Academic journal article Federal Reserve Bank of Atlanta, Working Paper Series

What Explains Differences in Foreclosure Rates? A Response to Piskorski, Seru, and Vig

Academic journal article Federal Reserve Bank of Atlanta, Working Paper Series

What Explains Differences in Foreclosure Rates? A Response to Piskorski, Seru, and Vig

Article excerpt

Working Paper 2010-8 March 2010

Abstract: In this note we discuss the findings in Piskorski, Seru, and Vig (2010) as well as the authors' interpretation of their results. First, we find that small changes to the set of covariates used by Piskorski, Seru, and Vig significantly reduce the magnitude of the differences in foreclosure rates between securitized and nonsecuritized loans. Second, we argue that early payment defaults (EPD) are not a valid instrument for the securitization status of the loans and that the empirical implementation chosen by the authors for using EPD is not a valid instrumental variables approach. Finally, we discuss the use of foreclosure rates as a measure of renegotiation and argue that explicitly using modification rates of delinquent mortgages is a better way of studying renegotiation activity. On balance, the evidence in Piskorski, Seru, and Vig indicates that there are at most small differences in the outcomes of delinquent loans, but whether those differences reflect accounting issues, willingness to renegotiate, or unobserved heterogeneity remains an open question.

JEL classification: D11, D12, G21

Key words: foreclosure, mortgage, house prices

1. Introduction

Piskorski, Seru, and Vig (2010) (hereafter referred to as PSV), use micro-panel data on the repayment behavior of mortgage borrowers to study the role of securitization in foreclosure loss mitigation efforts. The authors find a foreclosure bias in servicing decisions induced by the process of securitization. Specifically, they find large differences in foreclosure sale rates between seriously delinquent loans held on a mortgage servicer's portfolio (hereafter referred to as "portfolio-held" loans) and loans securitized by private trusts (hereafter referred to as "private-label" loans). According to their estimates, the probability of a portfolio-held, seriously delinquent loan being foreclosed upon is 3.8 to 7 percentage points lower in absolute terms and 18 to 32 percent lower relative to the unconditional foreclosure mean of private-label loans. They interpret this difference in foreclosure rates as evidence that servicers of private-label loans renegotiate with borrowers less often than servicers of portfolio-held loans do. In turn, they attribute such differences in renegotiation behavior to be the result of contract frictions present in securitization trusts that prevent an optimal amount of renegotiation to take place.

In this comment we discuss three issues that we believe shed serious doubts on the paper's empirical findings as well as the authors' interpretation of those findings. First, in section 1, we replicate the main PSV foreclosure rate results and show that they are extremely sensitive to the addition of covariates and are not particularly robust to a subsample analysis. Second, we discuss a "quasi experiment" that the authors use in an attempt to solve the endogenous selection issues inherent in the determination of whether or not a mortgage is securitized. A mortgage originator decides which loans to sell from its portfolio to issuers of mortgage-backed-securities (MBS) on the secondary mortgage market, and in turn, an MBS issuer decides which loans to purchase from mortgage originators. Since the dataset does not contain all of the variables that determine these decisions, there is the potential for omitted variable bias in the empirical estimation. PSV try to solve this problem by using early-payment default stipulations embedded in many securitization deals that require the originating institution to buy back loans that default within a specific number of months of being sold. They argue that these stipulations generate exogenous variation in the ultimate decision of whether or not a loan is securitized. In section 2 we provide a detailed list of reasons why early-payment default stipulations do not solve the securitization identification problem. Finally, in section 3, we discuss the authors' interpretation that differences in foreclosure sale rates between private-label and portfolio-held loans represent differences in the incidence of renegotiation between borrowers and servicers of each type of loan, respectively. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.