One wonders how a new paper, “STEM crisis or STEM surplus? Yes and yes,” in this month’s issue of the Bureau of Labor Statistics’ Monthly Labor Review, was deemed to have met the high standards of that publication. It consists of little more than a small amount of data, an unfulfilled promise of a sophisticated model, a few interviews of some probably biased sources, and the random musings of the authors. Even the authors’ theme, as summarized in the paper’s title — that there is much diversity among the STEM fields and thus the very question of whether there a shortage in “STEM” — is hardly new.
In addition, in my opinion the authors have not been completely forthcoming about their possible lack of impartiality in this research. In particular, the first author, Yi Xue, describes herself as a “former MIT grad student,” which though correct, fails to disclose that she works for Palantir, a firm that has hired a number of H-1Bs in the area of Big Data — a field that the paper claims as having a shortage. Indeed, from the spelling of her name in China’s pinyin system, and her LinkedIn page, she appears to be of either Chinese or Canadian citizenship, and thus a foreign worker herself. This would color her views, not necessarily in terms of overt employer-related bias, but also in terms of limited viewpoint: Many foreign tech workers see so many people like themselves being hired that they falsely assume that no Americans are available for those jobs.
My interest in the paper stems (no pun intended) from the authors’ repeated claim that though some STEM fields don’t have shortages, there is definitely a shortage in the software development area. More on this shortly.
The authors begin with a rather promising statement:
Although many studies have examined the science and engineering workforce in the aggregate, little analysis has been aimed at identifying specific areas of STEM worker shortage or surplus. Using a “taxicab queuing model” as a framing metaphor, this article examines the heterogeneous nature of STEM occupations by studying distinct STEM disciplines and employment sectors on the basis of current literature and statistical data, as well as anecdotal evidence from newspapers.
(The second author is a specialist in queuing theory.) But actually the authors don’t use that taxicab model in their paper at all! The hapless reader must wait until the Conclusion section at the end of the paper to learn that the authors finally admit that they never did use that taxicab model (or any other, for that matter). This reason alone should have been enough for the journal to reject the paper, or at least to insist that the authors not make such a misleading claim at the outset of the paper.
The authors’ primary source is interviews with recruiters. They concede that their sample size there was small (18), but that is not the real problem. Instead, the issue is that the authors don’t realize that when a recruiter tells them that he/she has trouble finding software developers, the authors don’t know the unspoken restrictions that are controlling the recruiter’s search. As I’ve mentioned, a big issue is age; the recruiter knows that the given job is open only to young programmers, either new grads or up to 5-10 years out of school, and thus his/her statement “I can’t find enough programmers” really means, “I can’t find enough YOUNG programmers.” Or worse, the recruiter knows that the employer wants “loyal” workers who won’t jump ship to another company, a euphemism for foreign. The authors of this paper aren’t aware of these dynamics.
The only quantitative support the authors offer for their claim of a programmer shortage is indirect, pointing out for example that programmer salaries are higher in states with big tech industries, such as CA, WA and TX. Fine, but there are lots of confounding factors there. The authors also cite a much-criticized job ads study by a think tank funded by the industry.
The biggest problem with this paper is that the authors ignore “the elephant in the room” — wages in the IT field aren’t rising. They cite the Salzman/Kuehn/Lowell paper, but ignore the latter’s finding that IT wages have been flat. You don’t have to be an economist to understand the basic principle: Flat wages means no shortage, period. Had the authors done their homework, they would have found that if anything, IT wages are declining. (See my January 23 blog post, “New CS Grads’ Wages Down 9%.”)
As someone who has often taught queuing theory, I look forward to seeing future work by the authors using their taxicab model. But I hope they do their homework this time.