The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
orKing raper series
The Value of Relationships Between
Small Firms and Their Lenders
P aula R. W orthington
1
Working Papers Series
Research Department
Federal Reserve Bank of Chicago
December 1999 (W P -99-29)
FEDERAL RESERVE BA NK
O F CHICAGO
The value of relationships between small firms and their lenders
Paula R. Worthington
Economic Research Department
Federal Reserve Bank of Chicago
230 South LaSalle
Chicago, Illinois 60604
(312)322-5802
prw@frbchi.org
December 1999
This paper investigates the impact of the relationship between a small firm and its lender on the
interest rate paid by the firm. I examine detailed data on loans made by small business
investment companies (SBICs) to small firms between 1936 and 1991, and m y estimates imply
that, other things equal, the interest rate charged by an SBIC to a new small business customer is
between 40 and 50 basis points higher than the rate charged to a repeat customer. These results
offer solid evidence that relationships between small firms and their lenders can be quite
valuable.
JEL Classification:
G20
G21
Financial Institutions and Services
Banks, Other Depository Institutions
The opinions expressed in thispaper are those of the author and do not necessarily represent the views of the Federal
Reserve Bank of Chicago or the Federal Reserve System.
In tro d u c tio n
Commercial banks offer many products and services to the nation’s small businesses, and
bankers, economists, and policymakers alike have long recognized the special nature of the
relationship between banks and their small business customers. In broad terms, financial
intermediaries serving small firms must screen firms and projects, to identify worthwhile
investment opportunities; design investment contracts (debt, equity, or hybrid) between
themselves and the small firms; and monitor the performance of the firms over the lifetime of the
investments. This paper addresses one aspect of the financing contract, namely the rate of
interest charged on loans. In particular, I investigate whether a pre-existing relationship between
a small firm and its lender affects the rate of interest charged on a loan made to the firm.
In brief, I examine detailed data on loans made by small business investment companies
(SBICs) to small firms between 1986 and 1991, and I find strong evidence that relationships
matter: m y estimates imply that other things equal, the interest rate charged by an SBIC to a new
small business customer is between 40 and 50 basis points higher than the rate charged to a
repeat customer. This effect remains after controlling for many other risk factors that in principle
affect loan interest rates. The main advantage of this paper over previous research is its
treatment of a potentially serious selection issue: firms who obtain loans from SBICs (ie., the
firms for w h o m I observe a loan interest rate) may differ from firms who obtain nondebt
financing (via equity or hybrid instruments). If some unobservable firm characteristic affects
both the loan decision and the rate of interest charged, then the error terms of the "security
choice" equation and the "interest rate” equation will be correlated; hence, O L S coefficients in
the interest rate equation will be biased.
In the next section of the paper, I briefly review selected previous research on this topic.
Next, I describe the data and empirical approach used. Section 4 presents and discusses the
results, and section 5 concludes.
P rev io u s re s e a rc h
Theory offers several insights into the determination of the equilibrium interest rate
charged by financial intermediaries to their small business customers. The premise of many
models is the notion that a relationship can generate valuable information about borrower quality.
Initially, the financial intermediary may know little about the firm, which itself may be young
and small, with little formal track record. Because the underlying quality (riskiness, managerial
quality, or "type") of the firm is unknown, the intermediary may charge a high interest rate early
in a relationship.1 Over time, as the firm’s "type" is revealed to the lender, the interest rate falls,
as there is less pooling of high and low quality firms (e.g., Petersen and Rajan 1995).
A n alternative line of research emphasizes the potential nontransferability of the
information learned over time by the intermediary. As the lender learns more about its small
business customer, the lender’s ability to extract monopoly rents from the relationship increases,
since it is costly/impossible for the small firm to transfer this knowledge to an alternative lender.
In this case, interest rates may actually rise over the course of a relationship, as the monopoly
power of the lender increases (e.g., Sharpe 1990).
Empirical work on these issues has yielded mixed results. For example, Berger and Udell
'Of course, the intermediary’s ability to charge a high rate of interest will be limited by the
adverse selection issue: charging too high a rate could force the better quality firms out of the
market, leaving the intermediary with only the lower quality firms. In extreme cases, this
mechanism implies that the market dries up completely.
(1995) analyze data from the 1987 National Survey of Small Business Finance (NSSBF) and find
that the duration of the relationship signficantly affects both the interest rate paid on the loan and
the probability of pledging collateral. Petersen and Rajan (1994), who also use the 1987 NSSBF,
find that the simple presence, as opposed to duration, of a relationship with a bank has only a
small effect on the interest rate paid by a small business. Berger and Udell (1995) reconcile these
somewhat conflicting results by noting that their study includes only lines of credit, thus
excluding most loans likely to be for the purchase of tangible, easily valued assets (e.g., motor
vehicles, equipment, and so on); in contrast, Peterson and Rajan (1994) include all loans in their
analysis, not just "relationship-type" loans. More recent work by Cole (1998) using the 1993
N S S B F finds that having a relationship affects the availability of credit through the
approval/denial process but that the duration of the relationship is unimportant for credit
availability.
A related empirical finding concerns the impact of the number of lenders from which a
small firm borrows, or, more generally, the number of banks from which a small firm receives
any financial services. Petersen and Rajan (1994) find that loan rates are increasing in the
number of banks from which a small business obtains financial services, and Cole (1998) finds
that the probability of loan approval (credit availability) is decreasing in the number of banks
from which a small business obtains financial services. These results suggests that the value of a
relationship between a firm and financial intermediary may be diluted as the number of other
intermediaries involved rises.
D ata a n d em p irical a p p ro a c h
D a t a sources
This paper uses data provided by the U.S. Small Business Administration
on the activities of its SBIC program. SBICs are SBA-chartered and regulated financial
intermediaries intended to provide long-term debt and equity capital to nonfinancial small
businesses in the United States. The program, which was created in 1959, has numerous
distinctive features which have been discussed elsewhere (Brewer et al., 1999). For the present
paper, it suffices to note that SBICs can invest in debt, equity, or convertible securities of small
businesses; that SBICs can choose to borrow long-term funds that are guaranteed by the S B A
("SBA leverage") and that the ability to prepay these funds was severely limited during the time
period under study; and that SBICs can be "stand-alone" entities or can be subsidiaries of larger
firms.
By law, SBICs file a report with the S B A each time they fund a small business; this
report contains descriptive information about the small firms and the transactions themselves.
M y dataset consists of these reports for all transactions occurring between 1986 and 1991. I then
match these data to the annual reports of condition that SBICs must file with the SBA; these
reports provide detailed balance-sheet and income statement information for the SBICs.
Restricting the sample to those transactions for which we have nonmissing data on key firm
characteristics and SBICs’financial conditions yields a sample of 4807 transactions.2
B a s i c features o f the d a t a
Figure 1 shows that the spread between the average loan rate
charged by SBICs and the rate on a constant maturity 5-year Treasury bond fell over the sample
period; the overall average spread was 4.79%, with the level of the interest rate charged
2I
restrict the sample by excluding the largest firms (500 or more employees) and all
proprietorships and partnerships. Furthermore, I drop records for which the intended uses of
funds are research and development, marketing, or "miscellaneous."
averaging 12.81%.3 To identify some simple correlates of the spread, I graph in figures 2 through
7 average spreads for selected types of firms and transactions. Figure 2 shows that average loan
spreads were in general lower when the transaction involved a "repeat" SBIC-small business pair,
as compared to a "first-time" or "initial" pair. Figure 3 shows that syndicated loans were, on
average, more expensive than loans involving only one lender. The effects of the small firm’s
intended use of funds are shown in figure 4. N o obvious pattern jumps out, but some of the cells
are sparse, and I combine some of these categories for the empirical analysis below.4 Firm size
has no clear-cut impact on the loan spread, as evidenced by figure 5, nor does firm age (figure 6).
One relationship that does emerge clearly is that between S B A leverage and loan spreads: on
average, loans made by leveraged SBICs (i.e., those who had borrowed through the S B A ’s special
program) had higher spreads than those made by unleveraged SBICs.
Sample means
Table 1 reports sample means for all the variables used here. In the
sample, 20.7% of the transactions were syndicated, i.e., involved more than one SBIC on a given
date, and 67.5% of the transactions were "repeat" transactions. Average firm age is 7.3 years,
and the size distribution of firms, as measured by number of employees, points to somewhat
larger firms than are sampled by the NSSBF.5
Table 1 also offers some evidence on the selection issue discussed earlier: are firms
receiving SBIC loans systematically different from firms receiving other forms of SBIC
3I use the daily average Treasury bond rate for the week in which the financing occurred.
4In the analysis below, I combine four categories (plant modernization, new building,
machinery and equipment, and land) into one "tangible assets" category.
5RecaIl that I exclude sole proprietorships from the sample.
financing? The figures in table 1 indicate that, compared to nondebt transactions, debt
transactions are less likely to be syndicated or to be "repeat" transactions; go to smaller, older
firms, perhaps with less growth potential; are more likely to be for "tangible assets" projects; are
smaller in dollar amounts; and are made by larger, more leveraged SBICs. Table 1 offers ample
evidence that the small firms receiving debt financing from SBICs differ substantially from those
receiving nondebt (equity or convertible funding) from SBICs.
Empirical a p p r o a c h
The main issue raised by table 1 is the possibility that the debt and
nondebt transactions and firms differ in some unobservable characteristics as well as the
observables documented in the table. If this is so, then O L S estimates of the coefficients in a
regression of the loan spread on covariates will be biased. Consequently, Iproceed by estimating
a selection model, in which Ijointly estimate a security choice equation, in which the type of
investment (debt or nondebt) is chosen, and a loan spread equation.
I rely on previous research to identify likely correlates of the loan spread. In the loan
spread equation, I include indicator variables for syndication and "repeat" status; firm age (linear
and quadratic); firm size (through size class indicator variables, based on number of employees);
the intended use of funds; the funding amount; the one-digit sector of the recipient small firm;
the leverage ratio of the funding SBIC; the size of the funding SBIC; and year d u m m y variables.
In the security choice equation, I include all of the independent variables from the loan spread
equation, plus three additional variables I view as likely to affect the security choice decision but
not the spread decision: bank-ownership status of the SBIC; corporate vs. partnership form of
the SBIC; and public vs. private status of the SBIC. M y assumption that these characteristics of
the funding SBIC do not affect the loan spread is plausible only if I control for the SBIC’s cost of
funds in the spread equation—this is accomplished by including an SBIC-specific measure of
leverage (the ratio of SBA leverage to private capital).
Results and discussion
B a s ic resu lts
Table 2 presents the main results of the paper. Column 1 contains the OLS
estimates of the spread equation, while columns 2 and 3 present the selection model coefficient
estimates. Because selection models tend to be sensitive to the somewhat arbitrary inclusion or
exclusion of variables in the two equations, I will discuss both sets of results. The OLS estimate
of the "repeat pair" coefficient is substantial: it implies that having a preexisting relationship
lowers the interest rate spread paid by the firm by nearly 40 basis points. This effect contrasts
with Petersen and Rajan (1994), who found only a weak effect of a binary relationship variable.
As mentioned earlier, however, that study included all loans, not just those "relationship-loans"
most likely to be affected by a previous relationship between the lender and the firm. Table 2’s
results seem broadly consistent with previous research pointing to the importance of
relationships-e.g., Berger and Udell (1995), who found that the length of relationship strongly
influenced loan interest rates, and Cole (1998), who found that a binary relationship variable
affected credit availability, though the length of relationship did not matter.
The impact of syndication is also substantial: the OLS results suggest that syndicated
loans have sharply higher loan spreads than nonsyndicated loans—over 57 basis points higher.
This seems consistent with results of Cole (1998), who finds that having more than one lender
tends to lessen the availability of credit, and Petersen and Rajan (1994), who find that increasing
the number of banks from which a firm borrows raises the interest rate paid. In contrast, Berger
and Udell (1990) find a "participation" variable has no impact on loan spreads.
The selection model estimates of columns 2 and 3 also point to significant effects on
spreads from the relationship and syndication variables. In fact, the point estimates for both
coefficients are noticeably higher in the selection model—an 87 basis point impact from
syndication, and a 51 basis point effect from the relationship variable.
The effects of firm age and size are not clear cut, as might have been predicted from
figures 5 and 6. The OLS estimates imply that loan spreads rise with firm age, in conflict with
previous empirical research and with theoretical models linking older firms to fewer information
problems and so on. On the other hand, the selection model estimates imply that loan spreads
fall with firm age. Specifically, the latter estimates imply that the loan spread falls 2.7 basis
points (at the sample mean) for each additional year of firm age; this is somewhat smaller than
the 3.3 basis points per year impact measured by Berger and Udell (1995). The impact of firm
size on loan spreads is also unsettled, as neither model yields a clear pattern.
Previous research
has typically found that larger firms pay lower rates or spreads (e.g., Petersen and Rajan (1994)
and Berger and Udell (1995)).
Table 2 shows that the intended use of funds has a large impact on loan spreads. In the
OLS model, projects generating tangible assets, or loans used to acquire an existing business,
have lower loan spreads than (the excluded category of) debt consolidation.6 The selection
model estimates also imply that tangible assets associated with lower spreads, and they further
point to substantially higher spreads for operating capital (vs. debt consolidation). The
6Berger and Udell (1990) find that collateralized loans are riskier than others; hence they
find a positive relationship between collateralization and loan pricing. There is no conflict with
the current results, as a "tangible assets" intended use of funds doesn’t directly imply the loan is
collateralized. Other things equal, a loan which will be used to buy tangible assets will be less
risky than a loan for working capital.
magnitudes involved are economically significant: operating capital loans are priced 46 basis
points higher than debt consolidation loans, while loans intended to finance the acquisition of
tangible assets are priced over 87 basis points lower than debt consolidation loans. This seems
consistent with results of Berger and Udell, who view working (operating) capital as a
"relationship-type" loan, so higher risk than, say, a car loan.
Both OLS and selection model estimates imply that loan interest rate spreads are
increasing in loan size. This conflicts with Berger and Udell’s (1990) results and argues against
the view that scale economies in lending are important.
Finally, the estimates imply that leverage raised loan spreads during the 1986-1991
period. During this time period, SBA leverage was "expensive," since levered SBICs could not
prepay their SBA borrowings, despite substantial declines in the levels of market interest rates
from their early 1980s levels. The magnitude of the effect is, again, large: the coefficient
estimates imply that moving from a 1:1 SBA leverage/private capital ratio to a 2:1 ratio raises
loan spread between 46 (selection model) and 77 (OLS) basis points.
R o b u stn e ss ch ecks
How sturdy are the results in table 2? For example, a surprising
number of firms have a reported age of 0 in the sample (1.7%); this could be simple reporting
error (recording the same dates for the financing and birth of the firm), or it could reflect a
practice of executing several legal documents at one time (e.g., establishing a new corporation
and signing loan agreements on the same day). I find that dropping these records from the
sample yields little change to the results.
Another issue is the role played by the intended use of funds variable. It seems plausible
that the effects of other variables, e.g., the relationship or syndication variables, may differ
depending on the type of project the small firm is trying to finance. To consider this, I
reestimated the OLS model on two mutually exclusive subsamples: one which excludes
operating capital loans, and the other including only operating capital loans. I choose operating
capital loans for two reasons: first, they represent a large fraction of my sample’s loans. Second,
as Berger and Udell (1995) have discussed, the availability and/or pricing of working capital
loans seem quite likely to depend on the extent of relationship between the small firm and its
lender. Thus, such loans may show more sensitivity to the relationship and syndication variables
used here.
The results contained in columns 4 and 5 of table 2 offer some confirmation of this view.
Results for the sample excluding operating capital loans (column 4) show that the relationship
variable’s impact is negative but no longer statistically significant. This sample, which includes
loans to generate tangible assets, acquire existing businesses, or consolidate debts, apparently
includes loans for which having some preexisting lender relationship is not that important. In
contrast, column 5’s estimates for the operating capital loans sample show that loans involving a
repeat small firm-SBIC pair are priced 54 basis points lower than loans involving a first-time
pair. One puzzle arising from these results is the behavior of the syndication variable, whose
coefficient is significantly larger in the nonoperating capital sample than in the operating capital
sample.
Discussion and conclusions
This paper has offered evidence that firms borrowing from a lender for the first time pay
significantly higher spreads than do firms borrowing from a "repeat" lender. This suggests that
relationships with their lenders can indeed be valuable to small businesses, especially when they
are borrowing for general working capital needs. Furthermore, the value of those relationships
seems to decline as the number of lenders increases, as evidenced by the syndication’s variable
positive impact on loan spreads.
The results presented here are suggestive, but more research remains to be done. In
particular, refining some of the firm size results is important, as is developing more detailed
measures to control for macroeconomic factors affecting loan spreads. Perhaps the most
important task ahead is to estimate a completely different econometric model, one that still
addresses the selection issue but that is a bit more robust. In particular, by exploiting the
dataset’s time series dimension more fully, I may be able to estimate the spread equation in first
differences, which would "wipe out" any unobservable firm effect that otherwise biases the
coefficient estimates in the spread equation. Future revisions will pursue this line of research and
may offer even stronger evidence on the value of relationships between small businesses and
their lenders.
References
Berger, Allen N., and Gregory F. Udell. "Some evidence on the empirical significance of credit
rationing." J o u rn a l o f P o litic a l E co n o m y 1992(100): 1047-1077.
Berger, Allen N., and Gregory F. Udell. "Collateral, Loan Quality, and Bank Risk."
M o n e ta ry E c o n o m ic s 25 (1990): 21-42.
J o u rn a l o f
Berger, Allen N., and Gregory F. Udell. "The economics of small business finance: The roles of
private equity and debt markets in the financial growth cycle." J o u rn a l o f B an kin g a n d
F in an ce 22 (1998): 613-673.
Berger, Allen N., and Gregory F. Udell. "Relationship Lending and Lines of Credit in Small
Business Finance." J o u rn a l o f B u sin ess 68:3 (July 1995): 351-381.
Berlin, Mitchell, and Loretta J. Mester. "On the profitability and cost of relationship lending."
J o u rn a l o f B an kin g a n d F in a n ce 22 (1998): 873-897.
Brewer, Elijah HI; Hesna Genay; William E. Jackson; and Paula R. Worthington. "When
government guarantees don’t help: the role of adverse selection." Manuscript, Federal
Reserve Bank of Chicago, June 1999.
Cole, Rebel A. "The importance of relationships to the availability of credit."
B an kin g a n d F in a n ce 22 (1998): 959-977.
J o u rn a l o f
Petersen, Mitchell A., and Raghuram G. Rajan. "The Benefits of Lending Relationships:
Evidence from Small Business Data." J o u rn a l o f F in an ce 49:1 ;(March 1994): 3-37.
Petersen, Mitchell A., and Raghuram G. Rajan. "The Effect of Credit Market Competition on
Lending Relationships." Q u a rte rly Jou rn al o f E co n o m ics 110:2 (May 1995): 407-443.
Sharpe, Steven A. "Asymmetric information, bank lending, and implicit contracts: a stylized
model of customer relationships." Jou rn a l o f F in an ce 45 (1990): 1069-1087.
Table 1 Sample means 1986-1991
Nondebt
Full sample
Debt
4.80
N.A.
Syndication dummy
0.208
0.147
0.259
Repeat firm/SBIC pair
0.675
0.655
0.693
7.3
9.3
5.6
Spread over 5-year Treasury
Firm age (years)
1-3 employees
0.096
0.140
0.059
4-7 employees
0.115
0.161
0.077
8-19 employees
0.216
0.262
0.177
20-49 employees
0.241
0.208
0.268
50-99 employees
0.176
0.118
0.224
100-249 employees
0.123
0.095
0.147
250-499 employees
0.033
0.016
0.048
Operating capital
0.788
0.739
0.830
Tangible assets
0.068
0.107
0.034
Business Acguisition
0.075
0.050
0.096
Debt consolidation
0.069
0.104
0.039
Log of funding amount 82$
11.02
10.84
11.17
Manufacturing sector
0.476
0.376
0.561
Transportation & utilities secto
0.053
0.070
0.039
Trade sector
0.173
0.253
0.106
Services sector
0.222
0.196
0.245
Other sector
0.075
0.105
0.050
Corporate SBIC
0.650
0.738
0.575
SBIC owned by bank
0.335
0.233
0.421
SBIC publically held
0.120
0.153
0.091
SBA leverage/private K
1.41
1.90
1.00
Log (SBIC total assets)
16.52
16.59
16.48
Observations
4807
2204
2603
All differences in means are significant at the 1% level.
source: n4.log, December 9, 1999
Table 2 Determinants of loan spread, 1986-1991
Syndication dummy
Repeat firm/SBIC pair
Firm age (years)
Firm age (squared)/100
1-3 employees
4-7 employees
8-19 employees
20-49 employees
50-99 employees
100-249 employees
Operating capital
Tangible assets
Business Acquisition
Log of funding amount 82$
SBA leverage/private capital
Loq of SBIG assets
OLS
spread
0.574
(0.136)**
-0.398
(0.120)**
0.025
(0.010)**
-0.033
(0.016)*
0.932
(0.401)*
0.516
(0.399)
0.587
(0.386)
0.454
(0.387)
0.127
(0.396)
0.958
(0.402)*
0.017
(0.164)
-0.738
(0.202)**
-0.563
(0.258)*
0.198
(0.044)**
0.775
(0.042)**
-0.149
(0.040)**
Corporate SBIC
SBIC owned by bank
SBIC publically held
Constant
3.443
(0.849)**
Selection model
spread
selection eqn
0.870
-0.251
(0.146)**
(0.051)**
-0.510
0.101
(0.131)**
(0.049)*
-0.029
0.045
(0.012)*
(0.005)**
0.014
-0.035
(0.019)
(0.008)**
-0.618
0.988
(0.429)
(0.137)**
-0.915
0.915
(0.425)*
(0.134)**
-0.697
0.785
(0.408)
(0.126)**
0.491
-0.431
(0.401)
(0.124)**
-0.509
0.350
(0.407)
(0.127)**
0.361
0.245
(0.414)
(0.130)
0.460
-0.329
(0.190)*
(0.084)**
-0.875
0.224
(0.234)**
(0.112)*
0.154
-0.469
(0.287)
(0.110)**
0.329
-0.123
(0.048)**
(0.016)**
0.460
0.196
(0.021)**
(0.051)**
-0.243
0.086
(0.017)**
(0.044)**
0.187
(0.042)**
-0.114
(0.049)*
0.016
(0.060)
7.031
-1.260
(0.937)**
(0.327)**
2204
4807
Observations
0.26
R-squared
Standard errors in parentheses
Year and sector dummies included but not reported
* significant at 5% level; ** significant at 1% leve
source: n4.log, December 9, 1999
OLS
spread
1.609
(0.240)**
-0.130
(0.177)
0.006
(0.015)
-0.002
(0.023)
2.371
(0.871)**
1.494
(0.856)
1.490
(0.841)
(0.836)
0.946
(0.852)
0.645
(0.870)
OLS
spread
0.372
(0.161)*
-0.541
(0.153)**
0.028
(0.012)*
-0.036
(0.022)
0.426
(0.445)
0.151
(0.445)
0.312
(0.427)
0.314
(0.429)
-0.107
(0.440)
0.845
(0.445)
-0.723
(0.182)**
-0.573
(0.245)*
0.180
(0.085)*
0.377
(0.101)**
-0.308
(0.075)**
0.218
(0.052)**
0.821
(0.046)**
-0.069
(0.046)
5.326
(1.596)**
2.457
(0.950)**
576
0.30
1628
0.30
1.011
o (mean) spread2
a
(p 50) spread2
sra T a '
I n itia l
(mean) spread2
R epeat
Figure
By type of relationship
Interest rate spread
2
STaTa™
No-.syn
Yes-synd
(mean) spread2
6.33714
1.68027
i
86
Figure
i
67
i
88
By syndication status
3
Interest rate s p r e a d
r
89
PlntMod
BusAcq
(mean) spread2
QperKap
Figure
By intended use of funds
A
Interest rate spread
STaTa1
1-3 employees
4-7 employees
2 0 -4 9 employees
5 0 -9 9 employees
6-19 employees
(mean) spread2
i------ 1------ 1------ 1------1------r
100-249 employees
2 5 0 -4 9 9 employees
Figure
5
By size of firm
Interest rate
spread
S T a ra 1
zpeajds (ueauj)
I n fa n t
T een
Midage
Old
i.9 5 9 19 H
3 .4 8 9 7 1 “L
»
86
87
88
89
Figure
90
6
QA
66
By age of firm
rate
Interest
87
68
spread
89
lev d u m = = 0
lev d u ra = = l
(mean) spread2
5.73291
2.6083
Figure
By leverage status of SBIC
7
Interest rate s p r e a d
@FedFRASER