Count Data Model of Work-related Training: a study of young men in Britain

This paper estimates models of training based on count data, in which the dependent variable takes only non-negative integer values corresponding to the number of work-related training courses occurring in the interval 1981 to 1991. The data set is the National Child Development Study. The raw data indicate substantial over-dispersion, and tests of the negative binomial model against the Poisson model indicate that the former better describes the data. Over half of the young men had no work-related training at all over the period 1981-1991, and there is evidence of a "skills-segmented" labour market. The estimates indicate strong complementarities between past general education and training, suggesting that reliance on job-related training to increase the level of skills of the British workforce will result in an increase in the skills to the already-educated, but will not improve the skills of individuals entering the labour market with a low level of education.