St. James Ballroom
A Maximum Likelihood Method for Correlated Discrete and Continuous Outcomes with Selection and Lagged Effects: Quantifying the Role of Past Season Yield on Improved Maize Seed Adoption (303852)
*Rhoda Nandai Muse, University of Arizona
Keywords: Maximum likelihood, selection/endogeneity, correlated discrete and continuous outcomes, longitudinal data
Agricultural technology adoption and production remains low in developing countries. Farmers are risk averse but quantifying this effect has been difficult. Empirical studies on risk have used two stage methods and expected yield and variance to capture elements of risk. It has been argued that past season (lagged) yield may affect the capacity of farmers to take on risky inputs in the next season. This value has not been quantified in empirical studies due to estimation challenges that arise. We use maximum likelihood method to quantify the effect of past season yield while accounting for the endogenous seed use effect on yield. Improved seed use (discrete) affects yield (continuous variable) but is correlated with the yield error term therefore it is an endogenous or selection term. Maximum likelihood methods exist for correlated discrete and continuous outcomes with lagged values only or the endogenous term alone but not both. We extend the method of lagged effects to include the endogenous term and apply it to household survey data from Kenya. We found that higher past season yield significantly increased the likelihood of improved seed adoption in the next season.