The Role of Family Background on Secondary School Choices

The aim of our work is to measure the impact of social origins on the choice of the academic track in order to allow for consistent cross-country comparisons. We analyze Italy, the Netherlands, and Germany, employing the data from PISA 2003. A substantive problem is that a good measure of individual ability before tracking occurs is not available, thus ability cannot be adequately kept under control. A simple model for school choice is proposed, but the model is not identified with cross-section data. The consequences of unobserved ability are assessed; in the absence of a measure of ability at time the track, the logit regression coefficient of social background is an estimate of the total effect of social background, given by the sum of direct and indirect effects. This is a measure of substantive interest because it represents the total causal effect of social origins on school track. Yet, given that regression coefficients in logit models with independent unobserved heterogeneity are biased towards zero, comparison across countries are difficult; the average sample derivative of the response probability is employed, and it is shown to be a valid alternative measure of the effect of explanatory variables in this context. Our main substantive finding is that the total effect of social background on the choice of the academic track is weaker in the Netherlands and stronger in Germany, with Italy somewhere in between, although, as the German case reveals, when access is regulated by formal restrictions based on ability tests the role of parental background is significantly reduced.