plains why Alesina and Wacziarg (1998!) rely on data for the 1980–1984 period to examine the effect of country size on different national account aggregates.49 Their results with regard to other national account aggregates as gov ernment consumption are more or less in line with theoretical pre- dictions. There is no significant relationship between log population and public investment, but comparability of data on investment across coun- tries may be doubted. They find a weak negative effect of country size on public expenditure on education (which is quite surprising, since edu- cation is generally regarded as a local public good exhibiting small or no economies of scale, with the notable exception of tertiary education) and on the broadest available measure of government expenditure, including transfers and interest payments. Transfers, for instance, are expected to rise proportionally with country size (population) and their inclusion in the proxy for government size results in less significant estimates. Alesina and Wacziarg, furthermore, find no relationship between log population and public expenditure on defense, which is a bit surpris ing at first but can be explained quite easily. Generally, public goods provided in connection with defense and national security are expected to display considerable economies of scale, but many smaller countries (like Costa Rica) do not even have military forces or only employ a small police-like frontier guard.50Therefore, we have two offsetting effects, since small countries, which are supposed to have the highest per capita costs for de - fense, sometimes simply avoid establishing a defense force. Further more, there might be a non-negligible part of defense expen diture which does not appear under this heading in budgets or national accounts. All in all, the strongest effects are associated with public consump- tion, which includes all current expenditure for purchase of goods and 50Does 
country size matter for public sector size? 49Note that there are serious principal problems in precisely measuring the size of the public sector in international comparisons. One has to keep in mind that public ex- penditure in international compilations is contingent on the extent of outsourcing to private enterprises and on to what extent privatized or «private» firms are run by the government. The result of different definitions is non-negligible, which has also been shown by the long discussions and detailed regulations in the EU Treaty of Maastricht and the Growth and Stability Pact. It is however impossible to check for all possible caveats in our data on public sector size, and one has to rely on and trust in the IMF, the source most of the data on public sector size, to have sufficiently har- monized the data. 50See Section 4.3.2 for details.