On the Size Distribution of Employment and Establishments

Recent arguments that employment growth occurs disproportionately at small establishments are fundamentally misleading because they confuse regression to the mean with structural shifts in the size distribution of establishments and with an aging effect within cohorts. The net growth usually observed in aggregate studies hides the gross flows; 13 percent of the jobs in existence in 1974 had disappeared by 1980, while 18 percent of the 1980 jobs had not existed six years previously. The variation observed here in labor demand over time within individual establishments may help to explain unemployment.