Low Salaries For Low Skills: Wages And Skill Levels For H-1B Computer Workers, 2005
This report updates a December 2005 Backgrounder, "The Bottom of the Pay Scale: Wages for H-1B Computer Programmers."2 The previous report examined Labor Condition Applications filed in FY 2004. The procedures used in this report are nearly identical to those used in its predecessor.
There were three reasons for producing a new report based on the same type of data. First, this new report confirms that the results from 2004 were not a fluke. Repeating the same measurement on the subsequent year’s data produced nearly identical results. Another reason for a new report is that new data became available. Last year, for the first time, the Department of Labor made available the skill-based wage data. This makes it possible to examine how employers classify the skills of the H-1B workers they are seeking. Finally, a second look allowed investigation in more detail on exactly how employers produce the extremely low prevailing wage claims.
The H-1B Visa Program
The H-1B program is, in theory, a non-immigrant program, although it is widely used as a stepping stone to permanent residency. H-1B visas are valid for up to three years and can be renewed once for an additional three years. H-1B visas also are tied to employment, so an H-1B visa becomes invalid if a worker loses his or her job. While employed, it is relatively easy for a worker on an H-1B visa to transfer the visa to another employer. However, H-1B workers seeking permanent residency are effectively bound to their employer because changing jobs requires restarting the application process, making it unlikely that it will be completed before the visa expires. Transfers do not extend the time limit on the original visa.
The first step in the H-1B visa process is the filing of a Labor Condition Application (LCA) by the employer with the Department of Labor. The employer attests that workers will be paid the prevailing wage and that there is no strike or lockout in progress. The LCA shows the number of workers covered, the prevailing wage, and the wage to be paid. The statutes specifically limit the Labor Department review process to checking for "completeness and obvious inaccuracies," making this step effectively a rubber stamp process. After the LCA is approved (less than 1 percent are rejected), the employer can file an H-1B petition (I-129) with United States Citizenship and Immigration Services (USCIS). In FY 2005, over 300,000 LCAs were filed, covering about 700,000 workers. USCIS approved 117,000 visas in FY 2005. While the law does not constrain the petition review process in the same manner as the LCA approval process, the approval rate for visa petitions (99.5 percent)3 is nearly as high as the approval rate for LCAs (99.7 percent). If the petition is approved, the employee can then apply for the H-1B visa, either in the country of residence, if living overseas, or at a USCIS office, if already living in the United States. The State Department has a slightly higher rate of H-1B refusals; its consulates refused between 5 and 10 percent of all H-1B applications in 2005.4
The disclosure data are divided into two databases, one representing electronic filings, the other representing fax filings, with the electronic filing database containing the bulk of the records. These two databases have different structures so integrating them is not as simple a process as it might be. For completeness, this report includes the data from both sources. However, other researchers could legitimately ignore the fax data as they increase the average H-1B wage by only $2.
The reason for using the disclosure data as a measure of H-1B wages is that they are the best data available on wages for H-1B workers. This is the data Congress mandates be made available to monitor the H-1B program. In the future the author hopes to be able to repeat this analysis using the data from visa applications. So far, USCIS has responded to requests for this data by stating it cannot find the information. The disclosure data provide the best possible source for employer prevailing wage claims. Unlike the measurements of H-1B wages, all the prevailing wage comparisons to U.S. wages are direct comparisons. As such, they leave no doubt that employer prevailing wage claims do not reflect the actual prevailing wage.
The source for U.S. wages was the May 2004 National Occupational Employment and Wage Estimates available at http://www.bls.gov/oes/2004/may/oes_nat.htm. The May 2004 data were the last released before the start of FY 2005. Therefore, they were the last published before any of the LCAs examined were submitted.
This study also uses the 2005 FLC wage data (available at www.flcdatacenter.com) to match specific employer prevailing wage claims to their sources. These data were used to identify the job title used for prevailing wage claims and to analyze the nature of specific claims. They were not used for H-1B/U.S wage comparisons.
The biggest challenge in analyzing the LCA data is to match employer job titles to the Standard Occupation Codes (SOC) used to classify jobs in the Occupational Employment Statistics (OES) produced by the Bureau of Labor Statistics. Rather than using SOCs, LCAs use broad numeric classifications in which managers of programmers get lumped in with programmers. Analysis of these data is made even more difficult because the job code entered by employers is frequently wrong. In assigning SOC codes to LCAs, this study tries to be conservative and take the path that would minimize the difference between H-1B and U.S. wages.
The first step in this process was to select all records with computer-related job codes ("030" to "039") on the LCA.
The next step was to match LCAs using OES as the prevailing wage source to the actual OES data to get an SOC code. The job title in each matching record was verified using pattern matching (e.g. using "*soft*eng*" for "software engineer"). For those records where pattern matching failed, they were individually checked to ensure that the SOC code actually reflected the job title.
Pattern matching was used both to eliminate LCAs that were not programming related and to assign job codes. For example a pattern like "*vice president*" would be used to eliminate records for certain managerial employees and "*database admin*" to assign the SOC for "database administrator" to other LCAs.
Some job titles created special difficulties. There are separate SOC codes for "Software Engineers, Applications" and "Software Engineers, Systems," with wages for the latter tending to be higher. In this study, all software engineers were classified as "Applications" unless "Systems" was specified in the job title or the prevailing wage claim had been matched to an OES record specifying systems software engineer.
The most common job title was "Programmer/Analyst" (44 percent of programming workers). That could be either a "programmer" or a "systems analyst." This study chose to treat "Programmer/Analysts" as programmers except for the few cases (4 percent of "Programmers/Analysts") where the LCA’s prevailing wage claim had been matched to a specific OES record with a different job title. Classifying "Programmer/Analysts" as "Programmers" rather than "Systems Analysts" reduces the average H-1B to U.S. wage difference by about $4,000.
LCAs with the job title "Consultant" were treated as systems analysts unless the LCA’s prevailing wage claim had been matched to a specific OES record with a different job title. While classifying "Consultants" as programmers would have given a smaller H-1B/U.S. wage difference, the SOC definition of a programmer is not applicable to consultants.
Many LCA job titles were variations of "Developer." This
could mean a software engineer, systems analyst, or programmer. In order to
A number of LCAs could not be matched to any job code because
of unusual job titles that are not common in the industry. Together, these LCAs
had a slightly higher average wage than the average wage of those that could be
classified. These LCAs were arbitrarily assigned the SOC code for "Computer
programmers" because this choice gave the lowest H-1B-to-U.S. wage
LCAs where the annual wage was less than $10,000 or greater than $300,000 were treated as errors and were excluded. The 130 records (out of 300,000 filed) excluded had an average wage of $11.5 million with the highest value $454 million. In addition, LCAs for part-time work were excluded.
The Special Case of Infosys
Employer Prevailing Wage Claims
This section also specifies that when the Department of Labor
provides a prevailing wage source, it must include at least four pay levels,
"commensurate with experience, education, and the level of
The regulations (20 C.F.R. § 655.731(a)(2)) specify how the employer determines the prevailing wage. As an illustration of how loosely defined the process is, the first section says "...the employer is not required to use any specific methodology to determine the prevailing wage…." One specific the regulations do give is that the prevailing wage is to be based on either the mean or median depending on the circumstances.
Despite the complexity of the requirements, the statute provides no mechanism for DOL to verify employer claims; the LCA process is essentially an honor system. DOL adjudicators may only check that the form is filled out correctly — they may not investigate whether the LCA data contain bogus prevailing wage claims. There is insufficient information within the LCAs to cross-reference most LCA claims to the source of a prevailing wage claim, so there is insufficient information to determine the possible extent of a bogus prevailing wage problem. The fact that employer claims are not verified does not necessarily mean that the claims are wrong or bogus. The more important question is: How do employer prevailing wage claims compare to actual U.S. wages? What is the result of our reliance on an attestation system to protect U.S. workers?
Table 1 shows how employer prevailing wage claims compare to actual U.S. wages. Employer prevailing wage claims on LCAs for computer programming workers averaged $16,000 below the median U.S. wage based on occupation and location. The median difference between employer prevailing wage claims and the median U.S. wage was $18,000. For employer prevailing wage claims, 90 percent were below the U.S. median wage and 62 percent were in the bottom 25th percentile of U.S wages. Figure 1 compares the distribution of U.S. wages to H-1B prevailing wage claims.
H-1B Worker Skills
These skill levels only apply when the Department of Labor is used as the prevailing wage source.
In FY 2005, along with defining prevailing wages with four skill levels, the Department of Labor made its skill-based data available in table form for the first time. This allows one to match prevailing wage claims using this data to the specific source record used to make the claim. From a matching record, one can identify the skill level the employer claimed. About two thirds of the computer programming LCAs using this wage source could be matched back to the source record.
An examination of the skill levels reported on the LCAs reveals that employers claim that the majority of their H-1B workers are entry level. In stark contrast to claims that the H-1B program is used for "highly-skilled" workers, employers claim the majority are less-skilled workers. Figure 3 shows the distribution of employer skill claims on LCAs.
Prevailing Wage Sources
If one weights the difference between U.S. wages and the claimed prevailing wage for each wage source by the number of workers requested, three wage sources account for about $14,000 of the $15,000 median difference between prevailing wages and the median U.S. wage: Occupational Employment Statistics (OES), State Employment Security Agencies (SESA), and Watson-Wyatt. Together, these wage sources are represented on 87 percent of the LCAs, covering 93 percent of the workers requested.
Occupational Employment Statistics
Since OES is the wage source this report uses to measure actual U.S. wages, it might appear contradictory that LCAs using OES as the prevailing wage source contribute the most to the difference between U.S. wages and employer prevailing wage claims. This apparent discrepancy is easy to explain.
When the H-1B program was created, it required employers to pay workers the higher of the prevailing wage for the occupation and location, or the wage paid to similar employees. By requiring H-1B workers to be paid the overall prevailing wage, in theory it would prevent employers from using the H-1B program to import young, lower-paid workers to displace older U.S. workers.
As discussed, in 2004 Congress added a new prevailing-wage option for employers. It mandated that the Department of Labor provide employers with four skill-based prevailing wages. To comply with this change, the Foreign Labor Certification Data Center took the OES data produced by the Bureau of Labor Statistics and used them to create four skill-based prevailing wages. This created a mechanism for employers to justify low wages. Regardless of the actual skills of an H-1B worker, employers need only assert that a worker is in the Level I category for entry level, trainee, or intern employees, and pay according to that prevailing wage.
The findings described above show that, according to employers’ LCA claims, the majority of H-1B computer programming workers are indeed at the lowest skill level (Level I). The difference between the Level I wage and the OES median wage averages about $14,000. While the use of Level I wages as the prevailing wage is by far the most significant reason for low prevailing wage claims using the OES data, the LCA data also contain a number of completely bogus prevailing wage claims that purport to use OES as the wage source. Because most LCAs do not contain the information necessary to match a prevailing wage claim back to its source, it is impractical to determine the extent of these bogus claims.
Table 3 shows some examples of OES-based prevailing wage claims that are lower than the Level I wage for the occupation and location. A small number of these low wage claims are the result of employers improperly using the prevailing wage for higher education, as in the last claim. The fact that these erroneous applications were processed at all illustrates that LCAs are not checked for accuracy, much less fraud. This should be a routine aspect of any government benefit issuing process, and is often automated. Clearly, maintaining the integrity of H-1B processing is a low priority for the Department of Labor.
Of LCAs, 9 percent used SESA as the prevailing wage source, representing about 13 percent of the workers requested. The LCAs using SESA account for about $3,000 of the average difference between prevailing wage claims and the median U.S. wage.
Like employer-made, skill level-based OES prevailing wage claims, SESA determinations overwhelmingly tend to be based on low skill levels. Figure 5 shows how SESA determinations classify skill levels.
Verifying Wage Claims
Salary.com’s prevailing wages are based on three levels for each job title (e.g., Programmer I, Programmer II, and Programmer III). Because most LCAs using Salary.com as the prevailing wage source do not include the level, it is impractical to match large numbers of these wage claims to their sources.
Even LCAs that match the Salary.com job title and level exactly are difficult to match to their source, because Salary.com does not maintain historical data. To illustrate, one LCA is for a "Programmer I" in Edison, N.J. The claimed prevailing wage is $56,500. Unfortunately, that LCA claims to use 2005 data. For August 2006, Salary.com says the median is $58,784. The discrepancy here is probably due to wage variations over time.
Another LCA is for a "Software Engineer I" in Greensboro, N.C., and claims the prevailing wage is $44,497. The August 2006 data from Salary.com give $51,372 as the median wage and $45,129 as the 25th percentile wage. In this case it looks like the employer has improperly given an artificially low prevailing wage by claiming the 25th percentile as the prevailing wage.
On nearly every LCA using Salary.com as the wage source and where the job title could be matched exactly, the claimed wage appeared to come from the 25th percentile rather than the median. Prevailing wage claims based on the 25th percentile are invalid under the regulations. Table 4 shows some typical examples.
Sources of Extremely Low
Prevailing Wage Claims
While the wage data suggest that a few employers use the H-1B program to import a small number of highly skilled workers, these are exceptional cases. Overwhelmingly, the H-1B program is used to import workers at the very bottom of the computer programming wage scale.
Many in the information technology industry have called for an increase in the number of H-1B visas available. However, given the very few H-1B workers earning salaries that highly skilled workers in their profession would make and the fact that employers classify most H-1B workers at low skill levels, this report concludes that the existing number of visas is more than ample for the nation’s needs. Contrary to industry claims, the perennial exhaustion of the H-1B visas due to the cap more likely reflects widespread preference for lower-paid workers and lax enforcement of program rules, not an insufficient supply of visas to meet a deficit of highly-skilled U.S. workers.
The appendices to this report contain more detailed breakdowns of the LCA disclosure data. The following appendices are available at www.cis.org/articles/2007/back407data.pdf :
Appendix A — H-1B Computer Programming Prevailing Wage Claims Compared to National Wages by Occupation FY 2005
Appendix B — H-1B Computer Programming Prevailing Wage Claims Compared to National Wages by Wage Source FY 2005
Appendix C — H-1B Computer Programming Wages Compared to National Wages by Occupation FY 2005
Appendix D — H-1B Computer Programming Wages Compared to U.S. Wages by Employer FY 2005
Appendix E — Geographic Distribution of H-1B Computer Programming Workers FY 2005
Appendix F — Top Employers of H-1B Computer Programming Workers FY 2005
1 See for example Norman Matloff, "On the Need for Reform of the H-1B Non-Immigrant Work Visa in Computer-Related occupations," University of Michigan Journal of Law Reform, Fall 2003.
3 U.S. Citizenship and Immigration
Services, Characteristics of Specialty Occupation Workers
4 U.S. State Department, Office of Visa Services, Non-Immigrant Visa Workload, 2005, at http://www.travel.state.gov/pdf/fy%202005%20niv%20workload%20by%20category.pdf .
5 Raghu, K., "Infosys net soars 49.2 per cent, guidance thrills Street," Daily News and Analysis (India), July 12, 2006.
John Miano is an author and expert on the software industry. He may be contacted at firstname.lastname@example.org
The opinions expressed in this article do not necessarily reflect the opinion of ILW.COM.