see that, with the possible exception of men age 65 and older, the changes were almost exclusively within occupations.5 As an example, the proportion of the workforce that is female and between ages 25 and 54 rose by .043 over this period; using the first decomposition, 93 percent of this change was within occupations.

This finding is even stronger for changes by race and ethnicity. The across-occupation shifts would have reduced slightly the proportions of blacks and Hispanics, as indicated by the negative values for the across-occupation component of the decomposition. This implies that the occupations in which blacks and Hispanics tended to be represented have shrunk, so the increase in the proportion of the workforce in these categories was all within occupations.

The changes by educational attainment are more balanced. The decline in the proportion with less than four years of high school is largely a within-occupation phenomenon, as is the increase in the proportion with some college. However, the smaller change (decline) in the proportion with 12 years of schooling occurred partly across and partly within occupations, whereas about half of the increase in the proportion with at least four years of college was across occupations.6

The decomposition for employer tenure, in particular, indicates growing diversity of the workforce within occupations, as the entire increase in the proportion both with low tenure and with high tenure occurred within occupations. Finally, the decomposition for training also indicates possible changes within occupations, as approximately 80 percent of the increase in both types of training arose within occupations, pointing predominantly to changing skill requirements of existing occupations, rather than growth of higher-skilled occupations.

To summarize, this evidence on demographic and workforce


A negative value for the across-occupation shift is not at all anomalous. It arises when occupations in which a particular group is represented shrink, so that a within-occupation increase in the representation of this group is necessary simply to hold the overall representation of the group constant.


This latter calculation must be regarded cautiously, given changes in the coding of education in the Current Population Survey between 1983 and 1993.

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement