A Mortgage Prepayment Study Using Survival Analysis and Time-Varying Determinants

Abstract

A critical role in the functioning of housing and financial markets, involving a complex interaction between various factors, is played by mortgage prepayment and its behavior. This study analyzes mortgage prepayment behavior in the United States using loan-level data from the Freddie Mac single-family dataset covering 1999-2023. We apply survival analysis techniques like Cox proportional hazard model for time-varying covariates, as well as Aalen-Johansen, Kaplan-Meier estimators, and the baseline cumulative hazard function, to estimate the effect of borrower characteristics and macroeconomic variables on prepayment hazard over time. Several important characteristics influencing prepayment behavior have been identified: the credit score, a borrower-specific feature; macroeconomic variables including housing price appreciation, interest rate spread, inflation, and unemployment. A statistically significant interaction between borrower credit score segments, mortgage interest rates, and government refinancing or stimulus policies has been found. These findings demonstrate the dynamic nature of mortgage prepayment behavior, highlighting the importance of both individual borrower characteristics like the credit score and shifting economic conditions in shaping prepayment risk over time.