Relaxed Credit Standards in the U.S. Housing Boom: Changes in Risk Characteristics of Mortgage Recipients*

• Published date: 01 March 2021
• Author: Seda Durguner
• Date: 01 March 2021
• DOI: http://doi.org/10.1111/irfi.12262

Relaxed Credit Standards in the

U.S. Housing Boom: Changes in Risk

Characteristics of Mortgage

Recipients*

SEDA DURGUNER

Clinical of Finance and Business Economics, University of Southern California,

Los Angeles, CA

ABSTRACT

This paper examines components of credit risk and how their ability to predict

the interface between households and mortgage market changed under the

relaxed lending standards prevailing during the U.S. housing boom of the 2000s.

Using data from the Federal Reserve Board’s 1998 and 2007 Survey of Consumer

Finances, the paper evaluates changes between 1998 and 2007 in the signifi-

cance of credit risk characteristics in explainingthree variables regarding the pur-

chase of a primary residence by households in low-, moderate-, and high-income

groups: the loan-to-value ratio of the purchase mortgage, the likelihood of pur-

chase, and the price paid. The study also analyzes the extent to which the period

saw increases in the values of those three variables. The findings strongly suggest

a decline in the ability of credit risk characteristics in predicting the interface

between households and mortgage market over the period.

JEL Codes: G01; G21; R31

Accepted: 26 February 2019

I. INTRODUCTION

The global financial crisis began in second half of 2007 with defaults in

U.S. subprime adjustable-rate mortgages (ARMs).

1

The default rate on such

loans hit 16% in August 2007, roughly triple that of mid-2005, and peaked at

* I thank the Federal Deposit Insurance Corporation (FDIC) for providing me the time to work on

the updates of the paper when I was a Visiting Scholar over there. The views expressed herein are

those of the author and do not necessarily reflect the views of the FDIC. I thank George

G. Pennacchi for his feedback and support. I also thank Andrea H. Beller, Allen N. Berger, Myron

Kwast, Oscar Mitnik, Nicholas N. Paulson, Gary D. Schnitkey, Robert L. Thompson, and seminar par-

ticipants at the FDIC, the Office of the Comptroller of the Currency, the U.S. Securities and

Exchange Commission, and the Bank of Canada for their comments. I thank Roger E. Cannaday

and Dawn Owens Nicholson for their support.

1 The term subprime refers to the credit quality of borrowers, who have poor credit histories

and high risks of loan default.

© 2019 International Review of Finance Ltd. 2019

International Review of Finance, 21:1, 2021: pp. 208–254

DOI: 10.1111/irfi.12262

25% in May 2008 (Bernanke 2007, 2008a, 2008b). In 2007, subprime ARMs

accounted for 43% of mortgage foreclosures. The number of mortgage foreclo-

sures jumped 75% from 2006 to 2007 and another 81% from 2007 to 2008.

2

By

August 2008, 9.2% of all outstanding mortgages were delinquent or in

foreclosure.

3

The defaults in the U.S. mortgage market were triggered by the interaction of

a decline in mortgage lending standards with a subsequent rise in interest rates

and decline in house prices. In the years leading up to the crisis, lenders had a

strong incentive to approve subprime mortgages at the expense of loan quality

because they could lower their risk of default loss by selling their mortgages

(including subprime) to Fannie Mae or Freddie Mac or to Wall Street dealers

that packaged them into mortgage-backed securities (MBS).

4

With the decline

in lending standards, the proportion of outstanding mortgages that were sub-

prime rose from 9% in 2004 to a peak of 21% in 2006.

The increase in interest rates that began in 2006 drove up rates on existing

ARMs; and with the decline in house prices, subprime borrowers especially

could not refinance their ARMs to obtain more favorable terms. The resulting

rise in subprime mortgage foreclosures imposed losses on lenders because

homes were worth less than the mortgage balance. And investors in subprime

MBS suffered significant losses as the spike in defaults drove down the value of

the securities.

5

A number of papers have found that the relaxation in lending standards on

home-purchase mortgages was a contributor to the crisis (Foote et al. 2008; Keys

et al. 2009, 2010; Mian and Sufi2009; Demyanyk and Van Hemert 2011;

Loutskina and Strahan 2011; Purnanandam 2011; Dell’Ariccia et al. 2012;

Goetzmann et al. 2012). But to date, the literature has not looked at the compo-

nents of credit risk and their ability to predict how households interface with

the mortgage market under the relaxed lending standards. The paper fills this

gap and looks at credit risk components (characteristics); and, with the relaxa-

tion in lending standards shows which credit risk characteristics—for example,

income, net worth, debt level, employment status, job stability—were no longer

important in predicting the interface between households and mortgage mar-

ket; that is, in predicting how much mortgage loan households receive from

creditors (financing levels) and which households purchase principal residences

(home purchase probability) and how much households pay for the principal

2 RealtyTrac, U.S. Foreclosure Market Reports, 2007 and 2008.

3 Mortgage Bankers Association, National Delinquency Survey, 2007 and 2008.

4 Some of the MBS had recourse options under which the original lenders would bear the losses

in the event of defaults on the underlying mortgages. Since 2008, Fannie Mae, Freddie Mac,

and others have been exercising the recourse options and putting defaulted loans back to

original lenders.

5 Financial Institutions invested heavily in MBS because they paid higher interest rates than

U.S. Treasuries and they had a triple-A rating from credit rating agencies. The great disparity

between ratings and performance on subprime MBS has led to widespread proposals for

changes in the way credit ratings in general are conducted.

© 2019 International Review of Finance Ltd. 2019 209

Relaxed Credit Standards in the U.S. Housing Boom

residences they purchase (price paid for the purchase of properties). Among the

loan applicants who decide to purchase home property, when the credit supply

is low, only low-risk applicants receive mortgage loans. In contrast, under

relaxed lending standards, credit supply is high and high-risk applicants are also

granted mortgage loans given the same pool of loan applicants. The expecta-

tion is that high-risk households will be granted similar mortgage amounts as

low-risk households and be as likely as low-risk households to purchase homes

and afford expensive homes. This will lead to a weaker correlation between

credit risk characteristics and mortgage financing levels, home purchase proba-

bility, and the price paid for the purchase of properties. By documenting the

weakening correlation, the overall goal of the paper is to find out how the aver-

age risk characteristics of households, who were granted mortgage loans, chan-

ged following the relaxation in lending standards. The goal is not to document

the changes in household decision-making conditional on their risk characteris-

tics and whether households cared less about their credit risk characteristics

when making decisions regarding the purchase of a primary residence; such as

how much mortgage loan to apply for or whether to purchase a principal resi-

dence or how much to spend on home purchases. Keeping track of changes in

the household risk characteristics, who were granted mortgage loans, will

inform supervisors about changes in mortgage lending by creditors and can act

as a warning mechanism regarding a decline in lending standards.

Although the present analysis captures mostly changes in supply side and

lenders’concern about borrower riskiness, it is not based on lender data but

rather on the triennial Survey of Consumer Finances (SCF), which is a house-

hold data sponsored by the Board of Governors of the Federal Reserve System.

The data shows a variety of borrower covariates, measured in different cross-

sections over time, which allows a decomposition of several of the different

components of credit risk. The analysis is based on household data as opposed

to lender data as losses in household wealth and employment were a major

determinant of the depth and duration of the subsequent recession and

extended period of low growth in gross domestic product (GDP).

The analysis compares households observed in 1998 to households observed

in 2007. Lending standards had not been relaxed in 1998 compared with earlier

years. Nonetheless, subprime lending (extending loans to borrowers with weak

credit characteristics) did exist by then, as stimulated by the federal govern-

ment’s goal of improving the home ownership rate for low-income house-

holds.

6

By 2007, relaxed lending standards had been in effect for some time.

Some may argue that 2007 is a tricky year to do the analysis because household

behaviors may have been affected from the rise in subprime defaults. An

intriguing feature of the 2007 survey results is that household expectations

about the economy were essentially the same as those reported in the 1998 sur-

vey and rises in the mortgage defaults did not impact the household

6 Subprime lending can be consistent with undiminished lending standards when subprime

borrowers must accept more stringent loan terms than other borrowers.

© 2019 International Review of Finance Ltd. 2019210

International Review of Finance