Stima dei parametri della funzione di ripartizione dei periodi trascorsi: il caso dell’anzianità aziendale

The duration data sets used to study employment or unemployment often come from samples of individuals at a point in time (Cross Population Survey, CPS) and they contains the lengths of ongoing spells of employment or unemployment for sample subjects who are currently employed or unemployed. The observations are right-censored and termed spells in progress or incomplete spells. The analysis of the raw data involves a length-biased estimate of spell durations. In this paper, we examine the procedure to analyze a data set which includes only censored or incomplete spells when the full lengths of spells, conditioned on their being in progress at the time of survey, are Weibull distributed. Particularly, the methodology of estimation of the parameters is described for both the distribution function of spell durations and the proportional hazard model which include a set of explanatory variables (personal and job characteristics, local labor demand). Then, the methods are applied to the lengths of ongoing spells of employment for both employed workers and self-employed workers.