degryse, H., Van Cayseele, P.J.G. (2000), “Relationship lending within a bank-based

system: evidence from European small business data”, Journal of Financial

Intermediation, Vol. 9, 90-109.

De la Torre, A., Martinez Peria, M.S., Schmukler, S.L., (2010), “Bank involvement with

SMEs: Beyond relationship lending”, Journal of Banking and Finance, Vol. 34,

2280-2293.

Frame, W. S., Srinivasan, A., Woosley, L., (2001), “The Effect of Credit Scoring on

Small-Business Lending”, Journal of Money, Credit and Banking, Vol. 33, 813-

825.

15

Guiso, L., Sapienza, P., Zingales, L. (2004), “The Role of Social Capital in Financial

Development”, The American Economic Review, Vol. 94, 526-556.

Kano, M., Uchida, H., Udell, G. F., Watanabe, W., (2011). “Information verifiability,

bank organization, bank competition and bank–borrower relationships”, Journal of

Banking and Finance, Vol. 35, 935-954.

Liberti, J., Mian, A., (2009), “Estimating the effect of hierarchies on information use”,

Review of Financial Studies, Vol. 22, 4057–4090.

Mocetti, S., Pagnini, M., Sette, E., (2010). “Information technology and banking

organization”. In: Cannari L., Pagnini M., Rossi P. (eds), Banks, local credit

markets and credit supply, Bank of Italy.

Pavitt, K., (1984), “Sectoral patterns of technical change: Towards a taxonomy and a

theory”, Research Policy, Vol. 13, 343-373.

Petersen, M. (2004), “Information: Hard and Soft”, mimeo, Northwestern University.

Rajan R.G. (1992), “Insiders and Outsiders: The Choice between Informed and Arm's-

length debt”, Journal of Finance, Vol 47, 1367-1400.

Scott, J.A. (2004), “Small business and the value of community financial institutions”,

Journal of Financial Services Research, Vol. 25, 207-230.

Stein, J. C. (2002), “Information production and capital allocation: Decentralized vs.

hierarchical firms”, Journal of Finance, Vol. 57, 1891-1921.

Uchida, H., Udell, G. F., Yamori, N., (2006), “SME financing and the choice of lending

technology”, RIETI Discussion Paper Series 06-E-025, Research Institute of

Economy, Trade, and Industry.

Uchida, H., Udell, G. F., Yamori, N., (2008), “How do Japanese banks discipline small-

and medium-sized borrowers? An investigation of the deployment of lending

technologies”. In: Choi J., Dow S. (eds.) Institutional Approach to Global

Corporate Governance: Business Systems and Beyond, Emerald Group Publishing

Limited.

Uchida, H., Udell, G. F., Yamori, N., (2012), “Loan officers and relationship lending to

SMEs”, Journal of Financial Intermediation, Vol. 21, 97-122.

Udell, G.F. (2004), Asset-Based Finance, The Commercial Finance Association, New

York.

Vos, E., Yeh, A. J., Carter, S., Tagg, S. (2007), “The happy story of small business

financing”, Journal of Banking and Finance, Vol. 31, 2648–2672.

16

Appendix

Table 1. Variables definition and sources

Variable Definition and source (in parentheses)

LT_FS

Index for financial statement lending technology. We use a question available in the

Survey: “In your view, which criteria does your bank follow in granting loans to

you?”. In answering this question the firm was required to give a weight (going, in

descending order, from 1, very much, to 4, nil) to fifteen factors. From the list of

fifteen criteria shown in the Appendix, we use the initial four. For each of the four

characteristics we constructed a dummy variable, which takes a value of one if the

firm chose 1. LT_FS is the average of these four dummy variables. (Survey on

Italian Manufacturing Firms)

LT_RE

Index for real-estate lending technology. We use a question available in the Survey:

“In your view, which criteria does your bank follow in granting loans to you?”. In

answering this question the firm was required to give a weight (going, in descending

order, from 1, very much, to 4, nil) to fifteen factors. LT_RE, is a dummy variable

that take the value one if the firm answered 1 (very much) to lending factor no. 5.

(Survey on Italian Manufacturing Firms)

LT_OF

Index for other fixed-asset lending technology. We use a question available in the

Survey: “In your view, which criteria does your bank follow in granting loans to

you?”. In answering this question the firm was required to give a weight (going, in

descending order, from 1, very much, to 4, nil) to fifteen factors. LT_OF, is an

average of the two dummy variables which take a value of one if the firm answered 1

(very much) to lending factors no. 6 and 8, respectively. (Survey on Italian

Manufacturing Firms)

LT_TRANS

Index for the transactional lending technology. To construct this index, we aggregate

the three transactional lending technologies (LT_FS, LT_RE and LT_OF) in a single

index. (Survey on Italian Manufacturing Firms)

Credit Rationed

Dummy taking a value of one if the firm answers yes to the question “In 2006 would

your firm have wished a larger amount of loans at the prevailing interest rate agreed

with the bank?”, and yes to at least one of the following two questions: “In 2006, did

the firm demand more credit than it actually obtained?” or “To obtain more credit,

were you willing to pay a higher interest rate?”. (Survey on Italian Manufacturing

Firms)

Soft

We use the following question of the Survey: “Which characteristics are key in

selecting your main bank?”. In answering this question the firm was required to give

a value, with descending order of importance, from 1 to 4 to the two following

characteristics (among others): “The bank knows you and your business” and

“Frequent contacts with the credit officer at the bank”. The variable Soft is a dummy

that takes value one if the firm chose the highest value for both the above two

characteristics. (Survey on Italian Manufacturing Firms)

LT_RL

Index for relationship lending technology. We use a question available in the Survey:

“In your view, which criteria does your bank follow in granting loans to you?”. In

answering this question the firm was required to give a weight (going, in descending

order, from 1, very much, to 4, nil) to fifteen factors. LT_RL, is an average of six

dummy variables which take a value of one if the firm answered 1 (very much) to

lending factors, no. 9, 10, 11, 13, 14 and 15, respectively. (Survey on Italian

Manufacturing Firms)

National bank Dummy variable taking value one if the main bank is either a national bank or a

foreign bank; 0 otherwise. (Survey on Italian Manufacturing Firms)

Local bank

Dummy variable taking value one if the main bank is a smaller-sized cooperative

mutual banks, a larger-sized Volksbank type cooperative banks, a saving bank or

“other type of bank”; 0 otherwise. (Survey on Italian Manufacturing Firms)

17

Number of Banking relations Log of the total number of firm’s reference banks. (Survey on Italian Manufacturing

Firms)

Corporation Dummy variable taking value one if firm is a public limited company; 0 otherwise.

(Survey on Italian Manufacturing Firms)

Leverage Ratio of firm’s total loans to the sum of firm’s total loans and firm's equity as of the

end of December 2006. (Survey on Italian Manufacturing Firms)

Size Log of the firm’s number of employees as of the end of December 2006. (Survey on

Italian Manufacturing Firms)

Age Log of the age of firm since foundation, in years. (Survey on Italian Manufacturing

Firms)

HHI Average value of the Herfindhal Hirschman index of concentration on bank loans in

the province during 1991-2004 period. (Statistical Bulletin of the Bank of Italy)

Provincial GDP Log of the value of the GDP in the province as of the end of December 2004.

(ISTAT)

Judicial inefficency Log of the length of the first-degree trial by the courts located in the province in

2004. (ISTAT)

Social Capital

Voter turnout at the province level for all the referenda before 1989. These include

data referenda on the period between 1946 and 1987. For each province turnout data

were averaged across time. (Guiso, Sapienza and Zingales, 2004)

Branches in the province Average number of branches per thousands inhabitants in the province during 1991-

2004 period. (Statistical Bulletin of the Bank of Italy)

Center

Dummy variable taking value 1 if the bank branch where the credit relationship with

the firm takes place is located in Central Italy; 0 otherwise. (Survey on Italian

Manufacturing Firms)

South

Dummy variable taking value 1 if the bank branch where the credit relationship with

the firm takes place is located in Southern Italy; 0 otherwise. (Survey on Italian

Manufacturing Firms)

ROA Average value of the ratio of firm’s EBIT to firm’s total assets during 2004-2006

period. (Survey on Italian Manufacturing Firms)

Relationship Length Log of the length of the firm-main bank relationship. (Survey on Italian

Manufacturing Firms)

Mainly relationship lending

Dummy variable taking value 1 if the index LT_RL is larger than the 75% percentile

of the distribution, while the index LT_TRANS is lower than 75%. (Survey on

Italian Manufacturing Firms)

Mainly transactional lending

Dummy variable taking value 1 if the index LT_TRANS is larger than the 75%

percentile of the distribution, while the index LT_RL is lower than 75%. (Survey on

Italian Manufacturing Firms)

Complementarity

Dummy variable taking value 1 if both the indexes LT_RL and LT_TRANS are

larger than 75% percentile of their distribution. (Survey on Italian Manufacturing

Firms)

18

Table 2. Summary statistics

Full Sample Local banks National banks

Mean St. Dev Mean St. Dev Mean St. Dev

LT_FS 0.216 0.332 0.092 0.239 0.193 0.31

LT_RE 0.119 0.324 0.034 0.181 0.093 0.291

LT_OF 0.113 0.259 0.035 0.157 0.090 0.237

LT_TRANS 0.173 0.274 0.068 0.18 0.149 0.237

LT_RL 0.145 0.284 0.053 0.180 0.087 0.189

Soft Information 0.097 0.297 0.025 0.157 0.072 0.259

Credit Rationed 0.426 0.496 0.435 0.498 0.420 0.495

Local bank 0.503 0.500 1.000 0.000 0.000 0.000

National bank 0.497 0.500 0.000 0.000 1.000 0.000

Number of banks 4.973 3.959 5.482 4.602 5.985 4.416

Number of banks (ln) 1.383 0.668 1.487 0.617 1.592 0.611

Corporation 0.332 0.471 0.390 0.488 0.381 0.486

Leverage 0.899 0.113 0.891 0.114 0.892 0.109

Size 3.553 1.118 3.757 1.279 3.797 1.370

Age 22.663 14.388 22.746 15.020 23.625 15.898

HHI 0.111 0.048 0.112 0.047 0.116 0.050

Audit 0.376 0.485 0.214 0.410 0.461 0.499

Provincial GDP 10.192 0.219 10.197 0.203 10.179 0.218

Judicial inefficency 5.893 0.276 5.883 0.254 5.916 0.272

Social Capital 0.845 0.055 0.851 0.050 0.846 0.058

Branches 0.530 0.125 0.544 0.119 0.522 0.121

North 0.719 0.449 0.736 0.441 0.674 0.469

Center 0.162 0.369 0.170 0.376 0.197 0.398

South 0.118 0.323 0.094 0.293 0.129 0.336

ROA 0.025 2.097 0.052 0.079 0.052 0.058

Relationship length 2.595 0.782 2.691 0.738 2.559 0.758

Mainly relationship lending 0.051 0.221 0.044 0.206 0.064 0.244

Mainly transactional lending 0.074 0.262 0.055 0.228 0.119 0.324

Complementarity 0.232 0.422 0.072 0.259 0.172 0.378

19

Table 3. Correlation matrix

LT

TRANS LT_RL

Credit

Rationed Soft Local bank

Number of

banks Length ROA Age Size Center South Lever

LT_TRANS 1.000

Credit Rationed -0.009 1.000

Soft 0.175* -0.118 1.000

LT_RL 0.774* 0.003 0.197* 1.000

Local -0.191* 0.015 -0.108* -0.092* 1.000

Number of Banks 0.014 0.037 0.051* 0.030 -0.056 1.000

Leverage 0.013 0.095 -0.033 -0.024 -0.004 -0.077* 1.000

Size 0.038 -0.019 0.065* 0.090* -0.015 0.394* -0.258* 1.000

Age -0.006 -0.081 0.084* 0.017 -0.028 0.174* -0.183* 0.225* 1.000

Relationship length -0.025 -0.119 0.010 0.009 0.088* 0.077* -0.080* 0.108* 0.333* 1.000

Mainly rel. lend. -0.089* -0.027 0.033 0.154* -0.043 0.005 0.053* -0.064* -0.012 -0.017 1.000

Mainly trans. lend. 0.264* 0.042 -0.001 -0.145* -0.113* -0.034 0.057* -0.0754* 0.005 -0.027 -0.066* 1.000

Complementarity 0.795* -0.007 0.172* 0.790* -0.153* 0.022 -0.027 0.056* -0.031 -0.022 -0.128* -0.155* 1.000

Note: The table shows the pairwise correlation coefficients. (*) correlation coefficients significant at the 5% level or better.

20

Table 4. The determinants of lending technologies

(1) (2) (3) (4)

Variables LT_FS LT_RE LT_OF LT_TRANS

LT_RL 0.717*** 0.587*** 0.598*** 0.664***

0.056 0.084 0.068 0.049

Local bank -0.083*** -0.043*** -0.036*** -0.064***

0.018 0.016 0.012 0.013

Number of banks -0.013 0.010 0.005 -0.005

0.018 0.017 0.012 0.013

Corporation 0.043* 0.022 0.008 0.030*

0.023 0.021 0.015 0.016

Leverage 0.101 0.028 0.011 0.064

0.078 0.080 0.060 0.054

Size -0.016* -0.010 0.000 -0.011*

0.009 0.009 0.006 0.006

Age -0.001 -0.001 -0.000 -0.001

0.001 0.001 0.001 0.001

HHI 0.002 -0.011 0.004 0.001

0.233 0.166 0.146 0.151

GDP 0.022 -0.071 -0.088* -0.023

0.078 0.061 0.047 0.050

Judicial inefficiency -0.005 -0.163*** -0.113*** -0.059

0.066 0.057 0.043 0.046

Social Capital 0.636** 0.053 -0.098 0.343

0.322 0.257 0.212 0.221

Branches in the province -0.034 -0.168 -0.061 -0.061

0.107 0.102 0.078 0.080

Center -0.004 0.003 -0.004 -0.003

0.028 0.024 0.016 0.018

South 0.080 0.031 -0.031 0.041

0.069 0.060 0.035 0.044

Constant -0.601 1.776** 1.700*** 0.396

0.881 0.770 0.621 0.607

Observations 816 816 816 816

R-squared 0.272 0.229 0.337 0.381

Note: The table reports regressions coefficients. The dependent variables are the three transactional

lending technologies taken also in aggregate. For the definition of the explanatory variables see Table 1.

The regressions are estimated by OLS. The regressions include sector dummies. Robust standard errors are

reported below coefficients. (*): coefficient significant at 10% confidence level; (**): coefficient significant

at 5% confidence level; (***): coefficient significant at less than 1% confidence level. The table also

reports, as goodness-of-fit tests, the R-squared.

21

Table 5. Robustness checks

(1) (2) (3) (4) (5) (6)

Variables LT_TRANS LT_TRANS LT_TRANS LT_TRANS LT_TRANS LT_TRANS

Local Banks National Banks Small Firms Large Firms Age < 21 years Age ≥ 21 years

LT_RL 0.699*** 0.642*** 0.648*** 0.676*** 0.707*** 0.616***

0.063 0.068 0.074 0.072 0.071 0.081

Local bank -0.074*** -0.060*** -0.051** -0.080***

0.023 0.018 0.021 0.020

Number of banks -0.002 -0.016 0.009 -0.010 0.011 -0.023

0.010 0.024 0.023 0.013 0.016 0.017

Corporation 0.021 0.040 0.006 0.034* 0.025 0.030

0.014 0.028 0.037 0.018 0.021 0.022

Leverage 0.018 0.114 0.035 0.055 -0.042 0.126**

0.052 0.105 0.083 0.066 0.092 0.061

Size -0.010* -0.009 -0.016 -0.003 -0.014 -0.007

0.006 0.012 0.018 0.009 0.008 0.009

Age -0.001** -0.000 -0.000 -0.001 -0.001 -0.000

0.000 0.001 0.001 0.001 0.001 0.001

HHI 0.190 -0.137 0.036 -0.112 0.331* -0.336**

0.196 0.213 0.259 0.151 0.194 0.157

GDP -0.014 -0.027 -0.013 -0.061 -0.005 -0.047

0.055 0.069 0.065 0.058 0.073 0.059

Judicial inefficiency -0.019 -0.089 -0.056 -0.063 0.004 -0.107*

0.039 0.077 0.054 0.057 0.063 0.064

Social Capital 0.259 0.250 0.229 0.480* 0.336 0.384

0.231 0.303 0.338 0.242 0.279 0.264

Branches in the

province 0.018 -0.119 0.021 -0.096 -0.062 -0.100

0.066 0.114 0.104 0.086 0.120 0.090

Center 0.014 -0.017 0.001 -0.001 0.014 -0.021

0.018 0.037 0.029 0.028 0.027 0.026

South 0.044 0.006 0.039 0.043 0.002 0.055

0.054 0.082 0.065 0.052 0.054 0.057

Constant 0.049 0.711 0.358 0.711 -0.102 0.930

0.587 0.850 0.801 0.707 0.924 0.736

Observations 411 405 344 472 398 418

R-squared 0.519 0.287 0.347 0.419 0.422 0.366

Note: The table reports regressions coefficients. The dependent variable is the transactional lending technologies taken in

aggregate. For the definition of the explanatory variables see Table 1. The regressions are estimated by OLS. The

regressions include sector dummies. Robust standard errors are reported below coefficients. (*): coefficient significant at

10% confidence level; (**): coefficient significant at 5% confidence level; (***): coefficient significant at less than 1%

confidence level. The table also reports, as goodness-of-fit tests, the R-squared.

22

Table 6. The role and production of soft information

(1) (2) (3) (4)

VARIABLES Credit Rationed Credit Rationed Soft information Soft information Coefficients Marginal Effects Coefficients Marginal Effects

Soft -0.530* -0.193**

0.277 0.093

Mainly relationship lending 0.750*** 0.153**

0.239 0.067

Mainly transactional lending 0.288 0.044

0.190 0.034

Complementarity 0.803*** 0.144***

0.108 0.026

Relationship length -0.190 -0.075 -0.010 -0.001

0.145 0.057 0.091 0.012

Number of banks 0.174 0.068 0.206 0.026

0.185 0.072 0.153 0.019

Leverage 1.398 0.549 -0.669 -0.086

1.227 0.482 0.464 0.062

ROA -3.686** -1.447** -0.793 -0.102

1.691 0.663 0.505 0.065

Size 0.085 0.033 0.035 0.005

0.101 0.039 0.054 0.007

Age -0.002 -0.001 0.007 0.001

0.007 0.003 0.006 0.001

HHI -1.355 -0.532 -1.116 -0.143

1.893 0.744 1.432 0.185

Provincial GDP 1.407** 0.552** 0.333 0.043

0.647 0.252 0.485 0.061

Branches in the province -1.164 -0.457 -1.277 -0.164*

1.103 0.432 0.776 0.096

Center 0.416* 0.165* 0.037 0.005

0.221 0.087 0.155 0.021

South 0.599 0.235 -0.386 -0.039

0.430 0.162 0.372 0.030

Constant -15.040** -4.191

6.596 5.152

Observations 186 186 1,112 1,112

Pseudo R-squared 0.123 0.123 0.111 0.111

Average predicted probability 0.428 0.066

Note: The table reports regressions coefficient and marginal effects. The dependent variables are the dummy

of credit rationing and the proxy for the production of soft information. For the definition of the explanatory

variables see Table 1. The regressions are estimated with Probit. The regressions include sector dummies.

Robust standard errors are reported below coefficients. (*): coefficient significant at 10% confidence level;

(**): coefficient significant at 5% confidence level; (***): coefficient significant at less than 1% confidence

level. The table also reports the Pseudo R-squared as goodness-of-fit tests and the average predicted

probability.

23

Survey questions

F1.15: Which of these characteristics are key in selecting your main bank?

1. The bank knows you and your business.

2. The bank knows a member of your Board of directors or the owners of the firm.

3. The bank knows your sector.

4. The bank knows your local economy.

5. The bank knows your relevant market.

6. Frequent contacts with the credit officer at the bank.

7. The bank takes quick decisions.

8. The bank offers a large variety of services.

9. The bank offers an extensive international network.

10. The bank offers efficient internet-based services.

11. The bank offers stable funding.

12. The bank offers funding and services at low cost.

13. The bank’s criteria to grant credit are clear.

14. The bank is conveniently located.

F1.17: In your view, which criteria does your bank follow in granting loans to you?

1. Ability of the firm to repay its debt (e.g. years needed to repay its debt).

2. Financial solidity of the firm (capital/asset ratio).

3. Firm’s profitability (current profits/sales ratio).

4. Firm’s growth (growth of sales).

5. Ability of the firm to post (not personal) real estate collateral.

6. Ability of the firm to post tangible non-real estate collateral.

7. Support by a guarantee association (e.g. loan, export, R&D, etc.).

8. Personal guarantees by the firm’s manager or owner.

9. Managerial ability on the part of those running the firm’s business.

10. Strength of the firm in its market (number of customers, commercial network).

11. Intrinsic strength of the firm (e.g. ability to innovate).

12. Firm’s external evaluation or its evaluation by third parties.

13. Length of the lending relationship with the firm.

14. Loans are granted when the bank is the firm’s main bank.

15. Fiduciary bond between the firm and the credit officer at your bank.