Abstract
The study calculates the probability of success at the centralized university entrance examinations for applicants from different districts of the capital city of Georgia, Tbilisi. Through linear regression, binary logistic regression and multinomial logistic regression analyses of a government-collected census database of 10,000 entrants, the study examines the extent to which residential or school location, gender, school type or age of the applicants impact their probability of success. Because their scores at the exam determine their admission and the value of the study voucher that they receive, by uncovering major differences in the chances of success among applicants from different socioeconomic, educational and cultural backgrounds, the study brings our attention to the possible efficiency- and equity-related threats of merit-based vouchers and centralized merit-based entrance examinations. A 3-stage analysis reveals a clear divide between the Western and Eastern parts of the capital city in terms of its affluence, average per-student school funding and outcomes at the exams. Students from the West, females, younger applicants and private school graduates have considerably higher scores and probability of success in obtaining vouchers. For example, linear regression analysis revealed that when keeping school status, gender and age category constant, applicants from the Western schools score 3.8 out of 80 points higher in the General Abilities Test. According to the results of logistic regression analysis, when keeping age category, gender and the status of the school constant, those studying in the West part of the city are 1.68 times more likely to obtain a voucher than those from the East. When keeping age category constant, female applicants from Western private schools have a 48% chance of obtaining a voucher while the probability of male applicants from Eastern public schools is almost 3 times lower – 17%. The study also suggests that the school location of the applicant has a slightly larger influence on outcomes than their residential origin.