THE POTENTIAL OF QUANTILE REGRESSION TO COMPARATIVE ANALYSIS OF HETEROGENEITY AND ACHIEVEMENT IN SCHOOLS ACROSS COUNTRIES

Montreal

Congreso; 55th Annual Conference of the Comparative and International Education Society (CIES); 2011

CIES

Comparative analysis of achievement in schools across countries makes extensive use of methods, such as HLM, in which a selected set of factors are said to explain ?on average? the resulting test score, measured at the mean. Unresolved questions are: do test score responses to these same factors have similar strength at the center as well as in the tails of the score distribution? Do the explanatory power of the covariates change along the distribution of test scores in any predictable direction? In what sense responses of students in the upper tail of the scores distribution are similar or different from that of students in the lower tail? Does the strength of responses vary across countries? Quantile regression analysis helps to explore precisely these questions by testing the potential existence of heterogeneous responses along the score distribution and across countries. This paper makes use of SERCE data for Argentina, Brazil, Chile, Paraguay and Uruguay to investigate heterogeneity in the relationship between students? performance in math and covariates that reflect personal traits and family attributes. Two school variables are introduced as control variables. If there is substantial heterogeneity in the relationship between the student performance and his personal and family characteristics, those individuals in charge of compensatory education policy design would benefit from investigating the size of these effects at different sections of the distribution of the dependent variable. The study of heterogeneous responses has the potential to help education to focalize programs towards empowerment and agency of the disenfranchised young.