Interest Rate and Collateral Differences in Loans — William D. Bradford
- Issuer
- Congressional materials
- Document type
- 2025 01 29 A32584 D276687 Uw Foster School Of Business Interest Rate And Collateral Di
- Date
- 2025-01-29
- Case
- 2025 01 29 A32584 D276687 Uw Foster School Of Business Interest Rate And Collateral Di
Summary
A research report from the Foster School of Business, University of Washington, by William D. Bradford, Ph.D., Emeritus Dean, on interest rate and collateral differences in loans to businesses owned by people of color and women. It analyzes survey data from 2,784 firms in 44 states on loans made between January 2022 and June 2023, across six lender types. The report states that, after controlling for risk factors, Asian American-owned firms paid 2.88, Hispanic-owned firms 2.91 and Black-owned firms 3.09 percentage points higher rates than White-owned firms, and woman-owned firms 2.38 percentage points more than man-owned firms. It estimates $8.0 billion more in annual interest paid by those minority-owned groups and finds outside co-signers were required more often for minority-owned firms. An appendix presents principal component analysis in Exhibit A-1 and Exhibit A-2.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Cite this document
“2025 01 29 A32584 D276687 Uw Foster School Of Business Interest Rate And Collateral Di.”
Original at app.leg.wa.gov, dated 2025-01-29. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/2025-01-29-2025-01-29-a32584-d276687-uw-foster-school-of-business-interest-rate-and-collateral-di/
, SHA-256 b89cf99facb4026b86665fe8d29600986ec048a02b0d8fde94bb60ea3c5199a5.Full text
INTEREST RATE
AND COLLATERAL
DIFFERENCES
IN LOANS TO BUSINESSES OWNED BY PEOPLE OF COLOR
AND WOMEN: OPPORTUNITIES TO IMPROVE PROFITABILITY
FOR BUSINESS BORROWERS AND LENDERS
BY WILLIAM D. BRADFORD, Ph.D. Emeritus Dean
INTEREST RATE
AND COLLATERAL
DIFFERENCES
IN LOANS TO BUSINESSES OWNED BY PEOPLE OF COLOR
AND WOMEN: OPPORTUNITIES TO IMPROVE PROFITABILITY
FOR BUSINESS BORROWERS AND LENDERS
By William D. Bradford, Ph.D. Emeritus Dean, We thank David Hinson for being the catalyst for
University of Washington Foster School of this study. We also thank the Social & Economic
Business and Project Director Sciences Research Center (SESRC) at Washington
State University for their management of the
Michael Verchot, Director, Consulting and survey. Ronald Howell, Edward Rice, Stephan
Business Development Center, Foster School Siegel, and Quentin Dupont made helpful
of Business, University of Washington comments on early versions of this report.
Rachel Atkins and Esther Uduehi contributed
Statistical Contributors: Magnus Lofstrom, to the development of the survey.
Ph.D. and Chunbei Wang, Ph.D.
The forementioned organizations and people
do not necessarily agree with nor are they
responsible for the contents and conclusions
of this report.
The Schultz Family Foundation provided
generous financial support for this project.
2
TABLE OF
CONTENTS
EXECUTIVE SUMMARY 4
I. INTRODUCTION 6
II. BACKGROUND 10
FOCUS OF THE RESEARCH 10
SURVEY SUMMARY STATISTIC 11
III. MISPRICING ANALYSIS 18
APPROACH 18
INTEREST RATE RESULTS 21
INTEREST RESULTS BY RACIAL/ETHNIC & GENDER GROUPS 24
COLLATERAL 27
PRINCIPAL COMPONENT ANALYSIS 29
IV. CONCLUSION 30
FINDINGS 30
COST TO AFFECTED BUSINESSES 30
REFERENCES 33
APPENDIX 35
3
EXECUTIVE
SUMMARY
The U.S. small business sector is important in Federal Reserve’s Survey of Small Business
job creation and other elements of economic Finances (SSBF) in 2003. This study provides
growth. This study examines the nature of the a picture of small business financing in the
credit provided to support the growth of these early post-COVID period. We used the
firms. We examine differences in interest rates survey data to compare the credit terms
and collateral requirements for businesses of minority-owned firms to those of
owned by people of color compared to White- White-owned firms, and woman-owned
owned businesses and differences between firms to man-owned firms.
businesses owned by men and those owned
by women. In comparing the terms of the loans, we
considered a range of factors that might
We have three goals for affect loan terms, such as the size of the
conducting this study: business, credit rating, owner experience,
industry, revenues, and more. These factors
Provide business borrowers with are important because they determine the
information about where they would risk for lenders when making a loan. Our
likely receive the best loan terms. statistical approach determined whether,
after considering that range of factors,
Provide lenders with information that minority ownership or gender-based
would enable them to more successfully ownership affected the terms of the loan.
compete for market share by providing Minority-owned or woman-owned status
more competitive loan terms. should not influence interest rates or collat-
eral requirements if the financial risk factors
Provide regulators with information have already been considered.
on the current credit environment for
minority- and women-owned businesses. We found that, after controlling
for firm risk factors:
Between January 2022 and June 2023, we
collected data from 44 states on loans made Our basic models indicate that after
to privately-owned businesses with 500 or controlling for firm risk factors:
fewer employees. This data set, collected Compared to White-owned firms,
through surveys, contained interest rate and Asian American-owned firms paid
collateral differences in loans to small busi- 2.88 percentage points higher rates,
nesses along racial, ethnic, and gender lines Hispanic-owned firms paid 2.91
in the US; it was the first such survey since the percentage points higher rates, and
4
Black-owned firms paid 3.09 percentage Black-owned firms paid higher rates than
points higher rates. their White-owned counterparts at credit
unions, fintech lenders and nonbank
After adjusting our statistical analyses finance companies. We did not find that
to reflect the national distribution of race was statistically significant in bank
firms, we estimate that Asian-American-, and CDFI loans: the Black firm-White firm
Black-, and Hispanic-owned businesses interest rate differences for these lenders
collectively pay, on average, $8.0 billion were explained by factors such as credit
more in annual interest than comparable score and size of firm, not race.
White-owned firms.
Hispanic-owned firms paid higher interest
Women-owned firms paid 2.38 percentage rates than their White-owned counter-
points higher rates than man-owned parts at each of the six lender types.
firms.
Asian-owned firms paid higher interest
Lenders were more likely to require an rates than their White-owned counter-
outside co-signer on loans to Asian- parts at large banks. There were too few
American-, Black-, Hispanic-, and Native cases in the sample to make inferences
American-owned businesses than their about the other lenders to Asian-owned
White-owned counterparts. The survey firms.
data was not sufficient to estimate the
economic impact. This is an important Woman-owned firms paid higher interest
topic for future research. rates than their man-owned counterparts
at credit unions, CDFIs, and fintech lend-
ers. We did not find that gender affected
We also examined the credit terms the interest rate in bank and nonbank
provided to these small firms across finance company loans; for these lenders,
six types of lenders: the interest rate differences between
woman-owned firms and man-owned
Large banks (deposits of at least $10 billion), firms were explained by factors such as
small banks, community development finan- credit score and size of firm, not gender.
cial institutions (CDFIs), credit unions, fintech
lenders, and nonbank finance companies. Disparities in borrowing experiences can have
While we earlier summarized the overall far-reaching implications beyond the econom-
interest rate differential relative to White- ic sphere, contributing to broader inequalities
owned firms for each minority group, the and perpetuating historical patterns of dis-
minority-owned groups’ experiences differed crimination and exclusion. If such disparities
across lenders. Consistent with the approach exist, addressing these disparities will require
that controls for a broad range of factors that concerted efforts from policymakers, lenders,
affect loan risk, we found that: and other stakeholders to improve access to
credit and reduce discrimination in lending.
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Executive Summary 5
I. INTRODUCTION
Borrowing difficulties can be one of the most period. We compared the terms of minority
important obstacles in operating and growing firms to White firms, and woman owned firms
a business. The purpose of this study is to to man owned firms.
examine mispricing in small business loans
regarding interest rate and collateral terms. The information gathered in the survey
Mispricing occurs when lenders adjust the includes loan terms (interest rate and collat-
terms for borrowers based on factors unre- eral required), firm financial information, CEO
lated to their financial risk, such as the race attributes and lender type (bank, credit union,
or gender of the business owner, rather than CDFI, fintech lender, and nonbank finance
basing decisions solely on economic indicators company).Data collected comprises over sixty
and creditworthiness. This study explores characteristics of the firms and their owners
whether mispricing is evident across racial including ethnicity/race of controlling owner-
and gender groups on loans made to small ship: White, Asian, Hispanic, Black, and Native
firms. (i.e., Native American and Alaskan Native), as
well as gender of the majority ownership of
The issue of discrimination in small business the firm.
lending has long been a concern in the United
States. This is the first data set containing Regarding the results of our analyses, we find
interest rate and collateral differences in loans the interest rates paid by Asian-, Black-, and
to small businesses along ethnic and gender Hispanic-owned firms and woman-owned
lines in the US since the Federal Reserve’s firms were higher than economically justified
Survey of Small Business Finances (SSBF) after considering creditworthiness indicators
in 2003. Research on the 2003 SSBF data and other financial attributes of the firms. In
indicated racial disparities: loan rejection rates contrast to comparable White-owned firms,
and interest rates were inexplicably higher Asian-owned firms paid 2.88 percentage
for some groups of minority borrowers (for points higher interest, Hispanic-owned firms
example, see Asiedo et al., 2012). Our study paid 2.91 percentage points higher interest,
collected and analyzed data on small business and Black-owned firms paid 3.09 percentage
loans made between January 2022 and June points higher interest. Below we provide
2023. This study provides a picture of small estimates of the dollar amount of mispricing
business financing in the early post-COVID across the minority firms.
6
Woman-owned firms paid that the firm’s minority group indicator either
has or does not have a statistically significant
2.38 percentage points impact on its interest rate. These models
higher interest than also include relevant economic and
creditworthiness indicators that lenders
man-owned firms. consider in making loans.
Woman-owned firms paid 2.38 percentage Considering large-bank lending to Hispan-
points higher interest than man-owned firms. ic-owned firms vs. White-owned firms, we find
We also find that lenders were more likely that Hispanic-owned firms paid higher interest
to require an outside co-signer on loans to rates to large banks than we can explain in our
Asian-American-, Black-, Hispanic-, and Native models. Also, the rate Hispanic-owned firms
American-owned businesses than to their paid to large banks does not differ statistically
White counterparts. However, the survey data from the rates Hispanic-owned firms paid
was not sufficient to estimate the associated to small banks, credit unions and nonbank
economic impact. This is an important topic finance companies; at the same time, the
for future research. rates Hispanic-owned firms paid to CDFIs
and fintech lenders are lower than those
This study also examines lending practices paid to large banks.
across six types of lenders: large banks (with
deposits greater than $10 billion), small banks, In the Black-owned firm vs. White-owned firm
community development financial institutions comparison, the rates paid by Black-owned
(CDFIs), credit unions, fintech lenders, and firms to large banks are explained within our
nonbank finance companies. These lenders models, and rates paid to large banks do not
differ in regulatory structure, sources of differ statistically from the rates Black-owned
funds, and lending specialization. By analyzing firms paid to small banks and CDFIs. However,
the differences in lending practices across the rates paid by Black-owned firms to credit
these lenders, the study identifies variations unions, fintech lenders and nonbank finance
in lending practices that may contribute to companies are higher than those our models
or soften disparities in lending outcomes for can explain.
diverse firms.
In the Asian-owned firm vs. White-owned firm
While we earlier summarized the overall comparison, we find that Asian-owned firms
interest rates relative to White-owned firms pay higher interest rates to large banks than
for each of the groups, the minority-owned we can explain in our models. Also, the rates
firms’ experiences differed across the six paid to large banks do not differ statistically
types of lenders. Large commercial banks are from the rates Asian-owned firms paid to each
the dominant business lenders in the U.S., so of the other types of lenders.
we conduct minority group vs. White-owned
firm comparisons of interest rates to gauge
differences (mispricing) between large banks
and other lenders. In these exercises, mis-
pricing is or is not evident when our models,
which predict interest rates paid by firms, find
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Introduction 7
We also compared woman-owned firms exist, addressing these disparities will require
to man-owned firms. Here, we find that concerted efforts from policymakers, lenders,
relative to man-owned firms, the rates paid and other stakeholders to improve access to
by woman-owned firms to large banks, small credit and reduce discrimination in lending.
banks, and nonbank finance companies are
explained by our models, while the rates paid The data presented in this study originated
to credit unions, CDFIs and fintech firms are from a national survey encompassing firms
higher than those predicted by our models. from 44 states across the U.S. These firms
come from a large data set of either existing
Finally, we compared various forms of or prospective suppliers to the U.S. govern-
collateral requirements on credit provided ment. While the more populated states in our
to the firms. sample tend to have more firms, these 2,784
firms do not fully mirror the U.S. distribution
of minority- and White-owned firms by num-
We find that co-signatures ber and industry. Accordingly, our analysis
was twofold. First, we examined the original,
from third parties are unaltered firm data to maintain the statistical
required more frequently authenticity of the responses. Second, for
the sake of methodological robustness, we
for each minority group included two separately weighted analyses
of the relationships examined in the main
compared to White firms...
analysis (additional robustness tests ex-
amined other aspects of the data). In both
We find that co-signatures from third parties weighting analyses, weights were applied
are required more frequently for each that mirrored the national distribution of
minority group compared to White firms White-owned and minority-owned firms
than is justified by our economic analyses. across the various states and industries.
However, we do not find that collateral
requirements are greater for woman-owned It is important to acknowledge, however,
firms compared to man-owned firms. Our that the weighting process was constrained
survey also asked the firm owners’ opinions by the limitations of available data. The U.S.
about their borrowing experience. Compared Census data, which is crucial for weighting,
to White and man owners, minority and is regrettably incomplete in its state and
woman owners 1) were less satisfied with industry-specific information, particularly
their loan terms, 2) more often felt that their for minority groups. Despite these con-
loan terms were mispriced, and 3) more often straints, a key finding of our study was
felt that compared to similar firms, they had the consistency in outcomes between the
more difficulty in borrowing. weighted and unweighted analyses. This
consistency suggests that our core findings
Disparities in lending practices can have are resilient and remain unaffected by the
far-reaching implications beyond the econom- chosen methodology of weighting. In this
ic sphere, contributing to broader inequalities report we present the basic (unweighted)
and perpetuating historical patterns of dis- analyses in detail and discuss the equivalent
crimination and exclusion. If such disparities results of the weighted analyses.
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Introduction 8
The remainder of this report is organized as reports the results on interest rate and
follows. Part II: Background, presents the collateral mispricing. Part IV: Conclusion,
focus of this study in terms of the credit issues provides an overview of the results and
examined, and provides summary statistics estimates of the economic cost of the interest
on the firms that are analyzed in the study. rate differentials found between minority-
Part III: Mispricing Analyses, introduces the and White-owned small business borrowers.
statistical approach used in the study and
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Introduction 9
II. BACKGROUND
FOCUS OF THE RESEARCH
Numerous studies have examined the Blanchard et al., 2008; Blanchflower et al.,
differences in financing patterns between 2003; Bostic & Lamani, 1999; Cavalluzzo et al.,
minority-owned firms and White-owned 2002; Cavalluzzo & Cavalluzzo, 1998).
firms. No set of data has been available to
examine mispricing in the terms of loans More recent research has had to rely on
granted to small firms in the U.S. since 2003, data sources that include less information
which is the final year of the Federal Reserve’s on business loans than SSBF. The first, the
series of cross-sectional surveys: Survey of Kauffman Firm Survey (KFS), contains data on
Small Business Finances (SSBF). The Federal the financing of new firms from 2004 through
Reserve’s SSBF covered the years 1989, 1993, 2011. The second, the Federal Reserve, has
1998 and 2003. completed annual Small Business Credit
Surveys (SBCS) for each year 2019–2022. Bates
Researchers have used SSBF to explore: and Robb (2015, 2016) and Fairlie et al. (2022),
a) the number of businesses discouraged using the KFS results from 2008–2011, found
from applying for a loan due to fear of that minority business owners were more
rejection; and b) the loan denial rates for likely to be discouraged from borrowing and
businesses applying for credit, comparing more likely to be rejected for credit. Barkley
minority-owned to White-owned firms and and Schweitzer (2023) had similar findings us-
woman-owned firms to man-owned firms. ing the SBCS. None of these data sets include
Overall, these studies concluded that: a) information on the terms of loans granted to
compared to White business owners, a higher minority- and woman- owned businesses.
fraction of bankable minority business owners
do not apply for loans because of fear of Since this study concerns loans granted to
rejection; and b) for those minority firms, it does not consider firms that needed
businesses that do apply, higher fractions credit but did not apply, nor firms that applied
of bankable minority-owned businesses are but were rejected for financing. A firm desiring
rejected for financing than comparable White- debt financing is taking the first step in the
owned businesses (Asiedu et al., 2012; sequence that ends in obtaining a loan.
10
However, not all firms that want debt financing suppliers with connections to firms and
take the second step and apply for a loan. organizations seeking to hire diverse
The third step is the lender accepting the suppliers, while concurrently providing
loan request and making the loan. However, these firms access to its list of diverse
the loan granted may differ from the loan suppliers. It also assists firms in developing
requested. Thus, step four concerns whether and growing their own programs for contract-
the loan was fully or partially accepted. We ing with minority suppliers.
collected data on both the loan requested
and the final loan committed. We explored Supplier.IO provided us with a list of over
the loan terms conditional on the firm having 300,000 firms, to which we sent an internet
received a loan with a focus on interest rates survey. The survey collected firm/owner/
and collateral. This covers the outcomes in lender information on non-government-
yellow below. guaranteed borrowing by small firms (< 500
employees) from January 2022 through May
2023. Duns credit score information was
added as a variable for a subset of firms that
completed the survey. To be eligible for the
survey, a business must have been operating
as a for-profit firm with employees at the
time of the survey. Also, during the specified
period, it must have borrowed through a line
of credit or loan that was not government
guaranteed. Following previous studies, firms
in the finance industry were ineligible.
The summary statistics of the responding
firms will be presented as follows: interest
rate on financing received, other loan terms
(including collateral), firm information, lender
information, loan terms, and other survey
information. The key outcomes of the loan
process that we examine are the rate paid in
borrowing and the collateral requirements
associated with borrowing. We winsorized
the reported interest rates at the bottom 1%
and top 2% to reduce the distortive effect
of outliers. Exhibit I shows that the average
interest rate is higher for each of the minority
groups than for the White-owned firms. Also,
the woman-owned firms pay a higher average
S U RV E Y S U M M A RY S TAT I S T I C S rate of interest than the man-owned firms.
This section presents summary statistics on
GR APH ON NE X T PAGE
the responses of the surveyed firms. Supplier.
IO is a private firm that provides diverse
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Background 11
S U M M A R Y S TAT I S T I C S O N
EXHIBIT I
L O A N R AT E B Y R A C E A N D G E N D E R
In Part III, we determine to what extent these It shows that White-owned firms committed
rates embody mispricing of minority- and the least collateral relative to the loan amount,
woman-owned firms, i.e., the extent to which were least often required to commit more
the rates for minority- and women-owned collateral than the amount of the loan, and
firms are higher than they should be after risk were least often required by the lender to
factors are considered. The comparisons at have an external party sign for the loan.
the top of Exhibit II show the loan rate spread, There are multiple reasons why these differ-
which is the interest rate paid minus the ences in interest rates and collateral might
prime rate at the time the loan was granted. exist, including the industry, geographical
The results of the loan spread mirror those location, and prior years’ profits that are
of the interest rate. Exhibit II also contains un-related to the race, ethnicity, and gender
various measures of collateral associated with of the owner. Our analyses control for
the loans, as reported by the firms. creditworthiness and other firm attributes
to see if minority-owned and White-owned
firms of like attributes were treated the same
...White-owned firms in defining the interest rate and collateral
requirements.
committed the least
collateral relative to
GR APH ON NE X T PAGE
the loan amount . . .
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Background 12
EXHIBIT II OTH E R L OA N O U TCO M E VA R I A B L E S
Exhibit III shows that firm characteristics highest credit scores. On the other hand,
vary across racial, ethnic and gender lines. White-owned firms report the lowest fraction
White-owned firms report the lowest fraction of firms in good or better condition at yearend
of firms with losses in 2022. White-owned 2022. Black- and Hispanic-owned firms have
firms also report the highest fraction of firms the lowest credit score profile. The other traits
with the same or higher revenue in 2022 vs. are mixed, and no group stands out as better
2021 and the highest fraction of firms with or worse overall in other firm measures
the same or more employees in 2022 vs. 2021.
GR APH ON NE X T PAGE
White-owned firms also report the
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Background 13
EXHIBIT III FIRM CHAR ACTERISTICS
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Background 14
Regarding gender analysis, woman-owned banks, CDFIs, credit unions, fintech lenders,
firms exceeded man-owned firms in business and nonbank finance companies. The lenders
age, owner’s experience, and owner age. and rates charged to each group of borrowers
are shown in Exhibits IV-A and IV-B.
The lender utilization figures in Exhibit IV-A
Nonbank finance companies show that, overall, large banks were used
more often than the other lender groups.
charged the lowest rate, However, only 13% of the White-owned firms
followed by fintech lenders; used large banks, while at least 60% of each of
the other four racial/ethnic groups borrowed
CDFIs and large banks from large banks. About nine out of ten Asian-
owned firms and Native-owned firms in our
charged about equal rates,
sample borrowed from large banks.
with small banks charging
Meanwhile, a higher fraction of White-owned
higher rates. Credit unions firms borrowed from small banks, credit
charged the highest average unions, CDFIs, and fintech lenders and non-
bank finance companies than did any of the
rates among the six lenders minority groups. In the case of gender, com-
pared to man-owned firms, higher fractions
in our sample.
of woman-owned firms borrowed from large
banks and small banks, and lower fractions
A higher fraction of woman-owned firms from the other lenders. Regarding the interest
suffered losses in 2021, and a higher fraction rate paid by the borrowing firms, nonbank
are family businesses. A lower fraction of finance companies charged the lowest rate,
woman-owned firms had a high credit score, followed by fintech lenders; CDFIs and large
and a higher fraction had a low credit scored banks charged about equal rates, with small
compared to the man-owned firms. banks charging higher rates. Credit unions
charged the highest average rates among the
The survey also collected information on the six lenders in our sample.
type of lender that provided the credit: large
GR APH ON NE X T PAGE
banks (deposits of at least $10 billion), small
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Background 15
S U M M A R Y S TAT I S T I C S :
E X HIB IT I V-A LENDER CHAR ACTERISTICS
BY R ACE AND GENDER
Regarding the borrower groups, Exhibit IV-B Hispanic firms paid higher than the average
shows that White-owned firms are the only of all groups. For the other three sources,
group that paid lower than average interest Hispanic-owned firm paid lower than the
rates at each of the lenders. At the other average rate of all groups. After Asian- and
end, Asian-owned firms are the only group Hispanic-owned firms, Black-owned firms
that paid higher than average interest rates paid the highest average interest rate.
at each of the lenders that provided them The difference between the Hispanic- and
credit. Asian-owned firms borrowed almost Black-owned firms’ borrowing rate is small.
exclusively from large banks (92 percent). While Black-owned firms were charged a
midrange average rate by large banks, they
Seventy-six percent of Hispanic-owned firms paid higher than average rates when borrow-
borrowed from large banks, small banks, or ing from the other five lender types.
credit unions. For each of these sources,
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Background 16
S U M M A R Y S TAT I S T I C S : L O A N R AT E
E X HIB IT I V- B
BY R ACE/GENDER AND LENDER T YPE
With regard to the gender analysis, that woman-owned firms borrowed from
woman-owned firms paid higher rates large banks and small banks more often
to each type of lender than did man- than did man-owned firms and borrowed
owned firms, with an overall higher from the other lenders less often than
average borrowing rate than that paid by man-owned firms.
man-owned firms. The lender traits show
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Background 17
III. MISPRICING
ANALYSIS
APPROACH
The previous sections describe the summary of mispricing of loans to minority- and wom-
findings, but they do not control for any risk an-owned firm borrowers.
factors among the borrowers. This “Mispricing
Analysis” section reports the results of statis- The following basic model provides the
tical tests to determine evidence of mispricing framework for the multiple regressions that
in borrowing terms for businesses owned by follow. R is the interest rate to be determined
people of color and women, after controlling by a basic rate β0 , and there are N control
for financial attributes that are used to price variables that determine the shifts in R, based
credit. Ordinary least squares regressions are on the number of control variables. Thus,
the standard approach to examine mispricing there are N control variables considered
of loan terms. Financial theory suggests that that affect the interest rate; the β i (i = 1 … N)
the required return lenders demand from represents the changes in the interest rate
borrowers should be positively related to the reflecting the value of the attributes affecting
loan’s default risk. the rate; β R measures the effect of race on the
interest rate, while β G is the effect of gender
Accordingly, we use the interest rate on the on the interest. Mispricing occurs when the
loan as the measure of return charged by the coefficient of race and gender are statistically
lender. An alternate measure is the interest significant in a model that is considering effect
rate on the loan less the prime rate at the time of attributes measuring loan risk. This predic-
the loan was granted. This is the “loan spread”. tion is based on the information available to
We report this in our analyses, and find the the lender concerning the creditworthiness
results using this measure are very similar to of the applicant and the business activity,
those using the interest rate. In this section, including the owner’s credit and resources,
the standard regression framework was used the firm’s credit and financial health, and the
to determine to what extent there is evidence environment in which the firm and lender
operate.
18
Our first step to gaining insight to potential
factors contributing to differences in loan
pricing across groups is to estimate linear
regressions with loan rates as the dependent
variable. In this way, we limit the sample to
When all the N attributes describing loan risk White-owned firms and each of the minori-
are included in the model, then β R and β G ty-owned firms, separately. For each of these
should be zero. We cannot prove, of course, regressions, we include an indicator variable
that we have controlled for every credit for the relevant minority group, which then
variable that lenders consider in evaluating captures the loan rate difference between
applications for business loans. However, the the two groups. We then examine changes
information gathered in the survey makes it in the estimated minority-White loan pricing
possible to control for a very wide range of difference (i.e., the estimated coefficient on
such variables, thereby greatly lowering the the indicator variable of the minority group) as
probability that our estimates are affected by we sequentially added controls. The controls
omitted variable bias. we used are in Exhibit V.
GR APH ON NE X T PAGE
As a result, we believe that
our results provide credible
estimates of mispricing.
As a result, we believe that our results provide
credible estimates of mispricing. We assess
mispricing by determining whether the inter-
est rate on the credit received differs between
the respective minority-owned firms and
White-owned firms after controlling for the N
variables that effect the interest rate. Included
in the N variables are the other features of the
loan, which include whether the interest rate
is fixed or variable and whether collateral is
required. These are determined at the same
time as the interest rate and may, therefore,
be endogenous. We will consider this as part
of our analysis by studying the importance
of this effect in our data. Exhibit V shows the
control variables used in our analyses.
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Mispricing Analysis 19
EXHIBIT V CO NTR O L VA R I A B L E S
We did the same exercise for man-owned captured the loan rate difference between
firms and woman-owned firms, where we woman- and man-owned firms with an
included owners of all minority groups and indicator for woman-owned.
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Mispricing Analysis 20
I N T E R E S T R AT E R E S U LT S
Interest Results by Minority Group
Exhibit VI shows our analysis of interest rates after controlling for equal creditworthiness
on approved loans for firms in our sample. and other firm attributes, the annual interest
Groups are defined based upon controlling rate paid by Hispanic-owned firms was 2.91
ownership. Racial/ethnic identity of a firm is percentage points higher than that paid by
defined by 50% or more controlling owner- White-owned firms. The asterisks indicate
ship, while all other racial/ethnic groups hold the statistical significance of the finding.
less than 50% combined ownership. Gender We consider a result statistically significant
groups are defined by the gender holding 51% when it is at least significant at the .10 level.
controlling ownership. We consider model 5 as Statistically significant results are also found
the most important since it considers all the for Black-owned firms, Asian-owned firms and
control variables. The interpretation is that, woman-owned firms in comparison to man-
compared to White-owned firms and owned firms.
GR APH ON NE X T PAGE
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Mispricing Analysis 21
REGRESSIONS CONTROLLING FOR
DIFFERENT CHAR ACTERISTICS MINORIT Y
EXHIBIT VI
VS. WHITE FIRMS AND MAN OWNED
VS. WOMAN OWNED FIRMS
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Mispricing Analysis 22
To assess the robustness of our analysis, we gender group and no other single racial/ethnic
examined the results for firms where at least or gender group owns more than 49.0%.
51% of the ownership is by one racial/ethnic or These results appear in Exhibit VII.
ROBUSTNESS CHECK USING R ACIAL/
EXHIBIT VII
ETHNIC FIRM 51% OWNERSHIP
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Mispricing Analysis 23
Using this check, we found that the statistical In summary, our final mispricing differential
significance differences across groups do not for the minority-White interest rate for each
change, while the percentage differentials are minority group is the average of estimates
moderately different. from three models, all of which used the main
data set (at least 50% ownership for that
To confirm our results, we performed two group with no other group holding more than
separately weighted analyses of the relation- 49%) with winsorized interest rate measures:
ships examined in the main analysis. In both the main model, the weighted-by-state model,
weighting analyses, weights are applied to the and the weighted-by-state/industry model.
winsorized data. The weights mirrored the The average of the mispricing differential
national distribution of White-owned and mi- across these three models is taken as the mis-
nority-owned firms across the various states pricing differential. Compared to White-owned
and industries. It is important to acknowledge, firms, the results showed that Asian-owned
however, that the weighting process was firms paid 3.34 percentage points higher
constrained by the limitations of available rates, Hispanic-owned firms paid 3.96
data. The U.S. Census data, which is crucial percentage points higher rates, Black-owned
for this analysis, is regrettably incomplete in firms paid 2.82 percentage point higher rates.
its state and industry-specific information, Woman-owned firms paid 2.29 percentage
particularly for minority groups. Despite these points higher rates than man-owned firms.
constraints, a key finding of our study was the
consistency in outcomes between the weight-
ed and unweighted analyses. This consistency
suggests that our core findings are resilient I N T E R E S T R E S U LT S B Y R A C I A L /
and remain unaffected by the chosen meth- ETHNIC & GENDER GROUPS
odology of weighting. We do not include the The preceding analysis reports the overall
weighted analyses in this report, but they are mispricing interest rate differential for each
available from the authors upon request. of the minority groups relative to White-
owned firms. However, the minority groups’
experiences differed across lenders. Since
minority-owned businesses most frequently
borrowed from large banks, we used the
interest rate differential with large banks as
the baseline. We then looked to see if other
lenders (i.e., small banks, FinTech, etc.) had
a higher or lower interest rate differential
between minority groups and White-owned
firms than did large banks.
In Exhibit VIII, the “minority” line is the racial
gap (difference between minority-owned
firms and White-owned firms) in interest rate
paid by that group at large banks. So, Hispan-
ic-owned firms paid 3.57 percentage points
higher interest rates than White-owned firms
at large banks, and that difference is at the .01
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Mispricing Analysis 24
level of statistical significance. Moving across However, the CDFI interaction term is a
the minority line, Asian-American-owned statistically significant -2.91 percentage points,
businesses paid 2.66 percentage points higher implying that at CDFIs, Hispanic-owned firms
than White-owned firms at large banks and paid (3.57 percentage points – 2.91 percent-
that difference is also statistically significant. age points =) .65 percentage points more than
There is no statistically significant difference white-owned firms. The premium paid by
in the interest rates that large banks charged Hispanic-owned firms over white-owned firms
Black- and Native American-owned businesses was lower at CDFIs. We conclude from addi-
compared to White-owned firms, after consid- tional tests that the.65 percentage points is
ering creditworthiness and other attributes of not statistically significant, and thus Hispanic
the firms. The summary is that at large banks, and White firms do not differ in interest rates
Hispanic-owned firms and Asian-owned firms paid at CDFIs.
paid higher rates than those predicted in our
model. For Black- and Native American-owned For brevity and clarity, we will outline the
firms, mispricing by race was not indicated. remaining differences for the other minority
groups and gender differences, without the
The terms noted by “#” are interaction terms specific numbers provided in Exhibit VIII.
that show the difference between that type of Black-owned firms paid the same interest
lender and large banks in the interest rate gap rates as their White-owned counterparts at
between minority- and White-owned firms. large banks, small banks and CDFIs, while
We add the interaction term to the Minority paying higher interest rates than their White-
term to measure if the racial gap in interest owned counterparts at credit unions, fintech
rate (minority–White-owned firms) differs lenders and nonbank finance companies.
between that type of lender and large banks. Asian-owned firms paid higher interest rates
Thus, Hispanic-owned firms paid higher rates than their White-owned counterparts at large
(3.06 percentage points + 3.57 percentage banks. Because of the small sample sizes of
points) than White-owned firms in small Asian-owned firms borrowing from the other
banks, meaning that the racial gap may be lenders, we will not make inferences about the
3.07 percentage points larger in small banks results in those cases.
than in large banks. However, the 3.06 per-
GR APH ON NE X T PAGE
centage points are not statistically significant.
So, we conclude that the difference at small
banks is the same as large banks: 3.57
percentage points. Likewise, the additional
.54 percentage points higher rates paid by
Hispanic-owned firms over White-owned firms
at credit unions compared to large banks are
not statistically significant.
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Mispricing Analysis 25
LENDER INTER ACTION EFFECTS
EXHIBIT VIII
ADDED TO BASIC MODEL
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Mispricing Analysis 26
Woman-owned firms paid the same rates as CO L L AT E R A L
their comparable male-owned counterparts Thus far, this report has focused on mispricing
at large banks, small banks and nonbank in the context of interest rates. The other
finance companies, and higher interest part of the credit agreement is the collateral
rates than their male-owned counterparts requirement. Collateral can be examined in a
at credit unions, CDFIs and fintech lenders. binary approach: it can be observed in anal-
The five lines below the interaction term yses of “yes” or “no” regarding various forms
indicate the difference between the lender of collateral committed for the loan. These
and large banks in the interest rate paid by binary outcomes require a different type of
White-owned firms. White-owned firms paid statistical regression. This study used various
1.197 percentage points less in small banks “probit regressions” of collateral decisions on
than large banks, but the difference is not the credit received by the responding firms.
statistically significant. We ran these probit regressions using all the
control variables to determine if, after consid-
In summary, in the Hispanic-owned firm to ering the control variables, the membership in
White-owned firm comparison, the rates paid a “minority” group-owned or woman-owned
by Hispanic-owned firms to each of the lender business had any impact on the collateral
types are higher than we can explain in our associated with the loan. The models deter-
statistical models. In the Black-owned firm mined the statistical significance of minority
to White-owned firm comparison, the rates or woman-owned business was impacted by
paid by Black-owned firms to large banks are the yes/no outcome regarding that measure
explained within our models (no mispricing of collateral. The probits showed that for the
indicated), and rates paid to large banks do minority-owned groups, the only consistent
not differ statistically from the rates Black- collateral term was that minority-owned
owned firms paid to small banks and CDFIs. groups had a higher probability of an external
However, the rates paid by Black-owned firms party being required to guarantee the loan
to credit unions, fintech lenders and nonbank (Exhibit IX).
finance companies are higher than our
GR APH ON NE X T PAGE
models can explain. In the Asian-owned firm
to White-owned firm comparison, we find that
Asian-owned firms paid higher interest rates
to large banks than we can explain in our
models. The sample of Asian-owned borrow-
ers at other lender types is too small
to make inferences.
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Mispricing Analysis 27
OTH E R O U TCO M E VA R I A B L E
COMPARISONS: MINORIT Y VS .
EXHIBIT IX
WHITE-OWNED FIRMS &
WOMEN- VS. MAN-OWNED FIRMS
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Mispricing Analysis 28
In addition, for Hispanic-owned firms more each other. For example, bundles of various
often the value of the collateral is higher than collateral values and interest rates are
the loan amount compared to white-owned frequently offered as part of a loan package,
firms. Regarding Black-owned firms, Exhibit and their determination is simultaneous.
II shows that 57% of Black-owned firms have Focusing solely on one outcome, however,
blanket liens on their assets, which is the would neglect the preferences of different
highest among all groups, while 30% of white- racial groups in relation to the other outcome.
owned firms borrowed with blanket liens, the To account for the simultaneity and correla-
lowest of all groups. Exhibit IX shows that the tion between these outcomes, we perform
black-white differences in blanket liens are a PCA on all variables related to interest rate
statistically significant after considering all and collateral, encompassing a set of eight
the controls. For Asian American-owned firms, variables, including loan rate, loan spread,
only the external party signature requirement collateral requirement, collateral value,
that is higher compared to white-owned firms blanket lien requirement, business collateral
after considering all the controls. All other requirement, personal collateral requirement,
differences are not statistically significant. and both business and personal collateral
Native American-owned firms are less likely requirements.
to have cases of both business and personal
collateral, than white-owned firms, after Using this analysis, we do not find that racial
considering all controls. Woman owned firms and ethnic minority groups had consistently
were less often required to put up personal greater collateral requirements than White-
assets as collateral than man owned firms. owned firms, and this finding is independent
of the interest rate component.
PR INCIPA L COMPONE NT
A N A LY S I S Hispanic-, Black-, and
As mentioned above, loan terms are jointly
determined and “packaged” together. For
Asian-owned firms have
this reason, we seek to add robustness to loans with higher interest
our results above using Principal Component
Analysis (PCA). PCA, originally introduced by rates than White-owned
Hotelling (1933), is a widely used technique
firms
for dimensionality reduction, with the
different terms creating the dimensions in
the PCA analysis. PCA synthesizes a series
of correlated variables into one or more Hispanic-, Black-, and Asian-owned firms have
composite variables, referred to as principal loans with higher interest rates than White-
components, which capture the most salient owned firms. Additionally, woman-owned
sources of variation in a dataset. firms have loans with higher interest rates
than man-owned firms. These results align
Loan interest rate and collateral requirements with our previous findings. The appendix
are important outcomes related to loan shows the statistical analyses related to PCA.
quality and they are often correlated with
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Mispricing Analysis 29
IV. CONCLUSION
FINDINGS
A major purpose of this study was to explore is consistently mispriced is that third party
the economic impact of loan mispricing on co-signatories are demanded more frequently
diverse-owned companies in terms of interest for minority-owned firms than is justified by
rates and collateral required in obtaining our economic analysis. Unfortunately, the
credit. survey data did not have sufficient detail for
us to pursue an estimate of the economic
We found that for Hispanic-, loss associated with the greater requirement
for third party signatures across the minority
Asian- and Black-owned firm borrowers. This is an important topic for
firms, the interest rate paid future research. Also, previous studies have
found that borrowers from minority-owned
was higher than for compa- firms are rejected for loans at a higher rate
than White-owned firms of the same credit-
rable White-owned firms...
worthiness. We did not look at loan rejections
in this study, therefore, this is another
We found that for Hispanic-, Asian- and Black- important topic for future research. We need
owned firms, the interest rate paid was higher to understand the issues that minority-owned
than for comparable White-owned firms, firms face in financing their survival and
after considering the controls for risk that growth. We as a nation should identify and
determine the interest rate on loans, including eliminate impediments to their growth that
the firm’s industry, financial attributes, owner should not exist.
traits, credit history and type of loan. We con-
ducted robustness tests to verify the strength
of these results. Another component of the COST TO AFFECTED
credit received by the firm is the collateral. We BUSINESSES
performed tests to determine if the collateral We were able to make estimates of the
requirements were justified given the charac- additional interest that the minority-owned
teristics of the firm as enumerated above. We firms paid for credit over and above the rates
found that the only collateral requirement that that comparable White-owned firms paid. We
found that the interest rate differential due to
30
the race of the business owner totaled $8.1 estimates from three models, all of which
billion annually (Exhibit X-A). The difference used the main data set (at least 50% own-
in the costs to each racial or ethnic group ership for that group, with no other group
was a result of the number of businesses that holding more than 49%): the basic model,
secure loans annually, the average loan size the weighted-by-state model, and the
of the group, and the mispricing differential weighted-by-state/industry model. Another
estimated in this study. approach would be to use the range of the
three minority-White differentials to quantify
As mentioned above, our final mispricing the range of values. These results are present-
differential in minority-White interest rate for ed in Exhibit X-B below.
each minority group was the average of
ADDITIONAL INTEREST
E X H I B I T X- A
PAID FROM MISPRICING
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Conclusion 31
The annual business survey (ABS) of the US small business loans that we consider in this
Census Bureau provides the number of small study. The average loan size for each group is
businesses by race/ethnicity in column A. The in column E. Column F shows the interest rate
latest figures refer to 2021. Column B is from differential we estimated in the study, and
the Small Business Credit Survey (SBCS) of Column G reports the estimated annual total
the Federal Reserve for 2023. It indicates the dollars paid by each of the minority groups
fraction of each of the groups that applied for through mispricing.
credit in 2022. It also shows, in Column C, the
fractions approved for loans in each of the Exhibit X-A presents the average cost of the
groups. Note that the fraction approved in- interest rate differential to different racial
cludes those credits that were fully approved and ethnic groups. Yet, in conducting the data
and those that were partially approved. By analysis, we used several analytical tools to
multiplying through each of the foregoing test the robustness of the results. This led
columns, we obtain the estimated number of to developing a range possibilities for the
approved loans for each group, as shown in interest rate differential and thus a range of
Column D. The average loan size is estimated potential impacts in excess interest payments
based on the fiscal year 2023 average size of paid by different racial and ethnic groups.
SBA 7(A) loans. These small business loans are These results are presented in Exhibit X-B
often competitors with the conventional below.
LENDER INTER ACTION EFFECTS
E X H I B I T X- B
ADDED TO BASIC MODEL
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Conclusion 32
REFERENCES
Asiedu, E. , Freeman, J.A., & Nti-Addae, A. Brick, I. E. & Palia, D. (2007), Evidence of
(2012). Access to credit by small businesses: jointness in the terms of relationship lending.
How relevant are race, ethnicity, and gender? Journal of Financial Intermediation, 16,
American Economic Review, 102 (3), 532-37. 451–476.
Barkley, B., & Schweitzer, M. (2023). Credit Cavalluzzo, K.S., & Cavalluzzo, L.C. (1998).
availability for minority business owners in Market structure and discrimination: The case
an evolving credit environment. Economic of small businesses. Journal of Money, Credit,
Development Quarterly, 37(3), 230-242. and Banking, 30 (4), 771–792.
Bates, T., & Robb, A. (2015). Has the Communi- Cavalluzzo, K.S. Cavalluzzo, L.C., & Wolken,
ty Reinvestment Act increased loan availability J.D. (2002). Competition, small-business
among small businesses operating in minority financing, and discrimination: Evidence from a
neighborhoods? Urban Studies, 52, 1702–1721. new survey. Journal of Business, 75 641–680.
Bates, T., & Robb, A. (2016). Impacts of owner Dunteman, G.H. (1989). Principal component
race & geographic context on access to small analysis. Sage Publications.
business financing. Economic Development
Quarterly, 30, 159–170. Fairlie, R.., Robb, A., & Robinson, D. T. (2022).
Black and White: Access to capital among
Biswas, Basudeb and Caliendo, Frank (2002) minority-owned start-ups. Management
“A Multivariate Analysis of the Human Devel- Science, 68(4), 2377–2400.
opment Index” (2002). Economic Research
Institute Study Papers. Paper 244. Haq, R., & Zia, U. (2013). Multidimensional
wellbeing: An index of quality of life in a de-
Blanchard, L., Bo Z., & Yinger, J. (2008). Do veloping economy. Social Indicators Research,
lenders discriminate against minority and 114, 997-1012.
woman entrepreneurs? Journal of Urban
Economics 63 (2), 467–97. Hotelling, H. (1933). Analysis of a complex of
statistical variables into principal components.
Blanchflower, D. G., Levine, P.B., & Zim Journal of Educational Psychology, 24(6),
merman, D.J. (2003). Discrimination in the 417–441.
small-business credit market. Review of
Economics and Statistics 85 (4), 930–43. McGilvray, M. (2005). Measuring non-eco-
nomic well-being achievement. Review of
Bostic, R. W., & Lampani, K. P. (1999). Racial Income and Wealth, 51(2), 337-364.
differences in patterns of small business
finance: The importance of local geography.
Federal Reserve Bank of Chicago Proceedings,
149-179.
33
Ram, R. (1982). Composite indices of physical Slotted, D. J. (1991). Measuring the quality of
quality of life, basic needs fulfilment, and life across countries. The Review of Economics
income: A ‘principal component ‘representa- and Statistics, 684-693.
tion. Journal of Development Economics, 11(2),
227-247.
Scott, M. Bone, S., Christensen, G., Lederer,
A., Mende, M., Christensen, B., & Cozad, M.
(2024). Revealing and mitigating racial bias and
discrimination in financial services. Journal of
Marketing Research. 61(4), 598-618.
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | References 34
APPENDIX
US ING PR INCIPA L COMPONE NT
A N A LY S I S T O C O N S I D E R T H E
J O I N T D E T E R M I N AT I O N O F
LOAN TERMS
As mentioned earlier, loan terms are jointly collateral values and interest rates are
determined and “packaged” together. For frequently offered as part of a loan package,
this reason, we seek to add robustness to and their determination is simultaneous.
our results above using Principal Component Focusing solely on one outcome would
Analysis (PCA). PCA, originally introduced by neglect the preferences of different racial
Hotelling (1933), is a widely used technique for groups in relation to the other outcome. To
dimensionality reduction, with different terms account for the simultaneity and correlation
creating the dimensions in the PCA analysis. between these outcomes, we perform a
PCA synthesizes a series of correlated vari- PCA on all variables related to interest rate
ables into one or more composite variables, and collateral, encompassing a set of eight
referred to as principal components, which variables, including loan rate, loan spread,
capture the most salient sources of variation collateral requirement, collateral value,
in a dataset. This method has found broad blanket lien requirement, business collateral
applications in constructing composite indices requirement, personal collateral requirement,
and summarizing complex data, including and both business and personal collateral
those related to economic development, quali- requirements.
ty of life, well-being (e.g., Dunteman, 1989;
Ram, 1982; Slotted, 1991; Biswas & Caliendo, In Panel A of Exhibit A-1, the principal com-
2002; McGillivray, 2005; Haq & Zia, 2013). ponents derived from the original variables
are presented. Given that eight original
The principal components are linear combina- variables are utilized, the PCA generates eight
tions of the original variables and are ranked components. Following the Kaiser rule, only
according to the proportion of variance they components with high eigenvalues (typically
account for. The first principal component exceeding 1) are deemed significant and
explains the most substantial portion of selected as principal components (PCs). In our
the variance in the original variables, while case, the first two components meet this crite-
the second captures the largest portion of rion. As indicated in the “proportion” column,
the remaining variance, and so on. These the first component explains 46.28% of the
components are orthogonal to each other and variance in the eight variables, followed by the
provide a more concise representation of the second component, which explains 24.78% of
original variables. the variance. Together, these two components
account for 71.06% of the variance.
Loan interest rate and collateral requirements
are important outcomes related to loan Panel B of Exhibit A-2 shows the loadings on
quality and are often correlated with each the principal components, which provides
other. For example, bundles of various information on which original variables
35
contribute most to the first and second In Exhibit A-2, Panel A shows the results using
principal components. It is evident that the the first principal component (the collateral
first component heavily loads on collater- component) as the dependent variable.
al-related variables, allowing us to define it as There is no evidence that racial and ethnic
the collateral component of the loan quality. minority groups have greater collateral
The second component exhibits significant requirements than majority White-owned
loadings on loan rate-related variables; thus, firms, and this finding is independent of the
we can define it as the loan rate component of interest rate component. However, majority
loan quality. Due to the orthogonality of the woman-owned firms do exhibit higher collat-
PCs, the second PC (the loan rate component) eral requirements. Panel B shows the results
is independent of the first PC (the collateral using the second principal components as the
component). This feature allows us to exam- dependent variable, especially focusing on the
ine the racial disparities in each component loan rate component, which is independent of
separately without concern for the correlation collateral. Firms owned by Hispanic, Black, and
between the two. Consequently, we use these Asian owners have loans with higher interest
two principal components as dependent rates than businesses owned by Whites.
variables instead of the original variables to Additionally, woman-owned firms have loans
assess racial differences. Because the princi- with higher interest rates than man-owned
pal components represent composite indices firms. These results align with our previous
of the original variables, our focus centers on findings.
the direction of the effect and its statistical
GR APH ON NE X T PAGE
significance in interpreting the results.
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Appendix 36
P R I N C I P A L C O M P O N E N T A N A LY S I S :
E X H I B I T A -1 USE OF PCA TO SUMMARIZE LOAN
QUALIT Y
PANEL B
Comp1 loads heavily on
coll related variables
Comp2 loads
heavily on loan rate
PANEL A
Proportion of variation
in the data explained by
each component
Kaiser Rule: eigenvalue>1
There is significant break
between 2 and 3. keep 2.
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Appendix 37
EXHIBIT A-2 REGRESSION MODELS USING PCA
FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON | Appendix | © 2024 | consult1@uw.edu 38
File and source
- File
- 2025-01-29_a32584_d276687_uw-foster-school-of-business-interest-rate-and-collateral-di.pdf
- Size
- 2,021,359 bytes
- SHA-256
- b89cf99facb4026b86665fe8d29600986ec048a02b0d8fde94bb60ea3c5199a5
- Original
- app.leg.wa.gov