Family
firms, soft information and bank lending in a
financial crisis
☆
Leandro D'Aurizio
a, Tommaso Oliviero
b,⁎
, Livio Romano
ca
Department of Economics and Finance, Bank of Italy, Italy bCSEF, University of Naples Federico II, Italy
cCentro Studi Confindustria, Italy
a r t i c l e i n f o
a b s t r a c t
Available online 15 January 2015 This paper studies differences in family and non-familyfirms' access to bank lending during the
2007–2009 financial crisis. The hypothesis is that the former's incentive structure results in less agency conflict in the borrower–lender relationship. Using highly detailed data on bank–firm relations, we exploit the reduction in bank lending in Italy following the crisis in October 2008. Wefind statistically and economically significant evidence that credit to family firms contracted less sharply than that to non-familyfirms. The results are robust to observable ex-ante differences between the two types offirms and to time-varying bank fixed effects. We show, further, that the difference is related to an increased role for soft information in some Italian banks' operations, following the Lehman Brothers failure. Finally, by identifying a match between those banks and familyfirms, we can control for time-varying unobserved heterogeneity among the firms and validate the hypothesis that our results are supply-driven.
© 2015 Elsevier B.V. All rights reserved.
JEL classification: D22 G21 G32 Keywords: Familyfirms Financial crisis Soft information Bank lending 1. Introduction
The globalfinancial crisis of 2008 and the subsequent global recession made it clear that capital markets can be a major source of
business cyclefluctuations1. Shocks to the banking sector are propagated to the real economy via reduced credit supply. In particular, a
heightening of problems of asymmetric information in bank–firm relationships tends to amplify the shocks, affecting some types of
borrower disproportionately (Bernanke et al., 1996). Problems of moral hazard (Holmstrom and Tirole, 1997) and adverse selection
(Stiglitz and Weiss, 1981) tend to discourage lenders from supplying credit tofirms with high agency costs. Information asymmetry is
typically less severe for banks than for bondholders; while the latter must rely chiefly on publicly available information (balance
sheets, ratings, etc.— so-called hard information), the former have access to “inside” information, which is transmitted through
re-peated interactions between the loan officer and the firm's manager (Diamond, 1989; Fama, 1985; Petersen and Rajan, 1994). Such
information relates to the lending officer's subjective evaluation of the firm's creditworthiness and is commonly labeled as soft
☆ An earlier version of this paper circulated under the title “Family Firms and the Agency Cost of Debt: The role of soft information during a crisis”. The views expressed are those of the authors and do not involve the responsibility of the Bank of Italy or Confindustria. We thank Mario Daniele Amore, Arpad Abraham, Jerome Adda, Alastair Ball, Elena Carletti, Vittoria Cerasi, Marco Cucculelli, Luigi Guiso, Anais Hamelin, Andrea Ichino, Giuseppe Ilardi, Steven Ongena, Marco Pagano, Nicola Pavoni, Jun Qian, Giancarlo Spagnolo, Hannes Wagner and an anonymous referee for their insightful comments. We also thank participants at: Journal of Corporate Finance Special Issue Conference on Family Business Renmin University Business School, June 2013; Workshop on SMEfinance University of Strasbourg, April 2013; Workshop on Italian bank lending Bank of Italy, March 2013; CSEF seminar, November 2012; EIEF seminar, October 2012; Workshop on Institutions, Individual Behavior and Economic Outcomes, September 2012; EUI Forum, May 2012. All remaining errors are our own.
⁎ Corresponding author at: CSEF, U Naples Federico II Dept. of Economics and Statistics Via Cintia, Monte S. Angelo 80126 Napoli, ITA.
E-mail addresses:leandro.daurizio@bancaditalia.it(L. D'Aurizio),tommaso.oliviero@unina.it(T. Oliviero),l.romano@confindustria.it(L. Romano). 1
SeeQuadrini (2011)for a review.
http://dx.doi.org/10.1016/j.jcorpfin.2015.01.002 0929-1199/© 2015 Elsevier B.V. All rights reserved.
Contents lists available atScienceDirect
Journal of Corporate Finance
(Berger and Udell, 2002; Petersen, 2004). Soft information is an important determinant of corporate lending, especially to small
busi-nesses (Garcia-Appendini, 2011). In addition, it has been shown that soft information mitigates the repercussions of aggregate credit
contractions (De Mitri et al., 2010; Jiangli et al., 2008). The reason is that hard information, such as past performance and standardized
risk measures, are less reliable in predictingfirm risk profiles during a crisis. Soft information about a firm's current results and future
plans, which is continuously updated and better targeted to the characteristics of the borrower, can reduce such uncertainty. Yet despite the academic interest in the importance of soft information in banks' lending decisions, it is still unclear which types of firm benefit most from an established banking relationship. We address this issue by focusing on the heterogeneity in firms' owner-ship structure, namely the presence or absence of a family block-holder. In particular, we pose an empirical question: does the
presence of a family block-holder mitigate bank–firm agency conflicts during a financial crisis? The answer is closely related to
differ-ences in the incentive structures of family and non-familyfirms, hence to the banks' potential risk-shifting problem (Jensen and
Meckling, 1976).
Burkart et al. (2003), and more recentlyBandiera et al. (2012), have observed that family block-holders attach a value to control that is not merely monetary but also comprises an amenity component, i.e. utility gained from the control per se. This amenity
com-ponent can be thought of as the personal status acquired thanks to the identification of the family name with the firm or the desire to
pass thefirm on to descendants. This translates into higher non-monetary costs of default, which reduces the incentive for strategic
default (Anderson et al., 2003). On the other hand, as is pointed out byVillalonga and Amit (2006),Ellul et al. (2009)andLins et al.
(2013), family block-holders may have more incentive to extract private benefits at the expense of the other shareholders and
stake-holders generally2. In contrast to the case of non-family block-holders, the gains from misconduct are concentrated in a single family
group.
In afinancial crisis, the lower expected return on investments can aggravate the incentive to divert resources out of the company,
reducing a familyfirm's investment in the future and decreasing the probability that it will repay its debt. On the other hand, family
firms may be perceived as more creditworthy because they have less incentive to default in the future. The evaluation of the overall
impact of family ownership on credit allocation thus depends on the relative numbers of“good” and “bad” family and non-family
firms. Therefore, even if the family status of firms is observable to all banks, only soft information gathered through personal contact
withfirms' managers can enable a loan officer to assess whether — for the same publicly available characteristics — a family firm is
more creditworthy than a non-family one. In other words, soft information supplements hard information by revealing the possibly
different objective functions of family and non-familyfirms.
We answer our empirical question on the basis of highly detailed data from the Italian Central Credit Register (CCR), which covers
all loans to non-financial firms by banks operating in Italy. These data are matched with firm-specific data, including family ownership
2All these papers focus on listedfirms, with their agency conflicts between controlling and minority shareholders. Our analysis, instead, concerns smaller firms that have generally not gone public, so this type of agency conflict is less of a concern here.
Table 1
Summary statistics for family and non-familyfirms, prior to the shock.
Non-Family Family Mean diff.
Mean St. dev. Median Mean St. dev. Median Obs.
Panel A: Firm characteristics
Foundation 1976.88 22.79 1981.00 1973.88 24.20 1979.00 3.00⁎⁎⁎ 2909 Employees (2008) 421.63 1324.20 100.00 170.06 422.20 60.00 251.60⁎⁎⁎ 2909 SMEs (%) .63 .48 1 .78 .41 1 −0.15⁎⁎⁎ 2909 North (%) .46 .50 0 .39 .49 0 0.07⁎⁎⁎ 2909 Center (%) .24 .43 0 .22 .41 0 0.02 2909 South (%) .31 .46 0 .40 .49 0 −0.09⁎⁎⁎ 2909 Roe (2007)a (%) 6.25 6.97 4.9 6.40 6.13 5.26 −0.15 2741 Leverage (2007)a (%) .44 .49 .31 .51 .51 .40 −0.07⁎⁎⁎ 2200 Cashflow/revenues (2008) .06 .12 .05 .04 .42 .05 0.02⁎ 2781 Change in sales2008− 09(%) −.14 .29 −.09 −.16 .27 −.12 0.02⁎ 2909
Panel B: Bank–firm relation
Zscore (2008) 4.50 1.82 4 4.30 1.76 4 0.20⁎⁎⁎ 2641
Bank Leverage (2007) (%) .39 .46 .27 .44 .42 .35 −0.05⁎⁎ 1710
N bank relations 6.64 5.01 5.00 7.60 5.03 6.00 −0.96⁎⁎⁎ 2848
Sharefirst bank (%) .56 .24 .51 .48 .21 .44 0.08⁎⁎⁎ 2909
Share second bank (%) .22 .11 .21 .22 .09 .22 −0.00 2763
Share third bank (%) .12 .07 .12 .13 .06 .13 −0.01⁎⁎⁎ 2535
Share fourth bank (%) .08 .05 .08 .09 .05 .09 .01⁎⁎⁎ 2253
Herfindahl index .45 .21 .30 .36 .23 .32 0.09⁎⁎⁎ 2909
SMEs are defined as having 250 employees or less and annual sales less than 50 million. Z-score takes values between 1 and 9. Extreme values were recoded at the 1st and the 99th percentiles to eliminate outliers. Leverage is measured as total debt over total assets in 2007; ROE is calculated as net profit over total equity in 2007. Num-ber of bank relations and Herfindahl index (measured in terms of loan concentration at the firm level) were measured at the end of SeptemNum-ber 2008.
⁎ p b 0.10. ⁎⁎ p b 0.05. ⁎⁎⁎ p b 0.01.
status. Here we are able to include familyfirms of different sizes, including small and medium-sized enterprises (SMEs), for which detailed information on corporate structure is not ordinarily available. We cover the period from 2007 to 2009, so that we can
com-pare results before and after the Lehman Brothers bankruptcy. The choice of October 2008 as the watershed date reflects the nature of
the shock represented by the Lehman Brothers failure. This event was exogenous and largely unexpected by Italian banks, and
prompted a decreased propensity to lend (Albertazzi and Marchetti, 2010). At the same time, the direct asset losses that hit the
bank-ing sector in many OECD economies were not a major concern for Italy (or for Japan), where banks concentrate more on traditional
credit activities (Panetta et al., 2009).
For the purpose of our analysis, Italy represents an ideal laboratory. First, bank lending is by far the most important type of debt for
both the family and the non-familyfirms in the sample (85% of total debt). Moreover, there was substantial heterogeneity in the
banks' use of soft information in the wake of the crisis. A survey conducted by the Bank of Italy in 2009 indicates that, as a result of
thefinancial crisis 35%, of the banks surveyed (representing 36% of total aggregate credit) increased the relative weight, in lending
decisions, that they attached to qualitative information and direct knowledge of the borrower.
The empirical analysis reveals that the aggregate growth rate of loans was lower, one year after the failure of Lehman Brothers, for
both family and non-familyfirms. But for family firms the contraction was about 5 percentage points less severe and the difference
was statistically significant. This result is robust to the inclusion of a rich set of observable characteristics to control for the correlation
of family ownership with otherfirm characteristics. By exploiting the presence of multiple lending relationships, we also control for
time-varying bankfixed effects. This differential effect does not depend on the nationality of the controlling shareholder, on group
affiliation, or on concentration of share ownership. We find no differences between family and non-family firms in interest rates or
the amount of collateral (see the Appendix). Thesefindings can be read to mean that, other things being equal, the existence of a
fam-ily block-holder reduced banks' expected default risk. Given that on average famfam-ilyfirms are smaller, this alternative flight-to-quality
mechanism favoring familyfirms is consistent with the findings ofPresbitero et al. (2012)who show that in the 2007–2009 financial
crisis smallerfirms in Italy suffered a less severe contraction in credit availability than larger ones.
Using the information on bank lending practices from the Bank of Italy survey, we further show that the banks that increased their
reliance on soft information accounted for the observed difference between family and non-familyfirms. Starting from this, we
esti-mate a time-varyingfirm-fixed-effect model, interacting a family firm dummy with an identifier of banks that increased such reliance
in their lending decisions. This empirical strategy controls for unobserved heterogeneity betweenfirms (e.g. demand shocks) and
enables us to validate the hypothesis that results are driven by changes in credit supply. The results suggest that those banks that
increased their reliance on soft information tended to re-allocate credit towards familyfirms.
Our results are important in two ways. First, familyfirms are common all over the world among SMEs, but also among listed
corporations (Bertrand and Schoar, 2006).3So their access tofinancial markets is potentially a significant factor in the performance
of the economy, insofar asfinancially constrained firms tend to lower investment and employment levels (Campello et al., 2010),
ag-gravating the impact of credit supply shocks.4Accordingly, in the last section of the paper we examine the extent to which differing
access to bank credit is mirrored in differences infirm performance from 2007 to 2009. We find no significant differences in capital
expenditure, but we dofind substantial differences in the employment policies of family and non-family firms, the former reducing
their workforce by 2.6 percentage points less. Finally, profitability, as measured by the ROE, declined less in family firms. Taken
togeth-er, thesefindings suggest that the credit re-allocation towards family firms had significant economic effects, and it appears to have
been efficient for the banking system. Second, the study contributes to the debate on the efficiency of relationship lending. In the
3In our representative sample of Italianfirms, family firms accounted for some 60% of firms by number and 40% by sales in 2008. 4
See alsoKahle and Stulz (2013)for a review of the empirical literature on the effects of the recentfinancial crisis. Fig. 1. Bank lending before and after the Lehman Brothers bankruptcy: overall adjustments.
last two decades, hard information has played an increasingly important role in lending practices, owing both to regulatory pressure
and to the spread of information technology in thefinancial industry. Our findings indicate that soft information can mitigate the
negative effects of an aggregate credit contraction, constituting a valuable resource for banks at times of uncertainty.
The rest of the paper is organized as follow:Section 2presents the data and gives some descriptive statistics of the samplefirms;
Section 3analyzes the dynamics in loans granted, showing how they differ between family and non-familyfirms;Section 4examines
bank–firm relationships, focusing on the interaction between soft information and family firm status;Section 5documents the
differ-ences in real outcomes between family and non-familyfirms, andSection 6concludes.
2. Data sources and descriptive statistics
In this paper we exploit information on bank–firm relationships, corporate governance, balance sheets and bank organization.
Ac-cordingly, our dataset comes from four main databases: Invind, Cerved, Central Credit Register (CCR) and a special survey of Italian
banks conducted by the Bank of Italy in 2009. Each single observation consists of afirm–quarter–bank triplet; the observations
cover the years 2007–2009.
The Invind survey is conducted yearly by the Bank of Italy (Survey of Industrial and Service Firms), on a representative sample of
Italian non-financial companies with at least 20 employees (Bank of Italy, 2011).5The survey collected information on family status of
thefirms in the three waves from 2007 to 2009.6
The“family firm” status is assigned according to the answer to the following
question:
“Is the firm controlled (directly or indirectly) by a single individual or a group of persons linked by family relationships?”7
This reliance on self-reported information can overcome the typical problem of identification of family firms, namely measuring
the stake of each shareholder in order to determine who controls thefirm (Ellul et al., 2010). Additionally, for a sub-sample consisting
in industrialfirms with at least 50 employees we can also assess controlling shareholders' stakes quantitatively. In order to recover
balance-sheet data (total assets, leverage, and ROE, etc.), we used the official data of the Italian Chambers of Commerce (Cerved
archives).
We match ourfirm-level information with the Central Credit Register (CCR) database, which has data on all loans granted by the
Italian banking system tofirms. These data enable us to construct unique variables based on each bank–firm relationship, with
quar-terly frequency. In the empirical analysis, we focus on revocable credit lines, which are relatively more homogeneous and can be
renegotiated unilaterally by banks.8The loans considered thus exclude long-term, collateralized loans. Borrowers may also have
contemporaneous relations (deposits, personal loans) with their bank that could affect the lending decision (Sapienza, 2004), and
for which we cannot control by using the credit lines.9
Finally, we integrated the abovefirm–year–bank observations with information from a special 2009 survey by the Bank of Italy's
regional branches of about 400 banks that accounted for 80% of outstanding bank credit to Italianfirms. This survey asked specifically
about changes in the respondents' use of soft information as a result of thefinancial crisis:
“Please indicate whether since October 2008, as a result of the economic and financial crisis, the importance you attach to
qualitative information and direct knowledge of the borrower has increased, decreased or remained the same”10
Excluding state-owned companies andfirms whose ownership structure could not be determined, we were left with 1808 family
and 1101 non-familyfirms. Panel A ofTable 1describes the characteristics of this sample, presenting family and non-familyfirms
sep-arately. Familyfirms were on average much smaller (in keeping with an abundant literature), slightly older, and relatively less
5This cut-off is set by the Bank of Italy in order to collect information for a representative sample offirms belonging to the industrial and service sectors: firms above this threshold account for 70.5% of the total payroll employment in industry and 59.2% in non-financial services.
6
When the information for afirm is not available in all three waves, we check the information from previous years, using Amadeus and on-line search from the company's websites. Amadeus is a European database that provides qualitative and quantitative information onfirms' ownership structure.
7
Translated from Italian. 8
The CCR database distinguishes between call loans and term loans. Call loans are those that banks can call unilaterally at any time, while for term loans they ordi-narily have to wait to the end of term for renegotiation.
9Unfortunately, this information is never observable, so all the results are subject to this caveat. 10
Translated from Italian. Table 2
Summary statistics of the change in log lending.
log(loans)09− log(loans)08 Mean St. dev. Median Observations
Aggregated at thefirm level −.08 .42 −.03 2851
At the bank–firm level −.15 1.01 0 15,212
log(loans)09− log(loans)0808 is the log difference of the average granted loans in the time windows 1 October 2007–30 September 2008 and 1 October 2008–30 September 2009. When aggregated at thefirm level, in each quarter all bank loans for each firm are summed, and then the ex-ante and ex-post averages computed. When considered at the bank–firm level, for each loan from a single bank to a single firm, the ex-ante and ex-post averages are computed. At the aggregate level, we cut the distribution at the 1st e 99th percentiles to control for outliers. At the bank–firm level we consider only those observations relative to firms analyzed at the aggregate level.
numerous in the North of Italy (and conversely more numerous in the South). They were also more indebted prior to the crisis,
suf-fered slightly more from the contraction in sales,11and generated less cashflow in relation to revenues (this last difference is weakly
significant). There was no significant difference between the two types of firm in profitability as measured by ROE.
Panel B ofTable 1provides summary statistics onfirm–bank relationships. Risk profiles were similar between the two groups, as
measured by the Z-score (the difference is statistically significant but economically negligible).12In line with other works using Italian
data (Detragiache et al., 2000; Ongena and Smith, 2000), multi-banking is common in our sample— over 87% of the firms have
rela-tions with at least three different banks. On average, familyfirms have more relationships, a result that accords with the findings of
Guiso and Minetti (2010).13This explains the differing degree of loan concentration for familyfirms, as measured by the Herfindahl
index and also by the relative share of the various banks infinancing the firm (in particular the first bank).
3. Bank lending and corporate ownership
This section investigates whether the amplitude of the contraction in bank lending tofirms was related to their ownership
structure. In addressing this empirical question, we lookfirst at the firms' overall exposure to the banking system, gauged by the
total volume of credit lines they have. We aggregate all the data from eachfirm's banking relationships into a single observation.
3.1. Graphical inspection
Fig. 1examines the bank lending channel non-parametrically by plotting the dynamics of average granted loans for family and
non-familyfirms separately. We take the mean of the logarithm of outstanding loans to family and non-family firms in each quarter
from September 2007 to September 200914and normalize the observations to zero, taking as base the end of the third quarter of 2008.
The vertical axis can then be interpreted as the rate of growth of outstanding loans relative to that quarter.
Thefigure confirms the soundness of dating the credit shock in Italy to the failure of Lehman Brothers, as the average growth of
loans outstanding turned negative in the third quarter of 2008. Thefigure also shows that there was no significant difference between
11
Controlling for sector, size, the square of the size, year of foundation and geographical area, the difference in sales contraction between family and non-familyfirms is not statistically significant.
12The Z-score is provided by the Company Accounts Data Service (Cerved) and is calculated by a complex procedure that refines Altman's original approach. The index is built on balance-sheetfigures and can take integer values from 1 to 9. Higher values imply a higher probability of default.
13They use concentrated ownership as a proxy for informational opacity and debt restructuring costs for banks in case of corporate reorganization. With both types of interpretation, more concentrated ownership predicts higher probability of multi-banking.
14
In each quarter we controlled for outliers by eliminating thefirst and last percentile of the distribution of the relative change in the logarithm of loans. Table 3
Granted loans and family ownership.
Dependent variable:ΔtlogLoansi
(1) (2) (3) (4) (5) Family 0.0577∗∗∗ 0.0618∗∗∗ 0.0477∗∗ 0.0427∗ 0.0509∗∗ (0.0168) (0.0204) (0.0203) (0.0242) (0.0234) log(size) 0.00691 −0.00724 −0.00699 0.359∗ (0.0645) (0.0633) (0.0633) (0.208) Square of log(size) −0.00265 −0.00126 −0.00129 −0.0457∗ (0.00641) (0.00628) (0.00628) (0.0252) Risk −0.0422∗∗ −0.0546∗∗∗ −0.0549∗∗∗ −0.0326 (0.0189) (0.0188) (0.0188) (0.0208) Leverage 0.0108 0.0131 0.0133 −0.000737 (0.0186) (0.0186) (0.0186) (0.0201) % Change in sales2008− 09 0.0582 0.0642 0.0871 0.0680 (0.0450) (0.0451) (0.0688) (0.0509) Year of foundation 0.000122 0.000154 0.000160 0.000922∗ (0.000446) (0.000440) (0.000440) (0.000471) Sharefirst −0.258∗∗∗ −0.259∗∗∗ −0.207∗∗∗ (0.0482) (0.0484) (0.0546)
Family X %Change in sales2008− 09 0.0346
(0.0843)
Constant −0.118∗∗∗ −0.297 −0.194 −0.200 −2.463∗∗
(0.0139) (0.907) (0.894) (0.895) (1.002)
Other controls No Yes Yes Yes Yes
Observations 2851 2026 2026 2026 1473
Robust standard errors in parentheses. Leverage is measured at the end of 2007. Sharefirst measured at the end of Sept. 2008. Other variables are measured at Dec. 2008. Other controls include 11 sector dummies, 3 geographical dummies, cash-flow over revenues and weighted length of the relationships. SMEs are defined as having 250 employees or fewer and annual sales less than 50 million. For all the specifications we cut the 1th e 99th percentiles of the dependent variable to control for outliers.
⁎ p b 0.10. ⁎⁎ p b 0.05. ⁎⁎⁎ p b 0.01.
family and non-familyfirms either up until or immediately after the sudden drop in October 2008. The divergence began after the first quarter of 2009.
3.2. Econometric analysis
In this subsection we test whether the different patterns observed inFig. 1can be explained by differences in the characteristics of
family and non-familyfirms. Given the nature of the exogenous shock considered, we estimate the following model:
L▪tlogLoansi¼ α þ β0Familyiþ β1Xiþ Ei ð1Þ
where subscript i refers to thefirm, and Xiis a vector of controls. The control variables capture channels that have been identified in
the literature as possible determinants of banks' lending behavior and that can be correlated with the family-firm status. Since family
firms are smaller, on average, we include firm size (log of number of employees) and square of firm size at the end of 2008, a
charac-teristic that may be correlated with the familyfirm dummy and may relate to differences in access to the credit market. We also
control for the location of thefirm with three geographical dummies for the North, the Center and the South of Italy. This is justified
by the uneven geographical diffusion of family and non-familyfirms, which could produce different demand shocks and different
terms for credit access, reflecting differing distance between firms' headquarters and their financing banks.
The control variables also include the share of thefirm's credit granted by the first bank (at the third quarter of 2008), as this may
affect the company's ability to substitute across banks and so to hedge bank-specific shocks. For each firm, we construct the weighted
average length of its bank relationships (up to October 2008), weighted by the share of each relationship in total borrowing. In this
way we control for the intensity of the bank–firm relationships, which may not be fully captured by the share of credit granted by
thefirst bank. The firm's age (year of foundation), sector, total leverage, cash-flow over revenues, and risk (defined as a Z-score greater
than 5) are also included as natural controls. Finally, the change in sales between 2008 and 2009 controls for possible differences in
loan demand dynamics due to differential impact of the crisis on the two types offirm.
Our dependent variable is the log difference between average loans granted in two time windows: 1 October 2007–30 September
2008 and 1 October 2008–30 September 2009. Within these “pre-crisis” and “post-crisis” windows, we summed up each firm's loans.
The windows are of the same length, avoiding problems of seasonal adjustment. The fourth quarter of 2008 is taken as the beginning of the post-crisis period both because the Lehman Brothers default came at the very end of the third quarter and because this avoids an
arbitrary choice of windows. An observation period immediately“after” helps to capture mostly supply-side effects in loan dynamics,
since credit lines respond relatively promptly to changes in banks' credit. We exclude the top and bottom percentiles of the distribu-tion of the dependent variable in order to control for outliers and improve estimadistribu-tion accuracy. Summary statistics of the change in log
of loans are reported in thefirst line ofTable 2.
Column (1) estimates the basic model with no control variables; column (2) includes the basic set of controls; column (3) adds the
share of thefirst bank; column (4) adds the interaction term between the family firm dummy and the change in sales; and column
(5) analyzes the sub-sample of SMEs.15The results are shown inTable 3.
Table 3shows that the decline in the credit growth rate was 5 percentage points less for family than non-familyfirms. The coef-ficient is robust to different specifications of the model, and is highly statistically significant as well as economically substantial. Note
that theβ0estimates in column (1), without controls, and column (3), with the complete set of controls, are quite similar, confirming
15
Firms with no more than 250 employees and 50 million in sales, as commonly defined in Italy and also in the rest of the European Union. Table 4
Robustness checks.
Dependent variable ΔtlogLoansi ΔtlogLoansij
(1) (2) (3) (4) (5) Family 0.0444∗∗ 0.0447∗∗ 0.0572∗ 0.0398∗∗ 0.0441∗∗∗ (0.0207) (0.0205) (0.0340) (0.0197) (0.0198) Foreign −0.0335 (0.0569) Group affiliation −0.0455∗∗ (0.0210) Concentration −0.0009∗ (0.0005)
Controls Yes Yes Yes Yes Yes
Bankfixed-effects No No No No Yes
Observations 2026 2009 911 15,212 15,212
For columns (1) and (3): robust standard errors in parentheses. Controls are those included in column (3) ofTable 3. We cut the 1th e 99th percentile of the dependent variable to control for outliers. For columns (4) and (5): controls are those included in column (3) inTable 3, plus the share of loans from bank j tofirm i, relative to total loans forfirm i, and the length of the bank–firm relation, both measured at the end of September 2008. In column (4) we compute robust standard errors clustered at the firm level, while in column (5) clusters are derived at the bank level.
⁎ p b 0.10. ⁎⁎ p b 0.05. ⁎⁎⁎ p b 0.01.
that the differences in the volume of loans granted cannot be fully explained by standard mechanisms; further investigation is
re-quired. Column (4) shows that the magnitude of the coefficient is not driven by any systematic difference in the demand for credit
by family and non-familyfirms, controlling for the contraction in sales.16Finally, column (5), where the sample is restricted to
SMEs, shows that our results are not driven by the change in loans to very largefirms (where family and non-family firms are less
comparable, as shown inTable 1): the coefficient of β0is almost unchanged.
Higher risk is associated with a smaller volume of lending, as theory predicts. Moreover, lending growth is lower when borrowing
is more concentrated with thefirst bank. The negative sign of this coefficient, as noted, may be explained by the fact that the higher
concentration of borrowing with thefirst bank affects the firm's ability to hedge bank-specific shocks. It is also consistent with the
empirical evidence that thefirst bank more frequently belongs to one of the top five Italian banking groups, which cut their credit
more sharply than the other banks during the crisis (Albertazzi and Marchetti, 2010).
Our thesis is that the estimated difference between family and non-familyfirms is driven chiefly by a change in the supply of credit.
True, the change in outstanding loans owing to thefinancial crisis cannot be directly interpreted as a credit supply contraction.
How-ever, the coefficient β0in the regression model captures any additional difference beyond that observed for non-familyfirms, so we
could interpret the difference between family and non-familyfirms as a supply-driven effect, assuming that our rich set of observable
characteristics capturesfirms' demand for credit.17
3.3. Robustness checks
We have documented divergent patterns in the aggregate dynamics of credit between family and non-familyfirms, controlling for
a set of observable characteristics that are potentially correlated with the existence of a family block-holder and able to influence
cred-it dynamics. Some concerns still need to be addressed, however.
3.3.1. Foreignfirms
Thefirst concern is the presence of foreign firms. In fact, the great majority of foreign firms (which account for some 8% of our
sample) is controlled by non-family block-holders and could have different borrowing patterns from Italian companies. In principle
foreignfirms can replace Italian with foreign credit by exploiting their multinational group affiliation, or they may be systematically
discriminated against by local banks. To make sure that our results are not driven by differences in nationality, column (1) inTable 4
adds a dummy for foreign status to the full specification in column (3) inTable 3. Reassuringly, our family dummy is still statistically
significant, although the coefficient is now slightly lower (foreign status is negatively correlated with the change in loans, but the
difference is not significant).
3.3.2. Group affiliation
A second, not unrelated concern, is the possibility thatfirms could substitute intra-group financing for bank credit. If group
affil-iation is negatively correlated with familyfirm status, then our results could be explained by lower demand for bank loans by
non-familyfirms. To control for this, column (2) ofTable 4supplements the full specification of column (3) inTable 3with a dummy
for group affiliation. Again, the results are robust; the group dummy has the expected negative — and statistically significant — sign.
16Insection 4.1we propose an empirical strategy includingfirm fixed effects to account for differences in credit demand more robustly. 17
The results are robust to measuring size by total assets rather than number of employees. Table 5
Summary statistics for non-soft and soft banks.
Non-Soft Soft Difference
Mean Obs. Mean Obs.
Pre-crisis capital ratio .126 210 .120 117 .006∗,∗∗,∗∗∗
[.062] [.0.05]
Pre-crisis weighted average length of the bank–firm relationship (measured in years) 7.353 213 6.852 119 .501
(3.334) (3.720)
Pre-crisis weighted average net interest rate (%) 2.882 213 3.127 119 −.245
(4.836) (4.458)
Pre-crisis share of granted credit to familyfirms (1) .614 213 .649 119 −.035
(.352) (.326)
Pre-crisis share of granted credit to familyfirms (2) .606 213 .626 119 .020
(.345) (.316)
Granted credit amounts are labeled with (1) or (2). (1) refers only to revocable loans. (2) refers to the sum of revocable loans and term loans. Pre-crisis period is from 1 October 2007 to 30 September 2008. Capital ratio is defined as total equity over total assets for each bank; it is measured at the second quarter of 2008. Weighted av-erage length of credit relationship is the number of years of each bank–firm relationship at October 2008. Weights are equal to the relative share of credit granted to firm i by bank j, over the total amount of loans granted to thatfirm, in the time window 1 October 2007–30 September 2008. Weighted average net interest rate is the av-erage interest rate in the time window 1 October 2007–30 September 2008 for each observable bank–firm relationship; weights are constructed as explained above. Standard deviations are in square brackets.
⁎ p b 0.10. ⁎⁎ p b 0.05. ⁎⁎⁎ p b 0.01.
3.3.3. Ownership concentration
We want to be sure that our results are not driven by a difference in the ownership concentration of the controlling shareholder,
which has been found to affect default risk negatively (Aslan and Kumar, 2012) and may differ between family and non-familyfirms.
The cleanest gauge is the fraction of shares held by the largest shareholder; unfortunately this information is only available for
rela-tively largefirms (50 employees or more), which more than halves the number of observations. Column (3) inTable 4adds ownership
concentration to the full specification in column (3) inTable 3. In keeping with theory, greater concentration reduces the amount of
credit granted (although the coefficient is only weakly significant), but, other things being equal, the presence of a family block-holder
significantly reduces this negative effect (the coefficient associated with family firm status is positive and equal to about 6 percentage
points).
3.3.4. Lock-in effects
The last hypothesis we want to test is whether the observed difference between family and non-familyfirms can be explained
sim-ply by ex-ante matching with different kinds offinancial institution. In other words, because it is costly to switch banks, and the
switching costs may be proportional to loan concentration, non-familyfirms might have been “locked in” with the banks that cut
their lending more sharply. In order to address this issue, we need to check whether the same bank treated family and non-family firms differently. By exploiting information on individual bank–firm relationships, we can compare the change in log loans for family
and non-familyfirms, controlling for bank fixed effects (and hence for time-varying bank fixed effects).18We estimate the following
model:
L▪tlogLoansi j¼ α þ β0Familyiþ γXi jþ fjþ Ei j ð2Þ
where L.tlogLoansijis the log difference in the average value of loans granted (the averaging procedure is the same as above) tofirm i,
by bank j. Xijincludes not only thefirm-specific characteristics used earlier but also the share of loans from bank j in total loans to firm i
18
Summary statistics on the change in log loans at the individual bank–firm level are reported in the second row ofTable 2. Fig. 2. Bank lending and heterogeneity in screening technologies.
and the length of the single bank–firm relationship, both at the third quarter of 2008. The addition of these two variables is important,
as they control for very large percentage variations in the dependent variable induced by small loans. Finally, fjrepresents the bank j
fixed effect. The results are reported in columns (4) and (5) ofTable 4. Column (4) gives the analog of the aggregate results presented
in column (3) ofTable 3at the individual bank–firm level. Column (5) adds bank fixed effects. The estimates of β0in the specifications
with and without bankfixed-effects are almost identical and very similar to those at the aggregate level. This confirms that
diver-gences in the amount of loans granted to family and non-familyfirms are not driven by “lock-in” effects of ex-ante sorting of family
firms with particular banks.
4. Heterogeneity among banks in lending practices
In the previous section, controlling for an ample set of observable characteristics, hence conditioning on hard information, we
observed that the credit contraction that followed October 2008 was significantly less severe for family firms. This is consistent
with the hypothesis that additional soft information, acquired through the personal interaction of loan officers with firms' managers,
played a substantial role in explaining the difference.19In particular, this kind of information might have enabled banks to better
assess borrower risk andfind that on average it was lower for the family firms.
To test this hypothesis, we rely on a special survey conducted by the Bank of Italy in 2009, which asked an explicit question on respondent banks' changing attachment of importance to soft information in their lending decisions after October 2008. There was an increase in its relative importance at around 35% of the banks surveyed (which accounted for 36% of total aggregate credit) and a decrease at less than 5%. An increase in the importance of this type of information in the wake of an adverse aggregate shock is
con-sistent with the idea, formalized byBolton et al. (2013), that soft information collected by bank branches can partially take the place of
hard information in assessing borrower risk, since it is continuously updated thanks to frequent contacts with thefirm. At the same
time, the extent of the change in screening methods over time depends on the bank's organizational structure. In particular, getting soft information is costly: in the extreme case of a bank that had used only hard information, a sudden shift to soft-information-based screening would not be likely to be feasible. Before the crisis, regulatory provisions played a role in shaping the optimal mix of hard and soft information for banks. The Basel II reform in 2004 recommended greater reliance on standardized criteria for gauging
company default risk in order to increase cross-border banking transparency and comparability. For ItalyAlbareto et al. (2008)
show that almost all the large Italian banks and most others accordingly adopted hard-information-based practices.
19Examples of soft information are such factors as the degree of cohesiveness amongfirm's shareholders, their personal history, or the existence of hidden personal assets.
Table 6
Soft information and family ownership.
Dependent variable:ΔtlogLoansi
“Non-soft”-type banks “Soft”-type banks
(1) (2) Family 0.00637 0.0676∗ (0.0343) (0.0347) log(size) −0.0824 0.179∗∗ (0.0895) (0.0768) Square of log(size) 0.00124 −0.0177∗∗ (0.00899) (0.00753) Risk −0.0737∗∗ −0.0153 (0.0295) (0.0297) Leverage −0.0318 0.00443 (0.0264) (0.0387) % Change in sales2008− 09 0.108∗ −0.0599 (0.0648) (0.0721) Year of foundation −0.000394 −0.000309 (0.000634) (0.000892) Sharefirst −0.292∗∗∗ 0.0381 (0.0808) (0.0919) Constant 1.179 0.132 (1.310) (1.763)
Other controls Yes Yes
Observations 1970 1827
Robust standard errors in parentheses, clustered at thefirm level. Length of the relation and share of the bank measured at the end of Sep. 2008. Other controls include 11 sector dummies, 3 geographical dummies, cash-flow over revenues and weighted length of the relations.
⁎ p b 0.10. ⁎⁎ p b 0.05. ⁎⁎⁎ p b 0.01.
Table 5shows that there is no difference between“soft-type” and “non-soft-type” banks — i.e. those that did and did not increase
the use of soft information, in pre-crisis level of capitalization (equity over total assets).20Similarly, the weighted averages of net
interest rates and the length of bank–firm relationships do not differ between the two groups.21
Finally, and most importantly for our purposes, the last two rows ofTable 5report the pre-crisis share of credit to familyfirms by
the two groups of banks.22Statistically, there is no difference whether revocable loans only or totalfinancial exposure is measured.
Familyfirms account for a relatively large share of the total both by “soft” and “non-soft” banks, mainly because they make up
some 60% of the sample and are, on average, more leveraged. This last piece of evidence supports the hypothesis that the magnitude of shocks to banks, which is likely to be correlated with the endogenous choice of relying more heavily on soft information, is invariant
to the family-firm characteristic.
Splitting the sample between banks that did and did not increase their use of soft information, inFig. 2we replicate the graphical
analysis ofFig. 1.
The two groups of banks display considerable divergences in lending growth both before and after the crisis. This suggests that the
decision to change lending technique may be correlated with the severity of the shock to the banks. While the credit granted by
“non-soft” banks was already declining before October 2008, that of “soft” banks did not begin to shrink until the end of 2008 with the onset
of the recession. Also, for these banks there was no difference between family and non-familyfirms until the Lehman Brothers
bank-ruptcy shock; the difference emerged thereafter. For the“non-soft” group, such a difference appears to have subsisted even before the
crisis and to have widened modestly following the shock.
To test whether the differences inFig. 2are statistically significant after controlling for heterogeneities between family and
non-familyfirms, we can re-estimate Eq.(1), splitting the total amount of lending to eachfirm into loans from “soft” and “non-soft” banks.
The results are reported inTable 6.
The difference between family and non-familyfirms is statistically significant only for those banks that increased their reliance on
soft information after October 2008. In particular, the decline in lending growth to familyfirms was 6.7 percentage points smaller than
that to non-familyfirms, other factors were constant. This validates our prior that soft information is crucial in explaining the observed
difference in credit access for familyfirms. Consistently with the previous discussion, we find that high risks negatively and
signifi-cantly affect the dynamics of credit granted by“non-soft” banks but do not affect the lending decisions of the “soft” banks. This
evidence jibes with the analysis ofGarcia-Appendini (2011), who shows that for banks lacking access to soft information the
propen-sity to grant a loan is more sensitive to changes in publicly available variables.
4.1. Controlling for unobservedfirm heterogeneity
Thefinal set of results controls fully for unobserved time-varying heterogeneity at the firm level, and in particular for demand-side
effects. We exploit the existence of multiple lending within our sample and we includefirm fixed effects in the following regression
model:
L▪tlogLoansi j¼ α þ β0L▪tSoftjþ β1FamilyiL▪tSoftjþ γZi jþ fiþ Ei j ð3Þ
20
Recall that we are measuring changes in the relative importance, not the absolute level, of soft information. Hence,“non-soft”-type banks could be those that even before October 2008 attached a high value to soft information. However, the estimated coefficient of risk variable inTable 5suggests a strong positive association be-tween the relative and the absolute measures of soft information.
21Weights are equal to the bank j's share in the total credit granted tofirm i during the time window 1 October 2007–30 September 2008. 22
Thefirst of the two rows considers revocable loans only, the second revocable plus term loans. Table 7
Banks' heterogeneity in the screening process.
Dependent variable:ΔtlogLoansi j
(1) (2)
ΔtSoft 0.0995∗∗∗ 0.0642∗
(0.0353) (.0352)
ΔtSoft × Family 0.0797∗ 0.0845∗∗
(.0424) (0.0420)
Share of the bank −1.0583∗∗∗
(0.0773)
Length of the relation −0.0094∗∗∗
(0.0030)
Constant −0.1804∗∗∗ −0.0530∗∗
(0.0082) (0.0240)
Firmfixed-effects Yes Yes
Observations 12,864 12,864
Robust standard errors in parentheses, clustered at thefirm level. Length of the relation and share of the bank measured at the end of Sep. 2008. ⁎ p b 0.10.
⁎⁎ p b 0.05. ⁎⁎⁎ p b 0.01.
where L.tlogLoansijis the change in log loans tofirm i by bank j; L.tSoftijis a dummy equal to 1 if bank j increased the importance
at-tached to soft information after October 2008; Zijdenotes the bank j's share in total loans tofirm i and the length of the bank–firm
relationship, both at the end of September 2008, and controls for loan-specific demand effects that may vary between banks in respect
to family and non-familyfirms, due to heterogeneity in the banks' screening processes; fiis thefirm fixed effect. This estimation
strategy is similar to that proposed byKhwaja and Mian (2008), as thefirm fixed effect controls for demand effects that are invariant
with respect to bank characteristics.
The focus is now on coefficient β1, which identifies the interaction between the family firm dummy and L.tSoftij. This parameter
captures whether banks' use of soft information affected the supply of credit to family and non-familyfirms respectively. Intuitively,
the coefficient measures whether the difference in the family firm dummies between the two columns ofTable 6is supply-driven.
Given the evidence inTable 5, the identifying assumption, similar in spirit toRajan and Zingales (1998), is that the different lending
behavior of the two types of bank to the two types offirm depends on the change in reliance on soft information. The results are given
inTable 7.
Columns (1) and (2) show that the“soft” banks decreased their lending much less sharply to family than to non-family firms. The
effect is sizable (a difference of about 8 percentage points) and statistically significant, offering substantial supporting for the thesis
that banks' soft information mitigated the repercussions of thefinancial crisis on family firms by revealing their greater reliability.
5. Investments, employment and economic performance
In this last section we analyze the economic effects of thefinancial crisis to detect possible differences between family and
non-familyfirms in the years from 2007 to 2009. Our results tie in with the growing body of literature since the financial crisis on the effects
of bank lending shocks on the real economy. Columns (1) to (4) inTable 8take as dependent variable the log difference in physical
capital expenditures, intangible asset investment, number of employees and average wage. In column (5) the dependent variable is
the absolute difference in return on equity. We use balance-sheetfigures at the end of 2007 and 2009. We control for the firm's
economic sector, geographical area, year of foundation, size, square of size, and overall leverage. Unfortunately, the data on some of
the dependent variables for somefirms are missing, so the number of observations is not constant.
There is no significant difference between family and non-family firms in terms of investment (either tangible or intangible), but
there is a significant difference in the change in the number of employees. The reduction in the number of employees was 2.6
percent-age points less among family than non-familyfirms. This difference is not paralleled by a change in average wages, suggesting that
downsizing was relatively uniform across classes of workers. These results are consistent with thefindings ofSraer and Thesmar
(2007),Bach and Serrano-Velarde (2015–in this issue)andD'Aurizio and Romano (2013), namely that employment levels in family firms tend to be less sensitive to changes in governance and to negative aggregate shocks. Finally, the fall in ROE was less severe for
familyfirms, by 2 percentage points, and this difference is statistically significant at conventional levels. These results tend to
corrob-orate the hypothesis that the credit re-allocation towards familyfirms has been efficient from the banks' perspective; they reject the
alternative hypothesis of collusion between family-firm owners seeking to funnel resources out of the company and opportunistic
loan officers gaining private benefits at the expense of the bank.23
Table 8 Real outcomes.
Dependent variable: ΔtlogTang. Invi ΔtlogIntang. Invi ΔtlogEmploym.i ΔtlogWagei ΔtROEi
(1) (2) (3) (4) (5) Family −0.0690 −0.0364 0.0259∗∗,∗∗∗ 0.0128 2.028∗ (0.0830) (0.108) (0.00998) (0.0134) (1.113) log(size) −0.0837 −0.363∗ 0.0955∗∗∗ 0.0149 −1.114 (0.178) (0.217) (0.0213) (0.0367) (2.720) Square of log(size) 0.00725 0.0265 −0.00644∗∗∗ −0.00178 0.0985 (0.0151) (0.0183) (0.00204) (0.00357) (0.2601) Leverage −0.0995 −0.0413 0.00376 −0.00184 0.381 (0.0889) (0.0777) (0.00773) (0.0114) (0.950) Year of foundation −0.00181 −0.000439 0.000414∗∗∗ 0.000227 0.0126 (0.00158) (0.00206) (0.000158) (0.000249) (0.0248) Constant 3.999 1.966 −1.116∗∗∗ −0.436 −25.17 (3.253) (4.242) (0.313) (0.499) (49.32)
Other controls Yes Yes Yes Yes Yes
Observations 1801 1046 2037 1395 1833
Robust standard errors in parentheses. Other controls include 11 sector dummies, 3 geographical dummies. ⁎ p b 0.10.
⁎⁎ p b 0.05. ⁎⁎⁎ p b 0.01.
23
6. Conclusions
We have studied the credit allocation decisions of Italian banks following the Lehman Brothers failure in October 2008 in relation
to differences in borrowingfirms' ownership structure. We have verified that ownership is an important source of heterogeneity
amongfirms. In particular, the presence of a family block-holder had a positive effect in attenuating the agency conflict in the
relation-ship between borrowers and lenders. This effect was closely related to the increased use of soft information by Italian banks in their
lending decisions following the Lehman Brothers failure. The principal result is robust to different specifications of our empirical
model. Thanks to the availability of highly detailed data on bank–firm relations, we were able to control for ex-ante observable
differ-ences between family and non-familyfirms and also to exclude significant “lock-in” effects that impede firms from hedging
bank-specific shocks. Finally we controlled for unobserved heterogeneity, confirming that credit allocation was driven by a change in the
credit supply. At the same time, this difference in credit availability was not paralleled by a difference in capital investment, while
it was associated with a smaller contraction in the total workforce at familyfirms. Our results demonstrate the importance, in studying
the propagation of adverse shocks through bank lending, of examining heterogeneity on both sides of the borrower–lender
relation-ship. Also, in line with other recent contributions, our paper highlights the importance of soft information in mitigating the repercus-sions of a credit crunch.
Finally, it is worth pointing out that our results are not inconsistent with the standard thesis of aflight-to-quality in lending away
from smaller and more opaquefirms to larger and more transparent ones as a consequence of shocks to the banking sector. We
complement the evidence on this point (including the recent contribution byIyer et al. (2014)) by showing that within the same
size class there is a difference in the dynamics of credit supply depending onfirms' ownership structure. However, our analysis
bears primarily on the effects of the shock to the Italian banking sector at the end of 2008, which did not involve direct asset losses. Accordingly, a natural extension of our analysis would be to explore whether the differences in patterns of lending to family and
non-familyfirms are also found when a negative shock to asset values impacts directly on banks' balance sheets. In this case, financial
markets' and regulators' demand for greater banking transparency could induce greater reliance on hard information in the loan
screening process, which would potentially diminish the relative advantage of familyfirms in accessing bank lending in times of crisis.
7. Appendix 7.1. Data construction
The CCR database records all the loans granted by Italian banks above a set reporting threshold. The threshold applies to the sum of
all credit granted to an individualfirm by a bank in three main categories:
1. short-term lines of credit (analyzed in this paper), 2. collateralized credit lines, mortgages, etc., 3. advances, etc.
The reporting threshold has changed over time: it was lowered from 75,000 to 30,000 in September 2008. For missing CCR obser-vations we proceed as follows:
• when an observation for a specific line of credit at the bank–firm level is missing in some quarters between 2007Q4 and 2009Q3, we consider the total value of loans issued by the individual bank,
• if the total amount of loans is above the threshold, we assign 0 to that observation,
• if the total amount of loans is below the threshold, we compute its expected value (37,500 before October 2008 and 15,000 after-wards) and divide it by three (the number of components in the total amount) and assign the resulting value to the observation.
The inclusion of zeros poses a problem in estimating the log difference in loans granted at the individual bank–firm level. We
there-fore exclude these observations from the sample, instead of arbitrarily changing their values to a positive integer. In any event, these
observations are relatively few and of only limited economic relevance, asTable 9clearly shows.
Table 9
Comparative statistics for the bank–firm loan observations (euro).
Mean Median Obs.
Before October 2008
Bank–firm relations disappeared after Sept. 2008 247,692 11,267 458
Bank–firm relations considered in the analysis 710,715 100,000 19,722
After October 2008
New bank–firm relations appeared after Sept. 2008 178,746 6250 438
Bank–firm relations considered in the analysis 672,147.8 100,000 19,722
Before October 2008 refers to bank–firm loan averages either for the period from 1 October 2007 to 30 September 2008; after October 2008 for the period from 1 Oc-tober 2008 to 30 September 2009.
7.2. Collateral channel
The observed differences between family and non-familyfirms in the change in credit granted could also stem from differences in
the ability to provide hard and verifiable collateral. Although our analysis turns on call loans only (which are not directly affected by
collateral), there may still be some substitutability with collateralized term loans. To dispel this concern, we re-estimated the
empir-ical models with the sum of call and term loans as dependent variable. The results confirmed our earlier findings: the estimates were
qualitatively similar to those ofSubsection 3.2and statistically significant. An alternative hypothesis is that there is some
complemen-tarity between call and term loans. That is, a bank may be more willing to grant call loans tofirms that have already pledged collateral
on their term loans. Since endogeneity problems prevent making the change in collateral for eachfirm a control variable, we address
this concern by estimating a model where the dependent variable is the log difference between the average amounts of collateral
pledged by eachfirm in our two time windows (1 October 2007–30 September 2008 and 1 October 2008–30 September 2009).
Table 10, columns (1) and (2), shows that the familyfirm dummy has no explanatory power. This finding reassures us that our results
are not driven by systematic differences in the elasticity of collateral provision between family and non-familyfirms.
7.3. Interest rate
We also analyze the cost of borrowing (Table 10, columns (3) and (4)), to check whether there are differences in the change in the
net interest rate charged to family and non-familyfirms. To this end we exploit a special survey conducted by the Bank of Italy on
about 200 Italian banks. Unfortunately, this procedure significantly reduces the number of observations in our estimation. In the
regression analysis, the dependent variable is the difference between the average interest rates charged on outstanding loans in
our two time windows, 1 October 2007–30 September 2008 and 1 October 2008–30 September 2009. Interest rates are weighted
by the relative volume of credit granted in each bank–firm relationship. Interest rates went down in the period under consideration
(probably as a result of ECB interventions), but no differences between family and non-familyfirms emerged.
7.4. Otherfinancing channels
Given that family and non-familyfirms differ in size, it is possible that the biggest firms may finance themselves by direct recourse
to the capital markets through equity or bond issuance. Therefore, even though a size control was included from the outset, for the
sake of completeness we have re-estimated the main model excludingfirms that made equity or bond issuances or payouts during
the period covered. No more precise information can be obtained from the Invind dataset on all thefirms in our sample. We find
that 16% of all ourfirms changed at least 0.1% of their capital financing structure 14.5% of family firms and 20% of non-family firms.
Re-estimating Eq.(1)without thefirms that accessed the capital market directly, our findings are still robust; the family dummy's
coefficient is always large and significant for all the specifications of the model. This result further reassures us that the size of the
firms is not the explanation for our findings. Table 10
Collateral channel, interest rates.
Dependent variable ΔtCollat. ratioi ΔtNet interest ratei
(1) (2) (3) (4) Family 0.0267 0.0502 0.190 −0.298 (0.0315) (0.0417) (0.800) (0.813) log(size) 0.176∗∗ 3.669 (0.0819) (2.782) Square of log(size) −0.0153∗∗ −0.446 (0.00737) (0.298) Risk 0.00262 0.359 (0.0371) (1.062) Leverage 0.00727 −0.923∗ (0.0361) (0.515) % Change in sales2008− 09 0.0137 −1.399 (0.0602) (1.165) Year of foundation −0.000271 0.0217 (0.000846) (0.0221) Sharefirst 0.00443 2.227 (0.0941) (2.242) Constant −0.0816∗∗∗ −0.0630 −2.907∗∗∗ −52.02 (0.0238) (1.717) (0.674) (47.24)
Other controls No Yes No Yes
Observations 1182 841 998 863
Robust standard errors in parentheses. We cut the 1th e 99th percentiles of the dependent variable to control for outliers. For columns (2) and (4): additional controls are those included in column (3) ofTable 3.
⁎ p b 0.10. ⁎⁎ p b 0.05. ⁎⁎⁎ p b 0.01.
References
Albareto, G., Benvenuti, M., Mocetti, S., Pagnini, M., Rossi, P., 2008.L’organizzazione dell’attivita` creditizia e l’utilizzo di tecniche di scoring nel sistema bancario italiano: risultati di un’indagine campionaria. Banca d’Italia occasional papers, n. 12.
Albertazzi, U., Marchetti, D., 2010.Credit supply, flight to quality and evergreening: an analysis of bank–firm relationships after Lehman. Bank of Italy's Working Pa-pers, n. 756.
Anderson, R., Mansi, S., Reeb, D., 2003.Founding family ownership and the agency cost of debt. J. Financ. Econ. 68 (2), 263–285. Aslan, H., Kumar, P., 2012.Strategic ownership structure and the cost of debt. Rev. Financ. Stud. 25 (7), 2257–2299.
Bach, L., Serrano-Velarde, N., 2015.CEO identity and labor contracts: Evidence from CEO transitions. J. Corp. Financ. 33, 227–242 (in this issue). Bandiera, O., Guiso, L., Prat, A., Sadun, R., 2012.Matching firms, managers and incentives. CEP Discussion Paper, n. 1144.
Bank of Italy, 2011.Survey of Industrial and Service Firms. Supplements to the Statistical Bulletin. Sample Surveys vol. XXI (n. 37).
Berger, A.N., Udell, G., 2002.Small business credit availability and relationship lending: the importance of bank organizational structure. Econ. J. 112 (477), 32–53. Bernanke, B., Gertler, M., Gilchrist, S., 1996.The financial accelerator and the flight to quality. Rev. Econ. Stat. 78 (1), 1–15.
Bertrand, M., Schoar, A., 2006.The role of family in family firms. J. Econ. Perspect. 20 (2), 73–96.
Bolton, P., Freixas, X., Gambacorta, L., Mistrulli, P.E., 2013.Relationship and transaction lending in a crisis. BIS Working Paper, n. 417. Burkart, M., Panunzi, F., Shleifer, A., 2003.Family firms. J. Financ. 53 (5), 2167–2201.
Campello, M., Graham, J.R., Harvey, R.C., 2010.The real effects of financial constraints: evidence from a financial crisis. J. Financ. Econ. 97 (3), 470–487. D'Aurizio, L., Romano, L., 2013.Family firms and the great recession: out of sight, out of mind? Bank of Italy Working Paper, n. 903.
De Mitri, S., Gobbi, G., Sette, E., 2010.Relationship lending in a financial turmoil. Bank of Italy's Working Papers, n. 772.
Detragiache, E., Garella, P., Guiso, L., 2000.Multiple versus single banking relationships: theory and evidence. J. Financ. 55 (3), 1133–1161. Diamond, E., 1989.Reputation acquisition in debt markets. J. Polit. Econ. 97 (4), 828–862.
Ellul, A., Guntay, L., Lel, U., 2009.Blockholders, debt agency costs and legal protection. FRB International Finance Discussion Paper, n. 908. Ellul, A., Pagano, M., Panunzi, F., 2010.Inheritance law and investment in family firms. Am. Econ. Rev. 100 (5), 2414–2450.
Fama, E., 1985.What's different about banks? J. Monet. Econ. 15 (1), 29–39.
Garcia-Appendini E. (2011). Lending to small businesses: The value of soft information, Mimeo.
Guiso, L., Minetti, R., 2010.The structure of multiple credit relationships: evidence from US firms. J. Money Credit Bank. 42 (6), 1037–1071. Holmstrom, B., Tirole, J., 1997.Financial intermediation, loanable funds, and the real sector. Q. J. Econ. 112 (3), 663–691.
Iyer, R., Lopes, S., Peydro, Alcalde J., Schoar, A., 2014.The interbank liquidity crunch and the firm credit crunch: evidence from the 2007–09 crisis. Rev. Financ. Stud. 27 (1), 347–372.
Jensen, M., Meckling, W.H., 1976.Theory of the firm: managerial behavior, agency costs and ownership structure. J. Financ. Econ. 3 (4), 305–360.
Jiangli, W., Unal, H., Yom, C., 2008.Relationship lending, accounting disclosure, and credit availability during the Asian financial crisis. J. Money Credit Bank. 40 (1), 25–55.
Kahle, K.M., Stulz, R.M., 2013.Access to capital, investment, and the financial crisis. J. Financ. Econ. 110 (2), 280–299.
Khwaja, A.I., Mian, A., 2008.Tracing the impact of bank liquidity shocks: evidence from an emerging market. Am. Econ. Rev. 98 (4), 1413–1442.
Lins, K.V., Volpin, P., Wagner, H.F., 2013.Does family control matter? International evidence from 2008–2009 financial crisis. Rev. Financ. Stud. 26 (10), 2583–2619. Ongena, S., Smith, D.C., 2000.What determines the number of bank relationships? Cross-country evidence. J. Financ. Intermed. 9 (1), 26–56.
Panetta, F., Faeh, T., Grande, G., Ho, C., King, M., Levy, A., Signoretti, F., Taboga, M., Zaghini, A., 2009.An assessment of financial sector rescue programs. BIS Papers, n. 48. Petersen M.A. (2004). Information: Hard and Soft, Mimeo.
Petersen, M.A., Rajan, R.G., 1994.The benefits of lending relationships: evidence from small business data. J. Financ. 49 (1), 3–37. Presbitero, A., Udell, G., Zazzaro, A., 2012.The home bias and the credit crunch: a regional perspective. J. Money Credit Bank. 46 (2), 53–85. Quadrini, V., 2011.Financial frictions in macroeconomic fluctuations. Econ. Q. 97 (3), 209–254.
Rajan, R.G., Zingales, L., 1998.Financial dependence and growth. Am. Econ. Rev. 88 (3), 559–586. Sapienza, P., 2004.The effects of government ownership on bank lending. J. Financ. Econ. 72 (2), 357–384.
Sraer, D., Thesmar, D., 2007.Performance and behavior of family firms: evidence from the French stock market. J. Eur. Econ. Assoc. 5 (4), 709–751. Stiglitz, J., Weiss, A., 1981.Credit rationing in markets with imperfect information. Am. Econ. Rev. 71 (3), 393–410.
