Date: Fri, 12 Sep 2003 12:23:21 -0700 From: Norm Matloff To: Norm Matloff Subject: absurd Fed study of H-1B [The posting has been updated here.--NM] To: age discrimination/H-1B/L-1 e-newsletter The CNN article mentions the FAIR paper which I cited a few days ago, and newly-released papers by the American Immigration Lawyers Assocation and by Madeline Zavodny at the Atlanta branch of the Federal Reserve Bank. I will not bother here with the AILA paper, since it is so obviously self-serving. It uses the old arguments, e.g. there must be no big problem with H-1B because the Dept. of Labor gets so few complaints. Once again, for the record, in addition to the fact that most potential complainers would not do so, out of fear of retribution or blacklisting, complaints to DOL are USELESS, because DOL's hands are tied, due to the gaping loopholes in the law. Most underpayment of H-1Bs is FULLY LEGAL. It is NOT an enforcement issue. Hence no complaints. The AILA (actually AILF) also innocently points out that underpaying the H-1Bs is against the law--when it isn't. So, I will focus here on the Fed study. I should point out first that the Fed has been a strong, public promoter of the H-1B program. For example, an economist at the Dallas branch wrote an op-ed in support of the 2000 H-1B expansion, and the Boston branch has also been supportive. As many of you know, Fed chair Greenspan has been very supportive of H-1B. Zavodny's study consists of a statistical regression analysis relating IT wage levels in a given region to the concentration of H-1Bs in that region. Other covariates, such as level of education, age (and even number of Indian immigrants!) are included as explanatory variables as well. In addition to the CNN article, I am enclosing below the Fed's press release on Zavodny's paper, along with the Web link for that press release. Zavodny's paper itself is at http://www.frbatlanta.org/filelegacydocs/ACF6CBD.pdf Finally, I am enclosing another news item on the Zavodny paper (source seems to be a labor union newsletter). I also mention that I have a "license" to critique the statistical analysis in the Zavodny paper, in the sense that I myself am a professional statistician. I am a former statistics professor and statistical consultant, and my specialty during that "past life" was regression modeling, the tool Zavodny is using here. (Of course, just as for economists, one could also make a joke about what happens if one laid all statisticians of the world end-to-end...) A summary of what I will say below is: * Zavodny did not study the main occupation that IT H-1Bs are in, programming. * Zavodny did not do a comparison of wages of H-1B and American workers. * Zavodny found a negative impact of H-1B on U.S. IT workers in terms of unemployment in 2001-2002 (though not in 2000-2001). Let's get one thing out of the way, first and foremost: The CNN author is wrong when he says, In fact, Zavodny said that in 2001, the latest data available, H-1B workers with bachelor's degrees were paid more, on average, than U.S. citizens with bachelor's degrees. Zavodny's study does NOT say this. There is no such comment in her paper; there is no such comment in the Fed's press release; there is no such comment in the news item at the end of this posting; and as you can see above, there is no way Zavodny's study even *could* measure such a thing, the way the study is structured. I will restrict my criticism of the study to only a few points. There are many more I could cite, but to me these are the salient ones. First, if the main issue is whether H-1Bs are underpaid (which Zavodny's remarks in the paper indicate is the issue of interest), then one must ask what the point is of doing an INDIRECT, HARD-TO-INTERPRET study like Zavodny's, when there are already so many DIRECT, EASILY-INTERPRETED studies which clearly answer the question of underpayment. The classical example asks what we would do if we wish to determine how many teeth horses have. The direct way would be to open a horse's mouth and count the teeth. Anyone could understand the meaning of that data, and that data would directly answer the question at hand. The indirect way would be to collect data on, say, chewing times for various kinds of horse feed, and then somehow infer from that the number of teeth in a horse's mouth. That would not directly answer the question, and would be fraught with all kinds of questions of validity and interpretability. We already have a more than ample number of studies on the question of underpayment of H-1Bs. These studies directly answer the question, by actually looking at H-1B salaries and American salaries. This is very direct, and very interpretable. I've cited the main ones before: the UCLA study; my own study; the Cornell study; the INS data; and two congressionally-sponsored studies, by the National Research Council and the GAO. They all showed underpayment of H-1Bs. Those last two are about as direct as you can get--they actually asked employers whether they paid H-1Bs less. Needless to say, most employers who do underpay H-1Bs aren't going to admit to it, but interestingly, a number of employers surveyed actually did admit it. The GAO study stated, Some employers said that they hired H-1B workers in part because these workers would often accept lower salaries than similarly qualified U.S. workers; however, these employers said they never paid H-1B workers less than the required wage. The NRC report contained a similar statement. Note also that the last statement in the GAO quote above shows that the prevailing-wage requirement in H-1B law is meaningless (due to huge loopholes). One does need to be careful. For instance, some studies (e.g. the one by Lowell which is cited by Zavodny) look at all foreign-born engineers, for instance, which is far too broad, because once a foreign worker gets a green card he is not exploitable anymore. But if done carefully, direct comparison of H-1B and American wages is the only real way to answer the question. Thus the fact that Zavodny resorted to such a screwy indirect way to study the problem, suggests that the Fed has a political motivation here. This is especially a concern in light of the fact that Zavodny has been very selective in her bibliographic citations. She does not mention most of the studies I listed above at all, and although she cites the NRC study, she makes no mention whatsoever any of the negative findings of that study--its survey in which employers admitted to underpaying H-1Bs, the finding that H-1Bs are de facto indentured servants, etc. Second, I am deeply disturbed by the data that Zavodny chose to analyze. She says that the average wage in California for "computer-related" jobs is $1,300 per week! This ridiculously low figure right away totally invalidates her study, so it's important to look at how she comes up with that figure. She is using the Current Population Survey data, and the occupation titles she chose from that survey were, according to the footnote in Zavodny's Table 4, "computer system analysts and scientists," plus workers in operations research, which is a completely unrelated field. What Zavodny did NOT include was another CPS occupation code--computer programmers! In other words, she left out the MAIN occupational group. As I said, that alone completely invalidates her study, quite apart from the problems of indirectness and interpretability I cited earlier. (Even inclusion of programmers would still not solve some severe methodological problems; see http://heather.cs.ucdavis.edu/Archive/GAO03.txt) Zavodny also looked at unemployment rates. In her data for Oct. 2000-Sept. 2001 she found that "H-1B workers...do not appear to have an adverse imapct on contemporaneous unemployment rates," while for Oct. 2001-Sept. 2002 she said "...the results are quite different...The unemployment rate IV regressions indicate a positive relationship between the [number of H-1B applications] and the unemployment rate for IT workers." What is really going on here? First of all, unemployment is not a good measure to begin with, because if people can't get engineering work they leave the field. As Gene Nelson famously put it, the former programmer who is now working as a security guard counts in the data as an employed security guard, not an unemployment programmer. See Carol Veneri, Can Occupational Labor Shortages Be Identified Using Available Data?, Monthly Labor Review, March 1999. Having said that, though, what is the explanation of the difference between Zavodny's findings in 2000-2001 vs. 2001-2002? The salient explanation is that 2000-2001 was the time of the dot-com bust. The effect of there being so many IT layoffs and business closings in 2000-2001 dwarfed any impact of the H-1B program. Zavodny herself suggests that there is a lag effect, with H-1B hiring not having an impact until a year later (which is the reason for her qualifying adjective "contemporaneous"). Norm http://money.cnn.com/2003/09/11/news/economy/visa_impact/index.htm Visas vs. jobs Many U.S. tech workers hate the H-1B program, but studies raise questions about its full impact. September 11, 2003: 3:28 PM EDT By Mark Gongloff, CNN/Money Staff Writer NEW YORK (CNN/Money) - A U.S. visa program that allows foreigners to temporarily work in technology and other high-paying industries is often blamed for taking American jobs and pushing wages lower. But two recent studies raise some questions about just how damaging this program is. The H-1B visa program allows U.S. employers to hire non-citizens for up to six years, as long as those workers have the appropriate experience and training, and are paid the same wages U.S. workers get. Though the number of H-1B visas is limited by law to 195,000 per year -- a number that will drop to 65,000 in 2004 -- amounting to just 0.13 percent of the total U.S. labor force, the programs have their greatest impact on high-tech jobs, which are among the best-paying in the economy. Labor groups, lobbyists and legions of unemployed tech workers have taken to the Internet and the airwaves to protest what they see as employer abuse of the visa system. They say the program keeps a lid on wage growth, subjects non-citizen workers to substandard working conditions and keeps U.S. citizens unemployed. "At a time of high unemployment, the high-tech industry is flooding the labor market by importing workers who are willing to work more cheaply than American high-tech workers," the Federation for American Immigration Reform, a Washington lobbying group, wrote in an August report called "Deleting American Workers." In response to the pressure, and in light of a lack of job growth in the broader economy, House Democrats and Republicans have proposed a bill to repeal the H-1B program. But a study published this week by Atlanta Federal Reserve Bank economist Madeline Zavodny said the effect of the H-1B visa program on wages was difficult to measure. In fact, Zavodny said that in 2001, the latest data available, H-1B workers with bachelor's degrees were paid more, on average, than U.S. citizens with bachelor's degrees. In that sense, then, it would seem that H-1B competition could actually put upward pressure on wages. "Study results provide little support for claims that the H-1B program has a negative impact on wages," Zavodny wrote. However, Zavodny also said that there was an apparent connection between the number of applications for visas and unemployment in the information technology industry a year later. A separate report released this week by the American Immigration Law Foundation, a Washington think tank and lobbying group, pointed out that H-1B visa issuance has fallen as the economy has weakened -- just 79,100 visas were issued in 2002, compared with 164,000 in 2001. The group also said visa processing fees have raised $692 million in the past five years, which has been used to train and educate thousands of U.S. workers. "Curtailing legal immigration to the United States or further impeding the flow of skilled foreign professionals to America will hurt the nation's competitiveness and its leadership in the world," the AILF said in a statement. "Such actions would slow U.S. labor-force growth, inhibit innovation inside the United States, reduce job growth, and encourage increased efforts to outsource and place overseas high-technology jobs and centers for research and development." The AILF also said an immigration specialist for ChevronTexaco suggested that delays in H-1B visa application processing could encourage U.S. companies to simply move jobs offshore. But that process is already well under way in many of the same white-collar industries served by the H-1B program, and critics of the visa system say it only gives overseas labor sources, such as India and China, a greater supply of well-trained workers to take more American jobs. In her study, Fed economist Zavodny also pointed out that a growing number of employers are turning to the L-1 visa program, which allows companies to rotate overseas employees into U.S. positions for up to a year, and the visas are renewable for up to seven years for high-level executives. The L-1 program has no mandated limit, and employers aren't required to pay "prevailing" wages to L-1 workers. The State Department will likely issue about 60,000 L-1 visas in 2003, according to an August study by the Information Technology Association of America, a worker lobbying group. Zavodny said the impact of the L-1 program on wages was still unknown and could "possibly merit analysis." Legislation that would limit the number of L-1 visas issued every year to 35,000 and place other restrictions on the program is pending in a House subcommittee. http://www.frbatlanta.org/invoke.cfm?objectid=59128295-7160-46B3-B963F9A123AB9F26&method=display_body ISSN 0732-1813 The H-1B Program and Its Effects on Information Technology Workers Madeline Zavodny Employment in the information technology (IT) field rose rapidly during the late 1990s. Many IT employees are foreign born and are working in the United States with H-1B visastemporary nonimmigrant visas issued for terms of up to six years. Critics of the H-1B program contend that it reduces job opportunities and wages for native workers. The programs supporters argue that H-1B workerswho typically have at least a bachelors degree in a specialty such as computer programminghelp address a shortage of skilled native workers and fill positions that would otherwise go vacant. This article examines whether the H-1B visa program negatively affects IT workers wages. The author uses data on labor condition applications (LCAs) filed with the U.S. Department of Labor to investigate whether the number of H-1B workers in an area, relative to the total number of IT workers in that area, is negatively associated with the level of and change in average IT wages and the unemployment rate among IT workers in that area. The results provide little support for claims that the program 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 signals that any effect is difficult to find, as previous studies concluded. ___________________________________ ---------------------------------------------- A study recently published by the Atlanta Federal Reserve Bank concluded that although the link may not be a strong one, there appears to be a connection between use of H-1B visas and higher unemployment among U.S. workers in information technology fields. Looking at H-1B visa figures and wage and employment data for U.S. information technology workers in fiscal year 2001, Atlanta Fed economist Madeline Zavodny said, "[T]he results suggest that the number of H-1B workers does not depress wages or wage growth but may have an adverse effect on unemployment." Her findings are described in an article published in the bank's Economic Review: Third Quarter 2003. In the context of the persistently jobless economic expansion, lawmakers and groups outside of Congress recently have raised questions about the H-1B visa program might hurt U.S. workers at a time when many technology employees are out of work. However, major changes in the program are not expected anytime soon given other matters on the legislative agenda. The time frame of the study--Oct. 1, 2000, through Sept. 30, 2001--included the recession that began in March 2001 and ended that November. Like most other sectors, information technology showed employment declines during that time and still remains weak. The Atlanta Fed economist studied Labor Department data from labor condition applications (LCAs) on which employers state the position, wage, and benefits to be offered to the H-1B visa worker. Rules require employers to pay whichever is higher--the prevailing wage for the position or actual wages paid to other individuals with similar experience and qualifications. Once the LCA is certified, the employer files a visa petition with the Department of Homeland Security's Bureau of Citizenship and Immigration Services, which decides whether to grant the visa. Job Impact Lags Zavodny's findings suggest that any negative effects on employment show up about a year after the visa applications are approved. In a Sept. 11 phone interview, Zavodny said she attributes the lag mainly to the bureaucratic process with its delay between DOL's approval of an application and BCIS issuing the visa. When she compared data on H-1B applications for fiscal year 2001 with employment data for fiscal 2002, the economist found a correlation between the visa application level and the unemployment rate for IT workers. Her study did not attempt to quantify how much of an effect the visas had on unemployment. Zavodny underscored that her analysis reflects the visa applications made to the Labor Department, not the actual number of new H-1B workers. However, the two sets of figures "are likely to be well correlated," she said, adding that no data are available on new H-1B visa holders working in information technology jobs. H-1B visa holders are concentrated in information technology fields. The study said in fiscal 2001, nearly 58 percent of H-1B visas were issued to workers in computer-related occupations, with computer systems design and related services the primary category. "The vast majority of H-1B visa holders are from Asia and are highly educated," the study said. In fiscal 2001, nearly half of H-1B visa recipients were from India, the country with the highest proportion of workers obtaining visas under the program. The median income of H-1B recipients was $55,000 in fiscal 2001, and about 98 percent of these workers at least had a bachelor's degree, the study said. In contrast, about 26 percent of U.S. residents age 25 and older had at least a bachelor's degree in 2000, and median earnings among these workers was $46,969, according to Census Bureau figures cited by Zavodny in the study. Although Zavodny concluded that there is "little support for claims that the program has a negative impact on wages," she added that such a connection could be "difficult to find." Hard to Discern Effects Zavodny told BNA that she finds it "troubling" that members of Congress could eventually change the visa laws without having solid data to assess the program's impact. "We have found very little evidence of its effects, and that is because it's very difficult to find the evidence," she said. Without congressional action, the limit on H-1B visas will revert to 65,000 (from 195,000) for fiscal 2004, which will begin Oct. 1, 2003. Zavodny pointed out that the cap was not binding in fiscal years 2001 and 2002 because of the recession followed by weak economic growth. She cited BCIS figures showing that about 164,000 H-1B visas were issued in fiscal 2001, and about 79,000 in fiscal 2002. The study also pointed out that apart from issues related to H-1B visas, there is ample evidence that employers are increasing using L-1 visas, which allow, for example, transfers of workers within a company from overseas branches to subsidiaries. "Unlike H-1B visas, L-1 visas are not capped and do not require employers to pay the prevailing wage," she said. Zavodny suggested that increasing use of L-1 visas relative to H-1B visas "may reduce the measured impact of H-1Bs on native workers." During a July Senate Judiciary subcommittee House hearing, lawmakers discussed possibly passing limited changes in the law to address potential fraud and misuse of the narrower L-1 program (146 DLR A-8, 7/30/03). A recent report by the Federation for American Immigration Reform called for repeal of the H-1B program, and cited what it called abuses of the L-1 program (172 DLR A-3, 9/5/03). The Atlanta Fed study, "The H-1B Program and Its Effects on Information Technology Workers," is available on the bank's Web site at: . Click on "Publications" and "Periodicals." By Pam Ginsbach