Concluding
Remarks This replication study revisits Rodrik’s (1998) claim that government spending plays a risk reducing role in open economies which had support in Ram’s fixed-effects results. We show that the positive association between government size and openness is dependent on the version of the PWT data and the composition of the sample. In particular, the positive coefficient of openness is insignificant when using the updated PWT data for Ram’s sample (154 countries, 1960 to 2000) but significant when expanding the sample to additional countries and years (189 countries, 1960 to 2010). This mixed result is evidenced in the literature where there is no uniform support for Rodrik’s hypothesis. For example, Benarroch and Pandey (2008, 2012) find that bigger government Granger causes lower openness, but openness does not Granger cause government size. This remains true whether one considers aggregated government spending or specific expenditure components. Shelton (2007) observes that government expenditures associated with increased openness are not in categories that explicitly insure for risk.
We also find some evidence in favor of Alesina and Wacziarg’s hypothesized negative association between country size and government size. While Ram finds a significant positive association between country size and government size, the country size coefficients using PWT 7.1 data are negative and attain significance in some instances. However, as was the case in Ram’s analysis, the country size coefficients using PWT 7.1 data are not significant in the country size–openness regressions. Because Alesina and Wacziarg’s hypothesis that country size mediates the positive relationship between openness and government size requires that there exists a negative association between country size and government size, and country size and openness, we replicate Ram’s result that fixed-effects estimates do not support the joint hypothesis.
