DECOMPOSING THE GAP IN SCHOOL ACHIEVEMENT BETWEEN FINLAND AND ROMANIA '" SOME METHODOLOGICAL ASPECTS

This paper analyzes the advantages and shortcomings of the Blinder-Oaxaca decomposition. Using PISA data for Finland and Romania, we focus on the drawbacks of the detailed decomposition, when the explanatory variables are categorial. From the best of our knowledge, this kind of analysis is performed for the first time using PISA data.We show that, using covariates which are categorial, the partial characteristics effects can be different when we use different reference categories of the respective variable. However, the overall characteristics effect of a categorical variable does not depend on the omitted category. The more critical aspect of the interpretation of detailed decomposition, when explanatory variables are categorical, regards the unexplained part of the gap. As we empirically show, the both components of the unexplained part are sensitive to choices of the reference category. These aspects should be taken into account when we perform detailed decompositions with categorial variables.