Testing local level labor force and unemployment projections

The labor force projections perform well, even for small areas, which suggests that changes in local labor force participation rates can be approximated by national changes. In fact, the mean absolute percent errors are low even when the previously calculated population projections are used. The unemployment projections do not do as well, and using 1980 census labor force data instead of previously calculated labor force projections offers no improvement in the results. A two-step study is needed to determine why the errors are so large. First, the state level changes from 1970 to 1980 should be compared with national changes to determine the difference in unemployment rate changes by race and sex. Second, state level data on the occupational mix should be examined for the relationship between this and the changes in state level unemployment rates by race and sex. It is hypothesized that the first step will show that even state level changes in unemployment rates by race and sex cannot be well approximated by the national changes and the second step will show that some of the variation can be explained by state differences in occupational mix. Further studies should be made to determine how the calculation of national changes affects the results. Changes in labor force participation rates seem to follow trends, and therefore extrapolation may not have much effect on the results. Although using the actual changes in unemployment from 1970 to the latest year available as a proxy for the changes from 1970 to the target year may have a serious detrimental effect on the unemployment projections, if some of the variation in the national changes in race and sex can be explained and some adjustments made, the actual changes may be the best proxy to use.