Date: Fri, 8 May 2009 22:40:20 -0700 From: Norm Matloff To: Norm Matloff Subject: University of Maryland paper on H-1B To: H-1B/L-1/offshoring e-newsletter (UPDATE: SEE http://heather.cs.ucdavis.edu/Archive/MithasLucaPublished.txt FOR MY ANALYSIS OF THE FINAL PUBLISHED PAPER.--NM) The enclosed article from InformationWeek reports on recent research by Sunil Mithas and Henry Lucas, two University of Maryland business professors. The following passage in the article sums it up: # The new research dispels findings of some other studies that assert H-1B # and other foreign workers are paid less than American IT professionals. # These other studied often use salary data from Labor Condition # Applications, or LCAs, said Mithas. However, analysis of pay data from # LCAs, which employers file with the U.S. Department of Labor, do not # reflect differences in education, experience levels, and other factors # among workers, said Mithas. The underpayment of H-1Bs is firmly established. It has been shown repeatedly in a number of different ways. This is not just anecdote, but the result of a number of statistical studies, employer surveys, etc. Moreover, there is the fundamental economic principle that the H-1Bs, with very limited mobility in the labor market, simply cannot get salaries as high as U.S. citizens and permanent residents, who are free agents in the job market. Indeed, even the Mithas and Lucas study confirms this, as I will note later. So, given all of this prior work showing underpayment of the H-1Bs, how did the Maryland findings wind up contrary to the others? This is the main question I'll address here. Before I begin, please note that, while Profs. Mithas and Lucas kindly provided me with a copy of their paper, they asked me not to comment on its contents until the peer review process is complete. They thus requested that I instead simply discuss the InformationWeek article. In response, I asked their permission to at least quote one particular table in their paper that is especially significant, but again the authors stated that they do not want me to cite any of the contents of the paper. I will of course honor their request. By the way, they have seen a draft of this posting, and will give me a response to post here in about a week. Now, as some mathematical riddles sometimes ask, "Where is the missing dollar?" Why are the findings in the Maryland study so different from what has been found before? Or, as the author of a book on debugging ("The Art of Debugging," N. Matloff and P. Salzman, NSP, 2008), I might phrase the question as, "Where is the bug?" Let's start with the authors' own explanation for the discrepancy (reproducing the above quote here for convenience): # The new research dispels findings of some other studies that assert H-1B # and other foreign workers are paid less than American IT professionals. # These other studied often use salary data from Labor Condition # Applications, or LCAs, said Mithas. However, analysis of pay data from # LCAs, which employers file with the U.S. Department of Labor, do not # reflect differences in education, experience levels, and other factors # among workers, said Mithas. # # "There's no way to know whether those applications were approved, what # the education levels were," he said. One can debate back and forth how reliable the LCAs are, but the key point is that other than John Miano's work, I am not aware of any other studies based on them. For instance, none of my own studies has used the LCAs; I use the census data and the PERM data. The UCLA study also is based on the census data. The census and PERM data are for actual workers, not applications. The NRC and GAO studies rely on employer surveys and the CPS mini-census. So Mithas and Lucas' explanation for their own results being different from the earlier ones, due to a supposed reliance on LCA data, is not correct. Now, what about the variables cited above, education and years of experience? The census studies do control for education and experience levels (the latter proxied by age), and the employer surveys specifically asked the employers to compare salaries they paid to H-1Bs with the wages paid to comparable Americans, i.e. those with the same education, experience and so on. So again, there is no problem here. Thus Prof. Mithas and Lucas' explanation for the differences in outcomes between their study and the others is invalid. Instead, the major culprit seems to be Mithas and Lucas' data source. As noted in the article, Mithas and Lucas' analyzed survey data on InformationWeek's readers, who are not representative. The magazine is mainly aimed at managers and nontechnical people affiliated with the IT industry. Its self-description states, "InformationWeek: InformationWeek.com is the industry-leading source of news, analysis, and perspective, serving business technology executives at a cross section of companies." Its job postings section currently lists the following positions: % D. E. Shaw Research seeking Chief of Staff in New York, NY. % Switch and Data seeking Sr Product Marketing Manager in Tampa, FL. % Kadrmas, Lee and Jackson seeking Network Architect in Bismarck, ND. % Osram Sylvania seeking Benefits Specialist in Danvers, MA. % AccuWeather seeking Business Development Manager in Atlanta, GA. Only one of these, the network architect position, is a technical job. To be sure, it must be pointed out that the magazine is indeed available in the coffee rooms of many major tech firms, and the reader survey data does include substantial numbers of programmers and software engineers. But it also includes plenty of marketing people, benefits specialists and so on. Furthermore, though some of these are H-1Bs too, they are not programmers and engineers, the core of the H-1B controversy. The basic problem is that InformationWeek is not something that the typical software engineer at, say, Apple or Google would read, much less respond to its survey. And of course this is an absolutely crucial point. Many of you have probably heard of the famous Literary Digest election poll fiasco of 1936, in which that magazine's poll "showed" that Republican presidental candidate Alf Landon would overwhelmingly defeat Democrat Franklin Roosevelt. In the end, Roosevelt was the victor, with 62% of the vote. Why was the Literary Digest poll so far off? The reason was that the population they polled--their own readers, automobile owners etc.--tended to be richer than average, and thus more Republican than average, skewing the results beyond recognition. So, the nature of the population sampled is of the utmost importance. Not only is there the issue of the population polled, but also the one of who responds. Studies based on survey data are always shaky, due to self-selection biases in the response. There is huge potential for self-selection problems in the InformationWeek data, as the 2009 survey, available on the Web, states a response rate of under 4%. In that context, one concern would be that USCIS (then INS) data show a wide variation in H-1B salaries by nationality, with workers from the Third World getting lower salaries. On average the Chinese and Indians make about $20,000 less than those from the UK in the computer-related jobs, for example. Based on my experience in the Chinese immigrant community (which by the way includes my conducting a survey, wearing my statistics professor hat, for the Chinese-language Sing Tao Daily), the response rate for the Chinese readers of InformationWeek would be markedly lower, thus artificially raising H-1B salaries in the survey. There are a number of other issues. Remember, the industry claims that the H-1Bs are hired primarily because they have special "hot" skill sets. These generally command a salary premium of 15-25% (see data in my University of Michigan Journal of Law Reform article), so even if one were to take Mithas and Lucas' 5-9% as accurate, that would indicate net underpayment, not overpayment. (The article below says that the authors analyzed "skills data," but that is not in the reader survey that was used. It is simply a reference to education and number of years of experience.) There is also the issue of contractors, who generally are the better-quality people who would be getting higher pay if they drew a salary. Their exclusion (no salary to compare to) biases the results too. Many readers of this e-newsletter will recall that in my studies I distinguish between Type I and II salary savings accrued by hiring H-1Bs. Type I savings compare hiring an H-1B to hiring an American of the same education, experience and so on, which is the issue addressed by the Maryland study. But Type II savings come from hiring a young H-1B instead of an older (35+) American. As many readers know, it is Type II savings that I have always emphasized as being at the core of the attraction of the H-1B program for employers. The Maryland study was not designed to deal with this issue, of course, but I wish to point out that though Type I salary savings is definitely real, my focus has been on Type II. To end, let's turn to an aspect in which the Maryland study CONFIRMS the previous work on underpayment of H-1Bs. Mithas and Lucas found that green card holders get paid more than H-1Bs of the same education and experience levels. This jibes completely with the central point of the H-1B critics, that H-1Bs, as de facto indentured servants, are exploited in terms of salary. Norm http://www.informationweek.com/story/showArticle.jhtml?articleID=217200260 InformationWeek Foreign IT Pros Working In U.S. Earning More Than Americans By Marianne Kolbasuk McGee, InformationWeek April 28, 2009 While opponents of H-1B and L-1 visas have long argued that the temporary work programs encourage employers to hire cheap foreign labor, a new study says noncitizen IT professionals earn pay that's on average 5% to 9% higher than American workers with similar education The report, "Does High-Skill Immigration Make Everyone Better Off? United States' Visa Policies And Compensation Of Information Technology Professionals," by two researchers at the University of Maryland, analyzed skills and pay data on more than 50,000 IT professionals who participated in InformationWeek salary surveys from 2000 to 2005. The researchers -- Sunil Mithas, an assistant professor, and Henry Lucas, chair of the department of decision, operations, and information technologies at the Robert H. Smith School of Business, University of Maryland, College Park -- found that foreign IT professionals, including those with H-1B visas, L-1 visas, or green cards, reported pay that ranged between 5% and 9% higher than pay received by U.S. citizens with similar attributes, including educational degrees and IT experience. "There was no downward pressure on the pay of U.S. citizens," said Mithas. The new research dispels findings of some other studies that assert H-1B and other foreign workers are paid less than American IT professionals. These other studied often use salary data from Labor Condition Applications, or LCAs, said Mithas. However, analysis of pay data from LCAs, which employers file with the U.S. Department of Labor, do not reflect differences in education, experience levels, and other factors among workers, said Mithas. "There's no way to know whether those applications were approved, what the education levels were," he said. For noncitizens, the biggest driver for pay premiums appears to be IT experience, while employers tend to reward U.S. citizens with pay premiums based on education level, said Lucas. "Employers pay a premium not for education of non-U.S. citizens, but for their IT skills as reflected in their IT experience," Lucas said. Mithas said it's likely that noncitizen IT professionals are paid premiums because of "intangible" attributes that employers seek that aren't as easily found among U.S. citizens, including exposure to various global and cultural experiences. "By hiring these people, it helps a company to deliver products and services for a global market, to make connections to those global markets," he said. The research also found that among noncitizen IT professionals in the United States, those with permanent U.S. residency, or green cards, earn the highest premiums. Green card holders earn about 6.1% more than noncitizen IT professionals with H-1B or other work visas. The pay premiums for IT professionals on work visas fluctuate in response to "supply shocks" created by the annual U.S. government caps on H-1B visas, said Mithas. In general, lower and fully utilized caps result in higher pay premiums. For instance, the average salary premium for IT workers on work visas was 8.4% in 2000 when the H-1B cap was 115,000. However, the premium fell to 3.3% in 2001 when the H-1B cap went up to 195,000. A similar pattern was observed for green card holders and noncitizens. The pay premium for noncitizens rose significantly to 17.5% in 2004 when the H-1B visa cap went down to 65,000, from a cap of 195,000 in 2003. The current cap for H-1B visas is 85,000, including 65,000 "general" H-1B visas, and 20,000 exemptions for foreign individuals who receive advanced degrees from U.S. universities. U.S. Citizenship and Immigration Services on April 1 began accepting from employers H-1B visa petitions to hire foreign workers for fiscal 2010, which starts Oct. 1. But to date, the agency has received approximately 45,000 petitions toward the 65,000 general cap and approximately 20,000 petitions for the advanced-degree category. USCIS is continuing to accept petitions for both categories, since not all of the applications received so far will end up being approved for actual visas. The recession is being blamed for the relative weak demand for H-1B visas this year. In recent years, the USCIS stopped taking new H-1B visa applications within days of accepting the petitions for the next fiscal year. Supporters of the H-1B visa program, including technology companies, have been lobbying Congress for years to raise the H-1B visa cap back to at least 115,000, the level in 2000. However, opponents to H-1B visas argue that abuse and fraud, including allegations of employers hiring foreign workers at cheap wages, mar the program. While the research by University of Maryland suggests that pay is actually higher on average for noncitizen IT professionals, others believe new legislation is needed to reform the H-1B and L-1 programs. In fact, late last week, Sens. Chuck Grassley, R-Iowa, and Richard Durbin, D-Ill., introduced a bill aimed at eliminating alleged H-1B and L-1 visa fraud and abuse. The proposed legislation is similar to a bipartisan bill the two senators introduced in 2007 when Congress was weighing a more comprehensive overhaul of U.S. immigration policy.