Multivariate Tobit Model: An Application to Farmer Adoption of Soil Erosion Practices in El Salvador

This chapter talks about the appropriate modeling framework for mixed discrete-continuous random variables which is popularly called the Tobit model and its extension the multivariate Tobit model which is essentially a Seemingly Unrelated Regression or multi-equation Tobit system. The model is illustrated through an empirical application on technology adoption problem in agriculture. In developing countries, failures of programs to formulate technological prescriptions for soil conservation raise doubt about the effectiveness of large-scale public interventions. Many farmers abandon program practices soon after an intervention ends. Is this crucial evidence that farmers engage in strategic adoption behaviors just for program incentives, and that public intervention is inept? Most such criticisms rely on secondary and/or cross-sectional data—which are of limited use for tracing adoption behaviors during and after public intervention. This study uses two-period data for participants in the Environmental Program for El Salvador, during and after the intervention. To model dynamics between multiple adoption decisions—which differ across two periods and four viable practices—eight Tobit models are jointly estimated in Stata software using the Geweke-Hajivassiliou-Keane simulator.