Introduction South Africa’s land redistribution programme faces a productivity–equity dilemma: redistribution is constitutionally mandated, yet agricultural productivity stagnates while sectoral inequality persists. Existing studies typically examine isolated variables, limiting their policy relevance. Addressing this gap, this study develops and applies a Predictive Analytics (PA) model within a General Systems Theory (GST) framework to evaluate how land reform policies interact with structural, environmental, and institutional variables. Methods Using a quantitative Computational Social Science (CSS) approach, the model simulates projected outcomes of nil-compensation expropriation under Section 12(3) of Act No. 13 of 2024 within scenario-based conditions. Longitudinal data (2014/15–2023/24) from the Department of Agriculture, Land Reform and Rural Development (DALRRD) were analysed using machine learning (ML) techniques. Key indicators included Environmental Performance, Structural Capacity, and Barrier Indices. Feature importance analysis, trend diagnostics, and a policy classification tree generated three principal insights. Results First, productivity is more strongly associated with utilisation efficiency than with land expansion. Second, equity outcomes are systematically associated with structural barriers, including tenure insecurity and the decoupling of growth from inclusion. Third, while Act No. 13 is conceptually aligned with targeting underutilised land, scenario simulations indicate that redistribution without post-acquisition support is unlikely to yield sustained equity gains. Discussion The study produces a GST-informed, prescriptive policy framework. By translating legal provisions into measurable predictive factors, it offers an evidence-based decision-support tool for designing conditional interventions. The findings indicate that equity outcomes are structurally linked to productive performance; effective reform therefore requires prioritising land activation rather than transfer alone.
Mzulwini et al. (Wed,) studied this question.