Date: Tue, 11 Mar 2008 21:45:10 -0700 From: Norm Matloff To: Norm Matloff Subject: another badly flawed NFAP study To: H-1B/L-1/offshoring e-newsletter (see http://heather.cs.ucdavis.edu/h1b.html for archive search and general material on problems with the H-1B visa) The National Foundation for American Policy, a grandiose name for the one-person "foundation" run by industry advocate Stuart Anderson, has released yet another "study" extolling the virtues of the H-1B work visa. Sadly, this one is even more badly flawed than its predecessors. The enclosed article notes that * Stuart, a former staff director of the Senate Immigration * subcommittee, says NFAP, is "a non-profit, non-partisan public policy * research organization that's independently funded by foundations and * private contributors, and does not lobby for legislation." This is a tax law technicality, folks. Many organizations (including by the those who promote reductions in immigration) operate under the 501(c)3 section of the tax code, which grants tax-exempt status to organizations that do not lobby but who are freely allowed to "educate" Congress. You get the drift here, I'm sure. And "non-partisan" simply means that NFAP is not formally associated with one party or the other. Anderson is hardly an unbiased researcher. He's been making a living writing articles supporting the H-1B program since the mid-1990s, hired for example by the Information Technology of America, one of the major lobbyists for an expanded H-1B program. Though he is not saying who his present employer is, the fact that the lobbyists who quote him the most are immigration attorneys would seem to point to the American Immigration Lawyers Association or their American Immigration Law Foundation as a major funder. Though Anderson cites his work in the Senate--where he authored the H-1B expansion bills of 1998 and 2000--he doesn't mention that he subsequently served as Executive Associate Commissioner in the INS Office of Policy and Planning during the first term of George W. Bush. The INS (now USCIS) stopped making data available to researchers at the time Anderson was there, and I suspect that this is no coincidence. A Washington Monthly article at the time (www.washingtonmonthly.com/features/2001/0205.confessore.html) contained this passage: ^ "The best analogy I can draw about Stuart Anderson is something that ^ an INS agent said to me: If you were going to hire someone to run the ^ DEA, you wouldn't pick somebody who favors legalizing drugs," says a ^ top Republican aide on the Hill. "And by putting Stuart Anderson in a ^ ranking position in the INS, you've essentially done the same ^ thing---you've got somebody who favors open borders running the ^ agency that regulates the borders." Now, what about Anderson's analysis in this latest report? His main results involve statistical regression analysis, so as a former statistics professor I will now give you a crash course in regression. (Don't worry; this will be quick and simple.) In its basic form, a regression model estimates a presumed linear relationship between a response variable Y and predictor variables X1, X2 and so on. Here the word "relationship" means that the mean of Y for fixed values of X1, X2 and so on is a linear function: mean Y = c0 + c1 X1 + c2 X2 + ... for some constants c0, c1, c2 and so on to be estimated by our sample data. For instance, we might analyze human mean weight as a linear function of height and age. (Note the word "mean," often overlooked.) By the way, the term "linear" refers to the c's, not the X's.) Many thick books have been published on this methodology, showing for example how to assess the validity of the linearity assumption etc. But really, you now know the basic notions. By the way, to learn more about statistics, see my online course, at http://heather.cs.ucdavis.edu/~matloff/probstatbook.html Anderson had Y = job growth for a firm, X1 = number of H-1Bs the firm applied to hire and X2 = current number of jobs at the firm. (Y and X1 are both given as percentages of X2.) His main finding is that c1 has an approximate value of 5, which he correctly interprets as meaning that the mean increase in jobs is about 5 times the increase in H-1B applications. But what does that say? It simply says that about 20% of the new jobs go to H-1Bs. It doesn't say that the H-1B workers caused the creation of new jobs. Suppose a company tends to hire about 1% of its workers as H-1Bs. Then by Anderson's reasoning, each H-1B is creating 100 new jobs at the firm! In other words, there is zero content in Anderson's analysis. As I said, that was Anderson's main point, so I could end this post right here. He made an embarrassing statistical error, and that's that. But it's important to bring up some other points here: 1. The R-squared value, which is the classical measure of predictive ability in regression, is only 0.0687 in Anderson's main regression. In other words, his X1 and X2 are only explaining less than 7% of the variation in Y. Anderson concedes that this figure is "relatively low," but it's actually absurdly low. Basically, an R-squared value of 7% is saying that X1 and X2 have almost NO relationship to Y. And by the way, the sample R-squared is known to have a tendency to be an overestimate, so the situation is even worse still. 2. Even if some of the jobs filled by H-1Bs have some sort of job-creation abilities, e.g. development of new products, one could fill the jobs with U.S. citizens and permanent residents and still get the same job-generating effects. 3. Anderson is egregiously misusing the term "statistically significant," which merely means "not due to random chance." It does not imply practical importance. 4. After spending a good part of the last two years blasting H-1B critic John Miano for using LCAs to study the H-1B issue--Anderson's point being that LCAs are only *applications* to hire H-1Bs and thus do not necessarily represent the number of H-1Bs actually hired--Anderson has the chutzpah here to base his analysis on LCAs! He does make a disclaimer along these lines in the appendix of his report, but no one will see that. 5. Anderson also claims, % If the proposition was true that companies hire H-1B professionals % because they're cheaper, then when businesses hit hard times they should % hire more H-1Bs to save money. However, the analysis shows that, % overall, H-1B filings at U.S. technologies declined when companies hit % hard times, undermining the perennial assertion that H-1Bs are hired as % "cheap labor." This is ridiculous. If companies don't have job openings, they can't hire anyone, including H-1Bs. Anderson has outdone himself with this report. Norm http://www.informationweek.com/story/showArticle.jhtml?articleID=206902716 U.S. Tech Companies Add Five Workers For Each H-1B Visa They Seek A report says large tech companies with more than 5,000 employers added an average of five workers for every H-1B position requested; smaller companies added 7.5 workers. By Marianne Kolbasuk McGee, InformationWeek March 10, 2008 For each H-1b visa position requested, U.S. technology companies increase their employment by an average of five workers, says a new report released today by the National Foundation of American Policy, a research organization. The NFAP examined all H-1B Labor Condition Applications, or LCA filings, to the U.S. Dept. of Labor from 2001 to 2005 by Standards & Poor's 500 technology companies. The NFAP report says researchers used "a regression model that controls for both general market conditions and firm size" in analyzing the association between number of positions required in H-1B LCA documents and the percentage of total employment. Through this analysis, NFAP determined that S&P 500 technology companies with more than 5,000 employers added an average of five workers for every H-1B position requested. For companies with fewer than 5,000 employees, the average increase in employment was 7.5 workers for each H-1B position requested. The S&P 500, which consists of mostly American companies, currently features about 76 technology companies. The NFAP's report findings dispute a common argument by critics of the H-1B visa program who assert that the hiring of foreign technology professionals reduce employment opportunities for American workers. "At the minimum, this shows that H-1Bs are complementary to other U.S. workers being hired, not displacing them," said Stuart Anderson, executive director of NFAP in an interview with InformationWeek. Stuart, a former staff director of the Senate Immigration subcommittee, says NFAP, is "a non-profit, non-partisan public policy research organization that's independently funded by foundations and private contributors, and does not lobby for legislation." The LCAs filings examined by NFAP are documents that companies file to the Dept. of Labor to specify details such as job location and pay that an employer plans to provide for a position they'd like to fill by an H-1B worker. The LCAs are "supporting evidence" employers provide to the U.S. Dept. of Labor before the company can apply for H-1B visa workers. LCAs need to be certified by the DoL before the employer can petition the U.S. Citizenship and Immigration Services for an H-1B visa. However, the NFAP research did not examine the types of job positions that S&P companies added to their employment, or the pay for those positions. Anderson said the research did not conclude that companies added to their employment because they were saving money by hiring H-1B workers. In fact, he said an opposite argument may be true. "We did not see increases in H-1B hiring by companies when they faced harder times and needed to reduce their payrolls," he says. On April 1, the U.S. government begins accepting H-1B visa petitions from employers looking to hire foreign workers for fiscal 2009, which starts Oct. 1. In another report NFAP released on Monday, the organization said its analysis of S&P 500 employers found that there are more than 140,000 job openings for skilled positions at those companies today. Major U.S. technology companies today average more than 470 U.S.-based job openings for skilled positions, while defense companies have more than 1,265 each. The skilled positions include all jobs that require at least an undergraduate degree. NFAP examined posting for U.S.-based jobs at all companies in the S&P 500 companies. The vast majority of the data was gathered in January 2008, with the rest compiled in December 2007 or February 2008, says NFAP's report. So, who's trying to fill the most jobs? According the NFAP report, S&P 500 employers with the most job openings as of January 2008 are Microsoft (4,005), Northrop Grumman (3,925), Lockheed Martin (3,901), General Electric (3,078), Countrywide Financial (2,415), JPMorgan Chase (2,164), Tenet Healthcare (2,050), United Health Group (1,927), Raytheon (1,694), IBM (1,670), Computer Sciences Corp. (1,666), Cintas (1,664), L-3 Communications (1,618), Bank of America (1,600), U.S. Bancorp (1,562) and Cisco Systems (1,504).