3.3 Summary of empirical results on country size and public sector size Chapter 3 assesses the relationship between country size and govern- ment size. Major findings are the following: –The negative relationship between country size and government size prevails even in the 90ies. No matter which control variables are included, the coefficient for the number of inhabitants is almost al- ways significant. However, it is noteworthy that the results hinge critically on the definition of country size. When GDP or GNP is used as a proxy for country size instead of the number of inhabi- tants, the picture is not clear-cut anymore. This means an economies of scale effect is more closely associated with population than with a country’s economic size, confirming prior theoretical considera - tions and our model. Furthermore, we have to emphasize that this result cannot serve as a basis for political advise, in the sense that larger entities would more or less automatically lead to lower per capita costs in the public sector. There are doubtlessly a lot of other, more effective organizational possibilities to lower costs in VSC, apart from size considerations. –Our basic results are in line with Alesina and Wacziarg, with two notable exceptions. Their result concerning the irrelevance of proxy for country size cannot be confirmed, as explained above. Additionally, one has to be skeptical concerning the sign of the co- efficient for the control variable per capita income. According to Alesina and Wacziarg, wealthier countries should have smaller gov - ernments, which is counterintuitive and contradicts theoretical reasoning. It can be shown that the summary statistics of the data used by them displays means which are far from being realistic, but it is not entirely clear whether their regressions are also based on questionable data. –The hypothesis of a decreasing size effect over time concerning sig- nificance as well as magnitude due to more open countries and glo- bal and/or transnational public goods has to be rejected. Some pos- sible explanations for this result have been presented above, but the 65