Foreclosure Contagion and the Neighborhood Spillover Effects of Mortgage Defaults

• Date: 01 October 2019
• Published date: 01 October 2019
• DOI: http://doi.org/10.1111/jofi.12821
• Author: ARPIT GUPTA

THE JOURNAL OF FINANCE •VOL. LXXIV, NO. 5 •OCTOBER 2019

Foreclosure Contagion and the Neighborhood

Spillover Effects of Mortgage Defaults

ARPIT GUPTA∗

ABSTRACT

In this paper, I identify shocks to interest rates resulting from two administrative

details in adjustable-rate mortgage contract terms: the choice of financial index and

the choice of lookback period. I find that a 1 percentage point increase in interest

rate at the time of adjustable-rate mortgage (ARM) reset results in a 2.5 percentage

increase in the probability of foreclosure in the following year, and that each foreclo-

sure filing leads to an additional 0.3 to 0.6 completed foreclosures within a 0.10-mile

radius. In explaining this result, I emphasize price effects, bank-supply responses,

and borrower responses arising from peer effects.

OVER 4MILLION COMPLETED FORECLOSURES TOOK place between January 2007

and December 2010, and another 8.2 million foreclosures were initiated dur-

ing the same period (Blomquist (2012)). While the foreclosure crisis remains

historic in its aggregate cost to homeowners and investors, understanding the

precise mechanisms and channels behind this wave of mortgage defaults re-

mains a challenge. In this paper, I examine the role of neighborhood spillover

effects from foreclosures as an important amplification mechanism behind this

foreclosure crisis.

There are several plausible mechanisms through which foreclosures can af-

fect the default patterns of their geographical peers. First, foreclosures reduce

the market price of neighboring homes, which may induce those borrowers

to default due to the negative equity that results. Alternatively, lenders may

deny refinancing opportunities to prospective lenders from areas that have

previously experienced foreclosure activity. A separate possible channel re-

lies on the role of information. This channel operates through borrowers’ re-

assessment of the costs of default after exposure to neighboring foreclosures,

∗Arpit Gupta is with the Stern School of Business, NYU. I thank Tomasz Piskorski, Edward

Morrison, Wei Jiang, and DanielWolfenzon for their encouragement and guidance. I also thank the

Editor, Amit Seru, the associate editor,and two anonymous referees for their valuable suggestions

that substantially improved the paper. Seminar participants at Columbia GSB, the New York

Federal Reserve Bank, UCLA Anderson, Princeton, MIT Sloan, Wharton, Chicago Booth, UT-

Austin McCombs, NYU Stern, and Copenhagen Business School provided helpful comments. I

am grateful to Equifax, BlackBox Logic, DataQuick, and Zillow for their data, research support,

and infrastructure that were invaluable for the analysis in this paper. The Chazen Institute at

Columbia provided critical funding to support this research. I have read the Journal of Finance’s

disclosure policy and have no conflicts of interest to disclose.

DOI: 10.1111/jofi.12821

2249

2250 The Journal of Finance R

or through their reassessment of the stigma associated with mortgage nonpay-

ment (as emphasized, for instance, in Guiso, Sapienza, and Zingales (2013)).

Finally, a direct treatment effect involves foreclosures leading to an increase in

crime, vandalization, and other forms of property depreciation that reduce the

amenity value of the neighborhood.1Though the price effects of foreclosures

have been previously studied,2we know comparatively little about the causal

implications of foreclosure activity on neighborhood default behavior or about

the precise mechanisms through which such peer effects operate.

The key contribution of this paper is the development of an empirical setting

that allows for the causal estimation of foreclosure spillovers. Understanding

foreclosure externalities is important to understanding the seemingly snow-

balling wave of foreclosures observed during the period of initial subprime

mortgage defaults and the ensuing financial crisis, as well as the policy inter-

vention designed to combat these spillovers. For instance, as Timothy Geithner

argued in a speech in February 2009 introducing the Financial Stability Plan,3

“As house prices fall, demand for housing will increase, and conditions will

ultimately find a new balance. But now, we risk an intensifying spiral in which

lenders foreclose, pushing house prices lower and reducing the value of house-

hold savings, and making it harder for all families to refinance.” My work

examines the role of foreclosure externalities and the extent to which they con-

stitute an amplification mechanism that potentially motivates these federal

housing relief efforts.

A central econometric challenge to understanding foreclosure contagion is

the issue of reflexivity and the endogenous assignment of default, as empha-

sized by Manski (1993). On the one hand, observing that foreclosures appear

to be geographically clustered is consistent with geographically proximate bor-

rowers suffering a common shock (for instance, a local plant closure). On the

other hand, clustering of foreclosures may be due to the fact that geographi-

cally proximate borrowers share common (possibly unobservable) characteris-

tics that predict mortgage default.4Understanding the contribution of foreclo-

sures to the default behavior of neighboring properties has thus proven to be a

key challenge in prior literature on this subject.

To address this econometric issue, I introduce a novel instrument based on

exogenous shocks to interest rates on adjustable-rate mortgage (ARM) loans. I

argue that while these shocks impact the foreclosure of resetting households,

they affect the default choices of neighboring properties only through the chan-

nel of default on treated households. ARMs in the United States—which were

quite common among subprime and jumbo-prime borrowers during the boom

1For instance, see Immergluck and Smith (2005,2006) on crime and local amenities.

2See, for instance, Campbell, Giglio, and Pathak (2011) or Mian, Sufi, and Trebbi (2015).

3This speech laid the groundwork for the Home Affordable Modification Program

(HAMP) and a variety of other government programs. See http://www.treasury.gov/press-

center/press-releases/Pages/tg18.aspx.

4For example, this seems to be the case among foreclosure completions in Phoenix, as shown

in Internet Appendix Figure IA3. The Internet Appendix may be found in the online version of

this article.

Foreclosure Contagion and the Neighborhood Spillover Effects of Mortgage Defaults 2251

in house price appreciation—are characterized by an initial teaser rate that

resets to a market interest rate (plus a margin term to account for risk) after

an initial period that typically lasts two, three, or five years.5I focus on two

previously unexplored aspects of the mechanics according to which ARMs re-

set: the choice of financial index and the choice of lookback date. When ARMs

reset, the market interest rate component of the new payment is derived from

the prevailing market interest rate according to an index (typically LIBOR

or Treasury), taken a certain number of (lookback) days from the reset date.

Importantly for my analysis, these interest rates are then fixed for a period of

time between 6 and 12 months after initial reset.

I find variation in these contract terms that drives borrower payment

amounts subsequent to mortgage reset. Focusing on the choice of financial

index, I find that while LIBOR and Treasury rates tracked each other quite

closely prior to the financial crisis, a large spread between rates emerged dur-

ing the financial crisis that resulted in large differences in payments among

borrowers linked to different indices. LIBOR borrowers resetting in January

2009, for instance, paid on average $11,000 more than otherwise identical

Treasury borrowers who reset the same month. Next, I find that substantial

interday volatility in interest rates led to variation in payments paid by bor-

rowers with different lookback terms—for instance, 15 lookback days instead of

45. I thus argue that both forms of interest rate variation, which are ultimately

determined by administrative details of loan contracts, are unlikely to be re-

lated to other aspects of loan performance, and lead to substantial variation in

payment terms after reset.

In the first stage of my analysis, I find that the size of the within-month

interest rate shock resulting from these contract differences drives default

and foreclosure rates among resetting mortgages. The first-stage results are

substantial, suggesting that a 100 basis point increase in interest rates corre-

sponds to a roughly 2.5% rise in the probability of experiencing a foreclosure in

the subsequent 12 months—a substantial increase relative to a baseline fore-

closure rate of 8%. These results are in line with existing work on mortgage

resets (see, for instance, Fuster and Willen (2015)). I contribute to this litera-

ture by obtaining a tighter empirical setting using the within-month variation

in interest rates as a shock to the reset window.

My setting provides a clean form of identification for local foreclosure

spillovers. Focusing on a broad sample of resetting ARM holders, I develop a

novel merge algorithm that links information on these loans—including their

contract terms, credit scores, and loan performance—to deeds records contain-

ing precise geographical information based on where the borrowers live. I then

construct neighborhoods that consist of all transacting properties within a ra-

dius of 0.10 miles around the resetting ARM holder to analyze the spillover

effects of default.

I contrast my first-stage results, which use the interest rate variation induced

by the idiosyncratic contract terms such as lookback date and index to predict

5Initial teaser lengths of 1, 7, or 10 years also exist but are less common.