expenditure and GDP, obviously, differ. The scatterplot in Figure 3.2 gives a first clue of the differences and Table 3.3 presents univariate re- gressions for various years. It can easily be seen that the variable population cannot be replaced by GDP or GNP, at least for the 90ies. This is an interesting result, since it establishes that the detected negative relationship between coun- try size and government size is conditional upon the selected operatio n - alization of the variable country size and does not even prevail in the two most widely used definitions. Why does GNP or GDP work contrary to our expectations as a proxy for country size? One promising explana - tion might be the fact that many smaller countries, especially in Europe, are performing extremely well from an economic viewpoint. In other words, they «grow» when GNP or GDP is applied as a proxy for coun- try size relative to the application of the number of inhabitants. But the explanation is not entirely convincing, because there are also quite a few VSC located in the Pacific and in the Caribbean, and some of them are not among the wealthier countries in the world. 62Does 
country size matter for public sector size? Figure 3.2: Government consumption and log GNP 25 20 15 
30 
35 5 
10 02,5 2 1,535 4,5 4 3,5 loggnp96 
goco93–97 GOCO9397:government consumption 1993-1997 from IMF (national accounts). LOGGNP96: logarithm of GNP from Baratta (1999). 120 countries, r = 0.29 (not significant).