STEM Performance and Supply: Assessing the Evidence for Education Policy

Although it is now fashionable to look at the first decades of the post-World War II period as the golden age of U.S. education—in which education was once winning the race against increasing skill demands of technology—this nostalgia is belied by the actual accounts of the time that decried national education crises and shortages, albeit ideologically rather than empirically driven arguments that, as we will show, are repeated in today’s narratives (Ravitch 1983, 2014; Rothstein 1998).

Comprehensive, nationally representative studies of college major and occupational entry are limited to the National Center for Education Statistics longitudinal survey “Baccalaureate and Beyond,” which were for graduating cohorts in 1993 and 2008. These estimates are consistent with other data sources such as the National Science Foundation’s SESTAT on similar populations. Although the 2008 cohort entered the job market during the Great Recession of 2008, the population analyzed is restricted to those who are employed. The recession may have had an impact on occupational transition (and/or employment rates), but it does not appear to have had a large impact on these historical trends identified in other cohorts.

There is no reliable or even reasonable method for estimating non-STEM occupations that require a STEM-degreed education, but even highly speculative estimates of jobs that might need to be filled by someone with a STEM degree still show 30 to 60% more STEM graduates than employed in those jobs. Here, we use “STEM” as excluding social sciences and adjust calculations from the Census Bureau classifications that included social sciences.

See Freeman 1976 for development of the “cobweb” model of this function; see Lynn et al. 2018 for a contemporary case study.

This is another case of industry executives using a public policy platform--a Presidential Advisory Council--to give credibility to a narrative of “shortages” and supposed education failures and avoiding mention of the effect of the Great Recession of 2008 (see Lynn and Salzman, 2011).

The computer occupations, with about 3.7 to 4 million workers, comprises approximately half of the STEM workforce, or about 2 to 2.5% of the total workforce. According to projections of IT job growth by the Bureau of Labor Statistics (BLS), the total hiring demand is for, on average, 124,000 new IT workers each year, of which there is demand for about 40,000 computer science (CS) graduates, or about two-thirds the number of CS graduates each year. Despite all evidence consistently showing college graduate supply exceeding industry hiring, Congressional testimony by Microsoft’s Washington representative and counsel, Brad Smith, statements by the trade organization Code.​org and echoed by the Computing Research Association, assert that college graduate supply of CS graduates is inadequate (Harsha 2014). In a notable misstatement of the BLS projections, Brad Smith testified before the Senate Committee on the Judiciary in 2013 that “The Bureau of Labor Statistics has projected approximately 122,000 new job openings each year in computing occupations requiring at least a bachelor’s degree through the end of this decade. Yet nationally, our universities are only producing approximately 51,000 bachelor’s degrees in computer science each year” (Smith 2013). In fact, as clearly stated in the BLS projections, these openings are for computer occupations at all education levels and fields of study, of which about one-third are for those with at least a bachelor’s degree in any field (Landivar, 2013). Annual computer science graduation, which has grown from 38,500 in 2011 to over 65,000 bachelor’s graduates in 2016, and from 19,000 to over 40,000 master’s degree graduates in 2011 and 2016, respectively, produces 40-50% more graduates than needed to meet demand for new IT workers. Nevertheless, these IT industry claims are used in the widely repeated and unsupported claim of a supply shortage and of the inability of our education system to keep pace (Salzman 2013; Teitelbaum 2014).

Since 1972 women have constituted between 42 and 48% of all bachelor’s-level mathematics graduates and have gradually been increasing their proportion in graduate programs. In recent years, women received 40% of the 6000 to 7000 master’s degrees in mathematics awarded each year, but only 29% of the 1700 to 1800 PhDs in mathematics.

As a hypothetical example, if two groups score 260 and 290, respectively, of which schools can affect 20% of the performance and schools of both groups achieve 15% improvement annually for 10 years, the gap would actually increase from 30 to 39 points; if the lower group’s schools improved by 20% annually and the higher (290 base score) group’s schools improved by only 15% annually, after 10 years the gap would be reduced to 8 points.

There is an extensive body of research examining these issues and we are not suggesting that schools are ineffective but, rather, expecting schools to have large impacts on education outcomes overall, or on specific differentials such as racial achievement gaps, is unsupported by the research.

The PISA organization does not appear to try to correct these misrepresentations of their findings and, in fact, often appears to promote these unsupported assertions in presentations and dissemination materials such as press releases; in particular, the rankings that are presented as meaningful ordering of performance levels fail to note that rather than a rank ordering of countries, the results should differentiate only between statistically significant differences. In other words, a “statistical tie” is, analytically, the same ranking. Using the statistically appropriate comparisons, the United States places in the second-ranked achievement group, though even that is a flawed measure, as discussed here.

As Ramirez et al. (2006:15) explain: “...much of the achievement ‘effect’ is not really causal in character. It may be, rather, that nation-states with strong prodevelopment policies, and with regimes powerful enough to enforce these, produce both more economic growth and more disciplined student-achievement levels in fields (e.g., science and mathematics) perceived to be especially development related.”