The result of the dependence on the right proxy for country size can, how ever, be interpreted in another way. Since from a theoretical view- point it is rather obvious that the number of inhabitants is a better proxy for country size than GDP, especially when the provision of public goods is at stake, the results may simply be considered as empirical evi- dence for the superiority of one of the two 
proxies. 3.2.7 Multiple regressions with politics and geography The preceding sections did not leave scope for political and geographic variables in the regressions. A lot of them have been proposed in related studies, from which we decided to choose only three: political stability, 63 
Empirical evidence Table 3.3: OLS regressions for government consumption and log GDP/ GNP in comparison to government consumption and log popu lation Dependent variable: Period/LogpopLoggdpAdj. R2Number of government consumptionyear(Loggnp)observations in percent of GDP PWT 5.61980–84–2.727**0.052147 (–2.860) PWT 5.61980–84–4.261**0.179142 (–6.047) PWT 5.61990–2.258*0.034115 (–1.984) PWT 5.61990–3.985**0.190110 (–5.195) PWT 5.61992–1.4010.00292 (–1.088) PWT 5.61992–3.481**0.11085 (–3.862) IMF1993–97–2.531**0.119125 (–4.372) IMF1993–97–0.159–0.008115 (–0.297) Sources: Penn World Tables (PWT) 5.6, UNDP and IMF (national accounts). ** significant at 1% level; * significant at 5% level; t statistics based on White heteroscedasticity-con - sistent standard errors in parentheses.