Revealing and Mitigating Racial Bias and Discrimination in Financial Services

Despite laudable legal, political, and societal efforts to eliminate inequalities, racial discrimination remains a reality in the marketplace in the United States and other countries (Gabbidon and Higgins 2020; Gunarathne, Rui, and Seidmann 2022; Korver-Glenn 2018). In marketing, racial discrimination has received limited attention (Arsel, Crockett, and Scott 2022; Grier, Thomas, and Johnson 2019), although it affects millions of consumers (Hanson et al. 2016; Perry 2019). For example, in the financial realm, Bartlett et al. (2022) investigated racial discrimination in consumer lending with Fannie Mae and Freddie Mac data; the results show that lenders charge risk-equivalent Black or Hispanic (vs. non-minority) borrowers more for purchasing and refinancing mortgages, costing Black or Hispanic borrowers more than $450 million in aggregate in extra interest per year. Similarly, for racial or ethnic minority entrepreneurs who seek financing for their small business, the journey is more difficult than it is for their equally situated White counterparts (Bone, Christensen, and Williams 2014), and their loan outcomes are less favorable, as minority-owned (vs. non-minority-owned) firms are rejected at higher rates and pay higher interest (U.S. Department of Commerce 2010).

Against this background, our work contributes to a growing empirical literature on racial discrimination in marketing. We present field data on financial service encounters in the United States (two field experiments and a field survey that investigate the service experiences of Black [vs. White] customers inquiring about small business loans). An additional laboratory experiment then explores the provider's perspective when evaluating such small business loan requests. The results reveal that racial discrimination by service employees permeates Black customers’ journeys, with fundamental consequences that further cement racial inequality. Our findings contribute to marketing, and to the theme of “Mitigation in Marketing,” in multiple important ways.

First, despite efforts to fight inequality, racial discrimination persists in the United States, and some evidence in marketing documents such discrimination (e.g., Perry 2019). However, with subtler and more covert racism in daily life rising (Kipnis et al. 2021), much less is known about what discrimination looks like in service encounters, and how it contributes to sustained inequality between Black and White customers. For example, Bartlett et al. (2022) show discrimination in financial lending based on secondary data, but, although these authors note that “how discrimination happens is an important question” (p. 55), they leave the exploration of this topic to future research. We address this gap and reveal key indicators of discrimination against Black (vs. White) customers: we show that (1) service employees provide better financially qualified Black (vs. White) customers with objectively inferior service outcome quality (i.e., financial solutions offered), (2) better qualified Black (vs. White) customers experience inferior service process quality (i.e., employees’ display of warmth and competence), and (3) Black (vs. White) customers report lower loyalty intentions toward the financial firm. These findings provide novel and timely evidence that racial discrimination affects both the service process and service outcomes for Black customers. Thus, racial discrimination by service employees creates a counterproductive (and even illegal, under the Equal Credit Opportunity Act [1974]) lose-lose situation for Black customers (e.g., via exclusion from financial services, or costly inclusion) as well as firms (i.e., lost revenue, legal risks).

Second, grounded in theory of spontaneous inference making in person perception (Fiske and Neuberg 1990; McCarthy and Skowronski 2014) and the justification-suppression model of prejudice (Crandall and Eshleman 2003), we provide novel insights that illuminate conditions under which discrimination occurs, and when and why discrimination is mitigated in the context of financial loan services: namely, when (1) Black customers (are perceived to) have relatively high socioeconomic status (SES), or (2) the Black customer's firm (for which they seek the loan) has a more complex and sophisticated legal structure (e.g., corporation vs. sole proprietorship). ¹

The results on these two moderating factors show that cues that are inconsistent with cruel stereotypes about Black people’s level of financial knowledge, credibility, and trustworthiness in commercial settings (Austin 2004, p. 1250) can mitigate discrimination in financial services. In parallel, these findings also reveal that service employees apply racial stereotypes to assess potential customers. The employees then hold Black (vs. White) customers to higher standards, which results in exclusion from financial services or costly inclusion. Notably, inconsistent (i.e., counterstereotypic) cues—that do not fit with stereotypical categorizations of Black customers, such as higher (vs. lower) SES or a more (vs. less) sophisticated business structure—are required by Black customers to mitigate discrimination. These novel insights into factors (namely, SES and business structure) that help suppress discrimination contribute to efforts of mitigating undesirable marketing practices (Gupta et al. 2022) and make crucial conceptual as well as practical contributions to alleviating disparities that prior research has uncovered (e.g., Bone, Christensen, and Williams 2014).

Third, our studies leverage a unique methodology, namely two matched-pair mystery shopping field experiments. Our results stem from marketplace interactions across dozens of banks and locations across the United States, indicating the robustness and generalizability of the findings. Not only do these results show that racial discrimination in the marketplace persists in the wake of prior research conducted nearly a decade ago (Bone, Christensen, and Williams 2014), but our work also responds to recent calls in the literature. For example, Gabbidon and Higgins (2020, p. 132) call for research employing covert studies in services where attitudes toward Black customers can be more naturally captured, because “it goes without saying that most retailers will deny animus toward racial/ethnic minorities if directly asked. Covertly studying the topic would likely garner their true sentiments.” Our data offer such unique insights from actual service encounters, which remain rare in marketing, and make additional contributions over prior work: specifically, our outcome measures reveal how and when frontline employees effectuate bias in financial services as they recommend or withold distinct offerings (i.e., a home equity line of credit [HELOC] vs. a business line of credit [BLOC]) to potential customers.

Fourth, we also explore the moderating role of business structure on racial bias in services through the lens of the individuals making the loan decision (Study 4). The results again show differential effects of a customer's race and business structure, such that Black applicants have a greater likelihood of being offered a BLOC when their business structure is sophisticated (vs. a sole proprietorship), while the likelihood of White applicants being offered a BLOC is relatively unaffected by their business structure. Importantly, Study 4 also explores the underlying process that influences participants’ evaluations of the focal customer: it shows a serial mediation path via trust and perceived default likelihood, which directly links back to the aforementioned derogatory stereotypes about Black people’s credibility and trustworthiness in commercial settings (Austin 2004). Specifically, for Black loan applicants, as business structure is more sophisticated, increased trust leads to a decrease in perceived default likelihood, and in turn, to an increased likelihood to offer a BLOC; this process does not emerge for White applicants as a function of business structure.

Finally, our work has actionable managerial and policy implications (which we further illustrate in the “General Discussion” section). For companies and governmental agencies to mitigate inequality in the marketplace, they need a more detailed understanding of what discrimination looks like. Discrimination is not only ethically wrong and potentially illegal; it also conflicts with a firm's profit goals. When employees discriminate against otherwise viable customers (e.g., by discouraging them from applying or rejecting their loan applications), they lose the firm money, and managers and policy makers should work to combat this behavior.

Spontaneous Inference Making and Race in the Marketplace

Upon encountering another person, people form spontaneous impressions, often using information that is easy to interpret (McCarthy and Skowronski 2014). For an instant evaluation, observers tend to default to automatic categorizations based on a target's salient, phenotypical characteristics (e.g., skin color) (Blair, Judd, and Fallman 2004; Fiske, Lin, and Neuberg 1999). Placement into a certain social category then triggers associations or prejudices related to that category (Bodenhausen, Macrae, and Sherman 1999). We focus on social inference making because loan application and approval processes typically “include a subjective informal process, often described as a ‘gut reaction’ that is framed by the capital provider's beliefs and values” (Eddleston et al. 2016, p. 492; also Wilson et al. 2007); thus, this process is vulnerable to discrimination (e.g., if an employee made stereotypical inferences about Black customers’ credibility or trustworthiness in commercial settings; Austin 2004).

Race as a Cue in the Marketplace

Extending prior work on racial discrimination in services (e.g., by real estate agents and retail employees) (Francis and Robertson 2021; Harkness 2016; Pager and Shepherd 2008), we theorize that employees use a customer's race (Black vs. White) as a cue to categorize and make inferences about this individual, including how desirable they are as a customer. In turn, these inferences influence the employee's behavior during the service process (e.g., warmth and competence displayed toward the customer) and which service solution is (or is not) offered to this customer. In short, we expect that employees rely on heuristic inferences about customers based on their race rather than the individual's merit (which is stereotyping), and that the employees then treat Black customers differently and more poorly (which is discrimination).

Illustrated in a financial loan application context, our theorizing suggests that a loan officer not only focuses on relevant financial indicators (e.g., income) but also notices an applicant's salient demographic (racial) cues. With regard to Black (vs. White) customers, the employee is then likely to make less favorable inferences about the loan applicant's status and abilities, which negatively affects the loan approval process. For example, gauging the perceived financial value and risks of a loan applicant (e.g., spending time with them, collecting information from them, and providing information to them throughout the loan process), service employees draw on stereotypes as they discourage, avoid, stonewall, or reject members of certain racial groups although they may be viable customers (thereby negatively impacting the financial performance of the firm).

Accordingly, we propose that service employees consider race as an initial cue for their overall assessment of and subsequent behavior toward potential customers. We expect that (1) employees interacting with Black (vs. White) customers will provide inferior service process quality (e.g., employees will be less welcoming, friendly, and warm) and (2) employees will provide inferior service outcome quality (e.g., they will offer financial solutions that are objectively less appropriate and favorable for the customer's needs). We therefore hypothesize:

H₁: Black (vs. White) customers experience less favorable service processes and service outcomes.

Inconsistent Cues to Mitigate Categorization and Prejudice

Impression formation is a function of the degree to which observed cues are consistent. That is, categorization and stereotyping do not always dominate the consideration of the personal qualities of an individual (Bodenhausen, Macrae, and Sherman 1999); rather, after initially categorizing a person, observers use additional information to either preserve the initial categorization or correct it (Fiske and Neuberg 1990). A focal category or stereotype will only be applied to the target when the additional information about this target is consistent with the implications of that category (Brewer 1988). Vice versa, if the initial categorization is deemed inappropriate due to inconsistent (i.e., counterstereotypical) cues, observers use the new information to recategorize the target. In short, initial categorizations are strengthened by the presentation of converging cues, but they are weakened by diverging cues (McCarthy and Skowronski 2014).

These insights inform our theorizing on mitigating discrimination in financial services as a function of inconsistent cues about Black customers. Because race is a salient cue, it is often used for initial categorization (Fiske, Lin, and Neuberg 1999), for example, via cruel stereotypes about Black people’s credibility and trustworthiness in commercial settings (Austin 2004). However, this theorizing suggests that an additional counterstereotypical cue—which provides observers with information that is inconsistent with stereotypes about Black people—might reduce discrimination. We propose (1) a Black customer's SES and (2) the complexity and sophistication of their legal business structure function as inconsistent cues that require additional processing from social observers and can mitigate discriminatory behaviors. Our rationale is that both of these factors counter the negative stereotypes about Black people’s sophistication and intelligence (Massey et al. 2016, p. 12).

The Moderating Role of Socioeconomic Status

SES, which is defined as one's place in the hierarchy (related to factors such as occupational status, income, or education) is one of the primary stereotypical indicators people use to organize social groups; the corresponding race–status stereotype characterizes White Americans as high in SES and Black Americans as low in SES (Dupree et al. 2021; Durante and Fiske 2017). Our theorizing suggests that service employees draw on such racial stereotypes. Indeed, qualitative research on racial stereotypes in services discovered that employees often make inferences about economic, moral, or cultural characteristics of non-White customers. For example, real estate agents were found to associate being “Black” with being occupationally inferior to Whites, being financially irresponsible and unstable, and not being knowledgeable about the U.S. financial system, as well as having a low income (Korver-Glenn 2018). Such stereotypes likely lead to unfavorable assessments of Black loan applicants (e.g., employees might stereotype Black customers to be less trustworthy and infer a higher default risk related to the focal loan) and undermine the loan application process for Black (vs. White) customers.

However, extending prior work on SES (see Web Appendix A), and answering calls for more research at the intersection of SES and race (Dupree et al. 2021), we predict a moderating role of SES on the relationship between customer race and progress toward a loan. Our prediction draws on the following insights: Overall, Black Americans are indeed lower in SES than White Americans (Dupree et al. 2021), which is one driver of the stereotype. Yet, prior research has found that exposure to counterstereotypical cues reduces stereotypical attitudes (e.g., Burns, Monteith, and Parker 2017; Dasgupta and Asgari 2004). Specifically, once a social category is activated and both stereotypic and counterstereotypic cues are presented, counterstereotypic information is likely to dominate, especially during the impression formation phase (Stangor and McMillan 1992). This pattern supports the notion that the mind “does not waste valuable attentional resources on familiar (i.e., expectancy redundant) items”; instead, it directs attention to the encoding of unexpected cues (Macrae and Bodenhausen 2000, p. 106). Thus, as perceivers recognize the inconsistency confronting them (i.e., a focal individual does not fit the stereotype), they make sense of the situation by resolving this discrepancy between their expectations and focal cues; they do so by individuating the target, engaging in relatively more deliberate and effortful processing that focuses on the individual rather than their group membership (Fiske and Neuberg 1990). Accordingly, we expect that when service employees interact with Black customers with relatively higher SES (i.e., customers who do not fit the stereotype of Black Americans having low SES), those employees will individuate the Black customer by focusing on their personal profile rather than their social group membership. Consequently, as the employees recognize a poor fit between the categorical expectations and the information related to the focal customer (higher SES is an inconsistent cue), they should be less likely to engage in racial discrimination. This theorizing also draws on research on prejudice suppression, which argues that not only forces such as social norms and personal standards, beliefs, and values but also situational attributions can help suppress prejudice (Crandall and Eshleman 2003). In summary, we predict that the negative effects experienced by Black customers are greater for those with lower (vs. higher) SES, but we expect that White customers are unaffected by their SES.

H_2a: A customer's SES moderates racial discrimination such that Black customers with low (vs. high) SES experience (i) less favorable service processes and (ii) less favorable service outcomes, but service experiences of White customers are unaffected by their SES.

The Moderating Role of a Sophisticated Business Structure

Similarly grounded in a cue inconsistency and prejudice suppression lens, we expect a moderating role of the legal business structure on the relationship between customer race and progress toward a loan. In examining whether risk-equivalent Black (vs. White) business owners are held to a different (discriminatory) standard in obtaining financing, we note that Black people face negative stereotypes about their competence, intelligence, sophistication, and education level (Dupree et al. 2021); they also encounter stereotypes about their effectiveness and success in leadership roles (Taylor et al. 2019), which further discounts perceptions of their managerial skills. This harsh reality suggests that bank employees are likely to question the desirability and viability of Black (vs. White) business owners and use different evaluative criteria to the detriment of Black customers.

However, drawing on our theoretical rationale of counterstereotypical cues and consistent with signaling theory (Connelly et al. 2011), we expect that certain characteristics of a business can signal its inherent quality to financial service providers, which then mitigates discrimination (e.g., in small business loan applications). Signaling theory (Spence 2002) has been used in entrepreneurial research to explain how lenders consider signals of a firm's underlying quality to evaluate the potential of an entrepreneur and reduce the lender's information asymmetry inherent in loan settings. The idea of such studies is that lenders—to assess loan applicants and their ability to pay back loans—consider signals that indicate a venture's viability and underlying quality, which in turn, help explain differences in loans provided (Eddleston et al. 2016). What has received less attention is how the meaning of signals varies as a function of the sender (Eddleston et al. 2016) and, especially, the intersection of the entrepreneur's race and the legal business structure.

As lenders use “viability signals” that indicate the stability and legitimacy of a business, we theorize that one viability cue is the legal business structure of a (Black-owned) small business. A sole proprietorship, ² for example, is the simplest and least expensive business structure, which is owned by one individual only and requires no formal action to establish (U.S. Small Business Administration [SBA] 2022b). (Notably, more than 80% of small businesses in the United States fall into this category [Federal Reserve 2021b]). Thus, a more complex structure, such as LLC status, might be an indicator of business credibility as it requires advanced understanding of legal and tax considerations. Accordingly, a relatively more complex business structure signals to lenders that the business is more “serious” (Storey 1994). A small business’s legal structure is relied on by loan officers as a relevant input in loan applications (Kim and Elias 2007); indeed, in professional trainings to become a commercial loan officer, candidates are instructed to identify the legal structure of the small business (Todd, Vannoy, and Marin 2005). More complex and sophisticated legal structures are linked to increased loan success (i.e., payment without default) from the lenders’ point of view (Page et al. 1977). Thus, choosing a more sophisticated and complex legal structure is an important strategic decision for the firm because it helps define its maturity (Chawla et al. 2007), in addition to other factors such as age (years in business) and size (sales, employees, etc.) (Khurana et al. 2021). Another challenge for a sole proprietorship is that lenders are more concerned about the business’s legitimacy related to its size, such that smallness undermines the ability to secure loans; notably, sole proprietorships are often very small, with few (if any) employees (Cassar 2004). Finally, sole proprietorships also tend to lack the reputation, economies of scale, and strategic partners that more sophisticated businesses have; these aspects can further increase the lender-perceived risk of financing a sole proprietorship (Eddleston et al. 2016).

Related to these insights, small businesses organized as sole proprietorships expose lenders to “potentially higher levels of risk since the risk of repayment depends on a single owner” and their personal resources; in contrast, alternative forms of ownership (e.g., partnerships and corporations) can spread the risk of loan repayment “among greater numbers of owners” (Van Auken and Neeley 1996, p. 239). Thus, the number of owners associated with certain business structures also signals to service providers that the owners bring social and financial capital to bear on the transaction, which should make them less risky (more desirable) customers. This notion is consistent with research that has shown how joint liability lending (i.e., providing a loan to self-selected groups of borrowers and making them jointly liable) can alleviate credit market failures caused by information asymmetry (e.g., Ghatak 2000).

In synthesis, we expect that the negative effect of race on small business loans is moderated by the customer's business structure. We predict that Black (vs. White) customers who apply for a loan related to a more (vs. less) sophisticated business are treated more favorably and receive superior opportunities because the additional cue (business sophistication) is inconsistent with the employee's race-related categorization of a Black customer.

H_2b: Business structure will moderate racial discrimination such that Black customers with less (vs. more) sophisticated business structure will experience (i) less favorable service processes and (ii) less favorable service outcomes, but service experiences of White customers will be unaffected by their business structure.

Taken together, H_2a and H_2b predict that Black customers of (1) high (vs. low) SES and (2) high (vs. low) business sophistication provide cues that are inconsistent with a core facet of racial discrimination, namely the negative stereotyping of Black people’s financial credibility, sophistication, and trustworthiness, as well as their level of risk as customers (Burns, Monteith, and Parker 2016). In other words, what both of our proposed moderators have in common on a conceptual level is that they provide service employees with such counterstereotypical cues, which should mitigate the discrimination that Black customers otherwise face.

Empirical Overview

In collaboration with a partner organization, we conducted studies to test our hypotheses. In three field studies and a lab study, we examine how racial bias influences product portfolio recommendations (Studies 1 and 4), and how customers’ SES (Study 2) and small business structure (Studies 3 and 4) can mitigate racial discrimination in financial services.

In field Study 1, we demonstrate how frontline service employees’ (FLEs’) selective presentation of the firm's financial product/service portfolio results in racially biased marketing (also termed racial steering, shutout, or lockout). In field Study 2, we demonstrate that SES can reduce racial discrimination: we find that Black (vs. White) customers are treated less favorably in terms of the FLEs’ behaviors (i.e., actual warmth and competence behaviors during the interaction); yet, these negative FLE behaviors are mitigated when a Black customer has higher (vs. lower) SES.

In field Study 3, we show that actual bank loan approval rates are lower for Black (vs. White) customers, but this is mitigated when a Black customer has a more sophisticated business structure (e.g., partnership or corporation). In Study 4, with participants who have professional experience in financial services, we demonstrate the interaction of race and business structure on product recommendations and uncover the role of employee-inferred trustworthiness and the corresponding inferred default risk of the customer as a key process underlying the effect.

These findings demonstrate racial discrimination in financial services, but they also help uncover actionable steps that can be taken by firms and policy makers to help mitigate and eliminate the occurrence of racially biased behaviors in financial services.

Study 1: Proposed Service Solution and “Steering Away” as Racial Discrimination by Frontline Employees in Financial Services

Prior literature has established that Black customers experience less favorable service in banking (Bone, Christensen, and Williams 2014), which might affect the service process and outcome quality. Employees in financial service contexts possess a great deal of discretion in terms of the types of products they can recommend to a customer from the firm's portfolio of offerings (Bone and Mowen 2010; Harkness 2016; Korver-Glenn 2018), and often, financial products vary in terms of their fit with customer needs, their desirability in terms of monetary cost, and the level of risk incurred by the customer. This study examines how racial discrimination is enacted in financial services, based on the mix of products in the firm's portfolio offered to Black (vs. White) customers. That is, we test (H₁) whether service providers tend to offer Black (vs. White) customers more often (1) inferior products, (2) superior products, or (3) some combination.

Design, Participants, and Procedure

For this study, we employed a matched-pair mystery shopping field experiment in which Black and White testers visited banks in the Atlanta metropolitan area over a four-month time window. This method is used in fair lending testing and has been previously used in academic research (e.g., Bone et al. 2019), and is also an accepted and relied-on method in U.S. courts as evidence of differential treatment and discrimination in the marketplace (Lubin 2010; Schochet 2009). Testers visited a bank branch location under the premise of being a potential customer; then the testers reported on how they were treated by bank employees (Bone et al. 2019). In total, there were 12 Black testers and 12 White testers. The procedure was for two Black testers and two White testers to visit the same bank branch within a two-week period (each racial group had a male tester and a female tester in counterbalanced order). Our target was 200 unique tests (100 per racial group). The total sample size of tests included here is N = 195 (some tests could not be completed due to bank locations being closed). Testers visited 52 separate bank branches, consisting of 14 different parent banks. Bank branches were selected for testing using a stratified random sampling technique. Banks in the test area were ranked according to the number of loans they made to small businesses in low- and moderate-income census tracts, using Community Reinvestment Act data collected by the Federal Financial Institutions Examination Council. This prevented our study from oversampling banks with particularly poor or excellent small business lending ratings. The ranked banks were then stratified by lending performance quintiles and then proportionally and randomly selected to be tested. The sample consists of banks that are considered large banks (at least $10 billion in total assets [Federal Reserve Board 2021]), as well as smaller “community” banks (less than $10 billion in total assets).

All testers were given small business profiles that were strong enough to easily qualify for a loan based on financial ratios, credit scores, revenues, accounts receivable, and years in business. To make this a more conservative experiment, the profiles of the Black (vs. White) testers were objectively superior in terms of having greater business income, more years in operation, more money in the bank, and higher credit scores. The type of business and industry (e.g., graphic design, home design) were assigned to the testers and were equally distributed across the racial groups. Based on these tester profiles, in an equally accessible small business lending marketplace, the Black testers should receive better outcomes than the White testers, based on their objectively stronger business profiles. To avoid detection by the bankers, some elements of each profile were varied, such as the tester's name and the company name. The testers were trained on the profile of their small business so that they could respond to questions from bank employees about the business. Testers were unaware of the study hypothesis.

Outcome Variables

Testers visited the bank and inquired about obtaining a loan for their small business. After their interaction at the bank, testers completed a survey about their experience, and logged the start and end time of the visit. A focus of this study is to understand the types of loan products service providers recommended to the testers. Two widely used lines of credit are available in the small business arena: a HELOC and a BLOC. From the point of view of a small business loan customer, a HELOC is generally less favorable than a BLOC as it requires the collateralization of their home and can be expensive. For most U.S. households, a home is their largest asset and the cornerstone of total household wealth (Pew Research 2011). For most small business loan customers, a HELOC will place their largest personal asset at risk, which can threaten overall household wealth. The Forbes Finance Council notes, “HELOCs aren't just risky; they’re significantly more expensive than other options” (Camberato 2019). In contrast, a BLOC tends to be more favorable for the customer because it offers more flexible payment terms, is less costly, and requires less personal risk. BLOCs are typically unsecured and do not require collateral (Bank of America 2022), and if they are collateralized, they are usually backed by business (not personal) assets.

We also confirmed the favorability of a BLOC (vs. HELOC) with a study of 301 financial service workers. When asked about whether a BLOC or HELOC was more favorable for the customer, a significant proportion of respondents (91%; χ²(N = 301) = 451.3, p < .0001) indicated that the BLOC was the more favorable option for the customer (see Web Appendix B).

In the main study, we instructed testers to indicate whether or not they were offered each type of loan using a binary “no” or “yes” response for each loan type. Testers also indicated their customer loyalty intentions based on the experience (average of whether they would return to the bank [no/yes] and recommend it to others [no/yes]). Testers also provided a written narrative of their experiences during the visit (see Web Appendix C for exemplary excerpts). ³

Results

Loan options offered (BLOC vs. HELOC)

We conducted a binary logistic regression with BLOC (no = 0, yes = 1) as the outcome variable and race as the independent variable. ⁴ Results revealed a race main effect (Black = 17.35%, White = 40.43%; Wald χ²(N = 192) = 11.94, p = .001); Black (vs. White) customers were offered the BLOC less often. We conducted a binary logistic regression analysis with HELOC as the outcome variable and race as the predictor variable. There was no significant difference in bank employees recommending HELOC loans to Black and White customers (Black = 12.77%, White = 15.79%; Wald χ²(N = 189) = .352, p = .55).

In other words, there was no significant difference in the rate at which bank employees recommended the less favorable HELOC loans to Black and White customers; however, the more favorable BLOC was recommended significantly less often to Black (vs. White) customers (supporting H₁), although the financial profile of the Black customer was objectively stronger.

Loyalty intentions to the company

We conducted a one-way analysis of variance (ANOVA) on loyalty, with customer race as the independent variable. Results revealed a significant race main effect (M_B = .6818, M_W = .8073; F(1, 193) = 4.50, p = .035). Black customers had significantly lower loyalty intentions toward the bank than White customers.

For all studies, probabilities, means, and standard deviations are reported in Web Appendix D; models with control variables included for HELOC, BLOC, and loyalty remain consistent and significant (and these are reported in Web Appendix E).

Discussion

Study 1 reveals insight regarding what racial discrimination looks like in small business lending: Black and White customers were offered unfavorable products (from the customer's point of view) at similar rates. However, Black (vs. White) customers were less often offered the more favorable BLOC, as predicted in H₁; that is, Black customers were less likely to be offered the less expensive, less risky, more flexible offering. The testers also provided open-ended responses about their experience. The following are two exemplary responses from testers, who visited the same branch (same parent brand and same branch location) within a two-week window of each other, following the same script to request a small business loan from an employee. (Recall that the Black customer was an objectively stronger candidate for a loan.) When asked whether they felt encouraged or discouraged to apply for a loan, two testers indicated the following:

Black male customer:

“I was not specifically asked to apply. It [I] was unable to get comprehensive information to consider a loan without providing personal information to be verified. [Employee name] was friendly and was helpful, although the information provided was insufficient and limited due to bank policies regarding the verification process.”

White male customer:

“She was definitely interested in getting me a loan as soon as possible and made me feel confident about getting the ball rolling.”

At a different bank in the same city, two different testers (who also visited the same parent brand at the same specific branch location) to solicit loans similarly indicated:

Black male customer:

“The representative described himself is [as] unqualified to help me.”

White male customer:

“I felt encouraged because the representative stated that my details were good and with the length of time in business that would be a positive for me. They really acted like they wanted my business.”

These different experiences are noteworthy because the profiles given to the testers were very consistent, with the exception that Black testers had stronger financial profiles. (See Web Appendix C for additional exemplary responses from the testers.) Thus, Study 1 answers the call for more research to reveal “how discrimination happens” in financial services (Bartlett et al. 2022, p. 55). From a managerial perspective, it is crucial to note that, in line with the cues they take from their service experience, Black (vs. White) customers report lower loyalty intentions (recommend, repatronize) toward the bank. This is an undesirable outcome for the company fourfold: first, its employees discriminate against Black people; second, these people are likely to share their service experience with others, which can harm the firm's brand and reputation; third, the company loses viable business, which harms its revenue/profitability; and, fourth, the company is legally exposed due to employees’ discriminatory marketplace behaviors. All four are strong reasons for management to combat such racial discrimination. Thus, the next two field studies examine customer characteristics that help mitigate racial discrimination.

Study 2: Small Business Loan Application Experience as a Function of the Customer's Race and SES

Study 2 tests the moderating role of SES; we predict (H_2a) that negative effects experienced by Black business owners seeking loans are greater for those with lower (vs. higher) SES, but we expect that White business owners are relatively unaffected by their SES.

Design, Participants, and Procedure

To examine this question, we designed a 2 (customer race: White, Black) × 2 (customer SES: low, high) between-subjects field experiment. Building on and extending the procedure used by Bone, Christensen, and Williams (2014), fair lending testers inquired about small business loans with banks in the Washington, DC, metropolitan area over six weeks. Because these tests were conducted during the COVID-19 pandemic, they took place over the phone rather than in person.

In the 2 (customer race: White, Black) × 2 (customer SES: low, high) between-subjects design, there were 40 calls in each condition consisting of White and Black testers using scripts to represent either high-SES or low-SES small business owners, for a total of 160 test attempts. Testers were instructed to contact, by calling on the phone, 40 separate bank branches. There were 12 instances in which testers encountered branch locations of banks that were unreachable via phone because they had been closed during the COVID-19 pandemic. There were 11 instances in which after three phone attempts the testers were not able to contact any bank employees. In total, there were 137 tests across 34 separate bank branches representing 22 different financial institutions.

Testers were blind to the hypothesis, and four testers were used with each bank location (the study employed a between-subjects design, and testers never met each other). All testers in this study were men. Testers were trained to contact loan officers via phone to inquire about obtaining a small business loan, by following a script. To make our test more conservative, the Black testers had objectively stronger financial profiles relative to the White testers. Banks tested were randomly selected to fill quotas of large and small banks using the $10 billion asset cutoff point in the Federal Deposit Insurance Corporation survey of small and large small business lenders.

Manipulations

We manipulated perceptions of Black or White race based on the voice of the tester (the tester was a member of the corresponding racial group) and the name used (selected based on previous research on linguistic profiling [Baugh 2005]). ⁵ A pretest confirmed that listeners could accurately identify the race of the four testers (low-SES Black man 85% of the time; high-SES Black man 83%; low-SES White man 94%; high-SES White man 97%) at a rate that was greater than chance (ps < .001). We manipulated SES based on the linguistics (i.e., grammar and accent; e.g., “hello” vs. “hey there” and “you” vs. “y’all,” respectively for high and low SES). A pretest confirmed that the testers in the high (vs. low) SES groups were perceived differently (M_LowSES = 4.96, M_HighSES = 7.02; t = 16.63, p < .001). See Web Appendix F.

Measures of employees’ actual warmth and competence behaviors

We asked the testers to document the actual behaviors that took place during the interaction, and all testers were blind to the hypotheses and study purpose. We measured whether the bank employees demonstrated warmth behaviors toward the prospective customer (e.g., “Did they ask for your name?”; “Did they use your name?”; “Did they introduce themselves?”; “Did they ask how you are doing?”; “Did they ask how they can help you?”). We also measured whether the bank employee exhibited competence behaviors toward the prospective customer (i.e., based on actions that demonstrate knowledge and competency relating to the terms of the loan, including whether they were informed of rates, fees, costs, personal guarantees, collateral, loan duration, and other loan terms). These behavioral indicators were developed based on literature on warmth and competence in financial services (Scott, Mende, and Bolton 2013) and based on the industry experience of managers in our partner organization. Testers answered each question with no (coded as 0) or yes (coded as 1), and these were summed into the corresponding warmth (0–5) and competence (0–7) indices.

Overall quality of the service experience

We also asked testers about their perceptions of the service experience (i.e., “Overall, I’d say the quality of my interaction with the bank's employees is excellent”; “I would say that the quality of my interaction with the bank's employees is high”; “How thorough and detailed was the bank representative in explaining the business loans to you?”). These were measured on a ten-point scale (1 = “strongly disagree,” and 10 = “strongly agree”; α = .84).

Results

We conducted 2 (race: Black, White) × 2 (customer SES: low, high) ANOVAs (between subjects, full factorial) on the outcome variables (warmth, competence, overall service quality).

Employees’ warmth behaviors

An ANOVA on warmth behaviors revealed main effects of race (White = 3.45, Black = 2.85; F(1, 133) = 6.23, p = .014) and SES (Low = 2.57, High = 3.70; F(1, 133) = 21.39, p < .001), qualified by a significant interaction (F(1, 133) = 14.17, p < .001); see Figure 1, Panel A. For Black customers, those with lower SES experienced significantly fewer warmth behaviors from employees (Black-Low = 1.95, Black-High = 3.83; F(1, 133) = 38.84, p < .001); White customers did not experience differences in employee warmth behaviors based on SES (White-Low = 3.35, White-High = 3.55; F < 1). Furthermore, among low-SES customers, Black (vs. White) customers experienced significantly fewer warmth behaviors from employees (White-Low = 3.35, Black-Low = 1.95; F(1, 133) = 19.95, p < .001); high-SES customers were relatively unaffected as a function of race (White-High = 3.55, Black-High = 3.83; F < 1).

OPEN IN VIEWER

Figure 1.

Employee Behaviors and Customer Perceptions as a Function of Customer Race and Customer SES.

Employees’ competence behaviors

An ANOVA on competence behaviors revealed main effects of race (White = 1.85, Black = 1.24; F(1, 133) = 5.71, p = .018) and SES (Low = .93, High = 2.13, F(1, 133) = 22.11, p < .001), qualified by the significant interaction (F(1, 133) = 4.77, p = .031); see Figure 1. For Black customers, those with lower SES perceived significantly fewer employee competence behaviors (Black-Low = .44, Black-High = 2.11; F(1, 133) = 26.17, p < .001); White customers perceived only a marginally significant difference in employee competence behaviors based on SES (White-Low = 1.55, White-High = 2.16; F(1, 133) = 2.90, p = .09). Furthermore, low-SES Black (vs. White) customers assessed competence to be lower (White-Low = 1.55, Black-Low = .44; F(1, 133) = 10.65, p < .001); high SES customers were relatively unaffected as a function of race (White-High = 2.16, Black-High = 2.11; F < 1).

Overall quality of the service experience

An ANOVA revealed main effects of race (M_Black = 5.28, M_White = 7.07; F(1, 131) = 18.11, p < .001) and (M_Low = 5.39, M_High = 6.83; SES F(1, 131) = 10.68, p = .001), qualified by a marginally significant race × SES interaction (F(1, 131) = 3.76, p = .055); see Figure 1, Panel C. For Black customers, those with lower SES reported a significantly inferior experience (M_Black-Low = 4.21, M_Black-High = 6.38; F(1, 131) = 14.75, p < .001); White customers did not experience a difference as a function of SES (M_White-Low = 6.79, M_White-High = 7.35; F < 1). Furthermore, low-SES Black (vs. White) customers perceived the overall service quality to be lower (M_White-Low = 6.79, M_Black-Low = 4.21; F(1, 131) = 19.30, p < .001); high SES customers were relatively unaffected as a function of race (M_White-High = 7.35, M_Black-High = 6.38; F(1, 131) = 2.67, p = .11). (The degrees of freedom relative to warmth and competence vary slightly due to missing values.)

Moderated serial mediation analysis

Because prior research has shown that inferred warmth and competence of service providers drive customers’ downstream responses (Scott, Mende, and Bolton 2013), we examined whether warmth and competence influence the customer's overall service evaluation. We conducted a moderated mediation analysis (Hayes 2013; PROCESS Model 8); the independent variable was race, the moderator was SES, and the outcome variable was the customer's perceived quality of the service experience. Mediators were FLE warmth behaviors and FLE competence behaviors. The overall model of moderated serial mediation was significant (a × b = .3357; 95% CI: [.0127, 1.2944]). For Black customers, both warmth and competence behaviors mediate (warmth: a × b = .7277, 95% CI: [.1470, 1.5412]; competence: a × b = .8105, 95% CI: [.2154, 1.5696]). The mediation was not significant for White customers (warmth: a × b = .0792, 95% CI: [−.2121, .3901]; competence: a × b = .3008, 95% CI: [−.0849, .7585]).

Discussion

Consistent with Study 1, Study 2 also demonstrates that Black customers have a worse service experience than White customers (H₁), driven (mediated) by lower levels of customer-perceived warmth and competence of FLEs. The mediating role of inferred warmth and competence is conceptually consistent with research on service quality; for example, the lack of perceived warmth and competence in our studies captures violations of service quality dimensions such as “empathy” (e.g., the employee gives you individual/personal attention, has your best interest at heart, understands your specific needs), “assurance” (e.g., the employee is courteous, instills confidence, makes you feel safe), and “responsiveness” (e.g., the employee is willing to help you) (Zeithaml et al. 2024). Furthermore, the results of Study 2 show that for Black business owners, those with lower (vs. higher) SES experience significantly less favorable treatment from service employees (e.g., loan officers), whereas the treatment received by White business owners is unaffected by SES, supporting H_2a. Furthermore, low-SES White customers experience significantly more favorable treatment than low-SES Black customers.

These results suggest that FLEs respond differently to Black customers when these customers provide cues that are inconsistent with stereotypes about Black people, in this case, having higher SES. We next turn to another type of customer cue: business structure, which may signal status in the form of social capital or business acumen and sophistication.

Study 3: Small Business Loan Application Experience as a Function of the Prospective Customer's Race and Business Structure

Study 3 tests the moderating role of legal business structure on the relationship between customer race and progress toward a small business loan (H_2b). Study 3 operationalizes sole proprietorship as less complex and sophisticated and distinguishes this business structure from more sophisticated forms, in terms of social capital sophistication (i.e., joint proprietorship, partnership), and business-legal sophistication (i.e., LLC, S corporation, C corporation).

Through their business structure, small businesses can send cues about their level of sophistication. Specifically, there are, at least, two areas of relevant consideration in terms of relatively higher sophistication: social capital sophistication and business-legal sophistication.

Social Capital Sophistication

In this field study, we examine business structures that, relative to sole proprietorships, have a greater number of owners (i.e., partnerships and joint proprietorships). Similar to a sole proprietorship, the legal requirements of partnerships and joint proprietorships are relatively low in complexity. However, because they involve two or more owners, these business structures may signal to the bank that the owners are not alone and bring social capital to the transaction, which should make them more desirable customers for potential lenders because repayment responsibility is shouldered by more than one owner (Ghatak 2000).

Business-Legal Sophistication

In Study 2, we found that the negative relationship between customer race and employee behaviors was mitigated by SES, such that Black (vs. White) customers received poorer treatment, but the employees’ behavior improved when the Black customer signaled higher SES. An entrepreneur's business structure that requires advanced understanding of additional legal and tax considerations may suggest relatively greater business acumen, and it may therefore function in a manner similar to higher SES. That is, utilizing a more complex business structure (e.g., an LLC or corporation) may provide a cue that the entrepreneur possesses greater levels of business and legal education. In this study, we operationalize LLCs, S corporations, and C corporations as business structures that require greater understanding of legal and financial infrastructures.

A pretest confirmed that the business structure categorizations significantly differ in perceived complexity and sophistication (see Web Appendix G). We predict a moderating role of the customer's business structure such that the negative effects experienced by Black small business owners seeking loans are greater for those with less (vs. more) sophisticated business structures; in contrast, we expect White business owners to be relatively unaffected (H_2b).

Design, Participants, and Procedure

We administered a survey to small business owners that employed 500 people or less. Data were collected over a two-month period during the first year of the COVID-19 pandemic, and after the first two phases of the Paycheck Protection Program (PPP) had been administered. ⁶ This survey provided first-person insight into the pandemic experience of business owners and whether they reached out to financial institutions for help. We recruited business owners using panel services and Chamber of Commerce groups. We solicited responses from 938 business owner participants, who were asked about their debts and liabilities at the onset of the COVID-19 pandemic and how they managed their debts and need for financial services during the pandemic. Specifically, participants were asked if they had inquired about a loan during the pandemic (from March 1 through December 1, 2020). Of the small business owners who met these criteria, 153 were Black and 113 were White. ⁷ Participants were recruited to participate from nine metropolitan statistical areas distributed across the country: Washington, DC; Los Angeles, California; Atlanta, Georgia; New York City, New York; Chicago, Illinois; Houston, Texas; Miami, Florida; Detroit, Michigan; and Seattle, Washington. This geographic distribution allows for a deeper understanding of the U.S. small business lending marketplace. The average age of the respondents was 38 years (93 women, 173 men). Businesses included in the sample spanned various North American Industry Classification System codes. Industries most represented in the sample include retail trade (17.4%), construction (9.6%), and information technology (9.5%).

Participants reported whether they had inquired about and received a loan product designed to help small businesses during the pandemic. We examined four different loan products, including the PPP, Economic Injury Disaster Loans, SBA debt relief, and SBA Express Bridge Loans. We computed a binary score of whether the participant had been approved for one of these loans (not approved = 0, approved = 1). Participants reported their identified race or ethnicity, along with other business and personal demographic variables, and they indicated legal ownership structure of their business (sole proprietorship; joint proprietorship, partnership; LLC, S corporation, or C corporation). We grouped the business structures based on their sophistication: simple (i.e., sole proprietorship), social capital sophistication (i.e., joint proprietorship, partnership), and business-legal sophistication (i.e., LLC, S corporation, C corporation).

Results

We analyzed the data with a binary logistic regression analysis, as a function of race (White = 0, Black = 1), business structure (simple: sole proprietorship; social capital sophistication: joint proprietorship, partnership; business-legal sophistication: LLC, S corporation, C corporation; dummy coded), and their higher-order two-way interaction, with loan approval (not approved = 0, approved = 1) as the binary dependent variable.

The overall model was significant (likelihood ratio χ²(df(5) = 30.92, p < .0001). The model revealed a significant effect of race (z = −3.09, p = .002), and two significant interactions (Interaction 1: race × W1; z = 2.96, p = .003) and (Interaction 2: race × W2; z = 4.48, p < .001). The other effects were nonsignificant; W1 (sole proprietorship vs. joint proprietorship/partnership; z = −1.01, p = .31) and W2 (sole proprietorship vs. LLC/corporations; z = −1.60, p = .11); see Figure 2.

OPEN IN VIEWER

Figure 2.

Loan Approvals as a Function of Customer Race and Business Structure.

Following up on the significant interactions, in terms of loan approvals, Black customers with more (vs. less) sophisticated business structures received significantly greater approvals; in contrast, there was no difference in loan approvals for White applicants based on their business structure. (Results are consistent when control variables are included; see Web Appendix E.)

Specifically, contrasts reveal that for sole proprietorships, Black (vs. White) entrepreneurs received significantly fewer loan approvals (M_W-SP = 60.00%, M_B-SP = 25.71%; p = .002). Yet, this negative effect is attenuated when the business structure signals greater social capital sophistication (M_W-JP/Part = 48.15%, M_B-JP/Part = 63.64%; p = .23) or greater business-legal sophistication (M_W-LLC/Corp = 41.94%, M_B-LLC/Corp = 75.29%; p = .001).

Looked at another way, Black entrepreneurs receive fewer loan approvals when their business structure is less sophisticated (M_SP = 25.71 vs. M_JP/Part = 63.64, z = 3.06, p = .002; M_SP = 25.71 vs. M_JP/Part = 75.29, z = 4.72, p < .001). In contrast, loan approvals for White entrepreneurs are relatively unaffected by the structure of their business (M_SP = 60.00 vs. M_JP/Part 48.15, z = −1.01, p = .31; M_SP = 60.00 vs. M_LLC/Corp = 41.94, z = −1.60, p = .11), supporting H_2b.

Discussion

In this study, we surveyed entrepreneurs who had sought loans during the COVID-19 pandemic. We found that among sole proprietorships (the predominant (>80%) small business structure in the United States), White entrepreneurs were approved for loans at more than double the rate of Black entrepreneurs (60.00% vs. 25.71%). However, this negative effect is attenuated when Black entrepreneurs’ business structures represent greater social capital sophistication and business-legal sophistication. Taken together with findings of Study 2, which found that higher levels of SES similarly attenuated the negative effects of racial bias, these results suggest that the additional sophistication that may be associated with SES (e.g., via higher education) may also be conveyed with more sophisticated business structures (linked to social capital and business-legal profile). These cues, higher SES and complex business structures, may signal to lenders that the focal Black customer does not fit the stereotype of Black entrepreneurs (as risky borrowers). If this is the case, then this points to policies and solutions that firms (and policy makers) can use to overcome the bias demonstrated in bank employees’ behavior. In further reflecting the preceding finding that loan applications from Black (vs. White) entrepreneurs with a sophisticated business structure were approved to a higher extent, we note insights from studies on helping behavior in the context of weight discrimination. Randall et al. (2017, p. 125) show that “discrimination overweight individuals experience extends to situations when they are asking for help” but also that “displaying stereotype-inconsistent cues benefit overweight individuals by increasing the likelihood of them being helped.” While we do not have process-related evidence from the employees who approved these loans, it is possible that loan officers engage in “helping behavior” in light of stereotype-inconsistent cues.

We also raise a final important point of clarification: Although on the surface, the main effect of race appears inconsistent with prior literature, this is merely a function of the design of this study. In the United States, more than 80% of all small businesses are sole proprietorships; however, in our study, sole proprietorships comprise only 33.8% of the sample because we were interested in studying a broad sample of business structures. The contrast within sole proprietorships is consistent with findings in prior literature and our other studies, showing that loan approvals are significantly lower for Black (vs. White) business owners.

Study 4: Exploring the Financial Service Providers’ Perspective

The purpose of this study is to examine the moderating role of business structure on racial bias in financial services through a different vantage point: financial service professionals making the decision about the loan. That is, in this study, all participants had professional experience in financial services and were asked to evaluate a small business loan application that varies based on the race of the applicant and the business structure of the firm. This study also examines the underlying processes that could influence financial service professionals’ evaluations, including inferences about riskiness, trustworthiness, and default likelihood of the customer; thus, Study 4 goes to the heart of stereotypes about Black people’s financial behavior and risk level as customers (Austin 2004).

Design, Participants, and Procedure

The study employed a 2 (customer race: Black, White) × 2 (business structure sophistication: sole proprietorship, S corporation) between-subjects design. The participants (N = 588; M_age = 37.77 years; 252 women, 336 men; Amazon Mechanical Turk and Prolific) had professional experience in financial services. Participants evaluated a small business loan application, which was designed based on real applications in the marketplace. The loan application manipulated race and business sophistication and presented the firms’ financial performance (see Web Appendix I for stimuli).

Race manipulation

Participants in the Black applicant condition saw an application submitted by “Mr. DeShawn Washington,” and those in the White applicant condition saw an application submitted by “Mr. Brad Anderson” (names adopted from March, Gaertner, and Olson 2021; Milkman, Akinola, and Chugh 2012). We further manipulated race via the “How did you hear about us?” section of the application; the response indicated “African American Chamber of Commerce” or “Chamber of Commerce,” according to the condition (Rattan, Steele, and Ambady 2019). The application forms showed that company performance was 10% higher for the Black (vs. White) applications, to offer an objectively more conservative test. A pretest confirmed that the race manipulation performed as intended; condition significantly predicted the corresponding intended race (χ²(1) = 58.60, p < .001; please see Web Appendix J).

Business structure sophistication manipulation

We manipulated business structure by indicating it on the manila folder that was shown to include the application form and on the loan application itself. We used sole proprietorship to represent a business structure with low sophistication and S corporation to represent a business structure with high sophistication.

Measures

After reviewing the loan application, participants indicated the type of loan, if any, that should be offered (i.e., BLOC, HELOC, no loan, or other). Participants indicated the perceived trustworthiness of the applicant with a ten-item index (Morgan and Hunt 1994; Tax, Brown, and Chandrashekaran 1998), default likelihood (adopted from Gonzales Martinez et al. 2020), and riskiness of the loan application (Kim and Lakshmanan 2021); see Web Appendix K for the items. Participants also responded to a social desirability bias measure (Reynolds 1982), demographic items, and a business structure manipulation check. The manipulation check showed that participants correctly recalled the business structure 99% of the time (z = 23.75, p < .001).

Results

We conducted analyses as a function of the applicant's race, business structure, and their higher-order interaction. We controlled for the participants’ age, race, gender, financial literacy (Lusardi and Mitchell 2011), social desirability bias, and the source of the data.

BLOC

The regression model included applicant's race as the independent variable, business structure as the moderator, and BLOC as the dependent variable, ⁸ controlling for participants’ age, race, gender, financial literacy, source, and social desirability bias. Results revealed the predicted significant main effect of race (b = −.70, z = −1.97, p = .049) and a significant two-way interaction (b = 1.16, z = 2.20, p = .028); see Figure 3. The main effect of business structure was not significant (b = −.22, z = −.59, p = .55).

OPEN IN VIEWER

Figure 3.

Likelihood of Offering a BLOC Loan as a Function of Applicant Race and Applicant Business Structure.

For Black applicants, the likelihood of being offered a BLOC increased with a more sophisticated business structure (Prob_{Black-SoleProp} = 82.82%, Prob_Black-SCorp = 92.51%; b = .94, z = 2.53, p = .01). White applicants were relatively unaffected by business structure type (Prob_{White-SoleProp} = 90.66%, Prob_White-SCorp = 88.61%; b = −.22, z = −.59, p = .55). Looked at another way, for sole proprietorships, White (vs. Black) applicants are more likely to be offered a BLOC (Prob_{White-SoleProp} = 90.66%, Prob_{Black-SoleProp} = 82.82%; b = −.70, z = −1.97, p = .049); for S corporations there was no significant difference (Prob_White-SCorp = 88.61%; Prob_Black-SCorp = 92.51%; b = .46, z = 1.19, p = .23).

Process underlying BLOC recommendation

We conducted a test of moderated serial mediation (PROCESS Model 85) in which business structure was the independent variable; race was the moderator; riskiness, trustworthiness, and default likelihood were the mediators; and BLOC was the outcome variable, controlling for the participants’ age, race, gender, and financial literacy; the sample source; and social desirability bias. The model did not reveal effects of riskiness as a mediator. However, the model revealed an overall model of moderated mediation for the path of business structure → trustworthiness → default likelihood → BLOC (a × b = .0531; 90% CI: [.0021, .1233]). For Black applicants with a corporation (vs. sole proprietorship), increased trust leads to a decrease in anticipated default likelihood, to an increased likelihood to offer a BLOC (a × b = .0525; 90% CI: [.0135, .1085]). The mediational path does not emerge for White applicants (a × b = −.0006, 90% CI: [−.0395, .0383]); see Web Appendix L for further details.

Discussion

Through the lens of financial service professionals, this study demonstrates differential effects of a customer's race and business structure on recommending a BLOC. That is, Black applicants have a greater likelihood of being offered a BLOC when their business structure is sophisticated (Black_{sole proprietorship}: 82.82%, Black_corporation: 92.51%); yet, the likelihood of White applicants being offered a BLOC is relatively unaffected by their business structure. For sole proprietorships, White (vs. Black) applicants have a greater likelihood of being offered a BLOC (White_{sole proprietorship}: 90.66%, Black_{sole proprietorship}: 82.82%), supporting H₁; the effect is attenuated when Black applicants have a more sophisticated business structure. This is consistent with Study 1, which showed that among testers visiting banks, a BLOC was recommended to White customers 40.43% of the time (vs. 17.35% for Black customers). Similar to Study 1, this study showed no effects of race or business structure on being offered a HELOC. In other words, race, moderated by business structure, predicts whether a person is offered a BLOC but not a HELOC.

The findings from this study also align with Study 3 (i.e., sophisticated business structure helps attenuate racial bias toward Black customers). However, the results are not identical to those of Study 3. First, recall that Study 3 focuses on actual approvals of small business COVID-19 relief loans, whereas Study 4 focused on BLOC offers. Second, although we made the scenario as realistic as possible (e.g., by adapting actual loan applications, recruiting participants who understood different loans and business structures), the effects were much stronger in the real world (Study 3) than in the hypothetical scenario (Study 4). Third, as in Study 1, this study employed a conservative test in which Black (vs. White) applicants’ financials are 10% stronger.

The design of this study also helps address a shortcoming of Study 2, which had only one tester per condition. In addition, this study provides insight into the process underlying the effects. We find a serial mediation path via trust and perceived default likelihood. That is, for Black applicants, when business structure is more sophisticated, increased trust leads to a decrease in anticipated default likelihood, and, in turn, to an increased likelihood to offer a BLOC; this process does not emerge for White applicants as a function of business structure.

General Discussion

This research provides novel insights regarding how, when, and why discrimination occurs toward Black (vs. White) customers; it also discovers factors that mitigate discrimination in financial loan settings (SES and business structure). These results enrich marketing theory and practice; inform customers, managers, and policy makers; and offer avenues for further research.

Theoretical Implications

Service quality and customer satisfaction as drivers of firm performance have been studied extensively in marketing (Otto, Szymanski, and Varadarajan 2020), and companies spend millions to ensure positive customer service journeys (Lemon and Verhoef 2016). Given such deep knowledge on excellent service, our data reveal a sobering reality: systemic racial bias at service frontlines undermines positive customer experiences. In fact, the interplay of two forms of discrimination seems to affect Black customers’ service experiences. Research in economics distinguishes taste-based discrimination (i.e., due to discriminators’ affective tastes against particular social groups) from statistical discrimination (due to discriminators’ expectations derived from previous interactions or external sources such as social networks, official statistics, or media) (Becker 1957; Schwab 1986). While it is difficult to demonstrate taste-based discrimination in the marketplace (Gabbidon and Higgins 2020), our findings seem to provide indirect evidence for it: note that, because the Black (vs. White) customers in Studies 1, 2, and 4 had objectively stronger financials, they should have elicited more positive responses from the lenders, as they were more valuable customers. Not only did that not happen, but instead, Black customers were penalized in terms of loan recommendations and approvals. In addition, Study 3 suggests that the reason for lower loan approvals is not (purely) “statistical” in nature, because there was no difference in the FICO scores between Black (vs. White) loan applicants (and no interaction with business structure); yet, Black applicants’ loans were approved to a lower degree. However, Study 4 indicates that statistical discrimination plays a role, because we find a serial mediation such that recommendations for a more favorable loan to Black customers with a sophisticated structure (vs. sole proprietorship) is mediated by increased trust and decreased anticipated default. Taken together, these insights identify the need for more research that describes, explains, and quantifies discrimination, and the need for work on mitigating the negative effects on vulnerable minority customers and company performance due to employee racial biases. The insights about types of discrimination also suggest a need for a broader scope of testing the mechanisms that affect why employees discriminate against vulnerable groups (e.g., besides racial biases, there might be biases linked to weight, age, or disabilities; Gneezy, List, and Price 2012).

The results also indicate the need for a reenergized focus on theorizing impression formation in service settings. For example, studies might treat race as a control variable rather than as a meaningful theoretical variable (Grier, Thomas, and Johnson 2019). Accordingly, marketing theory needs to be more nuanced in explaining the extent to which biases are implicit or explicit, as these can have different causes and mechanisms, and require different theory-based interventions. In addition, new theoretical lenses are needed. For example, in our context of financial loans, expectation states theory (Berger et al. 2018) might shed new light on the mechanisms that influence employees as they engage in discrimination. According to expectation states theory, the status of a person influences the resource allocations others make related to that person, and they tend to allocate more resources to a person of higher (vs. lower) status (Berger et al. 2018; Harkness 2016). This insight is consistent with our findings on the mitigating role of high SES and sophisticated business structures. This theoretical lens also suggests that additional interventions might elevate the status of minority consumers to help protect them from bias.

Managerial and Policy Implications

In deriving practical implications, we conducted exploratory interviews with financial service executives to gain their reactions to our findings (the interviewees included two vice presidents [VPs] from regional banks, a senior vice president [SVP] in a superregional bank, and an SVP in a global bank; all had experience in lending). We asked the executives which implications they would recommend if our results were true in their own bank. Insights from these interviews inform the proposed implications discussed subsequently (also see the Appendix, which presents illustrative quotes from the interviews).

Standardized processes and service products

Our studies show that service employees might present product portfolios in a racially biased manner to customers (e.g., Black [vs. White] customers are significantly less likely to be offered more desirable options [i.e., BLOCs]). This effect persisted after controlling for other factors such as race and gender match (see Web Appendix E). Bank employees have substantial discretion regarding which products (not) to offer (Bone and Mowen 2010). Thus, firms may need to develop policies to ensure that customer-favorable product options are uniformly offered to all customers (e.g., via checklists that employees have to follow throughout the service process). Another idea is that (at least) two employees have to independently evaluate the focal loan application. Furthermore (as described in the next section), firms can take specific actions such as making available special-purpose credit products for vulnerable customers that have experienced discrimination, to help ensure that they have full access to and are receiving desirable product options.

SES

We found that bank employees exhibit racially biased warmth and competence behaviors toward Black customers. Of note, regardless of their lower or higher SES, White customers were treated more favorably than low-SES Black customers. All of that was true although the Black (vs. White) prospective customers in our studies had objectively stronger financials (i.e., more valuable customer profiles). This racial bias was only mitigated when Black customers signaled higher SES. These findings suggest that (1) employees use SES cues to differentiate Black customers, but they do not use SES cues for White customers, and (2) they might place more value on SES cues than on objective customer profiles, to the economic detriment of their employer. Thus, firms can benefit from enacting policies, reward structures, and trainings (see Web Appendix M) that emphasize the use of objective financial data over subjective judgments about a customer, and build awareness regarding racial bias.

Business structure

Study 3 included a survey of small business owners with varied business structures, who were seeking loans during the pandemic. In the United States, more than 80% of small businesses are sole proprietorships, which is the most basic business structure. Among sole proprietorships, we found that White (vs. Black) business owners were approved twice as often for loans (60% vs. 26%). But, when Black applicants have more sophisticated business structures (e.g., corporations), they will benefit from signaling this fact early in the process with a lender.

Actions firms can take to mitigate discrimination

Our findings point to actionable implications for service firms to mitigate discrimination (see Web Appendix M). These actions range from increasing internal compliance with legal frameworks (e.g., the Equal Credit Opportunity Act), deliberately designing inclusive products (e.g., special-purpose credit products), training employees on fair lending legal compliance, and using technology to reduce bias (e.g., self-service technology so that Black customers have the option of avoiding face-to-face assessments).

Implications for small business owners

Certain cues increase the likelihood of success in accessing credit for customers who might be discriminated against. For example, Black small business owners should signal their sophisticated business structure early and explicitly in the loan application process, take advantage of technical assistance for entrepreneurs (e.g., CPA services), and seek to build a banking relationship early in the entrepreneurship process (e.g., with a community development financial institution) (see the Appendix).

Policy implications

Our results point to policy tools to reduce discrimination. These tools include measures such as creating standardized small business lending forms, funding programs that provide technical assistance and education to minority-owned businesses, and increased oversight and enforcement. As the Appendix illustrates, public policy making is a direct and necessary tool to mitigate racial discrimination in the marketplace.

Limitations and Future Research

Limitations of our work provide opportunities for future research. We test the moderating role of SES via the customer's language and name, consistent with prior work that used a variety of SES indicators, including targets’ body language, nonverbal cues (e.g., self-grooming, fidgeting) (Kraus and Keltner 2009), and certain names (Figlio 2005; Gaddis 2017; Landgrave and Weller 2022). However, future research can examine other indicators for SES beyond our approach of using customer names and language; similarly, we expect that there are factors beyond business structure that might help shield Black entrepreneurs from discrimination. We also note that SES is not always highly predictive in explaining the income gap between Black and White people (Assari 2018; Chetty et al. 2020). Thus, understanding when and why SES is predictive in the context of inequality is valuable (e.g., whether SES is measured via inferred [subjective] indicators or via objective indicators might be relevant). That said, our research responds to calls in marketing to better disentangle the distinct experiences of Black consumers from the experiences of other racial (e.g., White) and ethnic (e.g., Hispanic) groups (Grier, Johnson, and Scott 2021). Notably, in Study 3, our sample included Hispanic entrepreneurs (in addition to Black and White entrepreneurs). However, following the U.S. Census approach, participants selected either a racial category (e.g., Black, White) or an ethnicity category (e.g., Hispanic). Prior work found similarities between Black and Hispanic customers in financial services, based on these groups’ shared status as minorities in the United States (e.g., Bone, Christensen, and Williams 2014). The expanded analysis from Study 3 (Web Appendix H) shows that Hispanic entrepreneurs have some experiences that are consistent with experiences of White entrepreneurs and others that are consistent with those of Black entrepreneurs. This may be due to the fact that in addition to their ethnic identity, Hispanic people may also signal their identification with one or more racial groups (e.g., White, Black, both, neither). Alternatively, this may relate to research on colorism, which examines people’s physical features as residing on a continuum from more Afrocentric to more Eurocentric features; people experience preferential treatment as they are perceived to be less Afrocentric (Dixon and Telles 2017; Hunter 2007). We also note that historically in the United States, consumers of Asian descent have not experienced racial discrimination in financial services to the extent that Black customers have. This may be due in part to the, also harmful, “model minority” stereotype (Kim, Lu, and Stanton 2021). In addition, individuals of Asian descent have recently been the subject of discrimination due to hostility relating to perceptions about the origins of the COVID-19 pandemic (Cheng et al. 2021). Future research can examine how Asian customers’ experiences in financial services have evolved and how they can be improved. Finally, although Study 4 examines the perspective of financial service providers, another important extension of our work is to examine dyadic perspectives in the field (i.e., the perspective of both the matched customer and the matched service employee).

Supplemental Material

sj-pdf-1-mrj-10.1177_00222437231176470 - Supplemental material for Revealing and Mitigating Racial Bias and Discrimination in Financial Services

Supplemental material, sj-pdf-1-mrj-10.1177_00222437231176470 for Revealing and Mitigating Racial Bias and Discrimination in Financial Services by Maura L. Scott, Sterling A. Bone, Glenn L. Christensen, Anneliese Lederer, Martin Mende, Brandon G. Christensen and Marina Cozac in Journal of Marketing Research