This paper studies determinants of the geographic distribution of unemployment in the United States since 1950. We argue that “equilibrium” differences in unemployment among markets are partially supported by corresponding differences in the covariance structure of sectoral demands for labor. When workers are mobile, greater diversification of sectoral demands reduces unemployment. We confirm this point using pooled time-series-cross-section data on state unemployment and employment by industry. We also find evidence for nonneutrality of aggregate disturbances based on geographic differences in industrial composition, and that permanent changes in the sectoral composition of employment lead to transitory fluctuations in unemployment.

A stochastic dynamic model of labour supply is specified and analysed empirically. The theoretical model provides a sound structural economic basis for a class of empirical models which have been shown useful. The implications of the estimated model are explored and diagnostic checks are suggested and implemented.