Gerardi, Lambie-Hanson, Willen, Lessons Learned from Mortgage Borrower Policies and Outcomes during the COVID-19 Pandemic (Boston Fed Current Policy Perspectives (2022-07-07), 2022-07-07)
Summary
A July 7, 2022 Current Policy Perspectives article from the Federal Reserve Bank of Boston research department, Lessons Learned from Mortgage Borrower Policies and Outcomes during the COVID-19 Pandemic, by Kristopher Gerardi, Lauren Lambie-Hanson, and Paul Willen. The authors review the aid offered to roughly 50 million homeowners with mortgages included in a forbearance program and the Federal Reserve's actions that lowered mortgage rates. They deem these policies quite effective in relieving financial distress and note that minority mortgage borrowers were much more likely to miss mortgage payments. The authors state that Black and Hispanic borrowers were less likely than white or Asian borrowers to refinance. The paper appears as Part I of Chapter 5 of Recession Remedies, published in 2022.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Cite this document
“Gerardi, Lambie-Hanson, Willen, Lessons Learned from Mortgage Borrower Policies and Outcomes during the COVID-19 Pandemic (Boston Fed Current Policy Perspectives (2022-07-07), 2022-07-07).”
Original at www.bostonfed.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/gerardi-lambie-hanson-willen-lessons-learned-from-mortgage-borrower-po-bd133e08b3/
, SHA-256 f02d9ee55345c7625c166c9ac611b0882c827fc106322e483df75ea0c6b3c1c8.Full text
Research Department
July 7, 2022
Current Policy Perspectives
Lessons Learned from
Mortgage Borrower Policies and
Outcomes during the COVID-19 Pandemic
Kristopher Gerardi, Lauren Lambie-Hanson, and Paul Willen
This article reviews the aid offered to the roughly 50 million homeowners with mortgages included in a
forbearance program, and the Federal Reserve’s actions that pushed down mortgage rates, allowing
many mortgage holders to reduce their monthly payments by refinancing. We deem these policies to
be quite effective in relieving financial distress and allowing homeowners to stay in their homes,
especially in contrast with the policies pursued during the Great Recession. We emphasize that these
policies in part worked because of rising housing prices and home equity, before and during the
pandemic, and note that such conditions might not hold in future downturns. We observe that minority
mortgage borrowers were much more likely to miss mortgage payments, so forbearance was
particularly important to them. Black and Hispanic borrowers, however, were less likely than white or
Asian borrowers to refinance.
Kristopher Gerardi is a financial economist and senior policy adviser in the research department at the Federal Reserve Bank of Atlanta.
Lauren Lambie-Hanson is a senior advisor and research fellow for the Consumer Finance Institute at the Federal Reserve Bank of
Philadelphia. Paul Willen (Paul.Willen@bos.frb.org) is a senior economist and policy advisor in the research department at the Federal
Reserve Bank of Boston.
The views expressed herein are those of the authors and do not indicate concurrence by the Federal Reserve Banks of Atlanta, Boston,
and Philadelphia; the principals of the Board of Governors;, or the Federal Reserve System.
The authors thank Wendy Edelberg, Raven Molloy, Louise Scheiner, Jenny Schuetz, and David Wessel for excellent feedback.
This paper appears as Part I of Chapter 5 of Recession Remedies: Lessons Learned from the U.S. Economic Policy Response to COVID-
19, edited by Wendy Edelberg, Louise Sheiner, and David Wessel and published in 2022 by the Hamilton Project and the Hutchins Center
on Fiscal and Monetary Policy at Brookings.
1 Introduction
The COVID-19 pandemic, which has proven to be the worst public health crisis in a century,
has caused significant distress in the mortgage market. Widespread job loss in the early
stages of the pandemic resulted in waves of missed mortgage payments. As figure 1 (panel
A) shows, the share of loans past due approached levels last seen during the global financial
crisis (GFC) and subsequent Great Recession more than a decade ago.
In this article, we detail how the most important policy responses to the pandemic af-
fected the mortgage market. In particular, we focus on the Coronavirus Aid, Relief, and
Economic Security (CARES) Act of 2020; the follow-on American Rescue Plan (ARP) Act
of 2021, which extended many of the provisions in the CARES Act; and the Federal Reserve’s
large-scale asset purchase (LSAP) program that was announced in March 2020. Our analysis
considers the aggregate effects and the distributional effects of these policies on US mortgage
borrowers. Although there are numerous ways to study the data, we will focus primarily on
documenting differences across racial and ethnic groups. This decision is motivated by the
fact that the COVID-19 virus disproportionately affected minority communities both as a
disease and as a disruptive economic force. During the pandemic, Black and Hispanic indi-
viduals faced elevated risk of infection, hospitalization, and death.1 Furthermore, minorities
experienced significantly worse labor market outcomes during the pandemic. For example,
the unemployment rate peaked in April 2020 at 16.7 percent for Black workers versus 14.1
percent for white workers. Even more concerning, though, unemployment stayed elevated
much longer for minority workers than for white workers as the economy healed.2 By Septem-
ber 2020 the white unemployment rate had fallen by more than half, to 7.0 percent, whereas
in March 2021, almost a year after the pandemic started, the Black unemployment rate was
still close to 10 percent (BLS (2022)). Although most of our focus is on documenting racial
disparities, we also look at differential policy effects across genders, household income levels,
and county unemployment levels.
The CARES Act included a national forbearance mandate, a foreclosure moratorium,
significantly expanded unemployment insurance (UI) benefits, and economic impact pay-
ments (EIPs) to most households. We show that, although minority mortgage borrowers
were much more likely to experience distress and miss mortgage payments, conditional on
missing payments, forbearance uptake was similar across racial and ethnic lines. Further-
more, we argue that these CARES Act policies were quite effective in alleviating financial
1
See Van Dorn, Cooney, and Sabin (2020) as well as Centers for Disease Control and Prevention (CDC)
data on hospitalizations and death rates by race and ethnicity (CDC (2019)).
2
For simplicity, we use “white” and “Hispanic” to refer to “non-Hispanic white” and “Hispanic white,”
respectively.
1
Figure 1: Mortgage Status and Interest Rates
Panel A: Mortgage Nonpayment and Foreclosure, 2005–Present
10
60+ Past Due,
All
8
As % of all loans
6
4
2
Foreclosure Starts In Forbearance
1
0
04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21
Year
Panel B: Interest Rates, 2007–Present
8
7
6
5
in %
4
3 ←F RM 30
2 ւMBS Yield
←10-yr CMT
1
0
2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020
Notes: Panel A: Mortgages past due and in forbearance are from McDash and McDash Flash data,
respectively. Forbearance shares include only loans 60 days or more past due and in forbearance.
Foreclosure starts reported by the Mortgage Bankers’ Association (MBA) National Delinquency Survey:
https://www.mba.org/news-research-and-resources/research-and-economics/single-family-research/national-
60+ Past Due includes all past-due loans, including loans in foreclosure. Panel B: FRM30 is the note rate
on a 30-year fixed-rate mortgage as measured by the Freddie Mac Primary Mortgage Market Survey. MBS
Yield uses data from JPMorgan Markets to compute the yield on a security containing a 30-year FRM
paying FRM30. 10-year CMT is the constant-maturity yield on a 10-year bond as reported in FRB H-15.
For details, see Fuster, Hizmo, Lambie-Hanson, Vickery, and Willen (2021).
Sources: McDash and McDash Flash Data; Fuster et 2 al. (2021); Mortgage Bankers’ Association Delinquency
Survey; authors’ calculations
distress at the outset of the pandemic and in preventing longer-run problems in mortgage
and housing markets.
The Federal Reserve’s LSAP program focused on improving market functioning and low-
ering long-term interest rates. Mortgage-backed security (MBS) purchases were a significant
component of the program, and Fuster et al. (2021) show that they indeed lowered mort-
gage rates and spurred a significant wave of refinancing. Although borrowers who were
enrolled in forbearance were unable to refinance, we show that a large fraction of borrowers
who remained current on their loans during the height of the pandemic took advantage of
the refinancing opportunity and significantly lowered their payments. Unlike the case of
forbearance, however, there were large differences in refinancing behavior across racial and
ethnic groups. We estimate that, through March 2021, only 10.6 percent of Black borrowers
refinanced as compared with 15 percent of Hispanic borrowers, almost 19 percent of white
borrowers, and 22 percent of Asian borrowers. After controlling for basic underwriting vari-
ables including credit score, loan-to-value ratio, income at origination, loan amount, as well
as the potential amount of refinance savings, we find that Black borrowers were 67 percent as
likely as white borrowers to refinance. These refinance disparities could be driven by several
factors including, but not limited to, differences in income/employment risk, differences in
health risk, or possibly discriminatory practices. Unfortunately, though our mortgage data
allow us to precisely document the disparities, they do not contain enough detail to identify
the underlying factors. Our hope is that as better data become available, future research
will pinpoint the causal mechanisms.
An alternative way to measure inequality in refinances is to look at the payment sav-
ings. In Gerardi, Lambie-Hanson, and Willen (2021), we estimate that the typical refinance
reduced the borrower’s monthly payment by about $280, leading to a payment reduction of
$5.3 billion per year for all households that refinanced in the first 10 months of 2020. Of
those savings, we estimate, only $198 million, or 3.7 percent, went to Black households, who
held 5.9 percent of mortgage debt in our sample. To put these numbers in perspective, Black
households account for 13.3 percent of the population and 9.1 percent of all homeowners.
Finally, we conclude the chapter with a discussion of some of the lessons that we believe
policymakers should take away from the pandemic experience. We argue that forbearance
was an especially effective policy in reducing borrower distress because of its timeliness,
high accessibility, and incentive compatibility. However, we also acknowledge that the stars
simply may have been aligned, as the state of the pre-pandemic housing and mortgage
markets and the dynamic of the pandemic itself set up almost perfectly for forbearance to
be an especially effective policy. Specifically, the rapid labor market recovery in the late
spring and early summer of 2020 meant that most borrowers needed only a few months of
3
assistance. In addition, the majority of outstanding mortgage debt (65 to 70 percent) was
insured by the US government going into the pandemic (Institute (2021)), including that
held by the most financially vulnerable segments of the market, and thus most financially
distressed borrowers had direct access to the forbearance policy mandated by the CARES
Act. Finally, we note that the housing market was exceptionally healthy as a result of years
of robust house price growth and low incidence of default and foreclosure, which meant that
most borrowers exiting forbearance were not in danger of being evicted from their homes.
So although we argue that forbearance should remain an important tool in the policy kit
going forward, it is unclear if it will be as effective in a future crisis.
Although forbearance was very effective in mitigating mortgage market distress, we argue
that the Federal Reserve’s LSAPs, implemented at the onset of the pandemic, had more mod-
est effects. Empirical evidence suggests that LSAPs lowered mortgage rates (Fuster et al.
(2021)) and spurred a refinancing boom in the spring and summer of 2020, however, most
borrowers experiencing pandemic-related financial distress were likely unable to refinance.
A first-order impediment was forbearance itself, as borrowers enrolled in a forbearance plan
were required to exit the plan and make three consecutive mortgage payments in order
to qualify for refinancing. Combined with the high fees associated with refinancing, this
requirement meant that many borrowers facing pandemic-related financial distress and liq-
uidity constraints were unable to exploit rate declines to lower their debt burdens. This
factor likely played a role in the large racial disparities in refinancing described above. We
offer a few suggestions to ensure that the benefits of lower mortgage rates reach a broader set
of borrowers in future downturns. These suggestions include the development and marketing
of alternative mortgage products that automatically lower payments when rates decline as
well as more widespread adoption of streamlined refinance programs that do not require
employment or income verification.
2 Data
For much of the analysis in this article, we track mortgage performance over time by borrower
race and ethnicity by combining several sources of anonymized data. These sources are Black
Knight McDash mortgage servicing data, Home Mortgage Disclosure Act (HMDA) data,
and two credit bureau data sets from Equifax: one from Credit Risk Insight Servicing data
linked to McDash data (known as CRISM) and the other from the Federal Reserve Bank
of New York/Equifax Consumer Credit Panel.3 The McDash data provide information on
3
See Gerardi, Lambie-Hanson, and Willen (2021) for more information on the matching procedures and
match rate.
4
loan performance, while the Equifax data allow us to observe other mortgages the borrowers
have and to determine if any mortgages are in forbearance. The HMDA data enable us to
identify the race, ethnicity, and gender of the borrower and to capture borrower income at
the time of underwriting. We focus on 30-year, fixed-rate, first-lien loans originated during
the 2010–2019 period. Loans originated during that period made up about 75 percent of
active accounts and 85 percent of active loan balances in 2019. We restrict our sample to
mortgages secured by owner-occupied, single-family homes and condos. We further limit
the sample to Federal Housing Administration (FHA) loans and conventional loans held
by Fannie Mae or Freddie Mac (government-sponsored enterprises, or GSEs). Although we
exclude portfolio and private-label securitized loans from our analysis, they make up less
than 35 percent of loans active during the pandemic. As An, Cordell, Geng, and Lee (2021)
show, the forbearance rates of portfolio loans were similar to those of GSE loans, and the
rates among private-label securitized loans were similar to those of FHA loans.
We supplement the matched data set with data from Optimal Blue to estimate the interest
rate that borrowers in our sample would likely receive upon refinancing.4 To do this, we use
the median interest rate locked each month by borrowers with similar credit scores and loan-
to-value ratios, as captured in the Optimal Blue database.5 We use CoreLogic Solutions
house price indices at the Zip code, county, and state levels to analyze recent trends in home
price appreciation for our mortgage sample and to calculate updated monthly loan-to-value
ratios and home equity accumulation.6
3 Mortgage Market Policy Responses to the COVID-
19 Pandemic
A key premise of the policy response to the COVID-19 pandemic was that alleviating financial
distress at the household level had desirable macroeconomic consequences. A summary
measure of the financial burden faced by a mortgage borrower is the debt-service ratio
(DSR)
4
Optimal Blue data, as referenced throughout this chapter, is anonymized mortgage market/rates data
that do not contain lender or customer identities or complete rate sheets.
5
We calculate the rate assuming the borrower pays zero points (and receives zero credits) from the lender
at closing. We observe the borrower’s credit score in month t in the CRISM data, and we estimate the
loan-to-value ratio of their mortgage by dividing its unpaid principal balance by the estimated value of the
home.
6
We do this by adjusting the property value at origination by the growth in the CoreLogic Zip code home
price index. We then use the CoreLogic county-level index for loans located in Zip codes for which CoreLogic
does not provide an index, and we use the state-level index if neither Zip code nor county data are available.
5
m
DSR = ,
y
where m and y are the mortgage payment and income, respectively. All else being equal,
an increase in the DSR makes a household worse off, suggesting that an increase can be
a signal of distress. Borrower responses to a higher DSR can also have negative spillover
effects, particularly when increases in DSRs are widespread across households. For exam-
ple, borrowers can reduce spending on nonhousing goods and services, reducing aggregate
demand. Or they can default on their mortgages and weaken the financial system. Finally,
borrowers can list their homes on the market and flood the market with unsold property.
Absent any policy intervention, the COVID-19 pandemic would have led to a massive fall
in income and a consequent increase in the DSR. To reduce financial distress, policymakers
took three actions early in the pandemic that affected the DSR. The first two—forbearance
and asset purchases—lowered mortgage payments (m), the numerator. The third, income
support programs, raised income (y), the denominator.
We now discuss details of the three policy interventions.
3.1 Forbearance
The CARES Act, passed by Congress and signed into law on March 27, 2020, instructed
lenders to allow borrowers to postpone payments for up to a year, later extended to 18
months, without incurring any penalty. Specifically, the CARES Act stipulated that any
borrowers who had mortgages insured by the federal government could enroll in forbearance
by simply attesting to financial hardship caused by COVID-19; households did not need to
document this hardship.7 While the CARES Act forbearance mandate formally applied only
to federally backed loans, which accounted for approximately 65 to 70 percent of the market
at the time, servicers of portfolio and private-label securitized mortgages also routinely
granted forbearance (An et al. (2021), Cherry, Jiang, Matvos, Piskorski, and Seru (2021)).
Figure 2 shows the stock of loans in forbearance from the first quarter of 2020 through
the third quarter of 2021. The gray area in the chart corresponds to loans that remained in
forbearance, while the colored areas correspond to the stock of loans that exited forbearance
in various ways. The stock of loans in forbearance peaked early in the pandemic, in the
second quarter of 2020, and subsequently slowly declined. The figure clearly shows that
the flows into forbearance were heavily concentrated during the first few months of the
7
Section 4022 of the CARES Act mandated that borrowers of federally backed mortgages could request
forbearance for up to 12 months. It further stated, “No fees, penalties, or additional interest will accrue on
the loan beyond what is scheduled” (sec. 4022 (b)(3)). In February 2021, the Biden administration extended
the CARES Act forbearance mandate through June 2021.
6
pandemic. More than 80 percent of borrowers in our sample who missed mortgage payments
in the first three months of the pandemic (April through June 2020) enrolled in forbearance,
suggesting that the policy helped most borrowers who experienced financial distress due to
the pandemic. Furthermore, previous research (Lambie-Hanson, Vickery, and Akana (2021))
shows that forbearance was concentrated among borrowers who were employed in hard-hit
industries before the pandemic, such as leisure, hospitality, arts, and entertainment, as well as
among households who had experienced a job disruption or income loss due to the pandemic.
Interestingly, approximately one-third of borrowers who enrolled in forbearance during this
period stayed current on their mortgage payments, which suggests that forbearance was
also widely used by nondistressed borrowers as a form of insurance against employment
uncertainty early in the pandemic.8
The CARES Act further stipulated that forbearance resulting from the pandemic could
not negatively affect a borrower’s credit score, which meant that lenders were not allowed to
report borrowers in forbearance as being delinquent on their payments. We show in section
3.3 that this stipulation largely prevented significant declines in the credit scores of borrowers
who missed payments.
The CARES Act also included a moratorium on foreclosures. Initially, the moratorium
was in effect only through May 17, 2020, but it was extended twice and finally expired on
July 31, 2021. For borrowers covered by CARES Act forbearance provisions, the moratorium
was largely irrelevant, because forbearance prevents any action by the lender against a past-
due borrower. However, the moratorium did help borrowers who had preexisting payment
problems stay in their homes.
3.2 Asset Purchases
The first mortgage market policy response to COVID-19 came from the Federal Reserve.
On March 3, 2020, the Federal Open Market Committee (FOMC) cut the fed funds target
rate by 50 basis points. Less than two weeks later, on March 15, the FOMC cut the rate by
an additional 100 basis points, taking it essentially to zero. In addition, on the same date,
the FOMC initiated large-scale purchases of both mortgage-backed securities (MBS) and
Treasury securities. It initially committed to purchasing at least $200 billion of MBS and
$500 billion of Treasury securities. Panel B of figure 1 shows that on March 20—following
these activities—the 10-year nominal Treasury rate fell below 1 percent for the first time,
and MBS yields also fell to historically low levels.
Mortgage rates also fell, though more slowly than Treasury rates or MBS yields. The
8
The fraction of borrowers in forbearance but current on their mortgage payments quickly declined to
trivial magnitudes in the second half of 2020.
7
Figure 2: Forbearance Outcomes with Exit Codes
5000
In Forbearance
Other
# Loans in thousands
4000
No plan
Paid Off
3000 Modified
2000 Payment Deferral
Reinstated
1000
No Missed
Payments
0
2020 Q1 2020 Q2 2020 Q3 2020 Q4 2021 Q1 2021 Q2 2021 Q3
Note: This figure uses data from the Mortgage Bankers Association’s (MBA) Weekly Forbearance Survey
to measure the cumulative outcomes of all loans that entered forbearance at some point. “In forbearance”
measures the stock of all loans in forbearance at a moment in time. All other data series are cumulative
exits. Note that the MBA cannot distinguish between new entrants and reentrants to forbearance. Thus
the same loan may appear twice in the sample. For example, most of the loans that exited with “No plan”
most likely reentered forbearance, meaning that at the end of the sample, many loans appear in both “No
plan” and “In forbearance.” The MBA also does not track loans after the end of forbearance, so for example,
many of the “no missed payments” loans may have refinanced after exit but will not show up in the “paid
off” category. The size of the surveyed universe varies but is typically about 38 million loans per week. A
small number of borrowers exited using a repayment plan, and those are included in the “other” category.
The first week of data is from May 31, 2020
Source: Mortgage Bankers’ Association (MBA) Weekly Forbearance Survey; authors’ calculations
8
Freddie Mac Primary Mortgage Market Survey (PMMS) 30-year fixed-rate mortgage rate fell
at the beginning of March, reaching a historic low of 3.29 percent. However, disruptions in the
MBS market caused the PMMS rate to rise later in the month. Fed interventions in the MBS
market meant that rates fell again in the beginning of April. However, as documented by
Fuster et al. (2021), capacity constraints among originators meant that the spread between
the primary market rates charged by originators and rates in the MBS market remained
wide for an extended period, as illustrated in panel B of figure 1. Rates were historically low
but most likely about 20 or 30 basis points higher than they would have been in the absence
of binding capacity constraints in the mortgage origination industry, driven by a shortage of
qualified workers and operational frictions such as how to complete appraisals and closings
while maintaining social distancing.
Not surprisingly, historically low interest rates led to a wave of refinancing. In March
2020, the Mortgage Bankers Association (MBA) refinance index increased to its highest level
in more than a decade and remained elevated throughout the entire year.
3.3 Income support
In addition to directly affecting the mortgage market through forbearance, the CARES Act
indirectly affected the market through direct payments to households to make up for income
lost as a result of the pandemic. From the standpoint of households, the main program was
the expanded provision of UI. The key UI-related terms of the CARES Act included expanded
coverage to nonsalaried workers who normally do not qualify for UI as well as a supplemental
payment of $600 per week per household. Panel A of figure 3 shows that, starting in May
2020, expanded UI was enough to ensure that aggregate personal income remained at or
above its pre-pandemic trend for almost every month through February 2022. Additional
income support programs, including the Paycheck Protection Program (PPP), meant that,
in fact, personal income exceeded its pre-pandemic trend throughout most of that period.
Panel B of figure 3 shows that, during the GFC and subsequent Great Recession, income
support programs did not play a similar role. The 2008 stimulus program did lead to an
increase in income in May and June 2008, but government assistance from September 2008
to March 2009—the acute phase of the Great Recession—was minimal. Congress passed the
American Recovery and Reinvestment Act of 2009, which expanded UI and provided other
stimulus, but those measures do not compare to the relief provided by the CARES Act and
subsequent legislation. During the Great Recession, personal income never returned to its
pre-crisis trend.
9
Figure 3: Personal Income after the Great Recession and the COVID-19 Recession
Panel A: COVID-19 Pandemic
24
22 Unemployment
in trillions of
Insurance
Other
Stimulus
$, SAAR 20
Time Trend
18 Personal Income
less UI and
Other Stimulus
2018 2019 2020 2021
Panel B: Great Recession
14
Other Stimulus Time Trend
13
in trillions of
Unemployment
Insurance
$, SAAR
12
Personal Income
less UI and
Other Stimulus
11
2007 2008 2009 2010
Note: This figure shows data on personal income and its components from the monthly “Personal Income
and Outlays” release from the US Bureau of Economic Analysis. “Other stimulus” is “Government social
benefits to persons: Other.” Time trends are based on linear regressions of personal income estimated over
the period from January 2016 through February 2020, and the period January 2004 through December 2007
for the COVID-19 and Great Recession periods, respectively.
Source: US Bureau of Economic Analysis, data from 2006 to 2022
10
Figure 4: Evolution of the Mortgage Debt Service Ratio during the Pandemic
Panel A: Decomposition of the Mortgage DSR
Mortgage Debt Service Ratio in %
0.2
0.1
(1) Income Loss
0.0
(2) Forbearance
-0.1 (3) Interest Rate
Reductions
-0.2 (4) Income
Support
-0.3 ← Mortgage DSR
-0.4
-0.5
-0.6
-0.7
2020 Q1 2020 Q2 2020 Q3 2020 Q4 2021 Q1 2021 Q2
Year
Panel B: Sources of Variation in Mortgage Debt Service Costs
4.2
Note Rate
4.0
in %
3.8
COVID-19
Pandemic
3.6
240
230
in $ thous.
Principal
Balance 220
210
200
2018 2019 2020 2021
Note: In panel A, the four shaded areas represent deviations from a linear pre-pandemic trend. Panel B
shows an unweighted average of the note rate and principal balance on all outstanding mortgages in the
McDash data set.
Sources: Federal Reserve Board; McDash data; authors’ calculations
11
3.4 Outcomes
Policy was clearly successful at reducing household financial distress caused by income losses
due to the pandemic. The heavy, solid black line in panel A of figure 4 shows the mortgage
DSR, which the Federal Reserve Board defines as the ratio of scheduled mortgage payments
relative to personal disposable income from the National Income and Product Accounts.9
The figure shows that the DSR fell by about 55 basis points or roughly 13 percent over the
four quarters from the first quarter of 2020 to the first quarter of 2021.
Why did the mortgage DSR fall during the crisis? Panel A of figure 4 shows the results
from a series of counterfactual experiments we conducted to illustrate how policy improved
household budgets. Starting from the top, the area labeled “(1) Income Loss” shows what
would have happened without any direct assistance from the government. The DSR would
have risen by about 20 basis points and then drifted down as the economy recovered. Our
next counterfactual isolates the effect of policy by asking what would have happened if in-
come had remained at its pre-pandemic level and borrowers had benefited from the policy
changes. The area labeled “(2) Forbearance” shows that forbearance would have lowered the
DSR initially by about 20 basis points. Panel A of figure 4 shows that early in the pandemic,
forbearance and income loss were roughly the same size, which leads to a crucial point: For-
bearance alone was able to roughly offset the effects of the pandemic if we measure financial
distress using the DSR. Our next counterfactual exercise adds interest rate reductions while
holding income constant. The area labeled “(3) Interest rate reductions” shows that they
had an effect similar to that of forbearance in overall magnitude. However, the timing of
the benefits of forbearance was quite different from the timing of interest rate reductions
benefits. The benefits of forbearance were front-loaded and played only a small role by the
spring of 2021, whereas interest rate reductions’ effect was small initially but grew over time.
Our final counterfactual experiment consists of adding income support programs to in-
terest rate reductions and forbearance, still holding income constant at pre-pandemic levels.
The area labeled “(4) Income support programs” shows that income support programs had
a bigger effect on the DSR compared with forbearance and interest rate cuts combined in
all but one quarter of the pandemic.
Overall, panel B of figure 4 illustrates that the multipronged assault of different parts of
the CARES Act and monetary policy meant that, using the DSR as a measure, households
were actually better off after the start of the pandemic than they were before it. Either
forbearance alone or income support programs alone would have been enough to blunt the
effects of the job and income losses associated with the pandemic. Of course, it is important
9
For details see https://www.federalreserve.gov/releases/housedebt/about.htm.
12
to stress that our analysis ignores any general equilibrium effects of the policies. For exam-
ple, without forbearance, many households would have cut spending, which would have, in
equilibrium, affected the time path of household income.
It is perhaps somewhat surprising that the effects of the interest rate reductions were
so small. The bottom panel of the figure shows that the average mortgage rate paid by
borrowers did, in fact, fall significantly, dropping by 60 basis points, or about 15 percent,
over the pandemic period. But several factors prevented lower rates from translating into
correspondingly large reductions in monthly payments. The bottom panel shows that lower
rates were offset by an acceleration in mortgage balance growth. In addition, some refinancers
took advantage of exceptionally low rates on 15-year mortgages and, as a result, had higher
payments despite paying less interest.
3.5 What Happened to Borrowers in Forbearance?
Forbearance is fundamentally different from interest rate reductions and income support.
Interest rate reductions and UI do not need to be paid back; forbearance does. An important
concern of policymakers was that, when forbearance ended, borrowers would have to quickly
repay the arrears they had accumulated. The institutional evidence and the data suggest
that this was not a major problem. On the institutional side, the main government lending
programs did not demand immediate repayment of arrears but rather offered a waterfall
of options. First, lenders offered to convert arrears into a non-interest-bearing second lien
due on termination of the loan. This payment deferral option meant that the borrower
could resume making monthly payments as if they had not missed any payments, meaning
a restoration of the pre-pandemic status quo, at least as far as cash flow was concerned. If
the borrower had suffered a permanent reduction in income due to the pandemic, lenders
could then offer a modification of the existing loan in addition to payment deferral.
The data show that, for the most part, the waterfall worked as intended. Figure 2
uses data from the MBA Weekly Forbearance Survey (MBA) to track the evolution of all
loans that entered forbearance, including loans that exited and then reentered forbearance.
According to the MBA, there were about 5 million entries into forbearance. As of October
2021, about 1 million loans remained in forbearance. What happened to the rest? About
700,000 loans had no plan, meaning that forbearance expired without the borrower making
contact with the servicer to explore options. Although we cannot be sure, we think most
of those loans subsequently reentered forbearance because data from Black Knight show
that, starting in the fall of 2020, most entries into forbearance were, in fact, reentries.
Another large exit category, especially in 2020, was borrowers who requested forbearance
13
but never actually used it and exited with no missed payments. In addition, a significant
number of borrowers had missed only a small number of payments and were reinstated after
repaying those missed payments. But, overall, most exits involved either a payment deferral,
a modification, or a combination of the two.
3.6 Forbearance and Credit Scores
The CARES Act of 2020 includes language that protects borrowers who choose to use forbear-
ance from experiencing a negative impact on their credit scores. Specifically, the legislation
says that if a borrower is in forbearance, the lender must report the loan as current to the
credit bureaus (CARES Act 2020, sec. 4021). This stipulation dramatically affected the
credit scores of borrowers who missed mortgage payments during the pandemic.
In February 2010, about 90 percent of past-due borrowers of FHA and GSE loans had
credit scores (from Vantage 3.0) below 622, whereas the 90th percentile for past-due bor-
rowers in February 2021 was 788, a super-prime score. The majority of the latter borrowers
began missing payments in April and May 2020 and used forbearance under the CARES
Act, which enabled them to avoid the serious damage to their scores that would normally
accompany missing months of mortgage payments.
This difference in the distribution of credit scores is also partly a product of stricter
underwriting in the aftermath of the GFC. Specifically, the 90th percentile score among
borrowers current or up to 30 days past due was 812 in February 2010, as compared with 824
in February 2021. It is also possible that because the pandemic caused a very large swath
of borrowers to become unemployed, nonpayment in the pandemic was less concentrated
among low-score borrowers than it was in the GFC. Even if not driven entirely by the role
of forbearance in protecting distressed borrowers’ credit scores, the fact that VantageScores
of distressed mortgage borrowers were significantly higher a year into the pandemic than
they were in the last crisis has important implications. It suggests that borrowers exiting
forbearance should have more robust access to consumer credit markets and a greater ability
to tap their housing wealth. In addition, borrowers who have not been able to cure their
distress and are forced to sell will likely face an easier return to future home ownership
compared with similarly distressed borrowers a decade ago.
4 Distributional Impacts of Mortgage Policies
We now turn to a discussion of the distributional effects of the policies. In particular,
we focus on differences in outcomes by race and ethnicity as well as by household income,
14
household composition, and the growth in county-level unemployment rates. For our analysis
of race/ethnicity, we use information from HMDA and construct indicators for Black, white,
Asian, and Hispanic borrowers. For our income analysis, we use HMDA income, which is
reported by borrowers when they file their loan applications, along with US Census Bureau
data on metro area income. We then compute an indicator variable for whether a borrower
meets the US Department of Housing and Urban Development’s definition of either low or
moderate income.10
4.1 Mortgage Nonpayment and Forbearance
The top panels of figures 5, 6, and 7 display monthly, unconditional nonpayment rates for
federally insured mortgages from January 2019 through the end of our sample in October
2021, broken down by borrower race/ethnicity, whether household income falls in the low
or moderate category, and the amount by which unemployment increased in the borrower’s
county early in the pandemic. We use a 60-plus days past due (DPD) definition of non-
payment (in other words, at least two missed payments), which is common in the mortgage
default literature. The figures correspond to the stock of mortgage nonpayments (or the
share of active mortgages that are at least 60 DPD in each month).
The differences across race/ethnicity in the pattern of nonpayment hazards are striking.
Nonpayment rates spike for all borrowers beginning in May 2020 shortly after the onset of
the pandemic, but the increase is significantly larger for borrowers of color.11 Black borrow-
ers experienced the most distress; their nonpayment rate rose from around 3 percent just
before the pandemic to 13 percent in mid-2020. Hispanic and Asian borrowers experienced
a similarly sharp rise in nonpayments, from 1 percent to 11 percent and from 1 percent to 8
percent, respectively. White borrowers experienced less distress; their nonpayment rate rose
from 1 percent to 6 percent.
The time-series pattern of the stock of nonpayment rates in panel A of figure 5 suggests
that mortgage distress was concentrated almost entirely within a two- to three-month period
at the beginning of the pandemic. Indeed, new mortgage nonpayments for all borrowers
spiked in May 2020 and remained elevated in June, but they quickly declined in July. New
nonpayments flattened afterward at levels that were slightly more elevated relative to their
pre-pandemic levels. The fact that we see the stock of 60 DPDs remaining extremely elevated
through the end of the sample, despite the flows into nonpayment receding in the summer
10
HUD’s definition of low income corresponds to household income being less than or equal to 50 percent
of area median income, and its definition of moderate income corresponds to income that is greater than 50
percent but less than 80 percent of area median income.
11
The spike in 60 DPD in May 2020 corresponds to borrowers missing their first payment at the beginning
of April and their second payment in May.
15
Figure 5: Households Past Due on Mortgage Payments and in Forbearance, by
Race/Ethnicity
Panel A: More than 60 Days Past Due
14%
Black
12%
Share 60+ Days Past Due
10% Hispanic
8%
Asian
6%
White
4%
2%
0%
Panel B: In Forbearance Given 60 Days or More Past Due
100%
80%
Share in Forbearance
60%
40%
20%
0%
Sources: Home Mortgage Disclosure Act (HMDA) data; Equifax Credit Risk Insight Servicing-McDash data;
Federal Reserve Bank of New York/Equifax Consumer Credit Panel (CCP) data; authors’ calculations.
Borrower race and ethnicity are captured in HMDA. Forbearance indicators are derived using tradeline-level
data from the CCP.
16
Figure 6: Households Past Due on Mortgage Payments and in Forbearance, by Income
Panel A: More than 60 Days Past Due
8%
7%
6%
Share 60+ Days Past Due
5% Above moderate-
income
4%
3%
Low- or moderate-
2% income
1%
0%
Panel B: In Forbearance Given 60+ Days Past Due
100%
80%
Share in Forbearance
60%
40%
20%
0%
Note: Borrowers are classified as low- or moderate-income (LMI) if their real income at origination (measured
in 2021 dollars) is less than 80 percent of the 2021 median family income in their metro area (or state, for
borrowers outside metro areas).
Sources: Home Mortgage Disclosure Act (HMDA) data; Equifax Credit Risk Insight Servicing-McDash data;
Federal Reserve Bank of New York/Equifax Consumer Credit Panel (CCP) data; authors’ calculations.
Borrower income is captured in HMDA data.
17
Figure 7: Households Past Due on Mortgage Payments and in Forbearance, by Unemploy-
ment
Panel A: More than 60 Days Past Due
8%
7%
Top quartile increase
in unemployment
Share 60+ Days Past Due
6%
5%
4% Bottom quartile increase
in unemployment
3%
2%
1%
0%
Panel B: In Forbearance Given 60+ Days Past Due
100%
80%
Share in Forbearance
60%
40%
20%
0%
Note: Borrowers are classified as top-quartile if their county’s unemployment rate increased by more than
the 75th percentile of counties nationwide (10.6 percentage points) from February to April 2020. Bottom-
quartile borrowers resided in counties with unemployment rates that increased by fewer than 4.9 percentage
points during this period. Nonpayment indicators are derived using McDash data; forbearance is derived
from Equifax data.
Sources: Home Mortgage Disclosure Act (HMDA) data; Equifax Credit Risk Insight Servicing-McDash data;
Federal Reserve Bank of New York/Equifax Consumer 18Credit Panel (CCP) data; Bureau of Labor Statistics;
authors’ calculations.
of 2020, suggests that many borrowers who experienced distress at the beginning of the
pandemic were unable to quickly resolve their financial difficulties. We show below that
most of those borrowers obtained relief with the CARES Act forbearance policy and that
many remained in forbearance through the end of our sample.
Figures 6 and 7 (Panel A) clearly show that minority borrowers and lower-income bor-
rowers experienced significantly higher levels of mortgage distress compared with white bor-
rowers and higher-income borrowers, respectively, during the pandemic. This observation is
not surprising, given the fact that job loss was significantly higher for minority households
and that sectors characterized by lower-paying jobs such as leisure and hospitality were af-
fected more by the lockdown and social-distancing measures implemented in response to
the pandemic. Panel A in figure 8 shows that counties with top-quartile increases in the
unemployment rate from February to April 2020 experienced significantly higher nonpay-
ment rates compared with counties in the bottom quartile during the same period, which is
consistent with the idea that employment losses from the pandemic created a lot of financial
distress for some mortgage holders—despite expanded unemployment insurance benefits. We
now turn to an analysis of forbearance, the primary policy response to the distress in the
market, to see if it had a differential impact across racial/ethnic lines or across borrowers
with low versus high incomes.
The bottom panels of figures 5, 6, and 7 plot forbearance rates by race/ethnicity, by in-
come group, and by unemployment growth groups. Importantly, the figures show forbearance
rates conditional on the borrowers being behind on payments so that the large differences
in nonpayment rates do not influence the forbearance differences. Conditional on the bor-
rowers being past due on payments, similarly high fractions of minority and white borrowers
were enrolled in forbearance plans. For example, as of August 2020, 84 percent of all white
borrowers who were 30-plus DPD were enrolled in forbearance, compared with 88 percent
of Asian borrowers, 83 percent of Black borrowers, and 87 percent of Hispanic borrowers.
We see similar forbearance enrollment rates across the income distribution: Borrowers with
low or moderate incomes were only slightly less likely to enroll in forbearance compared
with higher-income borrowers. Finally, figure 7 shows that conditional forbearance rates are
nearly identical across counties with top-quartile versus bottom-quartile increases in unem-
ployment rates. Thus, although minority and low-income borrowers were much more likely
to miss payments during the pandemic relative to white and high-income borrowers, those
who missed payments were approximately equally as likely to take advantage of payment
relief offered through forbearance.
19
4.2 Refinancing
Using pre-pandemic data, Gerardi, Willen, and Zhang (2020) show that racial disparities in
refinance behavior are significantly exacerbated during periods of low interest rates and high
refinance volume. Since the pandemic was characterized by both historically low mortgage
rates and significant refinance activity, we might expect to find similarly large disparities
during this period.
Figure 8 (panel A) shows the evolution of refinance propensities during the pandemic
by plotting monthly, unconditional refinance rates for different racial and ethnic groups.
Refinance rates were similar across all groups in the first couple of months of 2020, before the
onset of the pandemic. Beginning in March 2020, however, a significant gap between white or
Asian borrowers and Black or Hispanic borrowers emerged. Asian borrowers had the highest
refinance propensities during the pandemic, while Black borrowers were the least likely to
refinance. Notably, the racial gaps in refinance activity persisted during the pandemic.
Panel B in the figure displays refinance rates for loans taken out by single male borrowers,
single female borrowers, and multiple borrowers. Panel C shows refinance propensities for
low- and moderate-income borrowers and higher-income borrowers. Finally, panel D shows
refinance hazards for loans originated in counties with top- and bottom-quartile increases in
the unemployment rate during the pandemic.
Refinance rates were significantly higher for loans with multiple borrowers during the
pandemic period compared with loans with only a single borrower. Among single borrowers,
males were slightly more likely to refinance than females. Although the difference in refinance
rates between higher-income and low-to-moderate-income borrowers was small in the pre-
pandemic period, higher-income borrowers were approximately twice as likely to refinance
during the pandemic. Agarwal, Chomsisengphet, Kiefer, Kiefer, and Medina (2020) also find
significantly lower refinancing activity among low-income borrowers, and we find them less
likely to apply. Differences in refinance propensities among loans in high-unemployment and
low-unemployment growth counties were small.
While Figure 8 shows unconditional refinance rates, the sizes of the disparities are not
materially affected if refinance rates are conditioned on observable borrower and loan charac-
teristics such as credit scores, whether the borrower has been current on mortgage payments,
loan-to-value ratios, the incentive to refinance (how much the borrower’s rate differs from
what is available in the market), and geographic location.12 That is, differences in loan
12
For more details about how controlling for observables affects refinance disparities, see
Gerardi, Lambie-Hanson, and Willen (2021). This finding is contrary to those of Gerardi, Willen, and Zhang
(2020), who show that differences in observable characteristics can account for approximately 80 percent of
the unconditional refinance gap between Black and white borrowers.
20
Figure 8: Share of Borrowers Who Refinanced Their Mortgage, February 2019–June 2021
Panel A: By Race Panel B: By Household Composition
3.0%
2.5%
Asian
Co-Borrowers
Share Refinanced
2.0%
White
1.5%
Hispanic Single Women
1.0%
Black
Single Men
0.5%
0.0%
Panel C: By Income Panel D: By Unemployment
3.0%
2.5%
2.0%
Share Refinanced
Above moderate Top quartile
Income increase in
unemployment
1.5%
1.0%
Bottom quartile
Low or moderate increase in
Income unemployment
0.5%
0.0%
Note: Panel C: Borrowers are classified as low- or moderate-income (LMI) if their real income at origination
(measured in 2021 dollars in HMDA data) was less than the 2021 median family income in their metro
area (or state, for borrowers outside metro areas). Panel D: Borrowers are classified as top quartile if
their county’s unemployment rate increased by more than that of the 75th percentile of counties nationwide
(10.6 percentage points) from February to April 2020. Bottom-quartile borrowers resided in counties with
unemployment rates that increased by fewer than 4.9 percentage points during this period.
Sources: Home Mortgage Disclosure Act (HMDA) data; Equifax Credit Risk Insight Servicing-McDash
(CRISM) data; Bureau of Labor Statistics; authors’ calculations. Borrower race, ethnicity, gender, and
income are captured in HMDA. Refinance indicators are derived using data from CRISM.
21
or borrower characteristics included in our data do not explain the difference in refinance
rates by group. An important factor that we cannot observe is how a borrower’s income
and employment status change over time. Black and Hispanic households lost their jobs at
higher rates during the pandemic, which likely contributed to the disparities in their ability
to refinance.
The racial disparities in refinance activity documented in figure 8 are significant and
lead to large differences in how the total benefits from the lower interest rate environment
are shared. Those total gains are a function of the probability that a borrower refinances
and how much borrowers who do refinance save. Gerardi, Lambie-Hanson, and Willen (2021)
find that the mean monthly payment reductions for borrowers who refinanced were generally
similar across groups. White borrowers generally had lower existing interest rates, which
lowered their gain from refinancing, but they also had bigger mortgages, which worked in
the opposite direction. Annualizing the savings and multiplying them by estimates of the
number of mortgages held by each racial and ethnic group, we estimate that US homeowners
who refinanced through October 2020 will save about $5 billion a year until they refinance
again or sell their homes. We estimate that Black homeowners account for only $198 million,
or 3.7 percent, of the savings despite holding roughly 5.9 percent of balances in our mortgage
sample. In contrast, white borrowers account for approximately 71.1 percent of the savings
($3.8 billion), which is a slightly larger percentage than their sample share (69 percent).
5 Lessons Learned
Mortgage borrowers, like all Americans, experienced significant turmoil during the COVID-
19 pandemic. According to the MBA, the percentage of mortgage borrowers who were past
due peaked at 6.7 percent in the second quarter of 2020. That rate fell to 3.5 percent
of mortgage borrowers in the fourth quarter of 2021, almost a 50 percent reduction in six
quarters. To put that in perspective, after the GFC, the rate peaked at 10.2 percent in the
first quarter of 2010 and took until the third quarter of 2014 for the past due rate to fall by
50 percent, roughly three times as long. What role did policy play in those outcomes? What
went right, and what went wrong? How important was it that the nature of each downturn
was so different? Does success in this episode provide us with a road map, or even useful
insights, for the future? Can we say that the policies targeting homeowners were a success?
We now review, in turn, the three policy levers: forbearance, interest rate reductions and
income support.
Forbearance was especially effective due to its timeliness and the ease with which bor-
rowers were able to take advantage of it. Unlike with the Home Affordable Modification
22
Program (HAMP), the primary mortgage market policy enacted in the aftermath of the
GFC, enrolling in forbearance required zero documentation on the part of borrowers and
only minimal contact with mortgage servicers. Borrowers simply had to contact their ser-
vicer and attest to experiencing financial hardship due to the pandemic. Thus, whereas the
HAMP program took about a year to get up and running at full capacity, forbearance was
heavily used almost immediately.
Furthermore, forbearance, unlike modifications and principal reduction, is incentive com-
patible, meaning it is most attractive to those who really need it: financially distressed
borrowers. The reason is that forbearance requires borrowers to pay back their missed
payments and thus does not significantly lower the net present value (NPV) of payment
obligations. The emerging empirical evidence on forbearance usage suggests that it was, in
fact, used by the borrowers who needed it the most, with little evidence that it was used
strategically by nondistressed borrowers. Employing a survey of more than 1,000 homeown-
ers, Lambie-Hanson, Vickery, and Akana (2021) find that borrowers who used forbearance
overwhelmingly had suffered a job loss or income disruption during the pandemic. They also
show that forbearance was concentrated among borrowers who were employed pre-pandemic
in industries hard hit by COVID-19, including leisure, hospitality, arts, and entertainment.
An additional piece of evidence that forbearance targeted borrowers in need is that as finan-
cial distress waned over the course of the pandemic, so did forbearance usage. Forbearance
was used most intensively in the second quarter of 2020, when labor-income losses were most
significant.
Due to incentive compatibility, forbearance contrasts favorably with the concessionary
loan modifications used to assist borrowers during the GFC. The most common loan modifi-
cations reduced interest rates, thereby significantly lowering the NPV of payment obligations
and making them appealing to distressed borrowers as well as nondistressed borrowers. Stud-
ies such as Mayer, Morrison, Piskorski, and Gupta (2014) present evidence that this moral
hazard was a nontrivial issue for some of the modification programs rolled out in the after-
math of the GFC. To avoid modifying loans for borrowers not in need, lenders demanded
extensive documentation of hardship yet still foreclosed on many borrowers even when it was
more costly to foreclose than to modify.13 In addition, due to the complexity of dealing with
these information problems, the flagship HAMP did not really start to make a difference
until several years after policymakers identified a foreclosure problem in the United States.
As our data show, forbearance was helping borrowers at the beginning of April 2020, days
13
See Adelino, Gerardi, and Willen (2013) for a discussion of why information asymmetries lead rational
lenders to foreclose rather than modify loans even when the loss from foreclosure exceeds the reduced NPV
from modification.
23
after Congress passed the CARES Act and even before expanded unemployment insurance,
which did not start to flow in earnest until May 2020.
Supporting these distressed borrowers also had spillover effects on their communities.
Normally, increases in area unemployment and corresponding negative income shocks would
lead to more houses being put up for sale, which pushes down prices. But Anenberg and Scharlemann
(2021) show that forbearance offset pandemic-related increases in unemployment, decreasing
the number of new for-sale listings and propping up county-level home prices.
As a result, one might conclude that policymakers should have turned to forbearance in
2008 and should do so in any future economic downturn. However, three important points
should be considered before settling on such a conclusion. First, forbearance is not costless.
Put simply, lenders are effectively extending interest-free loans to borrowers, which is costly
even in a low interest rate environment.
Second, although the government insures investors against any missed payments of in-
terest and principal on MBS, there is a lag between missed payments by borrowers and
insurance payments by the government. Loan servicers are contractually obligated to cover
this gap and can find themselves in a liquidity squeeze.14 Indeed, a sufficiently high rate
of forbearance could bankrupt mortgage servicers.15 To address this risk, federal agencies
changed their reimbursement policies in March and April 2020. Fannie Mae reduced the
number of months that servicers were responsible for covering missed payments from twelve
to four. Ginnie Mae set up the Pass-Through Assistance Program (PTAP), an emergency
credit facility that servicers could access to fund payments. In the end, lower-than-expected
forbearance uptake and an increase in highly profitable refinance activity meant that ser-
vicers had ample liquidity throughout the pandemic. However, if policymakers consider a
broad-based forbearance policy in response to a future crisis, servicer liquidity risk could
resurface as a first-order concern. The third point to keep in mind before concluding that
forbearance is a panacea is that certain features of the pandemic likely made a policy of
broad-based forbearance particularly advantageous. First, due to the extremely rapid jobs
recovery in the late spring and summer of 2020, many distressed borrowers who had lost
their jobs needed only a few months of assistance. Most recessions, especially the Great
Recession, are characterized by much longer labor market recoveries. Second, most mort-
gages were federally insured, making risks to private investors minimal. At the start of the
14
Before the pandemic period, Fannie Mae required servicers to forward principal and interest payments
for twelve months for loans in forbearance, while Freddie Mac required four months of advances before
reimbursement could occur. For Ginnie Mae loans, servicers were expected to forward mortgage-related
payments for the entire life of the loan.
15
This was especially true for the nonbank mortgage companies (NBMCs), which are primarily funded by
short-term wholesale debt, exposing them to greater liquidity and run risk than banks. NBMCs accounted
for the majority of loan originations (approximately 70 percent) in the prepandemic period.
24
pandemic, 62 percent of mortgages by value were held in Fannie Mae, Freddie Mac or Ginnie
Mae MBS, meaning that the US Treasury effectively guaranteed repayment of principal and
interest. By contrast, before the GFC, the comparable figure was 43 percent.
Finally, perhaps the most important reason forbearance was so successful was the strong
pre-pandemic housing market and, specifically, the robust house price growth that most areas
of the country experienced in the years before the pandemic and, more importantly, during
the pandemic. Among borrowers in our sample whose loans were still active in February
2020, the median house price appreciation in their area over the next year was 9.8 percent,
and the average was 10.2 percent. And house price growth was widespread, as even the
10th percentile of the growth distribution in our sample experienced more than 5 percent
appreciation during the first year of the pandemic. Strong house price growth before and
during the pandemic translated into significant amounts of accumulated housing wealth for
borrowers. We estimate that the median borrower in our sample had an equity position of
more than 45 percent as of February 2021. More importantly, unlike during the GFC and
Great Recession, negative equity was not an issue. Even borrowers at the 5th percentile
of the equity distribution in our sample had accumulated significant wealth in their homes.
This wealth accumulation meant that most borrowers were not at risk of foreclosure when
exiting forbearance, as they had the option to sell their properties if they were still unable to
resume making mortgage payments. In contrast, during the GFC, negative equity was a huge
problem, and temporary payment forgiveness was not as effective in preventing large numbers
of defaults and foreclosures. As documented in Adelino, Gerardi, and Willen (2013), most
loan modifications granted by servicers in the lead-up to the GFC mirrored forbearance
in that they did not change any of the loan terms but simply involved the capitalization
of arrears into the balance of the loan. Before the GFC, these modifications were often
successful in giving borrowers time to cure their delinquencies, but in the aftermath of the
GFC, nonconcessionary modifications proved to be ineffective as household distress resulting
from employment and income loss became more prevalent and persistent.
Despite these caveats, we believe that forbearance could be a useful tool in mitigating
mortgage market distress in a future crisis. Many of the factors that made forbearance such
an effective policy in the pandemic period are likely to be present in the next crisis. For
example, the share of mortgages insured by the government increased after 2020, reaching 67
percent in the second quarter of 2021. The severe national house price decline that resulted
in widespread negative equity was a phenomenon unique to the GFC. In most postwar
recessions, house prices did not significantly decline at the national level, and thus a future
recession accompanied by deep, broad-based negative equity is unlikely.
Turning to the Federal Reserve’s monetary policy and large-scale MBS purchases, the
25
resulting reduction in mortgage rates and boom in refinances did serve to reduce household
financial distress. However, as a method for offsetting the shock of the pandemic, its effec-
tiveness was limited. Low mortgage rates were slow to diffuse through the economy, and
intermediaries captured a significant portion of the benefits, at least initially (Fuster et al.
(2021)). Figure 4 shows that the benefits of lower rates went into effect gradually over the
course of six quarters. This lag arose for several reasons. The first reason, as discussed
above, is that lenders have limited capacity for processing refinances, a problem aggravated
by the pandemic. Lenders rationed by raising prices, as panel B in figure 4 shows. An-
other is that refinances take 45 days or more even in normal times, and higher volumes,
combined with pandemic-related constraints on production, elongated timelines even more.
Another reason for the slow uptake of low rates is borrower inattention, as documented by
Andersen, Campbell, Nielsen, and Ramadorai (2020).
As mentioned above, enrollment in a forbearance plan disqualified a borrower from refi-
nancing into a new loan, and most lenders required a borrower who had exited forbearance
to make three consecutive payments before they would approve a refinance. The refinancing
process is also quite costly, with high fees and taxes limiting uptake. In addition, as we
showed in section 4, Black and Hispanic borrowers were significantly less likely to benefit
from low interest rates.
There are a few possible ways to ensure that lower mortgage rates reach more borrowers
and do so more quickly. One possibility is to increase the prevalence of streamlined refinance
programs. Gerardi, Loewenstein, and Willen (2020) argue that during the early stages of
the pandemic, a streamlined refinance program not requiring documentation of employment
or income would have provided necessary payment relief to many borrowers who had experi-
enced financial hardship. Another possibility would be to expand the use of adjustable-rate
mortgages or other types of mortgage products that automatically pass interest rate declines
through to borrowers. Borrowers with adjustable-rate mortgages, which are more prevalent
outside the United States, would have seen more or less immediate payment relief in April
2020 rather than having to initiate a costly and time-consuming refinance. One promising
product, in our view, is the ratchet mortgage, which combines the benefits of both fixed-rate
loans and adjustable-rate mortgages. The ratchet mortgage allows downward adjustments
in the mortgage rate but does not allow increases. This type of product provides lower costs
to borrowers over the life of the loan and, in exchange for a potentially higher initial rate,
eliminates the subsidization of those who refinance more frequently by those who refinance
less frequently.
Finally, the income support programs during the pandemic clearly played a large role in
alleviating financial distress, especially the expansion of the UI benefits program. As detailed
26
in chapter 2 of Edelberg, Sheiner, and Wessel (2022), UI expansion fully restored income for
many unemployed individuals and in some cases more than restored it. Dettling and Lambie-Hanson
(2021)) construct a measure of income support (such as UI, stimulus checks, and PPP loans)
relative to pre-pandemic incomes in each state and county. They document significant varia-
tion in the extent to which these federal programs provided under the CARES Act replaced
lost income, and that geographic areas with more generous income support experienced
better mortgage outcomes. Controlling for unemployment, the share of mortgages that
are government-backed, COVID-19 cases, and social distancing policies, they find that a
one-standard-deviation increase in the index of CARES Act income support generosity is
associated with rates of mortgage nonpayment (delinquency and/or forbearance rates) that
were about 2 percentage points lower, or roughly a 25 percent reduction.
While income support programs are broad based and can help to alleviate distress in
both the rental and mortgage markets, they do have a few drawbacks. One issue is cost.
They are much more expensive to taxpayers compared with forbearance or interest rate
reductions. In addition, because income support is typically provided as a gift and not a
loan, it suffers from much more severe moral hazard problems. People who have lost a job
might have less incentive to seek a new job if they are receiving generous unemployment
benefits, which are never repaid. However, a borrower has less incentive to voluntarily skip
mortgage payments through forbearance, since that debt must ultimately be repaid. As a
result, it seems that few borrowers misrepresented themselves as negatively affected by the
COVID-19 pandemic so that they could gain forbearance. In contrast, fraud was a major
concern for both the PPP and expanded UI benefits programs. Finally, although income
support programs provide help to households much faster than rate cuts, they are not as
timely as forbearance. In some states (for example, Florida) it took several weeks for UI
benefits to reach newly unemployed individuals at the start of the pandemic.16
Our intention in writing this article is to provoke discussion and debate on the merits and
drawbacks of recent policies enacted to mitigate distress during the COVID-19 pandemic in
the US mortgage market. It is our hope that future research using better data will add even
greater perspective on what worked, and what didn’t during this tumultuous period.
References
Adelino, Manuel, Kristopher Gerardi, and Paul S Willen. 2013. “Why don’t lenders renegoti-
ate more home mortgages? Redefaults, self-cures and securitization.” Journal of monetary
Economics 60(7): 835–853.
16
See Mazzei and Tavernise (2020) for a discussion of this issue.
27
Agarwal, Sumit, Souphala Chomsisengphet, Hua Kiefer, Leonard C Kiefer, and Paolina C
Medina. 2020. “Inequality During the COVID-19 Pandemic: The Case of Savings from
Mortgage Refinancing.” Available at SSRN 3750133.
An, Xudong, Larry Cordell, Liang Geng, and Keyoung Lee. 2021. “Inequality in
the Time of COVID-19: Evidence from Mortgage Delinquency and Forbearance.”
Federal Reserve Bank of Philadelphia Working Paper (WP 21-09). Available at
https://www.philadelphiafed.org/-/media/frbp/assets/working-papers/2021/wp21-09.pdf.
Andersen, Steffen, John Y Campbell, Kasper Meisner Nielsen, and Tarun Ramadorai. 2020.
“Sources of inaction in household finance: Evidence from the Danish mortgage market.”
American Economic Review 110(10): 3184–3230.
Anenberg, Elliot, and Therese C Scharlemann. 2021. “The effect of mortgage forbearance
on house prices during COVID-19.” FEDS Notes (2021-03): 19–2.
BLS. 2022. “Labor force statistics from the Current Population Survey.” Technical Report.
Bureau of Labor Statistics, U.S. Department of Labor.
CDC. 2019. “Risk for COVID-19 Infection, Hospitalization, and Death by Race/Ethnicity.”
Technical Report. Centers for Disease Control and Prevention.
Cherry, Susan, Erica Xuewei Jiang, Gregor Matvos, Tomasz Piskorski, and Amit Seru. 2021.
“Government and Private Household Debt Relief During Covid-19.” NBER Working Paper
(w28357).
Dettling, Lisa, and Lauren Lambie-Hanson. 2021. “Why Is the Default Rate So
Low? How Economic Conditions and Public Policies Have Shaped Mortgage and
Auto Delinquencies during the COVID-19 Pandemic.” FEDS Notes. Available at
https://www.federalreserve.gov/econres/notes/feds-notes/why-is-the-default-rate-so-low
Edelberg, Wendy, Louise Sheiner, and David Wessel. 2022. Recession Remedies: Lessons
Learned from the U.S. Economic Policy Response to COVID-19. Hutchins Center on
Fiscal and Monetary Policy at Brookings, The Hamilton Project.
Fuster, Andreas, Aurel Hizmo, Lauren Lambie-Hanson, James Vickery, and Paul Willen.
2021. “How Resilient Is Mortgage Credit Supply? Evidence from the COVID-19 Pan-
demic.” Federal Reserve Bank of Philadelphia Working Paper (WP 21-20). Available at
https://www.philadelphiafed.org/-/media/frbp/assets/working-papers/2021/wp21-20.pdf.
28
Gerardi, Kristopher, Lauren Lambie-Hanson, and Paul Willen. 2021. “Racial Differ-
ences in Mortgage Refinancing, Distress, and Housing Wealth Accumulation During
COVID-19.” Federal Reserve Bank of Philadelphia Discussion Paper. Available at
https://www.philadelphiafed.org/-/media/frbp/assets/consumer-finance/discussion-papers
Gerardi, Kristopher, Lara Loewenstein, and Paul Willen. 2020. “Eval-
uating the Benefits of a Streamlined Refinance Program.” Fed-
eral Reserve Bank of Cleveland Working Paper. Available at
https://www.clevelandfed.org/en/newsroom-and-events/publications/working-papers/2020-w
Gerardi, Kristopher, Paul Willen, and David Hao Zhang. 2020. “Mortgage Prepayment,
Race, and Monetary Policy.” Federal Reserve Bank of Atlanta Working Paper. Available at
https://www.frbatlanta.org/research/publications/wp/2020/12/18/22-mortgage-prepayment-
Institute, Urban. 2021. “Housing Finance at a Glance: A Monthly Chartbook.” Technical
Report.
Lambie-Hanson, Lauren, James Vickery, and Tom Akana. 2021. “Recent Data on Mortgage
Forbearance: Borrower Uptake and Understanding of Lender Accommodations.” Con-
sumer Finance Institute Research Brief.
Mayer, Christopher, Edward Morrison, Tomasz Piskorski, and Arpit Gupta. 2014. “Mortgage
modification and strategic behavior: Evidence from a legal settlement with countrywide.”
American Economic Review 104(9): 2830–57.
Van Dorn, Aaron, Rebecca E. Cooney, and Miriam L. Sabin. 2020. “COVID-19 Exacerbating
Inequalities in the US.” Lancet (London, England) 395(10232): 1243.
29
File and source
- File
- bostonfed-cpp20220707.pdf
- Size
- 1,283,251 bytes
- SHA-256
- f02d9ee55345c7625c166c9ac611b0882c827fc106322e483df75ea0c6b3c1c8
- Our copy
- bostonfed-cpp20220707.pdf
- Original
- www.bostonfed.org