Source: Dissertation Abstracts International, Volume: 77-12(E), Section: A.

Adviser: Andrew Metrick.

Access restricted by licensing agreement.

This item is not available from ProQuest Dissertations & Theses.

In their responses to financial crises, governments often face a series of tradeoffs with unclear consequences. One frequent dilemma a banking regulator faces, for instance, is when to enforce the failure of financially troubled banks. Another pressing question, particularly in the recent crisis, is how to address mortgage borrower defaults and reduce the social cost of foreclosures.

Many of these questions lack relevant empirical studies. Given the endogeneity of policy, simple analysis of available data may provide misleading results. Theory does not provide clear guidance, typically offering more than one relevant model with conflicting predictions. What we'd like to have are more experiments to address these questions, and that is what I provide in this dissertation. Through a combination of natural arid field experiments, I study bank and household responses to treatments intended to alleviate financial distress, and contribute to our understanding of borrower behavior.

In the first chapter of my dissertation, "Does Going Easy on Distressed Banks Help the Macroeconomy?," I use a natural experiment in regulatory policy provided by the U.S. Savings and Loan Crisis of the 1980s to study the net effects of regulatory forbearance1 on aggregate output growth. Historically, regulators often err on the side of delay in recognising--what could be temporary--losses in the banking sector, and instead tend to "kick the can down the road." But what are the net effects of delay in recognising financial losses on real economic growth?

The expected net impact of regulatory forbearance on aggregate outcomes can be positive or negative depending on the relative importance of opposing channels. Enforcing capital requirements can lead to unnecessary losses if bank deleveraging results in forced asset sales (Shleifer and Vishny, 2010). Forced closures of banks is understood to disrupt the credit supply and dampen economic growth (Ashcraft, 2005; Bernanke, 1983). On the other hand, highly leveraged banks might engage in socially undesirable behavior with long-run costs. Under regulatory forbearance, banks are the borrowers, and deposit insurance funds are the creditors, exercising what may be excessive leniency and allowing banks to gamble on activities that may riot be socially optimal for long-run growth, owing to moral hazard or other misalignments in incentives (Akerlof arid Romer, 1993; Gorton arid Rosen, 1995; Rajah, 1994).

To measure the net effects of forbearance on aggregate output growth, I exploit the historical accident that during the Savings & Loan Crisis, we have two regulatory authorities pursuing two different policies within the same country in response to the same crisis. At the time, there are two significant federal deposit insurance funds in the U.S., and one is forced to be easier on troubled banks than the other, delaying failure and giving banks more room to expand their lending into new asset classes. This provides a difference-in-differences setting for a causal estimate of the effects of two different resolution approaches to undercapitalised banks. I compare the outcomes of states by the levels of forbearance their banks' receive. I also use the limited diffusion of a particular bank charter during the 1800s to instrument for the extent of regulatory forbearance in the 1980s.

I find that high forbearance states initially experience greater loan growth and real estate activity. But after regulatory oversight is consolidated and forbearance ends in 1989, the same states subsequently suffer larger declines in real estate, credit, and aggregate output growth. The economic magnitudes I estimate indicate that following the stricter of the two regulatory policies could have avoided the 1990-1991 recession.

In the second and third chapters of my dissertation, titled "No Such Thing as a Free Option? Offers to Modify Mortgages and Borrower Mistakes" and "Promoting Automated Repayments and Mortgage Borrower Outcomes," I present evidence of deviations from a traditional model of mortgage borrowers where behavior is purely and efficiently driven by financial incentives. I conduct two different field experiments with mortgage borrowers in the U.S. to study household debt repayment behavior. I randomize debt-forgiving mortgage modifications in one, and randomize informational treatments encouraging the automation of debt repayment in the other field experiment.

While there are many findings of inefficient stock market investor behavior, relatively less behavioral research has been produced on mortgage borrowing, which is surprising from the perspective that it is a larger part of the household balance sheet for most U.S. households. The appropriate model to predict repayment behavior was critical in the most recent crisis, however. At their peak, more than $800 billion of mortgages were in default, and over 5 million households faced foreclosure. Initial government-sponsored mortgage modification programs such as HAMP focused on affordability, and largely failed. Many critics suggested the reason for HAMP's lack of success was that LTV was unaltered by the modification programs, as borrowers do not have a reason to repay their debts unless their Loan-To-Value (LTV, the ratio of mortgage principal balance to the market value of the property) is low enough.

A borrower's decision of when to default or prepay has traditionally been modeled as the exercise a financial option (Kau and Keenan, 1995). Inspired by this view, consumer advocates, investors, and academics, called for a policy of debt forgiveness in the most recent crisis to reduce the principal balance of borrowers' mortgage debts. These calls influenced a 2012 National Mortgage Settlement over illegal foreclosure practices, which required five banks to administer $17billion of debt-forgiving modifications to delinquent borrowers.

A model of borrowers as option exercisers provides a clear prediction to the benefits of debt forgiveness modifications from the National Mortgage Settlement, regardless of delinquency or nature of the mortgage: borrowers should (weakly) benefit. In contrast, in a field experiment I run in cooperation with a loan servicer, I find that offers of debt forgiveness can financially hurt borrowers while benefiting the lender. Second mortgage borrowers who have fallen far behind on their repayments are randomly assigned offers to modify their loan in return for completing three payments and legal paperwork. Modification offers increase the fraction of borrowers making a repayment over the next four months from 1% to 14%. Dollar repayments also increase. However, 72% of responding borrowers do not complete the terms of the offer, fail to modify their mortgage, and are worse off than had they not responded. On average, offer recipients do not benefit financially and I hypothesize that this is a result of low financial sophistication.

In a second field experiment, I explore the role of non-strategic factors in borrowers falling behind in their repayments. Simple low-cost treatments, promoting the use of an automated payment system, were administered to over 3,000 contractually performing second mortgage borrowers in the U.S. Treatment increased automation of payments and, three years later, reduced the number of contractually delinquent borrowers by 7%.

These results are, again, inconsistent with standard models of mortgage repayment. The results are, instead, consistent with a view of ongoing monthly mortgage repayments as a task in perspective memory: the ability to remember and complete delayed intentions. I introduce relevant results in the prospective memory literature (Kvavilashvili and Ellis, 1996; McDaniel and Einstein, 2007) to inform this practical view of non-financial obstacles that mortgage borrowers face in initiating prepayments and maintaining repayment.

The two field experiments are, to the best of my knowledge, the first randomized controlled trials in a developed country to study the effects of automating repayment and debt-forgiveness. The natural experiment in regulatory policy I study (the U.S. Savings & Loan crisis) delivers the first causal estimates of regulatory forbearance's relationship to GDP growth. These results argue for and against the relevance of particular models of borrower behavior in different settings, and the novel empirical findings form the primary contribution of my dissertation.

1Regulatory forbearance refers to a regulator refraining from closing down troubled banks and instead temporarily relaxing requirements.