Econometric duration analysis (Q788453)

The paper considers formulation and estimation of continuous time econometric duration models: single and multiple spell; a general time inhomogeneous multiple spell. Limited size of samples available in economics, abundance of candidate observed explanatory variables make standard non-parametric procedures used in biostatistics of dubious value. It is necessary to control for observed and unobserved explanatory variables to avoid biasing inference about underlying duration distributions. Criteria are presented for inferring qualitative properties of conditional hazard and distributions of unobservables from raw duration data sampled in time homogeneous environments (no parametric structure need be assumed.) It is noted that current econometric practice overparametrizes duration models. Yet, it is possible to consistently estimate the distribution of unobservables non-parametrically. For single spell data it is impossible to separate the effect of duration dependence from the effect of time inhomogeneity by a fully nonparametric procedure. Next, it is shown how sampling schemes distort the functional forms of sample duration distributions. New formulae for the densities of commonly used measures are produced for duration models with unobservables in time inhomogeneous environments.