Lohnstrukturverzerrung und betriebliche Ausbildung : empirische Analyse des Acemoglu-Pischke-Modells mit Daten des IAB-Betriebspanels (Wage structure bias and company training : empirical analysis of the Acemoglu-Pischke model with data from the IAB establishment panel)

"The paper examines the determinants of firms' decisions regarding training, which follow from the Acemoglu-Pischke model (APM), for their empirical content for western German firms. For this the data from the IAB establishment panel are used. Contrary to Becker's ideas of the standard model of human capital theory (1962), according to which employers do not contribute under any circumstances to financing general human capital investments, in reality a financial involvement on the part of the firms in the initial vocational training can actually be observed. Acemoglu/Pischke (1999a, 1999b) try to resolve this contradiction with the aid of a formal-analytical model. They show that employers can by all means have an incentive to invest in human capital even if the qualifications imparted are entirely general in nature. The crucial point here is the existence of wage structure biases in the sense of wage compression. In this context the two authors identify mobility costs, asymmetric information, components of firm-specific human capital, efficiency wages and wage floors, such as e.g. collectively agreed wages, as possible causes of wage structure biases. A censored regression model is used as an econometric model, which is estimated firstly with the common Tobit ML estimator and secondly with the censored least absolute deviations estimator (CLAD). The CLAD approach has the advantage over a Tobit estimate that the estimates are consistent even with heteroscedastic error factors and deviations from normal distribution. The estimate results broadly confirm the implications of the APM. However, it can also be observed that obviously not all market imperfections and institutions which lead to wage structure biases are of equal importance for firms' decisions regarding training. Whilst the model estimates support e.g. an interpretation with regard to a measurable influence of mobility costs, wage floors and the complementarity between general and specific qualifications, in particular the finding on poaching externalities seems more to limit the relevance of asymmetric information. In contrast, those findings that confirm the considerations of Acemoglu/Pischke (1998) on the high (low) quit-low (high) training equilibrium are in agreement with the APM." (Author's abstract, IAB-Doku) ((en))