In my last post, on the NAS conference, I did not mention the survey paper , “International Migration and U.S. Innovation,” by Bill Kerr. He gives a good, careful overview of the research literature in the field, for example taking care to point out that some of the papers were “done for advocacy purposes.” One small quibble is that he uses the doubt-inducing word claim for my work, while using the approving word find for all the other papers he cites, including his own. 🙂 Kerr finds, Peri finds, etc., but Matloff only claims. Ouch!
Bill begins his Conclusions section with this:
Immigrants are of deep importance to U.S. innovation. This is most evident in terms of their sheer quantity for STEM work in the United States, and the disproportionate number of superstars who are immigrants speak to this.
Regular readers of this blog know that although I’ve often expressed my great respect for Bill and his coauthors — they definitely among the most thoughtful researchers in the area of foreign tech workers — I still often disagree with their findings/methods, even in cases in which their findings jibe with my own.
I’ve written plenty, e.g. in my EPI paper, on what Bill describes as “their sheer quantity” before — as forecast 25 years ago by the NSF, the influx of foreign STEM specialists held down PhD wages and therefore drove out the Americans from STEM graduate work — but Bill’s mentioning “he disproportionate number of superstars who are immigrants” caught my eye. This aspect will be the focus of my post here this evening.
I should remind readers that I have always strongly supported facilitating the immigration of the superstars (including those with outstanding promise). But Bill’s word disproportionate is provocative. While I’ve shown in various ways that most H-1Bs, especially those who first come to the U.S. as foreign students (the industry’s prized group, according to the lobbyists), are not “the best and the brightest,” I haven’t done much specifically on rates of superstardom. What I’ll do here is review my work in that regard for computer science (CS), and then follow with some new data.
In any such analysis, the key aspect is properly defining the numerator and denominator implied in use of the term disproportionate. Take a common example, Nobel laureates. What should we use for our denominator? Certainly not the general population. Maybe the population of those with STEM bachelor’s degrees? Master’s? PhDs? PhDs in academia? You can see the problem.
I will not address other superstardom papers in this posting (will do so later if I see it worthwhile), but in the case of my EPI paper, I believe my choices of numerator and denominator were quite reasonable: number of dissertation awards / number of dissertations, referring to the very prestigious ACM Dissertation Award. There I found that foreign students in North America received 48% of the awards, and that 48-51% of all CS dissertations were done by foreign students. (One of the awards was in Canada, and that one went to a foreign student.) In other words, the foreign students were no more and no less prone to win this award than the Americans.
So, Bill’s term disproportionate does not apply in the case of CS. What field to look at next? I have generally restricted my analyses of H-1B-related issues to CS, because that is my own field. I’ve often criticized work on both sides of the H-1B debate on the grounds that although the researchers are fine number crunchers, to them the numbers merely concern widgets, without understanding what these workers really do. I believe that my own contribution stems largely from the fact that I do have such insight.
Accordingly, I’ve now done something similar to my CS dissertation analysis, this time turning the field of Statistics. That was my original research area — I was one of the founding members of the UCD Department of Statistics — and much of my CS work has been statistical in nature. So I know what Statistics researchers do, who is outstanding in the field, and so on.
And as with CS, there is a handy data set to work from, The Committee of Presidents of Statistical Societies (COPSS) Presidents’ Award. COPSS’ description is quite accurate:
The COPSS Presidents’ Award is given annually by the Committee of Presidents of Statistical Societies to a person under the age of 40, in recognition of outstanding contributions to the profession of statistics. It is arguably the most prestigious award in statistics, the Nobel Prize of Statistics.
These are brilliant, truly outstanding people, many of them quite prominent in today’s Big Data and Machine Learning crazes. They are professors at places like Harvard, Stanford, Berkeley, CMU and so on. The U.S. is lucky to have them, natives and immigrants alike.
I limited my analysis to those who earned their PhDs in the U.S., again reflecting the industry lobbyists’ emphasis on the foreign grad students at U.S. universities. The results were that foreigners got 33% of the awards, but form 50% of the PhDs granted (the latter number from the NSCG data).
In other words, in the Stat field, foreigners receive a disproportionately SMALL number of awards.
I’ll have more to say on this in future posts.