Date: Sat, 25 Apr 2009 18:27:15 -0700 From: Norm Matloff To: Norm Matloff Subject: H-1B study withdrawn from public, authors not talking? To: H-1B/L-1/offshoring e-newsletter (Note to readers: This will be a long posting, but I believe you will find it well worth reading in its entirety.) Last week two researchers, Prasanna Tambe of the NYU Stern School of Business and Lorin Hitt of Penn's Wharton School, released a working paper that found that the presence of H-1Bs lowered wages for Americans by as much as 6%. (The phrasing here is misleading and not what was intended by the authors. More on this below.) Computerworld picked it up, with an article at http://www.computerworld.com/action/article.do?command=viewArticleBasic&articleId=9131729 But then things were suddenly shut down. According to the CNN broadcast transcript below, the paper was suddenly withdrawn from public access at the Social Science Research Network (www.ssrn.com), and the authors are no longer giving interviews. I will give my analysis of the paper below and it will be the bulk of this message, but first, what about the question raised by the Dobbs show? Was the paper suppressed due to pressure from the vested interests (industry, the American Immigration Lawyers Assocation) on the authors and/or their institutions? Subjecting the paper to academic peer review is of course the proper thing to do, as review will uncover flaws, produce improvements, and so on. But this can take a year or more (I once had one languish for three years), so it is customary to share the work informally in the mean time. In the social sciences, this is done through "working papers," informal drafts. The authors apparently placed their work on SSRN as a working paper for just that purpose. It is indeed quite odd that they suddenly withdrew it from the site, leaving only the abstract. The Dobbs show was correct to ask about the funding of the research, but pernicious influences can be more subtle than that. Universities, especially schools of engineering and business, want to keep good relations with industry and the business world. It's no accident that at Stanford the computer science building is named Bill Gates Hall (I noticed one at MIT last week too), with Hewlett Hall and Packard Hall across the street. I'm proud of the support my own university has given me for my rabblerousing on the H-1B issue, even profiling me (and five other faculty) in a magazine article, cleverly titled "Informed Dissent" (http://heather.cs.ucdavis.edu/Archive/UCDRecognition.txt). But other universities, especially private ones, may not be so tolerant. And Dr. Tambe, as an untenured assistant professor, could be especially vulnerable. So the sudden withdrawal of the paper--and its authors--from public access is troubling. There could be innocent explanations, such as the authors realizing that some aspects of the paper need major reworking, but it does seem odd. Now, what about the content of the paper itself? I was given a copy by someone who had the foresight to download it immediately upon its release, and do have some comments. You might ask why the authors would even study the topic in the first place. Isn't it a simple matter of supply and demand, with a larger supply causing lower price? As a general principle, yes. The 2000/2001 congressionally-commissioned National Research Council report, whose authors included several labor economists, stated % ...economic theory implies that an increase in the supply of IT workers, % including temporary nonimmigrant workers, will cause the corresponding % IT wage rates to be lower than they otherwise would have been. Theory % alone does not imply any particular numerical magnitude of this effect. % It is the committee's judgment that the current size of the H-1B % workforce relative to the overall Category 1 IT workforce is large % enough to exert a nonnegligible moderating force that keeps wages from % rising as fast as might be expected in a tight labor market. Former Fed chair Alan Greenspan told Bloomberg News on March 14, 2007 that he was worried that IT workers' high salaries (which aren't that high) would be resented by the rest of the populace, creating social unrest, and recommended raising the H-1B cap: % Our skilled wages are higher than anywhere in the world. If we open up a % significant window for skilled workers, that would suppress the % skilled-wage level and end the concentration of income. The Fed has over the years consistently called for expansion of the H-1B program. And as I often mention, the National Science Foundation advocated bringing in a lot of foreign students to U.S. university PhD programs, for the express purpose of holding down PhD salaries, as the students flood the market when they graduate (and become H-1Bs). See http://nber.nber.org/~peat/PapersFolder/Papers/SG/NSF.html So government experts do believe that H-1B brings down overall wages in the fields in which the foreign workers are numerous. But it would be nice to quantify the effect, which the Tambe/Hitt paper attempts to do. (Note carefully, and keep in mind, that the 5-6% gap they found, in addition to meaning something different from what showed up in the press, to be explained later, does not mean that that is the amount by which H-1Bs are underpaid, which is a different kind of number. My analyses have shown a 15-20% underpayment for H-1Bs having the same age and experience level as Americans, more when various ages are considered, and other studies have found similar, and in some cases larger, gaps.) Furthermore, an NFAP study claimed that H-1Bs have a job-creation effect, which might actually raise overall wages by increasing demand. As I explained at the time, the NFAP study was statistical nonsense, and recently a Wall Street Journal analysis essentially agreed (http://heather.cs.ucdavis.edu/Archive/WSJOnNFAPClaim.txt). But still, the authors here do have to allow for the possibility that there is such a job-creation effect in spite of the NFAP study's flaws, so the goal of the Tambe/Hitt study makes sense. One major issue I have with Tambe and Hitt's paper doesn't directly involve their analysis, but instead has to do with a related point they make a couple of times. In asking the question as to whether American tech workers are harmed by the H-1B program, they pair this with a presumed benefit to the program, in the form of supposedly increased innovation and entrepreneurship. On page 3 they write # Recent work has demonstrated that guest workers and skilled immigration # have positive effects on both innovation levels (Kerr & Lincoln, 2008; # Hunt & Gauthier-Loiselle, 2009) and entrepreneurship (Saxenian, 2002; # Wadhwa et al, 2008)... They make this presumed loss/gain pairing even more explicit (though with a somewhat milder word, "suggest" rather instead of "demonstrated") in their Conclusions section, on page 21: # Although our findings suggest that the negative effects of globalization # may be substantial for some workers, it is critical that policy makers # weigh these effects carefully against the macro-level economic effects. # Offshoring will most likely remain a necessary and important part of the # global economy, and there is substantial evidence that H-1B admissions # appear to directly improve levels of innovation and entrepreneurship, # which in the long term should create new jobs and raise demand for # technology workers in other areas. I've used the word "presumed" above twice because the authors' presumption is false. None of the four papers they cite claims that H-1B or immigration has produced a net gain in innovation or entrepreneurship. Indeed, they could not have made such a conclusion, as they did not have the type of data needed to address that kind of question. The H-1B program has squeezed many U.S. citizens and permanent residents out of this field, and has discouraged many of our top young people from even entering the field. Thus one cannot tell, without analyzing this aspect, whether H-1B/immigration has produced a net loss or a net gain in innovation and entrepreneurship. So, I must say I found Tambe and Hitt's statements above to be disturbing. Concerning the main analysis of the paper, in many respects the authors are more careful than most. For example, they break their analysis down to the firm level, allowing them to correlate wages paid by a firm with the percentage of H-1Bs at the firm; this turns out to be very important. In the end, they run regressions of wages against H-1B share, offshoring share, education, experience, job title, industry, specific firm, etc. Accordingly, the interpretation given to their pay-impact figures in the press, and indeed in the paper's abstract, is incorrect. Instead, the figures correspond to a 10% increase in the share of H-1Bs in a firm (or in an occupation). For example, their statement on page 5, # ...the introduction of firm-effects into our models indicates that H-1B # employment decreases wages by about 5-6%, primarily for computer # programmers and systems analysts... is intended to refer to what would happen if the percentage of H-1Bs among IT workers of this type at a firm were go to from, say, 20% to 30%, i.e. this would produce a 5-6% reduction in wages at the firm. Imagine, then, the scale of the impact of the doubling, then tripling, of the H-1B cap that Congress enacted in 1998 and 2000, sunsetting in 2004. For lots of reasons, one cannot apply the Tambe/Hitt figures directly, but it should be very clear that THOSE TWO PIECES OF LEGISLATION CAUSED A MAJOR SUPPRESSION OF IT WAGES. (See also the Lemieux paper below.) The authors make one error common to many previous works, a potentially very serious flaw related to job titles. This is a key point, so please bear with me while I explain. Most computer-related H-1Bs are programmers, with job titles typically being Programmer or Software Engineer, occasionally System Analyst. (I will use capital letters for job titles here, with the term "programmer" being broader, i.e. one who does programming.) Failure to understand this has led to badly flawed research analyses. I'll start with the 2003 Zavodny paper, cited by Tambe and Hitt. (See http://heather.cs.ucdavis.edu/Archive/Fed03.txt for my more detailed analysis.) Zavodny, a researcher with the Fed, came up with mixed results. She found a possible positive relationship between the number of H-1Bs and overall IT wages (statistically significant under one set of assumptions though not under another), but also found a positive relationship between the number of H-1Bs and IT unemployment. (Note that Zavodny looked at the number of H-1Bs in a region, as opposed to the percentage of H-1Bs at a given firm.) Though as I will explain below, Zavodny's study was fatally flawed, it is worth quoting from her final conclusions at the end of her paper: % The results provide little support for claims that the pro- gram has % a negative impact on wages. However, some results do suggest a % positive relationship between the number of LCA applications and the % unemployment rate a year later. The failure to find an adverse wage % effect does not necessarily indicate that H-1B workers do not depress % wages but perhaps that any effect is difficult to find, as concluded % by previous studies. This is quite different Tambe and Hitt's unequivocal phrasing that Zavodny "found that an influx of guest workers improves wages and employment levels for the native IT workforce." Sadly, often researchers do not read the papers they cite, which may be the case here. I was also disappointed to see Tambe and Hitt describe my paper in the University of Michigan Journal of Law Reform as merely "anecdotal." In any event that is not the major issue with Zavodny's paper. Instead, the real trouble is that Zavodny states that her analysis involves "computer systems analysts and scientists and operations and systems researchers and analysts." That wording is a bit odd, but the point is that she is for the most part not picking up the programmers. In other words, her "H-1B" study is not about H-1Bs, which renders it irrelevant to the H-1B debate. Unfortunately, Tambe and Hitt look only at Computer Programmers and System Analysts, and the 5-6% reduction in wages they find resulting from a 10% swelling of the labor market by H-1Bs is for those job titles. They thus exclude a large, and even more important category, Computer Software Engineers. Around the mid-1980s, the title Programmer began to be replaced by the fancier name Software Engineer. That newer title became standard in firms that produce computer hardware or software, but many employers (including my university) continue to use the Programmer title. Either way, it's still programming. A worker could switch from company A to company B, doing work of the same level of sophistication, and yet have the title Software Engineer at A but be called a Programmer at B. And indeed, the same person might do the same work as a System Analyst at company C. On the other hand, those with Software Engineer titles tend to be paid more than Programmers, for various reasons, such as working with newer technologies and being employed by firms that spend more freely on IT. Lack of awareness of this point, and other problems with job titles, has led to much flawed research. The 2003 GAO study had such problems. (See http://heather.cs.ucdavis.edu/Archive/GAO03.txt.) Though their employer survey showed some actually admitting to underpaying H-1Bs, the statistical section of the study had mixed results, an error stemming in part from not distinguishing between Programmer and Software Engineer titles. The question then arises at to what effects, if any, Tambe's and Hitt's omission of Software Engineers had on their analysis. As I said, in terms of work done, Programmers and Software Engineers do the same kind of work, giving rise to a loophole: An employer can give an H-1B a Programmer title, quite legally, while giving his American workers Software Engineer titles-- and thus pay the H-1B less, again fully legally. See the case study, http://www.programmersguild.org/archives/howtounderpay.htm for a detailed look at how job titles can be used to legally underpay H-1Bs. As I've explained before, this is just one of many loopholes employers use to legally employ H-1Bs as cheap labor. Yet it suggests that in this kind of scenario, the gap between H-1Bs and their American Software Engineer coworkers would tend to be larger than the 5-6% found by Tambe and Hitt for American Programmers. And this job title issue may seriously compromise the goals the authors describe at the top of page 5: # ...differences across firms, our estimates are identified through # variation among the wages of workers within the firm, under different # assumptions about the degree of substitution among domestic and foreign # workers based on the job titles listed in H-1B applications or reported # by offshore workers. For example, we assume that domestic computer # programmers are closer substitutes to H-1B and offshore IT workers that # describe themselves as computer programmers in their job titles. In some # regressions, we also explore whether the wages of IT workers in an # incoming cohort are more sensitive to supply shocks than the wages of # existing employees... On the other hand, if the industry lobbyists' claims are correct that the H-1Bs are doing work for which qualified Americans are not available, i.e. in Tambe/Hitt's economist language "complementarities exist between domestic and foreign workers," then it may be useful to redo their analysis, this time for H-1Bs and Americans with Software Engineer titles. Again, I consider the evidence against the "complementarities" argument to be overwhelming, but this other job title would be something worth considering. The authors address the H-1B underpayment issue on page 14: # To begin exploring how H-1B employment and offshoring affect the wages # of domestic IT workers, in Figure 4 we overlay the wages of our domestic # IT workforce sample against the distribution of wages reported in H-1B # applications. Although the prevailing wage may not be an accurate # reflection of the amount H-1B workers are paid by employers, the average # wage listed in the H-1B application appears to be substantially less # than the average wage paid to the American IT worker. This comparison # does not control for job classification, education, or experience, and # therefore provides little evidence about whether H-1B workers are paid # less than their American counterparts, or whether H-1B admissions # depress wages for American workers. Comparing the two distributions, # however, indicates that unless employers are generally paying # significantly more than the prevailing wage for H-1B workers, H-1B # employees are either 1) primarily hired to fill positions towards the # bottom of the IT wage distribution, or 2) are paid less than their # domestic counterparts for the same positions. What are the authors referring to here? As they note, their data source here is only the Labor Condition Applications (LCAs). The LCAs are only applications for permission to hire an H-1B, whether the employer has a specific person ready or not. And in some cases the employer is asking for permission to hire multiple foreign workers. Thus the wage listed that will be paid can only be taken as a lower bound; when the employer finally hires someone, it could be at a higher wage. Stuart Anderson, who is the "NFAP" referred to several times by Tambe and Hitt and who, as I've shown before, is closely tied to the immigration lawyers, has often criticized John Miano's LCA studies on this basis. Anderson even claims that the actual wage is "usually" (his exact word, in his March 2006 press release) substantially higher than what is stated in the LCA. As with many of his statements, he offers no supporting evidence for that claim, and the data show otherwise. I've looked at the distribution of wages stated in the LCAs and compared them to corresponding percentiles in the INS/USCIS H-1B data (which are for real people, not just applications to hire as-yet-unknown individuals) and found that the two track quite well. Tambe and Hitt report good tracking for numbers of H-1Bs hired as well (page 9). I make this point in the last paragraph simply to support Miano's and Tambe/Hitt's use of the LCAs. My own research has not used the LCAs. Instead, I have used the Census data, the INS/USCIS H-1B data, and the DOL green card (PERM) data, all of which are for real individual workers. In all cases I have found a pattern in which the H-1Bs are paid less than the Americans. The same holds for Ong and Blumenthal's analysis of the Census data. In addition, there are the NRC and GAO employer surveys mentioned earlier, in which employers actually admitted to paying the H-1Bs less. Furthermore, economic theory (or, really, common sense) says that the relative immobility of the H-1Bs means they will be underpaid on average; if you can't shop around among employers for the best deal, you will likely not get as high a salary as you could if you were a free agent. In other words, the evidence is overwhelming that the H-1Bs are generally underpaid (again, quite legally in most cases). Tambe and Hitt's Figure 3 (which they erroneously refer to as Figure 4 above) really shows that quite dramatically, with the distribution of H-1B salaries being substantially to the left of that of their domestic counterparts. The authors are correct to caution against overinterpretation of the figure, for many reasons, but it certainly is striking. Also striking, though not mentioned by Tambe and Hitt, is that H-1B is by far the strongest factor in overall wage levels. For their regression (3), for example, their coefficient -0.758 for H-1B share is about 20 times larger than the one for education. In other words, amount of education just doesn't play much role in wages. This is greatly at odds with the industry lobbyists' claim that the industry needs to hire H-1Bs because they have graduate degrees. Though it might be worth adding an interaction term for H-1B share and education, the reality is that although having a graduate degree does indeed give one a boost in salary at the time of graduation and for a few years afterward, it does not do so after this time. Hence the relatively small coefficient for education. I should mention here the recent work by Thomas Lemieux of UBC (but for U.S. labor), "The Changing Nature of Wage Inequality," at http://www.econ.ubc.ca/lemieux/papers/changing3.pdf He analyzed the growth in wage variation within an occupation over time, roughly speaking, comparing what the top earners made in an occupation relative to the bottom earners, during the last two decades. He found that the variation grew a lot for doctors and lawyers, but was puzzled about the computer professions: # Occupations at the higher end of the skill distribution (doctors, # lawyers, professors, scientists, etc.) almost all do better than # average. It is also highly plausible to argue that workers in these # occupations mostly perform non-routine work. A more disturbing # pattern is that among high-skill occupations, the only ones that # experience negative relative wage changes are precisely the ones most # closely linked to the computer revolution. In particular computer # programmers and engineers both experience negative relative wage # changes, despite the fact ­indicated by the regression line--that # occupations with similar skills generally experience a large positive # relative wage change.13 The case of computer programmers is # particularly striking since this is the two- digit occupation that # experienced by far the largest increase in "relative demand" as # measured by its employment share that more than tripled (from 0.7 to # 2.2 percent) over this period. So while Figure 8 shows that high # skill occupations where workers perform non-routine work did very # well in terms of relative wages, it is difficult to understand why # occupations at the core of the computer/technological revolution did # relatively badly compared to occupations like lawyers of even # professors that are relatively more peripheral to the computer # revolution. One possible explanation for this phenomenon discussed # below in Section 5c is offshoring. It's interesting that he didn't think of H-1B as a factor. Granted, he is in Canada but that nation has its own H-1B program. He may have dismissed the idea on the grounds that physicians can come to the U.S. on an H-1B visa too. My own explanation for this latter point involves the age issue. As I've written extensively, in the computer field H-1B is largely used as a tool for avoiding hiring the older (age 35+) workers. As a result, programmers have much shorter careers than do, for instance, civil engineers. (See my University of Michigan paper for details.) Physicians do not have short careers, and thus H-1B has less of an impact on them. Finally, a couple of quick points. First, they dichotomize the situation in natives and foreign-born, which is incorrect and would be offensive to some. Many Americans were born abroad but became naturalized citizens. Some came to the U.S. as kids, while other came as adults (some of whom came originally as H-1Bs). But ALL are potential victims of the H-1B program. Second, with such a large sample, the calculation of statistical p-values is rather meaningless; almost anything will be "significant." All in all, it's a very interesting paper. http://transcripts.cnn.com/TRANSCRIPTS/0904/20/ldt.01.html LOU DOBBS TONIGHT Aired April 20, 2009 - 19:00 ET THIS IS A RUSH TRANSCRIPT. THIS COPY MAY NOT BE IN ITS FINAL FORM AND MAY BE UPDATED. LOU DOBBS, CNN ANCHOR: Good evening everybody. ... DOBBS: This broadcast has reported literally for years on the problems and the abuses with the H-1B visa program, that's the federal government program that allows tens of thousands of foreign workers into this country each year. The H-1B program benefits corporate America at -- without question at the expense of middle class jobs. Tonight a new study from two respected universities confirms what we have been reporting here, that the H-1B visa program not only cost American jobs, but it depresses American wages as well. Bill Tucker joins me now and has the story for us -- Bill. BILL TUCKER, CNN CORRESPONDENT: Well, Lou, there's a hitch to this report tonight. That report is no longer available online. All that is left is the abstract. The researchers declined our interview request, but a copy of the draft report obtained by LOU DOBBS TONIGHT, the researchers found that the use of H-1B visa workers decreases the wages of computer programmers, system analysts and software engineers by up to six percent. Offshore outsourcing decreases the wages of I.T. workers and managers by two to three percent. The study is a joint project between New York University Stern School of Business and the University of Pennsylvania's Wharton School. A spokesman for the Stern School says the report was taken down after being posted for five days because they were not prepared for the media attention. And a decision was made to instead put the paper up for peer review, not public scrutiny. The researchers say the research was undertaken because the policymakers need to understand the importance and the impact of the H-1B visa program. And Lou, we should point out that one of the researchers, Tom Bay (ph), from the Stern School of Business has in fact done a meeting interview and appeared in "Computer World", so I don't know why... (CROSSTALK) TUCKER: ... not quite ready for it today. DOBBS: Well this is fascinating. Let's back up here a bit. TUCKER: Right. DOBBS: One: Who were the two professors who did the study? TUCKER: Professor Hit (ph) at Wharton and Tom Bay (ph) at the Stern. DOBBS: And Tom Bay (ph) did an interview with "Computer World" and revealed what? TUCKER: A very straightforward interview. Five to six percent depression in the wages, two to three percent depression in the -- all the off-sourcing side of this business and talked about why they felt it needed to be done because there's a lack of empirical evidence. DOBBS: A lack of empirical evidence. We have been reporting here... TUCKER: Right. DOBBS: ... for years on it, I wrote a book as a matter of fact in 2004... (CROSSTALK) DOBBS: ... called "Exporting America" in which we dealt in part with H-1B visas. This looks like absolute censorship by two universities supported by corporate money. What are the universities saying about the appearance of absolutely stripping two authors of their academic freedom? TUCKER: Well, we put that question to them this afternoon and they insist this is independent research, not funded by any corporate money, no technology industry money involved in this study, and it's the insistence of the spokesperson at the Stern School today. DOBBS: And Wharton School, at the University of Penn -- their response. TUCKER: The same and then their professor just straight up declined any request for an interview as well. DOBBS: So two universities on the same day making a decision to take down this research after five days on the Web. This looks straightforwardly, straightforwardly like two institutions succumbing to immense pressure. I am very sorry for these two professors. I am very sorry for the academy for having to confront these kinds of forces. This is absolutely an embarrassment for these two institutions. It's inexcusable on their part in my judgment. We would be delighted, by the way, to have the heads of both of these business schools join us here at any time. If you don't feel comfortable enough to provide academic freedom to your professors, we would be delighted to have you with us to further discuss it. Extraordinary, but our commendation to the professors for doing the research and let's see where we go and over what period of time. Bill Tucker, thank you very much. Well as I said, we have been reporting on the outsourcing of middle class American jobs to cheap foreign labor markets for literally years and the push by particularly the Chamber of Commerce, the U.S. Chamber of Commerce for more H-1B visas and other issues facing our hard working middle class for years now. In fact, I have written two best selling books "Exporting America" and "The War on the Middle Class" and here are just a few examples from our broadcasts over the years on this very important issue. (BEGIN VIDEO CLIP) DOBBS: The war on our middle class is real and it is intensifying and middle class working Americans are losing their struggle against special interests. The dominant political power of corporate America in Washington. Big business interests in this country are pressuring Congress to expand the H-1B visa program. Business leaders couldn't care less that this failed, mismanaged guest worker program is taking jobs away from hard working middle class Americans. The abundance of cheap foreign labor in this country right now is pushing wages lower for millions of middle class working Americans. (END VIDEO CLIP) DOBBS: And that was three years ago. We'll keep you up-to-date on the progress at the Stern Business School and the Wharton School at the University of Pennsylvania. Hopefully the standards of those two fine institutions will prevail over some rather misguided policy judgments rendered apparently today, when we simply asked the question like what are the facts.