A Bayesian approach to modeling economic growth: Variable selection and cross-sectional dependence

This paper introduces a Bayesian framework designed to address two critical econometric challenges in empirical growth studies: variable selection and cross-sectional dependence. We first propose a growth regression model that incorporates cross-sectional dependence through a multi-factor structure and accounts for variable selection with sparse coefficients. We then introduce a new Bayesian Adaptive LASSO estimator specifically designed for this model, capable of performing variable selection while simultaneously addressing cross-sectional dependence. To validate this framework, we conduct a Monte Carlo study, demonstrating that our Adaptive Bayesian LASSO growth regression model with a factor structure consistently outperforms three benchmark models across all evaluation metrics. Furthermore, we apply our model to a dataset encompassing 55 candidate variables across 71 countries from 1961 to 2019. Our empirical comparison confirms the superiority of our model over the benchmark models. Additionally, our model identifies five global factors and 16 key growth determinants, with robustness checks using Principal Component Analysis and instrumental variables confirming the reliability of our findings.