House Price Expectations And Mortgage Default Decisions
WISE & SOE, Xiamen University
November 25, 2015
Michele Ragazzi's Giano Capital returned 1.9% for March, taking the fund's year-to-date performance to 1.7%. Since its inception, Ragazzi's flagship fund has produced a compound annual return of 7.8%. According to a copy of the €10 million fund's March update, a copy of which ValueWalk has been able to review, Giano's most significant investment at Read More
If households default strategically, their future house price expectations should affect their present mortgage default decisions. Merging the Freddie Mac loan performance dataset with the Michigan Survey of Consumers data containing households’ subjective expectations, this paper provides direct empirical evidence that low (high) house price expectations will increase (decrease) default probabilities. This effect is greater for loans on investment properties than for loans on owner-occupied properties; is greater in nonrecourse states than in recourse states. The out-of-sample fit shows that incorporating subjective expectations can potentially improve default-risk forecasting accuracy by 1.88% overall and by 4.36% for loans on investment properties.
House Price Expectations And Mortgage Default Decisions – Introduction
Numerous studies have provided the theoretical foundations and empirical evidence for a number of factors that affect the likelihood of mortgage defaults by households. Based on the option theory for mortgage default and prepayment developed by Kau, Keenan, Muller and Epperson (1995) as well as the empirical evidence provided by Deng, Quigley and Order (2000) and others, negative home equity (positive default option value) has been shown to contribute to strategic default (ruthless default). There are other factors that contribute to illiquidity-triggered default, including the debt-to-income ratio, assets, and accessibility to other loans.
Theoretically, if households strategically choose to default, their future house price expectations should also affect their current mortgage default decisions. If households are forward-looking when making decisions, they are not only considering the payoff in the current period, but also taking into account the future. Suppose in the current period house prices are low and certain households’ houses are underwater, which gives the households a positive financial incentive to default. Given that a default will incur a large amount of fixed costs, including the loss of their homes, moving, and ruining their credit rating and reputation, if the forward-looking households forecast that the house prices will rebound in the future, they may choose not to default. On the other hand, if they forecast that the house prices will drop further in the future, they will be more likely to choose to default; given that they will default finally, defaulting earlier to avoid additional payments is better than defaulting later.
Under the assumption that households are forward-looking when making mortgage default decisions, several types of research have been conducted. First, Campbell and Cocco (2015) developed a theoretical model in which households solve an intertemporal utility maximizing problem under a liquidity constraint for mortgage default decisions. Second, using macro data, Corbae and Quintin (2014) calibrated a dynamic general equilibrium model with heterogeneous households making decisions on mortgage selection and default. Third, using loan-level data, with different focuses, Carranza and Navarro (2010), Zhang (2010), Laufer (2011), Bajari, Chu, Nekipelov and Park (2013), and Ma (2014) estimated dynamic choice models for household mortgage default behaviors, assuming that households’ expectation formations for future house prices are based on realized house prices in the past and follow firstorder auto-regression (AR(1)) processes.
However, to the best of my knowledge, no previous study has provided direct empirical evidence that households’ house price expectations affect their mortgage default decisions, partially due to the lack of data about households’ subjective expectations for future house prices. In this paper, I extract the households’ subjective house price expectation data from the Michigan Survey of Consumers, which asks approximately 500 respondents about their subjective expectations of certain economic variables each month, including expectations of house price appreciation rates over the next year. I then merge these data with the Freddie Mac mortgage loan performance data and estimate multinomial logit models with three outcomes (default, prepay, and continue to pay).
The empirical results indicate that after controlling the current home equity status, current economic conditions, as well as households’ expectations of other economic variables, expecting a high (low) house price growth rate in the future will decrease (increase) the likelihood of default in the current period. Moreover, the effect of house price expectations on default likelihood is greater for loans on investment properties than for loans on owner-occupied properties; the effect is greater in nonrecourse states than in recourse states. Incorporating households’ subjective expectations of future house price appreciations can potentially improve the accuracy of mortgage default risk predictions by 1.88% for all the loans, and by 4.36% for loans on investment properties. Because banks hold a substantial amount of mortgage loans, this improvement on the prediction accuracy can increase bank profits by a huge dollar amount.
See full PDF below.