Loan Officer Turnover and Credit Availability for Small Firms

Document Sample
Loan Officer Turnover and Credit Availability for Small Firms Powered By Docstoc
					Journal of Small Business Management 2006 44(4), pp. 544–562

Loan Officer Turnover and Credit Availability
for Small Firms
by Jonathan A. Scott

   This paper presents empirical evidence on the role loan officers play in facilitat-
ing small firm access to commercial bank loans. If loan officers use soft information
(for example, assessments of character, information from customers and suppliers)
to make lending decisions that would not otherwise be made on the basis of hard
information (for example, tax returns or financial statements), then, frequent
turnover in loan officers should be associated with an adverse effect on credit avail-
ability. This relationship is confirmed empirically using survey data of U.S. small
firms in 1995 and 2001, where loan officer turnover is positively related to the turn-
down rate on the most recent loan application. Although loan officer turnover could
be influenced by the turndown rate (for example, an owner changes banks and gets
a new loan officer as a result of a recent turndown), its negative effect on credit
availability persists under several different tests.

Background                                       more complete profile for credit deci-
   Small firm credit decisions are often          sions, especially for banks that rely on a
made on the basis of hard information            relationship banking strategy (Berger
that is easily quantifiable such as audited       and Udell 2002). Without the soft infor-
financial statements, credit bureau infor-        mation, many small firms could be
mation, or owner tax returns, and nonfi-          denied credit because of their limited
nancial or soft information, which is            operating history or incomplete financial
more difficult to quantify, such as assess-       statements, especially for proprietorships
ments of the owner’s character. The loan         or family-owned firms. In other words,
officer (or relationship manager) plays a         the numbers or hard information may
key role in producing and interpreting           not tell the entire story. Though there is
soft information that should provide a           some empirical evidence that nonfinan-

   Jonathan Scott is associate professor, Fox School of Business, Temple University and adjunct
scholar, National Federation of Independent Business Research Foundation.
   Address correspondence to: Jonathan A. Scott, Fox School of Business, Temple University,
Philadelphia, PA 19122. Tel: (215) 204-7605. Fax: (215) 204-1697. E-mail:

cial information affects small firm bank        using it to engage in profitable banking
credit availability (Cole, Goldberg, and       activities (for example, Boot 2000).
White 2004), the role of the loan officer       Private information can include any
in using soft information in small firm         information about the borrower not in
credit decisions is largely unexplored.        the public domain and includes both
   The paper provides the first direct          “hard” information (for example, tax
empirical evidence that shows an associ-       returns, customer lists) and “soft” infor-
ation between the role of loan officers         mation (for example, supplier comments
and credit availability for small firms.        about the borrower). Without the collec-
Increased loan officer turnover could           tion of private information, small firm
lead to decreased credit availability for      access to credit markets would be limited
small firms if the value of soft informa-       because of their information opacity.
tion in the credit decision is dependent       Information opacity refers to the inabil-
upon the loan officer’s interpretation.         ity of lenders to completely understand
Small firm owners need to understand            the risk characteristics of the enterprise,
this relationship, especially in the United    possibly due to short-operating histories
States where a consolidating banking           or lack of complete financial informa-
industry has resulted in fewer commu-          tion. With this private information, banks
nity banks serving small firms, and an          will be able to write contracts that
increased reliance by larger banks on          improve credit availability and/or terms
credit-scoring techniques for small busi-      that would not otherwise take place in
ness loans—the antithesis of relationship      its absence (Berlin and Mester 1999;
banking.                                       Bhattacharya and Chiesa 1995; Petersen
   Firm-level data are used from the 1995      and Rajan 1995; Rajan and Winton
and 2001 Credit, Banks, and Small Busi-        1995).
ness Surveys of U.S. small businesses that        Berger and Udell (2002) have refo-
are members of the National Federation         cused the definition of relationship
of Independent Business (NFIB) to test         banking on private information gathered
what role loan officers have on small firm       by the lender and how this information
credit availability. One of the survey         is communicated within the organiza-
variables, a report of account manager         tion. They describe relationship banking
turnover, is used as a proxy for soft infor-   as a lending technology where the loan
mation production. Account manager             officer produces soft information, that is,
turnover and proxies for hard informa-         generally nonquantifiable information
tion (for example, length of banking rela-     obtained through interactions with the
tionship), along with control variable for     firm, its owner, suppliers, customers, and
firm risk characteristics (size, years in       the community. Soft information can
business) and market structure (merger         include assessments of the owner’s char-
activity, market size, and deposit con-        acter, assessments of managerial ability,
centration), are used to explain the most      leadership characteristics, or ability to
recent loan application outcome at com-        handle business adversity. Soft informa-
mercial banks.                                 tion has little value without the loan
                                               officer’s interpretation because the loan
                                               officer is the only person who can verify
Relationship Lending:                          it (Stein 2003). Hard information,
Theory and Evidence                            however, such as financial statements,
   Relationship lending in the literature      tax returns, and observations of bank-
generally refers to the process of col-        ing activity (checking accounts, other
lecting private, customer-specific infor-       financial service usage) can be easily
mation on potential borrowers, and then        transferred throughout the lending

                                         SCOTT                                        545
organization and can be analyzed inde-        lead to fraud), or unexpected health
pendent of the gatherer, who could be         problems for an owner that lacks man-
the loan officer. Hard information can         agement depth—all factors that might
also include public information such as       lead to turndown. Finally, some banks
years in business or Dun & Bradstreet         could turn down applications based on
ratings. Thus, the loan officer adds more      noneconomic factors such as ethnicity or
value to the production of soft informa-      gender. Though some empirical work
tion than hard information.                   supports the existence of these noneco-
    Actual lending decisions by banks can     nomic factors, more recent evidence sug-
rely on both hard and soft information.       gests that premarket forces play an
A relationship banking “technology”           important role in statistically observed
would rely on both hard and soft infor-       discrimination (for example, Cavalluzzo,
mation, where the interpretation of the       Cavalluzzo, and Wolken 2002; Robb and
soft information by the loan officer is a      Wolken 2002). Thus, the importance of
critical input to the lending decision. A     soft information to small firms in the
loan approval process that relies on a        credit granting process is ultimately an
credit-scoring technology is driven           empirical issue.
strictly by hard information (private and        Empirical work related to the impor-
possibly public information such as the       tance of banking relationships for small
years in business), the antithesis of rela-   firms uses firm-level data in both the
tionship lending. From the bank per-          United States and Europe, and focuses on
spective, they have a choice of lending       the outcome of the most recent loan
technology and must determine the             request and loan terms (rate and inci-
trade-offs in developing their lending        dence of collateral), using length (or
strategy. Ultimately, the ease and objec-     duration) of the banking relationship, the
tivity of credit scoring versus more time     scale of the relationship (for example,
intensive soft information production         the number of lenders), and the scope of
have to be evaluated against the interest     the relationship (for example, the
rates that can be charged and the bank’s      number of products used as the primary
competitive position in their local           quantitative proxies for the strength of
market. From an owner perspective, they       the relationship) as proxies for strength
must choose the bank with a lending           of banking relationships. This work has
technology that maximizes their chance        frequently found that both credit avail-
of obtaining a reliable funding source.       ability and loan terms improve as the
    The use of soft information may not       length of the banking relationship
always be beneficial to small firms. A          increases (for example, Degryse and Van
bank could decide the owner is not very       Cayseele 2000; Cole 1998; Harhoff and
adaptable in the face of a new competi-       Korting 1998; Berger and Udell 1995;
tive threat, and, despite acceptable finan-    Petersen and Rajan 1994). Increases in
cial performance, turn down the next          the scope of the relationship have been
loan as a result of this soft information.    found in some cases to improve credit
The bank could decide to exit a particu-      availability and loan terms, whereas
lar line of business (for example, auto       increases in the scale of the relationship
leasing), in which case no amount of soft     have been found to have an adverse
information would alter the lending deci-     effect (for example, Degryse and Van
sion. The bank could become aware of          Cayseele 2000; Cole 1998; Harhoff and
an impending change of control from           Korting 1998).
an intergenerational transfer, a revised         None of the previous work cited
assessment of control systems that are        specifically examines how soft informa-
inadequate to support growth (that could      tion affects credit market outcomes for

small firms. Harhoff and Korting (1998)         control factors for firm risk characteris-
find a negative relationship between            tics and market structure.
trust and cost of borrowing and inci-
dence of collateral, but no significant         Data and Methodology
relationship with credit availability             The data used in this study come from
(although the variable has the correct         the 1995 and 2001 Credit, Banks, and
sign). They note, however, that the            Small Business Surveys conducted by the
measure of trust and outcomes may not          NFIB. The purpose of the surveys is
be independent, which complicates the          to collect information about the credit
interpretation of their results. In other      market experiences of a random sample
words, there is no way to determine if         of the NFIB’s 600,000 members. Details
the high trust or quality rankings are the     about the surveys can be found in Scott,
result of good credit quality or vice versa.   Dunkelberg, and Dennis (2003) and
Two recent papers examine the role of          Dunkelberg (1998). In the 1995 survey,
soft information in small firm lending          18,000 questionnaires were sent in the
decisions by focusing on the role of bank      initial mailing and 3,642 completed ques-
size (Berger et al. 2005; Cole, Goldberg,      tionnaires were available after the second
and White 2004). Both papers conclude          mailing for a response rate of 20 percent.
that larger banks are more likely to rely      In the 2001 survey, questionnaires were
on hard information in lending decisions       mailed to 12,500 firms and responses
to small firms, but neither investigates        were received from 2,223 after two mail-
the role of the loan officer in the lending     ings for a response rate of 18 percent.
decision.                                         The Appendix compares the distribu-
   This paper adds to the literature on        tion of the 1995 and 2001 NFIB sample
relationship banking by providing evi-         to the distribution of the 1993 and 1998
dence on the role of the loan officer and       National Survey of Small Business
soft information production in the credit      Finance (NSSBF) conducted by the Board
approval process. The role of the loan         of Governors of the Federal Reserve
officer is especially important in the          System. In both surveys the NFIB sample
United States because there are many           tends to be slightly larger than those in
community banks without the resources          the NSSBF survey in terms of employees,
to construct sophisticated internal rating     sales, and assets. The 2001 NFIB survey
systems based on hard information,             tends to be a somewhat older, less urban
nor the management depth to transfer           sample than the 1995 survey.
problem customers to a workout group              The number of questionnaires used in
under the supervision of another lending       the empirical analysis is restricted to
officer. If the value of soft information       those firms applying for a loan from a
lies with the loan officer and cannot be        commercial bank, which reduced the
easily transmitted within the organiza-        1995 survey sample to 2,300 observa-
tion (large or small), then an increase        tions and the 2001 survey sample to
in loan officer turnover should have an         1,126 observations. The restriction is
adverse effect on credit availability, all     made to eliminate potentially unwanted
else equal. The empirical tests will inves-    variation attributable to different dynam-
tigate how the outcome of the most             ics in finance company, credit union, or
recent loan application by independent         personal lending markets. The definition
samples of U.S. small firms in 1995 and         and summary statistics of the key vari-
2001 is related to loan officer turnover        ables used in each survey in this study
experience (a proxy for soft informa-          are shown in Table 1. Not all of the vari-
tion), length of time at the primary bank      ables in Table 1 have 2,300 (1995 survey)
(a proxy for hard information), and other      or 1,126 (2001 survey) observations

                                         SCOTT                                       547
                                                                              Table 1

                                            Variable Definition and Summary Statistics of National Federation of Independent
                                                                      Business Survey Samples
                                       Variable                       Definition                           1995 Survey                    2001 Survey
                                                                                                        Applied for Loana              Applied for Loana

                                                                                                 Mean    S.D.b    Number of     Mean     S.D.     Number of
                                                                                                                 Observations                    Observations

                                       Soft Information Production Proxy
                                       Loan Officer        “Within the last 3 years, how          1.73     0.89       2,268      1.74     0.92        1,088
                                         Turnover           many different account
                                                            managers have you dealt
                                                            with at your primary financial
                                                            institution?” 1 = one; 2 = two;
                                                            3 = three; 4 = four; 5 = five
                                                            or more
                                       Other Relationship Proxies
                                       Length of Time     “When was the last time you            4.97     1.60       2,279      4.90     1.33        1,078
                                         at Bank            changed principal financial
                                                            institutions?” 1 = within the last
                                                            year; 2 = 1–2 years ago; 3 = 2–3
                                                            years ago; 4 = 3–4 years ago;
                                                            5 = 4–5 years ago; and 6 = more

                                                            than 5 years ago
                                       Number of          Total financial institutions used       1.88     1.17       2,285      1.58     0.86        1,106
                                         FIs Used           to obtain financial services
                                       Credit Availability Proxy
                                       Turndown             1 if the respondent reported that    0.17     0.38       2,330      0.12     0.32        1,126
                                                              they did not get a loan the last
                                                              time that they tried at a
                                                              commercial bank
        Bank/Market Characteristics
        Herfindahl Indexc Herfindahl–Hirshman index of             2,202   1,126   2,328   2,374   1,403   1,121
                            deposit concentration computed
                            by county
        Rural Location   1 if the firm is located in an MSA       0.63    0.48    2,330   0.59    0.49    1,126
                            area (0 = non-MSA area)
        Bank Recently    “During the last 3 years, was your      0.27    0.44    2,287   0.35    0.48    1,126
          Merged            principal financial institution
                            bought out or absorbed by
                            another? 1 = Yes
        Firm Characteristics
        Years in Business   Years in business                    15.3    12.7    2,314   18.6    13.6    1,096
        FTE Employees       Full-time equivalent employees       17.2    45.3    2,330   19.8    58.1    1,088
        Sales Growth        “Which category best describes       2.80    1.23    2,210   2.72    1.17    1,075
                              your average annual change
                              in your gross sales over the
                              past 3 years?” 1 = declined

                              more than 5 percent; 2 = no
                              change (−5 to +5 percent);
                              3 = grew 6–10 percent;
                              4 = grew 11 to 20 percent;
                              5 = grew 20 percent or more
        Noncorporate        1 if organized as a proprietorship   0.34    0.47    2,298   0.30    0.23    1,118
          Business            or partnership
        S-Corporation       1 if organized as an S-corporation   0.22    0.41    2,298   0.27    0.45    1,118
        Corporation         1 if organized as a corporation      0.45    0.50    2,298   0.42    0.49    1,118
        Agriculture         1 if a firm’s primary activity is     0.10    0.30    2,229   0.08    0.27    1,118
        Construction        1 if a firm’s primary activity is     0.14    0.35    2,229   0.15    0.35    1,118
        Finance             1 if a firm’s primary activity is     0.06    0.24    2,229   0.06    0.24    1,118
                              financial services

                                                                                                 Table 1
                                       Variable                      Definition                             1995 Survey                     2001 Survey
                                                                                                         Applied for Loana               Applied for Loana

                                                                                                  Mean    S.D.b    Number of     Mean      S.D.     Number of
                                                                                                                  Observations                     Observations

                                       Manufacturer        1 if a firm’s primary activity   is     0.14     0.35       2,229       0.13     0.34        1,118
                                       Professional        1 if a firm’s primary activity   is     0.05     0.22       2,229       0.06     0.25        1,118
                                         Services            professional services
                                       Retail              1 if a firm’s primary activity          0.21     0.41       2,229       0.20     0.40        1,118
                                                             is retail
                                       Nonprofessional     1 if a firm’s primary activity   is     0.19     0.40       2,229       0.16     0.36        1,118
                                         Services            nonprofessional services
                                       Transportation      1 if a firm’s primary activity   is     0.04     0.18       2,229       0.04     0.19        1,118
                                       Wholesale           1 if a firm’s primary activity   is     0.07     0.26       2,229       0.12     0.33        1,118

                                         The sample is limited to those firms that applied for a loan at a commercial bank.
                                         S.D., standard deviation.
                                         The Herfindahl Index for the 1995 survey is computed using the Federal Deposit Insurance Corporation’s June 1994 Summary
                                       of Deposits Report; the 2001 survey uses the June 2001 report.
because the “no answer” responses are        cal studies: length of time at the owner’s
excluded in the summary statistics.          primary bank and the number of finan-
                                             cial institutions (FIs) used, which
Model and Variable Definitions                includes both banks and nonbanks.
  The general reduced form model             Almost two-thirds have been with their
underlying the empirical tests is            primary bank for six or more years, but
                                             11 percent had changed within the past
      Credit availability                    two years in the 1995 survey; about
        = f {soft information, hard          40 percent reported never changing,
          informati on, control              whereas 10 percent reported changing in
          variables}.                  (1)   the past two years. Using this truncated
                                             distribution (that is, all firms over five
The proxy for credit availability is the     years take the same value of 6), the 1995
incidence of turndowns from the last         mean response is 5.1 years and 4.9 years
loan application, which takes a value of     for the 2001 survey. Length of time is
1 if the firm was turned down on its last     negatively correlated with loan officer
loan application, and 0 if otherwise.        turnover in 1995 (r = −.131), but there is
Seventeen percent of the 1995 sample         no association in 2001 (r = .037). The
was turned down, as compared to 15.5         potential effect of this correlation is
percent in the 1995 NSSBF data used by       examined in one of the sensitivity tests.
Cole (1998), and approximately 12               Length of time at the primary bank
percent in the 2001 NFIB sample.             can also capture the loan officer’s success
   The soft information production           in collecting soft information over time
proxy is loan officer turnover, the           and should vary inversely with the turn-
number of different loan officers the         down rate on the most recent loan appli-
owner was assigned in the three years        cation. However, length of time could
prior to the survey at their primary bank.   also reflect the accumulation of hard
Half of the owners in the 1995 survey        information over time (for example,
saw no change in their loan officer in the    audited financial statements) so that the
past three years, whereas over 45 percent    variable by itself does not provide an
had two to three loan officers, 3 percent     unambiguous proxy for the value of soft
dealt with four loan officers, and 1          information production. However, the
percent dealt with five or more. The 2001     effect of loan officer turnover may dimin-
distribution was very similar, with 48       ish with length of time as soft informa-
percent reporting one loan officer, 48        tion becomes relatively less important as
percent reporting two to three loan offi-     hard information accumulates. This
cers, three percent reporting four loan      migration of soft to hard information is
officers, and one percent reporting five       examined in an alternative specification
or more. If the value of soft information    of the model.
production lies with the loan officer—           The number of FIs used to obtain
and the loan officer uses this information    financial services is the other proxy for
to make lending decisions that would not     the strength of banking relationships.
otherwise be made on the basis of hard       Forty-six percent of the respondents use
information—then an increase in loan         just one FI whereas 15 percent use four
officer turnover should be positively         or more, with a mean of 2.6. The 2001
related to being turned down on the last     survey saw a reduction in the number of
loan application.                            banks used, possibly because of the con-
   The surveys also have two other vari-     tinuing consolidation of the U.S. banking
ables that have been used as banking         system. Fifty-nine percent reported using
relationship proxies in previous empiri-     just one bank and only three percent

                                       SCOTT                                       551
reported using four or more, with a mean       wise might be spuriously picked up by
value of 1.6.                                  the relationship proxies. For example, if
    The number of FIs used captures the        an increase in loan officer turnover is
incentives that a loan officer has to invest    associated with a higher turndown rate,
in the production of soft information.         it could be attributed to credit admin-
The use of more banks by a small firm           istration or organizational complexity
increases the chance that any one bank         problems associated with the merger and
investing in the production of soft infor-     not due to the change in loan officers.
mation will not be able to realize an ade-     Although the survey includes information
quate rate of return on their investment       on bank size, it is the size of the current
if the firm plays one bank off against          bank of the respondent and not the bank
another, similar to the free-rider problem     that might have turned the firm down in
identified by Cole (1998). Alternatively,       its last attempt. Consequently, bank size
owners could be using more than one            is not included as a control variable.
bank if they do not want to concentrate            Four firm characteristics are included
all their borrowing with a large bank to       in the model: firm size (full-time equiva-
avoid the hold-up problem or because a         lent [FTE] employees), years in business,
single bank does not have enough capital       sales growth, industry, and organiza-
to meet their total borrowing needs.           tional form. Firm size and years in busi-
There is no a priori reason to believe         ness serve in dual roles. The first role is
that this effect has a time dimension.         to serve as a rough credit risk proxy,
However, the relationship of loan officer       where larger, older firms presumably are
turnover to the turndown rate may be           less risky and obtain better outcomes.
complicated by multiple bank relation-         They also serve as measures of informa-
ships. For example, if an owner is initially   tion opacity, where smaller, younger
turned down and seeks a loan at another        firms are assumed to have more infor-
bank in their network of relationships,        mation problems and therefore benefit
the effect of loan officer turnover at their    more from stronger banking relation-
primary bank on credit availability will be    ships. The effect of years in business and
attenuated. This issue is also addressed as    firm size as sources of public information
part of an alternative model specification      on credit market outcomes is expected to
in the sensitivity tests shown next.           decline over time as little new informa-
    Two categories of control variables are    tion is added after the firm becomes a
shown in Table 1: bank/market char-            viable entity, and thus the log of each
acteristics and firm characteristics.           variable is used (for example, see Cole
Bank/market characteristics include a          1998; Berger and Udell 1995). Other
Herfindahl index of deposit concentra-          control variables include sales growth as
tion computed by Metropolitan Statistical      the owner’s report of annual average
Area (MSA), rural location, which takes a      growth over the past three years; one-
value of 1 if the firm is located in a rural    digit standard industrial classification
area (non-MSA) and 0 if in an MSA, and         (SIC) codes and form of business are
a 1/0 variable for whether or not the          included as categorical variables with
owner’s primary bank merged within the         noncorporate businesses (partnerships
past three years. Deposit concentration        and proprietorships) and retail firms are
and market size are frequently used in         the omitted categories.
the empirical literature that examines
the determinants of relationship lending       Hypotheses and Research Design
(for example, Petersen and Rajan 1995).           Logistic regression is used to estimate
These variables also control for important     the following reduced form specification
market structure influences that other-         of the general model in equation (1).

Turndown i = α 0 + α1Loan officer                 The primary estimation challenge is
  turnover + α 2Length of time at             the potential lack of independence (or
  primary bank + α 3No. of FIs used +         endogeneity) between loan officer
                                              turnover and the turndown rate, which
  α 4Primary bank recently merged +
                                              complicates the interpretation of α1 in
  α 5Rural location + α 6Herfindahl +
                                              equation (2). This lack of independence
  α 7 Years in business + α 8FTE
                                              may arise from several sources. First,
  employment + α 9Sales growth +              owners may successfully lobby a loan
  α10Form of business +                       officer to approve their loan request,
  α11Industry + ε I .                 (2)     when all other information would
                                              suggest a rejection of their loan request,
The primary null hypothesis is α1 = 0. If     possibly because the loan officer bene-
the null hypothesis is rejected and α1 >      fits from making the loan whereas the
0, then the positive effects of soft in-      lending institution does not (for
formation production by loan officers          example, loan officer bonuses based on
dominate the experience of small firms         loan volume without having an adequate
in these surveys. However, if the null        stake in the ownership of the bank).
hypothesis is rejected and α1 < 0, then       Another example of endogeneity would
the negative effects of soft information      be if risky firms that had their most
production dominate the experience of         recent loan request rejected were moved
small firms in these surveys.                  to another loan officer with less experi-
   The other specification of the model        ence or to a loan workout area under the
tests for the time dimension of the loan      supervision of a different loan officer.
officer’s investment in soft information.      This outcome may be more likely at
                                              larger banks with the capacity to support
Turndown i = α 0 + α1Loan officer              loan workout groups, but less likely at
  turnover × long time at primary             community banks where there may be
  bank + α 2Loan officer turnover ×           only one lending officer. A final example
  short time at primary bank +                is where a firm rejected in the their
  α 3No. of FIs used + α 4Primary             most recent request moves to another
  bank recently merged + α 5Rural             bank, again leading to a situation where
  location + α 6Herfindahl + α 7 Years         denial leads to higher loan officer
  in business + α 8FTE employment +           turnover and the destruction of soft
  α 9Sales growth + α10Form of                information.
  business + α11Industry + ε I .      (3)        Three approaches are used to mitigate
                                              the bias from estimating a reduced form
The interaction terms are defined as           model shown in equation (2). The first
follows: long time at primary bank takes      approach is to use a subsample of firms
a value of 1 if length of time at primary     that do not change banks to control for
bank is five or more years; and short time     the situation where poor credit availabil-
at primary bank takes a value of 1 if         ity leads to higher turnover when a firm
length of time at primary bank is less        leaves to search for a new bank. The
than five years. The expectation is that α1    second is to find a set of instruments for
and α2 > 0 and α1 < α2, that is, longer       loan officer turnover that is strongly cor-
time with the primary bank should result      related with the soft information proxies,
in a weaker soft information effect,          but not correlated with the error terms
reflecting the migration of soft to hard       in equation (2). If the structural model is
information. A Wald linear restriction test   not identified, then it is well known that
is used to test the null hypothesis that      the estimation of the reduced form
α1 = α2.                                      model in equation (2) gives inconsistent

                                        SCOTT                                        553
estimates of α1 because loan officer           change in the independent variable ×
turnover is correlated with the error term    coefficient × p (1 − p) where p is the
(see Greene 2002). The last approach is       mean of the dependent variable reported
to test whether or not the coefficient on      in Table 1.
loan officer turnover is a function of            The estimate of equation (3), shown
owner risk (size or years in business).       in column (2) for 1995 and column (6)
The problem addressed here is whether         for 2001 in Table 2, reject the null
or not younger, smaller firms are more         hypothesis that the loan officer turnover
likely to experience higher turnover          effect on the turndown rate is inde-
because they are shifted to workout           pendent of the length of time at the
groups within the banking organization.       bank. The longer the relationship at the
If so, the coefficient on turnover would       owner’s primary bank, the less the effect
be higher for these younger firms than         loan officer turnover has on credit avail-
for older firms.                               ability. The differences in these coeffi-
                                              cients are statistically significant for both
Analysis of Results                           survey years. This outcome is consistent
Reduced Form Estimates                        with the migration of soft information to
   The reduced form estimate of equa-         hard information over time, reflecting a
tion (2) is presented in Table 2, column      growing importance of hard information
(1) for 1995 and column (5) for 2001. The     relative to soft information in credit
coefficient on loan officer turnover is sig-    granting decisions.
nificant and positively related to the turn-      The primary firm risk factors, years in
down rate as expected in each survey.         business and size also vary as expected.
Thus, the net effect of loan officer           The chance of a turndown decreases
turnover, the proxy for soft information      with increases in firm age and FTE
acquisition, is positive, with firms expe-     employment—an outcome also found by
riencing lower turnover having a lower        Cole (1998) and others. However, the
change of a turndown after controlling        negative association with years in busi-
for firm risk and market structure. In         ness is not significant in 2001. Firms in
addition, the significance of the turnover     rural locations are less likely to experi-
effect is independent of the length of        ence a turndown, but those in more
time at the current bank, which has a         concentrated deposit markets are more
negative and significant association with      likely to do so, although the latter effect
the turndown rate. The other relationship     is only significant in 2001.
strength variable, number of FIs used, is        What do these results mean for a small
not significantly associated with the turn-    firm owner? Not surprisingly, older,
down rate in either survey.                   larger firms are less likely to be turned
   Changes in loan officer turnover            down in their most recent loan request,
appear to have a meaningful economic          as are those firms located in nonmetro-
effect as well, but care must be taken in     politan areas. So how do younger firms
quantifying this effect, especially if the    offset this inherent disadvantage due to
coefficient on turnover is biased because      their information opacity or location—
turnover is not independent of credit         assuming that they cannot change these
availability. An increase in loan officer      factors? First, a longer time at their
turnover from one (no change) to five or       primary bank allows the accumulation
more (a four-category change) increases       of hard information that decreases the
the (log) odds of turndown by 0.21 and        chance of being turned down. Second,
0.13 in 1995 and 2001, respectively.          owners at banks with low account
These marginal changes are approxima-         manager turnover are less likely to be
tions estimated by multiplying the            turned down on their most recent loan

                                                                    Table 2
                                    Logistic Regression Results for Loan Application Turndown Rate as the
                                                             Dependent Variablea
        Explanatory                                                               1995 Survey                                                                                               2001 Survey

                                           Baseline                Loan Officer                  Sample                   Instrumental                 Baseline                 Loan Officer                  Sample                   Instrumental
                                           Results                   Turnover                 Limited to                   Variables                  Results                    Turnover                 Limited to                 Variables for
                                                                    Coefficient                Those That                   for Loan                                             Coefficient                Those That                 Loan Officer
                                                                    Varies with                 Did Not                     Officer                                              Varies with                 Did Not                    Turnover
                                                                   Time at Bank                 Change                     Turnover                                            Time at Bank                 Change
                                                                                                 Banks                                                                                                       Banks
                                             (1)                       (2)                        (3)                        (4)                        (5)                        (6)                        (7)                        (8)
                                    Coefficient   Standard     Coefficient   Standard      Coefficient   Standard      Coefficient   Standard      Coefficient   Standard      Coefficient   Standard      Coefficient   Standard      Coefficient   Standard
                                                   Error                     Error                      Error                      Error                      Error                      Error                      Error                      Error

        Loan Officer Turnoverb            0.382     0.035***                                     0.408   0.089***          2.176    0.704***         0.317     0.110***                                  0.339       0.125***       1.723       1.195
          Turnover × Long Time                                      0.320    0.068***                                                                                          0.621     0.342*
             at Bankc
          Turnover × Short Time                                     0.454    0.070***                                                                                          1.248     0.375***
             at Bank
        Time at Bank                    −0.123     0.063***                                                              −0.144    0.035***        −0.215     0.064***                                                            −0.207       0.065***
        Number of Financial              0.059     0.060            0.072     0.060             0.070   0.091             0.061    0.059           −0.083     0.129           −0.070     0.128         −0.067       0.161         −0.060       0.128

          Institutions (FIs) Used
        Bank Recently Merged             0.067     0.134            0.098    0.134             0.015    0.198            −0.484    0.249*          −0.029     0.208            0.005     0.206         −0.126       0.263         −0.314       0.353
        Herfindahl Index                  0.172     0.658            0.170    0.656             0.514    0.846             0.722    0.664            1.483     0.669**          1.455     0.666**        1.303       0.868          1.713       0.673**
        Rural Location                  −0.702     0.157***        −0.706    0.157***         −0.770    0.217***         −0.644    0.159***        −0.529     0.228**         −0.553     0.227**       −1.024       0.302***       0.501       0.241**
        Years in Business               −0.596     0.089***        −0.607    0.089***         −0.604    0.122***         −0.551    0.088***        −0.301     0.209           −0.301     0.210         −0.539       0.234**       −0.367       0.211*
        Full-Time Equivalent            −0.185     0.070***        −0.184    0.070***         −0.192    0.099*           −0.182    0.070***        −0.296     0.105***        −0.312     0.106***      −0.349       0.130***      −0.353       0.107***
          (FTE) Employees
        Sales Growth                    −0.249     0.052***        −0.247    0.052***         −0.316    0.077***         −0.233    0.051***        −0.007     0.078            0.005     0.077         −0.080       0.099         −0.020       0.078
        Corporation                     −0.148     0.128           −0.151    0.127             0.019    0.175            −0.135    0.127           −0.131     0.211           −0.109     0.210         −0.114       0.257         −0.048       0.212
        Agriculture                     −0.337     0.264           −0.339    0.263            −0.246    0.381            −0.301    0.262           −0.910     0.481*          −0.911     0.479*        −0.945       0.550*        −0.875       0.480*
        Construction                     0.050     0.215            0.033    0.214            −0.107    0.307             0.021    0.213           −0.204     0.312           −0.237     0.311         −0.708       0.436         −0.273       0.316
        Finance                         −0.096     0.288           −0.128    0.287            −0.240    0.432            −0.119    0.284           −1.150     0.563**         −1.132     0.559*        −0.800       0.594         −1.170       0.561**
        Manufacturing                    0.015     0.221            0.020    0.220             0.284    0.297            −0.039    0.219           −0.055     0.324           −0.069     0.322          0.063       0.380         −0.049       0.325
        Professional Services           −0.063     0.291           −0.087    0.291             0.228    0.402            −0.066    0.290           −0.283     0.437           −0.286     0.433         −0.054       0.485         −0.351       0.437
        Nonprofessional Services         0.513     0.178***         0.497    0.177***          0.566    0.245**           0.482    0.177***        −0.460     0.306           −0.466     0.304         −0.577       0.391         −0.515       0.311*
        Transportation                   0.572     0.331*           0.521    0.330             0.846    0.420**           0.429    0.329            0.232     0.499            0.179     0.495          0.159       0.602          0.168       0.499
        Wholesale                        0.009     0.281           −0.005    0.280             0.167    0.372            −0.070    0.279           −1.325     0.472***        −1.350     0.473***      −1.053       0.531**       −1.401       0.473***
        Number of Observations       2,330                      2,330                      1,425                      2,330                     1,126                      1,126                      849                      1,126
        −2 Log Likelihood             1889.5                     1908.0                     1037.9                     1922.8                     744.6                      749.8                    508.4                      734.7
        Pseudo R-squared                 0.096                      0.089                      0.082                      0.083                     0.066                      0.062                    0.065                      0.066

          No answer categories for the independent variables were included in the estimation but are not reported. Loan officer turnover, time at bank, number of FIs used, years in business, and FTE employees are entered in log form. The significance
        levels at the 1, 5, and 10 percent levels are denoted by ***, **, and *.
          Long time at bank = 1 time at bank is 5 or more years; short time at bank = 1 if time at bank is less than 5 years.
          The chi-square statistic for the null hypothesis of equal coefficients for the 1995 survey is 4.98 with p = .03 while for the 2001 survey it is 8.39 with p = .00.

request, although the magnitude of this       present, has a minor effect on the basic
benefit decreases with the length of time      conclusion of the analysis—credit avail-
the owner has spent with their current        ability increases with less loan officer
bank.                                         turnover.
                                                  The third test addresses the issue of
Sensitivity Tests of Baseline Results         whether or not turnover is a function of
   The first test is shown in column (3)       firm risk. This phenomenon would occur
and column (7) in Table 2 using the           if higher-risk firms are transferred to a
sample that is limited to those firms that     loan workout group as the result of an
did not change banks. This restriction        internal rating change. Although the
limits the loan officer turnover to those      survey does not include any data on bank
cases where the last loan attempt was         internal ratings, a set of interactive vari-
made at the current principal FI. With        ables with loan officer turnover is created
this restriction, instances where the         based on years in business and size of the
owner changes banks and gets a new            firm. These interactive variables are: loan
loan officer are eliminated, thus provid-      officer turnover × young firms, loan
ing a purer measure of turnover. This test    officer turnover × mid-age firms, and
reduces the sample size to 1,425 in the       loan officer turnover × old firms for years
1995 survey and 849 in the 2001 survey.       in business, and loan officer turnover ×
The coefficient on loan officer turn-           small firms, loan officer turnover × mid-
over remains positive, significant, and of     size firms, and loan officer turnover ×
approximately the same magnitude com-         large firms based on FTE employment.
pared to the baseline coefficient in both      The years in business and size categories
surveys. If turnover was largely driven       are defined in Table 3 for 1995 and 2001,
by owners switching banks who were            and each takes a value of 1 if the firm
turned down, the coefficient on the            falls within the range and 0 if otherwise.
turnover variable should be insignifi-         Younger, smaller firms are more likely to
cant or should at least have a lower          encounter financial distress and thus be
magnitude.                                    more likely candidates to be shifted to
   For the second test, a set of instru-      workout groups. The expectation is that
ments is used to create predicted values      the coefficients on the interactive terms
for loan officer turnover. These instru-       for younger, smaller firms will be posi-
ments include the occurrence of a             tive, whereas the coefficients on the inter-
merger, region, market size, MSA em-          active term for older, larger firms will be
ployment, and deposit concentration.          zero or, if positive, less than the coeffi-
The instrumental variable (IV) estimates      cient for the younger firms. The estimates
are presented in column (4) and column        of the interactive terms with years in busi-
(8) of Table 2. The coefficient on loan        ness are shown in column (1) and column
officer turnover is positive for both          (4) of Table 3, whereas the estimates of
surveys, but it is only significant in         the interactive terms with employment
the 1995 survey. The mixed IV results         are shown in column (2) and column (5).
may be partially attributable to the low      A Wald linear restriction test is used to
variability of loan officer turnover instru-   test the null hypothesis that all of the
ments. For example, the standard devia-       interactive coefficients are equal and is
tion of loan officer turnover in the 1995      presented at the bottom of the tables.
survey is 0.38, but the standard deviation        In 1995, younger, smaller firms have a
of the IV predicted loan officer turnover      significantly higher chance of reporting
is 0.08. A similar pattern is observed for    increased turnover compared to old
the 2001 survey. Still, both sensitivity      firms. However, older, larger firms still
tests strongly suggest that endogeneity, if   have a significantly worse experience

                                                  Table 3
           Logistic Regression for Loan Application Turndown Rate: Sensitivity of Loan Officer
          Turnover to Years in Business, Full-Time Equivalent (FTE) Employment, and Number of
                                     Financial Institutions (FIs) Used
        Independent Variables                                    1995 Survey                                                          2001 Survey

                                           Years in                  FTE                 Number of              Years in                  FTE                 Number of
                                         Business Test           Employment             Banks Used            Business Test           Employment             Banks Used
                                                                     Test                   Test                                          Test                   Test
                                               (1)                    (2)                    (3)                    (4)                    (5)                    (6)
                                      Coefficient Standard    Coefficient Standard    Coefficient Standard    Coefficient Standard    Coefficient Standard    Coefficient Standard
                                                   Error                  Error                  Error                  Error                  Error                  Error

        Loan Officer Turnover ×          0.667     0.083***                                                   0.588     0.173***
          Young Firm (<5 Years)b
        Loan Officer Turnover ×          0.347     0.080***                                                   0.265     0.128*
          Mid-Age (5–15 Years) Firm

        Loan Officer Turnover × Old      0.241     0.069***                                                   0.216     0.133
          (>15 Years) Firm
        Loan Officer Turnover ×                                 0.507     0.079***                                                   0.425     0.149***
          Small (<3 FTE) Firmc
        Loan Officer Turnover ×                                 0.369     0.071***                                                   0.219     0.129*
          Mid-Size (3–6 FTE) Firm
        Loan Officer Turnover ×                                 0.306     0.076***                                                   0.162     0.117
          Large (>6 FTE) Firm
        Loan Officer Turnover ×                                                        0.321     0.076***                                                   0.325     0.118***
          Use 1 FId
        Loan Officer Turnover ×                                                        0.362     0.066***                                                   0.312     0.121**
          Use More Than 1 FI
        Time at Bank                   −0.130     0.035***    −0.122     0.035***    −0.127     0.035***    −0.207     0.065***    −0.215     0.064***    −0.213     0.065***
        No. of FIs Used                 0.053     0.060        0.055     0.060                              −0.081     0.131       −0.108     0.126
        Bank Recently Merged            0.056     0.134        0.080     0.134        0.081     0.134        0.037     0.207        0.009     0.206       −0.035     0.208
        Herfindahl Index                 0.217     0.654        0.169     0.660        0.144     0.656        1.448     0.678**      1.389     0.665**      1.458     0.668**
        Rural Location                 −0.731     0.157***    −0.700     0.157***    −0.699     0.157***     0.534     0.229**      0.449     0.226**      0.525     0.228**
        Years in Business                                     −0.607     0.089***    −0.590     0.089***                           −0.469     0.205**     −0.297     0.210
        FTE Employees                  −0.221     0.070***                           −0.179     0.070***    −0.360     0.108***                           −0.302     0.105***
        Sales Growth                   −0.230     0.052***    −0.252     0.051***    −0.247     0.051***    −0.019     0.078       −0.005     0.078       −0.010     0.078

                                                                                                                            Table 3
                                       Independent Variables                                           1995 Survey                                                                    2001 Survey

                                                                             Years in                    FTE                     Number of                   Years in                    FTE                     Number of
                                                                           Business Test             Employment                 Banks Used                 Business Test             Employment                 Banks Used
                                                                                                         Test                       Test                                                 Test                       Test
                                                                                (1)                       (2)                        (3)                       (4)                        (5)                        (6)
                                                                       Coefficient Standard       Coefficient Standard        Coefficient Standard       Coefficient Standard        Coefficient Standard        Coefficient Standard
                                                                                    Error                     Error                      Error                     Error                      Error                      Error

                                       Corporation                        −0.166      0.128          −0.165     0.127           −0.139     0.127          −0.089      0.211          −0.155     0.210          −0.133      0.211
                                       Agriculture                        −0.408      0.263          −0.344     0.265           −0.331     0.264          −0.794      0.476          −0.816     0.476          −0.920      0.480
                                       Construction                        0.019      0.214           0.036     0.215            0.055     0.214          −0.173      0.313          −0.222     0.311          −0.208      0.312
                                       Finance                            −0.097      0.287          −0.084     0.289           −0.088     0.287          −1.120      0.562**        −1.060     0.561*         −1.179      0.562**
                                       Manufacturing                      −0.015      0.221          −0.016     0.220            0.010     0.220           0.012      0.325          −0.183     0.323          −0.046      0.324
                                       Professional Services              −0.038      0.293          −0.045     0.292           −0.058     0.291          −0.215      0.438          −0.258     0.434          −0.298      0.437
                                       Nonprofessional Services            0.529      0.178***        0.522     0.178***         0.509     0.177***       −0.454      0.309          −0.378     0.303          −0.466      0.305
                                       Transportation                      0.546      0.328*          0.580     0.329*           0.568     0.330*          0.258      0.500           0.165     0.497           0.214      0.499
                                       Wholesale                          −0.038      0.279          −0.002     0.280           −0.002     0.280          −1.263      0.470***       −1.354     0.468***       −1.346      0.473***
                                       Number of Observations          2,330                      2,330                      2,330                     1,126                      1,126                     1,126
                                       −2 Log Likelihood                1907.8                     1895.2                     1903.0                     744.0                      745.6                     745.0
                                       Pseudo R-squared                    0.089                      0.094                      0.091                     0.067                      0.065                     0.066
                                       Linear Restriction Test for        38.30      (0.00)           7.05      (0.03)           0.40     (0.53)           1.95      (0.38)           3.06     (0.22)           0.02      (0.89)
                                         Equality of Loan Officer
                                         Turnover Interactive

                                         No answer categories for the independent variables were included in the estimation but are not reported. Loan officer turnover, time at bank, number of FIs used, years in business,
                                       and FTE employees are entered in log form. The significance levels at the 1 percent, 5 percent, and 10 percent levels are denoted by ***, **, and *.
                                         Each interactive variable is constructed by multiplying the value of loan officer turnover by 1 if the years in business falls in that range (young, mid-age, or old) or 0 if otherwise.
                                         Each interactive variable is constructed by multiplying the value of loan officer turnover by 1 if the years in business falls in that range (small, medium, or large) or 0 if otherwise.
                                         Each interactive variable is constructed by multiplying the value of loan officer turnover by 1 if the years in business falls in that range (use 1 FI or use more than 1 FI) or 0 if other-
                                         The chi-square statistic for the null hypothesis of equal coefficients is presented first, followed by the significance level in parentheses.
accessing credit if loan officer turnover      processes such as credit scoring. Loan
is high in 1995. The fact that old and        officers add value to soft information
large firms still have a significant loan       because it cannot easily be transmitted
officer turnover coefficient is consistent      within a banking organization. With hard
with the strength of soft information pro-    information, loan officers add less value
duction on credit availability—its effect     because the interpretation can be made
is just less than it is on smaller, younger   independent of the person who may
firms. In 2001, no significant difference       gather it.
in turnover is found by years in business        Soft information can be a critical input
or firm size. Taken together, this alterna-    to the credit decision for many small,
tive specification does not provide clear      information opaque firms, especially
evidence that loan officer turnover            those without a sufficiently long operat-
results from the shifting of risky small      ing history or those with substantial
firms to loan workout groups under the         intangible assets. For these firms, loan
supervision of a different loan officer.       approvals based only on hard informa-
    The fourth and final test examines         tion could limit their ability to obtain
whether or not multiple banking rela-         credit. Soft information obtained by the
tionships may complicate the interpreta-      loan officer is likely to increase credit
tion of loan officer turnover on the           availability, but there are instances where
turndown rate as noted earlier. To            it may not (for example, the officer learns
address this issue, another set of inter-     about internal control problems at the
active variables are created for loan         firm) and result in decreased credit avail-
officer turnover: loan officer turnover ×       ability. Thus, the benefit of soft infor-
uses 1 FI and loan officer turnover × uses     mation to small firms is an empirical
more than 1 FI, where uses 1 FI takes a       issue.
value of 1 if the owner uses only one FI         A survey variable is used from a
and 0 if otherwise, and uses more than        sample of U.S. small businesses in 1995
one FI takes a value of 1 if the owner        and 2001 that addresses the loan officer’s
uses more than one FI. These estimates,       role in producing soft information. This
presented in column (3) and column (6)        proxy, the number of loan officers the
of Table 3, show that the number of FIs       small firm had in the past three years, is
used has no significant effect on loan         related to the outcome of the most recent
officer turnover.                              loan application. Reduced loan officer
                                              turnover is associated with lower turn-
                                              down rates and has a time dimension,
Summary and                                   with the effect of higher loan officer
Conclusions                                   turnover weakening, with the length of
   This paper investigates how credit         time that the owner remains at their
availability is affected by loan officer       primary bank. This weakening turnover
turnover, a proxy for the production of       effect could reflect the migration of soft
soft information by loan officers. Soft        information to hard information in the
information refers to nonquantitative         lending decisions. Although the results
information such as assessments of the        may be biased because loan officer
owner’s character, reputation, or man-        turnover is not completely independent
agement ability that can be used as input     of the credit decision, estimates from a
to the credit approval process. In con-       subsample of firms that have not
trast, many credit decisions rely on hard     changed their bank, IV estimates of loan
information such as tax return data or        officer turnover, and interactive coeffi-
other financial ratios that are amenable       cients that test for differential turnover
to statistically driven loan approval         effects by years in business and size of

                                        SCOTT                                        559
the firm all continue to document a per-           The Importance of Bank Organisa-
sistent loan officer turnover effect.              tional Structure,” Economic Journal
    Overall, the results are consistent with      112, F32–F53.
the idea that the value of relationship        Berlin, M., and L. Mester (1999). “Deposits
banking resides as much with the loan             and Relationship Lending,” Review of
officer’s production of soft information           Financial Studies 12, 579–608.
as the bank’s accumulation of hard infor-      Bhattacharya, S., and G. Chiesa (1995).
mation. What do these results mean for            “Proprietary Information, Financial
owners of small firms? First, they should          Intermediation, and Research Incen-
develop an understanding of the costs             tives,” Journal of Financial Interme-
and benefits of relationship banking.              diation 4, 328–357.
Credit-scoring technologies may give           Boot, A. (2000). “Relationship Banking:
quick answers to loan requests at a new           What Do We Know?” Journal of
bank, but the outcome could change in             Financial Intermediation 9, 7–25.
a future application if key financial ratios    Cavalluzzo, K., L. Cavalluzzo, and J.
change. Second, soft information matters          Wolken (2002). “Competition, Small
for small firms, especially those with a           Business Financing, and Discrimina-
short operating history or with few tan-          tion: Evidence from a New Survey,”
gible assets, whereas older, larger firms          Journal of Business 75, 641–679.
may benefit more from a credit-scoring          Cole, R. A. (1998). “The Importance of
technology based on traditional financial          Relationships to the Availability of
ratios. Third, not all banks will choose to       Credit,” Journal of Banking and
be relationship lenders. Owners of small,         Finance 22, 959–977.
information opaque businesses should           Cole, R., L. Goldberg, and L. White
seek banks where there is a history of            (2004). “Cookie-Cutter versus Charac-
stability in loan officers and local deci-         ter: The Micro Structure of Small Busi-
sion-making. For many firms, this rec-             ness Lending by Large and Small
ommendation may mean choosing a                   Banks,” Journal of Business 39,
locally owned, community bank that                227–252.
focuses on relationship lending. Finally,      Degryse, H., and P. Van Cayseele (2000).
if the characteristics of the business are        “Relationship Lending within a Bank-
such that soft information is important,          Based System: Evidence from Euro-
the earlier the choice of a relationship          pean Small Business Data,” Journal
bank is made, the better because the              of Financial Intermediation 9, 90–109.
length of time at a bank is also related       Dunkelberg, W. (1998). “Credit, Banks
to credit decisions.                              and Small Business in America,”
                                                  Journal of Banking and Finance 22,
References                                        1085–1088.
Berger, A., N. Miller, M. Petersen, R.         Greene, W. (2002). Econometric Analysis,
  Rajan, and J. Stein (2005). “Does Func-         5th ed. New York: Prentice-Hall.
  tion Follow Organizational Form? Evi-        Harhoff, D., and T. Korting (1998).
  dence From the Lending Practices of             “Lending Relationships in Germany—
  Large and Small Banks,” Journal of              Empirical Evidence from Survey
  Financial Economics 76, 237–269.                Data,” Journal of Banking and
Berger, A., and G. Udell (1995). “Rela-           Finance 22, 1317–1353.
  tionship Lending and Lines of Credit         Petersen, M. A., and R. G. Rajan (1995).
  in Small Firm Finance,” Journal of              “The Effect of Credit Market Com-
  Business 68, 351–382.                           petition on Lending Relationships,”
——— (2002). “Small Business Credit                Quarterly Journal of Economics 110,
  Availability and Relationship Lending:          405–443.

Petersen, M., and R. Rajan (1994). “The        Differences between Female- and
   Benefits of Firm-Creditor Relation-          Male-Owned Small Business,” mimeo.
   ships: Evidence from Small Business      Scott, J., W. Dunkelberg, and W. Dennis,
   Data,” Journal of Finance 49, 3–37.         Jr. (2003). Credit, Banks, and Small
Rajan, R., and A. Winton (1995).               Business: The New Century. Washing-
   “Covenants and Collateral as an Incen-      ton, DC: NFIB Research Foundation.
   tive to Monitor,” Journal of Finance     Stein, J. (2003). “Information Production
   50, 1113–1146.                              and Capital Allocation: Decentralized
Robb, A., and J. Wolken (2002). “Firm,         versus Hierarchical Firms,” Journal of
   Owner, and Financing Characteristics:       Finance 57, 1891–1922.

                                      SCOTT                                      561
        Selected Demographic Characteristics of National
    Federation of Independent Business (NFIB) Credit Banks
      and Small Business Survey versus National Survey of
         Small Business Finance (NSSBF) Respondentsa
                    NFIBb           NSSBFc                              NFIB          NSSBF

                 1995    2001    1993    1998                     1995     2001     1993   1998

Form of Business                                 Years in Business
Proprietorship   31       26      44      43     0–4               15          11    15     15
Partnership       6        8       8       8     5–9               21          14    27     28
Corporation      42       40      28      20     10–14             19          16    19     19
S-Corporation    21       24      20      28     15–19             14          14    14     14
No Answer         1        3                     20–24             11          13     9      9
                                                 25 or more        19          28    15     15
Full Time Equivalent Employees                   No Answer          1           5
One                7      7       39      18
2–4               30     29       29      39     Industry
5–9               27     25       15      23     Construction/     13          15    14     14
10–19             17     17        8      11       Mining
20–49             11     12        5       6     Manufacturing     12          11     8      8
50–99              3      4        1       2     Transportation     3           4     3      3
100–499            2      2        1       1     Wholesale          6          10     8      8
500 or more        *      *                      Retail            21          20    22     22
No Answer          3      5                      FIREd              7           6     7      7
                                                 Business          17          17    22     21
Gross Sales ($000)                                 Services
Under 25            2      3      13      11     Professional       8          8     16     17
25–49               2      1       8       9       Services
50–99               6      4      12      13     Agriculture        8          5
100–249            18     12      24      25     No Answer          4          4
250–499            18     14      15      16
500–999            16     15      11      12     Region
1,000–2,499        15     14       9       8     Northeast         15          15    23     22
2,500–4,999         7      7       4       4     Midwest           34          39    24     24
5,000–9,999         5      4       2       2     South             27          24    29     29
10,000 or more      4      6       2       2     West              26          22    24     24
No Answer           8     21                     No Answer          0           *

                                                 Urban Location
                                                 Yes               65          60    79     79
                                                 No                35          40    21     21
                                                 No Answer          *           *

  The NFIB data are unweighted, while the NSSBF data are weighted to reflect the population pro-
  Percent of total responses reported.
  No answer responses for empirically estimated in the NSSBF data but not for the NFIB data.
  FIRE, finance, insurance, and real estate.
*Less than 0.5 percent.


Shared By: