Access and returns to education in Kazakhstan

Abstract

This thesis explores the heavily under-researched topic of returns to education in Kazakhstan and how access to higher education a ects them. It starts with an analysis of the returns to education, as estimated by applying pseudo-panel techniques to cross-sectional data for 2002-2016 from the Kazakhstani national statistics. The returns are found to be relatively high and internationally comparable, though they decrease over the period considered. The cohort e ect estimated with Mundlak correlated random e ects model is found to be negative, suggesting the business cycle impact interpretation: those cohorts whose school-leaving age fell during the 1990s recession, when there was a lack of jobs and increased access to tertiary education, ended up with more schooling, but lower future earnings. The second and third empirical chapters employ administrative data on 90,329 recent university graduates. I start with an analysis of the returns to attending more selective national universities as opposed to other public HEIs. National universities bene t from relatively higher public funding, administrative support and higher tuition fees. In 2012, they were forced to increase the minimum admission test score in order to improve the quality of their student intake. With a fuzzy regression discontinuity design, this chapter revealed no returns premia was gained by attending a national university, at least during the rst year in employment and for the rst a ected cohort. Alternative explanations for this could arise from problems with the methodology or data constrains. The nal chapter examines the e ects of tightening access to higher education, which occurred in 2012, on programme-level enrolment, student quality and the a ected cohort's labour market outcomes. Although the entry test is uni ed, I expect the impact of tightening entry requirements to depend on the proximity of the pre-treatment quality of the student intake to the test cut-o point. I exploit this variation in treatment intensity to apply di erence-in-di erences estimation. I found that the enrolment per programme has decreased for the worse-quality programmes relative to the better-quality ones only at public universities, which is likely due to di erences in the subject composition and the intervening e ect of the increased test cut-o point at the national universities. However, the student quality was found to decrease in relatively better programmes for all universities. This possible long-term trend might re ect adverse demographic conditions and increased competition, with HEIs maximising pro ts through enrolment maximisation strategies due to limited alternative sources of funding. Regardless, it does not a ect the graduates' further labour market returns, at least during their rst year in employment.