The development of a student graduation prediction model as a funding allocation tool

Kloppers, J'Mari

The development of a student graduation prediction model as a funding allocation tool

Thesis / Dissertation

2026

Publisher

University of Cape Town

Department

School of Economics

Faculty

Faculty of Commerce

Abstract

Access to higher education in South Africa is limited due to the high cost associated with obtaining a university degree. Many students are dependent on financial aid such as grants, bursaries or student loans. Lenders, however, face challenges. Institutions responsible for the allocation of grants and bursaries are confronted with a demand that exceeds available funds. At the same time, prospective students often lack credit histories, a metric traditionally leveraged to assess creditworthiness, forcing banks to rely on other indicators, like a relative or guardian's wealth, to allocate student loans, which excludes students from disadvantaged backgrounds. Identifying students likely to complete their degree and enter the workforce would enable lenders to allocate loans more effectively, based on repayment potential. This approach will reduce the reliance on existing wealth when applying for funding. Furthermore, expected student performance can inform bursary and grant allocations, ensuring funds are allocated to the most promising students. The purpose of this study is to leverage different statistical models to evaluate whether students' academic performance for a particular academic year can be predicted given features such as gender, race, institute of tertiary study, degree type and current academic results. This study is based on data provided by StudyTrust, a South African student bursary management and mentorship program, which encompasses students from various South African tertiary institutions and degree programs. While predicting firstyear academic performance using Grade 12 data proved challenging, the models showed promising results for predicting subsequent academic years' results. Further results highlighted the difficulty predicting a specific grade point average, with models unable to adequately capture the variance seen in the dataset. In contrast, models predicting high-performing students who achieve distinctions demonstrated strong performance. These model frameworks can be leveraged by lenders as an additional insight for allocating student funding.

Keywords

funding

higher education

Reference:

Collections

Masters

Full item page