Racial disparities have been widely documented in U.S. consumer credit markets, prompting researchers to wonder: To what extent are evaluators of loan applications treating otherwise identical people differently? To date, the absence of much quality data on a broad scale has hindered the ability of researchers to find answers to this and related questions.

In their paper, “Using High-Frequency Evaluations to Estimate Discrimination: Evidence from Mortgage Loan Officers,” Marco Giacoletti, Rawley Heimer, and Edison Yu test for discrimination using a novel approach that sheds light on decision makers’ subjective assessments and personal biases. Using this approach, they analyze the monthly mortgage application decisions of loan officers, who typically have substantial discretion in granting loans.1 Specifically, they test for taste-based discrimination — that is, they test for whether minority groups receive disparate treatment because mortgage lenders are biased against them.2

The authors consider that evaluators of all kinds can make very different decisions over a short time period. For example: Transportation Security Administration agents might limit the screening of airline passengers when lines are long or when they are at the end of their shifts; and police officers working under a monthly quota system may be more apt to ticket drivers exceeding the speed limit toward the end of the month. The authors test to see if this same concept applies to the mortgage-lending market — specifically, they examine whether loan officers are incentivized by monthly performance targets to expand the pool of people they are willing to lend to at month-end.

The researchers rely on high-frequency, time-stamped Home Mortgage Disclosure Act (HMDA) data that include nearly all U.S. mortgage applications from 1994 to 2018, comprising 500 million loan applications across 28,000 lenders. Crucially, they are able to observe the exact application and decision dates for all mortgage applications. By focusing on whether the approval of loans varied over the course of the month, they can isolate the effect of loan officers’ decision-making independent of other factors. These other factors, which include credit market conditions, applicants’ characteristics, and firm-level attributes, generally take far longer than a month to change.

To uncover the extent of taste-based discrimination, they search the data for an “end-of-month effect” among loan officers eager to meet performance goals. Next, they test for discrimination by comparing the percentage of loans approved for Black applicants at the beginning of the month with the percentage approved at the end of the month. Finally, they determine if the level of discrimination differs depending on three variables related to market structure/type: (i) among fintech lenders that lead in industry innovation; (ii) with respect to mortgage bank concentration at the county level; and (iii) among shadow banks, which have lower regulatory requirements, allowing them to have a larger presence in underserved communities.3

Using monthly data over a 25-year period, the authors observe that 150 percent more mortgages were approved on the last day of the month than on the first day of the month. This pattern is consistent with the fact that individual loan officers typically have monthly performance targets tied to compensation. The uneven awarding of mortgage approvals occurred even though the number of mortgage applications submitted over the course of a month stayed constant.

In addition, the authors find striking evidence that the mortgage loan approval gap between Blacks and whites shrinks over the course of the month. This occurs even though the racial composition of applicants and the quality of the applications are essentially constant over the month. They find that Black applicants are about 7 percent less likely than white applicants to be approved on the first day of the month, but only 3.5 percent less likely on the last day of the month. They find no evidence that this gap is the result of market conditions or applicant attributes such as income, loan amount, and type of loan.

Their findings suggest that mortgage loan officers become more willing to lend to Black applicants toward the end of the month when the officers need to meet their monthly performance targets. Because loan officers have the same application information at the start of the month as at the end of the month, these results point to taste-based discrimination based on subjective preferences.

Moreover, the authors find that the fintech lending model — which by design is supposed to reduce subjective decision-making — did not affect the within-month loan approval gap between Blacks and whites. Also, the degree of concentration in local lending markets did not affect the extent of discrimination. They do find that for shadow banks, both the month-end effects and the approval gap between whites and Blacks are less pronounced, though still present. The authors suggest that fintech lenders, while operating largely online, still leave room for loan officer subjectivity, while shadow banks have presumably reduced discrimination by operating more in underserved communities.

Their results have important policy implications for the distribution of credit. Federal legislation, notably the Equal Credit Opportunity Act (1974) and the Community Reinvestment Act (1977), have been enacted to address inequities (such as redlining4). Such legislation is designed to alter the behavior within lending institutions that results in discrimination. The authors show that antidiscrimination policies targeted at these institutions will have a minimal effect as long as individual loan officers have discretion in lending decisions.

Further research efforts to disaggregate the data down to the individual loan-officer level would help to further clarify the extent of discrimination. This more granular data could particularly benefit minority consumers as they shop for a mortgage lender.

  1. Joseph Engelberg, Pengjie Gao, and Christopher A. Parsons, “Friends with Money,” Journal of Financial Economics, 103:1 (2012), pp. 169–188.
  2. Another broad category of discrimination is statistical discrimination, which occurs when lenders use traits that correlate with racial groups as a whole to form beliefs about specific applicants in that racial group.
  3. Andreas Fuster, Matthew Plosser, Philipp Schnabl, and James Vickery, “The Role of Technology in Mortgage Lending,” Review of Financial Studies, 32:5 (2019), pp. 1854–1899.
  4. Ian Appel and  Jordan Nickerson, “Pockets of Poverty: The Long-Term Effects of Redlining,” Boston College, Carroll School of Management Working Paper (2016).