To: H-1B/L-1/offshoring e-newsletter Mon Jul 1 21:48:59 PDT 2013 Earlier this year I noted the release of a working paper by Kerr, Kerr and Lincoln (KKL), "Skilled Immigration and the Employment Structures of U.S. Firms," available at http://www.people.hbs.edu/wkerr/Kerr_Kerr_Lincoln%20Feb2013.pdf I stated that I would prepare a detailed analysis of the paper, then run it by Bill Kerr for comments, then post a revised version of my analysis here in this e-newsletter. I am now finally carrying out this plan, and it is a pleasure to do so. Bill is one of the nicest people I've met on either side of the H-1B issue, and I highly appreciate his willingness to take part here. You can read Bill's rejoinder in full at http://heather.cs.ucdavis.edu/Archive/BillKerr.pdf My posting below consists of remarks I originally made in a draft I sent to Bill on Sunday night, but now with some modifications in response to Bill's rejoinder, and some new material here and there. My comments follow below. Since they are rather involved, here is a summary: * KKL use a controversial statistical technique that I am skeptical of. Thus even though some of KKL's findings support my writings on the age issue, I do not embrace those findings. * Concerning the age issue, KKL's data only involves what happens to workers after they are hired. The real issue is WHO is hired in the first place, i.e. whether most well-qualified older workers tend to be rejected. * KKL's scope is far too broad, mixing together very disparate groups. For example, under the category "immigrants" they include all foreign-born, e.g. workers who came to the U.S. as children, rather than just H-1Bs. This can induce serious distortions. * KKL claim to quantify the number of jobs created by immigrant workers begs the question--what kind of workers creates more jobs, natives or immigrants? If it is the former, then there would be a negative implication for skilled-immigration policy. * KKL fail to take into account the long-term displacement effects of the H-1B program. Note by the way that my comments also apply, along with additional and much stronger criticisms, to the recent paper by my UC Davis colleague Giovanni Peri, which has gotten a lot of attention in the press: "STEMWorkers, H1B Visas and Productivity in US Cities," http://economics.ucdavis.edu/people/gperi/site/papers/peri_shih_sparber_05_feb_2013_TEMPO.pdf One technical issue: The terms "immigrant" and "native" arise a lot below. In my own papers (with the exception of my EPI paper, for reasons explained there), I try to avoid comparisons of natives to others. However, that is KKL's setting, and the term will arise frequently here. Preamble: Even though the KKL paper is easily available on the Web at the URL given above, Bill felt it would be useful for me to begin with a summary of the paper, say by posting the abstract here amid my comments. I think it is more helpful to post the paper's Conclusions section (though I will excise some material I don't think important to the discussion here): "In summary, the results of this study provide a multi-faceted view of the impact of young skilled immigrants on the employment structures outcomes of U.S. firms. We find consistent evidence linking the hiring of young skilled immigrants to greater employment of skilled workers by the firm, a greater share of the firm's workforce being skilled, a higher share of skilled workers being immigrants, and a lower share of skilled workers being over the age of 40. Results on whether total firm size increases or not are mixed. There is also consistent evidence in our IV specifications that older native employment expands very little, which is different from the other employment groups. Unlike this lack of growth, however, there is limited evidence connecting actual departures of workers to the hiring of young skilled immigrants. The closest connection is a relative statement across occupations within a firm that suggests that departure rates for older STEM occupations may be higher. "...substantial portions of the U.S. immigration framework like the H-1B visa program have been designed to allow U.S. firms to choose the immigrants that they want to hire. Given this system and the fact that the size of the program is determined by legislation, it is imperative to understand the motivations of the firm and the economics of skilled immigration within the firm. Our results have important implications for the competitiveness of U.S. firms, the job opportunities of natives and immigrants employed by the firm, our larger national innovative capacity (e.g., Furman et al. 2002), and much beyond. We hope that future work continues in this vein. Rough translation: (a) Skilled immigrants in a firm create jobs in that firm. (b) Those jobs tend not to go to older natives, and may even result to a mild degree in pushing the older natives out. (c) The positive effects of the skilled immigrants have important implications for Congress to consider. Here is my critique: 1. Problems with breadth of scope. I have long stated my reservations about the very broad scope of many of the papers of this type. See for instance my review of the otherwise-excellent work by Prof. Jenny Hunt, http://heather.cs.ucdavis.edu/Archive/JenniferHunt.txt Mixing together lots of different occupations can mask some effects, dilute others and produce contradictory results. I've mentioned before the famous Simpson's Paradox (regression version), under which the algebraic sign of a relationship can change when data is amalgamated. This problem was once neatly summarized in the clever title of a medical research paper, "Good for Men, Good for Women, Bad for People." There a certain medical treatment was seen to have a beneficial effect within each gender group, but to have a negative impact once the male and female data were combined. KKL's scope is especially broad. Not only does it combine lots of occupations, but also defines "skilled" and "immigrant" extremely broadly: KKL define "skilled" to mean any worker making over $50,000 in 2008 dollars. They define "immigrant" to simply be non-native. KKL do make appropriate disclaimers, yes, but since their paper bills itself as aimed at informing public policy, especially on H-1B and related visas, I consider the paper's disclaimers to be woefully inadequate. The public debate concerns foreign nationals on work visas, NOT workers who came to the U.S. as children and are now either permanent residents or naturalized U.S. citizens. And the public debate primarily concerns tech workers, NOT marketing executives or financial analysts. Congress and the press will take KKL's research to have implications specific to the hiring of engineers on temporary work visas--even though KKL's work does NOT do this. People in Congress and the press are not going to read the disclaimers in the paper. Even more importantly, KKL omit what should be an obvious analysis. They spend 46 pages measuring what they claim to be the effect of hiring immigrants is on a firm's subsequent hiring. The obvious question is then what the effect of hiring NATIVES is, compared to hiring immigrants. If, for example, the native impact on hiring is greater than the immigrant one, then policymakers should be reluctant to expand foreign skilled immigration in any long-term sense; indeed, in such a scenario, H-1B would bring net harm to the economy. I asked Bill about this omission back in May. He replied, "The main rationale for pursuing the immigrant hiring calculation [but not] the native hiring is that the former is more proximate to a policy action and its consequences than the latter is. Given that the 'creates 4 jobs' statements are at the center of the policy debate, they will tend to get the most attention from economists." Proximate, schmoximate, I strongly disagree. The policy debate focuses in large part on the industry lobbyists' claim that the foreign workers are especially innovative, "the best and the brightest," and thus have special job-creating powers. It would thus be natural, indeed imperative, to compare the job-creating powers of the immigrants to the job-creating powers of the natives. Arguably, KKL's failure to do so suggests an unconscious bias on their part. 2. Fragility of the models. I am a former statistics professor, and though I haven't been a member of a Statistics Department for a long time, I continue to teach and do methodological research in the field. My specialty is regression analysis, a tool that pervades the entire KKL paper. I use regression analysis in my own research on H-1B, but I know that it is a very fragile tool, as Simpson's Paradox often vividly shows. Part of KKL's paper uses a particularly fragile form of regression analysis, using something called "instrumental variables." IVs are quite popular among economists, but are rarely used elsewhere. Even among economists IVs are controversial. See http://online.wsj.com/article/SB117131554110006323.html for a popular account of the controversy, and http://www.wiley.com/legacy/wileychi/eosbs/pdfs/bsa598.pdf for a technical one by the late David Freedman of UCB. At least the IVs enable KKL to attempt to narrow the definition of "immigrant" to H-1Bs, but again, the methodology is questionable. I won't critique KKL's choice of IV here, but just point out the fragility. Personally I've always been skeptical of IVs, but again, even ordinary regression is fragile. Thus, I always apply multiple approaches in my work on H-1B, rather than relying only on regression. This is exemplified in my recent paper in the academic journal Migration Letters, http://heather.cs.ucdavis.edu/MigLtrs.pdf in which among other things I adddress the question of whether H-1Bs are underpaid. I use several approaches to that issue, and in fact none of them use regression. In my recent EPI paper, I do use regression but again do a number of nonregression analyses as well. Given that KKL's implied policy recommendations are at odds with other research (see below), I think the fragility of the models is a major problem with the KKL paper. 3. Lack of integration with other research. I teach my students and my consulting clients that regression analysis is most useful if its results jibe with other information one has. For instance, my research findings have shown that the former foreign students now working in the computer science field in the U.S. are on average of somewhat lower quality than their U.S. native peers. That jibes with the fact that many of the foreign students come from Asian cultures that emphasize rote memory, thus impeding insight and innovation. (The governments of China, Japan, Korea and so on have all worried about this.) Unfortunately, KKL do not attempt to perform such a check, which in this case would mean attempting to reconcile their findings with the research on the "hot button" questions in H-1B: Are H-1Bs paid fairly, relative to their market worth? Is there a tech labor shortage, the putative raison detre for H-1B? Are the H-1Bs typically "the best and the brightest," as claimed by the lobbyists? There is quite strongly negative research regarding all three of these questions, so KKL's upbeat evaluation of H-1B produces a disconnect. KKL should have addressed that disconnect. For concreteness, let's focus the underpayment issue. This is the clearest, as it can be essentially settled by a "thought experiment." As I've pointed out before (and conceded even by some strong advocates of the H-1B program), H-1Bs who are being sponsored for green cards (and to some extent many others) are de facto immobile, and immobility means they can't get the best deal out there in the labor market. Therefore, on average they are underpaid relative to their true market value. So basic economic theory, not to mention plain common sense, establishes the underpayment issue. And I believe KKL, as economists, would agree with this analysis. Then how would they reconcile this with their paper's findings? Of course, there is no direct contradiction between this and KKL's findings. Nowhere in their paper do they claim the H-1Bs are paid fairly. However, again, the takeaway from KKL's paper by Congress and the press will be that "H-1B is good, they create jobs etc. so the naysayers are wrong e.g. in claiming the H-1Bs are underpaid." Bill's comment to the NYT last week, “People take an extremely one-sided view of this stuff and dismiss any evidence to the contrary,” didn't help in this regard. :-) More subtly on this same topic, KKL's findings on job creation, coupled with their presumed agreement that there is an underpayment problem, would seem to imply that they believe cheap is good! Employers get more bang for the buck! But that implies that the workers are mere commodities, no variation in quality, starkly contradictory to KKL's own claim (and Bill's claim in his paper on H-1B patenting) that H-1Bs are major contributors to innovation. One cannot optimize two variables at once. The more weight an employer puts on cheapness in hiring, the less weight it puts on quality. Of course, this also relates back to KKL's failure to compare job-creation powers of immigrants and natives. This is absolutely crucial. Similarly, if H-1B is such a boon to native skilled workers, why have their wages been flat for the last 10+ years, as shown by other research? KKL need to face this issue. In Bill's rejoinder, he dismisses these things as being beyond the scope of his paper. I disagree. The questions I'm raising here could (and should) arise in the peer reviewing process for the paper, and once again, a paper that is aimed at informing policymakers is not very helpful if it doesn't address these questions. Mind you, I'm not saying that KKL should perform additional analysis; but they should at least include a section that brings up the apparent disconnect, with a bit of discussion. 4. Failure to capture long-term structural effects. One big point that KKL miss is the fact that H-1B and similar programs have, in the last two decades or so, dramatically changed the way tech employers hire--and even the way they THINK about hiring, their basic premises. I mentioned this in my analysis of Jenny Hunt's paper cited above, I stressed that there has been a gradual decline in the number of American students pursuing graduate study in STEM due to displacement by foreign students, who brought down wages and thus made graduate work unattractive to Americans. (This was forecast, if not actually advocated, by our federal government's National Science Foundation.) Jenny had written, "...economists have an as yet incomplete picture of the aggregate benefits to natives of skilled immigration. In this paper, I address this by providing evidence not merely on skilled immigrants' private productivity, as measured by their wage, but also on their success in creating, disseminating and commercializing knowledge, activities with public benefits likely to increase U.S. total factor productivity...The subanalyses here in different visa categories make] the results directly informative to policymakers, who can use them to influence their decisions about which visa classes to expand or shrink and which transitions to legal permanent residence to facilitate." I continue to regard the Hunt paper as one of the best that has been written on the subject, but look at the mindset above--that immigration can only produce positive results, the only question being the degree of positiveness. My objection was that one can't measure the impact of skilled immigrants without taking into account the DISPLACEMENT of Americans from STEM, such as that correctly forecast by the NSF. Indeed, I've argued, e.g. in my EPI paper that at least in the field of computer science, the average quality of the immigrants (former foreign students) is lower than that of their American counterparts--so that the displacement means a NET LOSS to the U.S. economy. So displacement MATTERS. Again, the NSF position paper in 1989 correctly forecast that H-1B (then on the drawing board) would result in fewer Americans going to grad school in STEM. Many companies now take it for granted, without realizing it, that they are going to hire many or most of their R&D workers from the foreign student pool. This is a major structural change, but it occurred gradually, so it would have been difficult or impossible for Jenny's analysis to detect the impact. And the impact is definitely there. Since many companies tend to assign those with grad degrees to R&D work, and since the latter produces new jobs, patents and so on, the failure to account for this structural change greatly reduces the value of the research. And it certainly reduces the validity of the claims, such as quoted above, that the research should be useful for policymakers. The same thing is true for age. I applaud KKL for investigating the connection between hiring of immigrants and reduced job opportunities for older (age 35+) Americans. KKL found some tentative indications of an adverse impact, but nothing big, whereas I claim the real impact is indeed large. Why didn't KKL's work pick that up? Part of the reason is likely related to the methodological flaws I cited earlier. But there is a much more important reason, similar to the grad-school displacement I mentioned above: Over the years, the use of H-1Bs to avoid hiring older American workers has become entrenched, indeed institutionalized in the tech industry. HR departments routinely earmark positions with labels New College Graduate and Recent College Graduate. Though of course there sometimes are exceptions, basically the fundamental hiring structure is now set for the young. You may recall a Microsoft quote stating this, that I cite often. If employers run out of young Americans to hire and run out of H-1B visa allotments, they do not resort to hiring the older Americans. This is quite a contrast to the old days, in which experience was highly valued. Accordingly, the short-term demand for older workers is rather inelastic with respect to the number of H-1Bs, even though the long-term impact has been severe. So of course KKL won't detect much in their relatively short-term study. (By the way, KKL didn't cite any of the statistical studies that show the problems older engineers face in the job market.) Just as the industry has gotten into the habit of hiring the young, it has over the years also gotten into the habit of hiring Americans mainly at the bachelor's level. As I've mentioned before, staunch H-1B supporter Texas Instruments has publicly stated this, But this aspect works in ways one might not first think of. During good times for a firm, they greatly raise starting salaries for the new bachelor's graduates--giving the latter a great incentive to NOT pursue graduate study. Then if the firm concentrates its R&D on those with graduate degrees, it would appear in analyses like KKL's patents, job creation etc. would not have occurred if there had been no H-1B program. Again, I assert the opposite--that we would have a stronger tech industry today without H-1B. In short, KKL's paper does not/cannot detect harmful aspects of H-1B that have built up gradually over the years, via changes in hiring patterns related to H-1B. Again, I'm not saying KKL should have done more analysis in this regard, but they should have recognized it. Bill did use the qualifier "short term" in his remarks to the NYT. But the public policy changes being considered will be LONG-TERM if enacted. Again, since the takeaway from KKL's paper by Congress and the press as giving the green light to those long-term changes as being beneficial. Given that the paper explicitly bills itself as being aimed at informing public debate, KKL should have discussed potential long-term negative effects. I'm sorry for harping on this point, but clearly it's important. (Note similar phrasing in the Hunt quote above.) 5. Lack of accounting for the skills factor: Employers have said repeatedly that they hire H-1Bs because they possess "hot" skills that Americans lack. But such skills bring a wage premium of 20% or more in the open market. Any analysis that does not account for skills is thus off by something like 20%. 6. Restriction to employees: The authors had a great idea here, to conduct a study at the firm level. But it only allows one to look at the workers who are actually there, not the job applicants. In any discrimination study, one must look both at actual workers AND job applicants. To investigate possible gender discrimination, one can study the female workers in a company, but the bigger question is whether the firm is rejected qualified female applicants in favor of males. In the case here, it's fine to measure the impact of hiring young foreign workers on the older American workers, but again, the larger question is whether young foreign workers are hired instead of qualified American applicants. 7. Conclusion. KKL's paper makes a contribution in that it is focused on the firm level. (Some of my analyses have been on that level too.) It makes interesting use of a new data set, and is to be lauded for attempting to confirm the adverse impact of H-1B on older American workers. However, the paper uses just one methodology that even advocates concede can be misleading. The paper's rosy implications do not jibe with a number of other statistical research papers, and KKL make no attempt to reconcile, or even recognize, such clashes. In my opinion, KKL substantially overstate the relevance of their findings to formulation of public policy. Norm Archived at http://heather.cs.ucdavis.edu/Archive/KerrKerrLincoln.txt