We measure cost efficiency of 202 Czech municipalities of extended scope in period 2003-2008. The study is the first application of overall efficiency measurement of the local governments in the new EU member states, and the second in post-communist countries. We measure government efficiency through established quantitative and qualitative indicators of the provision of education, cultural facilities, infrastructure and other local services. First, we employ non-parametric approach of the data envelopment analysis and adjust the efficiency scores by bootstrapping. Second, we employ the stochastic frontier analysis and control for effects of various demographic, economic, and political variables. We compare scores under our preferred specification, i.e. pseudo-translog time-variant stochastic-frontier analysis with determinants, with alternative scores. The determinants that robustly increase inefficiency are population size, distance to the regional center, share of university-educated citizens, capital expenditures, subsidies per capita, and the share of self-generated revenues. Concerning political variables, increase in party concentration and the voters' involvement increases efficiency, and local council with a lower share of left-wing representatives also tend to be more efficient. We interpret determinants both as indicators of slack, non-discretionary inputs, and unobservable outputs. The analysis is conducted also for the period 1994-1996, where political variables appear to influence inefficiency in a structurally different way. From comparison of the two periods, we obtain that small municipalities improve efficiency significantly more that large municipalities.