There are two main creditors in commercial real estate: arm’s-length investors and banks. We model commercial mortgage-backed securities (CMBS) as the less informed source of credit. In equilibrium, these investors fund properties with a low probability of distress and banks fund properties that may require renegotiation. We test the model using the 2007-2009 collapse of the CMBS market as a natural experiment, when banks funded both collateral types. Our results show that properties likely to have been securitized were less likely to default or be renegotiated, consistent with the model. This suggests that securitization in this market funds safe collateral.
Using detailed bank balance sheet data we examine how banks respond to a net worth shock. We make use of variation in banks’ loan exposure to industries adversely affected by the oil price declines of 2014 and the implied variation in losses resulting from credit deterioration in those industries. In response to these losses, exposed banks reduced the risk of their balance sheets by shifting away from portfolio lending and towards assets with lower risk weights. Banks tightened credit on corporate lending and on mortgages that they would ultimately hold in their portfolio. However, they expanded credit for mortgages to be securitized. Our results imply that previous work suggesting that banks tighten credit in response to a shock provides only a partial story and is in some ways misleading. It appears that banks respond to a negative shock by de-risking rather than a uniform reduction in lending. In terms of the ultimate impact on borrowers, we find that the shock had only a minimal impact on the overall quantity of loans supplied to firms or households, reflecting substitution to other sources of financing.
We explore the question of optimal aggregation level for stress testing models when the stress test is specified in terms of aggregate macroeconomic variables, but the underlying performance data are available at a loan level. Using standard model performance measures, we ask whether it is better to formulate models at a disaggregated level (“bottom up”) and then aggregate the predictions in order to obtain portfolio loss values or is it better to work directly with aggregated models (“top down”) for portfolio loss forecasts. We study this question for a large portfolio of home equity lines of credit. We conduct model comparisons of loan-level default probability models, county-level models, aggregate portfolio-level models, and hybrid approaches based on portfolio segments such as debt-to-income (DTI) ratios, loan-to-value (LTV) ratios, and FICO risk scores. For each of these aggregation levels we choose the model that fits the data best in terms of in-sample and out-of-sample performance. We then compare winning models across all approaches. We document two main results. First, all the models considered here are capable of fitting our data when given the benefit of using the whole sample period for estimation. Second, in out-of-sample exercises, loan-level models have large forecast errors and underpredict default probability. Average out-of-sample performance is best for portfolio and county-level models. However, for portfolio level, small perturbations in model specification may result in large forecast errors, while county-level models tend to be very robust. We conclude that aggregation level is an important factor to be considered in the stress-testing model design.
We explore the sources of household balance sheet adjustment following the collapse of the housing market in 2006. First, we use microdata from the Federal Reserve Board’s Senior Loan Officer Opinion Survey to document that banks cumulatively tightened consumer lending standards more in counties that experienced a house price boom in the mid-2000s than in nonboom counties. We then use the idea that renters, unlike homeowners, did not experience an adverse wealth shock when the housing market collapsed to examine the relative importance of two explanations for the observed deleveraging and the sluggish pickup in consumption after 2008. First, households may have optimally adjusted to lower wealth by reducing their demand for debt and implicitly, their demand for consumption. Alternatively, banks may have been more reluctant to lend in areas with pronounced real estate declines. Our evidence is consistent with the second explanation. Renters with low risk scores, compared to homeowners in the same markets, reduced their levels of nonmortgage debt and credit card debt more in counties where house prices fell more. The contrast suggests that the observed reductions in aggregate borrowing were more driven by cutbacks in the provision of credit than by a demand-based response to lower housing wealth.
We explore the role of foreclosure inventories in a model of housing supply. The foreclosure variable is necessary to account for the steep and sustained drop in new construction activity following the U.S. housing market bust beginning in 2006. There is modest evidence that local banking conditions play a role in determining housing starts. Even with state-level foreclosures and banking variables in the model, there is a sizeable post-2006 residual common to all states. We argue that, in addition to observable macro and local factors, housing starts in the Great Recession have been weighed down in part by aggregate uncertainty factors
We study the interaction of borrower mortgage prepayment and mortgage delinquency during the period between 2001 and 2010. We show that when house prices flattened and began their subsequent decline, borrowers had increasingly slow prepayments and that this decline in prepayment rates roughly coincided with the sharp increase in their delinquency rates. Low credit score borrowers, in particular, display a pronounced negative correlation between default rates and prepayment rates. Shortfalls of actual prepayment rates from predicted rates based on an estimated prepayment model suggest that, in addition to the effects of declining house prices, tighter lending standards also may have played a role in weak prepayment activity.
We study optimal exercise by mortgage borrowers of the option to default. Also, we use an equilibrium valuation model incorporating default to show how mortgage yields and lender recovery rates on defaulted mortgages depend on initial loan-to-value ratios when borrowers default optimally. The analysis treats both the frictionless case and the case in which borrowers and/or lenders incur deadweight costs upon default. The model is calibrated using data on California mortgages. We find that the model’s principal testable implication for default and mortgage pricing–that default rates and yield spreads will be higher for high loan-to-value mortgages–is borne out empirically.
We evaluate the importance of three different channels for explaining the recent performance of subprime mortgages. First, the riskiness of the subprime borrowing pool may have increased. Second, pockets of regional economic weakness may have helped
push a larger proportion of subprime borrowers into delinquency. Third, for a variety of reasons, the recent history of local house price appreciation and the degree of house price deceleration may have affected delinquency rates on subprime mortgages. While we find
a role for all three candidate explanations, patterns in recent house price appreciation are far and away the best single predictor of delinquency levels and changes in delinquencies. Importantly, after controlling for the current level of house price appreciation, measures
of house price deceleration remain significant predictors of changes in subprime delinquencies. The results point to a possible role for changes in house price expectations for explaining changes in delinquencies.
We develop new techniques to assess the relationship between commercial bank performance and the economic conditions in the markets in which they operate. In the analysis, we allow for heterogeneity in the responses of banks to regional economic conditions. We find a statistically significant relationship between bank performance and shocks to the regional markets in which they operate. We find that region-specific shocks have a significant and persistent effect on the cross-sectional variance of bank performance in the market. That is, shocks affecting average performance of banks in a region also tend to increase the dispersion of their performance. We demonstrate that this effect is due to heterogeneity in the banks’ exposures to their regional economies. Moreover, by allowing for this heterogeneity, we find that systematic responses to regional economic effects are notably more important in explaining the variation in bank performance than suggested by analysis in which responses are constrain to be the same for all banks.
Innovations in the mortgage market since the mid-1990s have effectively reduced a number of financing constraints. Coinciding with these innovations, we document a significant change in the propensity for households to own their homes, as well as substantial increases in the
share of household income devoted to housing. These changes in housing expenditures are especially large for those groups that faced the greatest financial constraints, and are robust across the changing composition of households and their geographic location. We present evidence that young, constrained households may have used newly designed mortgages to
finance their increased expenditures on housing.
Mortgages are one-sided contracts under which the borrower may terminate the contract at any time, while the lender must commit to honoring the terms of the contract throughout its life. There are two aspects to this feature of the contract that are modeled in this paper. The first is that the borrower may choose between buying a house or renting. Given these alternatives, a contract between a household and a lender makes home ownership feasible, and provides insurance to the household against fluctuating rental payments. The second is that once in a contract, the household may terminate the contract by refinancing the future mortgage, and thus enter into a new contract. This option will be exercised whenever a combination of house price appreciation and declines in the mortgage rate is sufficient to increase the ex ante expected lifetime utility from the new versus the old contract.
Published Articles (Refereed Journals and Volumes)
We estimate a structural equilibrium model of the market for automatic teller machines (ATMs) in order to evaluate the implications of regulating ATM surcharges on entry, pricing and welfare. We use data on bank characteristics, potential and actual ATM locations and consumer locations; identify the model parameters with a regression discontinuity design (Iowa banned ATM surcharges while neighboring Minnesota did not); and develop methods to estimate the model without computing equilibria. We find that a ban on ATM surcharges reduces ATM entry by 12 percent, increases consumer welfare by 32 percent and lowers producer profits by 34 percent. Total welfare under either regime is 4 percent lower than the surplus maximizing level. The paper can help shed light on the implications of free entry for differentiated products industries.
We use a unique loan-level dataset to compare portfolio and securitized commercial real estate loans. The paper documents how the types of loans banks choose to hold in their portfolios differ substantially from the types of loans the same banks securitize. Banks tend to hold loans that are “non-standard” in some observable dimension. These loans are riskier and more likely to become delinquent or distressed. Conditional on default, we find that banks are significantly more likely to extend portfolio loans than is the case for securitized loans. Our results suggest that banks have a comparative advantage in funding risky assets with contracts that may require flexibility in the event of distress.
Evidence and Implications of Regime Shifts: Time-Varying Effects of the United States and Japanese Economies on House Prices in Hawaii
Real Estate Economics 41(3), September 2013, 449-480
We show that local house prices may be driven almost entirely by the demands of one identifiable group for several years and then by demands of another group at other times. We present evidence that house prices in Hawaii were subject to such regime shifts. Prices responded to demands associated with U.S. incomes and wealth for most years from 1975 through 2008. For about a decade starting in the middle of the 1980s, after the Japanese yen appreciated dramatically and Japanese housing and stock market wealth soared, however, house prices in Hawaii responded to Japanese incomes and wealth. Estimated models with these regime shifts outperformed conventional, constant-coefficient models. The regime-shifting model helps explain why, when and by how much the volatility and the elasticities of house prices in Hawaii with respect to the incomes and wealth of the United States and Japan varied over time.
We compare the ex ante observable risk characteristics, the default performance, and the pricing of securitized mortgage loans to mortgage loans retained by the original lender. In our sample of loans originated between 2000 and 2007, we find that privately securitized fixed and adjustable-rate mortgages were riskier ex ante than lender-retained loans or loans securitized through the government sponsored agencies. We do not find any evidence of differential loan performance for privately securitized fixed-rate mortgages. We find evidence that privately securitized adjustable-rate mortgages performed worse than retained mortgages, although other observable factors appear to be more economically important determinants of mortgage default. We do not find any evidence of a compensating premium in the loan rates for privately securitized adjustable-rate mortgages.
We develop an overlapping generations model of the real estate market in which search frictions and a debt overhang combine to generate price persistence and illiquidity. Illiquidity stems from heterogeneity in agent real estate valuations. The variance of agent valuations determines how quickly prices adjust following a shock to fundamentals. We examine the predictions of the model by studying depreciation in Japanese land values subsequent to the 1990 stock market crash. Commercial land values fell much more quickly than residential land values. As we would posit that the variance of buyer valuations would be greater for residential real estate than for commercial real estate, this model matches the Japanese experience.
We study the estimation of static games of incomplete information with multiple equilibria. A static game is a generalization of a discrete choice model, such as a multinomial logit or probit, which allows the actions of a group of agents to be interdependent. While the estimator we study is quite flexible, we demonstrate that in most cases it can be easily implemented using standard statistical packages such as STATA. We also propose an algorithm for simulating the model which finds all equilibria to the game. As an application of our estimator, we study recommendations for high technology stocks between 1998-2003. We find that strategic motives, typically ignored in the empirical literature, appear to be an important consideration in the recommendations submitted by equity analysts.
U.S. bank supervisors conduct comprehensive inspections of bank holding companies and assign them a supervisory rating, known as a BOPEC rating prior to 2005, meant to summarize their overall condition. We develop an empirical model of these BOPEC ratings that combines supervisory and securities market information. Securities market variables, such as stock returns and bond yield spreads, improve the model’s in-sample fit. Debt market variables provide more information on supervisory ratings for banks closer to default, while equity market variables provide useful information on ratings for banks further from default. The out-of-sample accuracy of the model with securities market variables is little different from that of a model based on supervisory variables alone. However, the model with securities market information identifies additional ratings downgrades, which are of particular importance to bank supervisors who are concerned with systemic risk and contagion.
Regional Economic Conditions and Aggregate Bank Performance
In Research in Finance, 24, ed. by A. Chen | Bingley, UK: Emerald Group Publishing, 2008. 103-127 | With Daly and Lopez
The idea that a bank’s overall performance is influenced by the regional economy in which it operates is intuitive and broadly consistent with historical bank performance. Yet, micro-level research on the topic has borne mixed results, failing to find a consistent link between various measures of bank performance and regional economic variables. This chapter attempts to reconcile the intuition with the micro-level data by aggregating bank performance, as measured by nonperforming loans, up to the state level. This level of aggregation reduces the influence of idiosyncratic bank effects sufficiently so as to examine more clearly the influence of state-level economic variables. We show that regional variables, such as employment growth and changes in real estate prices, are not particularly useful for predicting changes in bank performance, but that coincident indicators developed to track a state’s gross output are quite useful. We find that these coincident indicators have a statistically significant and economically important influence on state-level, aggregate bank performance. In addition, the coincident indicators potentially contribute to the out-of-sample forecasts of the relative riskiness of state-level bank portfolios, which should be of interest to bankers and bank supervisors.
House Prices and Consumer Welfare
Journal of Urban Economics 58(3), 2005, 474-487 | With Bajari and Benkard
We develop a new approach to measuring changes in consumer welfare due to changes in the price of owner-occupied housing. In our approach, an agent’s welfare adjustment is defined as the transfer required to keep expected discounted utility constant given a change in current home prices. We demonstrate that, up to a first-order approximation, there is no aggregate change in welfare due to price increases in the existing housing stock. This follows from a simple market clearing condition where capital gains experienced by sellers are exactly offset by welfare losses to buyers. Welfare losses can occur, however, from price increases in new construction and renovations. We show that this result holds (approximately) even in a model that accounts for changes in consumption and investment plans prompted by current price changes. We estimate the welfare cost of house price appreciation to be an average of $127 per household per year over the 1984-1998 period.
We examine whether equity market variables, such as stock returns and
equity-based default probabilities, are useful to U.S. bank supervisors for assessing the condition of domestic bank holding companies. We develop a model of supervisory ratings that combines supervisory and equity market information. We find that the model’s forecasts anticipate supervisory rating changes by up to four quarters. Relative to simply using supervisory variables, the inclusion of equity market variables in the model does not improve forecast accuracy. However, we argue that equity market information should still be useful for forecasting supervisory ratings and should be incorporated into supervisory monitoring models.
Forecasting Supervisory Ratings Using Securities Market Information
In Corporate Governance: Implications for Financial Services Firms. The 39th Annual Conference on Bank Structure and Financial Services Firms | Chicago: FRB Chicago, 2003 | With Lopez
Approximately once a year, bank supervisors in the United States conduct
a comprehensive on-site inspection of a bank holding company and assign
it a supervisory rating meant to summarize its overall condition. We develop an empirical forecasting model of these ratings that combines
accounting and financial market data. We find that securities market variables, such as stock returns and changes in bond yield spreads, improve the model’s in-sample fit. Both equity and debt market variables appear to be useful for explaining upgrades and downgrades. We conclude that stock and bond market investors possess different but complementary information about bank holding company condition. In an out-of-sample forecasting exercise, we find that the forecast accuracy of the model with both equity and debt variables is little different from the accuracy of a model based on accounting and lagged supervisory data alone.
Equilibrium Valuation of Illiquid Assets
Economic Theory 19(2), January 2002, 223-242 | With LeRoy
We develop an equilibrium model of illiquid asset valuation based on search and matching. We propose several measures of illiquidity and show how these measures behave. We also show that the equilibrium amount of search may be less than, equal to, or greater than the amount of search that is socially optimal. Finally, we show that excess returns on illiquid assets are fair games if returns are defined to include the appropriate shadow prices.
A Theory of Liquidity in Residential Real Estate Markets
Journal of Urban Economics 49(1), January 2001, 32-53
A “hot” real estate market is one where prices are rising, average selling times are short, and the volume of transactions is higher than the norm. “Cold” markets have the opposite characteristics–prices are falling, liquidity is poor, and volume is low. This paper provides a theory to match these observed correlations. I show that liquidity can be good while prices are high because the opportunity cost of failing to complete a transaction is high for both buyers and sellers. I also show how state varying liquidity depends on the absence of smoothly functioning rental markets.