Returns to scale in agriculture have profound implications on how effectively governments’ various efforts in developing countries lead to agricultural intensification and transformation. Historically, agricultural transformation has often accompanied increasing returns to scale. Little direct evidence exists, however, on what actually causes such an increase, despite knowledge about many factors that are associated with the increase. We fill the knowledge gap by testing whether the hiring in of tractor services has increased returns to scale in agriculture at the household level in the Terai of Nepal, an area that has undergone a rapid increase in tractor use through custom hiring services. Using the Roy–Rubin framework of identifying causal mechanisms, combined with endogenous switching regression methods and IPW-GMM, we address two sources of endogeneity involved with the estimation of returns to scale: (a) farmers’ self-selection on whether to hire in tractor services; and (b) use of inputs in the production function. Based on the assumptions of the Cobb–Douglas production function specification and perfect efficiency of farm households, we find that the hiring in of tractor services significantly increased the returns to scale in agricultural production for a certain segment of tractor-hiring farm households not owning tractors, for which suitable control groups are found. Findings are robust under various alternative specifications.