Racial disparities in unemployment benefits among U.S. mortgage borrowers during COVID-19

Abstract

This article describes racial and ethnic differences in mortgage payment difficulties during the COVID-19 pandemic and examines whether disparities exist in the benefits of the unemployment insurance (UI) program. The sample consisted of 80,797 jobless mortgage borrowers who received or waited for UI benefits between August 2020 and May 2022. Considering individual- and state-level variables in multilevel logistic regressions, we examined rates of mortgage delay in the last month and payment concerns about the next month by racial and ethnic group. Minority borrowers were more likely to have a difficulty in paying mortgage than White borrowers. UI recipients—regardless of race and ethnicity—were less likely to experience mortgage difficulties, but the positive unemployment benefit was reduced disproportionately among Blacks. Blacks were also at a higher risk of mortgage difficulties compounded by other pandemic-induced hardships—loss of household, lack of food, and mental illness—even after the receipt of UI. Findings on the intersection between race and ethnicity and UI suggest that pandemic policy interventions should be race conscious and consider the longstanding and systematic barriers experienced by minority mortgage borrowers.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10901-022-10006-w.

Introduction

The coronavirus disease 2019 (COVID-19) pandemic has abruptly changed the lives and livelihoods of billions of people across the globe since the World Health Organization’s (2020) declaration on March 11, 2020. In the United States, the pandemic has disrupted nearly every aspect of American society and among many consequences, the housing market has borne the brunt of the global health crisis (Joint Center for Housing Studies of Harvard University, 2022). This is due in large part to the unprecedented unemployment rate, which peaked at 14.7% in April 2020, a level not recorded since data collection started by the U.S. Bureau of Labor Statistics (BLS) in 1948 (BLS, 2022a). The historically high unemployment and extensive furloughs sharply reduced the fundamental ability to pay monthly housing costs among homeowners with a mortgage, the single largest type of household debt in the nation (Cornelissen & Hermann, 2020; Kolomatsky, 2020).

To respond in a timely fashion to the pandemic-driven economic fallout1 and job losses, the U.S. Congress passed the Coronavirus Aid, Relief, and Economic Security (CARES) Act. It became law on March 25, 2020, expanding eligibility for and benefits from the unemployment insurance (UI) program. This state-led funding program has supported workers who lost their job since 1935. During the COVID-19 pandemic, UI has been widely recognized as one of the earliest and major forms of federal assistance for numerous jobless mortgage borrowers to make ends meet. Beyond the seemingly universal UI benefits, however, there is an unanswered question of whether the benefit has been equal for everyone, particularly minority borrowers. Early COVID-19 studies have provided ample evidence that minority homeowners were more likely to experience difficulty in paying their mortgage, but they did not focus on whether and the extent to which the UI benefit eased these racial disparities (Hardy & Logan, 2020; Hermann & Cornelissen, 2020; Wong et al., 2020). Given that the pandemic has induced not only housing payment difficulties but other household hardships such as lack of access to food and mental illnesses, the possible role of UI in reducing the compounded types of pandemic hardships remains to be examined.

To fill the gaps, we examined the association between UI and mortgage payment difficulties in the pandemic, with an emphasis on the intersectionality between race and ethnicity and UI. We used the Household Pulse Survey (HPS), an experimental survey administered by the U.S. Census Bureau to produce near real-time data on how adult Americans—including mortgage borrowers—have fared with regards to their health and socioeconomic statuses. We first identified which racial and ethnic groups were more likely to be behind on mortgages and concerned about future payments during the pandemic, controlling for other intervening factors. Second, we examined whether and the extent to which UI benefits eased difficulties related to mortgage payments. Third, we focused on the intersectionality between race and ethnicity and UI to investigate whether UI benefits were equally available to different racial and ethnic groups in reducing the difficulties of paying a mortgage and other types of socioeconomic and health hardships at the same time.

In the next section, we conduct a brief overview of the COVID-19 trend and previous findings on racial disparities in the pandemic. We set research hypotheses and explain how we collected the HPS microdata and other state-level data. By adopting a multilevel modeling approach, we examined the associations between race and ethnicity and mortgage payment difficulties during the pandemic while emphasizing the intersectionality between UI and race. Last, we discuss our findings and conclude by drawing practical implications for housing policy and pandemic responses.

Background and research hypotheses

Race, unemployment, and mortgage market disruptions during COVID-19

Since the first COVID-19 case confirmed in January 2020, the ups and downs of the pandemic trend have continued throughout the end of 2022 (as of writing this article) even as life has returned to a semblance of normal and people inevitably have adapted to living with the virus. The continuing changes in the pandemic trend have had a significant effect on the job market and unemployment across the nation (Falk et al., 2021). Figure 1 displays how the national rate of unemployment peaked at an unprecedented and historically high level in April 2020 (14.7%) before declining to prepandemic level (around 4%) in the fourth quarter of 2021 (BLS, 2022a).

Fig. 1.

National trend in monthly unemployment rate and weekly Unemployment Insurance (UI) receipt rate, United States, January 1, 2020–June 1, 2022. Notes: Unemployment Insurance (UI) receipt rate is an indicator which is provided by the U.S. Department of Labor’s Weekly Regular State Data and is defined as the percent of U.S. total covered employees who received UI benefits in the past week. Monthly unemployment rate is marked on the 15th of each month. Sources: U.S. Department of Labor (DOL), Regular State Data within the Unemployment Insurance Weekly Claims Report; U.S. Bureau of Labor Statistics (BLS), National Labor Force Statistics.

Pandemic-driven unemployment has posed a threat to housing security among mortgage borrowers due to their lost and reduced income to pay the principal and interest (Mehdipanah, 2020). Particularly, minority and lower-income borrowers tend to be concentrated in the most vulnerable industries, including service, retail, and transportation, which were most affected by pandemic layoffs (Airgood-Obrycki, 2020) and government measures (e.g., stay-at-home orders, restaurant and bar closures; Davy, 2020; Froimson et al., 2020). An early UCLA report found that people of color experienced more severe mortgage delinquency and foreclosure during the pandemic than others (Wong et al., 2020). Studies also revealed a much higher prevalence of mortgage payment difficulties during  the COVID-19 pandemic compared to the pre-pandemic, with disproportionately severe disruptions among Black and Hispanic mortgagors (Airgood-Obrycki, 2020; Cornelissen & Hermann, 2020; Frost, 2020; Karpman et al., 2020; Wedeen, 2020).

Studies have noted possible reasons for the racial disparity, such as unequal health impacts of the pandemic and lockdowns (Chetty et al., 2020; Neal & McCargo, 2020), existing income gaps (BLS, 2022a; Keister & Lee, 2014), and different racial composition of the frontline workforce (Goldman et al., 2020; Saenz & Sparks, 2020). Compared to their proportion of the population, Blacks experienced higher infection and mortality rates related to the coronavirus (Neal & McCargo, 2020). Existing income gaps are reflected in the racial composition of the richest population in the United States; 92.5% of the top 1% of income for non-Hispanic Whites (Keister & Lee, 2014). Black workers work in the frontline industry, such as public transport and food retail, much more often than White workers (Ivanova, 2020). As essential workers, people of color had to go to work even if they were at risk of infection (Saenz & Sparks, 2020). According to the BLS, nearly 40% of Black workers were included in the mandatory worker category, which is about 10% points higher than that of Whites (BLS, 2022b). Lockdowns, designed to reduce or slow the spread of COVID-19, have paralyzed the housing market and businesses, affecting racial minorities more profoundly than others (Chetty et al., 2020). The racial disparities in mortgage insecurity in the pandemic are not a newly emerging problem due to the pandemic but rather a systemic and chronic issue that the pandemic has only exacerbated.

Even before the COVID-19 pandemic, a vast body of studies revealed factors related to the systemic racial disparity in the mortgage market. Particularly, studies revealed the role of racial gaps in income and debt in determining mortgage delinquency and forbearance. These studies found that people with lower incomes (Brent et al., 2011; Figueira et al., 2005), higher income variability in jobs (Webb, 1982), increased debt (Albanesi et al., 2017), and lower age (Deng & Liu, 2009) were more likely to become delinquent. In particular, the loan-to-value ratio, which indicates the equity effect, is a key variable influencing mortgage delinquency. A high initial loan-to-value has been shown to increase the likelihood of delinquency (Campbell & Dietrich, 1983; Deng et al., 2000; Mian & Sufi, 2010; Vandell et al., 1993). Also, interest rates, unemployment rates, and housing prices have been studied as variables affecting mortgage delinquency at the socioeconomic level. These studies found that high interest rates (Daglish, 2009; Figueira et al., 2005), high unemployment (Campbell & Dietrich, 1983; Deng & Liu, 2009; Figueira et al., 2005), and falling housing prices (Brent et al., 2011; Daglish, 2009) can increase the probability of mortgage delinquency. In this regard, Lee (2015) emphasized that the mortgage default is more problematic in areas of declining housing prices and high vacancy rates. Other regional explanatory variables related to mortgage delinquency include urban centrality (Immergluck, 2009) and geographic concentration of the Black population (Chan et al., 2013; Doviak & MacDonald, 2011). Based on these prepandemic and pandemic findings on racial disparities in the mortgage industry, we expected that minority borrowers would experience disproportionately severe difficulty in paying mortgages compared to Whites.

Research Hypothesis 1

During the COVID-19 pandemic, racial and ethnic minority borrowers are more likely to experience mortgage payment difficulties than Whites.

UI program

The UI program, first introduced in 1935 by the U.S. Department of Labor (DOL), is a federal and state program that helps workers who have lost their job by providing much-needed funding to temporarily replace part of their earnings while they look for a job (DOL, 2019b; Stone & Chen, 2014). Unlike Social Security and other federally operated programs, UI is administered mainly by the state government and can be seen as a package of different state-level programs due to its wide variation across states (Edwards, 2020). Each state has its own way of defining a worker’s preunemployment income, method to calculate how much income can be replaced by UI, and maximum benefits, among many other aspects of the program (Edwards, 2020).

For workers and their households that experienced a loss of income during the pandemic, UI has been a lifeline that offers funds for regular household expenses, including housing, food, transportation, education, and medical service (Mohanty, 2021). Since March 14, 2020, a day after the declaration of the national health crisis in the United States, more than 68 million UI initial claims have been filed (see Fig. 1; DOL, 2021d). Relatively to preceding recessions, this rapid upsurge in claims due to the pandemic is unprecedented (Urban Institute, 2021). UI has been widely noted as the early and major form of federal assistance for many jobless mortgage borrowers by easing liquidity constraints and providing a social safety net during the pandemic (Dieterle et al., 2020; Hsu et al., 2018. Other studies showed, due to UI benefits, a 2% increase in total consumption (Kaplan et al., 2020) and poverty declines (Han et al., 2020) during the pandemic. A prepandemic study also noted the important role of UI as a safety net to cope with the lack of liquidity, even for people with higher income (Hsu et al., 2018).

The CARES Act helped strengthen the UI program, which had become less generous after the Great Recession, as a more generous program that might help recipients pay for necessities including mortgage payments by including provisions to support families that lost their income (Page, 2020; Vroman & Kuehn, 2020). In addition to a one-time payment to every adult with annual income under $75,000, the CARES Act offered supplementary cash benefits each week in the UI program on top of existing state benefits. Also, the CARES Act extended coverage by additional weeks for recipients who had spent all of their existing benefits and expanded eligibility to self-employed people and gig workers. Given these extended UI benefits, according to Urban Institute’s estimates, the total UI benefit was as large as the monthly amount of median household income in nearly all of the 10 most populous states, implying that median mortgage borrowers might have a sufficient residual income for other basic necessities, including food and medical expenses, after their monthly mortgage payments.

The UI program may also have created an extended unemployment period, which has been known to impact some racial and ethnic groups more severely than others (Gruber, 2007; Rebollo & Rodríguez, 2020). Gruber (2007) confirmed that an increase in UI benefits by 10% has the effect of increasing the unemployment period by about 4% to 8%. Chetty (2008) found that UI benefits only affect jobseekers facing liquidity constraints and that it does not have a significant effect on job-seeking periods of jobseekers who are not faced with these constraints. In contrast, other studies suggested unintended negative effects of UI benefits on some population subgroups in that the program contributes to extending the unemployed period and moral hazard (Bell et al., 2020; Chetty, 2008; Dieterle et al., 2020; Gruber, 2007; Hagedorn et al., 2013; Parolin et al., 2020). Therefore, it is needed to examine the role of UI in determining personal liquidity constraints during the pandemic and consider regional issues such as the interrelationship of race and UI. Considering the litereature on the role of UI in pre-pandemic and pandemic periods, it is reasonable to hypothesize that mortgage borrowers who received the benefit would be less likely to experience housing payment difficulties than their counterparts.

Research Hypothesis 2

UI recipients are less likely to experience mortgage payment difficulties compared to those who applied but did not receive UI benefits during the COVID-19 pandemic.

Intersectionality of race and ethnicity and UI benefits

Experts have widely acknowledged that people of color, particularly Blacks, were more likely to lose their job and apply for benefits from the UI program than Whites during the COVID-19 pandemic (Bell et al., 2020; Ganong et al., 2022; Mohanty, 2021; Urban Institute, 2021). The unemployment rate among Blacks was recorded at 5.9%, relative to only 3.1% among Whites in the second quarter of 2022 (BLS, 2022b). According to federal data (from May 2022) on the characteristics of UI claimants, Blacks account for 19.1% of all benefits claims while their proportion in the total population is only 12.6% (DOL, 2021e; see Supplemental Table 1 and 2020 American Community Survey 5-year summary table B02001). In contrast, Whites account for only 46.1% of all UI claims whereas they are 70.4% of the total population. These racial statistics imply that unemployment and reliance on UI benefits were more acute among Blacks than Whites in the latest phase of the pandemic.

Pandemic studies about the intersectionality between race and ethnicity and UI primarily have examined racial discrimination and privilege in terms of receipt (among all unemployed people), uptake (receipt among eligible unemployed people), preunemployment incomes, and place of residence (Edwards, 2020; Kuka & Stuart, 2021). A study showed that Blacks who lost their job were 24% less likely to receive UI than Whites, which has been consistent between 1986 and 2015 (Kuka & Stuart, 2021). They also found racial disparities in UI receipt and uptake were mainly attributable to lower preunemployment incomes and a higher tendency among Blacks to live in the South (Kuka & Stuart, 2021). A RAND study (Edwards, 2020) also emphasized the importance of state of residence because 1 in 4 Black workers in the nation resides in one of only three states (Texas, Florida, and Georgia), which offer less generous UI benefits than other states, which is consistent with prepandemic findings (Nichols & Simms, 2012; Vroman, 2005).

Unemployed people from minority groups or with low levels of education may lack information and knowledge about UI program or have difficulty in completing the entire application process (Gould-Werth & Shaefer, 2012). Another possible reason for the racial disparity in UI benefits is that not all workers are eligible for UI although they become unemployed (Nichols & Simms, 2012). Low-educated workers tend to have shorter tenures and are more likely to be fired voluntarily or for other reasons, which means they are ineligible for UI benefits. In this context, Parolin et al. (2020) and Bell et al. (2020) argued that more governmental social security benefits and support are needed for low-income people or ethnic minorities.

COVID-19 studies further noted a doubled or tripled burden on mortgage borrowers who have experienced housing payment difficulties in addition to other pandemic-induced hardships (Mohanty, 2021; Park, 2021). According to the U.S. Census Bureau, 31.2% of households who received and used UI for basic household expenses reported that they still had serious difficulty in meeting normal spending needs, which included not only housing payments (mortgage or rent) but also food, transportation, K-12 education, student loans, and medical treatment (Mohanty, 2021). A pandemic study also found an overlap between housing payment difficulties and other pandemic-induced hardships, such as loss of employment income, food insufficiency, and perceived health issues (Park, 2021). Particularly, Blacks were 3 times more likely to experience all hardships at the same time than Whites, implying racial disparities in exposure to compounded hardships. According to these prior findings on racial disparities in UI benefits and pandemic-induced compounded hardships, we expected to find a moderating role of intersectionality between race and ethnicity and UI benefits in determining mortgage payment difficulties during the COVID-19 pandemic. We also hypothesized an additional racial disparity in which minority borrowers would disproportionately fall into mortgage payment difficulties accompanied by other pandemic-induced socioeconomic and health hardships.

Research Hypothesis 3a

Among UI recipients during the COVID-19 pandemic, racial and ethnic minority borrowers are more likely to experience mortgage payment difficulties than Whites even after the receipt of UI benefits.

Research Hypothesis 3b

Among UI recipients during the COVID-19 pandemic, racial and ethnic minority borrowers are more likely to experience mortgage payment difficulties compounded with pandemic-induced socioeconomic and health hardships than Whites, even after receiving UI benefits.

Data and methods

U.S. Census Bureau’s Household Pulse Survey

The Household Pulse Survey (HPS) is a COVID-19-specialized and nationally representative2 survey launched by the U.S. Census Bureau (2020a) in conjunction with the U.S. Department of Housing and Urban Development (2020) and several federal organizations. It measures the housing, socioeconomic, and public health impacts of the pandemic on Americans and their households, including mortgage borrowers. Particularly, the HPS provides unique data that allowed us to examine not only housing payment difficulties but other socioeconomic and health hardships experienced by mortgage borrowers who relied on UI benefits to meet basic needs (Mohanty, 2021).

We used a publicly available file of the HPS data, which contains individual respondents’ answers to questions (U.S. Census Bureau, 2020b). We compiled a pooled cross-sectional data that featured a sample of 80,797 adults (aged 18 or older) who met the following criteria during the COVID-19 pandemic: (a) resided in an owner-occupied housing unit with mortgage or loan, (b) household experienced a loss of employment income, (b) applied for UI benefits, and (d) answered all questions of interest. Even if the HPS data represented the adult population as discussed in the footnote, it may not be a good representation of mortgage borrowers who lost income and filed an application for UI benefits, particularly in small states such as Wyoming, Vermont, and Washington, DC (see Supplemental Table 3 for state-by-state sample size). Despite the small samples in some states, we used all observations to examine the nation as a whole and were cautious about any cross-state comparisons.

The study period included available survey weeks from August 19, 2020–May 9, 2022. The HPS data through May 2022 feature two phases: (a) Phase 2 (five survey weeks from August 19–October 26, 2020) and (b) Phase 3 and subsequent subphases (21 survey weeks from October 28, 2020–May 9, 2022; see Supplemental Table 4 for detailed information on survey phases and weeks). This article excluded Phase 1 (12 survey weeks from April 23–July 21, 2020) and Phase 3.2 (six survey weeks from July 21–October 11, 2021) because UI-related questions were not asked of survey participants. The study period allowed us to examine mortgage payment difficulties in various subperiods of the COVID-19 pandemic, such as repeated worsening and easing times and the peak (January 2022) of the health crisis. A notable turning point of the COVID-19 pandemic might be vaccine availability, which was initiated in December 2020. The study period covered both the initial vaccine uptake among vulnerable subpopulations (e.g., older adults and residents in long-term care facility) and frontline essential workers (e.g., medical doctors and nurses) and also the later period when around 70% of the entire population group received one or more vaccine doses (Centers for Disease Control and Prevention, 2020).

Multilevel modeling approach

Since the data consisted of individual mortgage borrowers clustered in aggregate contexts that differed by state, we adopted multilevel modeling approach to examine racial disparities in the UI benefits. Multilevel modeling is often used to analyze data with a hierarchical structure in which both lower level (e.g., mortgage borrowers in level 1) variables and higher level (e.g., states in level 2) variables may be related to outcomes(Raudenbush & Bryk, 2002). The primary advantage of the multilevel modeling over other models (e.g., ordinary least squares) is that we could analyze not only heterogeneities among borrowers (level 1) but also heterogeneities between states (level 2) to which those individuals belong, allowing us to consider random components at each level.  In addition, multilevel models may partially account for the fact that individuals in the same state are not independent to each other and result in unbiased estimates of standard errors compared to traditional models. This is especially important to COVID-19 research given that state governments have played an important role in responding to the pandemic, resulting in a large cross-state variation in the mortgage payment difficulty.

We estimated multilevel mixed-effects logistic models by using Stata (StataCorp, 2019). The person and household (level 1) model was expressed as:

$$Y_{\mathit{ijk}}={\beta }_{0jk}+{\beta }_{1jk}{\mathit{UI}}_{\mathit{ijk}}+{\beta }_{2jk}{D}_{\mathit{ijk}}+{\beta }_{3jk}{S}_{\mathit{ijk}}+{\beta }_{4jk}{F}_{\mathit{ijk}}+{\epsilon }_{\mathit{ijk}}$$

The state-level (level 2) model was expressed as:

$${\beta }_{0jk}={\gamma }_{0jk}+{\gamma }_{1jk}{P}_{\mathit{jk}}+{\mu }_{\mathit{jk}}$$

For person i in state j in week k in the level 1 model, $Y_{\mathit{ijk}}$ is a dependent variable denoting difficulties of mortgage payment during the COVID-19 pandemic, ${\mathit{UI}}_{\mathit{ijk}}$ is UI status; $D_{\mathit{ijk}}$ is demographic variables; $S_{\mathit{ijk}}$ is social and economic characteristics; and $F_{\mathit{ijk}}$ is fixed effects. For state j in week k in the level 2 model, $P_j$ is statewide pandemic conditions. ${\mu }_{\mathit{jk}}$ and ${\epsilon }_{\mathit{ijk}}$ are error terms at each level.Standard errors in the models were clustered to correct for heteroscedasticity across clusters (states) of individual samples (mortgage borrowers). Since the models in this article were unable to examine causality, we were cautious to make interpretations of the estimated results.

Variable specification

Mortgage payment difficulties during COVID-19

The HPS asks two kinds of mortgage payment evaluation questions. The first retrospective question is “Did you pay your last month’s mortgage on time? Select only one answer.” It provides the following two options: (a) yes, [I/we] paid last month’s mortgage on time and (b) no, [I/we] did not pay last month’s mortgage on time. We defined last month’s mortgage delay as a binary variable (0 = paid, 1 = did not pay). Table 1 presents descriptive statistics of the mortgage payment and all the other variables.

Table 1.

Descriptive statistics of variables

Variable	Full sample (n = 80,797, % of n or mean (SD))	Sig.	Last month's mortgage delay		Next month's payment concerns
			Yes (n = 11,184, % of n or mean (SD))	No (n = 69,613, % of n or mean (SD))	Yes (n = 16,359, % of n or mean (SD))	No (n = 64,438, % of n or mean (SD))
Level 1 (person and household) variables
Unemployment insurance (UI)
UI status
UI applied but not received yet	28.5		31.5	27.9	32.4	27.2
UI received	71.5		68.5	72.1	67.6	72.7
Demographic characteristics
Age
18–24 (Ref)	6.1		6.5	6.0	4.8	6.5
25–34	18.6		17.8	18.8	18.9	18.5
35–44	22.3		25.6	21.7	25.2	21.4
45–54	21.8		25.8	21.0	25.1	20.7
55–64	20.4		17.8	20.9	18.4	21.0
65+	10.8		6.5	11.7	7.7	11.9
Gender
Female (Ref)	53.8		53.5	53.9	52.9	54.2
Male	46.2		46.5	46.1	47.1	45.8
Race/ethnicity
Non-Hispanic White (Ref)	63.8		51.8	66.3	51.3	68.0
Non-Hispanic Black	9.5		14.5	8.5	13.5	8.1
Non-Hispanic A&PI	5.5		6.2	5.3	6.7	5.1
Non-Hispanic other	3.9		5.2	3.6	4.7	3.6
Hispanic	17.3		22.3	16.3	23.9	15.2
Marital status
Unmarried (Ref)	39.4		45.0	38.3	44.0	37.9
Married	60.6		55.0	61.7	56.0	62.1
Children in household
No child (Ref)	55.4		46.0	57.3	48.2	57.8
One or more children	44.6		54.0	42.7	51.8	42.2
Household size
Single person (Ref)	4.1		4.0	4.2	4.1	4.1
2-person	25.1		16.8	26.7	18.4	27.3
3-person	22.0		19.2	22.5	20.1	22.6
4-person	23.5		25.0	23.2	24.0	23.4
5-person	12.9		17.2	12.1	15.9	12.0
6 or more persons	12.4		17.8	11.3	17.4	10.7
Socioeconomic statuses
Education
Less than high school (Ref)	5.9		9.2	5.2	10.2	4.4
High school	31.1		34.5	30.4	35.2	29.7
Some college & AA	36.3		36.0	36.4	36.5	36.3
BA +	26.7		20.2	28.0	18.0	29.6
Household income
Less than $25,000 (Ref)	9.2		15.1	8.1	15.6	7.1
$25,000–49,999	23.2		30.4	21.8	31.7	20.5
$50,000–74,999	21.2		22.4	20.9	23.1	20.5
$75,000–99,999	16.7		14.3	17.1	13.7	17.6
$100,000–$149,999	17.7		11.0	19.0	11.1	19.8
$150,000 and above	12.0		6.8	13.1	4.8	14.4
Level 2 (state) variables
COVID-19 cases per 100 persons	6.3 (4.61)	***	6.38 (4.6)	6.29 (4.61)	6.44 (4.8)	6.25 (4.54)
% unemployment	6.67 (1.87)	***	6.74 (1.89)	6.66 (1.86)	6.7 (1.9)	6.66 (1.86)
UI claims	4.46 (2.99)	***	4.41 (2.97)	4.47 (3)	4.41 (3.03)	4.47 (2.98)
% delinquency	0.91 (0.33)	***	0.94 (0.34)	0.9 (0.33)	0.93 (0.34)	0.9 (0.33)
% forbearance	4.49 (2.74)	***	4.81 (2.96)	4.43 (2.68)	4.81 (2.95)	4.39 (2.65)
Spatiotemporal fixed effects
15 largest MSAs
None (Ref)	64.5		60.5	65.3	61.3	65.5
New York	5.6		7.1	5.3	6.8	5.2
Los Angeles	4.6		4.5	4.7	4.9	4.5
Chicago	2.9		3.3	2.8	2.9	2.9
Dallas	1.6		2.1	1.5	2.2	1.5
Houston	2.1		3.4	1.9	3.5	1.7
Washington, D.C	1.5		1.7	1.5	1.7	1.4
Miami	1.6		2.3	1.5	1.9	1.5
Philadelphia	2.6		2.7	2.6	2.9	2.6
Atlanta	2.2		2.9	2.0	2.7	2.0
Phoenix	1.4		1.2	1.4	1.3	1.4
Boston	1.7		1.4	1.8	1.4	1.8
San Francisco	1.3		1.3	1.3	1.2	1.3
Riverside	2.2		2.5	2.2	2.5	2.2
Detroit	2.3		2.1	2.4	1.9	2.5
Seattle	1.7		1.0	1.8	0.9	1.9
HPS week
Week 13 (8.19–31, 2020, Ref)	5.5		4.7	5.7	5.4	5.6
Week 14 (9.2–14)	5.2		5.4	5.2	4.9	5.3
Week 15 (9.16–28)	5.4		5.5	5.4	5.0	5.6
Week 16 (9.30–10.12)	5.4		5.2	5.5	5.0	5.6
Week 17 (10.14–26)	5.4		5.6	5.4	5.3	5.5
Week 18 (10.28–11.9)	5.8		6.6	5.7	6.8	5.5
Week 19 (11.11–23)	5.6		4.5	5.8	5.9	5.5
Week 20 (11.25–12.7)	5.6		5.6	5.6	6.4	5.3
Week 21 (12.9–21)	6.0		7.2	5.7	6.9	5.7
Week 22 (1.6–18, 2021)	5.2		4.8	5.3	5.1	5.3
Week 23 (1.20–2.1)	5.6		5.7	5.6	5.6	5.6
Week 24 (2.3–15)	5.5		5.0	5.5	5.1	5.6
Week 25 (2.17–3.1)	5.4		5.4	5.4	5.0	5.5
Week 26 (3.3–15)	5.0		5.2	4.9	4.2	5.2
Week 27 (3.17–29)	5.1		5.0	5.1	3.8	5.5
Week 28 (4.14–26)	2.1		2.1	2.1	2.1	2.1
Week 29 (4.28–5.10)	2.1		2.4	2.0	2.3	2.0
Week 30 (5.12–24)	2.1		1.8	2.1	2.2	2.0
Week 31 (5.26–6.7)	2.3		2.3	2.2	2.0	2.4
Week 32 (6.9–21)	2.0		2.0	2.0	2.1	1.9
Week 33 (6.23–7.5)	2.0		2.2	2.0	2.3	1.9
Week 40 (12.1–13)	1.3		1.2	1.3	1.5	1.2
Week 41 (12.29–1.10, 2022)	1.5		1.5	1.5	1.5	1.5
Week 42 (1.26–2.7)	1.5		1.3	1.6	1.8	1.4
Week 43 (3.2–14)	0.4		0.5	0.4	0.6	0.4
Week 44 (3.30–4.11)	0.4		0.6	0.4	0.4	0.5
Week 45 (4.27–5.9)	0.5		0.3	0.5	0.7	0.4

UI Unemployment Insurance, HPS Household Pulse Survey, MSA Metropolitan Statistical Area, A&PI Asians and Pacific Islanders, AA Associate degrees, BA+ bachelor and higher degrees, Ref reference

⁺p < 0.10; *p < 0.05; **p < 0.01; ***p < 0.001. Multiple-sample (i.e., 50 state and D.C.) tests on the equality of means confirm that there are statistically significant differences at the p < 0.001 level for each of the Level 2 (state) variables (mvtest in Stata)

The second prospective question in HPS is “How confident are you that your household will be able to pay your next mortgage payment on time? Select only one answer.” Respondents had the following four options: (a) no confidence, (b) slight confidence, (c) moderate confidence, and (d) high confidence. We specified next month’s payment concerns as another binary outcome (0 = moderate or high confidence, 1 = no or slight confidence) to indicate the confidence of respondents regarding making their next month’s mortgage payment. Given the pandemic-driven volatilityin the mortgage market, we further tested alternative outcomes to reflect narrower and broader concepts of mortgage payment difficulties as a robustness check: one denoting the experience of both payment delay and having concerns at the same time and another for either experiencing delay or concerns.

Panel (a) of Fig. 2 displays the change in mortgage payment difficulties in the nation from April 2020 to June 2022 through a trendline graph, with the percentage of mortgage payment difficulties on the vertical axis and the HPS week on the horizontal axis. The national rate of mortgage delays was recorded at 5.2% at the beginning of the pandemic. The rate increased rapidly until August 2020, when federal foreclosure protections were implemented by the CARES Act. The heightened rate continued for around half a year and finally declined in March 2021. In contrast, the rate of payment concerns was very volatile, between 6.4% and 10.4%, for the past 2 years. The nationwide trend paralleled changes in the number of newly confirmed cases of COVID-19 implying that psychological concerns about mortgage payments reflect the volatility of the health crisis.

Fig. 2.

Mortgage payment difficulties during the COVID-19 pandemic, United States, April 23, 2020–June 13, 2022. Notes: The displayed trendlines are exclusive to each other. Universe is adult population in owner-occupied housing units with mortgage. The survey question about the confidence in ability to pay mortgage next month was not asked in the survey week of June 1–13, 2022. Sources: U.S. Census Bureau, Household Pulse Survey (HPS), Public Use Files (PUFs). Notes: The displayed columns are exclusive to each other. Universe is adult population in owner-occupied housing units with mortgage. Sources: U.S. Census Bureau, Household Pulse Survey (HPS), Public Use Files (PUFs)

The fifty states and Washington, DC show a comprehensive picture of mortgage difficulties across the nation. Panel (b) of Fig. 2 shows the percentage of mortgage borrowers who delayed their last month’s mortgage payment and those who lacked confidence about their next payment. For the past 2 years, on average, 8.4% of mortgage borrowers were not confident about their next payment, whereas 6.0% did not make their last month’s payment. Some states presented a much higher level of mortgage difficulty than the national average while others were relatively lower. Many Southern states (e.g., Mississippi and Louisiana) and states in the West and Northeast experienced a more serious level of mortgage difficulties than the national average. Also noteworthy is that states with the greatest prevalence of concerns also had relatively greater proportions of mortgage borrowers being behind on mortgages.

UI status

The key independent variable was UI status, which identified whether a survey respondent received UI benefits. The variable was built by combining three questions in the HPS: (a) an income loss question, “Have you, or has anyone in your household, experienced a loss of employment income since March 13, 2020?”; (b) a UI application question, “Since March 13, 2020, have you applied for UI benefits?”; and (c) a UI receipt question, “Since March 13, 2020, did you receive UI benefits?” All had answer options of yes or no. We specified UI status as a binary variable (0 = did not receive UI benefits, 1 = received UI benefits).

The UI variable separated beneficiaries from others on the waitlist, allowing us to examine the association between UI receipt and mortgage payment difficulties. However, a notable limitation of the variable is it may not represent all UI recipients in the nation and across states. This is due to not only the small sample size of the HPS microdata but also differences in the survey design. The HPS asks respondents whether they had received UI benefits since the beginning of the COVID-19 pandemic (i.e., March 13, 2020) without detailed identification such as continued (or discontinued) receipt and initial claim case. Unlike the accumulated and simplified count of UI recipients, the widely used federal data feature a count with more detailed identifications every week, which prevents apple-to-apple comparisons with the HPS data (DOL, 2021d).

Level 1 (person and household) variables

We included level 1 variables known to be related to mortgage payment ability, all derived from the HPS data: (a) race and ethnicity and other demographic characteristics (age, gender, marriage, children, and the number of household members) and (b) social and economic statuses, such as education and household aggregate income (Sanchez-Moyano, 2021).

Level 2 (state) variables

We considered level 2 variables denoting elements of state-level conditions that may be related to individual mortgage borrowers’ payment difficulties during the COVID-19 pandemic. Using these state-level variables, it was possible to not only control for regional differences in the empirical analysis but also to find out whether the cause of the results was due to regional and individual differences. To reflect the volatile and frequent changes of the pandemic situation over time, we used high-frequency (weekly and monthly) data instead of low-frequency (annual) data for most variables (see Supplemental Table 5 for details about data sources and frequency).

Studies have noted the overall condition and performance of the mortgage market as one factor that affects individual borrowers’ successful payments (Gür, 2022; Lee, 2015). Mortgage forbearance is not automatically applied, so individuals had to contact their mortgage servicer. In the early period of the pandemic, there were difficulties connecting for such a request (U.S. Consumer Financial Protection Bureau, 2022). Difficulties related to forbearance varied from state to state and may have had different impacts on an individual’s liquidity constraints by region. In addition, because the regional mortgage delinquency situation is based on local economic conditions and prospects in the regional housing market, a sudden change in the local housing market situation, including delinquency rate, can affect the probability of individual mortgage delinquency in a chain (An et al., 2022; Wadud et al., 2020). By using the National Mortgage Database administered by the Federal Housing Finance Agency, we operationalized the state-level aggregate conditions of the mortgage market. We used a quarterly state-level database called U.S. State Level Mortgage Performance Statistics to specify mortgage market performances in two dimensions: (a) percentage of mortgages in delinquency of 30–89 days past the due date, defined as the count of active loans 30–89 days past the due date-subject to the state account rule-at the end of a given quarter divided by all of the active loans at the end of the same quarter multiplied by 100,3 and (b) percentage of mortgages in forbearance, defined as the count of active loans indicated as being in forbearance at the end of a given quarter divided by all of the active loans at the end of the same quarter multiplied by 100.4

Cross-state variations in the UI program may affect state residents’ receipt and use of UI benefits, even if the federal government oversees the program (DOL, 2019a, 2019b). Interstate differences occur mainly because state governments administer and determine details about benefits eligibility, maximum amount and duration, methods of financing, and many other key characteristics of the program (DOL, 2021a, 2021b, 2021c). For example, some states pay all unemployed people for the same period, but in some states, UI benefits vary depending on the unemployed person’s work history (DOL, 2021a). Because not all unemployed people face liquidity constraints, UI benefits are limited to only some unemployed people, so it is necessary to control for regional UI claims. By using the DOL’s Unemployment Insurance Weekly Claims Data, we considered the total number of weekly UI claims filed in each state (DOL, 2021d). We generated a biweekly UI claim variable, which was a natural log-transformed variable that reflected different populations across states.

The statewide prevalence of unemployment, as noted by other COVID-19 studies (Dettling & Lambie-Hanson, 2021), was also included  to consider labor market and job conditions. Rising regional unemployment rates can be a direct cause of changes in the probability of an individual’s declining ability to pay (Neal & McCargo, 2020). Because the 2020 annual average unemployment rates for states varied significantly from 4.1% (Nebraska) to 13.5% (Nevada) and the discretionary authority of UI programs of states differed (BLS, 2022a; DOL, 2021b), we included both UI claims and unemployment rates in the model. By using the unemployment data calculated and released every month by the BLS, we included the state-level monthly unemployment rate in the models.

The speed and pattern of the spread of the coronavirus varied across states, resulting in diverging experiences of the pandemic among state residents in different states. Because the coronavirus is infectious, individual activities are inevitably affected by the pandemic’s severity in each region (Li & Mutchler, 2020). Because the industrial structure and composition of the population of each region are different, the pandemic’s severity in a region not only directly affects an individual’s mortgage situation but also needs to be controlled to determine the effect of individual-level variables. By using the federal government’s COVID Data Tracker, we specified the accumulated count of confirmed COVID-19 cases relative to the total population in each state (Centers for Disease Control and Prevention, 2020). We aggregated and averaged the daily database to match the biweekly survey cycle of HPS data. Then, the average was divided by the total state population as of 2019 and multiplied by 100.5

Spatial and temporal fixed effects

A set of fixed effects were added in the model to partly account for time- and space-invariant factors related to individual borrowers' mortgage payment outcomes. We included a group of dummy variables identifying the 15 largest metropolitan statistical areas (reference = outside the metros). We examined the metropolitan fixed effect to consider locational differences in mortgage difficulties across the nation’s largest metros. In the study period (August 19, 2020–May 9, 2022), mortgage borrowers' difficulties might have changed over time for some unobservable reasons that were not accounted in the models. We included a series of dummy variables to identify survey weeks and partially reflect the temporal change (reference = August 19–31, 2020).

Empirical results

Spatiotemporal differences in mortgage payment difficulties

We first examined trends in mortgage payment difficulties during the pandemic by including only a set of dummy variables of survey week (reference = August 19–31, 2020) and the 15 largest metropolitan statistical areas (reference = outside the metros), with a constant term. These plain fixed effects can be seen as a basis for evaluating the effects of including key variables in the next subsections.

For the past 2 years of the pandemic, mortgage payment difficulties worsened and eased repeatedly in the study period, but most of the coefficients of weekly dummy variables were not significant (see Supplemental Table 6). This implies that the mortgage difficulty did not change over time simply because the pandemic occurred; rather, it was attributable to some time-variant factors, as reflected in the level 2 model in the next subsections. Despite insignificance of the weekly variables, the volatile trend observed in the model is consistent with the trendlines shown in panel (a) of Fig. 2.

Turning to spatial fixed effects, we found significant differences in mortgage payment difficulties in most of the largest metropolitan areas compared to nonmetros. The result shows that, for example, mortgage borrowers in Los Angeles were 18.4% and 15.1% more likely to be behind on payment and concerned about their next payment than their nonmetro counterparts, respectively. Also, metro areas in the South (e.g., Houston and Miami) showed the highest prevalence of mortgage payment delays and concerns. In contrast, metro areas renowned for high incomes, such as San Francisco and Seattle, showed a lower prevalence of mortgage difficulties compared to nonmetros. A notable finding is that the cross-metro difference was significant even when we used clustered standard errors at the state level, which indicates an acute difficulty of mortgage payments concentrated in the largest urban areas in the nation.

Full model results

Table 2 shows estimation results that include all variables at the level 1 (person and household) and level 2 (state). Most person and household variables appear to be significant determinants of mortgage payment difficulties. We found that Blacks were nearly 2 times (odds ratio [OR] = 1.966) more likely to be behind on their mortgage than Whites. Also, Blacks were 1.7 times more likely to be concerned about their next payment than Whites. These findings support Research Hypothesis 1, in which we assumed that compared to Whites, Blacks would be more likely to experience mortgage payment difficulties during the COVID-19 pandemic. Other racial and ethnic groups—Asians, Hispanics, and others—also had a higher risk of mortgage difficulties, but Blacks had the highest risk. Other significant determinants of mortgage difficulties included middle-age, male, unmarried status, having children in the household, greater household size, lower level of education, and lower level of household income, compared to their counterparts respectively.

Table 2.

Multilevel mixed-effect logistic regression results for mortgage payment difficulties during the COVID-19 pandemic

Variable	Last month's mortgage delay		Next month's payment concerns
	Odds ratio	Sig.	Odds ratio	Sig.
Level 1 (person and household) variables
Unemployment insurance (UI)
UI status (Ref = UI applied but not received yet)
UI received	0.850	***	0.845	***
Demographic characteristics
Age (Ref = 18–24)
25–34	1.297	***	1.690	***
35–44	1.683	***	2.169	***
45–54	1.987	***	2.553	***
55–64	1.601	***	2.014	***
65+	0.968		1.284	***
Gender (Ref = female)
Male	1.166	***	1.136	***
Race/ethnicity (Ref = non- Hispanic white)
Non-Hispanic black	1.966	***	1.706	***
Non-Hispanic A&PI	1.587	***	1.615	***
Non-Hispanic other	1.495	***	1.511	***
Hispanic	1.320	***	1.421	***
Marital status (Ref = unmarried)
Married	0.840	***	0.825	***
Children in household (Ref = no child)
One or more children	1.254	***	1.190	***
Household size (Ref = single person)
2-person	0.905	*	1.014
3-person	1.034		1.118	***
4-person	1.153	**	1.204	***
5-person	1.436	***	1.489	***
6 or more persons	1.616	***	1.716	***
Socioeconomic statuses
Education (Ref = less than high school)
High school graduate	0.909		0.783	***
Some college or associate degree	0.900	+	0.743	***
Bachelor's degree or higher	0.744	***	0.522	***
Household income (Ref = less than $25,000)
$25,000–49,999	0.783	***	0.720	***
$50,000–74,999	0.605	***	0.513	***
$75,000–99,999	0.486	***	0.391	***
$100,000–$149,999	0.347	***	0.265	***
$150,000 and above	0.246	***	0.159	***
Level 2 (state) variables
COVID-19 cases per 100 persons	1.019	*	1.017	+
% unemployment	1.046	***	1.027	**
UI claims	0.997		1.005
% delinquency	1.333	***	1.292	***
% forbearance	1.016	*	1.015	*
Spatiotemporal fixed effects
15 largest MSAs (Ref = none)
New York	1.338	***	1.328	**
Los Angeles	1.142	***	1.160	***
Chicago	1.194	***	1.101	*
Dallas	1.200	***	1.232	***
Houston	1.446	***	1.429	***
Washington, D.C.	1.253	***	1.419	***
Miami	1.442	***	1.210	***
Philadelphia	1.039		1.173
Atlanta	1.070	*	1.042	*
Phoenix	1.190	***	1.058	***
Boston	1.088	**	1.088	**
San Francisco	1.053	*	1.013
Riverside	1.075	***	1.010
Detroit	1.037		1.024
Seattle	0.916	***	1.020
HPS week (Ref = week 13, 8.19–31, 2020)
Week 14 (9.2–14)	1.102	+	0.939
Week 15 (9.16–28)	1.069		0.975
Week 16 (9.30–10.12)	1.073		0.918
Week 17 (10.14–26)	1.075		0.990
Week 18 (10.28–11.9)	1.111		1.075
Week 19 (11.11–23)	1.084		1.181	*
Week 20 (11.25–12.7)	1.088		1.103
Week 21 (12.9–21)	1.145	+	1.111
Week 22 (1.6–18, 2021)	1.065		0.997
Week 23 (1.20–2.1)	1.061		0.970
Week 24 (2.3–15)	1.017		0.937
Week 25 (2.17–3.1)	1.060		0.890
Week 26 (3.3–15)	1.079		0.835	*
Week 27 (3.17–29)	0.868		0.621	***
Week 28 (4.14–26)	1.110		0.952
Week 29 (4.28–5.10)	1.042		0.913
Week 30 (5.12–24)	1.117		0.996
Week 31 (5.26–6.7)	0.978		0.870
Week 32 (6.9–21)	1.026		0.967
Week 33 (6.23–7.5)	1.061		1.025
Week 40 (12.1–13)	0.953		1.029
Week 41 (12.29–1.10, 2022)	0.872		0.812
Week 42 (1.26–2.7)	0.846		0.782
Week 43 (3.2–14)	0.928		1.004
Week 44 (3.30–4.11)	0.756		0.952
Week 45 (4.27–5.9)	0.636	+	0.876
Constant	0.090	***	0.214	***
Number of observations	80,797		80,797

Standard errors were clustered at the state level

UI Unemployment Insurance, HPS Household Pulse Survey, MSA Metropolitan Statistical Area, A&PI Asians and Pacific Islanders, AA associate degrees, BA+ bachelor and higher degrees, Ref reference

⁺p < 0.10; *p < 0.05; **p < 0.01; ***p < 0.001

Another key finding is the significant association between UI and a lower rates of mortgage delay and payment concerns. Mortgage borrowers who received UI benefits were 15.0% less likely to be behind on their mortgage than those on the waitlist (reference group). UI beneficiaries were also 15.5% less likely to be concerned about upcoming payments than the reference group. The results lend support to Research Hypothesis 2, which posited that during the pandemic, UI recipients would be less likely to experience mortgage payment difficulties than those who applied for but had yet to receive benefits.

Focusing on state-level (level 2) variables, we found significant associations between statewide mortgage market conditions and individual borrowers’ payment difficulties after controlling for person- and household-level variables. We found that a 1% increase in the rate of mortgage delinquency in a state was related to higher likelihoods of delaying mortgage and being concerned about next payments in the state by 33.3% and 29.2%, respectively. Also, findings show that a higher rate of forbearance in a state was associated with a higher probability of experiencing mortgage difficulties (both delays and concerns) in the state. The associations were less significant than those for delinquency, which may be understood when the general difference between forbearance and delinquency is considered.

As for statewide labor market conditions, the results show that a 1% higher rate of unemployment in a state was linked with a 4.6% and 2.7% greater incidence of mortgage payment delays and concerns in the state, respectively. These findings can be interpreted as a negative relation between the pandemic-induced disruptions in the job market and individual borrowers’ mortgage payment difficulties. As for the statewide capacity of UI operation (measured by UI claims), we found an nonsignificant relationship between state capacity and individual borrowers’ experiences.

The spread of COVID-19 appears to have had a negative association with individual mortgage borrowers’ payment difficulties. We found that one additional cumulative case per 100 residents in a state was associated with a 1.9% and 1.7% higher probability of delaying and being concerned about mortgage payments in the state, which is consistent with other COVID-19 literature (Dettling & Lambie-Hanson, 2021).

Intersectionality of race and ethnicity and UI benefits

In this section, we consider Research Hypothesis 3a in which we hypothesized that among UI recipients in the pandemic minority borrowers would be more likely to experience mortgage payment difficulties than Whites despite the receipt of UI benefits. Table 3 presents a summarized estimation result from the models that included interaction terms between UI status and race and ethnicity. Most estimation results about the level 1 and 2 variables are consistent with Table 2. Thus, we focused our interpretation on the intersectionality between race and ethnicity and UI, whereby UI benefits may be common across racial groups and also moderated among specific groups.

Table 3.

Summarized multilevel mixed-effect logistic regression results with interaction terms between Unemployment Insurance (UI) status and race and ethnicity

Variable	Last month's ortgage delay		Next month's payment concerns
	Odds ratio	Sig.	Odds ratio	Sig.
UI status (Ref = UI applied but not received yet)
UI received	0.816	***	0.814	***
Race/ethnicity (Ref = non-Hispanic white)
Non-Hispanic black	1.687	***	1.511	***
Non-Hispanic A&PI	1.279	**	1.479	***
Non-Hispanic other	1.495	***	1.528	***
Hispanic	1.317	***	1.313	***
Interactions between UI status
× Non-Hispanic black	1.251	***	1.194	+
× Non-Hispanic A&PI	1.337	***	1.128
× Non-Hispanic other	0.997		0.982
× Hispanic	1.001		1.119
Constant	0.093	***	0.221	***
Number of observations	80,797		80,797

Full regression results are available upon request. Standard errors were clustered at the state level

UI Unemployment Insurance, A&PI Asians and Pacific Islanders, Ref reference

⁺p < 0.10; *p < 0.05; **p < 0.01; ***p < 0.001

We again found a positive UI effect of reducing mortgage delays during the COVID-19 pandemic. However, we found that the UI benefit decreased when the recipient was a person of color, particularly Black. That is, Blacks had a higher risk of mortgage payment difficulties than Whites despite the receipt of UI benefits. This finding is notable in that the intersectionality was strongly significant (p < 0.001) even after considering income, household size, and other level 1 variables. The lower level of UI benefit can be understood in that Black recipients may have had less money left to pay their mortgage because they were more exposed to other pandemic-induced expenses than Whites. For example, the unexpected COVID-19 pandemic exacerbated a long-lasting racial disparity in digital and technology literacy, imposing a higher cost on Blacks to manage telework and K-12 education. Other than Blacks, we found that Asians also appeared to experience a reduced benefit from the UI program, which ran counter to our expectation that they would be equal with Whites.

Turning to the results on concerns about future payments, we found consistent intersectionality between race and ethnicity and UI. We again found that UI lowered payment concerns, but the benefit was reduced only among Blacks. Unlike the result for payment delays, however, the significance of the intersectionality was marginal (p < 0.10). These findings from the interactions between race and ethnicity and UI largely support Research Hypothesis 3a.

Racial disparities in mortgage payment difficulties compounded by pandemic-induced socioeconomic and health hardships

In the COVID-19 pandemic, mortgage borrowers who lost their job or a part of their income have been prone to not only housing cost burdens but also other hardships such as limited access to food and mental health problems. Furthermore, compounded hardships may be more acute among some racial and ethnic groups than others. Thus, we tested Research Hypothesis 3b, in which we assumed that among UI recipients, Blacks would be more likely to experience mortgage payment difficulties compounded by additional pandemic-induced hardships than Whites.

To test the hypothesis, we examined mortgage payment difficulties accompanied by other health, social, and economic hardships due to the COVID-19 pandemic—loss of income, food insufficiency, and mental illnesses—which have been known to substantially overlap housing payment difficulties (Alhenaidi & Huijts, 2020; Mendez-Smith & Klee, 2020; Monte & O’Donnell, 2020; Park, 2021; Park & Ahn, 2022; Park & Kim, 2021; Park et al., 2022). To operationalize dependent variables, we built binomial variables for each pandemic-induced hardship (0 = no hardship, 1 = hardship): loss of income, food insufficiency, and mental illnesses. These binomial hardship variables were then combined with the existing mortgage payment variable, one at a time.

The upper panel of Table 4 shows that people of color, particularly Blacks, were much more likely to be behind on mortgages while also experiencing income loss, food insufficiency, or mental health problem. The receipt of UI appeared to lower the compounded types of mortgage delays, but the UI effect was reduced among Blacks except for the case of mortgage delay and mental illness. In the case of mortgage delays accompanied by food insufficiency, Black borrowers were 46.0% more likely to experience the doubled hardship than Whites. Though the receipt of UI lowered the risk of mortgage and food hardship, the UI benefit was reduced and largely offset among Blacks.

Table 4.

Summarized multilevel mixed-effect logistic regression results for mortgage payment difficulties compounded by pandemic-induced hardships, with interaction terms between Unemployment Insurance (UI) status and race and ethnicity

Last month's mortgage delay compounded with	Employment income loss		Food insufficiency		Mental health problem
	Odds ratio	Sig	Odds ratio	Sig	Odds ratio	Sig
UI status (Ref = UI applied but not received yet)
UI received	0.835	***	0.751	***	0.756	***
Race/ethnicity (Ref = non-Hispanic white)
Non-Hispanic black	1.568	***	1.460	***	1.428	***
Non-Hispanic A&PI	1.266	*	1.098		1.017
Non-Hispanic other	1.494	***	1.427	***	1.324	***
Hispanic	1.399	***	1.314	***	1.161	**
Interactions between UI status
× Non-Hispanic black	1.278	***	1.308	***	1.051
× Non-Hispanic A&PI	1.312	*	1.352	**	1.045
× Non-Hispanic other	0.993		1.045		1.094
× Hispanic	0.911		0.982		0.967
Constant	0.079	***	0.065	***	0.074	***
Number of observations	80,797		80,797		80,797

Next month's payment concerns compounded with	Employment income loss		Food insufficiency		Mental health problem
	Odds ratio	Sig.	Odds ratio	Sig.	Odds ratio	Sig.
UI status (Ref = UI applied but not received yet)
UI received	0.831	***	0.758	***	0.747	***
Race/ethnicity (Ref = non-Hispanic white)
Non-Hispanic black	1.390	***	1.227	**	1.155	+
Non-Hispanic A&PI	1.406	***	1.150		1.020
Non-Hispanic other	1.535	***	1.474	***	1.402	***
Hispanic	1.323	***	1.248	***	1.095
Interactions between UI status
× Non-Hispanic black	1.244	*	1.246	*	1.090
× Non-Hispanic A&PI	1.148		1.329	*	1.217	+
× Non-Hispanic other	0.976		1.026		1.018
× Hispanic	1.078		1.142	*	1.155	*
Constant	0.184	***	0.163	***	0.141	***
Number of observations	80,797		80,797		80,797

Full regression results are available upon request. Standard errors were clustered at the state level

UI Unemployment Insurance, HPS Household Pulse Survey, MSA Metropolitan Statistical Area, Ref reference

⁺p < 0.10; *p < 0.05; **p < 0.01; ***p < 0.001

Looking at the results on mortgage payment concerns compounded by added hardships, as shown in the lower panel of Table 4, we again found a significant and wide disparity among racial and ethnic groups. Compared to Whites, Blacks were more likely to experience mortgage payment concerns and income loss or food insufficiency. In addition, Blacks appeared to benefit from UI less than Whites. Hispanics also had a higher risk of experiencing mortgage concerns and food insufficiency at the same time and like Blacks, their expected benefits from UI were smaller than those of Whites. The results are in line with Research Hypothesis 3b in which we hypothesized that compared to Whites minority borrowers would have a higher probability of experiencing mortgage difficulties compounded by other hardships despite the receipt of UI benefits.

Robustness tests

This section describes robustness tests in which we tested if the empirical results in Table 2 (base models) are robust despite alternative choices regarding racial and ethnic subsamples, socioeconomically vulnerable subgroups, and the specification of dependent variables.

We extracted racial and ethnic subsamples (non-Hispanic (NH) White, NH Black, NH Asian and Pacific Islander, NH other, and Hispanic) and repeated the models in Table 2 to test the robustness of our results (see Supplemental Table 8). We found a consistent and significant role of the UI program in reducing mortgage payment difficulties (both delays and concerns) across racial and ethnic models except for Blacks. The coefficient of UI was not significant in the model with the Black subsample, which may be in line with the intersectionality between race and ethnicity and UI found in this study.

In addition to racial disparities, adults with children and low-income households have long been recognized as socioeconomically vulnerable subgroups in the mortgage market in terms of income gaps and difficulties due to household composition (Gür, 2022; Riley et al., 2015). They are also more likely to be people of color, and the overlap may have distorted our base results. We estimated the base models in Table 2 by using two separate subsamples: adults with a lower annual household income (< $50,000; n = 21,637) and adults with children (n = 34,196; see Supplemental Table 9). Estimation results on the subsamples were largely unchanged, supporting the consistent racial disparities in these vulnerable subpopulations.

The last test concerned the specification of mortgage payment difficulties (dependent variables). As described in the data and method sections, we tested two alternative outcomes to specify narrow and broad measures of mortgage experiences during the pandemic: one for both delay and concerns at the same time as the state of mortgage insecurity and the other for either delay or concerns (see Supplemental Table 10). The findings confirm that the model estimates were substantively unaffected.

Discussions and concluding comments

Unemployment and housing policy implications during COVID-19

This article found that during the COVID-19 pandemic, Black UI recipients were more likely to report delaying mortgage payments and being concerned about their next payment than Whites. Despite UI benefits, Blacks were found to experience mortgage payment difficulties accompanied by other pandemic-induced hardships such as loss of income, lack of food, and mental illnesses. These findings on the intersectionality between race and ethnicity and UI are consistent with previous literature showing that UI benefits, as nonhousing assistance, has helped mortgage borrowers pay their monthly principal and interest but the benefit was unequal among people of color.

The findings of this article may have implications for advancing race-conscious operation and management of UI program during and after the pandemic. The federal government rapidly conducted a historically large expansion of UI benefits to respond to the crisis. This inevitably resulted in administrative issues that affected racial and ethnic groups differently, such as an upsurge in applications, delays in insurance dissemination, excessive red tape, and overpayments (Ganong et al., 2022). In this article, we found that minority borrowers were significantly more likely to experience mortgage payment difficulties than Whites, even after the receipt of UI benefits. Given that UI payments are often used as a subsidiary measure to alleviate liquidity constraints among minority mortgage borrowers, a more race-sensitive UI system may shorten the long waitlist and expedite the dissemination of benefits to people with the most urgent needs. Our results do not necessarily imply the necessity of mortgage deferral for specific groups. Reciprocity toward certain population groups may involve reverse discrimination, unfairness, and social conflicts from another angle. Rather, the federal and state government may consider extending education programs and financial counseling for minority borrowers as an equal and race-sensitive intervention.

Our results also have implications regarding the multidimensional hardships experienced by minority borrowers. In the wake of the pandemic, the national forbearance mandate and foreclosure moratorium and the expanded UI program helped reduce financial distress for borrowers. More than 80% of mortgage borrowers who were behind on payments in early 2020 enrolled in forbearance. Minority borrowers, however, were still more likely to experience payment difficulties and less likely to benefit from refinancing than Whites (Gerardi et al., 2022). In the past 2 years of the pandemic, racial minorities lost their job or a part of their income more often than others and experienced subsequent constraints on their household budget that involved a wide range of expense difficulties. Also, the transmission of the coronavirus has been more prevalent among Blacks than other racial and ethnic groups. Our findings show that minority borrowers had a higher risk of experiencing both mortgage difficulties and other pandemic-induced hardships compared to Whites. As the pandemic abates, the geographic concentration of forbearance, foreclosure, and other health and socioeconomic problems may have lingering effects on the local housing market, such as foreclosure discounts, extensive subprime loans, and long-term vacancies (Barwick, 2010; Donner, 2020; Wang & Immergluck, 2019). Therefore, housing policymakers should prioritize assistance for minority mortgage borrowers, given that their experience of housing stress is intertwined with other hardships induced by the pandemic.

Limitations and future research directions

Although many studies have explored racial disparities in the housing and labor markets in the COVID-19 pandemic, the implications for UI benefits and housing security deserve further research. In this article, we emphasized understanding the intersectionality between race and ethnicity and UI across states.

An important gap in the literature may be the racial disparity in digital and financial literacy in the mortgage market. The use of mobile and online banking as a primary method of accessing personal bank accounts differs by race and ethnicity and may induce additional difficulties in paying mortgages in the context of the pandemic, when numerous brick-and-mortar bank branches were closed due to social-distancing measures (Broady et al., 2021). Also, the UI variable may not have represented the entirety of insurance recipients in each state, even if the data represented the total population and households. The sample size limitation requires careful interpretation and use of findings from this article when drawing state-level implications for the local mortgage market. Another research direction that would benefit from future examination is metropolitan-level analysis to develop more localized housing programs in the most populous and largest housing markets, such as Los Angeles and New York. Finally, future research based on longitudinal (or panel) data is crucial. Our HPS data was built by combining a series of weekly datasets and did not allow us to follow the same individuals over time. The survey data was also on a biweekly basis that did not perfectly match the monthly cycle of general mortgages. Based on these considerations, researchers may conduct further research on mortgage payment projections (e.g., Larrimore & Troland, 2020) by different racial and ethnic groups to predict how much more UI benefits are needed for which subpopulations.

Supplementary Information

Below is the link to the electronic supplementary material.