We decompose the underlying disturbances in total hours into three kinds: d
isturbances that shift the level of employment in the long run, those that
change the sectoral composition of employment in the long run, and those th
at cause temporary movement of hours. Our identifying restriction exploits
the distinctive nature of the two margins of labor: employment and hours pe
r worker. Based on the postwar U.S. data, we find that aggregate and sector
al disturbances are roughly equally important for the cyclical fluctuations
of aggregate hours. (C) 2001 Elsevier Science B.V. All rights reserved.