each individual. Differencing the identically censored observations for a fixed individual eliminates the fixed effect, just as time differences eliminate fixed effects in the standard linear panel data model. In fact, this approach yields consistent estimates even when correctly specified maximum likelihood would not.