Current and Alternative Sources of Data on the Science and Engineering Workforce
For more than 50 years, the NSF has attempted to meet the need for current information on the important S&E segment of the population. Although SESTAT provides substantial employment-related data for those with S&E degrees from U.S. educational institutions, some users have noted gaps in the data. For example, some users want more details concerning working environments and career paths. Also, the current definition of the S&E population excludes individuals with degrees below the bachelor's level or who lack degrees but have S&E occupations, but such individuals are of interest to some users. Another limitation of the current NSF system is that it has coverage gaps. For example, most individuals currently living in the United States who received their S&E degrees outside the United States after the 1990 decennial census are not included in SESTAT. Similarly, individuals with non-S&E degrees who were not working in S&E occupations at the time of the 1993 NSCG but who entered S&E occupations after that time are not included.
This report considers the feasibility and desirability of using other ongoing federal surveys, such as the ACS, the NIS, and the CPS as possible alternative sources of a sample of the S&E population. The use of an "establishment" approach is also examined.
On review, based on sample size considerations, the conclusion is that only the ACS and the NIS could be reasonably considered in their current form as alternative screening vehicles for an S&E workforce survey. The sample sizes from all the other federally sponsored household surveys are far too small to satisfy even much less stringent precision requirements than those currently specified for the S&E workforce estimates.
It is possible that the ACS or, less likely, the NIS, might offer the potential for NSF to develop a cost-effective S&E personnel data system that responds more fully to user needs. Even given the large sample sizes of these surveys, however, the minimally acceptable CVs established for the various S&E workforce estimates (appendix D) cannot be met from one year's data. Serious consideration should be given to relaxing the levels of some of the target CVs and to accumulating data over 2 or more years. Accumulation may be of particular relevance for producing estimates of required precision for some "rare" domains. It is important to note that both the ACS and NIS may have limitations on their availability imposed by the sponsoring agency. It should also be noted that the use of one of these surveys as the basis of an S&E workforce survey would result in the loss of the longitudinal analysis capabilities of the current design. There is much further work to be done before establishing that either survey could serve the needs of an S&E data system.
Establishment surveys do not seem to provide a feasible means of reaching the full S&E universe, although an establishment survey approach could be considered for the collection of limited information directly from employers, without requiring contact with the employees. However, it is not clear whether the existing government lists of establishments maintained by the Census Bureau and by the Bureau of Labor Statistics could be utilized easily, if at all. Existing private establishment lists, as noted, contain some limitations, and extensive research and investigation would be required before this approach could be deemed acceptable.
The limitations found in the alternative approaches examined in this report make their use for the 2003 SESTAT impractical. The sampling frames of the 1990s are the most cost effective and practical approach for the near term. The possible use of the ACS later in the decade after it has become fully operational deserves serious preparatory research in the immediate future.