50th Congress of the European Regional Science Association: "Sustainable Regional Growth and Development in the Creative Knowledge Economy", 19-23 August 2010, Jönköping, Sweden
This paper lies in the tradition of decomposition analysis of wage differentials based on the model set forth in Blinder (1973) and Oaxaca (1973), and aims to measure labour market discrimination against Roma in South East European countries (Albania, Bulgaria, Croatia, Serbia and Kosovo). We use microdata from 2004 UNDP household survey and a Bayesian approach, proposed by Keith and LeSage (2003), for the decomposition analysis of wage differentials. Statistical inference for both discrimination and characteristics effects estimates are based on Markov Chain Monte Carlo (MCMC) estimation. Variance estimates derived from this method of estimation are known to reflect the true posterior variance when a sufficiently large sample of MCMC draws is carried out. The results provide clear evidence for labour market discrimination against Roma in Albania and Kosovo, but not so in Bulgaria, Croatia, and Serbia. Nevertheless, there are significant differences in how individual characteristics are valued between Roma and non-Roma.