Cherry et al., Government and Private Household Debt Relief during COVID-19 (NBER w28357; BPEA Fall 2021, 2021-01)
Summary
NBER Working Paper 28357, dated January 2021, by Susan F. Cherry, Erica Xuewei Jiang, Gregor Matvos, Tomasz Piskorski and Amit Seru on household debt forbearance during COVID-19. The abstract says cumulative payments missed per individual in forbearance between March and October of 2020 were largest for mortgage ($3,200) and auto ($430) borrowers. It estimates that more than 60 million borrowers will miss $70 billion on their debt payments by the end of the first quarter of 2021 and that 60% of aggregate forbearance goes to above median income borrowers. One third of borrowers in forbearance continued making full payments, and more than a quarter of total debt relief came from the private sector outside the government mandates.
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
“Cherry et al., Government and Private Household Debt Relief during COVID-19 (NBER w28357; BPEA Fall 2021, 2021-01).”
Original at www.nber.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/cherry-et-al-government-and-private-household-debt-relief-during-covid-9d3d5f49b5/
, SHA-256 5b7f2dc0f472510383142d64e860a86a2d93721d140c099262e5364cc24bfa9f.Full text
NBER WORKING PAPER SERIES
GOVERNMENT AND PRIVATE HOUSEHOLD DEBT RELIEF DURING COVID-19
Susan F. Cherry
Erica Xuewei Jiang
Gregor Matvos
Tomasz Piskorski
Amit Seru
Working Paper 28357
http://www.nber.org/papers/w28357
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
January 2021, Revised January 2021
We thank seminar participants at the Housing Finance Policy Center, Columbia Leading through
Crisis seminar, AREUEA-ASSA meeting, and Kairong Xiao for helpful comments. Piskorski and
Seru thank the National Science Foundation Award (1628895) on “The Transmission from
Households to the Real Economy: Evidence from Mortgage and Consumer Credit Markets” for
financial support. The views expressed herein are those of the authors and do not necessarily
reflect the views of the National Bureau of Economic Research.
NBER working papers are circulated for discussion and comment purposes. They have not been
peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies
official NBER publications.
© 2021 by Susan F. Cherry, Erica Xuewei Jiang, Gregor Matvos, Tomasz Piskorski, and Amit
Seru. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted
without explicit permission provided that full credit, including © notice, is given to the source.
Government and Private Household Debt Relief during COVID-19
Susan F. Cherry, Erica Xuewei Jiang, Gregor Matvos, Tomasz Piskorski, and Amit Seru
NBER Working Paper No. 28357
January 2021, Revised January 2021
JEL No. G00,G01,G18,G2,G21,G23,G38,G5
ABSTRACT
We follow a representative panel of US borrowers to study the suspension of household debt
payments (debt forbearance) during the COVID-19 pandemic. Between March and October of
2020, loans worth$2 trillion entered forbearance. On average, cumulative payments missed per
individual in forbearance during this period were largest for mortgage ($3,200) and auto ($430)
borrowers. We estimate that more than 60 million borrowers will miss $70 billion on their debt
payments by the end of 2021:Q1. This large amount of debt relief significantly dampened the
household debt distress, which can help explain household delinquencies below pre-pandemic
levels—a significant difference from other economic crises when delinquencies sharply increased
along with unemployment. Forbearance thus may have had potentially large aggregate
consequences for house prices and economic activity. Relief flows more to higher income
individuals than those receiving stimulus checks, partially due to their higher debt balances: 60%
of aggregate forbearance is provided to above median income borrowers. On the other hand,
forbearance rates are higher among the more vulnerable populations: individuals with lower
credit scores and lower incomes. Borrowers in regions with a higher likelihood of COVID-19
related economic shocks and higher shares of minorities were more likely to obtain debt relief.
One third of borrowers in forbearance continued making full payments, suggesting that
forbearance acts as a credit line, allowing borrowers to “draw” on payment deferral if needed.
More than a quarter of total debt relief was provided by the private sector outside of the
government mandates. Exploiting a discontinuity in mortgage eligibility under the CARES Act
we estimate that implicit government debt relief subsidies increase the rate of forbearance by
about 25%. Government and private relief follow similar patterns across income and
creditworthiness, suggesting that borrower self-selection in requesting forbearance is an
important determinant of debt relief incidence, and drives the distribution of relief across different
population strata. Government relief is provided through private intermediaries, which differ in
their propensity to supply relief, with shadow banks less likely to provide forbearance than
traditional banks.
Susan F. Cherry Tomasz Piskorski
Stanford University Columbia Business School
Graduate School of Business (GSB) 3022 Broadway
scherry@stanford.edu Uris Hall 810
New York, NY 10027
Erica Xuewei Jiang and NBER
Marshall School of Business tp2252@columbia.edu
University of Southern California
Los Angeles, CA 90089 Amit Seru
erica.jiang@marshall.usc.edu Stanford Graduate School of Business
Stanford University
Gregor Matvos 655 Knight Way
Kellogg School of Management and NBER
Northwestern University aseru@stanford.edu
2211 Campus Drive
Global Hub 4361
Evanston, IL 60208
and NBER
gregor.matvos@kellogg.northwestern.edu
I: Introduction
Large economic crises such as the Great Depression and Great Recession are often accompanied
by significant household debt distress, which spills over to the rest of the economy (Mian and Sufi
2009; Keys et al 2013). Governments have therefore regularly intervened in household credit
markets during such times. This was also the case during the COVID-19 pandemic, where the
Coronavirus Aid, Relief, and Economic Security (CARES) Act included several provisions
mandating debt forbearance - temporary suspension of debt repayments – on a large share of
mortgages and almost all student debt. Debt forbearance differs from other government programs,
because it is also provided voluntarily by the private sector outside of government mandates and
is used to restructure debt payments of households outside of crises. We document the
extraordinary size and speed of debt forbearance in response to the COVID-19 pandemic and the
resulting low levels of household debt distress, which likely significantly dampened the potential
spillovers to the rest of the economy. We measure the incidence of the relief across population
strata, as well as the role that borrowers’ self-selection and government subsidies paid in
determining the amount and incidence of the relief.
We study government and private forbearance during the COVID-19 pandemic by
using a representative credit bureau panel of more than 20 million US consumers. The data allows
us to study which loans allow payment deferral (loans in forbearance) as well as the extent to
which households chose to miss payments. Some households, for example, decided to make full
payments despite being in forbearance. We use the data to measure the extent of forbearance, as
well as its distribution across households with varying credit quality, income, exposure to COVID-
19, and economic shocks. Finally, the data allow us to classify which loans were eligible for debt
relief under the CARES Act, and which relief was provided by the private sector.
We start by documenting the level of forbearance across different categories of debt. Between
March and October 2020, loans worth $2 trillion were in forbearance, affecting 60 million
individuals. Forbearance rates substantially increased across all categories of household debt, with
share of loans in forbearance ranging from 4.6% for revolving debt, to 92% for student debt. The
lion share of new forbearance initiated during the COVID-19 crisis was in the categories of
mortgages and student debt, accounting for, respectively $1.1 trillion and $580 billion, the first
because of the large balances, and the latter due to large take-up rates. While in forbearance,
individuals have the option to pause payments on their loans. We find that around a third of
borrowers in forbearance continued making full loan repayments. Therefore, forbearance partially
acts as a credit line, allowing borrowers to “draw” on forbearance if needed.
Forbearance actions resulted in substantial financial relief for households. On average, cumulative
payments missed by individuals in forbearance during the March-October 2020 period were largest
2
for mortgage debt ($3,200) and auto ($430) debt. By October 2020, debt forbearance allowed US
consumers to miss about $43 billion debt payments. At this rate, more than 60 million consumers
would miss about $70 billion of their debt payments by the end of the first quarter of 2021.
The extent of forbearance has significant aggregate consequences, by substantially dampening
household debt distress. In the Great Recession, for example, mortgage delinquencies rose from
their low of 2% to more than 8%, spilling into the rest of the economy through a decline in house
prices, as well as aggregate demand (Mian and Sufi 2011 and 2014a; Berger et al. 2017; Benmelech
et al. 2017; Kaplan et al. 2017; Piskorski and Seru 2020). During the COVID-19 pandemic, instead,
delinquency rates declined from 3% to 1.8%. This is especially striking given an unprecedented
increase in unemployment rate that reached almost 15% in 2020:Q2 and the strong historical
association between the unemployment rate and mortgage default (see Piskorski and Seru 2018).
A back of the envelope calculation suggests that most potential delinquencies in the mortgage
market were averted because of forbearance. We speculate that the low delinquencies explain at
least in part why the pandemic has not resulted in house price declines.
The incidence of household debt forbearance differs substantially from other relief programs
targeted at households. About 60% of the dollar amount of financial relief from forbearance was
received by borrowers with above median pre-pandemic incomes. Thus, forbearance provides
relief to higher income individuals than other CARES act policies such as stimulus checks. This
fact does not imply that forbearance relief flow is, on average, unrelated to pandemic induced
shocks. Lower income households are more likely to obtain forbearance relief. Because they have
lower debt balances, the dollar value of debt relief is also smaller. Therefore, conditional on
obtaining forbearance, higher income households obtain larger dollar values of relief. The rates of
forbearance also decline with creditworthiness. This observation highlights an important feature
of forbearance. Incidence is related to borrowers’ credit constraints, while other programs, such as
stimulus check programs, often target individuals based solely on their prior income. Debt
forbearance may have complemented other stabilization programs by providing significant relief
to financially vulnerable individuals with higher pre-pandemic incomes – i.e., individuals
ineligible for policies like the stimulus check program.
Forbearance has the potential to provide targeted relief to borrowers who are subject to shocks,
because it is up to the borrower to request it and is subject to lender approval in the case of private
forbearance. Either borrow self-selection or lender scrutiny could result in a large propensity for
relief among households who require it. We document that households with a higher likelihood of
COVID-19 related shocks were more likely to obtain debt relief. Forbearance rates are
significantly higher in regions that experienced the highest COVID-19 infection rates and the
greatest deterioration in their local economies, as reflected by unemployment insurance claims and
the concentration of industries most exposed to the pandemic. As has been documented, the
3
economic and health consequences of the pandemic have disproportionately impacted minorities,
especially Black Americans. Consistent with this observation, regions with higher shares of
minorities and Blacks received debt forbearance at higher rates. Thus, forbearance may have
reached its intended target, and might have especially helped households who were affected but
might otherwise not be eligible for income-based programs.
We conclude our analysis by studying the significant role of the private sector in extending debt
relief. We use the analysis to evaluate the role of implicit forbearance subsidies forbearance and
that borrower self-selection in determining the level and incidence of forbearance. The CARES
Act mandated forbearance of federally insured mortgages and student loans. We also find
substantial increases in forbearance in auto and credit card loans, as well as mortgage loans not
eligible under the cares act. Overall, approximately more than a quarter of total relief was provided
by the private sector for debt not eligible under CARES rules.
We exploit the mandated approval of CARES act eligible mortgages to understand the role of
borrower self-selection in determining the incidence of debt relief. There are generally two steps
in determining, which borrowers obtain debt relief. First, the borrower must request relief from
the lender—self-selection. Second, the lender must agree to provide relief, which was mandated
for government loans. More than 90% of borrowers decided not to take-up rate the option of debt
relief among eligible mortgages, suggesting that borrowers’ self-selection is a powerful force in
determining forbearance rates.
We also want to understand the degree to which self-selection is responsible for the distribution of
debt relief across households. Recall that forbearance rates are higher among lower income and
less creditworthy borrowers, but that the dollar amounts are larger among the higher income
individuals. To obtain mortgage forbearance under CARES ACT, eligible mortgage borrowers had
to apply for forbearance, but forbearance for those loans was mandated. In other words, borrower’s
self-selection is driving forbearance rates for these loans. Forbearance of non-eligible (private)
loans, on the other hand, must be approved by the lender. Therefore, private debt relief is the
outcome of a mutually beneficial renegotiation.
We document that forbearance rates decline in income and creditworthiness for private and
government loans across loan categories. These results suggest that borrowers’ self-selection is
important in determining how relief is allocated. In other words, unlike more blanket CARES Act
subsidies, such as stimulus checks, mortgage debt relief flows to more vulnerable households
precisely, because these types of households apply for it. This stands in stark contrast to federally
insured student loans that were automatically placed in the forbearance by the CARES Act,
resulting in more blanket relief, which was not necessarily correlated with borrower need.
Private debt relief is presumably a result of an (ex-ante) mutually beneficial renegotiation.
Government mandated relief does not need to be mutually beneficial: it can result in a transfer
4
from the lender (government) to the borrower. Such subsidies to household debt relief may very
well be warranted in the presence of renegotiation frictions and aggregate spillover that can result
from distressed household debt. We document that more than a quarter of relief is private,
suggesting that not all government relief is subsidized. On the other hand, CARES Act eligible
debt differs from private debt both in its type, and the types of borrowers, making it difficult to
evaluate the magnitude of subsidies.
To evaluate the importance of implicit government subsidies we exploit a size discontinuity in
eligibility of mortgages for relief under the CARES Act. While government-insured loans below
the conforming loan limit qualified for government mandated forbearance, loans above the limit
were not eligible.1 Restricting our analysis to mortgages with balances near the conforming loan
balance limits, we find that the percentage of loans in forbearance increases by about 25% for
loans covered by the mandate. This is opposite from the pattern outside of the pandemic, in which
loans issued without government guarantees have slightly higher forbearance rates at the
discontinuity. In other words, our back of the envelope estimates suggests that about 20% of
government forbearance is subsidized, and the rest is provided to borrowers who would have been
eligible under a private benchmark.
Debt relief is provided through loan servicers, which may not have ownership of the loan, nor did
they necessarily originate it. Moreover, government relief is explicitly provided by a variety of
private servicers, more than half of whom are shadow banks. Since relief of government loans is
mandated, one might expect that there are few differences between suppliers. Instead, even
accounting for borrower characteristics, we find lower rates of forbearance for loans serviced by
shadow banks relative to traditional banks. This result suggests that despite the blanket government
mandate, who implements forbearance has a meaningful effect on the amount of debt relief
provided.
An important policy question is how forbearance will be unwound after it expires. As of October
2020, more than half of 60 million Americans who entered forbearance during pandemic were
missing close to $40 billion on their debt repayments.2 At this rate, by 2021:Q1 when forbearance
mandates are currently set to expire, we estimate that these borrowers will be left with a
“forbearance overhang” of more than $60 billion in accumulated postponed repayments. This
estimated forbearance debt overhang amounts to about $1,800 per individual, which is more than
half of their average monthly income, and more than 80% for lower income borrowers. If deferred
payments are structured as a one-time (bullet) payment, which is due immediately after
1
Jumbo loans exceed the conforming loan balance limits set by the Federal Housing Finance Agency and cannot be
purchased, guaranteed, or securitized by the government sponsored enterprises (GSEs).
2
During March-October 2020, borrowers who entered forbearance during this period missed about $43.5 of their debt
payments. Accounting for debt repayments they have already done by October 2020, the net amount is $38 billion.
5
forbearance ends, then a significant share of low-income borrowers will likely enter distress, even
if this payment is anticipated. Alternative, deferred payments could be amortized over the life of
the loan or added as a one-time (bullet) payment at the end of the loan. Structuring forbearance by
delaying payments would alleviate households’ liquidity constraints, at a cost to the lender.
Depending on the degree of consumer foresight and ability to smooth their consumption, these two
implementations could have significantly different consequences for consumers (see Eberly and
Krishnamurthy 2014 and Piskorski and Seru 2018). Either way, the extent of forbearance overhang
suggests that the unwinding of forbearance could have first order consequences for household debt
distress, and through it, for the aggregate economy (Mian and Sufi 2014a).
The Great Recession was marked by a wave of household debt delinquencies and foreclosures,
which spilled over in the rest of the economy, and a significant passage of time before
implementation of major debt relief programs (Piskorski and Seru 2018). We document that a
quick and widespread implementation of debt relief during the COIVD-19 pandemic both by
policy makers as well as the private sector was accompanied by historically low debt
delinquencies, and the debt relief has been positively related with exposure to shocks and financial
vulnerability.
One possible reason for the quick implementation of debt relief actions is that the private sector
and policymakers may have internalized the lessons from the Great Recession pointing to
significant costs of widespread defaults and foreclosures and were more willing to provide
widespread and quick debt relief (Eberly and Krishnamurthy 2014; Campbell et al. 2020; Piskorski
and Seru, 2018). The large private response suggests that a substantial amount of debt forbearance
was mutually beneficial. Another alternative reason for such behavior could be that the COVID-
19 shock was perceived as more transitory relative to prior crises, which could have promoted a
more widespread deployment of temporary debt relief measures by the private sector. This is
consistent with the consumer debt design literature, which indicates that lenders should provide a
certain amount of debt relief during economic downturns to limit deadweight costs of default and
allow better risk-sharing between borrower and lenders, especially if the underlying shocks are
transitory (e.g., see Piskorski and Tchistyi 2010, 2011, 2017; Eberly and Krishnamurthy 2014;
Greenwald et al. 2020; Guren et al. 2020, Landvoigt et al. 2020; Campbell et al. 2020). Relatedly,
the COVID-19 shock is a textbook example of a rare aggregate “exogenous” shock that is largely
outside of the agents’ influence. This should alleviate concerns about the moral hazard effects of
debt relief on incentives to repay debt leading to a more widespread loan renegotiation efforts
during such times (Piskorski and Tchistyi 2010, 2011, 2017; Mayer el al. 2014).
Our paper is related to the literature on the role of household balance sheet channel in the
transmission of economic shocks (e.g., Mian and Sufi 2009, 2011, 2014a; Guerrieri and Uhlig
2016; Hurst et al. 2016; Agarwal et al. 2017, 2018, 2020; Berger et al. 2017 and 2019; Benmelech
6
et al. 2017; Kaplan et al. 2017; Favilukis et al. 2017; Di Maggio 2017, 2020; Greenwald 2018,
Guren et al. 2018; Auclert 2019; Beraja et al. 2019, Eichenbaum et al. 2019; Andersen et al. 2020).
Within this literature our paper contributes to the recent studies that analyze the effects of various
stabilization programs operating through the household balance sheet channel (e.g., Mian and Sufi
2012; Parker et al. 2013; Hsu et al. 2018; Berger et al. 2020) and especially the empirical studies
focusing on various forms of debt relief (e.g., Piskorski et al. 2010; Agarwal et al. 2010; 2017,
2020; Mayer et al. 2014; Scharfstein and Sunderam 2016; Ganong and Noel 2017; Di Maggio et
al 2017, 2020; Maturana 2017; Fuster and Willen 2017; Kruger 2018; Piskorski and Seru 2018;
2020; Auclert et al. 2019; Mueller and Yannelis 2020). Our work is also related to the recent
emerging literature on the effects of COVID-19 and the policy response to the pandemic (e.g.,
Baker et al. 2020; Chetty et al. 2020; Coibon et al. 2020; Cox et al. 2020; Elenev et al. 2020; Granja
et al. 2020; Guerrieri et al. 2020).
II: Institutional Setting: US Consumer Debt Market, Debt Forbearance, and the CARES Act
II.A US Consumer Debt Market
We study the forbearance actions in the $14 trillion US consumer credit market. We focus on the
four main categories of consumer debt: residential mortgages, auto, revolving, and student debt.
Figure A1 in the Appendix shows the evolution of the aggregate outstanding loan balance for these
four types of debt from 2006 until 2020. We describe each of these debt markets below.
Residential Mortgage Market
The residential mortgage market is by far the largest form of consumer debt in the United States.
As of 2020, total mortgage balances in the U.S. totaled roughly $10 trillion. There is significant
government involvement in the mortgage market, with roughly two-thirds of outstanding
mortgages guaranteed by the U.S. government (Buchak et al. 2018). The loans effectively
guaranteed by the US taxpayer consist of the conforming loans sold to government sponsored
enterprises (GSEs) including the Fannie Mae and the Freddie Mac and the Federal Housing
Administration (FHA) loans.
Conforming loans are eligible for securitization with the participation of GSEs that purchase and
package these loans into mortgage-backed securities (MBS) and insure MBS default
risk. Conforming loans are typically extended to borrowers with relatively high credit scores, fully
documented income and assets, and moderate loan-to-value (LTV) ratios. Additionally,
conforming mortgages loan balances must lie below the conforming loan limit. These origination
loan balance limits were $417,000 for a single-family home and $625,000 (depending on the area’s
cost of living) throughout most of the last decade, reaching $510,400 and $765,600, respectively
by 2020 after their progressive increases during 2017-2020 period. The Federal Housing
7
Administration (FHA) provides mortgage insurance on loans made by FHA-approved lenders
nationwide. FHA borrowers can make as little as 3.5% down payment and FHA loans are usually
considered the riskiest segment of the mortgage market as they mainly appeal to lower income and
less creditworthy households.
Jumbo loans, mortgages with balances exceeding the conforming loan limit, account for about
15% of the outstanding loan balances in our data. These loans are ineligible for government
guarantees and as a result they are much more difficult to securitize and are typically retained by
lenders on their balance sheets (Buchak et al. 2020). As we explain below, jumbo loans were
ineligible for the CARES Act forbearance mandates. We exploit this within debt category variation
in the applicability of the CARES Act to shed light on the role of these mandates for debt
forbearance actions.
Student Debt
Total student debt has increased significantly in recent years, as seen in Figure A1 in the Appendix.
In 2006, student loan debt was the smallest of the four consumer debt categories, but by 2020 total
student loan balances were exceeded only by mortgages. As with the mortgage market, there is
significant government involvement in the student loan market. The federal government is the
primary provider of student loans in the United States, with about 90% of outstanding student
loans held by the Department of Education (Looney and Yannelis 2019). 3 As we explain below,
all such loans were covered by the CARES Act forbearance mandates.
Auto Debt
Auto debt has increased faster than all other types of debt except for student loans over the past
ten years, totaling over $1.3 trillion in 2020. The majority of both new and used cars are financed
with debt – in 2018, over 85% of new cars and 55% of used cars were financed with a loan or a
lease (Di Maggio et al. 2017). Auto debt includes both auto loans and leases, with leases
accounting for roughly 30% of new passenger cars. 4
Unlike the mortgage and student loan markets, there is no significant government intervention in
the auto debt market. Auto loans and leases are typically made through either a bank or a car
dealership (Benmelechech et al. 2017). Dealership financing has been growing in recent years and
3
Two programs are responsible for the majority of federal student loans: The Federal Direct Loan program (DL) and
the Federal Family Education Loan program (FFEL). The FFEL program was implemented in 1965 but stopped
issuing loans after the Health Care and Education Reconciliation Act was passed in 2010. Funds for these loans were
provided by private lenders, though the federal government subsidized and guaranteed the loans. After 2010, all
federal loans were issued directly by the Department of Education under the DL program. The FFEL and DL programs
are similar, with identical interest rates, disbursement rules, and maximum loan amounts. The main difference between
the two programs is the source of funding, with FFEL loans funded by private lenders and DL loans funded directly
by the U.S. Treasury.
4
https://www.experian.com/blogs/ask-experian/research/auto-loan-debt-study/
8
accounted for 31% of new loans and leases in 2020, while banks, credit unions, and finance
companies accounted for 65.5% percent of car purchases and leases. 5
Revolving Debt
Revolving debt in our analysis refers to all accounts that allow individuals to borrow against a
credit line, with the exception of home equity lines of credit. This includes credit cards, as well as
retail and other revolving accounts like personal lines of credit (see Agarwal et al. 2018 for more
details). Borrowers can choose to pay off the full balance at the end of each billing cycle or
6
“revolve” the balance by carrying it over from one cycle to the next. Borrowers who choose to
revolve their balance are required to make minimum payments each month. Total outstanding
revolving debt was slightly over $1 trillion as of 2020 (see Figure A1).
II.B Consumer Debt Forbearance and the CARES Act
The primary focus of our paper is debt forbearance. Forbearance is the action of a lender waiving
a borrower’s contractual duties on the borrower’s payments for a fixed amount of time. This type
of agreement is present in all types of consumer debt including mortgage, auto, credit card, and
student debt. Agreements are structured somewhat differently in each sector, but all include a
short-term suspension or reduction of loan payments. Forbearance is a typical way for distressed
borrowers who are struggling with loan payments to avoid or delay foreclosure.
The structure of a forbearance agreement includes a halt or reduction in a borrower’s loan
payments for a fixed period. To enter a forbearance agreement the borrower must usually approach
the lender with satisfactory proof of distress and proof that the distress is temporary. The lender
and borrower may then design a forbearance agreement for a temporary period based on the loan
and the borrower’s payment history. Once the agreed-upon period begins, a borrower may stop or
decrease their loan payments without fear of foreclosure. Although the payments have been
delayed, the loan’s interest does not stop accruing over this period.
It is important to note that forbearance is not a debt forgiveness program or a loan modification.
Unlike in the case of loan modification (see Piskorski et al. 2010; Keys et al. 2013; Agarwal et al.
2017), the borrower is required to pay the lender the missed payments after the forbearance period
ends. Typical repayment plans following the end of the forbearance period consist of a lump sum
payment or increasing the regular payment amounts once forbearance is finished. Borrowers can
5
https://www.experian.com/blogs/ask-experian/is-it-better-to-finance-a-car-through-a-bank-or-dealership/
6
Credit cards are the most well-known form of revolving debt and are issued by banks or credit unions. Personal lines
of credit function similar to credit cards but are not linked to a physical card. Instead, the funds are typically obtained
through a check or direct deposits. Both credit cards and personal lines of credit are typically unsecured.
9
also attempt to obtain a loan modification agreement that allows them further deferral or permanent
reduction of at least part of their missed payments.
The Coronavirus Aid, Relief, and Economic Security (CARES) Act passed by Congress and signed
by President Donald Trump on March 27, 2020 included several provisions designed to help
individuals suffering from financial hardship as a result of the pandemic through loan forbearance.
Below, we discuss typical structures and standards for loan forbearance in each of the loan
segments and the implications of the CARES Act for them.
Mortgage Debt Forbearance and the CARES Act
Mortgage forbearance agreements have been historically granted on a case-by-case basis.
Mortgages are usually large so proof of distress, proof the distress is temporary, and proof the
borrower can repay the interest and missed payments has been typically required.
Under the CARES Act, borrowers with federally backed mortgages, including GSE, FHA, and
Veteran Administration loans, were allowed to pause their mortgage payments with no penalties
for six months with the option to extend forbearance for up to an additional six months. This means
that individuals with federally backed mortgages are entitled to up to a year of paused mortgage
payments beginning from the date they originally obtained forbearance. Under forbearance
policies, no fees, penalties, or additional interest may be added to a borrower’s account. In addition,
the CARES Act granted mortgage borrowers protections to help them avoid foreclosure, including
a 60-day foreclosure and eviction moratorium for borrowers with federally backed mortgages. This
moratorium was originally intended to expire in May 2020 but has been extended until at least
January 2021.
It is important to note that while the CARES Act guarantees individuals with federally backed
mortgages the right to pause their mortgage payments, it does not automatically place their
mortgages in forbearance. Borrowers must contact their loan servicer to put their payments on
hold, though the forbearance process is straightforward – borrowers simply need to claim they
have a pandemic related hardship and do not need to submit any documentation.
Borrowers with loans without the government guarantees such as jumbo loans were not covered
by the CARES Act forbearance mandates. However, like in the case of covered loans, the CARES
Act does place requirements on how payment information can be reported to credit bureaus. If a
borrower is current on their account and their lender agrees to allow a skipped payment, partial
payment, or other accommodation due to the COVID pandemic, then the creditor must report the
borrower as current on their loan to the credit bureaus. Similarly, if the account is already
delinquent, the creditor cannot report the borrower as more delinquent during the period of the
agreement.
Student Debt Forbearance and the CARES Act
10
Student loan forbearance agreements have historically been relatively easy to enter into because
forbearance provisions are usually built into the loan agreements. Federal student loan borrowers
may be entitled to a loan deferment (if they are unemployed) or a forbearance (if the amount owed
exceeds 20% of their gross income). These programs allow student loan borrowers to temporarily
defer making payments, and interest may or may not accrue depending on the type of loan and
specifics of the deferment or forbearance program (Mueller and Yannelis 2019).
The CARES Act offered relief to borrowers with qualifying federally held student loans, which
account for vast majority of all outstanding student loans. Federal student loans were automatically
placed into administrative forbearance. The interest rate on these loans is 0%, so any payments
that borrowers make during this period go directly towards reducing the loan principal. Student
loan forbearance was originally designed to expire in September 2020 but has been extended until
at least January 2021. Borrowers with private student loans are not covered under the CARES Act,
though private loan borrowers may negotiate with their loan servicers to be placed into
forbearance.
Auto Debt Forbearance and the Cares Act
Auto debt forbearance agreements are granted on a case-by-case basis and are subject to the
lender’s discretion. The CARES Act did not include any explicit forbearance mandates for auto
loans. Borrowers must contact their lenders for information about these forbearance or deferment
policies, with policies and eligibility varying by lender. While the CARES Act does not guarantee
forbearance policies for these loans, if a borrower is current on their account and their lender agrees
to allow a skipped payment, partial payment, or other accommodation due to the COVID-19
pandemic, then the creditor must report the borrower as current on their loan. Similarly, if the
account is already delinquent, the creditor cannot report the borrower as more delinquent during
the period of the agreement.
Revolving Debt Forbearance and the CARES Act
Credit card forbearance agreements are given on a case-by-case basis and is generally based on
the past history of a cardholder including payment history, length of time as a cardholder, and their
outstanding balance. It is given in many types of options including normal terms of suspended
payment, but may also include a lowered monthly payment, a lower interest rate, an increased
borrowing limit, or a deferment. These agreements are split into two types: “short-term” where the
term period is 12 months or fewer and “long-term” where the term period is longer than 12 months.
According to the Bureau of Consumer Financial Protection in August 2019, “The average quarterly
11
new enrollment rate among individual issuers ranged from a low of 0.2 percent to a high of 5.2
percent of all pre-charge-off delinquent balances.”7
As in the case of auto loans, the CARES Act did not include any explicit forbearance mandates for
revolving loans. However, revolving loans were covered by the same provisions that instruct the
lenders to not report missing payments under the forbearance plan as a consumer delinquency to
the credit bureaus.
III: Data Sources
In this section we describe our main data sources.
Equifax Analytic Dataset
Our main dataset is the Analytic Dataset provided by Equifax. Equifax is a credit-reporting agency
that provides monthly borrower-level data on credit risk scores, consumer age, geography, debt
balances, and delinquency status at the loan level for all consumer loan obligations and asset
classes. The Analytic Dataset is created from a 10% random sample of the U.S. credit population
from 2005 to 2020 across all U.S. geographical boundaries and consists of over 20 million
consumers. Randomization in the sample is based on social security numbers, ensuring that the
sample is representative of the U.S. credit population.
We use this data to investigate consumer forbearance status, delinquency status, payment history,
age, income, credit score, and location. Our analysis focuses on the months during and leading up
to the coronavirus pandemic. A challenge in our data is identifying consumers who are in
forbearance, as lenders are not required to report any special code when providing forbearance,
deferment, or other accommodations. We follow Equifax’s standard procedure for identifying
whether a loan is in forbearance. Specifically, we consider a loan to be in forbearance if it has a
narrative code indicating that it is in forbearance or deferment, if it is in a partial payment plan, if
the loan has been modified, or if the account has a positive balance with no reported scheduled
payment. We exclude all loans that have been refinanced or prepaid.
Table 1 provides descriptive statistics for more than twenty million borrowers in the Equifax
dataset. The first four columns of Table 1 contain information on all individuals in our sample,
while the final four columns show statistics for borrowers in forbearance. Columns 1, 2, 5, and 6
show statistics for the pre-COVID period just preceding the pandemic (January-February 2020)
and Columns 3, 4, 7, and 8 contain data for the COVID period (March-October 2020).
Fannie Mae Single-Family Historical Loan Performance Data
7
The Consumer Credit Card Market. Bureau of Consumer Financial Protection. August 2019.
https://files.consumerfinance.gov/f/documents/cfpb_consumer-credit-card-market-report_2019.pdf
12
Fannie Mae discloses loan acquisition and performance data on single-family mortgage loans that
it acquired since 2000. This loan-level monthly panel data provides detailed information on a rich
set of loan, borrower characteristics (e.g. fico scores, loan-to-value, debt-to-income, location of
the property, and interest rates), property, and monthly payment history.
Two sets of information are important to our supply of forbearance analysis. First, we observe the
name of the entity that delivered the mortgage loan to Fannie Mae and the name of the entity that
services as the primary servicer of the loan. For both sellers and servicers, we identify the entities
that represent at least one percent of volume within a given acquisition or reporting quarter. We
then identify whether a seller or servicer is a bank or a shadow bank by merging the Fannie Mae
dataset to bank regulatory filings (e.g. Form 031, 041, and FY-9C) and shadow bank call reports.
Second, Fannie Mae collects information about the type of assistance plan that the borrower is
enrolled in that “provides temporary mortgage payment relief or an opportunity for the borrower
to cure a mortgage delinquency over a defined period,” in which Forbearance Plan, Repayment
Plan, Trial Period Plan are the three major borrower assistance plan categories. We obtain all loans
that were acquired by Fannie Mae between January 1, 2000 and June 2020. The monthly
performance data runs through June 2020. We restrict our sample to active loans – loans that had
not been paid off, refinanced, or foreclosed – by January 2020.
Opportunity Insights Economic Tracker
In addition to the credit bureau data, we use regional data from the Opportunity Insights Tracker
to understand how local economic conditions and regional impacts of COVID-19 crisis relate to
forbearance actions. The Opportunity Insights Tracker provides real-time data on employment
rates, spending, mobility patterns, and Covid case rates across different counties, regions, and
income groups. Specifically, we are interested in total Covid case rates, total unemployment
insurance claims, changes in credit/debit card spending, and changes in time spent at workplaces
at the county level. Changes are relative to January 2020.
American Community Survey
We supplement the regional Opportunity Insights Tracker with socio-economic characteristics
from the U.S. Census Bureau’s American Community Survey 2018 5-year estimates. The 5-year
estimates are created from 60 months of collected data and are available at the Zip Code Tabulation
Area (ZCTA). In most instances, ZCTAs are the same as zip codes. However, we note that because
the Census Bureau creates ZCTAs by taking the most frequently occurring zip code in an area,
some addresses have ZCTAs that are different from their zip codes. We use the following variables
at the ZCTA level: percent Black, percent Hispanic or Latino, unemployment rate, and median
income, and percent of the population with a college education.
Other Data Sources
13
Our analysis also makes use of several other data sources. We use median house prices from
Zillow, unemployment claims and benefits from the Department of Labor, the number and amount
of Economic Impact Payments (stimulus checks) received by each state from the Internal Revenue
Service, and the number and size of Paycheck Protection Plan loans from the Small Business
Administration. We also gather information on the number of small businesses in a county and the
share of the workforce employed in certain industries from the Bureau of Labor Statistics.
IV: Aggregate Household Debt Forbearance and the Absence of Distress during COVID-19
We begin by analyzing the use of forbearance throughout the COVID-19 pandemic across loan
types and time. Since our dataset is a 10% random sample of the U.S. credit population, we also
estimate the total number of loans in forbearance in the United States and the aggregate dollar
amount of payments that have been missed as a result of forbearance actions undertaken during
the pandemic.
IV. Forbearance during the Pandemic
IV.B Forbearance Rates, Usage, and Amount of Relief per Borrower
We start by analyzing the forbearance rates on residential mortgages, the largest category of US
consumer debt. Residential mortgage forbearance rates increased from roughly 0.6% prior to
COVID-19 to nearly 7% in June following the declaration of the national COVID-19 emergency
and the implementation of the CARES Act in March 2020 (Figure 1, Panel b). Overall, about 9%
of mortgage borrowers were in forbearance during the period from March-October 2020 of which
about ninety percent entered forbearance during the COVID-19 period (Panel (a) of Table 2).
Forbearance rates during the COVID-19 period were also much larger than those during the Great
Recession, during which forbearance rates peaked at a little over 2%.
Figure 2 shows that auto, revolving (credit card), and student loans also saw significant increases
in forbearance rates during the pandemic and relatively low overall delinquency rates. For the most
part, these debt types featured low forbearance rates prior to the pandemic, with large spikes
occurring around April 2020. Student loans are the exception, with large numbers of loans in
forbearance or deferment prior to the pandemic (close to 50%). Nevertheless, student loans feature
a sharp increase in April 2020, when the percentage of loans in forbearance or deferment jumps
from roughly 50% to well over 90%. As we discuss in Section II.B, the large jump mainly reflects
the automatic forbearance mandate, which covered the vast majority of student loans. While
forbearance rates have declined for all debt types except for student loans since June 2020, they
remain elevated well above their historical averages. Overall, these findings show that debt
forbearance rates have dramatically increased during the COVID-19 pandemic on all types of debt.
Loan relief allows borrowers to stop payments on their loans, but they do not have to use this
option. Most borrowers in forbearance during the pandemic period missed their scheduled
14
payments: from more than 65% of borrowers in the case of mortgages to almost 100% in the case
of student debt (Panel (a), Table 2). On the other hand, a sizable minority of borrowers in
forbearance who continued to make full payments on their mortgage, auto, and revolving loans.
Therefore, forbearance partially acts as a credit line, allowing borrower to “draw” on forbearance
if needed.
Individuals in forbearance obtain substantial debt relief. Mortgage debt was the category with the
largest missed payments from individuals in forbearance, with the average borrower missing about
$3,200 from March-October 2020. This mainly reflects the fact that mortgage debt, and hence the
associated payments, are much higher relative to other debt categories for a typical consumer. The
second largest category for payments missed by individuals in forbearance is auto debt, with
borrowers missing on average about $430 during the same period. Revolving debt borrowers in
forbearance missed on average $70 over this period, while student debt borrowers missed on
average $140. While student loan is a large debt category of U.S. households, it also has a very
long maturity, lowering the payments, and thus the impact of forbearance substantially. Debt relief
is therefore substantial even considering other COVID-19 relief programs, such as stimulus checks
and unemployment benefits. The average recipient of a stimulus check received $1,696, the
8
average unemployment benefits going to an unemployed worker from mid-March to the end of
October 2020 ranged from around $16,000 to $32,000 . Therefore, payments missed through
9
mortgage debt forbearance were larger than stimulus checks on average and amounted for as much
as 20% of the total unemployment benefits received for some individuals during the pandemic. As
we discuss below, debt relief affected a substantially different population than income tested
programs, such as stimulus checks, and thus provided substantial relief to those households.
IV.C Aggregate Debt Forbearance
The substantial number of individuals who used forbearance to miss their payments means that
forbearance is associated with a significant amount of debt relief at the aggregate level. Our dataset
is a 10% sample of the entire U.S. credit population with randomization based on social security
numbers. Given this feature of our dataset, we can scale our estimates to the aggregate level
without concerns about sample representativeness (Panel (b), Table 2). From March to October
8
In April 2020, stimulus checks (Economic Impact Payments) were sent out to individuals with adjusted gross income
up to $75,000 and married couples with adjusted gross income up to $150,000. Individuals received up to $1,200,
while married couples were eligible for up to $2,400. An addition $500 was sent out for each qualifying child. A
number of unemployment programs have also been implemented to either increase the generosity of unemployment
benefits or extend benefits to individuals who would not usually be eligible. Pandemic Unemployment Assistance
extended unemployment insurance benefits to business owners, self-employed individuals, independent contractors,
and workers who have significantly reduced their services as a result of the pandemic. Additionally, a $600 payment
was added to each week of unemployment benefits for all recipients from March 29th through July 25th.
9
Our calculations assume that an individual was employed from mid-March to the end of August. We calculate these
numbers by taking the average unemployment benefits in each state and adding $600 per week for the weeks from
March 29 through July 25th.
15
2020, $2,7 trillion of dollars of debt were in forbearance, of which $1.9 trillion entered forbearance
during the pandemic, with most of the increase coming from mortgages ($1.1 trillion). This
increase reflected 139 million loans, and 61 million borrowers, suggesting that borrowers
experienced debt relief in more than one category. By October 2020, borrowers missed overall $43
billion of loan payments. At this rate by the end of the first quarter of 2021 more than 60 million
of borrowers have missed about $70 billion of payments due to debt forbearance actions
undertaken during the pandemic.
The aggregate expenditure on debt relief was somewhat smaller than other types of debt relief.
$267 billion were sent to Americans through stimulus checks, $659 billion was provided through
PPP loans, and an estimated $585 billion was spent on unemployment benefits. As we argue in the
next section, despite a lower cost, the sizeable household financial relief was targeted at
households in distress, and prevented substantial household distress, and with it, the spillover to
the rest of the economy.
VI.D Aggregate Implications: The Absence of Household Debt Distress
Large economic crises such as the Great Depression and Great Recession are often accompanied
by significant household debt distress, which spills over to the rest of the economy. During the
Great Recession, serious delinquency rates (60 days or more past due on payments) rose from less
than 2% in 2006 to more than 8% in 2010 (Piskorski and Seru 2020). A large body of work shows
how distressed household debt distress spilled over into aggregate house prices, employment, and
consumption. Governments have therefore regularly intervened in household credit markets such
times. During the Great Recession, the federal government created two large-scale debt relief
programs, which stimulated loan restructuring and refinancing activity and ameliorated some of
the impact of household debt distress (e.g., Agarwal et al. 2018, 2020, and Piskorski and Seru
2018, 2020).10 On the other hand, these programs were put in place after significant amounts of
household distress had already materialized, so they did not avoid a substantial spillover to the rest
of the economy (see Piskorski and Seru 2020). In other words, the household debt channel had
significant consequences in the Great Recession (e.g., Mian and Sufi 2009, 2011, 2014a).
In contrast to the Great Recession, the mortgage delinquency rates in the COVID-19 pandemic
declined! Mortgage delinquency rates (30 days or more past due on payments) had begun
decreasing prior to the pandemic and continued to decrease continuously throughout the pandemic
(Figure 1, panel c). Delinquencies have remained low and steady at roughly 1.5% since May 2020.
The pandemic has a devastating effect on the real economy with the unemployment rate reaching
almost 15% by April 2020 and a severe decline in GDP (Figure 1, Panel (c)). Extrapolating from
10
The Home Affordable Refinance Program (HARP) and the Home Affordable Modification Program (HAMP), were
aimed at a total of 12 million mortgage borrowers in distress.
16
the previous economic crises, and the strong historical association between the unemployment rate
and mortgage defaults, one would expect a significant rise in household debt distress (see Piskorski
and Seru, 2018).
The pandemic differs from the other crises in its rapid and intensive surge in private and
government debt relief. The relief materialized within months of the pandemic before household
distress realized. One way to evaluate the contribution of forbearance to the low delinquency rates
is to infer that households, which chose to miss payments under forbearance would have otherwise
been in financial distress. A simple back of the envelope suggests that if households in forbearance
that continued missing payments by October 2020 were seriously delinquent instead, we would
have about 2 million more mortgage borrowers in default. This simple calculation suggests that
debt relief may have decreased serious mortgage delinquency rate by about four percentage points
and avoided a significant amount of household debt distress.11 This might also help explain why
the standard household debt channel was largely absent from the COVID-19 pandemic.
V. Who are the Recipients of Forbearance?
As we established above, debt forbearance has provided significant financial relief to US
households throughout the initial stage of the pandemic. Ultimately, the impact of a given action
or policy and its cost-effectiveness crucially depends on who benefits from it, also in relation to
other programs. In this section, we exploit the richness of our data to shed light on the effectiveness
of debt forbearance actions by studying their distribution across consumers and regions based on
consumer financial vulnerability measures and the deterioration in local economic conditions
following the onset of the pandemic.
We begin by using individual-level credit bureau data to investigate forbearance rates across
borrower income and financial vulnerability measures such as borrower overall creditworthiness.
While our individual-level data includes a rich set of outcomes, it lacks information on certain key
characteristics, such as race and employment. To explore how forbearance is related to such
characteristics, we turn to regional data and explore the relation with socio-economic factors and
forbearance rates at the zip code level. We pay particular attention to the intensity of forbearance
across racial characteristics, since the distributional consequences of COVID-19 have been
documented to be highly uneven across these lines. In particular, the pandemic and its broader
economic and health consequences have been disproportionately impacting minorities, especially
Black Americans. We then look at the relationship between forbearance rates and the
industries/occupations that are most prevalent in a region, including the type of occupations that
were most affected by the COVID-19 induced shutdowns. Finally, we analyze the relation between
11
We note that this simple calculation is conditional on other CARES Act provisions being in place. In the absence
of these provisions along with debt forbearance the household delinquency rate would likely be even higher.
17
debt forbearance and the intensity of the crisis at the regional level more directly by studying the
association between forbearance and the pandemic case rates, unemployment insurance claims,
changes in time spent at workplaces since January 2020, and changes in consumption relative to
January 2020. We conclude by investigating the relation of forbearance to other stimulus measures
to shed light on the extent to which debt forbearance complements these other polices.
V.A Debt Forbearance across Creditworthiness and Income
We begin by investigating how forbearance rates relate to consumer income and financial
vulnerability measures. One potential distinct feature of forbearance actions is that they can target
individuals based on their creditworthiness and credit constraints. This may explain why
forbearance has such a large impact on household distress despite a smaller aggregate expenditure
than other programs, such as stimulus checks, or unemployment insurance.
Table 1 shows that borrowers in forbearance are more likely to be less creditworthy: they have
lower credit scores and are more likely to be delinquent on their debts. They also have higher debt
balances and are younger on average. Comparing columns (3) and (7) of Table 1, we see that
average Vantage credit scores are nearly 60 points lower for individuals in forbearance compared
to the overall population. Borrowers in forbearance have higher average balances on all debt types,
have lower average estimated annual incomes by about $6,000, and are younger by about 8 years.
Importantly, Columns (1) and (5) show that these differences in borrower characteristics were also
present prior to the pandemic (in January 2020) and hence are not driven by the impact of COVID-
19 and the response of individuals to the pandemic. In other words, consumers who ended up in
forbearance during the pandemic had on average much lower credit scores, lower incomes and
higher debt balances prior to the pandemic relative to those consumers that did not end up in debt
forbearance.
In Table A1 in the Appendix we look specifically at mortgages borrowers. Again, we compare
columns (3) and (7) and find similar patterns. Mortgage borrowers in forbearance had average
credit scores almost 70 points lower than the overall sample of mortgage borrowers, a very
significant difference. They also had higher mortgage debt, with average balances more than
$70,000 higher than the average mortgage borrower. Their average balances on other debt types
were also higher. Mortgage borrowers in forbearance were also much more likely to be delinquent
on all debt types – they were two times as likely to be delinquent on first mortgages and student
loans and over three times as likely to be delinquent on auto and revolving loans.
We next visually illustrate the relationship between mortgage forbearance rates and an individual’s
creditworthiness, income, and age in Panel (a) of Figure 3. Specifically, we split consumers into
four groups based on these characteristics. For income and age, we create four groups based on
the 25th, 50th, and 75th percentiles of these variables. For credit scores, we simply use the four credit
groups (“Low”, “Fair”, “Good”, “Excellent”) defined by Equifax based on VantageScore. The
18
impact of credit score is especially striking – an individual with a “low” credit score is more than
four times more likely to be in forbearance than an individual with an “excellent” score. In panels
(b) through (d), we show similar figures for auto, revolving, and student debt. These panels show
similar patterns for auto and revolving loans – forbearance and missed payments rates decrease
with income and credit score. In contrast, relative differences in forbearance and missed payment
rates for student loans are much smaller, reflecting the fact that all federally held student loans
were automatically placed into forbearance by the CARES Act, which resulted in less targeted
“blanket” financial relief across household characteristics.
Next, we move from the extensive to the intensive marginal and consider how the amount of
money missed by individuals in forbearance is related to an individual’s income, and credit score.
Panels (a) of Figures 4 and 5 focuses only on individuals who are in mortgage forbearance and
explores the heterogeneity in the percent of payment missed and the amount. Panel (a) of Figure 4
looks at the percent of scheduled payments that an individual missed. More than sixty percent of
individuals in mortgage forbearance missed their payments and percent of mortgage payment is
fairly similar across borrower characteristics.
Panel (a) of Figure 5 shows that there is only a slight relationship between the dollar amount of
missed payment and age and credit score, but there is a strong positive association between income
and the amount of payment missed. On average, individuals in the highest income group missed
over $1,000 more than individuals in the lowest income group.
Taken, together Panels (a) of Figures 3 to 5 highlight that there are important differences between
the patterns of heterogeneity of mortgage forbearance rates and the associated amount of debt relief
across households. Less creditworthy and low-income borrowers are more likely to be in the
mortgage forbearance program. However, conditional on being in forbearance, the dollar amount
of relief is much higher for higher income borrowers. This is mainly because higher income
borrowers have higher loan balances and hence higher associated loan payments.
We next turn our attention to other types of debt. The patterns observed in panels (b)-(d) of Figure
3 are broadly similar to those seen for mortgages in panel (a). Panel (d) of Figure 3 shows that
there are less relative differences in student forbearance rates across individuals. This is not
surprising, since federal student loan borrowers were placed in forbearance with no action required
on the part of the borrower. Panels (b) through (d) of Figure 4 show the patterns of percent and
amounts of scheduled payments missed for auto, revolving, and student loans are quite similar to
the mortgage debt patterns – conditional on being in forbearance, the percent of payments missed
is fairly uniform across all groups. Panels (b) through (d) of Figure 5 show that, as in the case of
mortgages, high income borrowers in forbearance receive significantly larger amounts of dollar
relief per individual compared to low-income borrowers (though in the case of revolving debt,
high income borrowers receive smaller amounts of relief).
19
To investigate the association between debt forbearance and individual characteristics more
formally, we next estimate the linear regression of the form
!",$,% = ' + )*",% + +,% + -",$,% (1)
where !",$,% is an indicator variable for whether individual i who lives in zip code z is in forbearance
or missed their monthly payment during month t, *",% is a vector of individual characteristics, and
,% is either a zip code fixed effect or a vector of zip code characteristics. Months include March
2020 to October 2020.
Panel (a) of Table 3 reports the results from these regressions for mortgages. Specifically, column
(1)-(2) show the estimates from the regression of an indicator for whether a mortgage borrower
was in mortgage forbearance during the pandemic on credit score, income, debt levels and debt-
to-income ratio, a small business owner indicator, the number of accounts past due, and age all
measured as of January 2020. Consistent with Figure 3, lower credit scores and income are
associated with a higher probability of mortgage forbearance. Higher mortgage debt balances are
also strongly correlated with the probability of being in forbearance. We also find that small
business owners are more likely to be in forbearance, as are individuals with higher debt-to-income
ratio and larger numbers of accounts past due as of January 2020. Next, we reduce our sample to
only individuals in forbearance on their mortgage accounts to understand which individuals are
taking advantage of forbearance to miss payments on their accounts. Columns (3) through (4) of
Table 3 show the estimation results from the specification where the dependent variable is a
dummy for whether an individual in mortgage forbearance missed their payment on individual-
level characteristics. We find that individuals with higher debt levels and lower credit scores are
not only more likely to be in forbearance but are also more likely to miss at least part of their
mortgage payment.
Panel (a) to (c) Tables A2 in the Appendix show results from the same analysis carried out for the
other three debt types. For the most part, similar patterns emerge. Individuals in forbearance on
their auto, revolving, and student loans tend to have higher balances, lower credit scores, and are
of younger ages. The patterns are somewhat different when we condition on a borrower being in
forbearance. Higher balances, lower credit scores, and lower income are positively associated with
individuals in forbearance missing payments for both auto and revolving debt. For student loan
borrowers, missing payments in forbearance is positively correlated with lower debt balances,
higher credit scores, and higher estimated incomes.
There are two main takeaways from this evidence. First, we find that overall debt forbearance rates
on consumer debt are much higher for less creditworthy and more financially constrained
borrowers. This differentiates debt forbearance actions from policy programs like the stimulus
checks that target individuals based on their income, regardless of their actual financial conditions.
20
Second, while we find that lower income borrowers have much higher forbearance rates,
conditional on being in debt forbearance, individuals with higher pre-pandemic incomes received
by far the largest dollar amount of debt relief per individual. This largely reflects much higher debt
balances and scheduled loan payments of higher income borrowers compared to lower income
ones. To illustrate the aggregate implications of this observation we quantity the aggregate dollar
amount of financial relief due debt forbearance that flowed to borrowers with above median
income, which for individuals in our data set is equal to about $37,000. Panel (c1) of Table 2 shows
that shows that 88% (15 billion) of missed mortgage payments can be accounted for by borrowers
with above median income during March-October 2020 period. Similarly, 61% (2.3 billion) of
missed auto payments come from borrowers with incomes greater than $37,000, as did 46% ($1.7
billion) of missed revolving payments. The percentage of student loan missed payments from
below median individuals is much lower, however, at just 33% (6 billion).
Next in panel (c2) of Table 2 we perform a similar assessment, when we now define a median
income within each product category. For example, mortgage borrowers in our sample have a
median income of $52,000, which is much higher than median in overall population of consumers.
Even comparing to this benchmark though, we find that higher income borrowers received over
61% ($10 billion) of the total amount of financial relief due to mortgage forbearance from March
2020 to October 2020. Similarly, we find that student loan borrowers with above median income
($22,000) account for over 64% ($12 billion) of student loan missed payments. High income
revolving and auto borrowers account for a lower percentage of missed payments, with above
median income auto borrowers ($41,000) accounting for 52% ($1.9 billion) of total auto missed
payments and above median revolving borrowers ($37,000) account for 46% ($1.7 billion) of
missed revolving payments. Across both definitions of median income, we find that above 60% of
the aggregate dollar amount of financial relief on all debt types has flowed to borrowers with
higher pre-pandemic incomes.
Overall, this evidence suggests that debt forbearance policies have mainly affected borrowers who
are less creditworthy and more likely to be liquidity constrained. In the case of the student debt,
the forbearance rates are more uniform across borrower characteristics, reflecting an automatic
forbearance mandate that applied to the vast majority of student loans. Our findings also suggests
that debt forbearance has importantly complemented other policies targeting US consumers during
the COVID-19 pandemic. Unlike polices based mainly on income, such as the stimulus check
program, debt forbearance allowed less creditworthy borrowers with higher pre-pandemic incomes
to obtain a significant amount of financial relief.
V.B Forbearance and Exposure to COVID-19 Shocks?
While our individual-level data includes a rich set of outcomes, it lacks information on certain
characteristics, such as race and employment. We also do not have individual data on which
21
households were directly infected with COVID-19. To explore how forbearance is related to such
characteristics, we turn to regional data and explore the association of forbearance rates with these
factors at the zip code level. In doing so we analyze whether forbearance rates are higher in the
areas that have been impacted relatively more by the COVID-19 pandemic.
We start by observing that there is significant regional heterogeneity in forbearance rates at the zip
code level, with dark red indicating high levels of forbearance and green indicating low levels of
forbearance. Panel (a) of Figure 6 shows that first mortgage forbearance rates are concentrated in
the Northeast, California, and certain parts of Texas and Florida. The Midwest and Great Plains
saw relatively lower mortgage forbearance rates. Panels (b) and (c) show forbearance rates for auto
loans and revolving accounts, respectively. These two debt types again show high forbearance
rates on both the East and West coasts, as well as in the South. Finally, panel (d) shows student
loan forbearance rates. Unlike other debt types, student loans exhibit uniformly high forbearance
rates across the country, because the CARES Act automatically placed all federal student loans in
forbearance.
We explore how this regional heterogeneity in the forbearance rates is related to the zip-code socio
economic characteristics through a series of regressions of the form:
!. = '. + )*. + -. (2)
where !. is either the average forbearance rate or the percent of debt payments missed due to
forbearance in region r and *. is a region r vector of socio-economic characteristics. Averages are
taken over the period from March 2020 to October 2020.
Debt Forbearance and Race
We begin our regional analysis by investigating the intensity of forbearance across zip-code racial
composition. This analysis is motivated by the observation that the distributional consequences of
COVID-19 have been highly uneven across these lines. The pandemic and its broader economic
and health consequences have disproportionately impacted minorities, especially Black
Americans.
The estimates in panel (c) of Table 3 show that racial composition is highly correlated with
forbearance rates – areas with larger Black or Hispanic/Latino populations have higher mortgage
forbearance rates. The forbearance rates are also higher in areas with higher debt-to-income ratio
and higher pre-pandemic house prices. In columns (3) and (4), we carry out similar regressions,
but with a measure of the amount of payments missed due to forbearance in each zip code.
Specifically, the dependent variable is the total amount of payments missed due to forbearance
divided by the sum of all scheduled payments in a zip code. Again, we find that areas with larger
Black and Hispanic and Latino populations have more missed payments, as do areas with higher
pre-pandemic debt-to-income ratios and higher house prices. This evidence combined with our
22
individual-level findings suggests that an important share of recipients of debt forbearance are less
creditworthy, lower income, minority borrowers that are living in areas with higher house prices,
where affordability and debt payment constraints are likely to be more binding.
In Table A3 in the Appendix, we carry out similar regressions but for auto, revolving, and student
loan forbearance. The results in these tables show that zip codes with a larger Black and Hispanic
and Latino populations have large debt forbearance rates and larger share of missed payments.
We explore the relationship between forbearance and racial composition visually in Figure 8. We
categorize zip codes into groups based on the 25th, 50th, and 75th percentiles of percent minority,
where racial minorities are all non-white races. As we observe, across all debt categories except
for student debt, forbearance rates are significantly higher in areas with larger shares of minorities
and a larger shares of Blacks.
Debt Forbearance across Industries and Occupation
The pandemic has largely impacted industries that require travel and face-to-face interaction, but
a much smaller impact on industries where employees are able to easily work from home. Because
we do not have individual-level data on borrowers’ occupations, we instead investigate whether
regions more exposed to industries that were more harmed by the pandemic have higher
forbearance rates.
Column (1) of Table 4 regresses mortgage forbearance rates on a variety of industry and
occupation characteristics. Our results show that zip codes with larger percentages of the
population employed in agriculture, education, health, construction, and manufacturing have lower
forbearance rates. This makes sense, as these occupations are typically considered “essential”, so
likely continued operations throughout the COVID-19 lockdowns. On the other hand, zip codes
with larger percentages employed in arts, recreation, and entertainment had higher forbearance
rates. Again, this is not surprising, since most of these industries were forced to either cease or
dramatically reduce operations throughout the pandemic. We also find that areas with more
workers able to work from home prior to COVID-19 have lower debt forbearance rates, while
regions with larger numbers of service and sales jobs have higher rates of forbearance.
In column (2) of Table 4, we explore how these industry and occupation characteristics correlate
with the percent of scheduled payments missed in a zip code. Our findings are similar to those in
column (1) – areas with higher concentrations of agriculture, construction, manufacturing, health,
and education have lower percentages of payments missed, while areas with more service, sales,
arts, recreation, and entertainment occupations have larger percentages of payments missed.
Table A4 in the Appendix shows that our findings for auto, revolving, and student loans are similar
to those of mortgages. Overall, these findings highlight that both forbearance rates and missed
payments are higher in regions with a larger presence of industries and occupations that have been
23
hit the hardest by the pandemic. Areas that have larger exposure to industries that involve travel
or face-to-face interactions have higher forbearance rates, while regions with larger concentrations
of “essential” industries or industries that can continue operations remotely have lower forbearance
rates.
Debt Forbearance and COVID-19 Impact
Our above results suggest that forbearance policies appear to be benefiting the regions that have
type of jobs most likely impacted by the pandemic. To investigate this further we directly analyze
whether the regions that were impacted the most by COVID-19 – both in terms of infection rates
and in severity of lockdowns – have higher forbearance rates.
We start by noting that there has also been significant regional heterogeneity in the impact of
COVID-19, both in terms of infection rates and in severity of lockdowns. Figure 7 plots average
COVID-19 case rates, the percent of workers in industries at risk of being impacted by COVID-
19, average changes in time spent at workplaces relative to January 2020, and average changes in
debit/credit card spending over relative to January 2020. Averages are taken over the period from
March 2020 to October 2020. Panel (a) shows that during March-October 2020, COVID case rates
were elevated in the Northeast, the South, and the Southwest. Decreases in spending and time spent
at workplaces were concentrated in the Northwest, California, and southern Florida. The percent
of workers in these industries is less concentrated by region, but the South and Southwest appear
to have higher concentrations of these workers.
Next, we investigate whether the regions that were impacted the most by COVID – both in terms
of infection rates and in severity of lockdowns – have higher forbearance rates. In column (1) of
Table 5, we regress county mortgage forbearance rates on a county-level characteristics capturing
the severity of the COVID-19 impact. Our results show that counties with higher average COVID
case rates have higher mortgage forbearance rates, as do counties with higher unemployment
insurance claim rates and higher percentages of workers in “at risk” industries12, where “at risk”
industries are those at most risk of having operations disrupted by the pandemic. On the other
hand, we find that counties with lower changes in credit/debit card spending and time spent at
workplace relative to January 2020 have lower forbearance rates. Therefore, our results imply that
counties impacted more by COVID infections and restrictions experience higher rates of mortgage
debt forbearance.
In columns (3) we carry out the same regressions, but with the percent of scheduled payments
missed in a county as the dependent variable. We find that higher COVID-19 case rates, more
unemployment insurance claims, and higher percentages of workers in at risk industries are
We define “at risk” industries according to criteria described here: https://www.brookings.edu/blog/the-
12
avenue/2020/03/17/the-places-a-covid-19-recession-will-likely-hit-hardest/
24
associated with more missed payments. Counties with lower declines in consumption and time
spent at workplaces experienced lower percentages of missed payments. For the most part,
relationships between COVID characteristics and forbearance rates are similar for the other debt
types (see Table A5 in the Appendix).
We illustrate visually the relationship between forbearance rates and COVID infections,
unemployment, and the percentage of workers in at risk industries in Figure 9. Once again, we
split counties into groups based on the 25th, 50th, and 75th percentiles of these variables. In panel
a) of Figure 9, we find that mortgage forbearance rates increase monotonically with COVID case
rates, unemployment insurance claim rates, and the percentage of workers in at risk industries.
Auto and revolving loans exhibit similar patterns. Unlike the other types of debt, student loan
forbearance is high in all high COVID, unemployment, and industry groups, again reflecting the
less targeted automatic forbearance mandate that applied to almost all student loans.
Taken together, our individual and regional analysis suggests that debt forbearance has reached its
intended target: financially vulnerable borrowers living in regions that experienced the highest
COVID-19 infection rates and the greatest deterioration in their economic conditions.
Debt Forbearance and other COVID-19 Relief Programs
We conclude this section by investigating the relation between debt forbearance and other stimulus
measures to shed light on the extent to which debt forbearance complents these other polices at the
region (state) level. In particular, we compare the amount of debt relief provided by forbearance
to other COVID relief programs, including stimulus checks, unemployment benefits, and
Paycheck Protection Plan loans.
Figure 10 shows the association between the total amount of money missed due to forbearance in
a state normalized by the number of adults against the total amount of money received from
stimulus checks, PPP loans, and unemployment insurance normalized by the number of adults at
the state level. Panel (a) shows the relationship between forbearance missed payments and the
amount of money from stimulus checks, while panels (b) and (c) show the correlation between
forbearance and money from unemployment benefits and PPP loans, respectively. This figure
shows that there is a positive relationship between money missed in forbearance and money from
unemployment benefits and PPP loans, but that states with higher amounts of stimulus checks
money per individual received relatively less relief through forbearance policies.
Overall, these findings are in line with our individual evidence, which suggests that debt
forbearance may have importantly complemented other stimulus measures. Unlike polices based
solely on income, such as stimulus checks, debt forbearance allowed less creditworthy and
financially constrained borrowers to receive financial relief. In doing so, debt forbearance provided
25
a significant financial relief to individuals and regions with higher pre-pandemic incomes who
were not directly targeted by the stimulus check program.
Section V: Government Mandates for Debt Forbearance
We conclude our analysis by studying the significant role of the private sector in extending debt
relief. We use the analysis to evaluate the role of implicit forbearance subsidies forbearance and
that borrower self-selection in determining the level and incidence of forbearance.13 We will also
shed light on the role of (supply-side) intermediary factors.
As we discuss above, the COVID-19 pandemic differs greatly from other economic crises by the
absence of household debt distress despite the relatively smaller size of the expenditures in the
program. We argue that forbearance may have been effective at preventing household debt distress,
because it was provided to households in need: those with lower credit scores, and affected by
COVID related shocks. In this section, we first document the substantial amount of private debt
forbearance. We then exploit the variation in government mandates across different types of loans
to better understand the role of two forces in determining the extent and incidence of debt relief:
self-selection by borrowers, and implicit forbearance subsidies by government programs.
As we observe from Figure 2, financial institutions provided substantial debt forbearance in
segments (auto and revolving debt) outside explicit government mandates as well as in mortgage
segments, which were not covered by the government mandate. In aggregate, more than a quarter
(28.7%) of financial relief due to forbearance was provided outside of the government mandates.
This evidence suggests that a considerable amount of debt forbearance would have existed even
in the absence of explicit government mandates embedded in the CARES Act.
V.A Debt Forbearance Mandates: The Role of Self-Selection
There are generally two steps in determining, which borrowers obtain debt relief. First, the
borrower must request relief from the lender—self-selection. Second, the lender must agree to
provide relief, which was mandated for government loans. As we discuss in Section II.B, the
13
The consumer debt design literature indicates that private debt contracts should allow for a certain amount of debt
relief during economic downturns to limit deadweight costs of default and allow better risk-sharing between borrower
and lenders (e.g., see Piskorski and Tchistyi 2010, 2011, 2017; Eberly and Krishnamurthy 2014; Greenwald et al.
2020; Guren et al. 2020, Landvoigt et al. 2020; Campbell et al. 2020). The potential to generate moral hazard problems
in repayments (Keys et al. 2013; Mayer et al. 2014) can limit the scope of debt relief that can be provided in an
incentive-compatible way by the private sector. Proponents of government intervention in debt relief argue that such
policies prevent excessive default and foreclosures that may not only lead to deadweight losses for borrowers and
lenders but also generate negative externalities for the society (e.g., Campbell et al. 2011; Guiso et al. 2011; Melzer
2010; Mian et al. 2015; Piskorski and Seru 2020; Gupta 2020; Diamond et al. 2020). In addition such interventions
may help circumvent variety of frictions including rigidity and incompleteness of debt contracts and limited ability
and incentives of financial intermediaries to modify loan terms that were documented to hamper the implementation
of debt relief by the private sector (see Piskorski et al. 2010; Agarwal et al. 2011; Agarwal et al 2017, 2020, Di Maggio
et al. 2017, Maturana 2017; Kruger 2018; Piskorski and Seru 2018 and 2020).
26
CARES Act guaranteed forbearance for all government-backed mortgages, accounting for about
two-thirds of outstanding mortgage debt, though forbearance was not automatic. Federally held
student loans, accounting for vast majority of student loans, were also guaranteed forbearance, but
unlike mortgages, these loans were automatically placed in administrative forbearance. On the
other hand, auto and revolving debt was not directly affected by the CARES Act, so lenders were
not required to grant forbearance on these debt types.
The take-up rate of mortgages covered by the government mandate is a priori evidence that self-
selection is a first order force in determining forbearance rates. Less than 10% of eligible borrowers
request relief, and of those, about two-third take advantage of the ability to defer payments. We
also want to understand the degree to which self-selection is responsible for the distribution of debt
relief across households. Recall that forbearance rates are higher among lower income and less
creditworthy borrowers, resulting in relief that is provided to households who are more likely
distressed. To obtain mortgage forbearance under CARES ACT, eligible mortgage borrowers had
to apply for forbearance, but forbearance for those loans was mandated. In other words, borrower’s
self-selection is driving forbearance rates for these loans.
Forbearance of non-eligible (private) loans, on the other hand, must be approved by the lender.
We already documented that forbearance rates decline in income and creditworthiness for private
and government loans across mortgages, part of which are covered by government mandates, as
well as auto loans and revolving credit. In Figure 12, we narrow the comparison to mortgages, and
compare patterns and consider how forbearance rates vary across borrower characteristics based
on whether the mortgage is covered by the government mandate in Figure 12. We that the
distribution of debt relief as well as missed payment percentage across borrower characteristics is
also quite comparable between government and private mortgages (Figure A3).
Finally, we investigate whether the relationship between the amount of forbearance flowing to a
region and the severity of the COVID crisis in a region differs depending on whether loans are
covered by the forbearance mandates. We re-estimate regressions from Table 5 for mortgages cut
by government mandates. Again, we find that forbearance patterns are similar for both types of
mortgages – regions with higher unemployment insurance claims, more COVID cases, larger
changes in spending and time spent at workplaces, and higher percentages of workers in at risk
industries have higher rates of forbearance and missed payments for both types of loans (though
some of these associations are statically insignificant). These results suggest that borrowers’ self-
selection is important in determining how relief is allocated. This self-selection might explain why
27
the program has been so effective at preventing household debt distress, despite its somewhat
smaller expenditures.14
V.B Implicit Forbearance Subsidies
Private debt relief is presumably a result of an (ex-ante) mutually beneficial renegotiation.
Government mandated relief does not need to be mutually beneficial: it can result in a transfer
from the lender (government) to the borrower. Such subsidies to household debt relief may very
well be warranted in the presence of renegotiation frictions and aggregate spillover that can result
from distressed household debt. Because CARES Act eligible debt differs from private debt both
in its type, and the types of borrowers, it is difficult to evaluate the magnitude of subsidies. To
evaluate the importance of implicit government subsidies we exploit a size discontinuity in
eligibility of mortgages for relief under the CARES Act.
We first start by considering all mortgages and re-estimate the specifications in Table 4 with an
additional control Mandate that is an indicator variable that takes on the value of 1 if the loan is
covered by forbearance mandates under the CARES Act (Fannie, Freddie, FHA, and VA loans)
and is 0 otherwise. Panel (a) of Table 6 shows that loans covered by the mandate are about 1.16
percentage points more likely to be in forbearance, implying about 30% higher forbearance rate
relative to loans not covered by the forbearance mandate. Columns (3) and (4) show that there are
not economically meaningful differences in the share of borrowers missing payments between
these two categories of loans.
To conduct a tighter comparison, we exploit a size discontinuity in eligibility of mortgages for
relief under the CARES Act. Unlike government-backed loans, the CARES Act does not require
lenders to extend forbearance to borrowers holding jumbo loans. Jumbo loans, unlike conforming
loans, exceed the limits set by the Federal Housing Finance Agency and cannot be purchased,
guaranteed, or securitized by GSEs. These origination loan balance limits were $417,000 for a
single family home and $625,000 thought most of the last decade reaching $510,400 and $765,600,
respectively by 2020 after their progressive increases during 2017-2020 period.
We first consider the relationship between forbearance rates and eligibility status by plotting
mortgage forbearance rates around the conforming loan limit. Panel (a) of Figure 11 shows that
forbearance rates are approximately 1.6pp lower to the right of the conforming loan limit, which
correspond to loans not covered by the mandate. Panel (b) of Figure 11 shows that prior to the
pandemic the forbearance rates were also similar across these two groups, hence the pre-existing
differences in the forbearance rates are unlikely to affect our above findings. In fact, forbearance
14
We note that as of January 2020, debt payments for loans in forbearance are deferred and not forgiven. Hence the
ultimate transfer to borrowers resulting from debt forbearance could be much smaller than the amount of their missed
payments during the first year of the pandemic.
28
rates were slightly higher for private loans, although the difference is economically small. Panel
(b) of Table 6 confirms these results and shows that accounting for borrower and loan
characteristics the loans to the left-hand side of the balance limit (covered by the mandates) have
about 1.5-1.6pp higher forbearance rates.
We can also examine these patterns over time. Panel (c) of Figure 11 plots the forbearance rates
on these mortgages from January 2019 to October 2020. This figure shows that prior to the
pandemic, forbearance rates were nearly identical for loans with balances below the conforming
loan limit and jumbo mortgages. Forbearance rates dramatically increased for both group of loans
in April and May, but the increase was larger for government loans starting in May and has kept
increasing over the period. This result suggests that the mandatory approval of loan forbearance
indeed increased debt relief beyond what would have been the outcome of private renegotiation
between the lenders. In other words, the estimate suggests that the private mortgage debt relief
incorporated an implicit subsidy to borrower.
We next formally estimate the amount of government loans, which obtained relief above what the
private sector would have provided. We exploit the size discontinuity within a difference-in-
differences design. Specifically, we consider only mortgages with balances that fall within 90% to
110% of the conforming loan limit, and restrict ourselves to mortgages owned by the government
below the loan limit. Loans of these sizes should be relatively comparable, though those with
balances below the conforming limit are covered by government forbearance mandates. We
estimate the following specification:
/0123414563",$,% = ' + )7 ,058",% + ): ;45<3=>6$ + )? ;45<3=>6$ ∗ ,058",% + +*".% + B C% + -",$,% (3)
where /0123414563",$,% is an indicator for whether loan i located in zip code z is in forbearance
during month t. ,058",% (conforming) is an indicator for whether the loan is a government-insured
and eligible for government forbearance, ;45<3=>6$ is an indicator that takes on the value of zero
prior to March 2020 and one from March 2020 to October 2020, *",% is a vector of borrower level
characteristics, and C% is either a vector of zip code controls or fixed effects.
The results in panel (c) of Table 6 show that the coefficient on the interaction term ;45<3=>6$ ∗
,058",% , which captures differential change in the forbearance rate between treatment group and
control group after the pandemic, is positive and statistically significant. This implies that during
the pandemic, mortgages covered by the government mandate loans were about 1.15-1.30pp more
likely to be in forbearance relative to ineligible (jumbo), implying about 25% higher forbearance
rate relative to mean forbearance rate of jumbo loans during the pandemic. Panel (d) of Figure 11
shows the pattern of these differences over time in our sample period. Again, like in the overall
sample of loans, we find little economic differences in the percentage of missing payments
conditional on being forbearance across these two group of loans.
29
Taken together these results suggest that implicit government debt relief subsidies imbedded in
the CARES Act mandates increase the rate of forbearance by about 25%. In other words, our back
of the envelope estimates suggests that about a quarter of forbearance provided on loans covered
by the government mandates is subsidized, and the rest is provided to borrowers who would have
been eligible under a private benchmark. Overall, this suggests, that the vast majority of the
mortgage relief provided by the CARES Act was provided without an implicit subsidy, and
mirrored the decisions of private lenders.
V.C The Supply of Debt Relief and the Regulatory Regime of Servicers
Debt relief is not implemented directly by debt holders. Instead, servicers, who oversee collecting
payments from borrowers, are in charge of choosing who obtains forbearance. This is the case
even for mortgages covered by the government mandate. Intermediary specific factors (e.g.,
capacity constraints) played a significant role in implementation of debt relief during the Great
Recession (e.g., Agarwal et al. 2017; Fuster et al. 2018, Piskorski and Seru 2018; 2020). One
important difference between government provided debt relief during the Great Recession and the
one provided in the CARES Act is eligibility. Forbearance during the CARES Act applies to all
government backed loans and the lender does not have to determine in any other way if the
borrower qualifies. In principle, all the borrower must do is to ask. So, there should be little scope
for differences in intermediary supply of forbearance. Despite that mandate, we document
differences in forbearance rates across different types of servicers.
We start by observing that since the Great Recession, there has been a significant growth of shadow
bank lending, with shadow banks – non depository institutions -- accounting for about half of loan
originations in the mortgage market (Buchak et al. 2018, Jiang 2019, Jiang et al. 2020). These
institutions are under different regulatory auspices from traditional banks. To investigate whether
the regulatory type of the financial intermediary is correlated with its provision of forbearance. We
turn to Fannie Mae loan level data that has detailed information on the identity of the financial
institution that originated the loan.
All Fannie Mae loans had mandated forbearance under the CARES Act. Despite that, Panel (a) of
Figure 13 shows that loans originated by banks have higher rates of debt forbearance than shadow
banks. Bank originated loans experienced about 2pp higher forbearance rates compared to similar
loans originated by shadow banks, and the difference persists throughout the period. One possible
source of this difference is that the composition of loans serviced by the two types of institutions
differs, either because they service different types of borrowers or markets. In Table 7, we show
that even controlling for extensive borrower characteristics, and locations, the difference persists.
During the pandemic, shadow banks loans we approximately 2pp less likely to be in forbearance.
We cannot completely rule out the alternative that borrowers serviced by shadow banks have
different demand for forbearance, despite their very similar characteristics. Nevertheless, this
30
result strongly suggests that despite the blanket government mandate, who implements forbearance
has a sizeable effect on the amount of debt relief provided.
Beyond regulatory differences, one can divide loan servicing in two categories. Servicing of loans,
which the lender originated, and servicing of loans which they did not originate, but for which they
acquired servicing rights. Panel (b) of Figure 13 shows that loans for which the servicing was
outsourced also have lower rates of forbearance. In unreported results we find that this effect is
mainly driven by bank originated loans: loans for which the banks retain their servicing rights have
higher forbearance rates than comparable loans for which banks transferred their servicing rights.
One potential mechanism to explain this fact is that loan originators have more information about
borrowers than outsourced servicers, so they can more easily tell which borrowers really require
relief. This would imply that even in government mandated relief, servicers try to evaluate whether
forbearance should be granted, which would be at odds with the formal mandate. Overall, the
evidence suggests that intermediary specific factors played a meaningful role in implementation
of the government mandates and may have affected the overall level of debt relief provided. Both
the regulatory framework of the servicers, as well as potential access to information may have
played a role in determining the supply response.
Section VI: Conclusion
We followed a representative panel of more than twenty million US consumers and analyzed the
widespread debt forbearance actions during the COVID-19 pandemic. While the debt delinquency
rates remained low, the rate of debt forbearance that allowed households to temporarily suspend
loan repayment increased dramatically following the CARES Act, providing households with
significant financial relief. Between March and October 2020, 5.2 million mortgages ($1.1
trillion), 8.7 million of auto loans ($130 billion), 66 million student loans ($580 billion) and 51
million revolving loans ($99 billion) were in forbearance. At this rate, more than 60 million US
consumers will miss about $70 billion by the end of 2021:Q1. We find that the debt relief reached
its intended target, since forbearance rates are higher in regions with the highest COVID-19
infection rates and the greatest local economic deterioration. Individuals with lower credit scores,
lower incomes, and higher debt balances, and regions with higher shares of minorities received
higher rates of forbearance. Conditional on being in forbearance, borrowers with higher “pre-
pandemic income” received the largest dollar amount of relief per individual. Forbearance actions,
therefore, complemented other income based policy measures, providing more than half of
aggregate forbearance relief to individuals who were ineligible for policies like stimulus checks.
Not all of the forbearance was mandated under the CARES act. The private sector provided
substantial debt forbearance in segments (auto debt, revolving debt and jumbo mortgages) outside
of explicit government mandates. Moreover, allowing borrowers a choice of whether to request
debt relief might have resulted in a potentially better targeted policy.
31
We estimate that borrowers in forbearance will be left with a “forbearance overhang” of more than
$60 billion in accumulated postponed repayments. This forbearance debt overhang amounts to
about $1,800 per individual, which is more than half of their average monthly income, and more
than 80% for lower income borrowers. The extent of forbearance overhang suggests that the
unwinding of forbearance – which could be done in several ways -- could have first order
consequences for household debt distress, and through it, for the aggregate economy.
Our findings suggests that massive consumer debt forbearance actions can help explain why,
unlike during the Great Recession, the standard household debt distress channel was largely absent
during the first year of the COVID-19 pandemic. There are a number of possible reasons for
quicker and more widespread implementation of debt relief during the COIVD-19 pandemic
relative to the Great Recession. First, the private sector and policymakers may have internalized
the lessons from the Great Recession pointing to significant costs of widespread defaults and
foreclosures and were more willing to provide debt relief.
Second, it is possible that the underlying adverse shock has been perceived as more transitory
relative to the prior crisis, which could promote a more widespread deployment of temporary debt
relief measures by the private sector. This is consistent with the consumer debt design literature
that indicates that lenders should provide a certain amount of debt relief during economic
downturns to limit deadweight costs of default and allow better risk-sharing between borrowers
and lenders, especially if the underlying shocks are transitory. Relatedly, the COVID-19 shock is
a rare aggregate “exogenous” shock that is largely outside of the agents’ influence. This should
also alleviate concerns about the moral hazard effects of debt relief on incentives to repay debt,
leading to a more widespread loan renegotiation effort.
Finally, the nature of financial intermediation has significantly changed since the last crisis. Prior
to the Great Recession most of the riskiest consumer debt, the so-called subprime mortgages, was
not retained on the balance sheet of intermediaries and instead was sold through securitization to
third parties (Mian and Sufi 2009; Keys et al. 2010, 2013). In contrast, the private securitization
market was very limited even prior to the COVID-19 pandemic and lenders retained the majority
of jumbo (non-GSE) mortgages on their balance sheets (see Buchak et al. 2020). As a result, the
lenders may have more economic incentives to undertake voluntary debt relief actions on these
loans, even though these fall outside explicit government mandate.
32
References (Preliminary and incomplete)
Auclert, A., 2019. Monetary policy and the redistribution channel. American Economic Review
109, 2333–2367.
Auclert, A., Dobbie, W., Goldsmith-Pinkham, P., 2019. Macroeconomic effects of debt relief:
Consumer bankruptcy protections in the Great Recession. Unpublished working paper. NBER.
Agarwal, S., Amromin, G., Ben-David, I., Chomsisengphet, S., Evanoff, D., 2011. The role of
securitization in mortgage renegotiation. Journal of Financial Economics 102, 559–578.
Agarwal, S., Amromin, G., Ben-David, I., Chomsisengphet, S., Piskorski, T., Seru, A., 2017.
Policy intervention in debt renegotiation: Evidence from the Home Affordable Modification
Program. Journal of Political Economy 125, 654–712.
Agarwal, S., Amromin, G., Chomsisengphet, S., Landvoigt, T., Piskorski, T., Seru, A., Yao V.,
2020. Mortgage refinancing, consumer spending, and competition: Evidence from the Home
Affordable Refinance Program. Review of Economic Studies, forthcoming.
Agarwal, S., Chomsisengphet, S., Mahoney, N., Stroebel, J., 2018. Do banks pass through credit
expansions to consumers who want to borrow? Quarterly Journal of Economics 133, 129–190.
Andersen, S., Campbell, J., Nielsen, K., Ramadorai, T., 2020. Sources of inaction in household
finance. American Economic Review 110, 3184-3230.
Baker, S. R. Farrokhnia, S. Meyer, M. Pagel, C. Yannelis, 2020, Income, liquidity, and the
consumption response to the 2020 economic
Benmelech, E., Meisenzahl, R., Ramcharan, R., 2017. The real effects of liquidity during the
financial crisis: Evidence from automobiles. Quarterly Journal of Economics 132, 317–365.
Beraja, M., Fuster, A., Hurst, E., Vavra, J., 2019. Regional heterogeneity and monetary policy.
The Quarterly Journal of Economics 134, 109–183.
Berger, D., Guerrieri, V., Lorenzoni, G., Vavra, J., 2017. House prices and consumer spending.
The Review of Economic Studies 85, 1502–1542.
Berger, D., Milbradt, K., Tourre, F., Vavra, J., 2019. Mortgage prepayment and path-dependent
effects of monetary policy, working paper.
Berger, D., Turner, N., Zwick, E., 2020. Stimulating housing markets. Journal of Finance 75, 277–
321.
Buchak, G., Matvos, G., Piskorski, T., Seru, A., 2018. Fintech, regulatory arbitrage, and the rise
of shadow banks. Journal of Financial Economics 130, 453–483.
Buchak, G., Matvos, G., Piskorski, T., Seru, A., 2020. Beyond the balance sheet model of banking:
Implications for bank regulation and monetary policy. Unpublished working paper. NBER.
Campbell, J., Giglio, S., Pathak, P., 2011. Forced sales and house prices. American Economic
Review 101, 2108–2131.
Campbell, J., Clara, N., Cocco, J., 2020. Structuring mortgages for macroeconomic stability.
Unpublished working paper. NBER.
33
Chetty, Raj, John N. Friedman, Nathaniel Hendren, Michael Stepner, and the Opportunity Insights
Team, 2020, “How did COVID-19 and stabilization policies affect spending and employment? A
new real-time economic tracker based on private sector data, working paper.
Coibion, O., Y. Gorodnichenko, and M. Weber, 2020, Labor Markets During the COVID-19
Crisis: A Preliminary View, working paper.
Cox, N., P. Ganong, P. Noel, J. Vavra, A. Wong, D. Farrell, and F. Greig, 2020, Initial mpacts of
the pandemic on consumer behavior, working paper.
Diamond, R., Guren, A., Tan, R., 2020. The effect of foreclosures on homeowners, tenants, and
landlords. Unpublished working paper. Unpublished working paper. NBER.
Di Maggio, M., Kermani, A., Keys, B., Piskorski, T., Ramcharan, R., Seru, A., Yao, V., 2017.
Interest rate pass-through: Mortgage rates, household consumption and voluntary deleveraging.
American Economic Review 107, 3550–3588.
Di Maggio, M., Kermani, A., Palmer, C., 2020. How quantitative easing works: Evidence on the
refinancing channel. Review of Economic Studies 87, 1498–1528.
Eberly, J., Krishnamurthy, A., 2014. Efficient credit policies in a housing crisis. Brookings Papers
on Economic Activity. (Fall): 73–119.
Eichenbaum, M., Rebelo, S., Wong, A., 2019. State dependent effects of monetary policy: The
refinancing channel. Unpublished working paper.
Elenev, V., T. Landvoigt, and S Van Nieuwerburgh, 2020, Can the COVID Bailouts Save the
Economy, working paper?
Favilukis, J., Ludvingson, S., Van Nieuwerburgh, S., 2017. The macroeconomic effects of housing
wealth, housing finance, and limited risk sharing in general equilibrium. Journal of Political
Economy 125, 140–223.
Fuster, A., Lo, S., Willen, P., 2017. The time-varying price of financial intermediation in the
mortgage market. Unpublished working paper. Federal Reserve Bank of New York.
Fuster, A., Willen, P., 2017. Payment size, negative equity, and mortgage default. American
Economic Journal: Economic Policy 9, 167–191.
Ganong, P., Noel, P., 2017. The effect of debt on default and consumption: Evidence from housing
policy in the Great Recession. Unpublished working paper. Harvard University.
Granja, Joao, Christos Makridis, Constantine Yannelis, and Eric Zwick, “Did the Paycheck
Protection Program Hit the Target?,” National Bureau of Economic Research Working Paper,
2020.
Guerrieri, V., G. Lorenzoni, L. Straub, and I. Werning, 2020, Macroeconomic implications of
COVID-19: Can negative supply shocks cause demand shortages?, working paper.
Greenwald, D., 2018. The mortgage credit channel of macroeconomic transmission. Unpublished
working paper. MIT Sloan, Cambridge, Massachusetts.
Greenwald, D., Landvoigt, T., Van Nieuwerburgh, S., 2018. Financial fragility with SAM?
Unpublished working paper. MIT Sloan, Cambridge, Massachusetts.
34
Guerrieri V., Uhlig, H., 2016. Housing and credit markets. In: Vol 2., Handbook of
Macroeconomics, Chapter 17. Elsevier, Amsterdam, Netherlands, pp. 1427–1496.
Gupta, A., 2019. Foreclosure contagion and the neighborhood spillover effects of mortgage
defaults. Journal of Finance 74, 2249–2301.
Guren, A., Krishnamurthy, A., McQuade, T., 2020. Mortgage design in an equilibrium model of
the housing market. Journal of Finance, forthcoming.
Guren, A., McKay, A., Nakamura, E., Steinsson, J., 2018. Housing wealth effects: The long view.
Unpublished working paper. NBER.
Hsu, J., Matsa, D., Melzer, B., 2018. Unemployment insurance as a housing market stabilizer.
American Economic Review 108, 49–81.
Hurst, E., Keys, B., Seru, A., Vavra, J., 2016. Regional redistribution through the U.S. mortgage
market. American Economic Review 106, 2982–3028.
Jiang, E., 2019. Financing competitors. Unpublished working paper. The University of Texas at
Austin.
Jiang, E., Matvos, G., Piskorski, T., Seru, A., 2020. Banking without deposits: Evidence from the
shadow bank call reports. Unpublished working paper. NBER.
Kaplan, G., Mitman, K., Violante, G., 2017. The housing boom and bust: Model meets evidence.
Journal of Political Economy, forthcoming.
Keys, B., Mukherjee, T., Seru, A., Vig, V., 2010. Did securitization lead to lax screening?
Evidence from subprime loans. Quarterly Journal of Economics 125, 307–362.
Keys, B., Piskorski, T., Seru, A., Vig, V., 2013. Mortgage financing in the housing boom and bust.
In: Glaeser, E., Sinai, T., (Eds.), Housing and Financial Crisis. University of Chicago Press,
Chicago, pp. 143–204.
Kruger, S., 2018. The effect of mortgage securitization on foreclosure and modification. Journal
of Financial Economics 129, 586–607.
Looney A., and C. Yannelis, 2019, The consequences of student loan credit expansions: Evidence
from three decades of default cycles, working paper.
Maturana, G., 2017. When are modifications of securitized loans beneficial to investors? Review
of Financial Studies 30, 3824–3857.
Mayer, C., Morrison, E., Piskorski, T., Gupta, A., 2014. Mortgage modification and strategic
behavior: Evidence from a legal settlement with Countrywide. American Economic Review 104,
2830–2857.
Mayer, C., Pence, K., Sherlund, S., 2009. The rise in mortgage defaults. Journal of Economic
Perspectives 23, 27–50.
Melzer, B., 2017. Mortgage debt overhang: Reduced investment by homeowners at risk of default.
Journal of Finance 72, 575–612.
Mian, A., Sufi, A., 2009. The consequences of mortgage credit expansion: Evidence from the U.S.
mortgage default crisis. Quarterly Journal of Economics 124, 1449–1496.
35
Mian, A., Sufi A., 2011. House prices, home equity-based borrowing, and the U.S. household
leverage crisis. American Economic Review 101, 2132–2156.
Mian, A., Sufi, A., 2012. The effects of fiscal stimulus: Evidence from the 2009 “Cash for
Clunkers” program. The Quarterly Journal of Economics 127, 1107–1142.
Mian, A., Sufi, A., 2014a. What explains the 2007–2009 drop in employment? Econometrica 82,
2197–2223.
Mian, A., Sufi A., 2014b. House of Debt: How They (and You) Caused the Great Recession, and
How We Can Prevent It from Happening Again. University of Chicago Press, Chicago.
Mian, A., Sufi, A., Trebbi, F., 2015. Foreclosures, house prices, and the real economy. The Journal
of Finance 70, 2587–2634.
Mueller, H. M., and C. Yannelis, 2019, The rise in student loan defaults, Journal of Financial
Economics 131, 1-19.
Mueller, H. M., and C. Yannelis, 2020, Reducing barriers to enrollment in federal student loan
repayment plans: Evidence from the Navient field experiment, working paper.
Parker, J., Souleles, N., Johnson, D., McClelland, R., 2013. Consumer spending and the economic
stimulus payments of 2008. American Economic Review 103, 2530–2553.
Piskorski, T., Tchistyi, A., 2010. Optimal mortgage design. Review of Financial Studies 23, 3098–
3140.
Piskorski, T., Tchistyi, A., 2011. Stochastic house appreciation and optimal mortgage lending.
Review of Financial Studies 24, 1407–1446.
Piskorski, T., Tchistyi, A., 2017. An equilibrium model of housing and mortgage markets with
state-contingent lending contracts. Unpublished working paper. NBER.
Piskorski, T., Seru, A., Vig, V., 2010. Securitization and distressed loan renegotiation: Evidence
from the subprime mortgage crisis. Journal of Financial Economics 97, 369–397.
Piskorski, T., Seru, A., Witkin, J., 2015. Asset quality misrepresentation by financial
intermediaries: Evidence from the RMBS market. Journal of Finance 70, 2635–2678.
Piskorski, T., Seru, A., 2018. Mortgage market design: Lessons from the Great Recession.
Brookings Papers on Economic Activity. (Spring): 429–513.
Scharfstein, D., Sunderam, A., 2016. Market power in mortgage lending and the transmission of
monetary policy. Unpublished working paper. Harvard Business School.
Shiller, R., 2008. The Subprime Solution: How Today’s Global Financial Crisis Happened, and
What to Do About It. Princeton University Press, Princeton, NJ.
Wheelock, D., 2008. The federal response to home mortgage distress: Lessons from the Great
Depression. Federal Reserve Bank of St. Louis Review 90, 133–148.
36
Table 1: Individual-Level Descriptive Statistics
This table shows summary statistics for all individuals in the Equifax dataset (Columns (1) through (4)) and all individuals in forbearance (Columns (5) through
(8)). Pre-COVID-19 statistics are based on January and February 2020, while COVID-19 statistics are based on March to October 2020. Odd columns contain
means while even columns contain standard deviations. Sources: Equifax 10% representative sample of the U.S. credit population.
All Individuals All Individuals in Forbearance
Pre-COVID-19 COVID-19 Pre-COVID-19 COVID-19
(1) (2) (3) (4) (5) (6) (7) (8)
Mean SD Mean SD Mean SD Mean SD
Vantage 701 122 708 119 645 122 665 113
Estimated Age 49 18 49 18 41 17 40 16
Estimated Income 39,992 19,272 40,099 19,382 34,063 19,250 34,126 18,607
Fraction with First Mortgage 0.22 0.22 0.18 0.21
Fraction with Auto Debt 0.35 0.35 0.38 0.40
Fraction with Revolving Debt 0.89 0.88 0.78 0.80
Fraction Student Debt 0.18 0.17 0.56 0.62
Average First Mortgage Balance (Nonzero Accounts) 210,020 240,631 214,788 246,309 232,542 284,857 234,805 266,933
Average Combined Mortgage Balance (Nonzero Accounts) 204,346 239,936 208,949 245,582 224,272 283,478 227,664 266,321
Average Auto Debt (Nonzero Accounts) 18,474 17,496 18,627 17,757 19,897 20,567 19,903 19,722
Average Revolving Debt (Nonzero Accounts) 5,047 9,178 4,498 8,690 7,369 12,370 6,651 11,388
Average Student Debt (Nonzero Accounts) 35,149 51,510 35,951 52,362 392,34 54,999 37,830 54,024
Average Credit Card Utilization 32% 28% 49% 42%
Average Available Credit 17,512 23,338 17,858 23,551 14,993 25,582 152,06 24,661
Number of Consumers 20,265,012 20,217,750 3,702,943 8,309,795
37
Table 2: Consumer Debt Forbearance during COVID-19
Panel (a) of this table shows summary statistics related to forbearance for different debt types during the COVID period (March through October 2020). We show
the average monthly delinquency and forbearance rates, the average monthly percent of loans in forbearance with missed payments, the average percent of monthly
scheduled payments missed, and the average dollar amount missed each month by loans in forbearance along with the standard deviation [S.D.]. The bottom three
rows show the percent of borrowers that were ever in forbearance during the COVID period, the percent of borrowers that entered forbearance during the COVID
period, and the average cumulative dollar amount of missed payments per individual in forbearance during March-October 2020 period along with their standard
deviation. Column (1) show results for first mortgages, Columns (2) for auto loans, Column (3) for revolving accounts, and Column (4) for student loans. Sources:
Equifax 10% representative sample of the U.S. credit population.
Panel A: Mean Delinquency and Debt Forbearance Rates during the COVID period
Mortgages Auto Loans Revolving Loans Student Loans
(1) (2) (3) (4)
Monthly Delinquency Rate 1.80% 2.98% 2.05% 1.35%
Monthly Forbearance Rate 4.54% 2.60% 2.50% 80.70%
% Missing Payments | In Forbearance 65.09% 76.49% 81.39% 95.84%
% of Scheduled Payment Missed | In Forbearance 61.50% 67.42% 80.90% 90.64%
$ Monthly Amount Missed per Individual | In Forbearance [S.D.] $1,071.6 $269.6 $35.6 $27.2
[$3,413.9] [$317.2] [$125.0] [$91.8]
% Ever in Forbearance 8.80% 8.80% 4.63% 92.20%
% Entered Forbearance 7.92% 6.95% 5.28% 36.40%
$ Cumulative Amount Missed per Individual | In Forbearance [S.D.] $3,282.4 $428.7 $69.1 $140.9
[$12,487.3] [$688.6] [$324.8] [$488.4]
38
Table 2: Consumer Debt Forbearance and COVID-19 [continued]
Panel B: Aggregate Debt Forbearance Statistics
This panel provides aggregate statistics on loans that entered forbearance during the COVID period (March 2020 through October 2020). It includes estimates of
the aggregate percent of loan balance in forbearance, the total dollar amount of loans in forbearance at least one month, the total of number of loans experiencing
forbearance, the average total amount of payments missed by loans in forbearance each month, and the total dollar amount missed due to forbearance. Column (1)
show results for first mortgages, Columns (2) for auto loans, Column (3) for revolving accounts, and Column (4) for student loans. Sources: Equifax 10%
representative sample of the U.S. credit population.
Mortgages Auto Debt Revolving Debt Student Debt
(1) (2) (3) (4)
% of Loan Balance in Forbearance 9.10% 7.20% 8.40% 42.60%
$ Amount of Loans in Forbearance 1.1 Trillion 130 Billion 99 Billion 580 Billion
Number of Loans in Forbearance 5.2 Million 8.7 Million 51 Million 66 Million
$ Missed in Forbearance (av per month) 2.2 Billion 501 Million 510 Million 2.7 Billion
$ Missed in Forbearance (March-Oct 2020) 17 Billion 3.7 Billion 3.8 Billion 19 Billion
All Debt Types
Number of Loans that Entered Forbearance 139 Million
Number of Individuals who Entered Forbearance 61 Million
39
Table 2: Consumer Debt Forbearance and COVID-19 [continued]
Panel C: Share of financial relief due to debt forbearance allocated to above median income borrowers
This table shows the percentage of dollar amount financial relief due to debt forbearance went to above median income borrowers along with the associated total
$ amount of payments missed from March to October 2020. Panel (c1) shows similar statistics but based on the median income of all individuals in our dataset.
Panel (c2) shows median income of borrowers within a given debt type. Column (1) show results for first mortgages, Columns (2) for auto loans, Column (3) for
revolving accounts, and Column (4) for student loans. Sources: Equifax 10% representative sample of the U.S. credit population.
Panel C1: Share of financial relief due to debt forbearance allocated to above median income borrowers (all consumers)
Mortgages Auto Debt Revolving Debt Student Debt
(1) (2) (3) (4)
Median Income 37,000
% of Missed Payments from Above Median 88% 61% 46% 33%
Total Amount Missed from Above Median 15 Billion 2.3 Billion 1.7 Billion 6 Billion
Panel C2: Share of financial relief due to debt forbearance allocated to above median income borrowers (within each debt type)
Mortgages Auto Debt Revolving Debt Student Debt
(1) (2) (3) (4)
Median Income 52,000 41,000 37,000 22,000
% of Missed Payments from Above Median 61% 52% 46% 64%
Total Amount Missed from Above Median 10 Billion 1.9 Billion 1.7 Billion 12 Billion
40
Table 3: Mortgage Debt Forbearance Regressions
This table reports the results from a regression of whether an individual is in forbearance or whether an individual
missed their payment while in forbearance each month on various borrower and zip code level variables. Months
include March 2020 to October 2020. Columns (1) and (2) are estimated from all individuals with first mortgages,
while Columns (3) and (4) are restricted to individuals who are in forbearance on their first mortgages. Columns (1)
and (3) contain individual-level characteristics and zip code fixed effects, while columns (2) and (4) contain individual
level controls and zip code controls. Zip code controls include zip code racial composition, educational attainment,
unemployment rate, median age, median income, average CLTV, average DTI, and median house price. Regression
inputs are scaled by standard deviation. Individual controls are taken as of January 2020. Standard errors are clustered
at the individual level. Sources: Forbearance status, missed payment status, and all other individual-level
characteristics come from Equifax. House price data comes from Zillow. The remaining variables come from the U.S.
Census Bureau American Community Survey 5-Year Estimates.
Panel A: Forbearance and Individual Characteristics
Missed Missed
In Forbearance In Forbearance
Payment Payment
(1) (2) (3) (4)
Mortgage Balance 0.405*** 0.402*** -0.219* -0.178*
(0.0454) (0.0398) (0.0960) (0.0905)
Auto Loan Balance 0.326*** 0.292*** 0.0892 0.120
(0.0139) (0.0141) (0.0568) (0.0692)
Revolving Debt Balance 0.450*** 0.464*** 1.017*** 0.960***
(0.0117) (0.0151) (0.0465) (0.0560)
Student Loan Balance 0.251*** 0.263*** 0.543*** 0.503***
(0.00967) (0.0125) (0.0427) (0.0535)
Small Business Owner 0.331*** 0.322*** 0.390*** 0.474***
(0.00865) (0.0108) (0.0538) (0.0659)
Credit Score -0.0249*** -0.0245*** -0.0424*** -0.0436***
(0.000134) (0.000170) (0.000702) (0.000905)
Income -0.186*** -0.153*** 0.106 0.184
(0.0190) (0.0194) (0.0866) (0.0995)
# Accounts Past Due in Jan. 1.220*** 1.189*** -0.0942** -0.143**
(0.0154) (0.0206) (0.0342) (0.0445)
DTI 0.189*** 0.239*** 0.116 0.276***
(0.0381) (0.0448) (0.0906) (0.0585)
Age -0.610*** -0.553*** -1.779*** -1.688***
(0.00903) (0.0103) (0.0693) (0.0875)
Zip Code Controls No Yes No No
Zip Code FE Yes No Yes No
Mean of Dependent Variable 4.542 4.348 69.28 68.96
Observations 39,500,895 23,755,189 1,792,278 10,32,948
Adjusted R-squared 0.035 0.028 0.037 0.013
41
Table 3: Mortgage Debt Forbearance Regressions [continued]
Panel B: Mortgage Forbearance and Race
This table shows the results from regressions of the average monthly forbearance rate and the percent of scheduled
payments missed in each zip code on various zip code-level characteristics. Average monthly forbearance rates are
from March 2020 to October 2020 and the percent of scheduled payments missed from individuals in forbearance is
calculated over the same time period. Analysis is restricted to zip codes with at least ten loans. Regression inputs are
scaled by standard deviation. Sources: Forbearance status, missed payment status, and all other individual-level
characteristics come from Equifax. House price data comes from Zillow. The remaining variables come from the U.S.
Census Bureau American Community Survey 5-Year Estimates.
Forbearance Percent of Scheduled
Rate Payments Missed
(1) (2)
% College Educated -0.0818** 0.0450
(0.0292) (0.0329)
% Black 0.595*** 0.360***
(0.0289) (0.0346)
% Hispanic 0.743*** 0.558***
(0.0280) (0.0298)
Pre-Pandemic Unemployment Rate 0.177*** 0.159***
(0.0322) (0.0337)
Median Age 0.139*** 0.195***
(0.0250) (0.0310)
Median Income 0.178*** -0.138
(0.0418) (0.0726)
Average DTI 0.293*** 0.947***
(0.0859) (0.212)
Average House Price in January 0.137*** 0.178***
(0.0318) (0.0348)
Mean of Dependent Variable 4.013 2.697
# Of Obs. 16920 16920
R-Squared 0.1413 0.1187
42
Table 4: Mortgage Forbearance Zip-Code Industry and Occupation Characteristic
Regressions
This table shows the results from regressions of the average monthly forbearance rate and the percent of scheduled
payments missed in each zip code on various zip code-level characteristics. Average monthly forbearance rates are
taken from March 2020 to October 2020 and percent of scheduled payments missed from individuals in forbearance
is calculated over the same time period. Analysis is restricted to zip codes with at least ten loans. Regression inputs
are scaled by standard deviation. Sources: Forbearance status, missed payment status, and all other individual-level
characteristics come from Equifax. The remaining variables come from the U.S. Census Bureau American Community
Survey 5-Year Estimates.
Forbearance Percent of Scheduled
Rate Payments Missed
(1) (2)
% Production Industry 0.0527 -0.0445
(0.0327) (0.0358)
% Agriculture Industry -0.326*** -0.302***
(0.0315) (0.0318)
% Finance, Insurance, Real Estate Industry -0.0741* -0.0576
(0.0295) (0.0374)
% Arts, Recreation, Entertainment Industry -0.0182** -0.00583
(0.00659) (0.00721)
% Education, Health, Public Administration Industry -0.517*** -0.423***
(0.0308) (0.0347)
% Construction and Manufacturing Industry -0.721*** -0.572***
(0.0349) (0.0393)
% Work From Home - Pre-Pandemic -0.158*** -0.141***
(0.0314) (0.0357)
% Self -Employed - Pre-Pandemic 0.0242 0.0341
(0.0333) (0.0362)
% Service Occupations 0.366*** 0.271***
(0.0311) (0.0337)
% Sales and Office Occupations 0.129*** 0.0855**
(0.0266) (0.0267)
Mean of Dependent Variable 4.182 2.841
# Of Obs. 21704 21704
R-Squared 0.0830 0.0497
43
Table 5: Unemployment Claims, and COVID Infections, and Other Stimulus
This table shows the results from regressions of the average monthly mortgage forbearance rate and the percent of
scheduled payments missed in each zip code on various county-level characteristics. Averages are taken over the
period from March 2020 to October 2020 and percent of scheduled payments missed by individuals in forbearance is
calculated over the same time period. Analysis is restricted to counties with at least ten loans. Regression inputs are
scaled by standard deviation. Sources: Forbearance status and missed payment status come from Equifax. UI claims,
COVID cases, credit/debit card spending, and time spent at workplace are averages of March 2020 to October 2020
values from Opportunity Insights Tracker. Remaining variables are calculated using data from the U.S. Census Bureau
American Community Survey 5-Year Estimates.
Forbearance Percent of Scheduled
Rate Payments Missed
(1) (2)
Total UI Claims (Check Rate) 0.572*** 0.457***
(0.0653) (0.0637)
Total COVID Case Rate 0.433*** 0.312***
(0.0672) (0.0655)
Change in Credit/Debit Spending 0.0304 0.0227
(0.0699) (0.0682)
Change in Time Spent at Workplace -0.416*** -0.192
(0.104) (0.102)
% Families Receiving Full Stimulus Check -0.153 -0.240**
(0.0919) (0.0895)
% Workers in at Risk Industry 0.566*** 0.552***
(0.151) (0.147)
Amount of PPP Received Normalized by # of
Small Businesses -0.309 -0.363*
(0.167) (0.163)
Mean of Dependent Variable 4.001 2.739
# Of Obs. 348 348
R-Squared 0.4245 0.3260
44
Table 6: Government Mandates and Mortgage Forbearance
Panel A: Forbearance and Government Mandates (All Loans)
Panel (a) of this table reports the results from a regression of whether an individual is in forbearance or whether an individual
missed their payment while in forbearance each month on various borrower and zip code level variables. Months include March
2020 to October 2020. Mandate is an indicator variable that takes on the value of 1 if the loan is covered by forbearance mandates
under the CARES Act (Fannie, Freddie, FHA, and VA loans) and 0 otherwise. Columns (1) and (2) are estimated for all individuals
with first mortgages, while Columns (3) and (4) are restricted to individuals who are in forbearance on their first mortgages.
Columns (1) and (3) contain individual-level characteristics and zip code fixed effects, while columns (2) and (4) contain individual
level controls and zip code controls. Zip code controls include zip code racial composition, educational attainment, unemployment
rate, median age, median income, average CLTV, average DTI, and median house price. Regression inputs are scaled by standard
deviation. Individual controls are taken as of January 2020. Standard errors are clustered at the individual level. Sources:
Forbearance status, missed payment status, and all other individual-level characteristics come from Equifax. House price data
comes from Zillow. The remaining variables come from the U.S. Census Bureau American Community Survey 5-Year Estimates.
(1) (2) (3) (4)
% Missed % Missed
In Forbearance In Forbearance
Payment Payment
Mandate 1.166*** 1.176*** -0.654*** -0.904***
(0.0166) (0.0197) (0.145) (0.185)
Mortgage Balance 0.452*** 0.451*** -0.246* -0.210
(0.0488) (0.0418) (0.0964) (0.148)
Auto Loan Balance 0.323*** 0.291*** 0.0926 0.126
(0.0140) (0.0141) (0.0568) (0.0747)
Revolving Debt Balance 0.442*** 0.454*** 1.023*** 0.971***
(0.0117) (0.0150) (0.0466) (0.0606)
Student Loan Balance 0.254*** 0.263*** 0.543*** 0.507***
(0.00966) (0.0125) (0.0427) (0.0552)
Small Business Owner 0.337*** 0.329*** 0.389*** 0.473***
(0.00866) (0.0108) (0.0539) (0.0704)
Credit Score -2.302*** -2.260*** -3.874*** -3.984***
(0.0123) (0.0156) (0.0642) (0.0851)
Income -0.168*** -0.136*** 0.0858 0.155
(0.0197) (0.0199) (0.0868) (0.115)
# Accounts Past Due in Jan. 1.223*** 1.193*** -0.0937** -0.142**
(0.0154) (0.0206) (0.0342) (0.0450)
DTI 0.189*** 0.240*** 0.116 0.276***
(0.0380) (0.0447) (0.0909) (0.0660)
Age -0.597*** -0.545*** -1.793*** -1.703***
1.166*** 1.176*** -0.654*** -0.904***
Zip Code Controls No Yes No Yes
Zip Code FE Yes No Yes No
Mean of Dependent Variable 4.542 4.348 69.28 68.96
Observations 39,507,016 23,758,561 1,792,391 1,033,005
Adjusted R-squared 0.036 0.029 0.037 0.013
45
Table 6: Government Mandates and Mortgage Forbearance [continued]
Panel B: Forbearance Rates around the Conforming Loan Limit
Panel (b) of this table reports the results from a regression of whether an individual is in forbearance or whether an
individual missed their payment while in forbearance on various borrower and zip code level variables, including an
indicator for whether the mortgage is a jumbo loan. Months include March 2020 to October 2020. Analysis is
restricted to mortgages with balances that fall within 95% to 105% of the conforming loan limit (90% to 110% in
Columns 5 to 8). Jumbo loans are loans that fall above 100% of the conforming loan limit and are not covered by
forbearance mandates under the CARES Act. Conforming loans are Fannie and Freddie mortgages that fall below
100% of the conforming loan limit. Standard errors are clustered at the loan level. Borrower level controls include
student debt balance, auto debt balance, revolving debt balance, age, income, credit score, DTI, number of accounts
past due, and small business owner. Zip code controls include zip code racial composition, educational attainment,
unemployment rate, median age, median income, average CLTV, average DTI, and median house price. Regression
inputs are scaled by standard deviation. Borrower control variables are taken as of January 2020. Sources:
Forbearance status, missed payment status, and all other individual-level characteristics come from Equifax. House
price data comes from Zillow. The remaining variables come from the U.S. Census Bureau American Community
Survey 5-Year Estimates.
Window -/+ 5% Window -/+ 10%
Forbearance Rate % Missing Payments Forbearance Rate % Missing Payments
(1) (2) (3) (4) (5) (6) (7) (8)
Conforming 1.616*** 1.514*** 2.015 2.942 1.446*** 1.412*** 2.134 1.401
(0.166) (0.191) (1.658) (1.614) (0.123) (0.144) (1.094) (1.179)
Mean of Dependent Variable 5.537 5.251 72.58 71.93 5.464 5.220 71.50 72.00
Zip Code Controls No Yes No Yes No Yes No Yes
Zip Code FE Yes No Yes No Yes No Yes No
Observations 493,227 301,936 26,768 15,856 837,247 512,103 45,751 45,110
Adjusted R-squared
0.107 0.035 0.228 0.017 0.087 0.032 0.013 0.183
46
Table 6: Government Mandates and Mortgage Forbearance [continued]
Panel C: Diff-and-diff Analysis
Panel (c) of this table shows results from regressions of the form:
!"#$%&#&'(%),+,, = . + 01 2"'3"#45'6),, + 08 9&':%45(+ + 0; 9&':%45(+ ∗ 2"'3"#45'6),, + =>)., + @ A, + B),+,,
where !"#$%&#&'(%),+,, is an indicator for whether loan i located in zip code z is in forbearance during month t.
2"'3"#45'6),, is an indicator for whether the loan is a conforming mortgage, 9&':%45(+ is an indicator that takes
on the value of zero prior to March 2020 and one from March 2020 to October 2020, >),, is a vector of borrower level
characteristics, and A, is either a vector of zip code controls or fixed effects. We restrict our analysis to mortgages
with balances that fall within 90% to 110% of the conforming loan limit. Jumbo loans are loans that fall above 100%
of the conforming loan limit and are not covered by forbearance mandates under the CARES Act. Conforming loans
are Fannie and Freddie mortgages that fall below 100% of the conforming loan limit. Standard errors are clustered at
the loan level. Borrower level controls include student debt balance, auto debt balance, revolving debt balance, age,
income, credit score, DTI, number of accounts past due, and small business owner. Zip code controls include zip code
racial composition, educational attainment, unemployment rate, median age, median income, average CLTV, average
DTI, and median house price. Regression inputs are scaled by standard deviation. Borrower control variables are taken
as of January 2020. Months include January 2019 to October 2020. Sources: Forbearance status, missed payment
status, and all other individual-level characteristics come from Equifax. House price data comes from Zillow. The
remaining variables come from the U.S. Census Bureau American Community Survey 5-Year Estimates.
Forbearance Rate % Missing Payments
(1) (2) (3) (4)
Conforming * Pandemic 1.148*** 1.312*** 13.54*** 16.03**
(0.116) (0.143) (3.665) (5.060)
Pandemic 4.349*** 3.991*** 6.201* 14.26***
(0.0985) (0.120) (2.964) (4.219)
Conforming 0.0760* -0.0354 -11.72** -14.75**
1.148*** 1.312*** 13.54*** 16.03**
Borrower Controls Yes Yes Yes Yes
Zip Code Controls No Yes No Yes
Zip Code FE Yes No Yes No
Observations 2,280,427 1,394,410 47,967 28,442
Adjusted R-squared 0.070 0.044 0.195 0.030
47
Table 7: Forbearance Rates: Bank versus Shadow Bank Loans
This table presents loan-level regression results of forbearance rate by bank and shadow bank. The dependent variable
indicates whether the loan is in forbearance. Pandemic indicates whether the reporting period is after March 2020.
The sample covers all loans sold to Fannie Mae since 2000 and that were still active as of January 2020. Since the
GSE do not disclose identity of servicers that service less than 1% of total unpaid outstanding balance, we remove
loans without identified servicers. The observations are at monthly level from January 2020 to June 2020. Standard
errors are clustered by zip code. ***, **, * represent 1%, 5%, and 10% significance, respectively.
(1) (2) (3)
Pandemic 7.07*** 7.29*** 7.27***
(0.16) (0.16) (0.16)
Shadow Bank 0.03*** -0.56*** -0.63***
(0.00) (0.02) (0.03)
Pandemic × Shadow Bank -1.85*** -1.98*** -1.96***
(0.06) (0.06) (0.06)
Borrower Controls No Yes Yes
Zip Code FE No No Yes
R2 0.03 0.05 0.06
Observations 51,276,016 43,646,272 43,646,272
48
Figure 1: Residential Mortgage Debt: Delinquency and Forbearance Rates
Panel (a) shows the US residential mortgage delinquency rate, while panel (b) shows the corresponding mortgage
forbearance rates around the beginning of the COVID-19 crisis. Panel (c) and (d) shows the corresponding patterns
over the longer-time period (Q1:2006 to Q3: 2020). In addition, panel (c) also displays the unemployment rate during
this period. Delinquency rates are 30 days past due or worse. The dashed vertical line shows the declaration of the
national emergency due to COVID-19 and the passage of the CARES Act in March 2020. Forbearance rates are
calculated according to Equifax standard procedure for identifying loans in forbearance. The quarterly unemployment
rates are form the US Bureau of Labor Statistics and are peak values in a respective quarters. Sources: Equifax 10%
representative sample of the U.S. credit population.
Panel (a): Delinquency rates and the COVID-19 crisis Panel (b): Forbearance rate and the COVID-19 crisis
3.5 7
6
3
5
2.5
4
2 3
2
1.5
1
1
0
Jan-19 Feb-19 Mar-19 Apr-19 May-19 Jun-19 Jul-19 Aug-19 Sep-19 Oct-19 Nov-19 Dec-19 Jan-20 Feb-20 Mar-20 Apr-20 May-20 Jun-20 Jul-20 Aug-20 Sep-20 Oct-20 Jan-19 Feb-19 Mar-19 Apr-19 May-19 Jun-19 Jul-19 Aug-19 Sep-19 Oct-19 Nov-19 Dec-19 Jan-20 Feb-20 Mar-20 Apr-20 May-20 Jun-20 Jul-20 Aug-20 Sep-20 Oct-20
Panel (c): Delinquency and unemployment rate (2006-2020) Panel (d): Forbearance rate (2006-2020)
16 7
14 6
12
5
10
4
8
6 3
4 2
2 1
0
Sep-06 Sep-07 Sep-08 Sep-09 Sep-10 Sep-11 Sep-12 Sep-13 Sep-14 Sep-15 Sep-16 Sep-17 Sep-18 Sep-19 Sep-20
0
Mar-06 Mar-07 Mar-08 Mar-09 Mar-10 Mar-11 Mar-12 Mar-13 Mar-14 Mar-15 Mar-16 Mar-17 Mar-18 Mar-19 Mar-20
Mar-06 Sep-06 Mar-07 Sep-07 Mar-08 Sep-08 Mar-09 Sep-09 Mar-10 Sep-10 Mar-11 Sep-11 Mar-12 Sep-12 Mar-13 Sep-13 Mar-14 Sep-14 Mar-15 Sep-15 Mar-16 Sep-16 Mar-17 Sep-17 Mar-18 Sep-18 Mar-19 Sep-19 Mar-20 Sep-20
Delinquency Rate Unemployment Rate
49
Figure 2: Auto, Revolving, and Student Debt: Delinquency and Forbearance Rates
This figure shows delinquency and forbearance rates of different debt types from January 2019 to October 2020.
Delinquency rates are 30 days past due or worse. The left column plots delinquency rates and the right column shows
plots of forbearance rates. The first row contains results for auto loans, the second row shows revolving debt, and the
final row shows student loans. The dashed vertical line shows the declaration of the national emergency and the
passage of the CARES Act in March 2020. Sources: Equifax 10% representative sample of the U.S. credit population.
Panel (a): Auto Debt Delinquency Panel (b): Auto Debt Forbearance
4.3 4
4.1
3.5
3.9
3.7 3
3.5
2.5
3.3
3.1 2
2.9
1.5
2.7
2.5 1
Jan-19 Feb-19 Mar-19 Apr-19 May-19 Jun-19 Jul-19 Aug-19 Sep-19 Oct-19 Nov-19 Dec-19 Jan-20 Feb-20 Mar-20 Apr-20 May-20 Jun-20 Jul-20 Aug-20 Sep-20 Oct-20 Jan-19 Feb-19 Mar-19 Apr-19 May-19 Jun-19 Jul-19 Aug-19 Sep-19 Oct-19 Nov-19 Dec-19 Jan-20 Feb-20 Mar-20 Apr-20 May-20 Jun-20 Jul-20 Aug-20 Sep-20 Oct-20
Panel (c): Revolving Debt Delinquency Panel (d): Revolving Debt Forbearance
3 3.3
3.1
2.5
2.9
2 2.7
2.5
1.5
2.3
1 2.1
1.9
0.5
1.7
0 1.5
Jan-19 Feb-19 Mar-19 Apr-19 May-19 Jun-19 Jul-19 Aug-19 Sep-19 Oct-19 Nov-19 Dec-19 Jan-20 Feb-20 Mar-20 Apr-20 May-20 Jun-20 Jul-20 Aug-20 Sep-20 Oct-20 Jan-19 Feb-19 Mar-19 Apr-19 May-19 Jun-19 Jul-19 Aug-19 Sep-19 Oct-19 Nov-19 Dec-19 Jan-20 Feb-20 Mar-20 Apr-20 May-20 Jun-20 Jul-20 Aug-20 Sep-20 Oct-20
Panel (e): Student Debt Delinquency Panel (f): Student Debt Forbearance
7 95
90
6
85
5 80
75
4
70
3
65
2 60
55
1
50
0 45
Jan-19 Feb-19 Mar-19 Apr-19 May-19 Jun-19 Jul-19 Aug-19 Sep-19 Oct-19 Nov-19 Dec-19 Jan-20 Feb-20 Mar-20 Apr-20 May-20 Jun-20 Jul-20 Aug-20 Sep-20 Oct-20 Jan-19 Feb-19 Mar-19 Apr-19 May-19 Jun-19 Jul-19 Aug-19 Sep-19 Oct-19 Nov-19 Dec-19 Jan-20 Feb-20 Mar-20 Apr-20 May-20 Jun-20 Jul-20 Aug-20 Sep-20 Oct-20
50
Figure 3: Forbearance and Missed Payment Rates by Income, Credit Scores, and Age
This figure plots first mortgage forbearance rates and missed payment rates by income, credit score, and age. Panel
(a) shows first mortgages, panel (b) shows auto loans, panel (c) shows revolving loans, and panel (d) shows student
loans. The left column plot forbearance rates by income quantiles. The center column plots similar results by credit
scores and the right column by age groups. Income, credit scores, and age are taken as of January 2020. Months
include March 2020 to October 2020. Sources: Equifax 10% representative sample of the U.S. credit population.
Panel (a): Mortgages
Income Credit Score Age
7% 16% 7%
6% 14% 6%
12%
5% 5%
10%
4% 4%
8%
3% 3%
6%
2% 2%
4%
1% 2% 1%
0% 0% 0%
<42,000 42,000-52,000 52,000-62,000 ≥62,000 Low Fair Good Excellent <41 41-52 52-62 62+
Panel (b): Auto Debt
Income Credit Score Age
6% 6% 6%
5% 5% 5%
4% 4% 4%
3% 3% 3%
2% 2% 2%
1% 1% 1%
0% 0% 0%
<32,000 32,000-43,000 43,000-60,000 >60,000 Low Fair Good Excellent <32 32-43 43-60 60+
Panel (c): Revolving Debt
Income Credit Score Age
7% 7% 7%
6% 6% 6%
5% 5% 5%
4% 4% 4%
3% 3% 3%
2% 2% 2%
1% 1% 1%
0% 0% 0%
<30,000 30,000-40,000 40,000-51,000 ≥51,000 Low Fair Good Excellent <40 40-52 52-65 65+
Panel (d): Student Debt
Income Credit Score Age
100% 100% 100%
90% 90% 90%
80% 80% 80%
70% 70% 70%
60% 60% 60%
50% 50% 50%
40% 40% 40%
30% 30% 30%
20% 20% 20%
10% 10% 10%
0% 0% 0%
<16,000 16,000-19,000 19,000-32,000 ≥32,000 Low Fair Good Excellent <26 26-32 32-42 42+
51
Figure 4: Percent of Scheduled Payments Missed by Income, Credit Scores, and Age
This figure plots the percent of scheduled payments missed due to forbearance by income, credit score, and age. Panel
(a) shows first mortgages, panel (b) shows auto loans, panel (c) shows revolving loans, and panel (d) shows student
loans. The figures show both the averages for each quantile based on income and age, and for each credit score group.
The left column plots by income quantiles, the center column by credit score quantiles, and the right column by age
quantiles. Income, credit score, and age are taken as of January 2020. Months include March 2020 to October 2020.
Sources: Equifax 10% representative sample of the U.S. credit population.
Panel (a): Mortgages
Income Credit Score Age
70% 70% 70%
60% 60% 60%
50% 50% 50%
40% 40% 40%
30% 30% 30%
20% 20% 20%
10% 10% 10%
0% 0% 0%
<42,000 42,000-53,000 53,000-71,000 ≥71,000 Low Fair Good Excellent <41 41-53 53-62 62+
Panel (b): Auto Debt
Income Credit Score Age
80% 80% 80%
70% 70% 70%
60% 60% 60%
50% 50% 50%
40% 40% 40%
30% 30% 30%
20% 20% 20%
10% 10% 10%
0% 0% 0%
<32,000 32,000-43,000 43,000-60,000 ≥60,000 Low Fair Good Excellent <32 32-43 43-60 60+
Panel (c): Revolving Debt
Income Credit Score Age
100% 100% 100%
90% 90% 90%
80% 80% 80%
70% 70% 70%
60% 60% 60%
50% 50% 50%
40% 40% 40%
30% 30% 30%
20% 20% 20%
10% 10% 10%
0% 0% 0%
<30,000 30,000-40,000 40,000-51,000 ≥51,000 Low Fair Good Excellent <40 40-52 50-65 65+
Panel (d): Student Debt
Income Credit Score Age
100% 100% 100%
90% 90% 90%
80% 80% 80%
70% 70% 70%
60% 60% 60%
50% 50% 50%
40% 40% 40%
30% 30% 30%
20% 20% 20%
10% 10% 10%
0% 0% 0%
<16,000 16,000-19,000 19,000-32,000 ≥32,000 Low Fair Good Excellent <26 26-32 32-42 42+
52
Figure 5: Missed Payment Amounts by Income, Credit Scores, and Age
This figure plots the dollar amount of payments missed due to forbearance by income, credit score, and age. Panel (a)
shows first mortgages, panel (b) shows auto loans, panel (c) shows revolving loans, and panel (d) shows student loans.
The figures show both the averages for each quintile based on income and age, and for each credit score group. The
left column plots by income quantiles, the center column by credit score quantiles, and the right column by age
quantiles. Income, credit score, and age are taken as of January 2020. Months include March 2020 to October 2020.
Sources: Equifax 10% representative sample of the U.S. credit population.
Panel (a): Mortgages
Income Credit Score Age
2000 2000 2000
1800 1800 1800
1600 1600 1600
1400 1400 1400
1200 1200 1200
1000 1000 1000
800 800 800
600 600 600
400 400 400
200 200 200
0 0 0
<42,000 42,000-53,000 53,000-71,000 ≥71,000 Low Fair Good Excellent <41 41-53 53-62 62+
Panel (b): Auto Debt
Income Credit Score Age
400 400 400
350 350 350
300 300 300
250 250 250
200 200 200
150 150 150
100 100 100
50 50 50
0 0 0
<32,000 32,000-43,000 43,000-60,000 ≥60,000 Low Fair Good Excellent <32 32-43 43-60 60+
Panel (c): Revolving Debt
Income Credit Score Age
70 70 70
60 60 60
50 50 50
40 40 40
30 30 30
20 20 20
10 10 10
0 0 0
<30,000 30,000-40,000 40,000-51,000 ≥51,000 Low Fair Good Excellent <40 40-52 50-65 65+
Panel (d): Student Debt
Income Credit Score Age
90 90 90
80 80 80
70 70 70
60 60 60
50 50 50
40 40 40
30 30 30
20 20 20
10 10 10
0 0 0
<16,000 16,000-19,000 19,000-32,000 ≥32,000 Low Fair Good Excellent <26 26-32 32-42 42+
53
Figure 6: Geographical Heterogeneity in Forbearance Rates
This figure shows heterogeneity in average monthly forbearance rates across zip codes. For each debt type, we limit
maps to zip codes with at least ten loans. Average monthly forbearance rates are taken from March 2020 to October
2020. Panel (a) shows forbearance rates for first mortgages, panel (b) shows forbearance rates for auto loans, panel
(c) shows forbearance rates for revolving accounts, and panel (d) shows forbearance rates for student loans. We restrict
to zip codes with at least ten loans. Sources: Equifax 10% representative sample of the U.S. credit population.
Panel (a): First Mortgages Panel (b): Auto Debt
Panel (c): Revolving Debt Panel (d): Student Debt
54
Figure 7: Geographical Heterogeneity in COVID Related Characteristics
This figure plots county level characteristics. Panel (a) shows average COVID case rates from March 2020 to October
2020, panel (b) shows the percent of workers in at risk industries, panel (c) shows the average daily change in time
spent at workplaces relative to January 2020, and panel (d) shows the average daily change in credit/debit card
spending relative to January 2020. Averages are taken from March to October 2020. Sources: COVID cases,
credit/debit card spending, and time spent at workplace come from Opportunity Insights Tracker. Industry composition
comes from the Bureau of Labor Statistics.
Panel (a): COVID Case Rates Panel (b): % of at Risk Workers
Panel (c): Change in Time at Workplace Panel (d): Change in Spending
55
Figure 8: Regional Evidence: Heterogeneity in Forbearance Rates based on % Minority
and % Black
This figure plots the average monthly forbearance rates for zip codes. Average monthly forbearance rates are taken
from March 2020 to October 2020. Plots are based on the percent of a zip code’s population that is a minority (non-
white) and the percent of a zip code’s population that is Black. Panel (a) shows first mortgage forbearance rates, panel
(b) shows auto loans, panel (c) shows revolving loans, and panel (d) shows student loans. The left column plots by
income quantile, the middle column by percent minority quantile, and the right column by pre-pandemic
unemployment rate quantile. We restrict to zip codes with at least ten loans. Sources: Forbearance rates come from
Equifax. All other data come from the U.S. Census Bureau’s American Community Survey.
Panel (a): Mortgages
% Minority % Black
6% 6%
5% 5%
4% 4%
3% 3%
2% 2%
1% 1%
0% 0%
<2% 2%-7.5% 7.5%-22% ≥22% <1% 1%-3% 3%-11% ≥11%
Panel (b): Auto Debt
% Minority % Black
4% 4%
3% 3%
3% 3%
2% 2%
2% 2%
1% 1%
1% 1%
0% 0%
<2% 2%-7.5% 7.5%-22% ≥22% <1% 1%-3% 3%-11% ≥11%
Panel (c): Revolving Debt
% Minority % Black
4% 4%
3% 3%
3% 3%
2% 2%
2% 2%
1% 1%
1% 1%
0% 0%
<2% 2%-7.5% 7.5%-22% ≥22% <1% 1%-3% 3%-11% ≥11%
56
Figure 8: Regional Evidence: Heterogeneity in Forbearance Rates based on % Minority,
and % Black – [Continued]
Panel (d): Student Debt
% Minority % Black
90% 90%
80% 80%
70% 70%
60% 60%
50% 50%
40% 40%
30% 30%
20% 20%
10% 10%
0% 0%
<2% 2%-7.5% 7.5%-22% ≥22% <1% 1%-3% 3%-11% ≥11%
57
Figure 9: Regional Evidence: Mortgage Forbearance Rates and Unemployment and
COVID-19 Infection rates
This figure plots zip code level average monthly forbearance rates. Average monthly forbearance rates are taken from
March 2020 to October 2020. Plots are based on COVID case rates, unemployment insurance claims, and the percent
of workers employed in “at risk” industries. We restrict to zip codes with at least ten loans. The left column plots by
COVID case rate quantile, the middle column by total UI claims quantile, and the right column by the percent of
workers in at risk industries quantile. COVID case rates are average case rates taken over the period from March to
October 2020. Sources: Forbearance and missed payments come from Equifax 10% representative sample of the U.S.
credit population. All other data come from Opportunity Insights.
Panel (a): Mortgages
COVID Cases Total UI Claims % Workers in At Risk Industries
4.5 4.5 4.5
4 4 4
3.5 3.5 3.5
3 3 3
2.5 2.5 2.5
2 2 2
1.5 1.5 1.5
1 1 1
0.5 0.5 0.5
0 0 0
<390 390-660 660-1080 ≥1080 <0.85% 0.85-1.2% 1-2-1.8% ≥1.8% <12% 12-15% 15-18% ≥18%
Panel (b): Auto Debt
COVID Cases Total UI Claims % Workers in At Risk Industries
3 3 3
2.5 2.5 2.5
2 2 2
1.5 1.5 1.5
1 1 1
0.5 0.5 0.5
0 0 0
<390 390-660 660-1080 ≥1080 <0.85% 0.85-1.2% 1-2-1.8% ≥1.8% <12% 12-15% 15-18% ≥18%
Panel (c): Revolving Debt
COVID Cases Total UI Claims % Workers in At Risk Industries
3 3 3
2.5 2.5 2.5
2 2 2
1.5 1.5 1.5
1 1 1
0.5 0.5 0.5
0 0 0
<390 390-660 660-1080 ≥1080 <0.85% 0.85-1.2% 1-2-1.8% ≥1.8% <12% 12-15% 15-18% ≥18%
Panel (d): Student Debt
COVID Cases Total UI Claims % Workers in At Risk Industries
90 90 90
80 80 80
70 70 70
60 60 60
50 50 50
40 40 40
30 30 30
20 20 20
10 10 10
0 0 0
<390 390-660 660-1080 ≥1080 <0.85% 0.85-1.2% 1-2-1.8% ≥1.8% <12% 12-15% 15-18% ≥18%
58
Figure 10: Debt Forbearance and other stimulus measures
This figure plots the total amount of money missed due to forbearance in a state normalized by the number of adults
against the total amount of money received from stimulus checks, PPP loans, and unemployment insurance normalized
by the number of adults at the state level. Panel (a) plots the amount missed through forbearance against the amount
received from stimulus checks, panel (b) plots the amount missed through forbearance against the amount received
from PPP loans, and panel (c) plots the amount missed through forbearance against the amount received from
unemployment benefits. Amounts are calculated from the months from March to October 2020. Sources: Missed
payments come from Equifax, stimulus amounts come from the Internal Revenue Service, unemployment benefits
come from the Department of Labor, PPP loans come from the Small Business Administration.
Panel (a): Stimulus Money Received per Adult Panel (b): Unemployment Benefits per Adult
Panel (c): PPP Loans per Adult
59
Figure 11: Government Mandates:
Mortgage Forbearance rates around the Conforming Loan Limit
This figure plots mortgage forbearance rates around the conforming loan limit. We restrict our analysis to mortgages
with balances that fall within 70% to 130% of the conforming loan limit. Panel (a) shows results during the COVID-
19 period (March 2020 to October 2020). Dots represent forbearance rates in 1% buckets by the loan’s percentage of
the conforming loan limit. Third degree polynomials are fit on either side of the conforming loan limit. Shaded bands
show the 95% confidence interval. Panel (b) shows the same plot during the pre-pandemic period (January 2020 to
February 2020). Jumbo loans are loans that fall above 100% of the conforming loan limit (solid line) and are not
covered by forbearance mandates under the CARES Act. Conforming loans are Fannie and Freddie loans that fall
below 100% of the conforming loan limit. Sources: Equifax 10% representative sample of the U.S. credit population.
6 6
5 5
Forbearance Rate Forbearance Rate
2 3 4 2 3 4
1 1
0 0
.7 .8 .9 1 1.1 1.2 1.3 .7 .8 .9 1 1.1 1.2 1.3
Panel (a): COVID Period Panel (b): Pre-COVID Period
60
Figure 11: Government Mandates:
Mortgage Forbearance rates around the Conforming Loan Limit
[Continued]
Panel (c) of this figure plots average monthly forbearance rates for jumbo and conforming loans from January 2019
to October 2020, while panel (d) shows the estimated differential changes in forbearance rates on conforming loans
relative to jumbo loans along with 95% confidence intervals. Jumbo loans are loans that fall above 100% of the
conforming loan limit and are not covered by forbearance mandates under the CARES Act. Conforming loans are
Fannie and Freddie loans that fall below 100% of the conforming loan limit. The dashed vertical line shows the
declaration of the national emergency due to COVID-19 and the passage of the CARES Act in March 2020. Sources:
Equifax 10% representative sample of the U.S. credit population.
10%
9%
8%
7%
6%
5%
4%
3%
2%
1%
0%
1/ 1/ 3/ 1/ 5/ 1/ 7/ 1/ 9/ 1/ 11 1/ 1/ 3/ 1/ 5/ 1/ 7/ 1/ 9/ 1/
19 19 19 19 19 /1/1 20 20 20 20 20
9
Jumbo Conventional
Panel (c): Forbearance rates of conforming and jumbo loans
2.5
2
1.5
1
0.5
0
1/
2/
3/
4/
5/1/
1/
1/
1/
1/19
19
19
19
19
6/
7/
8/
9/1/
1/
1/
1/
10 9
/1 19
19
19
1
/1
11 9
/1
12 9
/1
1/
2/
3/ /1
/1
1/
1/
1/ 9
20
20
20
4/
5/
6/
7/
8/1/
1/
1/
1/
1/20
20
20
20
20
9/1/
10 0
/1 2
/20
-0.5
Panel (d): Change in forbearance rates of conforming loans relative to jumbo loans
61
Figure 12: Mortgage Mandates and Forbearance Rates across Borrowers
This figure shows forbearance rates for conforming and jumbo rates, across age, income, and credit scores. Months include
March 2020 to October 2020. We restrict our analysis to mortgages with balances that fall within 90% to 110% of the
conforming loan limit. Jumbo loans are loans that fall above 100% of the conforming loan limit and are not covered by
forbearance mandates under the CARES Act. Conforming loans are Fannie and Freddie loans that fall below 100% of the
conforming loan limit. Panel (a) shows results for jumbo mortgages while panel (b) shows results for conforming mortgages.
The left column plot forbearance rates by income quintiles. The center column plots similar results by credit scores and the
right column by age groups. Income, age, and credit score are taken as of January 2020. Sources: Equifax 10% representative
sample of the U.S. credit population.
Panel (a): Jumbo Mortgages
Income Credit Score Age
12% 12%
20%
10% 10%
8% 15% 8%
6% 6%
10%
4% 4%
5%
2% 2%
0% 0% 0%
<76,000 76,000-93,000 93,000-111,000 ≥111,000 Low Fair Good Excellent <41 41-50 50-59 59+
Panel (b): Conforming Mortgages
Income Credit Score Age
12% 12%
20%
10% 10%
8% 15% 8%
6% 6%
10%
4% 4%
5%
2% 2%
0% 0% 0%
<76,000 76,000-93,000 93,000-111,000 ≥111,000 Low Fair Good Excellent <41 41-50 50-59 59+
62
Figure 13: Forbearance Rates and Intermediary Factors
Panel (a) of this figure plots the forbearance rate of loans sold to Fannie Mae splitting the sample by banks and
shadow bank originators. The banks and shadow banks are defined according to classification in Buchak et al.
(2018). Panel (b) plots forbearance rate of loans sold to Fannie Mae splitting the sample based on whether a loan
is serviced by the originator or by other servicers. In-House is defined as the loan being serviced by its originator.
Outsource is if the loan is serviced by an institution that was not its originator. The sample covers all loans sold to
Fannie Mae since 2000 and that were still active as of January 2020. The dashed vertical line shows the declaration
of the national emergency and the passage of the CARES Act in March 2020. Sources: Fannie Mae Loan Level
Performance Data.
Panel (a): Forbearance Rates: Banks vs Shadow Banks
Panel (b): Forbearance Rates: In-House vs Outsourced Servicing
63
Appendix
64
Table A1: Mortgage Borrowers: Descriptive Statistics
This table shows summary statistics for all mortgage borrowers in the Equifax dataset (Columns (1) through (4)) and all borrowers who are in forbearance on their
mortgages (Columns (5) through (8)). Pre-COVID-19 statistics are based on January and February 2020, while COVID-19 statistics are based on March to October
2020. Odd columns contain means while even columns contain standard deviations. Sources: Equifax 10% representative sample of the U.S. credit population.
Mortgage Borrowers Mortgage Borrowers in Forbearance
Pre-COVID COVID Pre-COVID COVID
(1) (2) (3) (4) (5) (6) (7) (8)
Mean SD Mean SD Mean SD Mean SD
Vantage 732 95 738 93 656 133 679 100
Estimated Age 50 14 50 14 49 14 47 13
Estimated Income 53,867 20,162 54,03 20,283 51,131 20,799 51,164 20,369
Fraction with First Mortgage 1.00 1.00 1.00 1.00
Fraction with Auto Debt 0.50 0.50 0.49 0.56
Fraction with Revolving Debt 0.91 0.91 0.82 0.89
Fraction Student Debt 0.14 0.14 0.16 0.19
Average First Mortgage Balance (Nonzero Accounts) 210,020 240,631 214,788 246,309 292,910 401,836 274,105 310,893
Average Combined Mortgage Balance (Nonzero Accounts) 211,549 242,103 216,254 247,719 294,797 402,919 276,147 312,695
Average Auto Debt (Nonzero Accounts) 215,97 20,640 21,824 20,942 221,60 21,836 24,225 23,219
Average Revolving Debt (Nonzero Accounts) 7,717 11,862 6,878 11,220 7,471 13,390 10,216 14,869
Average Student Debt (Nonzero Accounts) 40,589 56,288 41,537 57,339 46,536 61,293 49,192 64,746
Average Credit Card Utilization 28% 25% 42% 42%
Average Available Credit 24,955 27,555 25,506 27,887 19,984 36,292 19,666 27389
Number of Consumers 4,549,218 4,796,029 41,796 502,942
65
Table A2: Individual-Level Forbearance Regressions
This table reports the results from a regression of whether an individual is in forbearance or whether an individual
missed their payment while in forbearance on various borrower and zip code level variables. Months include March
2020 to October 2020. Columns (1) and (2) are estimated from all individuals with first mortgages, while Columns
(3) and (4) are restricted to individuals who are in forbearance on their first mortgages. Columns (1) and (3) contain
individual-level characteristics and zip code fixed effects, while columns (2) and (4) contain individual level controls
and zip code controls. Zip code controls include zip code racial composition, educational attainment, unemployment
rate, median age, median income, average CLTV, average DTI, and median house price. Regression inputs are scaled
by standard deviation. Individual controls are taken as of January 2020. Sources: Forbearance status, missed payment
status, and all other individual-level characteristics come from Equifax. House price data comes from Zillow. The
remaining variables come from the U.S. Census Bureau American Community Survey 5-Year Estimates.
Panel A: Auto Debt
(1) (2) (3) (4)
Missed Missed
In Forbearance In Forbearance
Payment Payment
Mortgage Balance -0.0164** -0.0151* 0.206* 0.263**
(0.00501) (0.00679) (0.0966) (0.0860)
Auto Loan Balance 0.771*** 0.746*** -0.185*** -0.285***
(0.0109) (0.00859) (0.0472) (0.0471)
Revolving Debt Balance -0.0417*** -0.0589*** -0.210*** -0.209***
(0.00503) (0.00657) (0.0574) (0.0587)
Student Loan Balance -0.0577*** -0.0523*** -0.238*** -0.286***
(0.00405) (0.00523) (0.0438) (0.0570)
Small Business Owner 0.0817*** 0.0853*** 0.141** 0.102
(0.00424) (0.00533) (0.0432) (0.0568)
Credit Score -1.258*** -1.252*** -3.021*** -3.126***
(0.00780) (0.0102) (0.0474) (0.0591)
Income 0.00540 0.0119 -0.298*** -0.460***
(0.00580) (0.00703) (0.0887) (0.0941)
# Accounts Past Due in Jan. 0.222*** 0.238*** 0.255*** 0.299***
(0.00781) (0.0107) (0.0310) (0.0402)
DTI -0.0135 -0.0171 -0.314 -0.179
(0.00707) (0.0122) (0.260) (0.170)
Age 0.109*** 0.117*** -0.349*** -0.318***
(0.00465) (0.00588) (0.0523) (0.0691)
Zip Code Controls No Yes No No
Zip Code FE Yes No Yes No
Mean of Dependent Variable 3.052 2.969 76.84 76.96
Observations 54,690,391 32,320,318 1,666,572 959,712
Adjusted R-squared 0.011 0.009 0.039 0.012
66
Table A2: Individual-Level Forbearance Regressions - [Continued]
Panel B: Revolving Debt
(1) (2) (3) (4)
Missed Missed
In Forbearance In Forbearance
Payment Payment
Mortgage Balance 0.209*** 0.213*** 0.0340* 0.00326
(0.0108) (0.0135) (0.0141) (0.0180)
Auto Loan Balance 0.613*** 0.622*** -0.767*** -1.044***
(0.00909) (0.00950) (0.0214) (0.0314)
Revolving Debt Balance 3.379*** 3.316*** -1.169*** -1.127***
(0.0163) (0.0241) (0.0237) (0.0362)
Student Loan Balance -0.105*** -0.108*** -0.322*** -0.316***
(0.00558) (0.00717) (0.0200) (0.0266)
Small Business Owner 0.109*** 0.112*** 0.601*** 0.693***
(0.00507) (0.00638) (0.0194) (0.0257)
Credit Score -2.316*** -2.309*** -3.084*** -3.231***
(0.00965) (0.0129) (0.0197) (0.0264)
Income 0.218*** 0.177*** 1.482*** 1.378***
(0.00799) (0.0102) (0.0246) (0.0326)
# Accounts Past Due in Jan. 1.794*** 1.848*** 1.800*** 1.905***
(0.0177) (0.0237) (0.0253) (0.0353)
DTI 0.0888*** 0.0772* 0.00826 -0.0245
(0.0230) (0.0302) (0.0202) (0.0298)
Age 0.122*** 0.142*** -0.789*** -0.558***
(0.00536) (0.00677) (0.0254) (0.0340)
Zip Code Controls No Yes No No
Zip Code FE Yes No Yes No
Mean of Dependent Variable 8.097 7.855 80.35 80.56
Observations 136,744,753 81,031,024 11,070,465 6,365,101
Adjusted R-squared 0.045 0.040 0.092 0.043
67
Table A2: Individual-Level Forbearance Regressions - [Continued]
Panel C: Student Debt
(1) (2) (3) (4)
Missed Missed
In Forbearance In Forbearance
Payment Payment
Mortgage Balance -1.335*** -1.345*** 0.0539 0.0870
(0.197) (0.234) (0.0402) (0.0501)
Auto Loan Balance 0.0678 0.0550 0.0123 -0.000456
(0.107) (0.129) (0.0220) (0.0279)
Revolving Debt Balance -1.086*** -1.139*** -0.0943*** -0.0815**
(0.103) (0.124) (0.0214) (0.0270)
Student Loan Balance 4.580*** 4.624*** 1.283*** 1.309***
(0.0481) (0.0584) (0.0109) (0.0139)
Small Business Owner 0.744*** 0.761*** 0.166*** 0.170***
(0.0180) (0.0236) (0.0107) (0.0142)
Credit Score -6.745*** -7.071*** -3.512*** -3.621***
(0.0512) (0.0648) (0.0133) (0.0173)
Income -4.391*** -4.266*** -1.185*** -1.182***
(0.152) (0.182) (0.0326) (0.0412)
# Accounts Past Due in Jan. -2.359*** -2.416*** -0.337*** -0.350***
(0.0429) (0.0554) (0.00856) (0.0109)
DTI -1.340 -1.215 -0.209 -0.199
(0.732) (0.878) (0.124) (0.154)
Age -6.560*** -6.725*** -0.696*** -0.730***
(0.0480) (0.0588) (0.0157) (0.0205)
Zip Code Controls No Yes No No
Zip Code FE Yes No Yes No
Mean of Dependent Variable 78.23 77.81 94.94 94.81
Observations 27,911,213 16,451,526 21,835,306 12,801,269
Adjusted R-squared 0.123 0.120 0.038 0.034
68
Table A3: Zip-Code Socio-economic Characteristics Regressions
This table shows the results from regressions of average monthly forbearance rate and the percent of scheduled
payments missed in each zip code on various zip code-level characteristics. Averages are taken from March 2020 to
October 2020. Regression inputs are scaled by standard deviation. Analysis is restricted to zip codes with at least ten
loans. Sources: Forbearance status, missed payment status, and all other individual-level characteristics come from
Equifax. House price data comes from Zillow. The remaining variables come from the U.S. Census Bureau American
Community Survey 5-Year Estimates.
Panel A: Auto Debt
Forbearance Percent of Scheduled
Rate Payments Missed
(1) (2)
% College Educated -0.0818** 0.0450
(0.0292) (0.0329)
% Black 0.595*** 0.360***
(0.0289) (0.0346)
% Hispanic 0.743*** 0.558***
(0.0280) (0.0298)
Pre-Pandemic Unemployment Rate 0.177*** 0.159***
(0.0322) (0.0337)
Median Age 0.139*** 0.195***
(0.0250) (0.0310)
Median Income 0.178*** -0.138
(0.0418) (0.0726)
Average DTI 0.293*** 0.947***
(0.0859) (0.212)
Average House Price in January 0.137*** 0.178***
(0.0318) (0.0348)
Mean of Dependent Variable 4.013 2.697
# Of Obs. 16,920 16,920
R-Squared 0.1413 0.1187
69
Table A3: Zip-Code Socio-economic Characteristics Regressions – [Continued]
Panel B: Revolving Debt
Forbearance Percent of Scheduled
Rate Payments Missed
(1) (2)
% College Educated 0.200*** 0.0297*
(0.0120) (0.0150)
% Black 0.333*** 0.168***
(0.00954) (0.0117)
% Hispanic 0.1000*** 0.206***
(0.00787) (0.0118)
Pre-Pandemic Unemployment Rate 0.0284** 0.0553**
(0.0105) (0.0171)
Median Age -0.0465*** -0.00352
(0.00944) (0.0134)
Median Income -2.316*** -2.309***
(0.00965) (0.0129)
Average DTI 0.218*** 0.177***
(0.00799) (0.0102)
Average House Price in January 0.165*** 0.0992***
(0.0118) (0.0129)
Mean of Dependent Variable 2.322 2.065
# Of Obs. 18,616 18,616
R-Squared 0.2068 0.0676
70
Table A3: Zip-Code Socio-economic Characteristics Regressions – [Continued]
Panel C: Student Debt
Forbearance Percent of Scheduled
Rate Payments Missed
(1) (2)
% College Educated -0.765*** -1.274***
(0.0871) (0.150)
% Black 1.741*** 3.185***
(0.0562) (0.116)
% Hispanic 0.764*** 1.482***
(0.0579) (0.114)
Pre-Pandemic Unemployment Rate 0.258** 0.356*
(0.0857) (0.156)
Median Age -0.512*** -0.655***
(0.0761) (0.131)
Median Income -1.187*** -2.927***
(0.107) (0.178)
Average DTI -0.196* 0.311
(0.0783) (0.161)
Average House Price in January 0.0892 1.017***
(0.0777) (0.123)
Mean of Dependent Variable 79.45 50.39
# Of Obs. 17,691 17,667
R-Squared 0.1614 0.1606
71
Table A4: Zip-Code Industry and Occupation Characteristic Regressions
This table shows the results from regressions of average monthly forbearance rates and the percent of scheduled
payments missed in each zip code on various zip code-level characteristics. Averages are taken from March 2020 to
October 2020. Regression inputs are scaled by standard deviation. Analysis is restricted to zip codes with at least ten
loans. Sources: Forbearance status, missed payment status, and all other individual-level characteristics come from
Equifax. The remaining variables come from the U.S. Census Bureau American Community Survey 5-Year Estimates.
Panel A: Auto Debt
Percent of
Forbearance Scheduled
Rate Payments Missed
(1) (2)
% Production Industry 0.0527 -0.0445
(0.0327) (0.0358)
% Agriculture Industry -0.326*** -0.302***
(0.0315) (0.0318)
% Finance, Insurance, Real Estate Industry -0.0741* -0.0576
(0.0295) (0.0374)
% Arts, Recreation, Entertainment Industry -0.0182** -0.00583
(0.00659) (0.00721)
% Education, Health, Public Administration
Industry -0.517*** -0.423***
(0.0308) (0.0347)
% Construction and Manufacturing Industry -0.721*** -0.572***
(0.0349) (0.0393)
% Work From Home - Pre-Pandemic -0.158*** -0.141***
(0.0314) (0.0357)
% Self -Employed - Pre-Pandemic 0.0242 0.0341
(0.0333) (0.0362)
% Service Occupations 0.366*** 0.271***
(0.0311) (0.0337)
% Sales and Office Occupations 0.129*** 0.0855**
(0.0266) (0.0267)
Mean of Dependent Variable 4.182 2.841
# Of Obs. 21,704 21,704
R-Squared 0.0830 0.0497
72
Table A4: Zip-Code Industry and Occupation Characteristic Regressions - [continued]
Panel B: Revolving Debt
Percent of
Forbearance Scheduled
Rate Payments Missed
(1) (2)
% Production Industry 0.0449** 0.0749***
(0.0138) (0.0181)
% Agriculture Industry -0.183*** -0.124***
(0.0101) (0.0159)
% Finance, Insurance, Real Estate Industry 0.0495*** -0.0394*
(0.0133) (0.0153)
% Arts, Recreation, Entertainment Industry 0.00695** 0.00156
(0.00257) (0.00334)
% Education, Health, Public Administration
Industry -0.0358* -0.110***
(0.0148) (0.0156)
% Construction and Manufacturing Industry -0.294*** -0.181***
(0.0143) (0.0185)
% Work From Home - Pre-Pandemic -0.00126 -0.0563**
(0.0138) (0.0188)
% Self -Employed - Pre-Pandemic -0.0325** 0.0244
(0.0124) (0.0263)
% Service Occupations 0.00455 0.0955***
(0.0127) (0.0153)
% Sales and Office Occupations -0.0517*** 0.0288*
(0.0123) (0.0137)
Mean of Dependent Variable 2.392 2.120
# Of Obs. 23,874 23,874
R-Squared 0.1112 0.0366
73
Table A4: Zip-Code Industry and Occupation Characteristic Regressions - [continued]
Panel C: Student Debt
Percent of
Forbearance Scheduled
Rate Payments Missed
(1) (2)
% Production Industry 1.461*** 2.335***
(0.0990) (0.180)
% Agriculture Industry 0.104 0.301
(0.0944) (0.157)
% Finance, Insurance, Real Estate Industry -0.465*** -0.841***
(0.103) (0.172)
% Arts, Recreation, Entertainment Industry -0.0108 0.000968
(0.0181) (0.0324)
% Education, Health, Public Administration
Industry -0.0912 -0.503**
(0.0879) (0.157)
% Construction and Manufacturing Industry -0.908*** -1.688***
(0.111) (0.193)
% Work From Home - Pre-Pandemic -0.719*** -1.045***
(0.110) (0.160)
% Self -Employed - Pre-Pandemic -0.316** -0.476**
(0.0981) (0.160)
% Service Occupations 1.201*** 2.307***
(0.0864) (0.159)
% Sales and Office Occupations 0.609*** 0.871***
(0.0766) (0.143)
Mean of Dependent Variable 79.75 50.81
# Of Obs. 22,679 22,648
R-Squared 0.0960 0.0829
74
Table A4: Forbearance Rates, Unemployment Claims, and COVID Infections
This table shows the results from regressions of average monthly forbearance rates and the percent of scheduled
payments missed in each county on various county-level characteristics. Averages are taken over the period from
March 2020 to October 2020 and percent of scheduled payments missed by individuals in forbearance is calculated
over the same time period. Regression inputs are scaled by standard deviation. Analysis is restricted to counties with
at least ten loans. Sources: Forbearance status and missed payment status come from Equifax. UI claims, COVID
cases, credit/debit card spending, and time spent at workplace are averages of March 2020 to October 2020 values
from Opportunity Insights Tracker. Remaining variables are calculated using data from the U.S. Census Bureau
American Community Survey 5-Year Estimates.
Panel A: Auto Debt
Forbearance Percent of Scheduled
Rate Payments Missed
(1) (2)
Total UI Claims 0.156*** 0.0951***
(0.0324) (0.0220)
Total COVID Case Rate 0.181*** 0.136***
(0.0333) (0.0226)
Change in Credit/Debit Spending -0.0708* -0.0222
(0.0347) (0.0235)
Change in Time Spent at Workplace -0.141** -0.112**
(0.0518) (0.0351)
% Families Receiving Full Stimulus Check 0.0278 -0.0124
(0.0456) (0.0309)
% Workers in at Risk Industry 0.123 0.118*
(0.0749) (0.0508)
Amount of PPP Received Normalized by # of Small
Businesses 0.0386 -0.0507
(0.0828) (0.0561)
Mean of Dependent Variable 2.362 1.493
# Of Obs. 348 348
R-Squared 0.2076 0.2508
75
Table A4: Forbearance Rates, Unemployment Claims, and COVID Infections –
[Continued]
Panel B: Revolving Debt
Percent of Scheduled
Forbearance Rate Payments Missed
(1) (2)
Total UI Claims 0.190*** 0.185***
(0.0363) (0.0181)
Total COVID Case Rate 0.145*** 0.0842***
(0.0373) (0.0187)
Change in Credit/Debit Spending -0.0579 0.0170
(0.0388) (0.0194)
Change in Time Spent at Workplace -0.254*** -0.109***
(0.0580) (0.0290)
% Families Receiving Full Stimulus Check 0.0771 0.0372
(0.0510) (0.0255)
% Workers in at Risk Industry 0.220** 0.119**
(0.0838) (0.0419)
Amount of PPP Received Normalized by # of Small
Businesses -0.238* -0.231***
(0.0927) (0.0463)
Mean of Dependent Variable 2.421 2.154
# Of Obs. 348 348
R-Squared 0.2316 0.3674
76
Table A4: Forbearance Rates, Unemployment Claims, and COVID Infections –
[Continued]
Panel C: Student Debt
Percent of Scheduled
Forbearance Rate Payments Missed
(1) (2)
Total UI Claims 1.625*** 3.524***
(0.142) (0.274)
Total COVID Case Rate 1.078*** 1.618***
(0.146) (0.281)
Change in Credit/Debit Spending -0.0235 0.0702
(0.152) (0.293)
Change in Time Spent at Workplace -0.113 -1.310**
(0.227) (0.437)
% Families Receiving Full Stimulus Check 0.758*** 1.543***
(0.200) (0.385)
% Workers in at Risk Industry 1.135** 1.902**
(0.366) (0.703)
Amount of PPP Received Normalized by # of Small
Businesses 0.297 -0.0617
(0.339) (0.651)
Mean of Dependent Variable 79.84 50.44
# Of Obs. 348 348
R-Squared 0.4624 0.4458
77
Table A5: Government Mandates and Forbearance:
Unemployment Claims, COVID-19 Infections, and Other Stimulus
This table shows the results from regressions of the average monthly mortgage forbearance rate and the percent of
scheduled payments missed in each zip code on various county-level characteristics for both jumbo mortgages and
conforming mortgages. Averages are taken over the period from March 2020 to October 2020 and percent of scheduled
payments missed by individuals in forbearance is calculated over the same time period. The first two columns show
results based off of jumbo loans and the final two columns show results for conforming loans. Our analysis is restricted
to loans that fall within 90% to 110% of the conforming loan limit. Jumbo loans are loans that fall above 100% of the
conforming loan limit and are not covered by forbearance mandates under the CARES Act (are not Fannie, Freddie,
VA, or FHA loans). Conventional loans are Fannie and Freddie loans that fall below 100% of the conforming loan
limit. Regression inputs are scaled by standard deviation. Analysis is restricted to counties with at least ten loans.
Sources: Forbearance status and missed payment status come from Equifax. UI claims, COVID cases, credit/debit
card spending, and time spent at workplace are averages of March 2020 to October 2020 values from Opportunity
Insights Tracker. Remaining variables are calculated using data from the U.S. Census Bureau American Community
Survey 5-Year Estimates.
Jumbo Loans Conforming Loans
Forbearance Percent of Scheduled Forbearance Percent of Scheduled
Rate Payments Missed Rate Payments Missed
(1) (2) (3) (4)
Total UI Claims 0.294 0.203 0.633** 0.371*
(0.196) (0.152) (0.206) (0.161)
Total COVID Case Rate 0.310 0.154 0.267 0.0758
(0.196) (0.152) (0.207) (0.161)
Change in Credit/Debit Spending 0.577** 0.277 0.291 0.345*
(0.204) (0.158) (0.218) (0.170)
Change in Time Spent at Workplace 0.266 0.215 -0.123 -0.119
(0.321) (0.249) (0.345) (0.269)
% Families Receiving Full Stimulus Check -1.882*** -1.394*** -1.351*** -0.951***
(0.327) (0.254) (0.352) (0.275)
% Workers in at Risk Industry 0.676* 0.487* 0.254 0.192
(0.295) (0.229) (0.320) (0.250)
Amount of PPP Received Normalized by # of
Small Businesses -0.274 0.167 -0.725 -0.544
(0.534) (0.415) (0.567) (0.442)
Mean of Dependent Variable 2.187 1.269 3.507 2.244
# Of Obs. 328 328 345 345
R-Squared 0.1801 0.1511 0.1534 0.1146
78
Figure A1: Consumer Debt Outstanding in Trillions of Dollars
This figure plots the total outstanding debt in the United States from the first quarter 2006 to the third quarter of
2020. Amounts are in trillions of dollars. Debt types include mortgages, auto loans, revolving loans, and student
loans. Source: Equifax data.
14
12
10
8
6
4
2
0
3/1/06 9/1/06 3/1/07 9/1/07 3/1/08 9/1/08 3/1/09 9/1/09 3/1/10 9/1/10 3/1/11 9/1/11 3/1/12 9/1/12 3/1/13 9/1/13 3/1/14 9/1/14 3/1/15 9/1/15 3/1/16 9/1/16 3/1/17 9/1/17 3/1/18 9/1/18 3/1/19 9/1/19 3/1/20 9/1/20
Mortgages Auto Revolving Student
79
Figure A2: Initial Policy Response to COVID-19
This figure shows the number of new daily COVID cases in the United States from January 2020 to July 2020 plotted
against major policy events of the pandemic.
CARES
Act
National
emergency
declared
Fed slashes rates President Trump’s
to near zero, executive order
restart of QE
80
Figure A3: Mortgage Mandates and Missed Payment Percentage across Borrowers
This figure shows the percent of scheduled payments missed for conforming and jumbo mortgages. We restrict our analysis
to mortgages with balances that fall within 90% to 110% of the conforming loan limit. Jumbo loans are loans that fall above
100% of the conforming loan limit and are not covered by forbearance mandates under the CARES Act (are not Fannie,
Freddie, VA, or FHA loans). Conventional loans are Fannie and Freddie loans that fall below 100% of the conforming loan
limit. Months include October 2020 to March 22020. Panel (a) shows results for jumbo mortgages while panel (b) shows the
same for conforming mortgages. The left column plots by income quintiles, the center column plots similar results by credit
scores, and the right column by age groups. Income, credit score, and age are taken as of January 2020. Sources: Equifax
10% representative panel of US credit population.
Panel (a): Jumbo Mortgages - % of Scheduled Payments Missed
Income Credit Score Age
80% 80% 80%
70% 70% 70%
60% 60% 60%
50% 50%
50%
40%
40% 40%
30%
30% 30%
20%
20% 20%
10%
10% 10%
0%
0% 0% <41 41-49 49-58 58+
<76,000 76,000-93,000 93,000-111,000 ≥111,000 Low Fair Good Excellent
Panel (b): Conforming Mortgages - % of Scheduled Payments Missed
Income Credit Score Age
80% 80% 80%
70% 70% 70%
60% 60% 60%
50% 50% 50%
40%
40% 40%
30%
30% 30%
20%
20% 20%
10%
10% 10%
0%
0% 0% <41 41-49 49-58 58+
<76,000 76,000-93,000 93,000-111,000 ≥111,000 Low Fair Good Excellent
81
File and source
- File
- nber-w28357.pdf
- Size
- 2,186,794 bytes
- SHA-256
- 5b7f2dc0f472510383142d64e860a86a2d93721d140c099262e5364cc24bfa9f
- Our copy
- nber-w28357.pdf
- Original
- www.nber.org