Postdoctoral Research in Support of the National Center for Science and Engineering Statistics

The techniques developed in this project make maximal use of the data, because they are not restricted to using only people who were in the SDR over the entire time period under consideration. This modeling technology was demonstrated by studying the evolution of individuals' salaries over a 15-year time period, as a function of their own characteristics such as gender, race and field of degree, and characteristics of their employers, for instamce, whether the employer is an academic, industrial or government organization, and where the employer is located. Had the study been conducted using those who were in the SDR for the entire 15 years, the sample would have shrunk by two-thirds.

The most striking finding is that adjunct faculty earn only one-half as much as other doctoral recipients with comparable characteristics, even after accounting for the numbers of weeks per year and hours per week they work. Race does not affect salary; however, woman are paid less than men with comparable characteristics who are in comparable positions. This is not, however, evidence of discrimination against women. Statistical associaton does not imply causality, and women's career paths differ from men's in important ways (especially "time off" to care for family members) that are not captured by the model. Salaries are highest in California, the New York City metropolitan area and the Washington, DC metropolitan area, which is consistent with studies of other workforce groups.

The second principal methodological contribution of the project is a way to estimate meaningfully the sizes of small grouups of people (for example, black women aged 35-39 in North Dakota with degrees in environmental engineering) by combining data from the NCSES' National Survey of Recent College Graduates (NSRCG) with data from the Census Bureau's American Community Survey (ACS). Especially in times of extreme pressure on the federal budget, this kind of leveraging of datasets by combining them is central to maximizing the value of NCSES' data collection activities. The methodology is complicated and poses significant computational burdens, but the concept is simple. Imagine having to prepare a cross-tabulation ("contingenchy table") of sex versus race when only the totals for men, women and each racial category are known, and come from different data sources.

The impact is more detailed understanding of the U.S. scientific workforce, which is essential to improving the nation's competitiveness in today's global economy. The human resource impact of the project on the three postdoctoral fellows who participated in it is also impressive. Two of them were so attracted by the opportunity to work with NCSES datasets that they left full-time postions to join the project. Of the three, one has effectively joined the federal statistical system, albeit indirectly, as an employee of a major private-sector contractor. One of the other two now holds an academic position in the U.S. that would not have been available to her without the experience she gained in the project. The third has returned to his home country to pursue an academic career there.