The study examines conservation agriculture technology within a sustainable agriculture context. In particular, the study uses dynamic discrete time approach to examines the role of social and institutional networks in the adoption and diffusion of conservation agriculture technology in Zambia. To correct for potential reflection problems in estimating the diffusion process, a pricipal component approach is used to generate latent variables for peer and institutional effects. A selectivity-corrected metafrontier approach is employed to assess the impact of conservation agriculture on technical and environmental efficiency and the key determinants of efficiency. The study also explores household allocative and scale efficieny using a zero-inefficiency stochastic frontier framework. An endogenous switching regression model and a recursive bivariate probit model are employed to assess the impact of the technology on farm output, farm throughput accounting ratio and household poverty status. The empirical results reveal that conservation agriculture has a role to promoting economic, environmetal and social indices of sustainability. Also, unlike the stochastic frontier model, the zero-inefficiency stochastic frontier successfully accounts for both fully efficienct and inefficient firms in the estimation strategy. The results also demonstrate the presence of scale econmies. Furthermore, farmers practicing conservation agriculture are technically and environmetally more efficient compared to conventional farmers.The adoption of the technology increases output and throughput accounting ratio and reduces poverty.Moreover, social and insitutional networkw have positive impact on adoption and diffusion of the technology. The results further indicate that years of schooling, risk preferences,access to credit, extension ad machinery positively influence adoption and diffusion.Inefficiency is explained by level of education, access to extension and credit as well as distance to markets.