This article analyzes mortgage terminations using a national individual loan data set for the 1986–1992 period. The standard option-choice-theoretic framework is supplemented with variables to proxy for non-option-related termination determinants. Separate multinominal logit models are estimated for three mortgage types: 30-year FRMs, 15-year FRMs, and 30-year ARMs. The results indicate substantial differences in the response of the mortgage types to variables included in the model. FRM15 prepayments are the most responsive to prepayment option values; FRM30 prepayments are less responsive to option values and are dirven by local area housing market and economic conditions; ARM prepayment rates are higher but defeult rates are lower relative to the FRMs. A noteworthy finding is that teaser discounts reduce the likelihood of ARM defaults.