Comparing Classical and Bayesian Panel Kink Regression Frameworks in Estimating the Impact of Economic Freedom on Economic Growth

Abstract

This study aims to accomplish three main tasks. Firstly, it seeks to determine the more appropriate choice between classical and Bayesian methods in estimating a pooled panel kink regression model under the condition of a known but bounded policy variable choice that serves as a kink point. Secondly, as a product of the first target, the study seeks to provide empirical evidence for the economic growth–economic freedom nexus in five top-performing economies in Sub-Saharan Africa. Using index explanatory variables, which are bounded between 0 and 100, and using both numerical and graphical methods, the findings show that the use of the Bayesian method is more appropriate in characterizing the data than the classical OLS framework, as the former better accounts for randomness via the use of posterior distributions. Finally, the study further employed both threshold and Bayesian pooled panel kink regressions, with mixed results. The Bai–Perron test confirmed that the economic freedom index has a single threshold value of 56.70. Whereas the threshold estimates show a negative impact of economic freedom on growth in both low and high regimes, the Bayesian estimates reveal that economic freedom has a negative impact on growth in a low regime but a positive impact in a high regime. Our novel findings show that there exists a nonlinear impact of economic freedom on growth. This provides some guidance and caution in charting policy paths that seek to achieve economic growth.