Determinants of Housing Mortgage Prepayment in Korea: Using a Loan-Level Dataset

Abstract

This study empirically investigates housing mortgage prepayment behavior decisions in Korea. Using the loan-level data from 2013 to 2014, a Cox proportional hazard model and a shared frailty model which controls the effect of unobserved heterogeneity among local markets are estimated. The results show that there exists significant heterogeneity among the local markets. Ignoring this heterogeneity results in errors in estimating the prepayment behavior of borrowers. Estimation results show that the higher expected housing price change, the lower the hazard rate of prepayment and the higher collateral value change, the higher the hazard rate of prepayment. The sale for capital gains as well as refinance by option value may also cause prepayment. The greater the degree where the collateral value declines, the higher the hazard rate of prepayment. It means that distressed prepayment is an alternative when the collateral value declines. Estimation results also indicate that loan characteristics like LTV, loan amount, maturity period, grace period, and borrower characteristics like borrower’s age, income, credit rating are determinants of prepayment. Especially, low credit serves as a constraint on refinance.