EXTENSION OF PANEL DATA MODELS: RANDOM UTILITY MODELS FOR ORDERED CHOICES

Resumo

Health satisfaction serves as a key indicator of perceived quality in health care services and is often used as a proxy for well-being in empirical research. This study explores the application of ordered probit models within the framework of random utility theory, particularly in scenarios where respondents’ preferences regarding health satisfaction are ordinally ranked. Using Stata, we replicate and analyze the results presented in Tables 18.21 and 18.22 of Greene (2018), focusing on the signs and marginal effects of explanatory variables on reported health satisfaction. Our replication confirms the original findings for pooled models, traditional random effects models, and the Mundlak-adjusted random effects model. However, limitations emerged when attempting to replicate the conditional and unconditional fixed effects estimators, primarily due to data constraints and computational challenges. Nevertheless, we observe that the core differences across models lie mainly in coefficient magnitudes, while the direction (sign) of the estimated effects remains consistent. This suggests that individual heterogeneity influences the intensity—but not the direction—of health satisfaction responses.