scatterplot for the relationship between government consumption as a percentage of GDP and log population. Table A.5 in the Appendix lists those countries for which data on government consumption from 1993 to 1997 are available. As can be detected fairly easily, more than 60 coun- tries are missing, and the data situation is – as expected – especially poor for smaller countries. After all, there are 17 countries with fewer than one million inhabitants, of which twelve provide data on government consumption for the whole period from 1993 to 1997. As to government consumption and population, a statistically sig- nificant negative relationship can be detected. Not surprisingly, the in- tuition of Table 3.1 is herewith confirmed. The univariate OLS regres - sion between government size and country size in Table A.6 in the Appen dix is significant at the 1% level for the 90ies. We find that a doubl ing of popula tion corresponds to a 0.76% decrease of government con sump tion relative to GDP (2.532*log2). Therefore, government con- sumption relative to GDP is expected to be 4% higher in the U.K. than in the U.S.A., for instance, solely due to the size effect. This result can, 52Does
country size matter for public sector size? Figure 3.1: Government consumption and log population 25 20 15
30
35 5
10 05 469 8 7 logpop96
goco93–97 GOCO9397:geometric mean of government consumption 1993–1997 in percent of GDP from IMF (national accounts). LOGPOP96:logarithm of population from Baratta (1999) (country profiles). 126 countries, r = –0.354 (significant on the 1% level).