Thesis (MEcon)--University of Stellenbosch, 2006. / Education production function analysis is widely recognised as one important area of research that needs to inform education policymaking, specifically policy relating to the mix of funded inputs in a schooling system. Arriving at production functions is a complex task, and is fraught with methodological pitfalls. This thesis sets out to establish a framework for undertaking education production function analysis, and in discussing its various elements, including its pitfalls, recommendations for good practice are arrived at. The material analysed is of four types: texts on econometric theory; existing production function analyses; documentation relating to three dataintensive school monitoring programmes, namely Brazil’s SAEB, South Africa’s Systemic Evaluation and the international SACMEQ programme; and lastly data, relating mainly to South Africa, from the 2000 run of SACMEQ. The thesis is organised according what can be regarded as seven key analysis steps. These steps include a focus on the importance of a ‘mental model’, the relative benefits of the one-level regression model and the hierarchical linear model (HLM), and the formulation of actual production functions for South Africa based on the SACMEQ data, using both one-level and HLM models. Key conclusions are, firstly, that the HLM, though still under-developed, offers great analysis potential and, secondly, that production function analyses ought to be translated into budgetary terms in order for them to become fully meaningful to the policymaker.