Monthly data for the aggregate U.S. economy are used to assess competi
ng hypotheses concerning the relationship between sectoral employment
shifts and fluctuations in the unemployment rate. It is shown that sec
toral shifts are caused by major work stoppages, aggregate fluctuation
s unrelated to permanent sectoral shifts, a reallocation-timing effect
, and allocative shocks at the sectoral level. Larger employment shift
s are associated with higher unemployment during slow growth periods a
nd lower unemployment during above-average growth periods. Models are
presented which demonstrate that both aggregate and allocative shocks
are causes of cyclical increases in unemployment.