There is widespread concern that an inefficiently low number of mortgages have been modified during the current crisis, and that this has led to excessive foreclosure levels, leaving both families and investors worse off. We use a large dataset that accounts for approximately 60 percent of mortgages in the United States originated between 2005 and 2007, to shed more light on the determinants of mortgage modification, with a special focus on the claim that delinquent loans have different probabilities of renegotiation depending on whether they are securitized by private institutions or held in a servicer’s portfolio. By comparing the relative frequency of renegotiation between private-label and portfolio mortgages, we
are able to shed light on the question of whether institutional frictions in the secondary mortgage market are inhibiting the modification process from taking place.
Our first finding is that renegotiation in mortgage markets during this period was indeed rare. In our full sample of data, approximately 3 percent of the seriously delinquent borrowers received a concessionary modification in the year following their first serious delinquency, while fewer than 8 percent received any type of modification. These numbers are extremely low, considering that foreclosure proceedings were initiated on approximately half of the loans in the sample and completed for almost 30 percent of the sample. Our second finding is that a comparison of renegotiation rates for private-label loans and portfolio loans, while
controlling for observable characteristics of loans and borrowers, yields economically small, and for the most part, statistically insignificant differences. This finding holds for a battery of robustness tests we consider, including various definitions of modification, numerous subsamples of the data, including subsamples for which we believe unobserved heterogeneity to be less of an issue, and consideration of potential differences along the intensive margin of renegotiation.
Since we conclude that contract frictions in securitization trusts are not a significant problem, we attempt to reconcile the conventional wisdom held by market commentators, that modifications are a win-win proposition from the standpoint of both borrowers and lenders, with the extraordinarily low levels of renegotiation that we find in the data. We argue that the data are not inconsistent with a situation in which, on average, lenders expect to recover more from foreclosure than from a modified loan. At face value, this assertion may seem implausible, since there are many estimates that suggest the average loss given foreclosure is much greater than the loss in value of a modified loan. However, we point out that renegotiation exposes lenders to two types of risks that are often overlooked by
market observers and that can dramatically increase its cost. The first is “self-cure risk,” which refers to the situation in which a lender renegotiates with a delinquent borrower who does not need assistance. This group of borrowers is non-trivial according to our data, as we find that approximately 30 percent of seriously delinquent borrowers “cure” in our data without receiving a modification. The second cost comes from borrowers who default again after receiving a loan modification. We refer to this group as “redefaulters,” and our results show that a large fraction (between 30 and 45 percent) of borrowers who receive modifications, end up back in serious delinquency within six months. For this group, the
lender has simply postponed foreclosure, and, if the housing market continues to decline, the lender will recover even less in foreclosure in the future.
We believe that our analysis has some important implications for policy. First, “safe harbor provisions,” which are designed to shelter servicers from investor lawsuits, are unlikely to have a material impact on the number of modifications and thus will not significantly decrease foreclosures. Second, and more generally, if the presence of self-cure risk and redefault risk do make renegotiation less appealing to investors, the number of easily “preventable” foreclosures may be far smaller than many commentators believe.
Ideally, a lender would like to know whether a borrower will either (1) self-cure; (2) be cured by a loan modification that has no reduction in balance; (3) be cured by a loan modification that has a balance reduction or (4) is beyond help. If anyone can figure out how to write a contract that induces the borrower to reveal their type, they will have the solution to the crisis. The closest I can think of is one that allows borrowers to short-sell their houses (a la Hancock and Passmore) with reference to some sort of price index, so as to prevent gaming of the sales process.