Using a cross-sectional data collected on 400 cocoa farmers from the Ashanti and Western Regions of Ghana, this paper provides empirical evidence on the impacts of cocoa agroforests on yield and household income. The propensity-score matching model was employed. The heterogeneity of high, medium and low shade adopters is statistically addressed. The empirical results generally indicate that adoption of cocoa agroforests has significant positive impacts on yield and household income. The impact on yields for low shade adopters was higher than medium shade and high shade adopters of cocoa agroforests. The paper provides useful policy recommendations based on the empirical magnitudes and directions on sustainable cocoa production and household welfare.