When multiple durations are generated by a single unit, they may be related
in a way that is not fully captured by the regressors. The omitted unit-sp
ecific variables might vary over the durations, They might also be correlat
ed with the variables in the regression component. We propose an estimator
that responds to these concerns and develop a specification test for detect
ing unobserved unit-specific effects, Data from Malaysia reveal that concen
tration of child mortality in some families is imperfectly explained by obs
erved explanatory variables, and that failure to control for unobserved het
erogeneity seriously biases the parameter estimates. (C) 1999 Elsevier Scie
nce S.A. All rights reserved, JEL classification: C14; C41; J13.