We study the volatility of growth rates and find that it differs systematically across countries. Our empirical investigation reveals that there is a high correlation between disparity in political regimes across countries and differences in volatility. This is not the case for some of the commonly cited reasons like initial income, inequality or instability of regimes. We find that less democratic countries are more volatile. To explain this observation we use a dynamic model in which democracy is parameterized by the fraction of people who benefit from being in power. The government in this model maximizes the utility of the group in power using a redistributive tax scheme – setting uniform income taxes but transferring lump sum amounts and providing goods and services to the favored group only. When there is a bad shock in this economy, the marginal utility of consumption of agents in power is high. When the transfer is divided among a few, gains from increased transfer outweigh distortionary costs of higher tax. Thus, the optimal tax policy in non-democratic countries, in contrast to that in democratic countries, is such that tax rates are high when there is a bad shock and low when there is a good shock (we refer to this as procyclical tax policy). Further, we show that procyclical tax rates will lead to higher volatility of growth rates than under alternative tax policies. Thus, our model is successful in explaining why tax policies are pro-cyclical in some countries, a commonly observed phenomenon, in addition to providing reasons for differences in volatility of growth rates across countries. The model’s predictions are borne out by data in a number of other dimensions also.