DP1582 Modelling Work-Related Training and Training Effects Using Count Data Techniques
|Author(s):||Wiji Arulampalam, Alison L Booth, Peter Elias|
|Publication Date:||February 1997|
|Keyword(s):||Count data models, Hurdle, Selectivity bias, Training endogeneity, Wages Growth|
|JEL(s):||C25, I21, J24, J30, J42|
|Programme Areas:||Human Resources|
|Link to this Page:||cepr.org/active/publications/discussion_papers/dp.php?dpno=1582|
This paper estimates the determinants of the number of work-related training courses, and their impact on expected wages growth, using longitudinal data from the British National Child Development Study. The analysis covers a crucial decade in the working lives of a cohort of young men ? from the age of 23 to 33. We use hurdle negative binomial models to estimate the number of work-related training events. This approach allows us to account for the fact that half of all sample members experienced no work-related training over the period from 1981 to 1991. We then estimate a wages growth model where the returns from the first training experience are allowed to differ from subsequent training experiences. The results generated from the hurdle count model are used to control for training endogeneity in the wages growth equation. This has not been done before in the training literature. The sensitivity of the wage growth estimates to alternative modelling strategies is also examined. Since we find a strong correlation between education and subsequent training experiences, we experiment with estimating the joint impact of previous education and subsequent training on wages growth, in order to tease out the combined effect of these variables. We find that the biggest returns to training are to highly educated men.