[Clyde Schechter]

Yes, I see the problem. It is not clear to me how to fix it, and I am realizing that I probably don't understand what you are trying to do here. The -snapspan- problem arises because your companies go in and out of exporting mode, so that their duration as an exporter does not strictly increase from one observation to the next, which leads to multiple observations with the same duration, but different covariates. Also, if I understand the situation correctly, a company is only at risk for experiencing a failure event while in export mode. And you envision that the values of the covariates experienced while not in export mode nevertheless might influence the risk of "death" later on (hence, the non-export years cannot simply be dropped). Is that right? If so, I think we have a situation that is beyond the scope of Stata's survival analysis commands. I don't think they can handle situations were entities repeatedly go in and out of the at-risk condition (as opposed to simply gaps in the observation period). If I am wrong about this, hopefully, somebody who knows better well step in.