Our study uses a multinomial logit model to analyze the concurrent termination experience of adjustable-rate and fixed-rate mortgages. A new set of ARM-specific interactive determinants expands the conventional FRM specification to isolate the unique termination behavior of ARMs. We find that expected rate adjustments and large lifetime caps are positively related to ARM termination probabilities while long adjustment frequencies are inversely related. Caps, both periodic and lifetime, have a secondary, inverse effect on termination probabilities when interest-rate movements exceed cap limits. The model also shows that interest-rate expectations affect FRM terminations more strongly than ARM terminations.