level of personal freedom and remoteness of a country. Among the many others which were considered are the number or length of wars in which a country was or is involved, the number or length of revolutions, an in- dex for bureaucratic efficiency as well as an index for ethnolinguistic fractionalization. Given the high correlation coefficients between some of these measures, we decided to rely on only one indicator for politics (political stability), one for social live (freedom index) and one for geo- graphy (remoteness dummy). Table A.7 in the Appendix presents the results. The findings in Table A.7 clearly meet expectations. It is worth mentioning that coefficients for log population remain significant at least at the 5% level in all model specifications (with the exception of model (2)), which is an indication that country size, at least with respect to pop - ulation, matters.65In passing note that some of the regressions contain few observations, which diminish their explanatory power, especially due to the omission of many VSC, for which data are not available. In spite of this limitation it is really noteworthy that the remoteness dum- my, which differentiates between islands and landlocked countries, is also significant in all regressions. This clearly corroborates the conjec - ture that «international outsourcing» or the transnational provision of public goods play a crucial role in determining the costs of publicly pro- vided goods, although we could not exemplify decreasing coefficients or decreasing statistical significance of the size variables.66If a country is sufficiently small and has no chance to source out parts of the produc- tion and/or provision of necessary public goods because distances are considerable and/or transportation costs are unbearably high, then public goods have to be produced and provided at inefficient scales at home, and we will observe a larger government sector leading to higher public expenditure. The effect of remoteness would probably be even more severe if many small island economies of the Pacific and Caribbean did not have to be neglected in the regressions due to the lack of data. 64Does 
country size matter for public sector size? 65In model (2) the effect of population density is strong enough to lead to the insigni- ficance of the size variable. 66See Section 3.2.4.