Imputing Borrower Heterogeneity and Dynamics in Mortgage Default Models

The determinants of mortgage default have been an area of rising interest since the Great Recession of 2008. One of the distinguishing features of mortgage default analysis is that predictor variables are often only recorded at origination. However, variables such as credit scores and income vary over time. Our focus is to develop a ridge regression model to impute the dynamics of time-varying predictors and to capture unobservable borrower heterogeneity. After allowing for imputed dynamics and borrower heterogeneity, we find increased importance for full documentation relative to the loan-to-value ratio.