With the high-skill immigration debate about to heat up again thanks to Senator Orrin Hatch’s (R–UT) 13 January introduction of the Immigration Innovation Act of 2015, nicknamed “I-Squared,” we are likely to hear familiar arguments for raising the numbers of foreign science, technology, engineering, and mathematics (STEM) workers admitted to this country. One of the most influential of those arguments has been the claim that foreign STEM workers increase employment for natives. The advocacy group Compete America, composed of several dozen major employers of STEM talent, proclaims on its Facebook page, for example, that “each foreign born worker with an advanced STEM degree” from a U.S. institution “creates” 2.62 jobs for Americans.
A recent reanalysis casts doubt on the source of that claim, according to University of California (UC), Davis, statistician and computer scientist Norman Matloff. The 2.62 figure comes from a 2011 study by economist Madeline Zavodny of Agnes Scott College in Decatur, Georgia, which was sponsored and published by the American Enterprise Institute and the Partnership for a New American Economy, an advocate of increased high-skill immigration. Well known as a critic of STEM shortage claims, Matloff writes that an independent researcher named R. Davis has found a defect in Zavodny’s work in the form of a statistical oddity called Simpson's paradox, which can lead to drawing incorrect inferences about correlations. Zavodny, meanwhile, stands by her work. “No, there isn't an error,” she tells Science Careers by e-mail.
What conditions increase employment, and how can those conditions be created or encouraged?
Even before the new analysis there was a reason to question the study’s implication of causality, as Zavodny herself acknowledges in the e-mail: “I agree that the question of causality is fundamental.” That is a statement beyond dispute, she writes.
The points at issue are mostly too technical for this column (and this columnist). The issue that really matters to scientists, STEM workers, and the economy at large is the meaning of the correlations: In a market where large numbers of very smart, well-educated scientists and others are struggling to find work commensurate with their abilities and extensive training, how are new jobs, both high skilled and otherwise, created? What conditions increase employment, and how can those conditions be created or encouraged? Those policy issues require more research by researchers from a range of disciplines.
Correlation and causation
“Immigrants with advanced degrees boost employment for US natives,” Zavodny declares in her study. Specifically, “[t]he data comparing employment among the fifty states and the District of Columbia show that from 2000 to 2007, an additional 100 foreign-born workers in STEM fields with advanced degrees from US universities is associated with an additional 262 jobs among US natives.” This correlation (noted in the latter statement), enhanced by its implication of causation (in the former) has gotten a lot of traction; a New York Times columnist even prescribed additional STEM immigration as a cure for the nation’s persistent unemployment.
How this purported connection, which other economists dispute, might work in the real world rather than the world of correlations has never been clear to me. None of the scholars I’ve asked have provided a convincing explanation.
It’s not that foreign-born scientists can’t create jobs. As I’ve mentioned before, my own father, the late Morris Lieff, Ph.D., was one who did. In keeping with the best only-in-America mythology, he earned patents that became the basis for a business that created dozens of jobs. Utilizing American capital and market know-how—it could have come from anywhere—my father and his colleagues turned an idea born in a Chicago lab into factories in New Jersey and Virginia and, ultimately, paychecks for him and others. How could his place of birth have had anything to do with this?
If the employers who hired the foreign STEM workers in Zavodny’s studies had instead hired comparable Americans (or kept on the older workers who often lose their jobs to younger ones, especially in the computer industry), it seems unlikely that the economic effects would be any different. The kind of correlation—not necessarily causation—that Zavodny’s study suggests could merely show that when jobs are available, people both native and foreign get hired. An unwarranted leap to causality could mask effects that some researchers argue decrease jobs for American STEM workers. What we need is less rhetoric and more study of how exactly job creation works so that we can adopt policies that encourage more of it.
Solving a paradox
Simpson’s paradox, Matloff explained in an interview, arises when the relationship between two variables is “badly distorted or reversed” because “there’s another variable that’s relevant but not being factored in” or given proper weight. A well-known example in academe is a 1973 sex discrimination suit against UC Berkeley alleging discrimination against women in graduate school admissions. For the graduate school as a whole, 44% of male but just 35% of female applicants won admission.
Department-by-department analysis, however, revealed a “small but statistically significant bias in favor of women,” wrote UC Berkeley’s P.J. Bickel, E.A. Hammel, and J.W. O’Connell in Science in 1975. Women, it turned out, tended to apply to departments that attracted more applicants and thus had more competitive admissions. Somewhat counterintuitively for us in the science world, “[t]he graduate departments that are easier to enter” and attract higher percentages of male applicants “tend to be those that require more mathematics in the undergraduate preparatory curriculum,” Bickel and co-authors note. The missing variable was the relative competitiveness of the individual departments.
Zavodny’s inquiry “asks whether having a higher share of workers who are foreign born in a given state increases or decreases the employment rate among US natives in that state. … A positive relationship would indicate that more immigration creates jobs for US natives, while a negative relationship would indicate that more immigration decreases employment for US natives.”
But, looking at a graph Davis made from data Zavodny readily and generously provided (“Foreign STEM Workers, 2000-2007”), “you see the dominant variable is not the number of foreign workers. … The dominant variable is the state,” Matloff says in our interview. Because “the data for each state” is “essentially flat,” he adds by e-mail, “you can see that there is essentially no relationship between jobs and foreign workers. … But there is a big difference from state to state.” This means, he says, that something else is at work, possibly including culture, atmosphere, or economic conditions within the individual states, an idea that other research also supports. By far, the largest numbers of foreign STEM workers and jobs are in California. The workers may very well be there because of the jobs—and, Matloff suggests, industry hiring policies—rather than vice versa.
Those interested in looking more deeply into Zavodny’s and Davis’s studies and correlations can find them here and here, respectively. Even more useful, though, would be serious, creative, open-minded research into the circumstances, conditions, and factors that lead to job openings and people getting hired. Those conditions probably involve giving smart, creative, technically trained people the resources they need to pursue their best ideas. It’s hard to see why it matters whether those people were born in the United States or overseas; surely there are plenty of good people with good ideas to be found in the United States and abroad. We need to understand better how to translate those ideas into good jobs and innovations.