Authors：Xiaofeng Lv，XuRan Sun，Yue Lu，and Rui Li

Abstract：Treatment assignment in the survival literature is often assumed to be allocated simultaneously and independently of prospective treatment gains. This paper relaxes these restrictions by introducing dynamic treatment assignment for survival data in a regression discontinuity design. Conditional on a pretreatment duration, we identify two survival functions of the remaining potential durations under treatment and no treatment. Conditional treatment effects can be identified by the difference between the integrals of the two functions, and we aggregate conditional treatment effects over pretreatment durations to identify unconditional ones. Accordingly, nonparametric estimates are proposed.

Key Words：Nonparametric identification;Treatment effects;Regression discontinuity;Survival analysis;Dynamic treatment assignment