Bank-firm relations – Empirics

I. Indicators of bank-firm relationship strength

II. Impact of bank-firm relationships on firms

III. Relationship banking and firms’ credit

constraints
I. Indicators of bank-firm relationship strength

• Source: Kysucky and Norden (2015), The benefits of

relationship lending: a meta analysis, forthcoming in
Management Science
I. Indicators of bank-firm relationship strength

• Duration of relationships (e.g. Degryse and Van Cayseele

(JFI2000)
• Number of years since inception of relationship

• A temporal banking relationship:

• Allows banks to learn (observe payment ability, observe
behaviour through other services)
• Improves contracting choices (contract changes depending
on outcome)
• May induce ex-post monopoly
I. Indicators of bank-firm relationship strength

• Duration of relationships (e.g. Degryse and Van Cayseele

(JFI2000)
• Number of years since inception of relationship

• A temporal banking relationship:

• Allows banks to learn (observe payment ability, observe
behaviour through other services)
• Improves contracting choices (contract changes depending
on outcome)
• May induce ex-post monopoly
 I. Indicators of bank-firm relationship strength
• Scope of relationships (e.g. Degryse and Van Cayseele (JFI2000)
• number of products bought from a bank
• Information intensive products or not
• Number of Relationships
• Single relationships might be stronger relationships
• Social interaction:
• E.g. Distance between borrowers and lenders (e.g. Petersen and
Rajan, JF2002; Degryse and Ongena JF2005); shorter distance might indicate
more intense relationships ( due to more social interations for example )

• Bank control: equity stake, bankers on board…

 I. Indicators of bank-firm relationship strength
• Some more recent methods
• Bank or branch size: larger banks might be less relationship
intense
• Infer from borrower population
• compare borrowers of different banks that are assumed
to use different lending technologies (Mian, JF2006)
• Infer from loan contracts (so reverse engineer)
• variables that can explain pricing; different loan
conditionalities (Beck, Ioannidou and Schaefer (2012))
• Ask banks…(see Beck, Degryse, de Haas and van Horen)
II. Impact of bank-firm relationships on firms
 II. Impact of bank-firm relationships on firms

• Cost of credit
• Typical regression equation:
Interest rate or spread = f(controls, indicators of
relationship strength)+ error term
• Evidence is mixed:
• Duration of relationships: increase or decrease
• Scope: typically reduces cost of credit
• Credit availability
• Petersen and Rajan (JF1994) identify the impact of stronger
relationships through trade credit: firms with stronger
relationships should make less use of costly trade credit
• Evidence suggests relationship banking increases credit
availability
 II. Impact of bank-firm relationships on firms

• Other financing
• Often obtaining bank loans may increase access to other
types of financing such as bond markets, equity markets, or
trade credit

• Firm performance
• Cheaper and better access to credit may improve firm
performance (profitability, growth, investment, innovation)
• evidence:
• Longer, wider, fewer and more intense bank-firm
relationships result in better firm performance
 III. Relationship banking and firms’ credit
constraints
When Arm’s Length is Too Far.
Relationship Banking over the Credit Cycle

Thorsten Beck (Cass Business School)

Hans Degryse (KU Leuven)
Ralph De Haas (EBRD, Tilburg University)
Neeltje Van Horen (De Nederlandsche Bank)

Any views expressed are only those of the authors and should not be attributed to the EBRD, De Nederlandsche Bank or the Eurosystem
 Motivation

Aftermath of the Great Recession: SMEs continue to experience

credit constraints, potentially delaying the long-awaited economic
recovery

Policy makers to the rescue…
 President Obama signs the Small Business Jobs Act
 BoE launches a subsidized SME funding and guarantee scheme
 ECB initiates targeted LTRO aimed at increasing lending to SMEs in
Europe
…
 Motivation

These initiatives may alleviate firm’s funding constraints but are
unlikely a long-term panacea

Open question: how to best protect entrepreneurs in more

structural way from the cyclicality in credit?

We ask: do banks’ lending techniques contribute to the cyclicality

of credit?
 Bank business models and SME lending

Two core lending techniques

① Relationship: bank repeatedly interacts with clients to obtain and exploit

“soft” proprietary borrower information

② Transaction: typical one-off loans based on “hard” verifiable information

and collateralizable assets
 Relationship lending over the credit cycle

“Learning model” (Bolton, Freixas, Gambacorta and Mistrulli, 2013)

① Relationship banks compete with transaction banks
② R-banks incur higher costs due to monitoring and the need to hold more capital.
Charge higher lending rates than T-banks in normal times
③ R-banks learn about the borrower over time, so can continue to lend at more
favorable terms when a crisis hits
④ R-banks relax firms’ credit constraints more than T-banks in crisis times

Test model using Italian credit-registry data and confirm these

theoretical predictions
 Note: R-bank is defined as bank whose headquarter is located in the same
province as the firm
 This paper

Identify relationship banks in a novel way: ask bank CEOs!
 No need for (simplifying) assumptions about lending technologies

Merge with new data on geographic location of bank branches

 Detailed picture of bank branches in the vicinity of each firm

Use direct measure of whether a firm is credit constrained

 Observe whether firms were turned down or discouraged

Information for 2005 and 2008/09

 Observe credit constraints in credit boom and bust

Banks and firms active in 21 countries in Eastern Europe and Caucasus

 Broadens external validity
 Credit cycle in emerging Europe

40
Average credit growth across emerging Europe (%)

35

30

25

20

15

10

-5
2005 2006 2007 2008 2009 2010 2011 2012 2013
Total credit Corporate credit
 Main take away

1. During a credit boom, SME access to credit does not depend on

the local presence of relationship lenders

2. But the presence of relationship lenders alleviates firms’ credit

constraints during a cyclical downturn
3. Positive impact is strongest for opaque firms, firms with no other
sources of external finance, and firms that lack tangible assets

4. Firms in regions where the economic downturn is more severe

also benefit more from the presence of relationship banks
5. Reduction in credit constraints due to relationship lending helps
mitigate the negative impact of financial crisis on firm growth
(not an evergreening story)
 Contribution to the literature

Theoretical and empirical literature on relationship lending
Sharpe, 1990; Rajan, 1992; Von Thadden, 1992, 2004; Detragiache, Garella and Guiso, 2000; Bolton,
Freixas, Gambacorta and Mistrulli, 2013; Petersen and Rajan, 1994; Berger and Udell, 1995; Agarwal
and Hauswald, 2010
 We examine the relevance of relationship lending over the business cycle in a
cross-country setting using a unique micro dataset

Literature studying the cyclicality of banks’ credit supply

Rajan, 1994; Ruckes, 2004;
 We examine the impact of lending techniques on the cyclicality of credit

Literature studying the factors that affect firms’ financing constraints

Fazzari, Hubbard and Petersen, 1988; Beck, Demirguc-Kunt and Maksimovic, 2005; Brown, Ongena,
Popov, Yesin, 2011; Popov and Udell, 2012
 We examine how bank lending techniques affect financing constraints of
SMEs
Data
 Dataset

We merge and combine three datasets:

① Firm characteristics
② Overview of bank branches
③ Bank characteristics incl. lending techniques
 1. Firms

EBRD-World Bank Business Environment and Enterprise

Performance Survey (BEEPS):
 21 countries in Eastern Europe and Caucasus
 Different levels of economic and financial development

Purpose of survey: gauge the extent to which different features of the

business environment constitute obstacles to firms’ operations
 Information on whether firm is credit constrained
 Large number of firm characteristics
 Geographical location of each firm

Sample
 2005 (6,948 firms): credit boom
 2008-09 (6,901 firms): turn of the credit cycle
 1. Firms

Survey data allow us to directly observe

1. Firms that do not need at loan: not credit constrained
2. Firms that need a loan and got it: not credit constrained
3. Firms that need a loan but are discouraged: credit constrained
4. Firms that need a loan but are rejected: credit constrained

Allows for the identification of the impact of credit-supply shocks

 Firms

Identify credit constrained firms (follow Popov and Udell 2012)
 BEEPS question K16 “Did the establishment apply for any loans or lines of
credit in the last fiscal year?”
 If NO, question K17 “What was the main reason that the establishment did
not apply?”
 If YES, question K18a “In the last fiscal year, did this establishment apply for
any new loans or credit lines that were rejected?”

Needs loan and not credit constrained:
 K16: “Yes” and K18a: “No”

Needs loan and credit constrained:
 K18a: “Yes” (= rejected firms)
 K17: “Interest rates not favorable”; “Collateral requirements too high”;
“Size of loan and maturity are insufficient”; “Did not think would be
approved” (= discouraged firms)

Does not need loan:
 K16: No and K17: “does not need loan”
 Firms

Loan needed Constrained

2005 2008 2005 2008

Share firms 0.70 0.62 0.34 0.40

Substantial share of firms constrained

Tightening of financing constraints in 2008

Substantial variation across countries:
 Slovenia: 12% firms credit constrained in 2005 and 17% in 2008
 Azerbaijan: 64% in 2005 and 78% in 2008
 2. Bank branches

Data hand-collected
 Directly contacting banks, bank websites, central banks
 Cross-checked with (more limited) information in SNL database

Sample
 Geo-coordinates of 38,310 branches of 422 banks (96.8% of bank assets)
 Opening and closing dates so time-varying information (2005, 2008, and
historical)
 Connection between banks and firms

Combine geographic location of the firm with location of branches to
determine which banks are physically present in vicinity of the firm

Firms tend to do business with nearby banks (< 6 km) (Petersen and
Rajan, 2002; Degryse and Ongena, 2005; Agarwal and Hauswald, 2010)

Two methods:
① Locality (main method)
• E.g. link all BEEPS firms in Czech city of Brno to all bank branches in Brno
• If firm in locality without bank branches, we link to branches in nearest locality
• Total 2,478 localities with on average 21 bank branches

② Circles with 5 (10) km radius

• Draw circles with a radius of 5 or 10 kilometers around the geo-coordinates of
each firm and then link the firm to only those branches inside circle
• A 5 (10) km radius contains on average 18 (30) branches
 3. Bank lending techniques

Use 2nd Banking Environment and Performance Survey (BEPS II)
 Face-to-face interviews with almost 400 CEOs of the banks operating in our
sample of countries (80.1% of bank assets)

BEPS II question Q6 “Rate on a five point scale the frequency of use

of the following lending techniques when dealing with SMEs”:
 Relationship lending
 Fundamental and cash-flow analysis
 Business collateral
 Personal collateral

Bank is relationship bank if it considers “relationship lending” very

important.
 Other definitions in robustness tests
 3. Bank lending techniques

For both 2005 and 2008-09, we identify for each branch in the
vicinity (circle or locality) of each firm whether it is a relationship
bank or not

Variable Share relationship banks:

 Number of branches of relationship banks to total number of
branches in the locality of the firm

On average, 52% in 2005 and 50% in 2008

 Bank lending techniques
Substantial differences across countries ...
Share relationship
banks
2005 2008
Albania 0.92 0.83
Armenia 0.35 0.46
Azerbaijan 0.36 0.45
Belarus 0.26 0.27
Bosnia 0.59 0.56
Bulgaria 0.84 0.77
Croatia 0.74 0.71
Czech
Republic 1.00 0.90
Estonia 0.57 0.53
Georgia 0.18 0.19
Hungary 0.60 0.58
Latvia 0.49 0.45
Lithuania 0.61 0.59
Macedonia 0.40 0.39
Moldova 0.27 0.28
Poland 0.60 0.59
Romania 0.58 0.55
Serbia 0.81 0.79
Slovak
Republic 0.27 0.31
Slovenia 0.67 0.64
Ukraine 0.11 0.27
 3. Bank lending techniques
… and substantial variation within-country
Empirical methodology
 Empirical methodology

Model:

Dependent variable: Yijkl =1 if firm i in locality j of country k in

industry l is credit constrained (rejected of discouraged), 0 otherwise

Share relationship bank: Share of bank branches in locality j of
country k that belong to banks for which relationship lending is “very
important” when dealing with SMEs

β3: The impact of the intensity of relationship banking on firms’ credit
constraints

Sample period: 2005 and 2008/09
 Evaluate importance of relationship banking over the credit cycle
 Empirical methodology

Covariates:
 Firm variables: small firm, large firm, publicly listed, sole
proprietorship, privatized, exporter, audited  control for
observable firm-level heterogeneity

 Locality variables: bank solvency (Tier 1), share foreign banks,

wholesale funding, economic activity locality (capital or city) 
control for bank and locality characteristics
• Constructed analogously to bank relationship variable

 Country and industry fixed effects  control for (un)observable

variation at country and industry level
 Empirical methodology

Probit with and without first-stage Heckman selection:

 Heckman first-stage dependent variable: D=1 if firm needs a loan, 0 otherwise
 Controls for the fact that being credit constrained is only observable if the firm
has need for a loan
 Selection variables: ‘competitive pressure’ and ‘applied for subsidy’ (Popov
and Udell, JIE2012; Hainz and Nakobin 2013)

Standard errors: cluster at country level

Results
 Relationship lending and demand for credit
2005 2008
Locality 5 km 10 km Locality 5 km 10 km
[1] [2] [3] [4] [5] [6]
Share relationship banks -0.082 0.024 0.028 0.046 0.051 0.089
(0.157) (0.141) (0.163) (0.139) (0.122) (0.138)
Competition 0.317*** 0.309*** 0.311*** 0.250*** 0.246*** 0.239***
(0.045) (0.044) (0.042) (0.043) (0.041) (0.042)
Subsidized 0.264*** 0.278*** 0.266*** 0.297*** 0.294*** 0.288***
(0.084) (0.083) (0.084) (0.086) (0.086) (0.081)
Firm controls Yes Yes Yes Yes Yes Yes
Locality controls Yes Yes Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes
Number of obs. 6,451 6,739 6,631 6,616 6,670 6,821
Pseudo R2 0.052 0.052 0.052 0.054 0.055 0.054

First stage of Heckman selection model with Demand for credit as dependent
variable
 Relationship lending and demand for credit
2005 2008
Locality 5 km 10 km Locality 5 km 10 km
[1] [2] [3] [4] [5] [6]
Share relationship banks -0.082 0.024 0.028 0.046 0.051 0.089
(0.157) (0.141) (0.163) (0.139) (0.122) (0.138)
Competition 0.317*** 0.309*** 0.311*** 0.250*** 0.246*** 0.239***
(0.045) (0.044) (0.042) (0.043) (0.041) (0.042)
Subsidized 0.264*** 0.278*** 0.266*** 0.297*** 0.294*** 0.288***
(0.084) (0.083) (0.084) (0.086) (0.086) (0.081)
Firm controls Yes Yes Yes Yes Yes Yes
Locality controls Yes Yes Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes
Number of obs. 6,451 6,739 6,631 6,616 6,670 6,821
Pseudo R2 0.052 0.052 0.052 0.054 0.055 0.054

No relationship between share of relationship banks and the demand for credit

Unlikely that relationship lending is endogenous to local demand conditions
 Relationship lending and credit constraints

2005 2008
Probit Heckman Probit Heckman
Locality Locality 5 km 10 km Locality Locality 5 km 10 km
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
Share relationship banks 0.017 0.191 0.169 0.240 0.159 -0.431*** -0.470*** -0.439*** -0.427*** -0.403**
(0.246) (0.270) (0.244) (0.200) (0.202) (0.134) (0.152) (0.156) (0.162) (0.182)
Firm controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Locality controls No Yes Yes Yes Yes No Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Number of obs. 4,162 4,059 4,059 3,655 4,055 4,052 4,022 4,022 3,868 4,126
Pseudo R2 0.14 0.15 0.15 0.14 0.15 0.10 0.10 0.10 0.10 0.10
 Relationship lending and credit constraints

2005 2008
Probit Heckman Probit Heckman
Locality Locality 5 km 10 km Locality Locality 5 km 10 km
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
Share relationship banks 0.017 0.191 0.169 0.240 0.159 -0.431*** -0.470*** -0.439*** -0.427*** -0.403**
(0.246) (0.270) (0.244) (0.200) (0.202) (0.134) (0.152) (0.156) (0.162) (0.182)
Firm controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Locality controls No Yes Yes Yes Yes No Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Number of obs. 4,162 4,059 4,059 3,655 4,055 4,052 4,022 4,022 3,868 4,126
Pseudo R2 0.14 0.15 0.15 0.14 0.15 0.10 0.10 0.10 0.10 0.10

In 2005 (credit boom) no significant relationship between the local

importance of relationship lending and firms’ financing constraints
 Relationship lending and credit constraints

2005 2008
Probit Heckman Probit Heckman
Locality Locality 5 km 10 km Locality Locality 5 km 10 km
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
Share relationship banks 0.017 0.191 0.169 0.240 0.159 -0.431*** -0.470*** -0.439*** -0.427*** -0.403**
(0.246) (0.270) (0.244) (0.200) (0.202) (0.134) (0.152) (0.156) (0.162) (0.182)
Firm controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Locality controls No Yes Yes Yes Yes No Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Number of obs. 4,162 4,059 4,059 3,655 4,055 4,052 4,022 4,022 3,868 4,126
Pseudo R2 0.14 0.15 0.15 0.14 0.15 0.10 0.10 0.10 0.10 0.10

In 2008 (credit crunch) firms in locality with more relationship banks less
likely to be credit constrained
 Relationship lending and credit constraints

2005 2008
Probit Heckman Probit Heckman
Locality Locality 5 km 10 km Locality Locality 5 km 10 km
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
Share relationship banks 0.017 0.191 0.169 0.240 0.159 -0.431*** -0.470*** -0.439*** -0.427*** -0.403**
(0.246) (0.270) (0.244) (0.200) (0.202) (0.134) (0.152) (0.156) (0.162) (0.182)
Firm controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Locality controls No Yes Yes Yes Yes No Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Number of obs. 4,162 4,059 4,059 3,655 4,055 4,052 4,022 4,022 3,868 4,126
Pseudo R2 0.14 0.15 0.15 0.14 0.15 0.10 0.10 0.10 0.10 0.10

In 2008 (credit crunch) firms in locality with more relationship banks less
likely to be credit constrained:
 A move from a locality with 20% relationship lenders to one with 80%
relationship lenders reduces the probability of being credit constrained by 10
percentage points
Control variables 2005
Control variables 2008-09
Robustness
Controlling for local competition and small banks
Additional controls local credit markets
2005 2008-09 2005 2008-09 2005 2008-09
[1] [2] [3] [4] [5] [6]
Share Relationship Banks 0.182 -0.421*** 0.182 -0.436*** 0.169 -0.425***
(0.259) (0.149) (0.263) (0.157) (0.247) (0.153)
HHI -0.167 0.348**
(0.141) (0.153)
Lerner index -0.415 0.504
(0.846) (1.084)
Share small banks -0.072 -0.100
(0.404) (0.165)
Firm controls Yes Yes Yes Yes Yes Yes
Locality controls Yes Yes Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes
Number of obs. 4,527 4,085 4,519 4,084 4,525 4,083
Pseudo R2 0.132 0.100 0.132 0.099 0.132 0.099
 Alternative relationship-lending measures
Relationship banks Relationship banks Relationship banks Share transaction
(continuous) (relative to other (relative to retail banks
lending techniques) borrowers)
2005 2008-09 2005 2008-09 2005 2008-09 2005 2008-09
[1] [2] [3] [4] [5] [6] [7] [8]
Share relationship banks 0.089 -0.233** -0.115 -0.520* 0.209 -0.455** -0.203 0.413**
(0.153) (0.116) (0.268) (0.279) (0.242) (0.219) (0.256) (0.192)
Firm controls Yes Yes Yes Yes Yes Yes Yes Yes
Locality controls Yes Yes Yes Yes Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes
Number of obs. 4,527 4,085 4,527 4,085 4,527 4,083 4,527 4,085
Pseudo R2 0.131 0.098 0.131 0.098 0.132 0.099 0.132 0.099

Continuous variable: takes each bank’s score (0 to 4) on importance of relationship lending and
then takes the branch-weighted average of this score by locality

Relative variable I: takes each bank’s score (0 to 4) on importance of relationship lending divided
by its score (0 to 4) on importance fundamental/cash flow analysis and then takes the branch-
weighted average of this ratio by locality

Relative variable II: no. branches of banks for whom relationship lending is "Very important" for
SME but not for retail lending/total no. bank branches in the locality.

Transaction bank: banks for which fundamental/cash flow analysis is “very important”, while
relationship lending is not “very important”
 Also robust to…

Clustering standard errors at the locality level or using wild cluster

bootstrap-t procedure (Cameron, Gelbach and Miller, 2008)

Use of linear probability OLS instead of probit

Pooling 2005 and 2008 observations and including 2008 interaction

Excluding Ukraine

Excluding banks with ownership change

Endogeneity
 Self-selection banks

Results unlikely driven by R-banks self-selecting into certain localities:
1. Share relationship banks does not affect demand for loan
2. Results unchanged if we use share R-banks in 1995/2000 as regressor
3. Results hold in IV procedure with R-banks in 1995 as instrument
4. Average characteristics of firms in locality independent of share R-banks
5. Negative Altonji ratio (Altonji et al. 2005; Bellows and Miguel 2008)
Share relationship banks: 1995 Share relationship banks: 2000

2005 2008 2005 2008

[7] [8] [9] [10]
Share relationship banks 0.044 -0.346*** 0.178 -0.299**
(0.211) (0.073) (0.146) (0.128)
Firm controls Yes Yes Yes Yes
Locality controls Yes Yes Yes Yes
Country FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
Number of obs. 4,063 3,537 4,137 3,683
Pseudo R2 0.134 0.099 0.134 0.100
 Self-selection firms

Results unlikely driven by firms self-selecting into certain localities:
1. Results also hold for old(er) firms

Firms 5 years and Firms 10 years and Firms 12 years and

older older older
2005 2008 2005 2008 2005 2008
[1] [2] [3] [4] [5] [6]
Share relationship banks 0.125 -0.478*** 0.202 -0.390** 0.147 -0.464**
(0.237) (0.157) (0.254) (0.193) (0.262) (0.212)
Firm controls Yes Yes Yes Yes Yes Yes
Locality controls Yes Yes Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes
Number of obs. 4,174 3,738 2,776 2,904 2,153 2,525
Pseudo R2 0.134 0.103 0.150 0.106 0.158 0.111
Differences across firms and countries
 Firm heterogeneity
2005
Firm type → Employees Age Exporter Audited External Publicly Asset
funding listed tangibility

[1] [2] [3] [5] [6] [7] [8]

Share relationship banks 0.055 0.089 0.148 0.296 0.102 0.173 -0.008
(0.380) (0.514) (0.262) (0.240) (0.265) (0.244) (0.347)
Share relationship banks * Firm 0.028 0.032 0.082 -0.278** 0.304 -0.233 -0.034
type (0.070) (0.165) (0.295) (0.138) (0.270) (0.579) (0.243)
Firm type -0.262*** 0.088 -0.269* -0.116 0.094 -0.002 -0.339**
(0.080) (0.076) (0.157) (0.076) (0.153) (0.381) (0.144)
Firm controls Yes Yes Yes Yes Yes Yes Yes
Locality controls Yes Yes Yes Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes No
Number of obs. 4,527 4,520 4,527 4,527 4,527 4,527 1,929
Pseudo R2 0.146 0.134 0.132 0.132 0.136 0.132 0.168

Virtually none of the main or interaction effects significant in 2005

 Firm heterogeneity
2008
Firm type → Employees Age Exporter Audited External Publicly Asset
funding listed tangibility
[9] [10] [12] [13] [14] [15] [16]
Share relationship banks -1.040*** -1.065*** -0.572*** -0.598*** -0.532*** -0.535*** -0.431*
(0.312) (0.364) (0.192) (0.182) (0.170) (0.165) (0.257)
Share relationship banks * Firm 0.181** 0.244** 0.409* 0.333* 0.448*** 0.594** 0.448**
type (0.078) (0.123) (0.219) (0.188) (0.167) (0.250) (0.205)
Firm type -0.282*** -0.139* -0.391*** -0.363*** -0.184** -0.045 -0.372***
(0.062) (0.073) (0.116) (0.115) (0.089) (0.132) (0.090)
Firm controls Yes Yes Yes Yes Yes Yes Yes
Locality controls Yes Yes Yes Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes No
Number of obs. 4,085 4,023 4,085 4,085 4,085 4,085 1,652
Pseudo R2 0.107 0.101 0.100 0.100 0.101 0.100 0.122

In downturn, especially smaller and opaque firms benefit from the

presence of relationship banks
 Geographical heterogeneity
Country GDP growth Regional GDP growth Regional GDP
growth if available;
country GDP
growth otherwise
[1] [2] [3] [4] [5] [6]
Share relationship banks -0.324* -0.400*** -0.546*** -0.631*** -0.362** -0.444***
(0.189) (0.151) (0.206) (0.198) (0.153) (0.150)
Share relationship banks*Output growth 2008-09 1.869 2.510** 2.451**
(1.464) (1.237) (1.093)
Share relationship banks*Output growth 2007-09 1.711** 1.151** 1.229**
(0.863) (0.576) (0.481)
Firm controls Yes Yes Yes Yes Yes Yes
Locality controls Yes Yes Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes
Number of obs. 4,085 4,085 3,099 3,099 4,085 4,085
Pseudo R2 0.099 0.099 0.095 0.093 0.101 0.100

Stronger impact in countries and regions hit harder by the Great Recession

Consistent with interpretation that collecting soft information by relationship
banks enables them to continue lending when a crisis hits
Relationship banking: helping hand or
evergreening?
 Real economic impact

Results indicate that the presence of relationship banks alleviates local

firms’ credit constraints during an economic downturn:
A. Does the presence of relationship banks help sound firms to bridge
difficult times and recover more quickly? (Chemmanur and Fulghieri, 1994:
“helping hand”)

B. Or does this reflect an evergreening story where banks roll over loans to
underperforming firms? (Cabellero, Hoshi and Kashyap, 2008: “zombie
lending”)
 Real economic impact

Need information on firms’ balance sheets:

 Match firms in BEEPS 2008-09 sample with Orbis database
 Match almost 50% of firms (2,966 firms)

Analyze growth in assets, operational revenue and number of employees

in 2008-2010 and 2005-2007 (placebo)

2SLS: Credit constrained instrumented by Share relationship banks:
 First stage: explain firm-level credit constraints (in 2008) by relationship
lending at the locality level (and other covariates)
 Second stage: explain real firm growth through the exogenous variation in
credit constraints
 Identifying assumption: Presence relationship banks only affects firm growth
through its impact on these firms’ ability to access credit
 Real economic impact

Dependent variable → ∆ Net Debt ∆ Net Debt ∆ Net Debt 2007-09

2005-07 2007-09
Safe firm proxy: Low Ohlson Safe firm proxy: Low default
O-score probability
2007 2009 2007 2009
[1] [2] [3] [4] [5] [6]
Share relationship banks -0.002 0.055** -0.075 -0.096* -0.137* -0.143**
(0.014) (0.018) (0.048) (0.044) (0.073) (0.053)
Share relationship banks * Safe firm 0.188* 0.196** 0.222** 0.229**
(0.086) (0.086) (0.100) (0.083)
Safe firm -0.123** -0.346*** -0.160** -0.374***
(0.047) (0.064) (0.055) (0.059)
Firm controls Yes Yes Yes Yes Yes Yes
Locality controls Yes Yes Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes
Number of obs. 62,793 79,261 57,162 69,356 57,162 69,355
R2 0.036 0.016 0.016 0.110 0.017 0.113

Higher share of relationship banking reduces drop in net debt, but only for safe
firms

Support for helping hand hypothesis
 Real economic impact

2007-2009 2005-2007
Inve stme nt Growth Growth Inve stme nt Growth Growth
total asse ts e mploye e s total assets e mployee s

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

∆ Net De bt 2007-09 10.339** 1.446** -0.697 78.284 3.309 1.68
(5.004) (0.633) (0.847) (121.731) (4.617) (4.294)
Firm controls Yes Yes Yes Yes Yes Yes
Locality controls Yes Yes Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes
F-stat 7.610 7.610 7.610 0.651 0.651 0.651
R2 (first stage) 0.019 0.019 0.019 0.011 0.011 0.011
Share relationship banks 0.044*** 0.044*** 0.044*** 0.010 0.010 0.010
(first-stage) (0.015) (0.015) (0.015) (0.011) (0.011) (0.011)
Number of obs. 71,009 71,009 71,009 49,596 49,596 49,596

Firms with a higher change in net debt (instrumented by share relationship

banks) have a better real performance
 Conclusions

How to shield SMEs from credit cycle downturns?

We examine the impact of relationship lending on firms’ credit constraints
at different parts of the cycle, using banks’ own views about which
lending technique they use. We find that:
 During ‘good times’, relationship and transaction lending techniques act as
substitutes
 During ‘bad times’, relationship lending helps alleviate credit constraints
 Especially for opaque firms
 Especially in countries that suffer more during the crisis

Helping banks to move from transaction to relationship lending may assist

SMEs during business-cycle downturns and hence soften such downturns

Relationship banking is helping hand and not evergreening