This paper explores the implication of the rural non-farm employment and waged agricultural employment for income and risk mitigation of agricultural households in Ethiopia. We use a nationally representative panel dataset, and employ endogenous switching mixed multinomial logit model. After we control the selection bias that can arise from employment selection in rural households stemming both from observed and unobserved characteristics, we do find an evidence that the non-farm sector improves the income of the agricultural households, and serve as risk mitigation tool in rural Ethiopia. Combining the non-farm sector with hiring labor for agricultural activities at times of labor shortage increases the variance and contributes to positive skewness of income. Finally, we suggest that the rural non-farm sector could serve as a key development pathway for improving livelihood in the predominantly smallholder production system in Ethiopia.