Improving scholarship assignment using approximate dynamic programming: A Chilean case study

Abstract

A solution approach is proposed for a problem of assigning scholarships under budget constraints and uncertainty regarding the duration of previously awarded scholarships as determined by future renewals. The problem objective is to maximize the sum of the scores of accepted applicants subject to two conditions: (1) no applicant can be assigned a scholarship until assignments have been made to all applicants with higher scores; and (2) the annual budget allocation must first cover all renewals of previous awards. The approach is built around the formulation and approximate solution of a Markov decision process that provides a systematic method for identifying scholarship assignment policies which make efficient use of the annual scholarship budget. The benefits of the approach are analysed in a case study involving a scholarship offered annually in Chile that compares via simulation the solutions generated by various alternative assignment procedures with those of the proposed model. The results suggest that under the latter, 7% to 9% more students would be awarded scholarships each year without increasing the yearly budget.