Job (Creation and Destruction) Hazards

In firm level applications, empirical duration analysis is an exercise in
misspecification. The implicit model of the firm does not contain time as a
state variable – the firm is an infinitely lived entity – which implies
that in empirical duration analysis time is significant because there is
some state variable we cannot observe or measure. In fact, duration analysis
is a tool for inference about these unobservables. So, what can we conclude
when we observe upward or downward sloping hazards of employment adjustment?
How good is this inference? We turn the exercise on its head and start from
the model to see what creates different hazards. We find that upward sloping
hazards are notoriously hard to generate.