Notes

[2] Although BLS labor force projections do a reasonable job of forecasting employment in many occupations (see Alpert and Auyer 2003), the mean absolute percentage error in the 1988 forecast of employment in detailed occupations in 2000 was 23.2%.

[3] Since their growth rate projection is near the overall average, engineers and physical scientists are classified as having average growth by BLS.

[4] Not all analyses of changes in earnings are able to control for level of skill. For example, data on average earnings within occupation over time may not be a good indicator of labor market conditions if the average experience level was to fall for workers in a rapidly growing occupation.

[5] Many comparisons using Census Bureau data on occupations are limited to looking at "nonacademic S&E occupations" because the occupation of "postsecondary teacher" has not been broken out into subjects in most recent census surveys.

[6] Specifically presented here are coefficients from linear regressions using the 2003 Scientists and Engineers Statistical Data System (SESTAT) data file of individual characteristics on the natural log of reported full-time annual salary as of October 2003.

[7] "Underrepresented ethnic group" as used here includes individuals who reported their race as black, American Indian/Alaska Native, or other, or who reported Hispanic ethnicity.

[8] In the regression equation, this is the form: age, age2, age3, age4; years since highest degree (YSD), YSD2, YSD3, YSD4.

[9] Included were 20 dummy variables for NSF/SRS SESTAT field-of-degree categories (out of 21 S&E fields; the excluded category in the regressions was "other social science").

[10] Variables added here include 34 SESTAT occupational groups (excluding "other non-S&E"), whether individuals said their jobs were closely related to their degrees, whether individuals worked in R&D, whether their employers had fewer than 100 employees, and their employers' U.S. census region.

[11] Variables added here include dummy variables for marriage, number of children in the household younger than 18, whether the father had a bachelor's degree, whether either parent had a graduate degree, and citizenship. Also, sex, nativity, and ethnic minority variables are included in all regression equations.