Information sharing between banks

Information sharing between banks:
theory and evidence from transition
countries
Marco Pagano
Università di Napoli Federico II,
CSEF and CEPR
Conference on “The Changing Geography of Banking”
Ancona, 22 – 23 September 2006
1
1. Information sharing between banks


Rather than collecting data about loan applicants,
banks may get it from other banks, via:

private arrangements: credit bureaus (CB)

public arrangements: public credit registries (PCR)
Information sharing (IS) arrangements feature:

Obligation to (i) reciprocity; (ii) timely and accurate data

CB or PCR consolidates data by borrower and, when
requested, feeds data back to banks (credit reports)

Type of information reported:

Negative: past defaults (black lists)

Positive: credit exposure, maturity structure, collateral, etc.
2
The main issues

IS systems have become increasingly
widespread, and their activity has increased

Often CBs and PCRs coexist: complementarity
(reporting threshold, degree of detail)

Three main research questions:

How does IS affect credit market performance?

Why (and when) do banks agree to do this?

Does the type of information shared matter?
3
The rest of this talk


Brief account of past research:

Theoretical predictions

Empirical findings
Two new pieces of research:

Information sharing with non-exclusive lending with
Alberto Bennardo and Salvatore Piccolo

Information sharing and credit market performance: firmlevel evidence from transition countries with Martin
Brown and Tullio Jappelli
4
2. Past research: theory

Three main effects of IS among lenders:
1. IS about borrowers’ types can reduce adverse selection
towards non-local applicants  raises efficiency, but has
ambiguous effect on lending (Pagano-Jappelli, 1993).
2. IS about types fosters competition  reduces lenders’
future informational rents  mitigates hold-up problems
 sharpens borrowers’ incentives, lowers default and
interest rates, increases lending (Padilla-Pagano, 1997).
3. With both adverse selection and moral hazard, default
signals bad quality to outside banks  IS about past
defaults (but not about types) raises incentives to repay
 lowers default and interest rate (Padilla-Pagano, 2000).
5
Past research: theory (2)

These models indicate also when banks should
find it more worthwhile to share information:

High geographic mobility of potential credit applicants 
many non-local customers.

High entry barriers arising from non-information related
factors (e.g., switching costs, regulation).

Low resource cost of setting up information sharing
arrangement.

Poor protection of creditor rights, low accounting
transparency  low incentives to perform for borrowers.
6
Gap in the theory

So far IS always modeled under exclusive lending,
i.e. each borrower borrows only from a single lender.

But:

In practice, multiple lending relationships are widespread.

And, even if a borrower patronizes a single bank in
equilibrium, the possibility of borrowing from others can
affect equilibrium strategies and outcomes.

Particularly problematic for models of IS: most IS systems
relay data about a borrower’s exposure to lenders. Perhaps
this reveals concerns arising from multi-bank lending…
7
3. IS with non-exclusive lending



Bennardo-Pagano-Piccolo (2006): entrepreneur can borrow
simultaneously from many banks to fund fixed-size project.
But if he can borrow in excess of his needs, he consumes
the “free cash flow” and defaults on corresponding lender.
Two settings:



Each bank can fund whole project  exclusivity is possible but
cannot be enforced without IS  bank knows that entrepreneur
has the incentive to borrow from other banks too.
No bank can fund whole project  exclusivity ruled out  bank
must consider that the entrepreneur has the incentive to borrow
in excess of his needs.
In both settings, each bank produces a negative contracting
externality  may cut or deny loan to avoid “overlending”.
8
A new role for information sharing


Consider situations where IS would entail a unique
perfectly competitive equilibrium and efficiency.
Without IS, instead, we have:



IS plays two roles:



Multiple equilibria, many of which imperfectly competitive. Even
monopoly can be sustained: if one bank plays the monopolistic
strategy, others banks may not undercut it for fear of overlending.
Insufficient lending: for fear of overlending, banks may offer so
little that the project is under-funded and cannot be implemented.
fostering competition
ensuring financing of investment project
Borrowing from several banks may also produce positive
contracting externalities. If so, IS helps exploit them.
9
4. Past research: empirical findings

IS helps to predict defaults:




IS sharpens borrowers’ incentive to repay:


Chandler-Parker 1989,
Barron-Staten 2003
Kallberg-Udell 2003, etc.
experimental evidence by Brown and Zender (2006)
IS increases lending and lowers default rate:

aggregate data:



Jappelli-Pagano (2002): 43 countries
Djankov-McLiesh-Shleifer (2006): 129 countries
firm-level data:

Love-Mylenko (2003), using 1999 World Bank data.
10
Problems with the existing evidence

Mostly based on aggregate data:




effects on individual firms may be confounded by
composition effects
impossible to explore whether effects of IS differ depending
on firm characteristics
Neglect possible correlation between IS and other
country-level macro or institutional variables.
Little work on transition countries, in spite of their
interest due to:


poor creditor protection  IS might play a substitute role
rapid and uneven change in institutions and credit markets
11
5. Firm-level study on transition
countries

Brown-Jappelli-Pagano (2006) tap firm-level
data for transition countries:


EBRD/World Bank survey designed for transition
countries (BEEPS data)  data on credit access and
cost of credit for 2002 and 2005
Merge them with


country-level data from World Bank / IFC “Doing
Business” survey to measure information sharing
data on other country-level macroeconomic and
institutional variables, such as banking reform
12
0
.5
1
1.5
2
2.5
3
Rapid development of IS in transition
countries
1996
1998
2000
Information Sharing Index
PCB score
2002
2004
PCR score
13
Dependent variables
Access to credit:


Answer to question “how problematic is your access to
financing for the operation and growth of your business?”
Cost of credit:


Answers to question: “how problematic is the cost of
financing for the operation and growth of your business?”
Construction of dependent variables:


Code answers on a scale from 1 to 4 (1=major obstacle,
2=moderate obstacle, 3=minor obstacles, 4=no obstacles)

Higher values indicate an improvement in the terms at
which credit is available: easier access and lower cost.
14
Approach 1: cross-sectional estimates

Cross-sectional estimates for 2002 data:



largest number of observations (9,655)
richest in terms of variables
Specifications:



Baseline: explanatory variables include IS, other
macro variables, and firm-level characteristics
Expanded: add interaction effects
Country effects: control for unobservable
heterogeneity, IS enter only via interaction effects
15
Crosssectional
estimates
Information sharing (IS)
Firm's age
Medium firm
Large firm
Dependent
variable:
Privatized company
State-owned firm
Transparency
Access to
finance
Per capita GDP
Inflation
Index of bank reform
Baseline
Interactions
0.125
(4.11)**
-0.002
(2.36)*
0.129
(3.35)**
0.176
(4.28)**
0.053
(0.79)
0.089
(1.24)
0.131
(4.72)**
0.040
(1.89)
-0.231
(2.37)*
-0.213
(2.01)*
0.144
(4.96)**
-0.003
(2.50)*
0.115
(2.70)**
0.135
(3.82)**
0.057
(0.70)
0.106
(1.27)
0.162
(5.51)**
0.041
(2.04)*
-0.233
(2.39)*
-0.218
(2.10)*
0.001
(1.82)
0.013
(0.89)
0.046
(1.25)
-0.010
(0.24)
-0.011
(0.25)
-0.036
(3.07)**
Firm's Age × IS
Medium firm × IS
Large firm × IS
Privatized company × IS
State-owned firm × IS
Transparency × IS
Country dummies
Observations
No
5392
No
5392
Country effects
Country effects
and further
interactions
-0.002
(1.55)
0.141
(3.21)**
0.154
(4.27)**
0.024
(0.29)
0.069
(0.81)
0.141
(5.20)**
-0.002
(1.56)
0.140
(3.11)**
0.151
(4.16)**
0.029
(0.35)
0.072
(0.84)
0.238
(2.91)**
0.001
(1.25)
0.006
(0.36)
0.048
(1.25)
-0.002
(0.06)
-0.006
(0.12)
-0.029
(2.38)*
0.001
(1.24)
0.005
(0.30)
0.049
(1.27)
-0.003
(0.08)
-0.007
(0.15)
-0.023
(2.01)*
Yes
5392
Yes
5392
16
Crosssectional
estimates
Information sharing (IS)
Firm's age
Medium firm
Large firm
Dependent
variable:
Privatized company
State-owned firm
Transparency
Cost of
credit
Per capita GDP
Inflation
Index of bank reform
Baseline
Interactions
0.139
(2.99)**
-0.002
(1.82)
0.055
(1.28)
0.126
(1.81)
0.045
(0.73)
0.175
(2.69)**
0.054
(1.89)
0.019
(0.74)
-0.282
(2.02)*
-0.196
(1.12)
0.146
(2.86)**
-0.003
(1.67)
0.051
(1.02)
0.108
(1.46)
0.070
(0.87)
0.210
(2.62)**
0.060
(1.70)
0.020
(0.77)
-0.282
(2.01)*
-0.198
(1.13)
0.000
(0.41)
0.004
(0.15)
0.019
(0.37)
-0.026
(0.98)
-0.040
(1.15)
-0.006
(0.40)
Firm's Age × IS
Medium firm × IS
Large firm × IS
Privatized company × IS
State-owned firm × IS
Transparency × IS
Country dummies
Observations
No
5450
No
5450
Country effects
Country effects
and further
interactions
-0.001
(0.57)
0.075
(1.28)
0.099
(1.39)
-0.003
(0.03)
0.160
(2.06)*
0.068
(2.68)**
-0.001
(0.56)
0.072
(1.21)
0.092
(1.30)
0.001
(0.02)
0.162
(2.07)*
0.126
(1.76)
-0.000
(0.24)
-0.003
(0.09)
0.025
(0.49)
-0.000
(0.01)
-0.021
(0.69)
-0.016
(1.10)
-0.000
(0.24)
-0.004
(0.13)
0.026
(0.49)
-0.001
(0.04)
-0.022
(0.72)
-0.017
(1.19)
Yes
5450
Yes
5450
17
Approach 2: panel estimates



Panel estimates for a subset of 1,457 firms
sampled both in 2002 and 2005.
The panel allows us to introduce country
fixed effects and still estimate the effect of
information sharing (not its effect via
interactions with firm characteristics, as in
cross-sectional estimate)
Note: this is the first firm-level panel data
evidence with country effects to relate IS
and credit market performance!
18
Panel estimates
Information sharing index (IS)
Medium firm
Large firm
Transparency
Per capita GDP
Index of bank reform
Inflation
Observations
Access to finance
Cost of finance
0.121
(1.96)
-0.154
(1.67)
-0.103
(0.65)
0.034
(0.81)
0.009
(0.37)
0.257
(1.04)
-0.001
(0.86)
0.113
(2.40)*
-0.125
(1.20)
-0.085
(0.47)
0.046
(1.09)
0.037
(1.58)
0.047
(0.42)
-0.004
(4.57)**
1208
1218
19
Overall results




IS is associated with improved availability and lower
cost of credit.
Firm-level accounting transparency has qualitatively
similar effects on credit market performance.
IS and accounting transparency appear to be
substitutes at the firm level: effect of IS stronger for
more opaque firms.
Results appear robust to controlling for possible
spurious effects arising from other country-level
variables.
20