Forced Asset Sales and the Concentration of Outstanding Debt: Evidence from the Mortgage Market

• DOI: http://doi.org/10.1111/jofi.12494
• Published date: 01 June 2017
• Date: 01 June 2017

THE JOURNAL OF FINANCE •VOL. LXXII, NO. 3 •JUNE 2017

Forced Asset Sales and the Concentration

of Outstanding Debt: Evidence

from the Mortgage Market

GIOVANNI FAVARA and MARIASSUNTA GIANNETTI∗

ABSTRACT

We provide evidence that lenders differ in their ex post incentives to internalize price-

default externalities associated with the liquidation of collateralized debt. Using the

mortgage market as a laboratory, we conjecture that lenders with a large share of

outstanding mortgages on their balance sheets internalize the negative spillovers

associated with the liquidation of defaulting mortgages and thus are less inclined to

foreclose. We provide evidence consistent with our conjecture. Arguably as a conse-

quence, zip codes with a higher concentration of outstanding mortgages experience

smaller house prices declines. These results are not driven by unobservable zip code

or lender characteristics.

FORCED SALES OF REAL AND FINANCIAL ASSETS fetch prices below their funda-

mental values (Shleifer and Vishny (1992)). When markets are illiquid, such

sales generate price spillovers that reduce the value of similar assets held by

other market participants. This may lead to further defaults and price-default

spirals (Kiyotaki and Moore (1997), Gromb and Vayanos (2002), Brunnermeier

and Pedersen (2009)). While a large literature documents the existence of fire

sales and their spillover effects (Shleifer and Vishny (2011)), little attention

∗Favara is with the Federal Reserve Board. Giannetti is with the Stockholm School of Eco-

nomics, CEPR, and ECGI. We thank Bruno Bias (the Editor); Michael Roberts (the Coeditor);

an anonymous Associate Editor; two anonymous referees; Manuel Adelino; Gene Amromin; El-

liot Anenberg; Eric Engstrom; Julian Franks; Paul Kalem; Anastasia Kartasheva; Amir Kermani;

Benjamin Keys; Elena Loutskina; Steven Ongena; Karen Pence; Anthony Pennington-Cross; Uday

Rajan; AvriRavid; Amit Seru; Steve Sharpe; Shane Sherlund; Greg Udell; Vikrant Vig; and seminar

participants at the American Finance Association, Western Finance Association, NBER Financing

Housing Capital Conference, London Business School Finance Summer Symposium, New York

University Stern School of Business, University of British Columbia Sauder School of Business,

Federal Reserve Bank of Chicago, Board of Governors of the Federal Reserve System, Boston

University, Georgia State University, University of California Irvine, University of Lausanne and

EPFL, University of Amsterdam, Goethe University in Frankfurt, IDC Annual Conference in Fi-

nancial Economics Research in Herzliya, EFA in Lugano, FIRS Conference in Quebec City, SED

Conference in Seoul, and the Summer Meeting of the Econometric Society at University of Southern

California. We also thank Mihir Gandhi for outstanding research assistance. Giannetti acknowl-

edges financial support from the Jan Wallander and Tom Hedelius Foundation and the Bank of

Sweden Tercentenary Foundation. This paper represents the views of the authors and not those

of the Federal Reserve System or its Board of Governors. The authors do not have any conflicts of

interest, as identified in the Disclosure Policy.

DOI: 10.1111/jofi.12494

1081

1082 The Journal of Finance R

has been paid to the ex post incentives of market participants to avoid the

negative spillovers associated with forced asset sales.

The purpose of this paper is to show that lenders with a high share of collat-

eralized debt in their portfolios internalize the spillover effects of liquidation

decisions on collateral values and are inclined to renegotiate their debt to avoid

price-default spirals. Using differences in U.S. local housing markets (census

tracts or zip codes) during the 2007 to 2010 housing crisis, we find evidence

that such incentives are at work and are economically significant.

The U.S. housing crisis is an ideal laboratory for testing this conjecture for

three reasons. First, mortgages, the standard debt contracts in the housing

market, entitle lenders to seize the houses and sell them through a foreclosure

process if borrowers default. Second, as the housing market is illiquid, fore-

closures may generate price discounts that tend to spill over to nondistressed

neighboring houses (Harding, Rosenblatt, and Yao (2009), Campbell, Giglio,

and Pathak (2011), Anenberg and Kung (2014), Hartley (2014)). Third, the re-

cent crisis has seen an unprecedented increase in foreclosures and decline in

house prices, with feedback loops between foreclosures and prices contributing

to the severity of the crisis. For instance, it has been shown that foreclosures

led to a generalized decline in house prices (Mian, Sufi, and Trebbi (2015)),

which in turn caused additional foreclosures as borrowers moved into negative

equity positions (Elul et al. (2010)), triggering further price declines (Guren

and McQuade, 2013).

We begin the analysis with a stylized model of the housing market in which

negative income shocks force distressed homeowners to default on their mort-

gage obligations, and foreclosures trigger a decline in house prices as liquida-

tions create an imbalance of housing demand and supply. When mortgages are

held by many (atomistic) lenders, each lender places no weight on the effects

of its foreclosure decisions on local house prices, and therefore foreclosures are

likely to be followed by further foreclosures. In contrast, when lenders hold

a large share of the mortgages outstanding in a neighborhood on their bal-

ance sheets, they internalize the adverse effects of their liquidation decisions

on house prices in that neighborhood and have stronger incentives to avoid

foreclosures, mitigating the adverse effects of foreclosures on house prices.

To test this theoretical prediction, we perform two sets of tests. First, for a

subset of lenders with available data on mortgage performance, we use loan-

level data to test whether a lender’s incentives to foreclose on defaulting mort-

gages depend on the proportion of the outstanding mortgages it has retained

on its balance sheet in a census tract. Second, we use aggregate zip code-level

data to test whether lenders’ incentives have implications for house prices.

In the loan-level analysis, we measure the share of outstanding mortgages

kept on lenders’ balance sheets at the census tract level. We expect this mea-

sure to capture lenders’ incentives to internalize foreclosure externalities in a

neighborhood. Consistent with the predictions of our stylized model, we find

that, while this share is negatively correlated with lenders’ propensity to fore-

close, there is no such correlation for securitized mortgages. We obtain these

results in specifications that include lender and local market fixed effects to

Forced Asset Sales and the Concentration of Outstanding Debt 1083

purge unobserved lender and local market characteristics that might correlate

with the share of mortgages retained by the lender.

To mitigate the concern that lenders retain more mortgages in census tracts

in which they have an informational advantage, all of our tests are conducted

conditional on mortgages being 90+days delinquent. The very fact that we

focus on seriously delinquent mortgages should reduce concerns that lenders

had a prior informational advantage on the quality of these mortgages. In

addition, we exploit exogenous variation induced by mergers of nonfailing large

banks to instrument the share of mortgages retained in a neighborhood. The

nature and size of these mergers make it unlikely that they are related to local

market characteristics. Our results are qualitatively robust and quantitatively

stronger when we use this instrumentation strategy.

We also provide evidence consistent with other predictions of the model. For

instance, the model implies that lenders’ foreclosure decisions are strategic

substitutes. This means that the likelihood that a lender forecloses on a de-

faulting mortgage depends negatively on the share of outstanding mortgages

on its own balance sheet and positively on the share of mortgages retained by

other lenders. The intuition is that house prices drop less when other lenders

renegotiate defaulting mortgages, strengthening a lender’s incentives to fore-

close in order to maximize its payoff of liquidation. Our model also implies that

lenders should be more inclined to foreclose in areas in which it is easier to

resell foreclosed properties and less inclined to foreclose in highly distressed

areas. In line with these predictions, we find that the share of outstanding

mortgages retained by a lender reduces the propensity to foreclose to a lesser

extent in areas with a higher share of local mortgages retained by other lenders

and in highly desirable neighborhoods, and to a greater extent in areas with

more distressed households. We find no evidence supporting alternative mech-

anisms, including differences in individual lenders’ organizational capabilities

to foreclose or renegotiate mortgages in different geographical areas.1

Next, we consider the aggregate implications of our hypothesis using zip

code–level data. We perform tests at the zip code level because zip codes are

the smallest geographical areas for which we can obtain comprehensive data

on house prices, and arguably the largest areas within which foreclosures are

likely to generate negative externalities on house prices.2We construct a zip

code–level measure of the concentration of mortgages on lenders’ balance sheets

using data on mortgages retained by the four largest holders in a zip code. Zip

codes with a higher concentration of outstanding mortgages are expected to

experience lower foreclosure rates. We find empirical support for this conjecture

even after instrumenting the concentration of mortgages on lenders’ balance

sheets with exogenous bank mergers. A one-standard-deviation change in the

1The finding that within-lender differences in organizational capabilities do not drive our results

is not at odds with the findings of Agarwal et al. (2015) that loss mitigation in mortgage markets

is explained largely by differences in renegotiation capabilities across lenders.

2See Mian, Sufi, and Trebbi (2015) for empirical evidence on foreclosure externalities at the zip

code level.