A Two-Step System for Dynamic Panel dealing with Endogeneity Issues and Causal Relationships

Abstract

The paper addresses a computational method implementing a standard Dynamic Panel Data model with Generalized Method of Moment estimators to deal with endogeneity issues, because of omitted factors and unobserved heterogeneity, and causal relationships in large and long panel databases. The methodology takes the name of Two-step System Dynamic Panel Data that combines a first-step Bayesian procedure for selecting potential candidate predictors in a static linear regression model with a frequentist second-step procedure for estimating the parameters of a dynamic linear panel data model. An empirical example to the effects of obesity and socioeconomic factors on labor market outcomes among Italian regions is performed. Potential prevention policies and strategies to address key behavioral and diseases risk factors affecting labor market outcomes and social environment are also discussed.