The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Federal Reserve Bank of Chicago
Supplier Relationships and Small
Business Use of Trade Credit
By: Daniel Aaronson , Raphael Bostic,
Paul Huck and Robert Townsend
WP 2000-28
Supplier Relationships and Small Business Use of Trade Credit
December 2000
Daniel Aaronson
Federal Reserve Bank of Chicago
Raphael Bostic
Board of Governors, Federal Reserve System
Paul Huck
Federal Reserve Bank of Chicago
Robert Townsend
University of Chicago and
Federal Reserve Bank of Chicago
----------------------------------------------------------------------------------------------------------------------The authors thank Ing-Haw Cheng for research assistance. The views expressed are the authors’
and do not necessarily reflect the views of the Federal Reserve Bank of Chicago, the Board of
Governors, or the Federal Reserve System.
Abstract
This paper sheds some light on the empirical importance of supplier relationships, including
ethnic ties, for the use of trade credit by minority-owned small businesses. Results based on the
1993 National Survey of Small Business Finance (NSSBF) indicate that ethnic differences in the
use of trade credit are present after conditioning on an extensive list of control variables. This holds
especially for Black-owned businesses, and we find that they use less trade credit, are less likely to
take advantage of discounts for early payment, and are more likely to have payments past due.
We use neighborhood survey data to explore the importance of supplier relationships for the
use of trade credit by Black- and Hispanic-owned businesses. Although Black and Hispanic owners
are equally likely to be offered trade credit, the relationship effects vary by ethnicity. Closer
relationships with suppliers as measured by ethnic ties and geographical proximity are associated
with more trade credit for Hispanic-owned businesses. In contrast, this result does not hold for
Black-owned firms.
The neighborhood survey results suggest the idea of looking for ethnic differences in the
effects of relationships at the national level as well. Although good supplier-level measures of
relationships are not available in the NSSBF, we use census data to construct MSA-level measures
of the prevalence of minority-owned businesses. We then explore how location in an MSA with a
higher proportion of businesses of the same ethnicity is associated with the use of trade credit by
minority owners relative to White-owned firms. We find that a higher MSA share for Hispanicowned businesses is generally associated with a reduction in differences in the use of trade credit by
Hispanic owners relative to White owners. No clear association is apparent between the MSA share
for Black-owned businesses and their use of trade credit.
Thus, the ethnic differences in the effects of relationships evident in the neighborhood
surveys seem to be consistent with the results from the national survey.
Introduction
Understanding access to capital and credit for small businesses is not a simple matter
because capital, credit, and insurance markets are not complete and frictionless. Thus, a great
variety of contractual arrangements, explicit and implicit, formal and informal, are observed as
economic agents and organizations devise ways of dealing with the frictions that hinder economic
exchange. For example, a lender may monitor the borrower’s operation, take an equity position, or
require collateral. These varied ways of doing business may include a role for ongoing
relationships between economic agents or perhaps networks of agents. Alongside this thicket of
financial arrangements, there is a sense that race or ethnicity may play a role in how an owner
finances the business. For example, self-employment rates vary across ethnic groups in ways that
are not fully understood, ethnic networks may be important in some communities, and some ethnic
groups may face discrimination. It is important that we know more about small business finance
because of the importance of these businesses for the economy as a whole and because many policy
efforts are aimed at promoting their access to capital and credit in one form or another.
However, we have much to learn, both empirically and theoretically, about the wide variety
of ways entrepreneurs finance their businesses and the role played by relationships and networks,
including ethnic connections. In order to further our understanding of these issues, researchers from
the University of Chicago and the Federal Reserve Bank of Chicago have cooperated in surveying
businesses and households in two neighborhoods in Chicago, one predominantly Hispanic and the
other predominantly Black. A primary goal of these neighborhood studies is to measure financial
arrangements and the relationships, ethnic and otherwise, among agents in order to better inform
policy discussions and theoretical work in this field. This paper focuses on one source of credit for
businesses, trade credit, as part of a larger research agenda based on the neighborhood studies.1
Trade credit is an important part of the balance sheets of many small businesses.2 Results
of a national survey show that it accounted for 31.3 percent of the total debt for small businesses in
1993, and 60.8 percent of the firms had outstanding credit from suppliers.3 Trade credit is also
interesting because it is a good place to look for the effects of relationships and networks. For
example, in a world of imperfect information, a supplier may learn about a firm’s creditworthiness
and future prospects in the course of their ongoing business relationship. Thus, the strength of the
ties between a business and its suppliers may play a role in the terms upon which trade credit is
offered or whether it is offered at all. Supplier relationships in the developing world have recently
received some attention, as Fafchamps and Minten (1999) and McMillan and Woodruff (1999) find
that relationships play an important role in access to trade credit in Madagascar and Vietnam,
respectively. This finding is analogous to the result that relationship measures are related to the
availability and terms of credit from U.S. financial institutions.4
The purpose of this paper is to report some empirical regularities in the use of trade credit
and in the effects of some measures of supplier relationships on trade credit, paying particular
attention to ethnic differences. We report results based on two sources of data. First, we establish
an empirical picture of the use of trade credit using the 1993 National Survey of Small Business
Finance (NSSBF), a nationally representative survey. For the first time, the NSSBF includes an
oversample of minority businesses, so we are able to measure ethnic differences in the use of trade
credit. Second, we use the neighborhood surveys to explore the empirical association between
several measures of supplier relationships and the offer of trade credit. The national survey and the
1
See Huck, et al (1999) for a general overview and previous results of the neighborhood surveys.
See Mian and Smith (1992), Petersen and Rajan (1997), and Ng et al (1999) for more discussion of the theory and
practice of trade credit.
3
The figures come from the 1993 National Survey of Small Business Finance, which defines small businesses as
businesses with fewer than 500 employees. See Cole and Wolken [1995, Table A.2] and Berger and Udell [1998, Table
1] for the cited figures on the use of trade credit.
2
neighborhood surveys complement each other in that the NSSBF is designed to be representative of
small businesses in the entire nation. On the other hand, the local focus of the neighborhood
surveys provides more information about the relationships between business owners and their
suppliers than is available in the national survey.
Briefly, the empirical results of the national survey section establish the fact that ethnic
differences are present in the use of trade credit, even after conditioning on an extensive list of
control variables. Although we find some differences for other minority groups, this finding
especially holds for Black-owned businesses.5 Some of the more striking results are that if we look
at businesses that make at least some purchases on account, Black-owned businesses use less trade
credit, are less likely to take advantage of discounts for early payment, and are more likely to have
payments past due.6
We use the neighborhood survey data to explore the correlation between supplier
relationships and the offer of trade credit for minority-owned small businesses. Although Black and
Hispanic owners are equally likely to be offered credit, both with and without conditioning on
control variables, the relationship effects vary by ethnicity. We find that working with a Hispanic
supplier and working with a supplier relatively close to home are associated with more credit for
Hispanic-owned businesses. In contrast, this result does not hold for Black-owned businesses.
4
Petersen and Rajan (1994), Berger and Udell (1995), and Uzzi (1999).
There is also recent evidence that Black owners are more likely to be denied bank credit relative to White owners with
comparable observable characteristics. See Cavalluzzo and Cavalluzzo (1998), Bostic and Lampani (1999),
Cavalluzzo, Cavalluzzo, and Wolken (1999), and Blanchflower, Levine, and Zimmerman (1998).
6
Businesses that do not take advantage of a discount for early payment pay a substantial implicit interest cost. Typical
trade credit terms, such as the 2/10 net 30 contract, implies an implicit annual interest rate of 44 percent. Delaying
payment after the due date may also entail a penalty, including perhaps a reputational cost, although the penalty for late
payment varies by supplier and may not be substantial. It can be argued that firms that use such a high cost source of
funds are constrained in their access to cheaper sources of funding, such as bank credit. Thus, not taking advantage of
discounts for early payment is a good indicator that a business faces credit constraints (Peterson and Rajan 1994).
Under this interpretation, our findings for early payment discounts indicate that Black (and Asian) owners face credit
constraints for lower cost funding sources. However, we do not necessarily accept this interpretation because it depends
on how one models the use of trade credit. It is possible to imagine models for which a supplier relationship involves
some benefits that compensate for paying a high interest cost for trade credit.
5
The neighborhood survey results suggest looking for ethnic differences in the effects of
relationships at the national level as well. Although good supplier-level measures of relationships
are not available in the NSSBF, we use census data to construct MSA-level measures of the
prevalence of minority-owned businesses. We then explore how location in an MSA with a higher
proportion of businesses of the same ethnicity is associated with the use of trade credit by minority
owners relative to White-owned firms. We find that a higher MSA share for Hispanic-owned
businesses is generally associated with a reduction in differences in the use of trade credit by
Hispanic owners relative to White owners. No clear association is apparent between the MSA share
for Black-owned businesses and their use of trade credit. Thus, the ethnic differences in the effects
of relationships evident in the neighborhood surveys seem to be consistent with the results from the
national survey. Furthermore, differing prospects for developing financial ties help to explain part
of the observed ethnic differences in trade credit usage.
Why do relationships arise?
Recent work on the theory of collective organizations suggests new ways to think about
some questions related to the use of trade credit. Why might some businesses choose to operate
without trade credit, whereas others form close-knit relationships with suppliers, including the
extension of credit? If there is trade credit, should we expect to see homogeneity or heterogeneity
in the characteristics of suppliers and creditors? One class of models, as exposited in the work of
Prescott and Townsend (2000), builds on an earlier mechanism design literature and can help us to
think through the many complicated forces that make all these forms of organization endogenous.
Suppose that a household can go into business and operate a technology producing output as
a stochastic function of labor and capital -- either with the owner’s own wealth or borrowed funds
from a competing set of financial institutions. The firm can also purchase insurance to cover some
of the fluctuations in its output or sales. Within this basic set-up, we can imagine various
impediments to production and exchange in financial, credit, and insurance markets. First, the labor
input may be unobserved by outsiders in the market. This creates the usual moral hazard problem.
Moral hazard would hinder full insurance of fluctuations in sales, for otherwise the owner has no
incentive to be diligent. Moral hazard would also limit the amount of credit; an owner who has
financed his operation with costly capital may need to use much of his revenue to repay, causing a
decline in labor effort and an implicit increase in the interest rate. This would limit the scale of
operations and conceivably preclude the operation of the potential business in the first place. A
second impediment would be the possibility of default. If an owner with borrowed capital can
default, that is, take off with revenues or direct too much compensation to the owner, this too would
limit the firm’s financing or again preclude production entirely.
Within this basic set-up, we can imagine alternative forms of organization. For example,
another household can form a close-knit relationship with a proprietor, possibly monitoring the
diligence of the proprietor at the cost of some labor effort. In the limit, suppliers may almost appear
to be partners, fully engaged in input decisions, the financing of those decisions, and the sharing of
output fluctuations. This third aspect has an interesting interpretation -- the supplier absorbs the
“internal” default of the proprietor, lessening the likelihood of external default to the market. More
generally, the advantage of network relationships is that they can mitigate impediments to
exchange. That is, the supplier, partner, or network member not only has better information on the
diligence of the proprietor, but also can in one way or another supply that information to the larger
credit and insurance market. Similarly, the supplier, partner, or network member can make default
on the part of the proprietor more difficult, or can make better use of the proprietor’s capital given
that the latter does default. One caveat, however, is that trade credit and other close-knit
relationships do not allow full recovery of the usual neoclassical efficiency properties. Indeed,
internal relationships may appear constraining relative to those neoclassical norms. Access to
outside credit and insurance on the part of the proprietor may appear to be overpriced or otherwise
constrained. For example, the proprietor might need to pay the supplier more, depending on the
circumstances of the latter.
Modest variations in the underlying characteristics of households or business owners can
produce large variations in organizational outcomes. We provide an example using the ideas from
Prescott and Townsend (2000) where wealth varies within an otherwise homogenous population.
Single proprietorships engaged in the larger credit and insurance market but without close-knit
suppliers are more likely to emerge for relatively wealthy entrepreneurs. These firms can finance
much of their own operation and hence for them moral hazard is less severe. They reap most of the
benefits of their own high efforts. However, a relatively wealthy firm may take the benefits of their
wealth by reducing work effort. This lowers the moral hazard problem, making partnerships less
fruitful but increases the probability of default. More generally, however, high economy-wide
wealth makes labor the limiting factor, but this also makes single proprietorships more likely, as
less labor is expended in supervisory or joint-production relationships.
In economies where capital is scarce, it makes more sense to use labor in monitoring.
Hence, collective network forms are more likely to emerge there. However, holding economy-wide
wealth constant, the distribution of wealth can be associated with the existence and nature of these
networks. Results here are sensitive to specific assumptions. We can show that higher inequality in
the wealth distribution can be associated with either an increased or a decreased likelihood of
network organization, depending on whether networks or collective organizations are defined by
collusion, coordination, risk-sharing, or by the joint operation of technologies. Increased inequality
in the wealth distribution can lead to homogenous matching in multi-agent networks, of poor to
poor for example, where one agent will be the proprietor and the other supplier/creditor. Decreased
wealth inequality can lead to heterogeneous matching, with the relatively wealthy taking their utility
benefit in the form of less onerous supervision and increased consumption compensation.
Other than wealth, we might imagine that households and (potential) firms vary in talent or
(potential) productivity, either in production directly or in the efficiency of supervision. Similarly,
households and (potential) firms may vary in preferences for the disutility of work effort or in
aversion to risk. Indeed, households or (potential) firms may vary in their aversion to being paired
with others, according to the characteristics of others, as in the literature on clubs. In this regard,
space and ethnicity may enter the picture. Proximity in space may facilitate the mitigation of
impediments to trade. Information and the ability to inflect penalties on default may be better
locally, in which case we might imagine that networks would be more likely within rather than
across neighborhoods. Finally, ethnicity may also be correlated contemporaneously with some of
the above-mentioned attributes: risk aversion, work aversion, or affinity aversion. The point is that
relatively simple considerations can lead to a great variety of endogenous outcomes. That is,
models may produce a variety of correlations between space, ethnicity, and other observables on the
one hand, and network relationships or more autarkic arrangements on the other.
National Use of Trade Credit Among Ethnic Groups
The relevance of ethnic relationships takes on particular consequence in an environment
where ethnic disparities in access to credit exist. In terms of trade credit, no such evidence has been
recorded. Therefore, this section briefly describes some facts about the use of trade credit among a
nationally representative sample of small businesses.
The data comes from the National Survey of Small Business Finances (NSSBF), a survey
conducted periodically by the Board of Governors of the Federal Reserve and the Small Business
Administration of for-profit, nonfarm, nonfinancial businesses with fewer than 500 employees. The
latest survey, conducted in 1994 and 1995 to approximate the population of businesses in operation
in 1993, includes a minority oversample, allowing us to more precisely account for the business
practices of Hispanic, Black, and Asian-owned firms. After excluding firms in the finance, real
estate, and insurance industries, our final sample includes 4,318 firms, of which approximately 9.7
percent are primarily owned by Blacks, 6.7 percent by Hispanics, and 6.8 percent by Asians.7 When
weighted to represent a national sample, Black, Hispanic, and Asian-owned businesses account for
3.0, 4.5 and 3.6 percent, respectively, of all small nonfinancial, nonfarm firms.
The NSSBF contains detailed information about the primary owner and the firm. Firm
characteristics include finances, performance, financial relationships, industry, organizational form,
and location. Owner information includes education, experience, gender, past financial problems,
and race. The survey also reports a rich set of trade credit questions, including:
•
•
•
•
•
•
•
Did the firm purchase any goods or services on account in the last year?
Has any supplier that offers trade credit denied a request by your firm?
From how many suppliers did the firm make purchases on account during 1993?
What percent of purchases were made on account in 1993?
What portion of suppliers offered cash discounts for prompt payment?8
What portion of the cash discounts offered did the firm take advantage of?
What portion of payments on account was made after the due date in 1993?
Table 1 contains weighted means by ethnic group for a number of firm and owner
characteristics and the trade credit measures. The indicators of statistical significance in Table 1
represent tests of differences in means relative to White-owned businesses. These tests clearly
show statistically and economically important differences in the use of trade credit between Whiteand minority-owned firms. Minority owners, on average, are less likely to use trade credit for their
7
Financial firms are excluded because their balance sheets are hard to compare with those of other businesses. The
ethnicity of a firm is defined as that which owns at least 50 percent of the firm.
purchases, have fewer suppliers, and are less likely to take up cash discounts for early payment. In
general, these differences relative to White-owed firms are larger for Black-owned businesses than
businesses owned by Hispanics and Asians. Finally, given our interest in the Black-Hispanic
comparisons in the Chicago neighborhood data, we note that there are some statistically significant
differences between these two minority groups. Hispanic firms are less likely to have any trade
credit last year from a supplier, but have more suppliers on account, are more likely to take
advantage of cash discounts, and are less likely to have payments past due.
However, as can be seen at the bottom of Table 1, there are many other differences between
the groups. For example, minority firms tend to be smaller, younger, have more financial problems,
and have fewer strong ties to financial institutions. Many models of trade credit imply that these
covariates could be correlated with both ethnicity and credit usage. For example, an empirical
implication of models that feature credit rationing are that measures of credit quality, such as size,
cash flow, and access to bank lending, should result in more trade credit being offered to a buyer
(Smith 1987; Biais and Gollier 1997). However, these models predict that buyers with more cash
flow and access to bank financing will use less trade credit, which is defined as a high-cost source
of credit. Petersen and Rajan (1997) find empirical support for these propositions. To the extent
that there are inter-industry differences and intra-industry similarities in the severity of information
and adverse selection problems, we would predict relatively wide variation in credit terms offered
across industries and little variation within industries. Ng, et al (1999) confirm this prediction and
conclude that it is indeed related to information problems.9 Therefore, a more detailed statistical
analysis is needed to measure the association between ethnicity and the use of trade credit.
8
The possible responses to the first two questions are yes and no. The five possible responses to the cash discount
question are none, fewer than half, half, more than half, and all/almost all.
9
Models of trade credit that emphasize the ability of suppliers to more effectively salvage collateral in the wake of
default, such as Frank and Maksimovic (1998), imply that arrangements that increase the value of collateral should
encourage the use of trade credit. Since finished goods have been transformed from the original purchased inputs, a
higher proportion of inventory held as finished goods implies a lower value as collateral for a supplier. Petersen and
The multivariate analysis controls for basic differences in firm and owner characteristics,
including the gender and education level of the owner, and the geographic region, urban status, and
two-digit industry of the business. We also include firm characteristics that may be associated with
the use of trade credit. Many of these factors are potentially endogenous but were included to be
comparable to the literature. Some, such as the firm's age, assets, number of employees, profits,
sales growth, and whether the business is incorporated, publicly traded, or a franchise, reflect the
size and quality of the firm. More directly related to credit quality are a series of questions about
owner and firm credit history. Respondents are asked whether the firm was delinquent on business
obligations in the last three years, the principal owner declared bankruptcy in the last seven years,
the owner was delinquent on personal obligations in the last three years, or a legal judgement was
rendered against the owner in the last three years. Finally, we include measures of the extent of a
firm's lending relationships with banks that have been used in previous research on trade credit,
including the length of the longest relationship with a financial institution, the size of the firm's
financial network, a measure of the concentration of the firm's banking relationships, and the
number of banking services used.10 Unfortunately, there is no information in the NSSBF that can be
used to directly measure ties with suppliers. We return to this issue later in the paper.
We do not report the marginal effects for the control variables for each regression because of
space limitations. However, we can make the general statement that asset size, owner and firm
credit history, and industry type are consistently important correlates with the various measures of
Rajan (1997) find that a measure of this proportion at the industry level is negatively related to the supply of trade
credit. However, data limitations force the use of a measure of the proportion of finished goods inventory at the level of
the industry rather than at the level of the firm. Unfortunately, this means that the inventory measure may be picking up
a variety of industry effects. Turning to a model that emphasizes nonfinancial reasons for extending trade credit, one
possible explanation is that suppliers with market power use trade credit to increase profits by price discrimination
(Brennan, Maksimovic, and Zechner 1988). Lee and Stowe (1993) and Ferris (1981) also present models that focus on
operational reasons for trade credit. Empirically, Petersen and Rajan (1997) find that account receivables are positively
related to gross profit margins, which is consistent with the price discrimination model of trade credit.
10
Peterson and Rajan (1997) interpret some of these measures as proxies for relationships with financial institutions in
statistical models of trade credit. The concentration measure is a Herfindahl index of the firm's banking connections,
defined in Uzzi (1999).
trade credit usage. This is seen in appendix 1, which presents background calculations on the
relative importance of each of the observable firm and owner factors in accounting for the raw
ethnic differences in five representative trade credit variables.11
Table 2 reports our main findings on ethnic differences in trade credit usage. Each row
represents a different regression. The regressions take into account whether the measures are
discrete (probits), discrete and ordered (ordered probit), or censored (tobits) and are weighted to
account for the sampling design. All of the reported results are marginal effects (with robust HuberWhite standard errors in parentheses), which are interpreted as ethnic differences when the control
variables are held constant.
A number of ethnic differences are indicated by the results in the table. There appears to be
no statistical difference between Black-, Asian- and White-owned firms in whether any trade credit
at all is used over the last year. However, there is a large, statistically significant, difference
between Hispanic- and White-owned businesses. Hispanic firms are 7.7 percent (standard error of
4.4 percent) less likely than White firms and 5.5 percent less likely than Black firms to have used
any trade credit in the last year, although the latter point estimate is not significant at conventional
levels. The sample size of the Hispanic and Black sample does not allow us to measure this gap
very precisely. However, the results suggest that differences in access to the first dollar of trade
credit may exist between Black and Hispanic firms, and, more strongly, White and Hispanic firms.
11
The computations are derived from weighed linear probability models using standard Blinder-Oaxaca
decompositions. The ethnic gap can be written as Y W − Y B = ( X W − XB )βW + X B (βW − βB ) where Y is the
dependent variable, X is the vector of independent variables, and W and B index the two ethnic groups. In appendix 1,
the first row reports the raw ethnic gap or Y W − Y B . The second row shows the fraction of that gap that is explained
by differences in population characteristics or ( X W − XB )βW . The third row reports the unexplained portion of the
ethnic gap, the share that is due to differences in the coefficient estimates or XB (βW − βB ) . Finally, the bottom panel
reports the share of the gap that is attributable to differences in the mean characteristics of each independent variable
across ethnic groups. An alternative and equally valid representation of this decomposition is to use the base case
However, conditional on a single dollar of trade credit being offered to the firm, Blackowned businesses use trade credit less than White- and perhaps Hispanic-owned firms in most
measures that we analyze. White and Hispanic differences tend to be small and statistically
insignificant. White-Asian gaps are less easily explained, as they appear to depend on the trade
credit measure being analyzed.
For example, relative to White and Hispanic owners, Black owners have 11.6 and 12.5
fewer suppliers and make 6.5 and 6.1 percent fewer purchases on account, after conditioning on
firm size, resources, and industry.12 These results are statistically significant at the 6 percent or
better level, with the exception of the Black-Hispanic gap in percent purchases. The latter estimate
is again suggestive but lacks the sample sizes to obtain finer estimates. Like Black-owned firms,
Asian firms have a smaller set of suppliers, but the percent of purchases that they have on account
look more like White or Hispanic firms.
The Black-White and Black-Hispanic differences in the use of cash discounts and
prevalence of overdue payments are perhaps the strongest ethnic disparities shown in the table.
Among firms that are offered cash discounts, Black owners are 4.9 percent more likely to never use
these discounts and 13.2 percent less likely to always use them relative to White owners.13 Relative
to Hispanic owners, Black owners are 3.2 percent more likely to never use discounts, 8.0 percent
less likely to always use discounts, and 9.6 percent less likely to never have payments past due.14
Additionally, Black owners are 11.0 percent less likely than White owners to never have payments
ethnic group B to compute the explained portion ( X W − XB )βB . However, because of the small minority sample sizes,
this decomposition is not as precisely estimated.
12
We report weighted least squares estimates of the percent purchases regression for those firms with some trade
credit and tobit estimates to account for zero censoring in the full firm sample. The tobit standard errors are derived
using the delta method (Greene 1999).
13
The marginal effects are reported in brackets. The numbers in [ ] brackets are marginal effects at the never response.
The numbers in { } brackets are marginal effects at the always response. Marginal effects computed at the less than
half, half, and more than half responses are available upon request.
14
For example, the 3.2 percent Black-Hispanic difference in never using cash discounts is calculated as the Black
marginal effect (4.9 percent) minus the Hispanic marginal effect (1.7 percent).
past due. All of the Black-White estimates are statistically significant at the one percent level, and
the Black-Hispanic differences are significant at the ten percent level. Moreover, the large BlackWhite and Black-Hispanic gaps are conservatively measured in that the regressions include controls for
owner and firm financial distress. Whether the firm has been delinquent in payments over the last three
years may, by definition, be related to whether the firm is past due on trade credit. Not surprisingly,
when the owner and firm delinquency variables are excluded from the regressions, the ethnic gaps are
substantially larger, and the Black-Hispanic differences are significant at better than the 5 percent level.
The Asian-owned results are mixed. Cash discounts used are at a level similar to Black
firms but past payments due are similar to White or Hispanic firms. However, the lack of cash
discount used by, as well as offers to, Asian-owned firm is unlikely to be due to short-term credit
problems. Asian firms are less likely to self-report a need for additional short-term credit (as shown
in the row labeled short-term credit needed) relative to all other groups.
In sum, the evidence points to some differences between ethnic groups in the use of trade credit
that cannot be accounted for by observed characteristics of the firms and owners. The differences are
most striking when comparing Black- and White-owned businesses, but we also observe some
differences between the other ethnic groups, including Black and Hispanic firms. The next section
turns to new evidence from the neighborhood business surveys. We use the neighborhood data to
examine the importance of supplier relationship measures that are not observed in the national
survey. In particular, we know the ethnicity of suppliers, their proximity to the neighborhood
businesses, and how long the supplier and business owner have worked together. Although these
variables may be rough proxies for the depth of supplier ties, they have a bearing on the existence of
relationships between buyers and suppliers. Our strategy is to use this information to assess the
importance of supplier relationships in delineating ethnic differences in the use of trade credit.15
Neighborhood survey results
In order to shed some light on small business finance in ethnic communities, the Federal
Reserve Bank of Chicago and researchers from the University of Chicago cooperated in conducting
surveys in two Chicago neighborhoods, Little Village, a predominantly Hispanic community, and
Chatham, which is predominantly Black.16 These communities were chosen as the sites of these
studies because they are distinct and well-recognized ethnic neighborhoods with viable small
business sectors. Although the bulk of the owners interviewed are either Black or Hispanic, other
ethnic groups are represented. The survey instruments are designed to elicit information about
ethnic relationships, informal sources of financing -- such as loans or gifts from family and friends,
and formal sources of funds for both households and businesses.
Little Village is a predominantly Hispanic area, mostly of Mexican origin, on the southwest
side of Chicago with a population of 81,155 persons and a median family income of $23,259, as of
the 1990 census. Substantial numbers of Hispanics migrated into the community beginning in the
1960s and the area became predominantly Hispanic in the 1970s. Chatham is a mostly Black
community on the south side of Chicago with a 1990 population of 36,779 persons and a median
family income of $29, 258. Chatham became predominantly Black during the 1950s (Chicago Fact
Book Consortium 1995).
In both communities, the survey universe was constructed by canvassing and enumerating
all identifiable existing businesses. A stratified random sample was then drawn in which relatively
common businesses, such as eating places and hair salons, were undersampled. In both surveys,
15
The surveys also ask about family relationships between businesses and their suppliers, but, in practice, almost no
firms have such family ties.
medical and legal professionals were excluded from the sample on the grounds that the educational
requirements for these fields result in entrance and financing decisions that have little in common
with those of other small businesses. Field staff, bilingual in the case of Little Village, then
contacted the businesses in the selected samples for an interview that required about one-and-a-half
hours. The fieldwork resulted in response rates of 70 percent for Little Village and 57 percent for
Chatham. About one-third of all enumerated businesses were interviewed in Little Village, and the
corresponding figure for Chatham is about one-quarter.17
Business and owner characteristics
The types of business by ethnic group are shown in Table 3. Asian owners are primarily
Korean, and Other is made up of owners from the Middle East, India, and Pakistan. Because
relatively few White, Asian, and Other owners are sampled, we focus on Black and Hispanic
business owners in the following discussion. For all ethnic groups combined, the bulk of the firms
fall into some variety of the retail or service sector. Within groups, Black owners have a relative
concentration in the service sector. Hispanic firms are relatively balanced across the industry types,
as no one category contains more than 25 percent of the total. The average age of the current
business for all groups is about 9 years, and firms owned by Blacks (11.6 years) tend to be older
than Hispanic-owned firms (6.9 years). Most of these firms employ relatively few workers, as the
average number of employees for all groups is 4.5 workers.
About one-third of all owners are women, and Hispanic and especially Black owners are
more likely to be women. Overall, the bulk of the firm owners are at least high school graduates,
and about a quarter of them have a college degree. However, educational attainment varies across
the racial/ethnic groups. The proportion of Hispanics in the sample who do not have a high school
16
See Bond and Townsend (1996) for a description and some findings from the Little Village Surveys for households
and businesses. See Huck et al (1999) for an overview of business finance in these neighborhoods.
17
The survey fieldwork was conducted during 1993-94 and 1997-98 in Little Village and Chatham, respectively.
degree (49.4 percent) is over three times as high as the corresponding proportion for Blacks (16
percent), the group with the next highest figure. Hispanic owners (9 percent) are least likely to have
a college degree and a relatively low proportion (70.5 percent) are, by their own assessment,
moderately or extremely proficient in English.
Because of our interest in business networks, we include two measures of interaction
between business owners. The first is an indicator variable for whether or not an owner is a
member of a formal business association. The second is an indicator variable for whether or not an
owner reports that he/she regularly meets with other business owners to talk about business issues.
Hispanic owners (22.9 percent) are somewhat less likely to be members of a formal association
relative to Black owners (32.1 percent), but they are somewhat more likely to have other business
owners to talk with informally, although the latter difference is not statistically significant.
Relative to the NSSBF sample, the neighborhood sample businesses are more likely to be
owned by women, are headed by owners with somewhat less education, are somewhat older firms,
and have fewer employees. One of the most important differences between the two samples is that
the neighborhood survey is much more heavily weighted towards retail establishments (66.7
percent) compared to the NSSBF sample (23.3 percent). It is important to keep in mind that since
almost all of the Hispanic owners are in Little Village and all of the Black owners are in Chatham,
these ethnic categories combine location and ethnic effects.
Some measures of the use of trade credit and supplier relationships are also shown in Table
3. Information for up to three suppliers was elicited from an owner, and the results are tabulated by
the ethnicity of the business owner. We measure the use of trade credit by an indicator variable for
whether or not a supplier offers trade credit to a business owner. Trade credit is available to many
of the businesses in Little Village and Chatham, as 49.7 percent of the suppliers in the sample offer
credit.18 Similar proportions of Hispanic (44.4 percent) and Black owners (42.4 percent) are offered
credit by their suppliers; owners in the other ethnic groups are more likely to be offered credit.
Hispanic (32.9 percent) and Black owners (30.8 percent) are also about equally likely to work with
a supplier of the same ethnicity. Hispanic-owned businesses (5.6 years) have a shorter relationship
with suppliers on average than do Black-owned businesses (7.6 years), and the difference may in
part reflect the fact that Hispanic-owned businesses in the sample are younger than Black-owned
businesses. The supplier locations are divided into three categories that form the basis for indicator
variables. In the first category, the supplier is in the same or adjacent neighborhood as the business.
In the second category, the supplier is outside of the neighborhood but within the Chicago MSA. In
the last category, the supplier is outside the MSA. The table shows that Hispanic owners (38.5
percent) are more likely to deal with suppliers in their neighborhood relative to Black owners (19.9
percent), and correspondingly less likely to deal with suppliers elsewhere in the MSA or beyond.
Trade credit offered results
The first point to make is that Hispanic and Black owners are about equally likely to be
offered trade credit in our two neighborhoods. As shown in Table 3, Hispanic owners are offered
credit by 44 percent of their suppliers and the corresponding figure for Black owners is 42 percent.
This is without conditioning on any other variables. If we condition on the industry and
demographic control variables noted below, we get the same result. However, the way that
businesses in these neighborhoods are tied to their suppliers seems to be quite different depending
on the ethnicity of the owner. The purpose of the regression analysis is to shed some light on how
supplier relationships are associated with the offer of credit.
We have information on up to three suppliers for each business, and for regression models
using each business and supplier pair as the unit of observation we would expect that because of
18
Of businesses that do have trade credit offered to them, a majority (67.9 percent) owe a supplier at the time of the
business-specific unobservables the error terms within a business are probably correlated.
Accordingly, we present probit results with robust standard errors adjusted for multiple
observations on a business. We also report the results of a random effects probit model as a
robustness check.19
The regressions include some control variables that can be grouped into a number of
categories. Owner demographic variables include indicators for education, gender, and proficiency
in English. We also include an indicator variable for whether or not an owner previously owned a
business and a self-reported measure of an owner’s disposition towards risk on a 1-to-5 scale, 5
being the most willing to risk all in a new business. Standard measures of firm characteristics
include industry type (not reported in the results), size as measured by the natural log of the number
of full-time equivalent employees (and the square of this term), and business age. Measures related
to firm quality or need for credit include indicator variables for a business reporting that it was in
danger of failing within the last three years, for reporting sales growth as an important challenge,
and for providing credit to customers. Indicator variables for having an account relationship with a
bank and for using an accountant are included as measures of a relationship with financial
institution and financial transparency.20
Our measures of supplier relationships include the ethnic tie between an owner and a
supplier, the geographic proximity of the supplier, and the length of time the owner has worked
with a particular supplier.21 Indicator variables for being a member of a trade association and for
survey. The median amount owed for those owners who do have trade credit outstanding is $3,095.
19
We find that the empirical results are qualitatively similar for both estimation models. Monte Carlo results reported
by Guilkey and Murphy (1993) suggest that the probit estimator with robust standard errors performs reliably for a
variety of parametric configurations. Their results also suggest that the random effects probit estimator is not as reliable
when the number of observations in a cluster equals two, which is similar to our application. We thus focus our
discussion on the results of the robust errors probit estimator.
20
Unfortunately, we do not have information on the credit history of the owner and firm, which are important
determinants of the credit worthiness of the business.
21
The length of the supplier relationship and working with a co-ethnic supplier are potentially endogenous to the trade
credit decision. We were unsuccessful in an attempt to use migration history for Hispanics as instruments for working
with a co-ethnic supplier.
being part of a group to talk about business issues are included as general measures of networking
with other business owners. We include interaction variables in some specifications that allow us to
test for differences across the ethnic groups in the impact of whether or not the supplier is the same
ethnicity as the owner and for supplier location. The marginal effects and standard errors for
several specifications of a probit model of trade credit outcomes are reported in Table 4. Since we
are focusing our attention on the supplier relationship measures, coefficients for the control
variables are not reported in the table and appear in Appendix 2. Note that we have combined the
businesses not owned by Blacks or Hispanics into the White/Asian/Other category because of the
low number of businesses for these ethnic groups. Accordingly, we focus on Black and Hispanic
owners in the discussion that follows.
The first two columns of Table 4 report alternate specifications for the length of the supplier
relationship. The first column shows the results of using the log of years with supplier and the
square of this term to measure relationship length. However, a close look at the data suggests that
the relationship between the offer of trade credit and years with supplier is better captured with
dummy variables allowing breaks at three and seven years.22 Accordingly, the second column
presents results for a specification using indicator variables for years with supplier less than three
and greater than seven. We focus on this measure of years with supplier for the rest of the
discussion. The results suggest that a shorter relationship (under three years) and a longer
relationship (over seven years) increase the probability of being offered credit by 11.6 and 18.3
percent, respectively.23
Note that the marginal effects for the other variables of interest are little affected by the
construction of supplier tenure. Columns 1 and 2 show that the marginal effect of a Black business
owner dealing with a Black supplier of the same ethnicity is -0.129 and –0.146, suggesting that a
22
Our conclusions are not affected by using cutoffs other than 7 years.
Black-owned firm working with a Black-owned supplier is 13 to 15 percent less likely to get trade
credit offered. However, this effect is not statistically significant in either specification. On the
other hand, the comparable marginal effect for Hispanic-owned firms is a large and significant
0.19.24 Finally, working with local suppliers in the neighborhood or within Chicago has no
appreciable impact on trade credit offers.
However, column three indicates that there are important ethnic differences in the effect of
supplier location as well. This regression supplements column 2 by including interaction terms
between the ethnicity of the owner and supplier location. This alteration reveals a more
complicated story than the specification reported in the second column, for which supplier location
effects are small for the combined ethnic groups. The marginal effect for supplier in neighborhood
of –0.392 applies to a Black owner with a supplier in the neighborhood and is measured relative to a
Black owner with a supplier outside of the MSA. Thus, dealing with a supplier in the neighborhood
rather than one outside the MSA reduces the probability of a Black owner being offered trade credit
by about 40 percent. This is an extremely large effect and is also statistically significant. The
marginal effect of 0.001 for supplier in the Chicago MSA indicates that the effect of a Black owner
dealing with a supplier outside the neighborhood but still in the MSA relative to a supplier outside
the MSA is essentially zero. Thus, dealing with a supplier in the neighborhood is associated with a
lower probability of a Black-owned business being offered credit relative to suppliers elsewhere.
The marginal effect of a Hispanic owner dealing with a supplier in the neighborhood relative
to dealing with one outside of the MSA is to increase the probability of being offered credit by 16.5
percent (-0.392 + 0.557 = 0.165), which is economically large and statistically significant.
Likewise, the effect of a Hispanic owner working with a supplier out of the neighborhood but in the
MSA is to increase the chances of being offered credit by 35.6 percent (0.001 + 0.355 = 0.356).
23
These marginal effects are statistically significant at the 10 percent and the 5 percent level, respectively.
Thus, as opposed to Black-owned firms, for Hispanic owners, dealing with a supplier closer to
home relative to outside of the MSA is associated with being more likely to be offered credit.
Finally, the indicator variables for being a member of a business association and having
someone to talk about business issues can be thought of as proxies for general networking apart
from supplier relationships. The association between networking and the offer of trade credit is of
interest in its own right, and the regression results indicate that networking is associated with being
more likely to be offered trade credit.25
In order to confirm these results, we did several additional experiments. First, we report
results that control for random firm effects in column 4. If there are unobserved factors that are
specific to the firm and correlated with our supplier ties measures, we may confound these latent
factors with our measures of interest. Second, we respecified the unit of observation to be each
business rather than each business and supplier pair; therefore, trade credit offered is an indicator
variable that equals one if at least one supplier offer credits to a given business. Third, we ran OLS,
WLS, and tobit regressions with an alternative measure of the use of trade credit, the dollar amount
owed (results available upon request). In all three cases, the results are qualitatively similar to those
discussed above.
We also estimated separate regressions for Hispanic and Black businesses, thus allowing all
of the variable coefficients to vary by ethnic group. The marginal effects are reported in Table 5
and are qualitatively similar to those presented for the pooled sample. We also ran regressions for
the Black and Hispanic subsamples with interactions between the ethnic supplier and location
variables (not reported in the table). This allows the effect of working with a supplier of the same
ethnicity to vary by location, and vice versa. Our findings that closer ties are associated with the
24
This is the sum of the first two rows (ie. 0.318+-0.129 in column 1 or 0.340-0.146 in column 2.
It is not clear which way causation runs between networking measures and the offer of trade credit. It may be, for
example, that owners of higher quality firms tend to be members of business associations.
25
offer of credit for Hispanic owners are quite robust to this alternative specification. However, the
finding that working with a neighborhood supplier reduces the probability of being offered credit
for Black owners is less robust in that it only holds for owners working with non-Black suppliers.
Lastly, an extension of our empirical description of the relationships between businesses and
their suppliers would be to add more information about the supplier side. For example, Petersen
and Rajan (1997) have shown that larger asset size is associated with a firm offering more credit as
measured by accounts receivables. Our finding that ethnic ties are associated with the offer of trade
credit raises the question of whether supplier characteristics also vary systematically by ethnicity.
We cannot explore this issue directly because the national and neighborhood surveys do not provide
more information about suppliers of the sample businesses beyond what we have already discussed.
However, we can get some idea of what might be found by looking at the NSSBF businesses from a
different perspective. That is, we look at how the firms provide credit to customers, as measured by
accounts receivable, rather than how they receive trade credit.
We report some measures of credit offered to customers in Table 6. In order to sharpen the
focus on providers of trade credit, we restrict the sample to firms in the manufacturing, wholesale
trade, and transportation sectors. We can see that a lower proportion of Black- and Asian-owned
businesses report having any accounts receivable, relative to White-owned businesses. More
startling, Black-owned businesses have less than half of the accounts receivables (in dollars) of any
other ethnic group, both with and without conditioning on having any receivables at all. However,
if we normalize by asset size, we see that Black-owned businesses have a ratio of receivables to
assets similar to that of White-owned businesses. This suggests that the relatively low levels of
receivables reflect the lower asset size of Black-owned firms. Although Black-owned suppliers of a
given size look similar to other firms in terms of their accounts receivables, the fact that they tend to
be smaller is asset size may have an effect on a buyer’s chances of being offered credit.26
Tests for Ethnic Networks in National Data
The results from the Chicago neighborhoods are striking. To determine whether these
results can be generalized, we employ the NSSBF to test whether ethnic differentials in trade credit
usage are reduced when there are more businesses of the same ethnicity in the local area. Using
data from the 1992 Census of Minority-Owned Businesses, we construct six MSA-level measures
of the presence of Hispanic-, Black-, and Asian-owned businesses for each urban NSSBF firm.27
The measures are the share of all (1) firms, (2) sales, (3) firms with paid employees, (4) sales of
firms with paid employees, (5) employees at firms with paid employees, and (6) payroll of firms
with paid employees represented by firms owned by members of a particular ethnic group.28
Using these data, we reestimated selected regressions from table 2. Because information on
minority representation in non-MSA areas was not available, the sample was restricted to the 3441
urban firms in the NSSBF. In addition, the base specification in the table 2 regressions was
augmented to include a variable interacting a firm’s ethnic ownership (e.g., Hispanic) and the
measures of the presence of firms owned by members of the same ethnicity (e.g., share of Hispanic
firms in the MSA). Because there are six measures of minority-firm presence, regressions are run
separately for each of the six MSA measures. The coefficient on the interaction terms provide some
evidence as to whether being in an MSA with a higher fraction of firms that are of the same
ethnicity is associated with the extent of trade credit use.29
26
Furthermore, the 1992 Census of Minority Owned Businesses report over 15,000 Hispanic-owned businesses with
paid employees in the manufacturing, wholesale trade, and transportation (less taxi services) sectors compared to less
than 7,000 Black-owned firms with paid employees in the same sectors.
27
We obtained special access to the proprietary MSA identifier for each firm in the NSSBF.
28
The six MSA measures are highly correlated (greater than 0.92) for the Hispanic and Asian measures. For the Black
firms, the correlations, which range from 0.60 to 0.95, are lower but still strong,
29
All six measures of minority business presence are used because there is little guidance suggesting which measure
might be best. For example, the adjusted R-squared in the regressions are all quite similar.
This is admittedly a very blunt instrument. Unlike the Chicago neighborhood data, which
includes information about each firm’s suppliers, the NSSBF data only allows for the use of data on
the pool of co-ethnic firms operating in the local MSA market without regard to whether any of
them actually provides trade credit. Furthermore, the broad ethnic categories used in the NSSBF
and Census may obscure potentially important differences in countries of origin. For example, if
Korean businesses rarely use non-Korean suppliers, the share of Asian firms may be a poor proxy of
potential Korean suppliers in the area. These mismeasurement problems will tend to bias the MSA
interaction coefficients towards zero.
Table 7 summarizes the six regression runs for each aspect of trade credit usage.
The six Black, Hispanic, and Asian interactions are grouped by sign and significance in the first,
second, and third group of columns, respectively. In each column group, the first subcolumn,
labeled “differential reduced,” reports the number of interaction coefficients that are statistically
significant at the 10 percent level and that associate a higher MSA share for co-ethnic businesses
with a smaller racial differential in access to trade credit. The second subcolumn, labeled
“differential increased,” reports the number of interaction coefficients that are statistically
significant at the 10 percent level and that associate a higher MSA share for co-ethnic businesses.
Insignificant interaction coefficients are reported in the third subcolumn.
Thus, the first row shows that none of the interaction coefficients for Black-owned
businesses are statistically significant in the six probit regressions of whether the firm used trade
credit last year. This suggests there is no association between a location in an MSA with more
Black businesses (however measured) and differences in the use of trade credit in the last year by
White-owned and Black-owned firms.
As the table indicates, this exercise is not always definitive. This is not surprising, given
that we subject a relatively weak proxy to a tough test. However, some patterns emerge.
There is some evidence suggesting that the presence of other Hispanic-owned businesses in
an MSA helps individual Hispanic-owned firms obtain and use trade credit relative to White-owned
firms. In particular, the location of an Hispanic-owned small business in an MSA with more
Hispanic-owned businesses is associated with reductions in the frequency of cash discount
offerings, the use of cash discounts, and the likelihood of having ever been rejected for trade credit.
The size of these ethnic neighborhood effects vary, but can be substantial. For example, the racial
differential for being offered a cash discount faced by a Hispanic-owned small business located in
an MSA with an average presence of Hispanic-owned firms was almost 17 percent lower (in
percentage points) than the differential the same firm would face if it were located in an MSA with
no Hispanic-owned firms.30 A much larger ethnic network effect is observed for the use of a cash
discount (a 44 percent reduction), but a smaller effect is observed for trade credit rejection (a 9
percent reduction). Only in one instance— trade credit last year— is there a significant relationship
between the interaction term and a firm’s trade credit relationships.
The national survey results for Hispanic-owned firms are thus generally consistent with the
neighborhood findings. That is, a higher MSA share for Hispanic-owned businesses, which offers
more opportunity for supplier ties, is generally associated with reductions in the trade credit
differentials between Hispanic-owned and White-owned firms, and very rarely associated with
increases in such differentials.
For Black-owned firms, as in the case of the neighborhood findings, the results are mixed.
Only one trade credit measure— cash discount used— suggests that location in an MSA with a
higher Black-owned business presence is associated with reductions in Black-owned/White-owned
firm trade credit differentials. For two other measures— ever rejected for trade credit and percent of
30
This is the average reduction in the racial differential over the 6 interaction terms used.
purchases on account— it appears that a higher Black-owned business presence is associated with
increases in differentials.
Finally, the results show no statistically significant correlations for Asian-owned firms.
However, among 6 of the 8 trade credit measures, there is a positive association between trade
credit usage by Asian-owned firms and the local presence of Asian-owned businesses.31
Conclusion
A goal of this paper is to provide better measurement of supplier relationships and ethnic
ties in order to guide theoretical treatement of these issues. We document several facts about the
empirical importance of supplier relationships for the use of trade credit by minority-owned small
businesses. First, using nationally representative data, we show economic and statistically
important disparities in the use of trade credit among ethnic groups. In particular, Black-owned
businesses use less trade credit, are less likely to take advantage of discounts for early payment, and
are more likely to have payments past due. Second, we use our Chicago neighborhood data to show
that closer relationships with suppliers as measured by ethnic ties and geographical proximity are
associated with more trade credit for Hispanic-owned businesses but not Black-owned firms. We
also show that a relatively long relationship with a supplier is associated with more trade credit for
both Black and Hispanic businesses. Finally, turning back to the nationally representative data, we
explore how location in an MSA with a higher proportion of businesses of the same ethnicity is
associated with the use of trade credit by minority owners relative to White-owned firms. We find
that a higher MSA share for Hispanic-owned businesses is generally associated with a reduction in
differences in the use of trade credit by Hispanic owners relative to White owners. No clear
31
We also ran the regressions with standard error adjustments for MSA clustering. The results are roughly similar; if
anything, the cluster analysis strengthens the previously observed ethnic network differences between Hispanics and
Black firms. In particular, the cluster adjustment increases the Black-White differential among the ever rejected,
percent purchases on account, and cash discount offered measures but decreases the Hispanic-White differential among
the ever rejected and cash discount used variables. The adjustment also reduces the Asian-White ever rejected for trade
credit differential.
correlation is apparent between the MSA share for Black-owned businesses and their use of trade
credit. Thus, the ethnic differences in the effects of relationships evident in the neighborhood
surveys seem to be consistent with the results from the national survey.
One way minority firms may deal with disadvantages relative to White firms could be to
cultivate ties to suppliers in an ethnic network in order to ameliorate ethnic disparities in access to
trade credit. The two surveys provide little evidence that closer relationships or ties with suppliers
are associated with better access to trade credit for Black owners, whereas we do find evidence that
closer supplier relationships is tied to trade credit for Hispanic owners. These results lead us to
offer the conjecture that the ethnic differences in the use of trade credit in the national survey
sample may potentially be due, in part, to differences in relationships between owners and their
suppliers. We find some support for this proposition using the rough relationship proxies available
in the national data.
References
Berger, Allen N., and Gregory F. Udell, 1998, “The economics of small business finance: The roles
of private equity and debt markets in the financial growth cycle,” Journal of Banking and Finance,
Vol. 22, pp. 613-673.
Berger, Allen N., and Gregory F. Udell, 1995, “Relationship lending and lines of credit in small
firm finance,” Journal of Business, Vol. 68, No. 3, pp. 351-81.
Biais, Bruno, and Christian Gollier, 1997, “Trade credit and credit rationing,” The Review of
Financial Studies, Vol. 10, No. 4, pp. 903-937.
Blanchflower, David, Phillip Levine, and David Zimmerman, 1998, “Discrimination in the small
business credit market,” National Bureau of Economic Research, Working paper, No. W6840.
Bond, Philip, and Robert Townsend, 1996, “Formal and informal financing in a Chicago ethnic
neighborhood,” Economic Perspectives, July/August, pp. 3-27.
Bostic, Raphael and K. Patrick Lampani, 1999, “Racial differences in patterns of small business
finance: The impact of local geography,” Business Access to Capital and Credit: A Federal
Reserve System Research Conference: Proceedings of a Conference Held in Arlington, VA, March
8-9, 1999, editors, Jackson L. Blanton, Alicia Williams, Sherrie L.W. Rhine, Chicago: Federal
Reserve Bank of Chicago.
Brennan, Michael J., Vojislav Maksimovic, and Josef Zechner, 1988, “Vendor Financing,” The
Journal of Finance, Vol. 43, No. 5, December, pp. 1127-1141.
Cavalluzzo, Ken S., and Linda C. Cavalluzzo, 1998, “Market structure and discrimination: The case
of small businesses,” Journal of Money, Credit, and Banking, Vol. 30, No. 4, pp. 771-792.
Cavalluzzo, Ken S., Linda C. Cavalluzzo, and John Wolken, 1999, “Competition, small business
financing, and discrimination,” Business Access to Capital and Credit: A Federal Reserve System
Research Conference: Proceedings of a Conference Held in Arlington, VA, March 8-9, 1999,
editors, Jackson L. Blanton, Alicia Williams, Sherrie L.W. Rhine, Chicago: Federal Reserve Bank
of Chicago.
Cole, Rebel A., and John D. Wolken, “Financial services used by small businesses: evidence from
the 1994 national survey of small business finances,” Federal Reserve Bulletin, July, 1995, pp. 629667.
Fafchamps, Marcel and Bart Minten, 1999, “Relationships and traders in Madagascar,” The Journal
of Development Studies, Vol. 35, No. 6, pp. 1-35.
Frank, Murray, and Vojislav Maksimovic, 1998, “Trade credit, collateral, and adverse selection,”
University of British Columbia, working paper.
Greene, William, 1999, "Marginal Effects in the Censored Regression Model," Economic Letters,
Vol. 64, pp. 43-49.
Guilkey, David K., and James L. Murphy, 1993, “Estimation and testing in the random effects
probit model,” Journal of Econometrics, Vol. 59, No. 3, October, pp. 301-317.
Huck, Paul, Sherrie L.W. Rhine, Philip Bond, and Robert Townsend, 1999, “Small business finance
in two Chicago minority neighborhoods,” Economic Perspectives, Second Quarter, pp. 46-62.
Lee, Yul W., and Stowe, John D., 1993, “Product risk, asymmetric information, and trade credit,”
Journal of Financial and Quantitative Analysis, Vol. 28, No. 2, June, pp. 285-300.
Mian, Shehzad L., and Clifford W. Smith, Jr., 1992, “Accounts receivable management policy:
Theory and evidence,” The Journal of Finance, Vol. 47, No. 1, March, pp. 169-200.
McMillan, John and Christopher Woodruff, 1999, “Interfirm relationships and informal credit in
Vietnam, The Quarterly Journal of Economics, Vol. 114, No. 4, pp. 1285-1320.
Ng, Chee K., Janet Kiholm Smith, and Richard L. Smith, 1999, “Evidence on the determinants of
credit terms used in interfirm trade,” The Journal of Finance, Vol. 54, No. 3, June, pp. 1109-1129.
Petersen, Mitchell A., and Raghuram G. Rajan, 1994, “The Benefits of Lending Relationships:
Evidence from Small Business Data,” The Journal of Finance, Vol. 49, No. 1, March, pp. 3-37.
Petersen, Mitchell A., and Raghuram G. Rajan, 1997, “Trade credit: Theories and evidence,” The
Review of Financial Studies, Vol. 10, No. 3, pp. 661-691.
Prescott, Edward S. and Robert Townsend, 2000, “Inequality and boundaries in collective
organizations,” University of Chicago, Working paper, May.
Smith, Janet Kilholm, 1987, “Trade credit and informational asymmetry,” The Journal of Finance,
Vol. 42, No. 4, September, pp. 863-872.
Uzzi, Brian, 1999, “Embeddedness in the making of financial capital: How social relations and
networks benefit firms seeking financing,” American Sociological Review, Vol. 64, August, pp.
481-505.
Table 1
Descriptive Statistics, NSSBF 1993 1
White owned firms
Sample
Mean
Hispanic owned firms
Black owned firms
Sample
Mean
Trade credit variables
Trade credit last year
Ever rejected for trade credit
Number of suppliers on account 2
Number of suppliers on account 3
Percent of purchases on account 2
Percent of purchases on account 3
Cash discount offered 2
Cash discount used 4
Cash discount used 2
Payments past due 2
3,293
3,293
2,353
3,293
2,353
3,293
2,353
1,770
2,353
2,353
0.67
0.06
31.62
21.05
0.73
0.49
2.48
3.50
2.75
1.82
418
418
276
418
276
418
276
156
276
276
0.63
0.13
13.11
8.20
0.60
0.38
2.23
2.70
1.94
2.22
Demographic and firm variables
Female owned
High school dropout
High school graduate
Some college
College graduate
Post college graduate
Owner declared bankrupcy last 7 years
Judgments rendered against owner, 3 yrs
Owner delinquent 1-2 times last 3 years
Owner delinquent 3 times last 3 years
Firm delinquent 1-2 times last 3 years
Firm delinquent 3 times last 3 years
Firm age
Firm was founded by owner
Firm was purchased by owner
Firm was inherited by owner
Firm is publicly traded
Number of FT employees
Log assets
Corporation
Franchise
3,293
3,293
3,293
3,293
3,293
3,293
3,293
3,293
3,293
3,293
3,293
3,293
3,293
3,293
3,293
3,293
3,293
3,293
3,293
3,293
3,293
0.20
0.05
0.24
0.26
0.25
0.21
0.03
0.04
0.05
0.08
0.07
0.12
14.49
0.74
0.20
0.06
0.00
9.02
11.12
0.50
0.02
418
418
418
418
418
418
418
418
418
418
418
418
418
418
418
418
418
418
418
418
418
0.23
0.03
0.15
0.36
0.25
0.20
0.05
0.15
0.15
0.22
0.15
0.19
11.99
0.87
0.10
0.02
0.00
6.32
10.60
0.34
0.02
Sample
* *
* *
* *
* *
* *
* *
* *
* *
* *
**
**
**
**
**
**
**
**
**
**
**
**
**
**
**
**
Asian owned firms
Mean
289
289
166
289
166
289
166
104
166
166
0.56
0.10
19.86
11.13
0.67
0.38
2.21
3.09
2.26
1.88
289
289
289
289
289
289
289
289
289
289
289
289
289
289
289
289
289
289
289
289
289
0.24
0.11
0.31
0.21
0.26
0.11
0.04
0.09
0.07
0.15
0.09
0.16
12.02
0.78
0.19
0.02
0.00
6.70
10.91
0.34
0.01
Sample
* *
*
* *
* *
* *
* *
*
*
* *
* *
*
* *
* *
* *
* *
* *
* *
* *
* *
293
293
181
293
181
293
181
88
181
181
293
293
293
293
293
293
293
293
293
293
293
293
293
293
293
293
293
293
293
293
293
Mean
0.59
0.07
17.11
10.04
0.67
0.40
2.04
2.65
1.76
1.91
0.24
0.05
0.13
0.19
0.34
0.29
0.02
0.05
0.09
0.06
0.06
0.08
9.64
0.66
0.31
0.02
0.01
7.57
11.20
0.46
0.04
*
* *
* *
*
* *
* *
* *
* *
* *
*
* *
* *
*
*
*
*
*
*
*
*
*
*
*
White owned firms
Sample
Mean
Longest relationship with a financial
institution
Financial network size
Concentration of banking services
Complexity of services
Profits/assets
Median sales growth, 1990-1992
MSA
Hispanic owned firms
Black owned firms
Sample
Mean
Sample
Asian owned firms
Mean
Sample
Mean
3,293
9.91
418
7.95 * *
289
8.41
* *
293
7.20
* *
3,293
3,293
3,293
3,293
2,555
3,293
1.92
0.88
0.51
0.94
0.09
0.77
418
418
418
418
307
418
1.91
0.82
0.36
0.70
0.25
0.90
289
289
289
289
212
289
1.78
0.86
0.24
1.12
0.09
0.89
*
293
293
293
293
190
293
1.81
0.89
0.43
0.98
0.07
0.96
* *
**
**
**
**
* *
* *
Notes:
1
**(*)= significantly different from White owned firms at the 5 (10) percent level. All means are weighted using the NSSBF sampling weights.
34 mixed race or Native American firms are not included in table 1 but are included in tables 2 and 3. Six firms are identified as Hispanic and Black and
three as Hispanic and Asian. These nine firms are included in the table.
2
Sample of firms with any trade credit in last year.
3
All firms.
4
Sample of firms that were offered a cash discount.
Table 2
Supply and Use of Trade Credit by Minority-Owned Businesses, NSSBF 1993
Marginal effects (robust standard errors in parentheses) 1
Dependent variable
Estimation
Method
Weighted
Hispanic
Black
Trade credit last year
probit
-0.022
(0.032)
-0.077
(0.044)
Ever rejected for trade credit
probit
0.024
(0.013)
* *
Log (no. of suppliers on acct +1)
(trade credit users only)
WLS
-11.640
(2.040)
Log (no. of suppliers on acct +1)
(full sample of firms)
WLS
Perc of purchases on account
(trade credit users only)
Sample
Size
Asian
*
Wald P-statistic
Black vs
Black vs
Hispanic
Asian
-0.039
(0.041)
4,318
0.312
0.744
0.019
(0.017)
0.005
(0.012)
4,318
0.815
0.283
* *
0.930
(2.700)
-14.430
(2.910)
* *
2,986
0.001
0.227
-7.320
(2.040)
* *
-4.860
(3.150)
-11.160
(2.550)
* *
4,318
0.512
0.240
WLS
-0.064
(0.024)
* *
-0.004
(0.030)
-0.016
(0.034)
2,986
0.121
0.249
Perc of purchases on account
(full sample of firms)
tobit
-0.122
(0.036)
* *
-0.088
(0.030)
* *
4,318
0.468
0.296
Cash discount offered
ordered probit
-0.109
(0.094)
[0.043]
{-0.009}
Cash discount used
ordered probit
(sample of firms with discounts offered)
-0.351
(0.116)
[0.049]
{-0.132}
* *
* *
-0.071
(0.033)
-0.121
(0.113)
[0.048]
{-0.010}
-0.253
(0.150)
[0.098]
{-0.019}
*
2,986
0.935
0.416
-0.143
(0.161)
[0.017]
{-0.052}
-0.534
(0.151)
[0.085]
{-0.205}
* *
2,126
0.295
0.337
Dependent variable
Estimation
Method
Weighted
Hispanic
Black
Cash discount used
(full sample of trade credit firms)
ordered probit
-0.347
(0.095)
[0.133]
{-0.126}
* *
-0.090
(0.125)
[0.033]
{-0.034}
-0.558
(0.125)
[0.217]
{-0.191}
Payments past due
ordered probit
0.302
(0.094)
[-0.110]
{0.001}
* *
0.040
(0.123)
[-0.014]
{0.000}
0.183
(0.146)
[-0.065]
{0.001}
-0.445
(0.090)
[0.176]
{-0.144}
* *
-0.128
(0.122)
[0.050]
{-0.045}
-0.502
(0.127)
[0.198]
{-0.159}
0.490
(0.082)
[-0.193]
{0.021}
* *
0.135
(0.102)
[-0.053]
{0.004}
0.051
(0.139)
[-0.020]
{0.001}
Without firm or owner deliquency controls 2
Cash discount used
ordered probit
(full sample of trade credit firms)
Payments past due
ordered probit
Sample
Size
Asian
* *
* *
Black vs
Hispanic
Black vs
Asian
2,986
0.102
0.179
2,986
0.091
0.493
2,986
0.037
0.714
2,986
0.007
0.007
Notes:
**(*)= significant at the 5 (10) percent level. All marginal effects are relative to a white small business. A fifth (unreported) racial indicator includes the 34
owners who are Native American or mixed race. The ordered probit models report coefficient estimates in the first row, marginal effects at the never response
in [ ] brackets, and marginal effects at the always response in {} brackets. Standard errors are Huber-White except the tobit regression. The Delta method is
used to compute standard errors for the tobit. FIRE firms are excluded. Controls include the gender and education of owner, two digit industry, region, whether
in an MSA, log assets, number of employees, firm age and firm age square, whether the firm was acquired or inherited, whether the firm is publicly traded, the
profit to asset ratio, sales growth positive, sales growth negative, whether incorporated, whether firm is a franchise, whether the owner was deliquent on
personal obligations 1-2 or 3 times in the last three years, whether the owner declared bankruptcy in the last 7 years, whether the firm was deliquent on
obligations 1-2 or 3 times in last 3 years, dummies for whether the sales growth and Herfindahl variables are missing, longest relationship with a bank, the
number of financial relationships, the complexity of those relationships, and the degree to which they are with the same instititions . All regressions are
weighted using NSSBF sample weights.
2
Excludes whether the firm or owner was deliquent on payments in the last 3 years, whether the owner declared bankruptcy in last 7 years, and whether
judgements have been levied against the owner in the last 3 years.
1
Table 3
Characteristics of Owners and Businesses in the Neighborhood Survey 1
All
Hispanic
Black
White
Asian
Other
5.6
6.9
2.3
*
21.6
*
2.5
66.7
18.4
11.4
8.5
28.5
27.7
8.5
4.5
31.2
29.6
70.2
24.2
14.0
10.9
21.1
22.9
6.9
3.9
31.1
49.4
51.1
13.0
8.4
7.6
22.1
46.6
11.6
5.3
40.5
16.0
**
**
*
**
**
49.3
20.2
4.7
0.0
24.4
29.1
14.9
11.1
16.1
9.8
**
95.0
5.6
2.8
2.5
84.1
2.5
3.7
2.3
18.1
3.6
43.4
27.0
84.9
41.6
9.0
70.5
46.6
37.4
100.0
**
**
28.0
62.1
100.0
**
**
49.0
47.4
88.8
Number of businesses
361
171
116
21
Supplier offers credit
Supplier of same ethnicity
Years with supplier
Supplier in neighborhood
Supplier elsewhere in MSA
Supplier outside of MSA
49.7
35.1
6.5
27.1
51.9
21
44.4
32.9
5.6
38.5
46.3
15.2
42.4
30.8
7.6
19.9
54.7
25.4
60.8
59.9
12.6
20.4
65.0
14.7
Number of suppliers
838
403
246
Manuf/wholesale
constr/transport
Retail total
Eating/drinking places
Food stores
Auto service/sales
Other retail
Business/personal serv.
Age of business in years
Number of employees
Female
No high school degree
High school degree or
some college
College degree or beyond
Proficient in English
**
**
**
**
**
**
55
**
*
4.9
**
**
**
95.1
22.8
25.5
11.0
35.7
0.0
5.9
3.1
15.2
14.5
**
**
43.3
42.2
90.1
**
**
**
31
**
**
**
**
**
77.6
56.1
4.3
5.3
64.1
30.7
79
Notes:
1
** (*)=difference from Hispanic firms is statistically significant at the 5 (10) percent level. These
results are weighted to reflect sample stratification. The Other category is made up of owners from
the Middle East, India, or Pakistan.
**
**
*
**
**
**
22
**
**
*
**
**
**
67.0
17.8
5.7
21.1
47.5
31.4
55
**
**
**
**
Table 4
Trade Credit Offered
Marginal Effects (Robust Standard Errors in Parentheses) 1
By supplier
Probit
Owner and supplier same
ethnicity
Hispanic owner and supplier
Wh/As/Other owner and
supplier
Log years with supplier
Log years with supplier
squared
Years with supplier less than 3
Years with supplier greater
than 7
Supplier in neighborhood
Supplier in Chicago MSA
-0.129
(0.093)
0.318
(0.089)
0.076
(0.135)
-0.151
(0.157)
0.034
(0.043)
Probit
*
*
-0.146
(0.090)
0.340
(0.085)
0.080
(0.135)
0.117
(0.064)
0.161
(0.064)
-0.061
(0.072)
0.094
(0.062)
-0.061
(0.072)
0.092
(0.062)
Probit
*
*
*
*
*
Supplier in
neighborhood*Hispanic
Supplier in Chicago*Hispanic
Supplier in
neighborhood*Wh/As/Other
Supplier in
Chicago*Wh/As/Other
Hispanic
Other
Member of trade association
Someone to talk about
business
Full Controls
-0.026
(0.080)
0.301
(0.079)
0.182
(0.058)
0.099
(0.055)
Yes
*
*
*
*
*
-0.031
(0.078)
0.293
(0.080)
0.178
(0.058)
0.095
(0.055)
yes
Random
effects
probit
*
*
*
*
*
-0.098
(0.098)
0.289
(0.099)
0.033
(0.141)
0.116
(0.064)
0.183
(0.064)
-0.392
(0.102)
0.001
(0.094)
0.557
(0.077)
0.355
(0.112)
0.060
(0.189)
-0.128
(0.156)
-0.378
(0.118)
0.355
(0.121)
0.165
(0.059)
0.104
(0.054)
yes
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
-0.171
(0.127)
0.402
(0.151)
0.165
(0.206)
0.094
(0.101)
0.236
(0.104)
-0.007
(0.117)
0.132
(0.087)
0.028
(0.133)
0.461
(0.152)
0.353
(0.111)
0.137
(0.099)
yes
By firm
probit 2
*
*
*
*
*
*
*
*
-0.147
(0.098)
0.386
(0.080)
0.221
(0.128)
0.164
(0.072)
0.148
(0.083)
-0.079
(0.077)
0.098
(0.072)
-0.034
(0.101)
0.252
(0.095)
0.205
(0.061)
0.038
(0.063)
yes
*
*
*
*
*
*
*
*
*
*
Table 5
Trade Credit Offered, by Race1
Marginal Effects (Robust Standard Errors in Parentheses)
Black owners
Random
Effects
Probit
Probit
Owner and supplier same ethnicity
Years with supplier less than 3
Years with supplier greater than 7
Supplier in neighborhood
Supplier in Chicago MSA
Member of trade association
Someone to talk about business
Sample size
Number of firms
Log likelihood
-0.081
(0.093)
0.150
(0.113)
0.323
(0.133)
-0.266
(0.097)
0.053
(0.092)
0.044
(0.098)
0.263
(0.098)
246
116
-111
*
*
*
*
*
*
-0.030
(0.075)
0.216
(0.162)
0.331
(0.206)
-0.158
(0.079)
-0.085
(0.091)
0.613
(0.233)
0.471
(0.224)
246
116
-97
Hispanic owners
Random
effects
Probit
probit
*
*
*
*
*
*
0.175
(0.066)
0.034
(0.085)
0.171
(0.099)
0.262
(0.109)
0.292
(0.099)
0.255
(0.084)
0.126
(0.077)
*
*
*
*
*
*
*
*
*
403
171
-220
Notes:
1
* (**) = significant at the 10 (5) percent level. All regressions include the variables listed in Appendix 1
and are weighted using sampling weights. Standard errors are Huber-White and are corrected for multiple
firm observations.
0.209
(0.091)
-0.008
(0.114)
0.285
(0.150)
0.353
(0.178)
0.382
(0.146)
0.393
(0.138)
0.121
(0.122)
403
171
-207
*
*
*
*
*
*
*
*
*
Table 6
Accounts receivable, by race
Manufacturing, wholesale trade, and transportation sectors only
White
Black
1
Hispanic
Asian
0.712
(0.017)
0.578
(0.050)
*
0.662
(0.066)
0.582
(0.062)
Accounts receivable
148,592
(9,011)
55,687
(11,879)
*
170,057
(63,052)
124,803
(44,186)
Accounts receivable
(conditional on >0)
208,558
(12,685)
96,372
(19,704)
*
256,792
(92,662)
214,424
(74,486)
0.172
(0.007)
0.187
(0.026)
0.213
(0.034)
0.141
(0.021)
Percent of firms with any
accounts receivable
Accounts receivable / assets
Notes:
1
* = signficantly different from White group at 5 percent level. The sample is weighted by the
NSSBF sampling weights. Standard errors are in parentheses. The sample includes all
manufacturing, wholesale trade, and transportation sector industries except taxi cab drivers.
*
Table 7
Effect of MSA-level Racial Firm Composition on Trade Credit Usage
Black
Differential
No
increased 1 effect
Hispanic
Differential
Differential
No
reduced1
increased 1 effect
Differential
reduced1
Asian
Differential
No
increased 1 effect
Dependent Variable
Differential
reduced1
Trade credit last year
0
0
6
0
6
0
0
0
6
Ever rejected for trade credit
0
3
3
3
0
3
0
0
6
Log (no. of suppliers on acct. +1) 2
0
1
5
0
0
6
0
0
6
Pct. of purchases on account2
0
3
3
0
0
6
0
0
6
Cash discount offered 2
1
0
5
6
0
0
0
0
6
Cash discount used 2
4
0
2
5
0
1
0
0
6
Payments past due 2
0
0
6
0
0
6
0
0
6
Total, each regression=1
Total, each trade credit measure=1 3
5
1
7
1
30
5
14
2.5
6
1
22
3.5
0
0
0
0
42
7
Notes:
1
“Differential reduced” (“Differential increased”) means the interaction coefficient was statistically significant at the 10 percent level and had a sign indicating
that an increase in the MSA-level own-ethnicity business presence was associated with a reduction (increase) in the difference between white-owned firms and
firms of the relevant ethnic ownership.
2
Sample restricted to trade credit users only.
3
Each dependent variable counts once. If the sum for a row is 4 or more, the trade credit measure counts as one. If the sum is 3, then the trade credit measure
counts as 0.5. Thus, for each column group, the total sum is 7.
Appendix 1
Decomposition of Ethnic Gap Among NSSBF Trade Credit Measures
Percent purchases on account
Trade credit users
Full sample
Black
Hispanic
Asian
Black
Hispanic
12.8
5.9
5.4
11.0
11.0
Racial gap
Fraction due to differences in
Characteristics
54.1
Coefficients
45.9
100.3
-0.3
Contribution to racial gap from differences in:
Female
0.4
2.1
Education
-0.2
2.7
Owner bankruptcy
-0.1
-0.0
Owner deliquent on
11.0
14.3
Debt
Judgements against
8.4
6.9
Owner
Firm deliquent on
2.2
3.2
Debt
Firm age
1.7
4.4
Firm acquired
4.0
-2.0
Firm inherited
1.0
1.0
Firm is traded
-0.1
0.1
Employees
0.6
1.1
Assets
7.7
11.1
Incorporated
1.6
5.7
Franchise
0.1
0.3
Profit/assets
0.3
-0.4
Sales growth
-0.2
-0.3
Region
3.6
7.6
MSA
1.2
2.7
Industry
2.9
33.5
Longest bank
1.9
2.8
Relationship
Network size
-0.2
1.1
Herfindahl of ties
6.0
0.7
Complexity of ties
0.3
1.7
Asian
8.9
Cash discount offered
Black
Hispanic
Asian
0.24
0.26
0.43
Payments
past due2
Cash discount used
Black
Hispanic
Asian
Black
0.80
0.40
0.85
-0.40
67.4
32.6
52.6
47.4
50.4
49.6
54.9
45.1
70.6
29.4
48.2
51.8
58.3
41.7
40.4
59.6
40.7
59.3
14.9
85.1
49.0
51.0
1.4
-5.6
0.0
-0.1
0.0
-4.0
-0.2
3.2
0.0
4.4
-0.1
2.1
0.0
-8.6
0.1
-1.6
1.7
-2.5
-3.1
12.1
4.0
3.1
-0.6
7.7
1.5
-8.0
0.0
-2.0
-0.2
2.0
-0.3
6.6
1.1
0.9
-0.4
6.1
-0.5
-7.4
-0.0
-0.2
0.3
0.6
0.6
22.2
1.7
8.2
4.1
0.7
-2.5
-0.9
-0.1
-3.0
-0.9
-1.7
1.7
-0.9
-8.2
-3.8
5.5
-2.3
-0.4
1.0
25.1
31.0
-5.2
30.6
9.8
-6.5
1.8
-0.1
0.4
-6.3
0.2
-1.0
0.2
3.2
16.3
4.4
42.9
7.8
2.1
1.6
0.1
-0.4
0.1
15.1
4.0
-0.4
0.5
0.5
2.8
4.2
9.2
0.2
2.3
0.2
0.1
-0.1
0.0
6.2
3.9
-0.8
-0.4
-0.5
0.5
3.8
13.3
0.2
5.0
-2.1
0.1
2.5
0.0
-2.7
1.1
0.8
-0.1
4.1
-4.7
7.2
43.5
0.3
3.8
7.4
-0.3
-0.1
-0.9
4.3
-6.3
-0.3
-1.3
0.2
13.5
9.0
39.4
6.3
2.3
-1.5
-0.1
0.1
-0.7
2.6
-9.6
-0.4
0.7
0.6
8.7
8.3
32.0
3.9
7.4
-2.8
-0.1
-0.0
-0.1
-0.8
-0.2
0.9
-0.3
-6.6
3.8
7.5
51.4
6.2
4.3
1.0
0.5
0.1
0.0
1.2
-0.7
0.0
0.0
-0.2
7.4
1.5
-6.6
1.0
6.8
-0.5
0.9
0.2
0.1
0.3
-1.6
-0.0
-0.0
-1.3
-1.9
3.6
-10.4
0.7
11.7
-0.5
0.2
0.1
0.0
-0.2
0.0
-0.1
0.0
0.0
-1.3
3.7
15.9
1.7
0.4
0.7
-1.3
-0.1
0.4
-1.3
-3.4
0.1
-0.3
0.2
-2.7
1.3
-5.9
0.7
0.3
-2.6
0.1
0.2
8.8
5.0
2.2
3.8
8.9
2.1
-1.8
3.3
-4.8
2.8
-5.7
9.1
-5.7
-14.9
1.5
-1.1
-0.6
1.2
-0.9
0.3
-1.7
0.3
7.7
-0.5
-1.2
0.3
1.7
2.8
-0.2
Notes:
1
Independent variable's contribution to racial gap is computed using the white sample as the base case. All regressions are estimated with linear probability models.
2
There is no raw Hispanic-White or Asian-White gap.
Appendix 2
Selected Full Regression Results 1
Marginal Effects (Robust Standard Errors in Parentheses)
Full sample, by supplier
Random
effects
probit
probit
Owner and supplier same ethnicity
Hispanic owner and supplier
Wh/As/Other owner and supplier
Years with supplier less than 3
Years with supplier greater than 7
Supplier in neighborhood
Supplier in Chicago MSA
Hispanic
Other
Member of trade association
Someone to talk about business
Less than HS degree
College degree
Proficient in English
Female
-0.146
(0.090)
0.340
(0.085)
0.080
(0.135)
0.117
(0.064)
0.161
(0.064)
-0.061
(0.072)
0.094
(0.062)
-0.031
(0.078)
0.293
(0.080)
0.178
(0.058)
0.095
(0.055)
-0.032
(0.064)
-0.189
(0.063)
0.146
(0.072)
-0.189
**
*
**
**
**
*
**
**
**
-0.171
(0.127)
0.402
(0.151)
0.165
(0.206)
0.094
(0.101)
0.236
(0.104)
-0.007
(0.117)
0.132
(0.087)
0.028
(0.133)
0.461
(0.152)
0.353
(0.111)
0.137
(0.099)
-0.092
(0.116)
-0.273
(0.114)
0.262
(0.134)
-0.320
Black
owners
probit
-0.081
0.093
Hispanic
owners
probit
0.175 * *
(0.066)
**
**
0.150
(0.113)
0.323 * *
(0.133)
-0.266 * *
(0.097)
0.053
(0.092)
0.034
(0.085)
0.171 *
(0.099)
0.262 * *
(0.109)
0.292 * *
(0.099)
0.044
(0.098)
0.263 * *
(0.098)
-0.117
(0.137)
-0.239 * *
(0.099)
0.255 * *
(0.084)
0.126
(0.077)
-0.046
(0.083)
-0.132
(0.123)
0.148 *
(0.081)
-0.095
**
**
**
*
**
-0.333 * *
Log number of employees
Log number of employees squared
Business in danger of failing
Business faces growth challenge
Has deposit relationship
Business age less than 3
Business age greater than 7
Uses accountant
Gives consumers credit
Previously owned another business
Disposition towards risk
(0.055)
(0.103)
-0.039
-0.005
(0.094)
(0.026)
0.045
0.028
(0.030)
(0.020)
0.174 * *
0.292
(0.058)
(0.105)
0.119
0.234
(0.084)
(0.158)
0.165 * *
0.251
(0.076)
(0.122)
-0.036
0.042
(0.081)
-(0.145)
-0.121 *
-0.201
(0.068)
(0.119)
0.005
0.050
(0.069)
(0.128)
0.097 *
0.192
(0.058)
(0.111)
Full sample, by supplier
Random
effects
probit
probit
-0.007
(0.055)
-0.041 * *
(0.015)
**
**
*
*
-0.114
(0.099)
-0.051 *
(0.028)
(0.099)
-0.024
(0.137)
0.023
(0.043)
0.126
(0.116)
0.106
(0.124)
0.294 * *
(0.102)
-0.422 * *
(0.084)
-0.192
(0.124)
0.138
(0.111)
0.022
(0.123)
Black
owners
probit
-0.108
(0.112)
-0.045
(0.030)
(0.075)
0.083
(0.147)
0.028
(0.050)
0.223 * *
(0.082)
0.069
(0.162)
0.108
(0.094)
0.253 * *
(0.105)
-0.125
(0.102)
-0.081
(0.104)
0.130
(0.086)
Hispanic
owners
probit
0.042
(0.078)
-0.043 * *
(0.021)
Notes:
1
* (**) = significant at the 10 (5) percent level. All regressions include industry dummies and are
weighted using sampling weights. Standard errors are Huber-White and are corrected for multiple firm
observations.
@FedFRASER