ABSTRACT
This study examines the potential determinants of state-level economic freedom for a panel of the 50 U.S. states from 1994 to 2020. To address model uncertainty in the identification of robust determinants, we use Bayesian model averaging to test the robustness (to the inclusion and exclusion of other determinants) of 17 potential determinants that have been recognized in the extant literature. The results show robustness with respect to the positive impact of the level of per capita real income and its growth, fiscal decentralization and neighbouring state economic freedom, whereas per capita fossil fuel production, population density, unemployment rate and democratic governorships have a negative impact. The remaining determinants were not robustly associated with economic freedom.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 While data on economic freedom at the U.S. state level starts in 1981, the availability of the educational attainment data for all the states restricts our analysis to begin in 1994.
2 For details on the construction of the indices, Stansel, Torra and McMahon (Citation2021) see https://www.fraserinstitute.org/sites/default/files/economic-freedom-of-north-america-2022.pdf.
3 In Delaware, Maryland, Mississippi, and Vermont, the 2020 economic freedom index was lower than in 1994, while the economic freedom index was the same in 1994 and 2020 for Nevada. The economic freedom index was higher in 2020 than 1994 in the remaining states.
4 See Olson (Citation1996, Citation1998), de Haan and Siermann (Citation1998), de Haan and Sturm (Citation2000), Sturm and de Haan (Citation2001), Dawson (Citation2003), Keseljevic and Spruk (Citation2013), Cebula and Clark (Citation2014), Lawson and Murphy (Citation2018), and at the U.S. state-level, see Stansel and Tuszynski (Citation2017) and Ihlenfeld, Hall, and Zhou (Citation2022), among others, regarding the growth effects associated with economic freedom.
5 As a robustness check, we also considered the benchmark prior for k number of determinants for the hyperparameter on Zellner’s g prior suggested by Fernandez, Ley, and Steel (Citation2001) and the random theta prior for the model prior choice from Ley and Steel (Citation2009). The main results are robust to these variations in the estimation approach and are available upon request from the authors.
6 To address potential simultaneity issues the determinants are lagged one period.