This paper uses microlevel data to examine recent prepayment performance of adjustable rate mortgages (ARMs) employing the competing risk methodology developed by Deng, Quigley and Van Order (2000). We find support for the teaser rate and adjustment date effects implied by the theoretical model of Kau et al. (1993). In addition, we find that teased ARMs bear prepayment risk related to their discount, contrary to results reported by VanderHoff (1996) and Green and Shilling (1997). Finally, and contrary to the usual finding for fixed-rate mortgages, we find that loan age has a negative effect on prepayment risk for ARMs, consistent with the phenomenon that borrowers with high mobility and/or propensity to refinance exit the pool early.