Abstract
In Asian Countries or some countries with high saving rates, partial prepayment, curtailment, is a dominant form of prepayment behavior. When borrowers have extra money more than the basic living needs, they usually use that additional amount of money to repay the principle payment more than schedule payment in order to curtail the loan balance which will further reduce the interest payment. Based on mortgage loan-level data extracted from a local commercial bank in Taiwan, we find the curtailment effect is significant. Therefore, we incorporate the curtailment factor as one of the explanatory variables in our analysis. More specifically, the aging patterns of prepayment rate and default rate are different. The patterns of prepayment rates are sigmoid curves; whereas the default rates initially increase progressively and then decrease thereafter which is similar to Gamma distribution. Our approach is first to use Begg and Gary (1984) their separate binomial logit model with different age functions to estimate conditional prepayment and default rates. Then the parameter estimates must be combined to compute multinomial probabilities to account for mortgage competing risks. Doing so, we can capture the borrowers' payment behavior and accurately estimate the prepayment and default rates.