During the current foreclosure crisis, lenders have seemed far more willing to foreclose on delinquent borrowers rather than offer them loan modifications. Some commentators have argued that this was not always the case. They claim that loan modifications are infrequent today because so many loans have been securitized, and thus are not owned by any one person or firm. They also say that the modern securitization process reduces loan modifications because securitization separates the entity that makes the modification decision—that is, the mortgage servicer—from the entities that gain the most if a foreclosure is avoided—that is, the mortgage investors. As we pointed out in our last post, Yale economist John Geanakoplos and Boston University law professor Susan Koniak argued in a March 2008 New York Times op-ed that the uncomplicated relationship between banks and borrowers in the good old days allowed the banks to work out modifications when their borrowers ran into trouble.
The Congressional Oversight Panel, created by Congress in October 2008 to "review the current state of financial markets and the regulatory system," expressed a similar belief in a March 2009 report on the state of the U.S. housing market:
For decades, lenders in this circumstance [that is, with troubled borrowers] could negotiate with can-pay borrowers to maximize the value of the loan for the lender (100 percent of the market value) and for the homeowner (a sustainable mortgage that lets the family stay in the home). Because the lender held the mortgage and bore all the loss if the family couldn't pay, it had every incentive to work something out if a repayment was possible.
Even in the good old days, lenders reluctant to restructure
Such claims, however, have usually been made with little or no reference to supporting research. Fortunately, a recent paper by Andra Ghent of Baruch College exploits a new data set to shed considerable light on this topic. Her findings argue against the idea that lender reluctance to modify is a recent phenomenon.
Ghent uses a data set from the National Bureau of Economic Research (NBER) that covers mortgages from 1920 to 1939, a period that encompasses the massive housing turmoil of the Great Depression. The data set consists of "mortgage experience cards," which the NBER collected in the 1940s from mortgage lenders in the New York metropolitan area. On the cards are the answers to short questionnaires about the characteristics of individual mortgage loans (see page 5 of Ghent's paper for an example). The cards also contain explicit information about any loan modifications, including the date of the modification and whether it was principal reduction, interest-rate reduction, change to the amortization schedule, or something else. The cards include loans from three types of mortgage lenders: life insurance companies, savings and loans, and commercial banks.1
Ghent finds few modifications in these cards, and these few were not particularly generous. Using a fairly conservative definition of what constitutes a concessionary modification, Ghent finds that approximately 5 percent of loans originated between 1920 and 1939 were modified, while 14 percent were terminated by foreclosure or a deed-in-lieu of foreclosure (the latter occurs when the owner surrenders the house to the lender without going through the foreclosure process). Of the loans that received a concessionary modification, about 40 percent received an interest rate reduction, which Ghent defines as an interest rate cut of at least 25 basis points (relative to origination) resulting in a new rate that is at least two standard deviations below the average interest rate on newly originated loans. The average rate reduction was only 78 basis points below the prevailing interest rate of new originations, suggesting that interest rate cuts were not particularly generous.
Another 40 percent of the modified loans received reductions in their amortization schedules, which would have likely decreased the required mortgage payments. However, Ghent points out that most of these extended amortizations occurred before 1930. In the period 1930–32, when house prices fell and unemployment rose the most, this type of modification was rare.
Principal balance reductions—and increases
Ghent also finds that less than 2 percent of all loans received principal balance increases. She argues that such increases may correspond to instances of forbearance. Forbearance occurs when a lender reduces the required mortgage payment for a short period. At the end of the period, the lender adds the arrears back to the loan balance. We have a minor quibble on this point: today, forbearance is not considered a permanent concessionary modification when the lender does not have to write down any debt.
What about principal reductions? Perhaps the most surprising finding is that the data set shows no instances of principal reduction in the New York City metropolitan area and only a handful of instances in a broader sample that includes the entire states of Connecticut, New Jersey, and New York over a similar period. To us, this low number of principal reductions is compelling evidence that even Depression-era lenders were averse to renegotiating with troubled borrowers, just as lenders are today.
Balloon mortgages sank some borrowers
Another interesting finding concerns the refinancing decisions of lenders. Short-term balloon mortgages were more common in the 1920s and 1930s than they are today, and various scholars have linked the high foreclosure rate of the Depression to the unwillingness of lenders to refinance these mortgages when they came due. In fact, lender reluctance to refinance maturing mortgages is often used to explain the existence of the Home Owners' Loan Corporation (HOLC), a government organization set up in the early 1930s to refinance troubled mortgages. Ghent revisits this hypothesis with her data, measuring the frequency at which short-term balloon mortgages ended in foreclosure. She finds that balloon mortgages that were about to expire did indeed experience increased rates of foreclosure (see Ghent's table 4). However, this relationship only exists during the years when HOLC was purchasing a great many loans (1933–35). In other years, balloon mortgages were no more likely to end in foreclosure than other loans.
To us, this finding suggests a "HOLC effect." While HOLC was actively buying loans, private lenders may have refused to roll them over so that the borrowers would qualify for a HOLC refinance. If they did, then the lenders would be paid close to par for the loans by the government (see our previous post about the generosity of the HOLC program). In particular, the lenders received what were effectively government bonds in return for their mortgage. While these bonds carried lower interest rates, they carried vastly less credit risk as well.
To explain her findings, Ghent points to information problems between borrowers and lenders. In particular, lenders may not have known which borrowers were likely to truly need modifications, nor did they know with certainty which borrowers were likely to re-default if a modification were offered. Note that these information problems must have been quite severe. The national unemployment rate hit 10.8 percent in November 1930 and stayed in double digits for more than a decade. In this environment, a borrower asking for a modification was quite likely to really need one. The fact that lenders made few modifications suggests some strong intrinsic hurdles to renegotiation when information between borrowers and lenders is less than perfect.
Old problems, new analysis
The crucial policy question is what the Depression-era reluctance of lenders to renegotiate teaches us about today's foreclosure crisis. Ghent surmises that the information problems are less of an issue in the current environment, but we disagree. Even with better data and screening technology, today's lenders face significant information problems when deciding on modifications. Moreover, Ghent's paper is also informative on the role of securitization in reducing modifications. Even when individual lenders owned entire loans, modifications were rare.
All told, Ghent's paper is full of solid analysis on a topical subject. And while she doesn't quite go this far, we believe that her findings not only confirm the importance of information problems, but also they may bury the notion that securitization is the primary obstacle to renegotiation in the current foreclosure crisis.
Kris Gerardi
Research economist and assistant policy adviser at the Federal Reserve Bank of Atlanta
Chris Foote, senior economist and policy adviser at the Federal Reserve Bank of Boston
1 Ghent argues that the data set probably provides a representative sample of loans held by life insurance companies and commercial banks in the 1920s and 1930s, but is less likely to be representative of loans held by savings and loans due to a survivorship bias. Unlike life insurance companies and commercial banks, savings and loans were not able to reliably report data on their inactive loans at the time of the survey.