Instrumental Variable Estimation for Duration Data: A Reappraisal of the Illinois Reemployment Bonus Experiment

We propose an Instrumental Variable method for Generalised Accelerated Failure Time (GAFT) models that adjust for possible endogeneity of the intervention of interest, without suffering the problems of the intention-to-treat method. We develop an estimatiom procedure that collapses to the linear Rank Estimation procedure for GAFT models with only exogenous variables This is accomplished by using a model that transforms the duration such that for the population parameters, this transformed duration is independent of the instrument. The estimator is used to re-analyse data from the Illinois reemploymnet Bonus experiment