A decomposition of training probabilities

The determinants of workers' training probabilities are analysed. A distinction is made between the unconditional probability that a worker is employed by a firm that provides any training opportunities at all, and the conditional probability that the worker receives training given that s/he works for a firm that provides training. For this analysis a censored version of the bivariate probit model is applied. The results indicate that establishment size, industry dummies, type of schooling, age and experience affect only the unconditional probability, whereas the type of labour contract, working hours and job level affect only the conditional probability. Years of formal schooling affect both probabilities; more highly educated workers are both more likely to work for training-providing firms and, given that they work for such a firm, to be selected for training programmes.