4. 3 Using other measures of government size
We now examine the growth implications of two other measures of the overall size of government. In column 4 of Table 2, we report results when we use the growth rate of total government expenditure as a share of GDP, Δgovexp, as a measure of size, while in column 5 we use the growth rate of total tax revenues as a share of GDP, Δtrev, as a measure of government size. Both variables are insignificant and the associated R2 drop by almost 50% (both regressions pass the serial correlation, the Hausman and the RESET tests). Obviously, shorter time periods do not help to make precise inference here (we remind that data for Δgovexp and Δtrev are available for 1973-1998 only), but we believe that this non-significance should be attributed to other factors. As we argued above, tax revenue is a problematic variable. In the case of total expenditure, this is a very general measure consisting of many components, each of which may have a different effect on the growth rate. This explains the popularity of Δgovshare in this type of studies. Finally, we report that we have also tried Δbudget as a measure of the size of government in our growth regression. This variable is again insignificant.
4. 4 Summary of the section
When we follow common practice by using the government share in GDP as a measure of the overall size of government, there is robust evidence that larger sizes hurt macroeconomic growth. Specifically, a rise in the growth rate of government
11
share in GDP by one standard deviation (or 7.655%) decreases GDP growth by 1.38%. On the other hand, alternative measures of the overall size, like total government expenditures or tax revenues, both as shares of GDP, are insignificant. Actually, the significance of the government share in GDP, combined with the insignificance of total government expenditure as a share of GDP, imply that different components of fiscal policy may have different effects on growth. This possibility is explored in the next section. Finally, there is evidence that endogeneity of fiscal policy does not introduce significant biases in the above results.