Credit Market Imperfections: Theory and Empirics

1
Credit Market Imperfections Theory and Empirics

• Lecture 3 Credit market relations the optimal
  number of creditors

2
Contents

• Theoretical literature/overview of findings
• The Bolton-Scharfstein model
• Description of the data sources and descriptive
  statistics
• Models and estimation results
• Conclusions

3
Up to now

• we analyzed theory and empirics of transmission
  with special attention for credit
• we focused on macro, individual bank, and firm
  balance sheets
• we neglected the endogeneity of firm decisions
  to some extent

4
Now we..

• analyze one specific firm financing decision
  the optimal number of banks
• try to dive deeper into relationship lending,
  which is crucial to the credit view
• take a specific example Japan

5
Theoretical background (1)

• One of the key problems in finance is what is
  the optimal number of suppliers of financial
  capital? Or more in particular what is the
  optimal number of creditors?
• Should firms use arms length finance or
  relationship borrowing? Bonds or credit?
• And if credit is used, how many borrowing
  channels should be operated (if there is a
  choice, so maybe not for SMEs)?

6
Theoretical background (2)

• A detailed analysis of credit relations at the
  micro level is helpful in understanding the
  working of the macro credit channel
• If a firm has multiple credit lines, the
  probability of being rationed in equilibrium is
  probably lower
• If credit lines are thin, liquidity-rationed
  small firms will probably react stronger in a
  period of monetary contraction

7
Theoretical background (3)

• An analysis of credit relations maybe also sheds
  some light on competition issues in banking
  fewer credit lines probably coincide with a less
  competitive banking market
• In case of a financial crisis, based on bad loans
  (like Japan), a detailed analysis of the market
  for loans might be helpful

8
Empirical background (1)

• The case of Japan is interesting from multiple
  point-of-views (1) long-term credit is essential
  for financing rapid growth, (2) long-term debt is
  partly responsible for the current economic
  recession, (3) governance of loans in Japan is at
  least covered by smoke

9
Empirical background (2)

• Data sets on multiple bank loans are typically
  not widespread available
• For the Netherlands e.g. we only know the
  aggregated balance sheet credit totals per firm
• The data set has rather unique detailed
  information content

10
Goals

• Exploration of credit relations of Japanese
  listed firms. How many banks are involved, and
  what role does credit play in financing?
• What determines the number of credit relations?
  Are Japanese group structures relevant? Can we
  find a confirmation of the relevance of
  standard economic determinants?

11
Theory of the optimal number of bank relations

• Why do firms want to borrow from a single bank?
  There are five major classes of explanations
• I cost minimization (ex ante screening,
  monitoring, and ex post situations like debt
  renegotiations or bankruptcy). Examples Diamond
  (1984), Bolton-Scharfstein (1996)

12
Optimal number . (2)

• II competition on the banking market fewer
  banks give less choice. Examples Von Thadden
  (1994), Petersen and Rajan (1995).
• III liquidity insurance. A firm wants more
  credit lines if it faces higher liquidity risk.
  Nice example Detragiache et al. (2000)

13
Optimal number.(3)

• IV coordination of lending activities. A firm
  with a high default risk will observe an increase
  of the coordination costs. Example Dewatripont
  and Maskin (1995)
• V type of business. An innovating firm will try
  to keep its inventions secret to the market and
  look for fewer loan contacts. Examples Yosha
  (1995), Von Rheinhaben and Ruckes (1998).

14
Optimal number of creditors

• Two types of default liquidity (no cash) or
  strategic (cash diversion by the manager)
• Optimal contract should minimize the costs of
  financial distress, but it should also discourage
  firms from defaulting
• Bolton-Scharfstein model (1) model of
  liquidation threat (2) comparison of 1 and 2
  creditor borrowing

15
Bolton-Scharfstein (1)

• Two-period investment project invest K at date 0
  to purchase two physical assets A and B. At date
  1 the project returns a random cash flow x with
  probability ? or zero with 1-?.
• The manager has no wealth at date 0. If the
  manager continues the project the date 2 cash
  flow is y. There are no assets left at date 2

16
Bolton-Scharfstein (2)

• The project can be separated from the manager
  (liquidation) in which case there is no cash flow
  at date 2.
• The assets ca also be managed by another manager,
  generating ?y, ??1
• With complete financial contracts the project
  would never be liquidated and investors receive
  an expected payment K

17
Bolton-Scharfstein (3)

• Problem is that cash flows are not verifiable
  (but they can be observed). Managers are able to
  divert corporate resources
• Contracts can be made contingent on physical
  assets though
• Contract specifies that is the firm repays Rt at
  date t, creditors have the right to liquidate
  some fraction zt?1 with probability ?t?1

18
Bolton-Scharfstein (4)

• Special case is a standard debt contract if R1ltD
  z1?11, and if R1?D, z1?10, and z2?20
• A standard debt contract is not optimal too much
  liquidation
• Better contract if the manager makes repayment
  Ri for a cash flow ix,0, investors have the
  right to liquidate with probability ?i

19
Bolton-Scharfstein (5)

• Expected pay-off for the firm
  ?x-Rx(1-?x)y(1-?)-R0(1-?0)y
• Expected pay-off for the investor
  ?Rx?x Lx(1-?)R0?0L0-K
• Li liquidation value of the assets when the cash
  flow is ix,0.
• Payments cannot exceed funding R0?0 and Rx?x
• Incentive compatibility the manager must have an
  incentive to pay Rx if the cash flow is x
  x-Rx1-?x)y?x ?0S(1- ?0)y, where S is
  the managers utility from paying R0 where the
  CF0

20
Bolton-Scharfstein (6)

• Maximize expected firm profits, given the
  constraints and the fact that credit profits are
  nonnegative, leads to ?x0 and R00
• It is never optimal to liquidate if the firm
  makes the repayment Rx and the repayment is 0 in
  case there is no cash flow
• Next we need to determine ?0 and Rx

21
Bolton-Scharfstein (7)

• The IC-constraint turns into Rx ?0(y-S).
  Remember that there should be some positive
  probability of liquidation in the zero cash flow
  case
• Using this in the profit conditions we get
• ?0K/?(y-S)(1-?)L0, which must be smaller than
  1. The denominator is the maximum feasible gross
  profit of the creditors. The nominator is the
  investment outlay
• So now we have a full description of the contract

22
One creditor

• What is the liquidation value L0(1)? Suppose that
  the buyer of assets incurs costs c, unknown at
  date 0 but uniformly distributed on 0,cmax, to
  get control of assets
• If the assets are liquidated the buyer gets
  ?y/2ltcmax. The buyer will negotiate if c??y/2, so
  the probability of asset sale is ?y/2cmax, so
  L0(1) (?y/2cmax)?y/2

23
One creditor (2)

• Two parties (manager and creditor) split y
  equally S(1)y/2. It is supposed that the
  outside buyer does not participate here
• So now we can describe ?0(1)K/?(1)
• ?(1)?y/2(1-?)?2y2/4cmax
• So ?0(1) is decreasing in ? and ?. So there is
  less probability of liquidation for low risk
  cases and better reusability of assets. L0(1) is
  increasing in ?

24
Two creditors (1)

• Now we use the two assets. The firm uses capital
  from two creditors to buy A and B. Creditor a is
  secured by A and b by B.
• We assume that yA is the date 2 cash flow from
  using A without using B and yB equivalently
• Synergy is assumed ?y-yA-yBgt0

25
Two creditors (2)

• Creditor as Shapley value is ?yA/2??/3. The sum
  of all coalitions the agent might belong to. Idem
  for b ?yB/2??/3. How can we see this?
• With probability 1/3 creditor a is in a
  coalition with b and the buyer. Marginal value of
  a is ?(y-yB)
• With prob. 1/6 coalition with the buyer ?yA.
  With prob. 1/6 in a coalition with b no value
  (you need outside finance). With prob 1/3
  coalition with himself no value!
• Total creditor Shapley value ?y/2 ??/6

26
Two creditors (3)

• So the two creditors get more than a single
  creditor, due to the synergy value
  (complementarity)
• So the outside buyers Shapley value must be
  ?y/2-??/6 if he bargians. So he will bargain with
  probablity (1/cmax)?y/2- ??/6. This is lower
  than in the 1-creditor case!

27
Two creditors (4)

• L0(2) (1/cmax)?y/2- ??/6 (?y/2- ??/6)
  L0(1)-(?2?2/36cmax)ltL0(1)
• S(2) the manager will be the most efficient user
  of the assets and buy them back from the
  creditors. In that case S(2) y/2-?/6, so this is
  lower than S(1)!
• We know that ?0(2)K/?(y-S(2))(1-?)L0(2)

28
Two creditors (5)

• ?0(1)K/?(1) and ?0(2)K/?(2)
• ?(1)?y/2(1-?)?2y2/4cmax
• ?(2) ?(1)??/6(1-?)?2?2/36cmax
• So the two creditorsmaximum gross profit could
  be greater or less than the single creditors
  gross profit

29
Comparing one and two creditors

• Comparing the liquidation values the manager will
  always choose to borrow from a single creditor
  L0(2)ltL0(1)
• Low values of ?0 reduce the inefficiency
  originating from a liquidation probability
  compare ?0(1) with ?0(2)

30
Comparing 1 and 2 creditors (2)

• I the firm borrows from two creditors when
  default risk is low (high ?) and from one
  creditor when risk is high
• II The firm borrows from two creditors when
  asset complementarity (?) is low
• III The frm borrows from two creditors when
  outside buyers have a low valuation of the assets
  (? low)

31
Empirical findings

• Size and age mixed results
• Bank market concentration relevant
• Liquidity risk relevant
• Type of activity leads to significant results
  RD, home or foreign orientation, etc
• Financial structure mixed results

32
International evidence on the number of banking
contacts

• Very few for US small firms and firms in
  Scandinavia 1 or 2 (median) relations
• Large numbers for e.g. Italy 33!
• Ongena and Smith (2000) more relations if
  creditor rights are poor, legal systems are
  inefficient, lowly concentrated banking sector,
  and stable private bond markets
• What about Japan?

33
Japanese Loan Data

• Sources Development Bank of Japan. Financial
  Statement Data (2000 listed firms from 1957
  onward) and Sources of Long-Term Loans Data
• Long-term loans 1982-1999 (about 35 thousand raw
  data points) this gives a nice time-series
  feature!
• Short-term loans 1998-1999 (about 2 thousand raw
  data points)

34
Measurement

• We define the number of banking contacts (e.g.
  for long-term loans) as the number of banks that
  provided a long-term loan in year t. We do not
  check whether the same bank continues the loan in
  t1
• We dont go back to the individual loan itself!

35
Percentage of firms with a single bank long-term
loan relation
36
Mean and median number of long-term loans
37
Time-series pattern

• Decrease in the number of banking contacts during
  the boom of the 1980s
• Increase again after the burst of the bubble
• Apparently in times of prosperity there is a
  lower spread due to liquidity risk and cost
  arguments are more pronounced

38
Short-term loans (1998-1999)

• About 4 has a single short-term loan (is 10 for
  long-term loans)
• Mean number of relations is 8, median is 7
• Mean Herfindahl index 0.28 (median 0.22).
  Corresponding values for long-term loans are 0.38
  and 0.28
• So short-term loans are typically less
  concentrated

39
Descriptive statistics

• Number of contacts increases in size of the firm
  (no matter how we measure it)
• Manufacturing firms typically have fewer loan
  contacts (although the difference is not so big)
• It seems that more profitable firms have fewer
  bank contacts (no hard relation)
• Firms with more debt have multiple contacts

40
Special feature Main Bank Relations

• MBD1 if the largest equity owner is also the
  largest debt owner
• MBD2 if the largest equity owner resorts under
  the top-3 debt owners
• MBD3 if the largest equity owner resorts under
  the top-10 debt owners
• MBD4 if the largest debt owner resorts under the
  top-3 equity owners
• MBD5 if the largest debt owner resorts under the
  top-10 equity owners
• MBD6 if one of the top-3 equity owners resorts
  under the top-3 debt owners
• MBD7 if one of the top-10 equity owners resorts
  under the top-10 debt owners.

41
Modelling the loan decision

• Typical discrete choice models y1 for a single
  and y0 for multiple loans a logit-model, or a
  multinomial logit model that allows for multiple
  class choices. On the agenda an ordered
  multinomial logit model
• Side-result for a tobit model on the H-index
• We present results for long-term loans in two
  subperiods

42
Determinants

• Size (employees, total assets, sales, debt)
• Profitability ROA and Tobins q
• Solvability debt-to-assets
• Liquidity liquid assets/total assets
• Alternative financing forms
• Firm activity RD, exports/sales, industry
• Main-bank relations

43
Results for the single versus multiple banks
decision (1)

• Size rather unimportant (except for debt and
  total loan amount). We work on an age
  indicator.
• Highly profitable firms seem to opt for a single
  short-term loan, but want multiple long-term
  loans
• Low solvability leads to multiple bank contacts

44
Single versus multiple (2)

• Liquidity rich firms like a single long-term loan
  contact
• More corporate bond financing coincides with
  fewer loan contacts
• A main bank relation reduces the number of
  creditors
• RD-intensive firms had multiple loans during the
  bubble period!

45
Multinomial logit model

• Classes 1, 2-4, 5-7, 8-10, 11-15, gt16
• We get mixed results for two classes up to 8
  loans and over 8 loans
• For LT-loans firms with fewer than 4 loans want
  fewer loans, firms with over 4 loans want to
  increase this number (for ST-loans we find a
  threshold of 10 loans)
• More profitable firms want more LT-loans and
  fewer ST-loans

46
Multinomial logit (2)

• A high debt-to-asset ratio decreases the number
  of loan contacts for firms with less than 8 loans
  (but increases it above this value)
• Cash-rich firms reduce the number of long-term
  loans (no impact on ST-loans)
• RD-intensive firms generally prefer more
  long-term loans in the 1980s

47
Continuous y Herfindahl index

• Size matters large firms have lower
  concentration of loans
• More profitable firms have more long-term
  relations
• More debt and less liquidity lead to multiple
  loan contacts
• A main-bank relation leads to fewer loan contacts

48
Conclusions

• Japanese firms have multiple credit relations
  6-7 for long-term and 7-8 for short-term loans on
  average
• Main determinants of multiple loan contacts are
  (1) size of debt, (2) lack of liquid assets, (3)
  profitability, (4) a main bank relation, and (5)
  RD activity in the 1980s