Comparative Evaluation of Housing Delivery Models in Gauteng Metropolitan Municipalities: Towards a Conceptual Governance Framework

Africa 0710431w@students.wits.ac.za Persistent failures in subsidised housing delivery across South Africa’s metropolitan municipalities have exposed a governance deficit that prevailing project-management remedies have not resolved. This paper develops an integrated conceptual framework for the comparative evaluation of two competing housing delivery models in Gauteng’s three metropolitan municipalities, namely implementing-agent-led delivery and municipally led delivery, across the City of Johannesburg, the City of Tshwane and the City of Ekurhuleni. The framework synthesises five theoretical lenses: Principal-agent theory, governance theory, transaction cost economics, project management maturity theory and contextual institutionalism. From this synthesis the paper derives a four-construct causal architecture in which the delivery model (independent variable) influences project performance outcomes (dependent variables) both directly and indirectly through governance effectiveness (mediating variable), with institutional and organisational project management maturity moderating the governance performance relationship. The paper argues that delivery-model choice is properly understood not as a technical procurement decision but as an institutional governance decision whose performance consequences are contingent on oversight intensity, accountability mechanisms and consequence management. The contribution is a context-sensitive governance framework that addresses a five- level research gap spanning global, African, South African municipal, metropolitan-specific and governance- framework-deficiency scales. Methodologically, the framework is built through a systematic narrative synthesis of the principal-agent, governance, transaction-cost, project-management-maturity and institutionalist literatures, using structured theory triangulation to integrate five previously disconnected theoretical lenses into a single, testable four-construct causal architecture rather than through primary empirical data collection. Analysis of Post-Retrofit Changes in Window Operation Behavior in