GGT: Stata module to implement Geweke, Gowrisankaran, and Town Model Quality Estimator

This program estimates the parameters of the Geweke, Gowrisankaran, and Town (Econometrica, 2003), GGT model. The GGT model estimates the posterior distribution of organizational performance where there are many organizations from which individuals can choose to receive services. In this framework, individuals may select organizations based, in part, on information that is unobserved to the researcher and is correlated with the binary outcome. If this is the case, then standard approaches to inferring organization performance will yield biased estimates. The GGT model corrects for this unobserved selection allowing for flexible correlation in the error structure across the organizational choice and outcome equations. The estimation approach is Bayesian. In sum, the model combines an organization choice multinomial probit model with an individual outcome binary probit model, allowing for correlation across equations for each individual. As noted in GGT, some possible applications for this model include: hospital quality based on mortality, school performance based on graduation rates, prison rehabilitation programs based on recidivism rates, and job training programs based on incidence of harassment complaints.