as lending shifts from soft to hard technology

BANK SIZE, LENDING TECHNOLOGIES, AND
SMALL BUSINESS FINANCE
Allen N. Berger
University of South Carolina
Wharton Financial Institutions Center
Lamont K. Black
Board of Governors of the Federal Reserve System
The opinions expressed do not necessarily reflect those of the Federal
Reserve Board or its staff. The authors thank Dan Grodzicki and Phil
Ostromogolsky for valuable research assistance.
1
OBJECTIVES
• Discuss current paradigm for small business
lending research.
– Focus on the most restrictive assumptions regarding
the lending technologies used by different sized banks.
• Show how we may relax these assumptions to
generalize the basic paradigm.
• Test implications of the current paradigm by:
– 1) Identifying the lending technologies used on bank
loans in the 1998 SSBF.
– 2) Analyzing comparative advantages of large and small
banks in using these technologies on firms of different
sizes.
• Draw some research and policy conclusions.
2
THE CURRENT PARADIGM IN A NUTSHELL
Bank size
Lending technologies
Large banks
Comparative advantage Advantage in
in hard-information
serving
larger,
technologies,
more
represented by
transparent
firms.
Firm size
financial statement
lending.
Small banks
Comparative advantage
in soft-information
Advantage in
serving
smaller
technologies,
usually just relationship
lending.
more opaque
firms.
3
OUR RELAXATION OF FOUR
ASSUMPTIONS OF THE PARADIGM
• 1) Allow for possibility that large banks may not
have comparative advantages in all hard
technologies.
– All technologies employ combination of hard and soft
information.
• 2) Relax (often implicit) assumption that hard
technologies as a whole may be represented by
financial statement lending.
– Hard technologies may be based primarily on other
types of hard information.
4
RELAXATION OF ASSUMPTIONS
(CONTINUED)
• 3) Allow the comparative advantage of large
banks in hard technologies as a whole to be
increasing or decreasing in firm size.
– Depends on the relative abilities of large and small
banks to employ hard versus soft technologies as firm
size increases.
• 4) Ease assumption that relationship lending is
only important soft-information technology.
– Judgment lending – Loan officer may use judgment
based on training and personal experience to lend to
firms without strong banking relationships and without
significant hard information.
5
IDENTIFICATION OF THE
LENDING TECHNOLOGIES
• Identify a technology by principal source of
information used to evaluate the loan.
• Main variables used in identification:
– Loan contract – contract type (lease versus loan), type
of collateral pledged (if any), and credit size.
– Firm – firm size and leverage.
– Firm owner – personal bankruptcy/delinquency.
– Relationship strength – combination of relationship
length and breadth.
6
PRINCIPLES USED IN IDENTIFICATION
PROCESS
• The bank chooses the lending technology that is most
efficient for that firm based on the available information.
– The bank generally chooses a hard-information technology over
a soft-information technology if hard information is available.
– Soft-information techniques tend to be labor-intensive.
• Lending based on the values of fixed assets (“immovables”)
that are leased or pledged as collateral is generally more
efficient than other hard-information lending technologies if
this collateral is available.
– Real estate, motor vehicles, and equipment.
– Strong incentive for firms to pay and bank can usually collect.
• Thus, we first identify the fixed-asset technologies (Step 1),
then other hard-information technologies (Step 2), then softinformation technologies (Step 3).
7
Step 1: Identifying Fixed-Asset Technologies
•
•
•
Identify over 50% of loans as made using fixed-asset lending technologies.
Very clean identification – uses very simple loan contract terms only.
High degree of certainty because of collection priority.
8
Step 2: Identifying Other Hard-Information Technologies
•
•
9
Identify about 30% of loans as made using other hard-information technologies.
Identification not as clean and certain – requires information on the firm and bank and our
intuition.
Step 3: Identifying Soft-Information Technologies
• Identify about 12% of loans as made using soft-information
technologies.
• Identification the least clean and certain.
– Requires that hard-information technologies total is accurate.
– Strong relationship is defined somewhat arbitrarily.
10
Frequency Distribution of Technologies Used by Banks to Lend to Small Businesses by Bank Size
Conditional on Lending Technology
Small Banks Large Banks
All Banks
GTA ≤ $1B
GTA>$1B
(4)
(5)
(6)
Hard-Information Lending Technologies
Fixed Asset Technologies
Leasing
Commercial Real Estate Lending
Residential Real Estate Lending
Motor Vehicle Lending
Equipment Lending
MV or EQ Lending
Fixed Asset Totals
LEASE
CRE
RRE
MV
EQ
MV or EQ
16.81
49.74
43.62
46.47
48.93
54.30
45.73
83.19
50.26
56.38
53.53
51.07
45.70
54.27
100.00
100.00
100.00
100.00
100.00
100.00
100.00
ABL
FSL
SBCS
27.48
32.27
0.00
18.01
72.52
67.73
100.00
81.99
100.00
100.00
100.00
100.00
36.37
63.63
100.00
75.00
65.43
68.75
25.00
34.57
31.25
100.00
100.00
100.00
Column Totals
40.16
59.84
100.00
Percent of Total
40.16
59.84
100.00
988
1472
2460
Other Hard Information Technologies
Asset-Based Lending
Financial Statement Lending
Small Business Credit Scoring
Other Hard Information Totals
Hard-Information Totals
Soft-Information Lending Technologies
Relationship Lending
Judgment Lending
Soft-Information Totals
RELATE
JUDGE
Number of Loans
• “Informal” test of comparative advantage – large banks have about 60% of loans, about 60% of bank branch offices.
• By convenience alone, expect about 60% of loans using each technology should be made by large banks under the null of no advantages.
• If >> 60%, then comparative advantage for large banks, if << 60%, then advantage for small.
• Large banks – significant comparative advantage in leasing, “other” hard technologies.
• Note: Advantage in SBCS is “engineered in” by assumption. Run regressions two ways to account for this.
• Small banks – significant comparative advantage in soft technologies.
11
METHODOLOGY – EMPIRICAL TESTS OF
CURRENT PARADIGM
• Logit model of probability that a given bank loan is made by
a large bank:
ln [P(loan is from a large bank) / (1 - P(loan is from a large bank))]
f(firm size, lending technology, firm size ▪ lending technology,
large bank branch market share, bank market concentration,
MSA dummy)
=
• Most general null hypothesis – no comparative advantage
or disadvantage of large banks in using any technology or
serving any size class.
– Coefficients of all the technology, firm size, and interaction
variables would be zero.
– P(large bank) determined only by competitive conditions.
• We interpret a significantly higher probability of a loan
being made by a large bank, conditional on competitive
conditions, as evidence of a comparative advantage for
large banks.
12
Table 6: Tests of Hard versus Soft Technologies
Dependent variable: Large Bank
(1)
Firm Size
Medium Firm
Large Firm
(2)
-0.428
[3.376]***
0.044
[0.347]
Technology
Hard
(3)
-0.442
0.182
[3.380]***
[0.549]
-0.157
1.603
[1.181] [3.293]***
1.389
1.344
2.073
[9.528]*** [9.003]*** [6.759]***
Interactions
Medium Firm * Hard
-0.753
[2.075]**
-1.927
[3.799]***
Large Firm * Hard
Market Characteristics
Large Bank Branch Market Share
Herfindahl
MSA
Pseudo R^2
Observations
•
•
•
(4)
3.300
3.293
3.33
3.341
[16.072]*** [15.788]*** [15.905]*** [15.896]***
0.571
0.544
0.466
0.457
[1.133]
[1.065]
[0.911]
[0.889]
0.294
0.334
0.307
0.297
[2.370]** [2.651]*** [2.430]** [2.341]**
0.135
2434
0.157
2434
0.161
2434
0.165
2434
(1) Nonmonotonic effect of firm size on probability of borrowing from a large bank. Inconsistent with
paradigm.
(2) Large banks have comparative advantage in hard – consistent with paradigm.
(4) Comparative advantage of large banks is decreasing in firm size. Essentially 0 for large firms.
13
Inconsistent with current paradigm’s prediction of increasing in firm size.
Table 6: Tests of Hard vs. Soft Technologies
(Panel B) Tests of Predicted Probabilities of Large Bank by Firm Size and Lending Technology
(1)
(2)
(3)
Soft
Hard
Hard - Soft
Small Firm
0.247
0.723
0.476
[6.76]***
Medium Firm
0.283
0.596
0.313
[6.76]***
Large Firm
0.620
0.654
0.034
[0.36]
F-Test for differences across size classes (2 restrictions)
•
•
•
14.46***
For small firms, predicted probability of loan being made by large
bank increases from 24.7% to 72.3% – statistically significant rise of
47.6% – as lending shifts from soft to hard technology.
For large firms, not statistically significant.
Decreasing comparative advantage for large banks in hard
technologies is statistically significant (F test).
– Bottom line – for small firms, large banks do well with hard technologies
and poorly with soft technologies. For large firms, technology does not
14
matter as much.
Table 8: Tests of Other Fixed-Asset Lending Technologies vs. Leasing
(Panel A) Regression Results
Dependent Variable: Large Bank
(1)
Firm Size
Medium Firm
Large Firm
(2)
-0.530
[2.949]***
-0.026
[0.143]
Technology
Commercial Real Estate Lending (CRE)
-1.439
[5.060]***
-1.154
[3.756]***
-1.246
[4.375]***
-1.361
[4.575]***
Residential Real Estate Lending (RRE)
Motor Vehicle Loan (MV)
Equipment Loan (EQ)
(3)
(4)
-0.431
[2.340]**
0.086
[0.451]
0.908
[1.336]
1.527
[2.323]**
-1.404
[4.900]***
-1.056
[3.401]***
-1.182
[4.128]***
-1.349
[4.503]***
0.065
[0.102]
0.178
[0.317]
-0.171
[0.318]
-0.510
[0.765]
Interactions
Medium Firm * CRE
-1.877
[2.281]**
-1.685
[2.177]**
-1.097
[1.481]
-1.351
[1.575]
-1.870
[2.335]**
-1.628
[1.994]**
-1.716
[2.359]**
-1.085
Medium Firm * RRE
Medium Firm * MV
Medium Firm * EQ
Large Firm * CRE
Large Firm * RRE
Large Firm * MV
Large Firm * EQ
Pseudo R^2
Observations
•
•
•
•
•
0.126
1269
0.135
1269
0.143
1269
0.151
1269
(1) Nonmonotonic effect of firm size on probability of borrowing from a large bank, inconsistent with
paradigm.
(2) Comparative advantage of large banks in leasing relative to other fixed-asset lending technologies.
(4) Comparative advantage of large banks in leasing relative to other fixed-asset lending does not hold for
smallest firms.
(4) Other differences in comparative advantage for larger firms by size class (e.g., equipment lending
15
interactions not significant).
Results are inconsistent with current paradigm that effectively treats all hard-information technologies as
if the comparative advantages were the same.
Table 9: Tests of Relationship Lending vs.
Judgment Lending
Dependent Variable: Large Bank
(1)
Firm Size
Medium Firm
Large Firm
(2)
0.139
[0.686]
1.585
[0.002]***
Technology
Relationship Lending (RELATE)
(3)
0.152
0.059
[0.661]
[0.880]
1.597
1.283
[0.002]*** [0.030]**
-0.102
[0.741]
-0.133
[0.678]
Interactions
Medium Firm * RELATE
•
0.204
287
0.169
287
0.204
287
0.207
287
RELATE and interaction terms are not statistically significant – cannot reject null of no
differences in comparative advantages between two soft technologies.
–
•
-0.562
[0.448]
0.425
[0.612]
1.130
[0.336]
Large Firm * RELATE
Pseudo R^2
Observations
(4)
May be due in part to small numbers of observations.
Finding suggests that there may be a significant bias in current research that just looks at
the effects of relationship strength – in effect groups judgment lending with hard
16
technologies.
CONCLUSION
• We relax some of the current paradigm’s most restrictive
assumptions regarding bank size, lending technologies,
and firm size.
• We show that:
– Large banks’ comparative advantages extend beyond lending
to large, transparent firms.
• Hard information is available in forms other than financial
statements.
• Large banks may be able to lend to opaque small firms using hard
information about the firm’s collateral or owner without using
significant hard information about the firm itself.
– Small banks’ comparative
relationship lending.
advantages
extend
beyond
• All lending technologies have both hard and soft components, so
small banks may have advantages in some hard-information
technologies based on the soft-information component.
• Small banks also have a comparative advantage in an important
soft-information technology that is neglected by the paradigm –
judgment lending.
17
CONCLUSION (2)
• Policy implications.
– The current paradigm implies that the consolidation of
the banking industry may result in reduced credit to the
smallest, least transparent firms, as large banks are
disadvantaged in serving these firms.
– When we relax some of the most restrictive
assumptions, we allow for the possibility that large
banks can and do lend to these firms using hardinformation lending technologies.
18