Abstract
The high correlation between aggregate unemployment and the dispersion of industrial unemployment can be explained by both aggregate disturbances and sectoral shifts. This paper separates the effects of these two shocks by decomposing each industry's unemployment into aggregate and idiosyncratic components, so that the two new unemployment dispersions constructed by only the aggregate and idiosyncratic components measure respectively the influence of aggregate disturbances and sectoral shifts. It is found that this high correlation is almost entirely attributed to aggregate disturbances and that sectoral shifts are not important in explaining unemployment fluctuations in the U.S. for the period 1948–1990.