This paper explores the relationship between output volatility and economic development. We develop a methodology to assess countries' extent of sectoral diversification. The productive structure of a country tends to be risky when the country i) specializes in highly volatile sectors, ii) has high sectoral concentration, and/or iii) specializes in sectors highly affected by country-specific fluctuations. We document the following regularities. First, sector-specific risk declines monotonically with development. Second, at early stages of development, sectoral concentration declines with development, whereas at later stages the relation flattens out. Third, country-specific risk declines with development. Fourth, the covariance between country- and sector-specific risks increases along the development path. We derive the implied mean-variance frontiers both for individual countries and for the world, and compute countries' distances to each. We find that poor countries are typically inside their mean-variance frontier, that is, they could achieve the same level of productivity at lower risk by modifying their sectoral composition. We discuss the implications of our findings for existing theories of volatility and growth.

This paper was revised in March 2004.