Purpose. The study aims to examine the effect of non-farm employment decisions on the adoption decision of rice hybrid seeds or improved variety in Vietnam. Methodology / approach. This study uses panel data from the Vietnam Access to Resources Household Survey (VARHS) 2008–2016 dataset. The study uses the correlated random effect Probit model with the Mundlak approach to control unobserved heterogeneity of panel data and the endogenous switching Probit model (ESPM) to solve the endogeneity problem and self-selection of the non-farm participation variable. Results. There has been increasing interest that the development of rural non-farm employment has effects on agricultural production as well as agricultural growth. However, still relatively poor understanding of how non-farm participation affects the farmers’ decision to adopt modern technologies in the face of market failure. Our findings indicate that non-farm employment has a positive effect on the adoption of rice hybrid seeds or improved varieties in Vietnam. The value ATT is predicted from the endogenous switching Probit model, which implies that farm households who engage in non-farm employment had a 35.1 % of probability of modern varieties adoption, vs. 19.0 % in the sample overall. Originality / scientific novelty. This study adds evidence from a developing country (using the example of Vietnam) to the broader literature on the role of non-farm employment participation on farmers’ adoption behaviour under market imperfections. In addition, the research addresses the limitations of unobserved heterogeneity of an unbalanced panel by applying the Mundlak approach and contributes to the literature by controlling the endogenous problem and self-selection problem of non-farm participation by using the endogenous switching Probit model. Practical value / implications. Based on the empirical results of this paper, some policy implications are provided to develop the rural non-farm sector and to diffuse modern technologies among rural farmers.