*Names in bold indicate Presenter
This analysis directly informs this policy need, with a unique dataset of more than 30,000 seniors counseled for reverse mortgages between 2006 and 2011, 60 percent of whom took out HECM loans. Our data includes comprehensive financial and credit report attributes, not typically available in analyses of reverse mortgage borrowers. In partnership with HUD, we link this data to loan-level HECM data, containing information on originations, withdrawals as well as termination outcomes. Our analysis builds on a small body of existing literature modeling reverse mortgage terminations (e.g. Szymanoski, Enriquez, and DiVenti 2007; Bishop and Shan 2008, Shan 2009). However, previous literature does not incorporate the probability of tax or insurance delinquency, as this data is not publicly available and has only recently been collected by HUD. Further, previous literature lacks many important characteristics of borrowers, including income, debt and credit report attributes, that we include in our analysis.
Our analysis allows us to isolate specific factors at the time of closing that predict termination outcomes of HECM borrowers, thereby improving upon the existing literature and informing the current debate about necessary policy changes. Using a three tiered nested logistic regression model, we estimate the likelihood of tax and insurance default and loan prepayment, conditioned on (1) selection of a HECM loan; and (2) whether or not the borrower withdraws the majority of available proceeds at the time of closing. Preliminary analyses by HUD suggests that borrowers who withdraw the majority of their available equity up front are more likely to default; however, it is difficult to separate out these effects from risk characteristics of borrowers. Further, policy changes to the program in 2009 created a new product that required a full draw of available equity at closing, potentially increasing default risk. The proportion of borrowers withdrawing the majority of their available equity up-front increased from about 50 percent prior to 2009, to more than 70 percent in 2010 and 2011. We employ regression discontinuity design to explore the impact of this policy change on the probabilities of tax and insurance default, conditioned on whether or not the borrower withdraws the majority of proceeds at closing.