Effects of higher education and the role of admission selection
Sammanfattning: Study I Using Admission Selection Rules to Estimate the University Earnings Premium - A Regression-Discontinuity DesignThis study estimates the earnings premiums for university admission and graduation, using unique data from the admission selection process. Applicants and non-applicants for university in Sweden in 1982 are studied in terms of earnings in 1996. Admitted applicants are compared with four groups of non-admitted individuals. The choice of comparison group proves crucial when estimating the admission earnings premium. A comparison between admitted applicants and a random sample of non-applicants yields a significantly higher admission effect than a similar comparison using withdrawals or screen-outs. When the admission selection variables are added to the model, the effect of admission is completely eliminated. When admitted applicants are decomposed into no-shows, dropouts and completers, the same pattern emerges as in the case of the effect of admission. The university completion premium falls from 18 per cent when non-applicants provide the comparison group to 5 per cent when screen-outs are used instead and admission selection variables are controlled for. Hence, two thirds of the estimated effects are attributable to unobserved differences, while only one third can be seen as a causal effect. In order to find unbiased estimates of the effects of different programmes or interventions, information about all the selection mechanisms seems to be crucial.Study II Estimating the Return to Higher Education Using a Social ExperimentApplicants for a university education in Sweden who possess equal qualifications have been subject to selection by lottery, i.e. there has been a social experiment at the margin. This study uses a unique data set of over 1,600 individuals who in the autumn of 1982 were subject to randomized admission to university. Annual earnings between 1982 and 1996 are used to compute lifetime earnings. To adjust for dropout and substitution bias in the estimated treatment effects, two new estimators are proposed. They allow for non-zero effects of partial treatment and for alternative treatment for all treatment group members.The choice of estimator proves crucial. If no adjustments are made, only about half the estimated returns to university training are positive. Accounting for dropout bias raises the estimates. The result is not sensitive to alternative identifying assumptions. When the earnings differences between treatment group and control group members are adjusted for both dropout bias and substitution bias, the estimated effects becomes much higher. Allowing all treatment group members to receive alternative treatment reduces the estimates substantially. The estimates suggest that the internal rate of return to completed university education is about 20 per cent, significantly different from zero.Study IIIDoes Pre-University Background Matter? This study analyses the completion probability and the effects of university on labour market performance for groups with different pre-university background. The data includes detailed information on the admission selection process for most applicants to university education in Sweden in 1982. The results reveal considerable heterogeneity in the probability of obtaining a degree and the effect of university studies on labour market performance. Applicants with long upper secondary schooling are more likely to graduate from university than other students are. The best indicator of degree completion is grade point averages from upper secondary school and the field of educational interest. It is interesting to note that age is negatively related to student performance.In terms of employment and earnings, returning adults have more to gain from being admitted to university or receiving a degree there than students with any other pre-university background. Hence, the pay-off for allowing individuals a 'second chance' seems substantial. Another finding is that completers generally do better than no-shows, who in turn do better than dropouts. Further, the effects of attending university on employment and earnings are small for all groups, particularly for students with short upper secondary schooling. The return to a completed programme for students with long upper secondary schooling is estimated at 5 per cent.
Denna avhandling är EVENTUELLT nedladdningsbar som PDF. Kolla denna länk för att se om den går att ladda ner.