Extension of the EM test to endogenous censoring

Extend the EM test for unobserved heterogeneity in censored Weibull duration models to settings with endogenous censoring, using additional structure such as instruments or partial-identification methods.

Background

The proposed EM test assumes that the duration and censoring time are conditionally independent given the observed covariates. The conclusion explains that under endogenous censoring, the duration distribution is not point identified without further structure.

The authors note that existing work uses instruments or identified sets in such settings and explicitly leave extending the EM test in that direction for future work. This is therefore an unresolved methodological extension rather than a result established in the paper.