BETTER BORROWERS, FEWER BANKS?

Christophe J. GodlewskiFrédéric Lobez

Jean-Christophe StatnikYdriss Ziane

1

Outline

1. Introduction2. Literature3. Model4. Empirical design5. Results6. Discussion

2

Introduction

• Multiple bank relationships = common and significant economic phenomenon

• European firm has more than 5 bank relationships• Various (theoretical & empirical) arguments to

explain multiple banking / optimal number of banks• Monitoring / hold-up problem / external financing

sources diversification / limit bank liquidity risk…• This article: novel theoretical explanation based on

signaling + empirical validation (Europe)3

Literature

• What drives the optimal number of banks ?• Benefits / costs of an exclusive bank relationship• => Multiple banking can lead to …• [-] duplication of transaction costs + free riding in

monitoring (Diamond 1984)• [-] dissemination of strategic information to

competitors (Yosha 1995)• [-] less flexibility in loan terms setting (Dewatripont &

Maskin 1995)4

Literature (cont.)

• [+] mitigate the hold-up problem (Sharpe 1990, Rajan 1992)

• [+] reduce liquidity risk (Detragiache et al. 2000)• Multiple banking = pool of banks with different

structures• => + / - homogenous depending on relative power of

some pool’s members among others• Banking pools structure related to borrower quality /

information asymmetry / agency costs / coordination5

Literature (cont.)

• Multiple banking => weak monitoring / increases early project liquidation risk (Bolton & Scharfstein 1996)

• => smaller / concentrated pool => better monitoring (Elsas et al. 2004, Brunner & Krahnen 2008)

• => bank syndicate => mitigate coordination and moral hazard problems

• Negative relationship between syndicate size and borrower quality (Lee & Mullineaux 2004, Sufi 2007)

6

Model

• Economy

7

Banks

Investors

Managers

Model (cont.)

• Timeline

8

T=0 T=1 T=2

Investment in a risky project

(size 1)

Private informationon project’s success / failureÞ positive info. =>

project continuation

Þ negative info => strategic default & assets’ diversion

Project outcome=> k : probability x

=> 0 : probability (1-x)

Model (cont.)

• Firm’s financial structure• Investment financed by n potential banks• => n : observable by other investors• => μ(n) : monitoring by n banks • Manager’s utility function• 2 components• => firm’s market value : V(x)• => strategic default value

9

Model (cont.)

• Proposition• The number of banks in the pool = credible signal of firm’s

quality• Signalling equilibrium => size of the banking pool = decreasing

with the quality of the firm• Intuition• Signaling cost => greater monitoring by banks• Good quality firm’s manager is less sensitive to a tighter

monitoring than a bad quality firm’s manager• => Spence condition

10

Empirical design

• Data• Information on banking pools’ size + loan terms =>

Dealscan (Reuters)• Information on firms => Amadeus (Bureau Van Dijk)• Information on country level data => Beck et al.

(2007) + Djankov et al. (2007)• 3303 bank loans to 616 firms from 19 European

countries over the 1999-2006 period

11

Empirical design (cont.)

• Dependant variable = Number of lenders in the banking pool (mean = 8.79 / std dev. = 8.52)

• Main explanatory variable = empirical proxy for the borrower quality signal

• => use of bankruptcy / business risk indicator = Altman Z-score

• => X1= working capital / TA; X2= retained earnings / TA; X3= EBIT / TA; X4= equity / liabilities; X5= sales / TA

12

,

Empirical design (cont.)

• Different Z-score measures

13

,

Variable Definition Mean Std dev.

Z score (t) Altman (2000) Z score computed on the same fiscal year as the bank loan 1.9061 1.4641

Z score (t, S1)Altman (2000) Z score computed on the same fiscal year as the bank loan including loans granted on the first semester of the year only

1.9067 1.4767

Z score (t+1) Altman (2000) Z score computed on t+1 fiscal year with respect to the bank loan 2.0886 1.5866

Empirical design (cont.)

• Control variables

14

,

Loan sizeLogarithm of the loan facility amount in USD

Loan maturityLogarithm of the loan maturity in months

Syndication =1 if loan is syndicatedTerm loan =1 if loan is a term loanEbit margin EBIT / Operating revenue

Bank concentrationShare of 3 largest banks in total banking assets

Creditor rightsIndex aggregating creditor rights (0:poor creditor rights to 4)

Results

• Borrower quality => banking pool size (= Number of lenders)• OLS with standard errors clustered at borrower level / sector + year

dummies / coefficient for main variables displayed only

15

,

Variables Model 1 Model 2 Model 3Z score (t) -0.2824**

(0.1286)Z score (t, S1) -0.4691***

(0.1444)Z score (t+1) -0.2708

(0.4023)N 2474 1184 603R² 0.3843 0.4313 0.4599

Results (cont)

• Banking pool organization => banking pool size / borrower quality

16

,

Variables Model 1a Model 2a Model 3aZ score (t) -0.9887***

(0.2938)Z score (t, S1) -1.7015***

(0.4500)Z score (t+1) -1.3409**

(0.5920)Z score (t) x Syndication 0.7737**

(0.3024)Z score (t, S1) x Syndication 1.3564***

(0.4242)Z score (t+1) x Syndication 1.2649**

(0.5462)N 2474 1184 603R² 0.3787 0.4192 0.4539

Results (cont)

• Robustness checks• Regressions by firm and loan size• => large firms / loans = less information asymmetry

between firm and investors• => banking pool structure less informative• Split sample according to medians (TA & loan size)• => coefficient for Z score / interaction term remains

negative / positive but becomes weaker for large firms or large loans

17

,

Results (cont)

• Use of alternative European Z Score• Z scores as above computed with different

coefficients of the Z function • => re-estimation of the scoring function using same

variables as Altman but on a sample of 365 000 European firms

• [firm’s default defined by rating category and default probability provided by Amadeus]

• => similar results18

,

Discussion

• Alternative theoretical foundations for the existence of banking pools

• => signaling equilibrium model where firms voluntary limit asset substitution through smaller banking pool (better monitoring)

• Theoretical prediction = better firms borrow from fewer banks

• Empirical validation on a sample of more than 3000 loans to 600 European borrowers

• Use of Altman Z score to measure firm quality19

,

Discussion (cont.)

• Reduced size of the banking pool funding a loan to better quality borrower

• => banking pool structure = signal of borrower quality• Signal less important when• => coordination, hierarchy, and organization of the pool

are stronger (syndication)• => less information asymmetry between firm and

investors (large firms and loans)

20

,