The aim of this paper is to select a duration model to assess whether the
unemployment duration for Colombian people under 26 years old has decay
since the application of a law that incentives hiring first time or without legal
contract employees. Parametrical and non-parametrical models of duration are
compared to investigate empirically the sensitivity of unemployment duration
regression estimates to distributional assumptions. It is found that parametric
results are sensitive to the choice of the distribution used to model unobservable
component. Less duration of unemployment is explained by another variables
different from the application of the first employment law.