Identifying Future Entrepreneurs in Southern Africa: Insights from the Allan Gray Orbis Foundation’s Second Validation Study

Context
Given South Africa’s current socio-economic challenges, the early identification and development of entrepreneurs that contribute positively to socio-economic development remains key to creating economic wealth and equality.

Objective
This case study presents the second validation of the Allan Gray Orbis Foundation (AGOF) selection processes for its Scholarship and Fellowship programmes across 2019–2024. The study aimed to evaluate the reliability, validity, fairness, and predictive utility of the selection tools, and to optimise decision matrices for identifying candidates with strong academic potential and entrepreneurial aptitude.

Theoretical Foundation
The validation was grounded in psychometric theory and best practice in personnel selection. Reliability was assessed using Cronbach’s Alpha and McDonald’s Omega, construct validity through Confirmatory Factor Analysis (CFA), and predictive validity via regression modelling. Bias and fairness were examined using ANOVA, while range restriction corrections were applied using Thorndike’s Case A method to estimate true predictor–criterion relationships. These analyses align with the principles of evidence-based selection and fairness in high-stakes educational contexts.

Methodology
Data from over 69,000 applications were cleaned, standardised, and analysed. The Scholarship process involved a Foundation Exam (Numeracy and Literacy) and an Interview. The Grade 12 Fellowship intake comprised an Application Form, Interview, Entrepreneurial Camp, and Entrepreneurial Readiness Quiz (ERQ), while the University intake included an Interview and ERQ. Analyses included descriptive statistics, reliability and validity assessments, regression models for predictive validity, and optimisation of predictor weights based on standardised beta coefficients.

Results
Reliability analyses showed strong internal consistency for Scholarship exams (Numeracy α=0.83–0.88; Literacy α=0.70–0.81). Fellowship Grade 12 tools varied: Application α=0.90, Interview α=0.76, Camp α=0.79, while ERQ demonstrated weak reliability (α=0.57). For the University intake, the Interview showed acceptable reliability (α=0.822), but ERQ remained weak (α=0.56). CFA indicated excellent fit for interviews (CFI≥0.998) but mixed results for exams and camp; ERQ validity was uncertain due to aggregated data. Predictive validity analyses revealed that prior academic achievement consistently predicted later academic performance (β≈0.34 for Grade 11; β≈0.44 for first-year marks). Entrepreneurial engagement was best predicted by interview and ERQ scores in the university intake (R²≈0.59), while camp and ERQ (Grade 12) showed negligible predictive power for academic outcomes. Bias analyses revealed small-to-moderate differences by race and socio-economic status, with minimal gender and age effects. Correcting for range restriction substantially increased correlations, confirming attenuation in observed validity. Optimisation analyses suggested that current weightings undervalue academic predictors and recommended increasing weights for academic scores and ERQ.

Implications and Insights
From a scientist-practitioner perspective, the findings highlight the need to capture item-level data for robust psychometric evaluation, rebalance decision matrices to emphasise academic predictors while retaining entrepreneurial measures, and address demographic disparities. Longitudinal tracking of outcomes such as graduation and entrepreneurial success is essential for future validation. These refinements will enhance fairness, predictive accuracy, and alignment with AGOF’s mission to develop entrepreneurial leaders who drive socio-economic change.