This article empirically assesses the effect of unobserved transaction costs for Nigerian farmers when investing in irrigation pumps. The unobserved stochastic threshold (UST) model popularly used in labor economics literature is applied, and is compared with two models, the Tobit and the Heckman sample selection model, which are slightly more restrictive versions of the UST model. The results indicate that the unobserved transaction costs are higher for female farmers, landless farmers, those who have a comparatively high dependency ratio, and those who live far from town. The results suggest that the unobserved transaction costs can be as important as the factors determining the profitability of irrigation pumps, which emphasizes the importance of reducing transaction costs and providing the environment for higher returns from investment in irrigation. From a methodological perspective, the UST model is found to be more accurate and informative than the Tobit and Heckman models.