In various empirical studies so-called tax ratios (tax revenues expressed as a ratio of some aggregate tax base) are employed as approximations for tax burdens. The most difficult problem in calculating tax ratios is the way in which personal income tax revenues are attributed to labour and capital. We argue that the methodology of mendoza et al. (1994) is seriously flawed in this respect. Using information from national sources, we calculate more accurate tax ratios for eight oecd countries that differ substantially from those of mendoza et al. (1997). Still, the results of the empirical analysis of mendoza et al. (1997) do not change significantly if we use our tax ratios instead of those of mendoza et al. However, the results change once country specific effects are taken up in the model. Capital taxes are, e.g., shown to reduce economic growth. We find that the results of daveri and tabellini (2000) are neither sensitive to the use of the tax ratios nor to the specification of the model: high labour taxes have increased unemployment in europe.