Lender-reported reasons for mortgage denials don’t explain racial disparities

Lender-reported denial reasons conflict with some common hypotheses about racial disparities in conventional mortgage approvals. These findings could inform next steps for lenders looking to understand and potentially close such racial disparities.

The Federal Reserve has privileged access to confidential Home Mortgage Disclosure Act (HMDA) data, which include credit scores and other financial information from millions of mortgage applications. An analysis of a confidential HMDA dataset our team conducted previously, along with others’ analysis of the same data, found that an applicant of color is more likely to have their application denied¹ than a White applicant with the same income and credit score who applies for a conventional mortgage of the same size for a similar home.

After releasing our analysis, we interviewed mortgage lenders and underwriters for their perspectives on our findings. Most offered hypotheses for the racial disparities that were variations on two common themes. First, our interviewees pointed out that employment and credit histories, not just current incomes and credit scores, are important in lenders’ decisions and may vary by race. Second, our interviewees suggested that mortgage officers might not provide the level of service that applicants of color sometimes need, resulting in more denials for procedural reasons.

HMDA data don’t include credit histories or measures that tell us about applicants’ experience with loan officers. However, under HMDA, lenders must report the reasons they denied loan applications—and these reported denial reasons allow for some exploration of the hypotheses our interviewees offered. We recently examined reported denial reasons listed in our confidential HMDA dataset, and our findings show that even after accounting for racial differences in applicant and property characteristics, the reasons lenders give for denying mortgages to people of color differ from the reasons they give for denying mortgages to White applicants. In other words, the rate at which a lender gives a specific denial reason like “excessive debt-to-income ratio” for a given racial group cannot be fully accounted for by the actual debt-to-income ratios we observe in the data.

Data challenge common narratives about denial reasons

Under HMDA, lenders select from eight standard reasons for denying an application. The eight reasons cite debt-to-income ratio, employment history, credit history, collateral, insufficient cash (for the down payment or closing costs), unverifiable information, incomplete credit application, and mortgage insurance denial. There’s also an option for lenders to mark “Other” and provide a custom description.² Figure 1 depicts the prevalence of these reasons in our sample.

Just as overall denial rates vary greatly by race, the prevalence of denial reasons varies by race and ethnicity.³ As Figure 2 shows, the top reason for Asian applicants is incomplete credit application, reported on 24.1 percent of denied applications, followed by unverifiable information at 18.0 percent. For Black applicants, the top reason is credit history at 23.2 percent, followed by insufficient collateral at 19.4 percent. And for Latino and White applicants, the top reason is insufficient collateral, at 21.7 percent and 25.0 percent, respectively.

Some differences in the denial reasons across racial groups are to be expected. For example, if one racial group has more applicants with a high debt-to-income ratio, we might expect debt-to-income to be a more common denial reason reported for this group. To adjust for these differences in application characteristics, we developed a regression model using variables from our previous analysis. We examined racial and ethnic differences in listed denial reasons to see if differences in denied applicants’ incomes, credit scores, debt-to-income ratios, and other key factors can account for them. This approach allowed us to analyze whether lender-reported denial reasons vary by race for denied applicants whose applications are otherwise similar. The results, detailed below, challenged some of the narratives we heard from lenders about racial disparities.

Insufficient collateral (22.8 percent of denials)

The collateral for a mortgage is the property itself. Collateral denials can occur when property appraisals are lower than expected, or when home inspections reveal significant physical issues. Based on the reasons lenders report under HMDA, White applicants are more likely than Asian, Black, or Latino applicants to be denied for reasons related to collateral.⁴ As shown in Figure 3, after adjusting for the application characteristics, among denied applications, Latino applicants are 5.6 percent less likely to be denied for insufficient collateral, Black applicants 14.2 percent less likely, and Asian applicants 22.8 percent less likely.

These findings don’t contradict evidence of bias in assessments or appraisals—but based on stated denial reasons, these biases are unlikely to be the primary drivers of conventional mortgage approval disparities.

Incomplete application (21.2 percent of denials) or unverifiable information (12.4 percent of denials)

Mortgage lenders require extensive documentation of prospective borrowers’ financial histories, including their employment histories and credit histories. If an application lacks all the required information and isn’t withdrawn or closed before reaching the point in the process where the lender must approve or deny it,⁵ then the lender would deny the application and report “Incomplete credit application” as a denial reason.

After applicants supply information, lenders must verify it. The “Unverifiable information” denial reason suggests the lender can’t confirm everything on an application. A denial for unverifiable information can also apply when a lender finds information that an applicant hasn’t explained—like undisclosed loans or large account transfers.

If our interviewees’ theories were true, applicants of color would receive less support than they need from lenders as they gather or fill out the many documents and forms necessary for their applications. As a result, lenders might report “Incomplete credit application” or “Unverifiable information” as a more frequent reason for denial among people of color.

As the adjusted numbers in Figure 4 show, among denied applications, Black, Latino, and Asian applicants are less likely than or about as likely as similar White applicants to have incomplete credit application listed as a reason for denial. According to the reasons lenders report under HMDA, Black denied applicants are only slightly more likely to have unverifiable information in their application than similar White denied applicants. However, both Asian and Latino denied applicants are about 36 percent more likely to be denied for loans because of unverifiable information compared to similar denied White applicants.⁶

“Other” (15.5 percent of denials)

In cases where the eight HMDA denial reasons don’t apply, lenders can choose “Other” as a denial reason and provide a brief explanation. As Figure 5 shows, among denied applications, those submitted by applicants of color are more likely to have “Other” as a cited reason compared to those submitted by White applicants.⁷ After adjusting for the application characteristics available in HMDA data, Black applicants are 10.0 percent more likely to have “Other” as a cited reason for denial relative to White applicants, Asian applicants 15.8 percent more likely, and Latino applicants 21.1 percent more likely.

Applicants denied for “other” reasons are overrepresented among applicants with multiple denial reasons. About 15 percent of all denials have at least two reasons listed; almost one-third of denials due to “other” reasons cite at least one other denial reason.

We investigate what terms are most commonly used in the explanations across all denied applications where “Other” is listed as the reason or one of the reasons. We find that terms related to program eligibility, such as “program,” “product,” and “ineligible,” to name a few, are among the most-used terms in the description. In fact, over a quarter of “other” denial reasons cite at least one of these three terms in their descriptions. Unfortunately, we don’t know which specific programs or products are being referenced.

Other frequent terms are less informative. Some of the top terms include “requested,” “credit,” and “terms,” but these are often associated with phrases that don’t provide further insight, like variations of “We do not grant credit to any applicant on the terms and conditions you have requested.” Over 28 percent of “other” denial reasons cite at least one of these three terms.

We heard from lenders that applicants of color might be more likely to be denied for immigration-related reasons. While we find that only about 3 percent of the “other” denials mention immigration-related terms,⁸ such as “permanent resident” or “citizenship,” patterns across races and ethnicities are telling. Among applications that are denied due to “other” reasons, over 7 percent of those submitted by Asian and Latino applicants mention at least one of these immigration-related terms compared to a little over 1 percent among applications submitted by Black and White applicants. This is consistent with the fact that a larger share of Asian and Latino individuals in the United States are foreign-born compared to Black and White individuals. Still, these roughly 7 percent of immigration-related “other” denials account for only 1.3 percent and 1.2 percent of all denials for Latino and Asian applicants, respectively.

Credit history (14.6 percent of denials)

People whose mortgages are denied due to credit history may lack long-term accounts or regular, on-time account payments. Their current credit score may meet a lender’s requirement, but their longer-term credit record may give that lender pause.

Based on the reasons lenders report under HMDA, denied Black applicants are the only racial group more likely to be denied due to credit history than denied White applicants. Black applicants, on average, have lower credit scores and less income than White applicants—traits that could be associated with worse credit histories. When we use our regression analysis to compare similar Black and White denied applicants, the gap in denial rates shrinks considerably: as shown in Figure 6, Black denied applicants are 11.6 percent more likely to be denied due to credit history.⁹ That’s still substantial, but relatively small compared to the overall Black-White gap in denials. In our previous analysis, after accounting for borrower and loan differences, Black applicants are about twice as likely to be denied a conventional mortgage.

Debt-to-income ratios (14.1 percent of denials)

The debt-to-income ratio compares an applicant’s total monthly debt service payments, including their expected mortgage payment (and that of their co-applicant, if relevant), to the applicant’s income. Lenders are more likely to deny loans that would result in the application having a high debt-to-income ratio.

After adjusting for all the available application characteristics, including the debt-to-income ratio itself, and using the reasons lenders report to HMDA, a Black denied applicant or a Latino denied applicant is still more likely to be denied because of their debt-to-income ratio than a White denied applicant. Figure 7 shows this comparison. The difference between White and Asian applicants, however, could be mostly explained by the differences in their application characteristics.¹⁰

If two applicants have the same debt-to-income ratio but housing accounts for different shares of their overall debt, lenders likely view their applications differently. For example, consider two applicants with a debt-to-income ratio of .33. If one applicant’s mortgage payment would represent half of that monthly debt service, and the other’s would represent a third, lenders may make a different decision on the loan based on their assessment of risk of non-payment.

HMDA data do not indicate the size of denied applicants’ potential mortgage payments. However, we can estimate a monthly payment based on the data we have. When we do, we don’t find large differences in the housing share of debt-to-income ratios between Black and White denied applicants—but we do between Asian or Latino denied applicants and White denied applicants. Figure 8 shows the relationship between denied applicants’ estimated housing payments and their incomes. This suggests that denied White and Black applicants, particularly at higher debt-to-income ratios, have relatively more non-housing debt than denied Asian and Latino applicants.

Insufficient cash (11.2 percent of denials)

A mortgage application could also be denied if the applicant does not have enough cash for the down payment or other costs. Wealth, held in liquid assets, can help a prospective home buyer maintain enough cash to cover such costs. Nationally, Black and Latino households hold much less wealth than White households. Wealth isn’t measured in the HMDA dataset.

As Figure 9 shows, among denied applications, lenders are more likely to report denying Asian and Black applicants due to insufficient cash.¹¹ While application characteristics explain the majority of the disparities between White applicants and Black and Latino applicants, they don’t explain the disparities between White and Asian applicants at all. In fact, the disparities increase slightly after accounting for those characteristics. This is consistent with Asian applicants in our sample, on average, having preferable characteristics such as higher income and credit score.

Employment history (4.9 percent of denials)

Loan underwriters use employment histories to assess applicants’ income stability. Lenders named employment history as a reason for denial on nearly 5 percent of denied applications. As Figure 10 shows, Black denied applicants are the least likely to be denied due to employment history, while Asian denied applicants are 32.1 percent more likely to be denied for this reason compared to similar White applicants.¹²

Denial reasons underline need for lenders’ perspective

Lender-reported denial reasons don’t support arguments that credit and employment histories are driving an outsize part of racial disparities in mortgage denials. Our analysis of data on denials due to debt-to-income ratios introduces more questions than it answers.

HMDA data don’t contain any details on applicants’ actual credit and employment histories, but lenders might retain such information on their own. A next step for such lenders, if they are interested in decreasing any racial disparities in their own mortgage practices, may be to examine their own pool of rejected applicants, as one mortgage lender did after talking with us. Our analysis suggests there may be important lessons to learn.

Endnotes

1 Sample includes 30-year conventional, conforming, first-lien, single-dwelling, primary-residence, home-purchase applications that meet government-sponsored enterprises’ underwriting requirements. We exclude open lines of credit and loans for commercial or business purposes. See our previous analysis, described in our working paper, for more detail on the sample.

2 Lenders can select up to four denial reasons per denied application. In our sample, about 85 percent of denied applications are given a single reason. About 12 percent of denied applications have two reasons listed, about 2 percent have three, and 0.3 percent have four.

3 These shares, however, are calculated among denied applications. Since the denial rate is considerably lower among White applicants compared to applicants of color, even after adjusting for application characteristics, the shares among all applications would be significantly smaller among White applicants across all denial reasons compared to applicants of color.

4 In addition to the estimated rate ratios, we also perform a linear probability model (LPM) to estimate the rate difference in the likelihood of being denied due to insufficient collateral across races and ethnicities. Compared to similar White applicants, Asian, Black, and Latino applicants are 5.0, 3.3, and 1.6 percentage points less likely to be denied due to insufficient collateral, respectively.

5 Applications can be withdrawn or closed, precluding a lender’s credit decision. This is slightly more common among White applicants compared to applicants of color. In the HMDA dataset, withdrawn and closed applications lack important information such as credit score, debt-to-income ratio, and loan-to-value ratios. Withdrawn and closed applications are not included in our sample. Among the prospective borrowers in our sample, 15.2 percent of Asian applicants, 15.0 percent of Black applicants, 13.1 percent of Latino applicants, and 11.6 percent of White applicants withdrew their application before a credit decision was made by the lending institution. Among the same pool of loans, 2.7 percent of applications submitted by Asian applicants, 2.3 percent of applications submitted by Black applicants, 2.0 percent of applications submitted by Latino applicants, and 1.3 percent of applications submitted by White applicants were closed for incompleteness.

6 In the alternative LPM model, we estimate that compared to similar White applicants, Asian, Black, and Latino applicants are 0.1, 1.8, and 2.8 percentage points less likely to be denied due to incomplete applications, and are 4.7, 0.7, and 3.8 percentage points more likely to be denied due to unverifiable information, respectively.

7 In the alternative LPM model, we estimate that compared to similar White applicants, Asian, Black, and Latino applicants are 2.3, 1.5, and 3.1 percentage points more likely to be denied due to “other” reason, respectively.

8 The full list of terms we used for this immigration-oriented analysis is “permanent resident,” “legal resident,” “lawful resident,” “EAD” (or “employment authorization document”), “resident/non-resident alien,” “citizenship,” “temporary resident,” “length of residence,” and “DACA” (or “Deferred Action for Childhood Arrivals”).

9 In the alternative LPM model, we estimate that compared to similar White applicants, Asian and Latino applicants are 2.6 and 2.4 percentage points less likely, respectively, and Black applicants are 2.8 percentage points more likely to be denied due to credit history.

10 In the alternative LPM model, we estimate that compared to similar White applicants, Black and Latino applicants are 2.5 and 0.6 percentage points more likely to be denied due to debt-to-income ratios, respectively, whereas Asian applicants are just as likely.

11 In the alternative LPM model, we estimate that compared to similar White applicants, Asian and Black applicants are 1.3 and 0.8 percentage points more likely to be denied due to insufficient cash. Latino applicants are just as likely as White applicants to be denied for this reason.

12 In the alternative LPM model, we estimate that compared to similar White applicants, Asian applicants are 1.4 percentage points more likely to be denied due to employment history, Black applicants are 0.6 percentage points less likely, and Latino applicants are just as likely.