United States Senate
Committee on the Judiciary
Washington, D.C. 20510-6275

December 22, 1999

The Honorable Janet Reno
Attorney General
Department of Justice
10th and Pennsylvania, N.W.
Washington, DC 20510

Dear Madam Attorney General:

As we discussed last week, the Immigration and Naturalization Service (INS) has contracted with KPMG Peat Marwick to conduct an audit of the H-1B count for FY 1999 and to recommend improvements for the future. As I told you, I strongly support having an audit conducted and indeed, I believe I have suggested one in the past.
I am very concerned, however, that as currently conceived, the scope of the audit may not be broad enough to provide answers either to how many visas INS issued in FY 1999 that should have counted toward the cap or as to what kinds of improvements are needed. INS has consistently described the audit as charged with identifying "systems errors" with the count. But unless a broad construction is given to the term "systems errors" I believe that simply will not be adequate. In order to be of use, the audit must not only identify any computer glitches that led to a miscount, but also produce an accurate and reliable count of H-1B visas INS issued in FY 1999 that should have counted toward the cap.

Specifically, that requires looking not only at any errors the computers may have made in counting what INS wanted them to count, but also at whether INS's instructions on what it wanted the computers to count were right or wrong. There are several persuasive reasons to believe that these instructions were seriously flawed in a way that systematically inflated the count. Accordingly, I am convinced that the KPMG audit cannot possibly be a useful or reliable tool in determining either last year's count or future steps or actions unless it also examines and resolves these questions.

I. How INS Counts H-1B Visas and the Issues This Raises

The first aspect of this dispute that is important to understand is that INS does not "count" H-1B visas in any conventional sense. That is to say, INS does not determine whether each visa it issues counts toward the cap, put that visa on a list, and total the number of visas on that list. Instead, as the matter has been explained to me, it does something quite different. It has computers count either all the H-1B visa applications it has approved or some subset of these visas, and then subtracts out those that do not count toward the cap. Moreover, as I understand it, INS does not even at that point create a list of all the visas that it is "counting" in this fashion, so the results of this approach cannot be verified by checking whether the visas on the list belong there, or whether a substantial number of them do not.

The reliability of INS's count thus depends on the answers to six questions. 1) What is INS trying to count in performing these various operations? 2) Is what it is trying to count what it should be trying to count? 3) What does INS actually capture in the first operation, where it adds up either all or certain H-1B applications approved in a Fiscal Year? 4) What does INS capture in the second operation, where it subtracts out the H-1B applications that it believes should not count toward the cap? 5) If the computers performed these operations properly, would the result actually represent what INS is trying to count? And finally, 6) are the computers performing these operations properly?

I believe we are all in agreement as to the answers to the first two questions. Section 214(g) of the Immigration and Nationality Act places a cap not on visas issued but on "the total number of aliens who may be issued visas or otherwise provided [H-1B] nonimmigrant status during any fiscal year" (emphasis added). As I understand it, INS agrees that this means that generally speaking, in determining whether the cap has been hit, it should count only individuals receiving an initial grant of H-1B status and not all visas issued. This means that a visa counts toward the H-1B cap if and only if it confers H-1B status on a person for the first time, and that the visa counts in the fiscal year that the person first receives H-1B status. Visas or changes in visas awarded to individuals already in H-1B status who change employers, renew their visas, or take concurrent employment (a second job) do not count toward the cap.

The problem comes in the third and fourth steps: what INS counts in the first instance, and what it then subtracts. Despite repeated requests from my staff for information on counting procedures, to date we have not been informed by INS exactly how these steps are actually performed, and I have many questions about them. I do not know who is charged with collecting the data, what visas that person or persons are instructed to include in the initial overinclusive count, or what visas that person or persons are instructed to subtract. I believe my staff has requested this information, along with any other instructions issued by INS or any of its agents relating to how H-1B visas are to be counted or any other documents describing how they are in fact counted, but I do not believe they have received most of this material. Accordingly, I am hereby reiterating that request more formally. Given the timetable on which I understand the audit is supposed to be proceeding, I would hope I might have that material by January 7, 2000. If any portion of the information cannot be provided by January 7, I would appreciate an explanation as to why it is not available, whether it has been provided to KPMG, if it has not how the audit can proceed without it, and a date by which it can be provided. A list of the specific materials I am requesting is attached.

The above assumes that there is and has been a centralized policy and approach. Other information I have received, however, suggests that individual offices, and sometimes different units within offices, may have each handled these issues in their own ways. It also appears that procedures and instructions used in these offices may have changed from time to time. Thus, the information requested should cover all offices, units, and/or contractors who participated in the count and should span, at a minimum, the period from FY 1996 to the present.

II. Evidence From INS's Own Instruction Sheets and Briefings That It Has Been Inflating the Visa Count

Even on the basis of the materials that INS has provided, however, as well as other information that has come to my attention, I believe there is serious reason to doubt that the way in which INS performs these two steps produces an accurate count of H-1B visas it has issued that should be counted toward the cap. First, as to what INS counts in its first, overinclusive count: INS has provided to my staff a document dated December 9, 1999 to the INS Services Division Service Center Operations, styled "Operational Guidance Memorandum No. FY00-003 Subject: Form I-129 and H-1B Cap Definitions and Proper Usage of form I-824." Although I am not entirely sure what this document is, what it appears to be is not directions about what to count to those charged with performing the actual count of H-1B visas but rather guidance to individuals in the service centers charged with adjudicating H-1B petitions who need to know for various reasons whether a particular application would if approved fall under the cap. Nevertheless, I assume that the guidance given to these adjudicators on this question is likely to conform with guidance given to those charged with performing the actual count.

Remarkably, this guidance, issued after considerable staff discussions between my office and INS and dated the same day you and I talked, nevertheless still contains one directive (out of two) on the cap question that is plainly wrong. It tells Service Center officials that two kinds of cases are to be treated as "cap cases": 1) cases where the employer has checked box 2a, "new employment" and box 4a, "request for consulate notification"; and 2) cases where the employer has checked box 2a, "new employment" and box 4b, "request for a change of nonimmigrant status". A blanket rule treating the first category of cases as "cap cases," however, cannot be justified, and is guaranteed to result in counting a significant number of cases against the cap that should not be counted. This is because an employee already in H-1B status may nevertheless be abroad shortly before his or her new employment is starting and may therefore require admission. The person may be taking a vacation between a change of jobs, or may have had to travel abroad on an assignment for his or her current employer, or may have some other personal or business reason for being out of the country. Regardless of the reason, however, if the person is already in H-1B status, the approved visa should not be treated as a cap case. Yet that is the treatment that INS's December 9, 1999 memo directs it to be given. Instead, since other information the employer is required to provide either elsewhere on the application form or in accompanying documentation will inform INS what the current status of the person is, the directive should point the INS officials to that information. [See ENDNOTE 1]

Assuming the instructions to INS officials charged with producing the actual FY 1999 count track the guidance in the December 9 memo, it is therefore highly likely that INS is counting all such approved applications toward the cap in its overinclusive initial count, and that it has no mechanism for subtracting these applications. As I understand it, INS has acknowledged in staff briefings that this is likely to be the case. I do not know how many applications of this type INS approved in FY 1999, but immigration lawyers advise me that it is almost certainly several thousand.

Second, and even more importantly: the documents INS has provided to my staff and other materials made available to the Committee lead me to believe that at least up until December 9, INS was also including in its initial overinclusive count not only most of these visas, but visas issued to anyone already in H-1B status who changed jobs, and others perhaps as well, and that INS had no systematic mechanism for subtracting these out at a later point in the counting process. Immigration attorneys estimate that about a third of H-1B beneficiaries change employers at some point during their stay. Even discounting that estimate somewhat, inclusion of these visas is likely to have inflated the count by 20% to 25%, or somewhere between 23,000 and 28,000 visas.

Here are my reasons for this belief. My staff has only received from INS one other document providing guidance on counting, a memo dated December 3, 1999, which appears to be the predecessor guidance to service center personnel that the December 9, 1999 memo superseded. To date, we have not received any information on what the instructions were before that date (which, it is perhaps worth mentioning, is also the date on which my staff first advised INS how deeply concerned I am about these issues). The December 3 document differs from the December 9 document in several respects, the most significant of which, for our purposes, is in telling employers hiring someone already on an H-1B visa away from another employer to fill out 2c "change in employment" rather than 2a "new employment" (along with 4c "extend or amend stay" if the person is already in the U.S. or 4a if the person is not already in the U.S.).

The December 3, 1999 memorandum at least tries to avoid counting these cases toward the cap by directing that applications filled out in the manner described above are not cap cases. The problem with this approach is that I am told by employers who use the H-1B program and some of the top practitioners in this area that none of them filled out the application form in the way the December 3, 1999 memorandum advised and that instead, they filled it out by checking the 2a "new employment" box. In other words, they followed the approach set out in the December 9 memorandum rather than the December 3 memorandum. They have also informed me that this was generally the advice INS gave when asked how the form should be filled out, notwithstanding any contrary suggestion in the December 3 memorandum. Unlike the December 9, 1999 memorandum, which states clearly that at least an application filled out 2a, 4c is not to be counted, the December 3, 1999 memorandum contains no instructions on whether an application filled out in this fashion should or should not be treated as a cap case. In light of these facts, it seems quite likely to me that up until December 9, 1999, INS counted anything where the 2a box was checked as a "cap case," or at the very least, that whether it did or not was entirely random. [SEE ENDNOTE 2]

Moreover, while the December 3 document is the earliest set of instruction sheets that INS has provided my office, we have also received from others a copy of pages from what appears to be a predecessor document that provides the same guidance as the December 3 memorandum for "change of employment" cases, i.e. fill out 2c and 4c or 4a. I am attaching a copy of these pages to this letter. Thus, it appears that the misimpression among INS officials that employers who were hiring someone on H-1B status away from another employer were filling out the form by checking 2c rather than 2a predates December 3. The predecessor document contains no instructions at all on what to treat as a "cap case," giving me no additional reason to doubt the hypothesis either that all applications where 2a was checked were so treated or that at best the decision whether to do so or not was random.

Assuming that at least a significant number of change of employer cases were erroneously included in INS's initial overcount, it is still possible that INS had a mechanism for subtracting them out on the back end. From the briefings that my staff has received, it appears that INS did have a mechanism for subtracting some of them, but one that was sure not to capture all of them. I am told that INS routinely had the computer cross-check for multiple visas issued to the same individual in the same fiscal year, and if it was confident on the basis of the computer match that it had found such instances, it only counted one of these multiple visas toward the cap. Matching programs of this type, however, can be easily confused by the use of a middle initial instead of a middle name, or the omission of either, or differing expressions of the same names (a particular problem with respect to some of the ethnic groups that comprise a large proportion of the H-1B beneficiary pool), or even handwriting problems such as a numeral in a social security number that one person reads as a "0" but another person reads as a "9". Indeed, I am told that even in the case of individuals who are working for a single employer, employers have discovered multiple file numbers approved for the same individual that are not caught by matching programs of this type. Even more importantly, INS did not look for matches across fiscal years, so if the person's initial employment was in a previous fiscal year, as is likely to be true about five times out of six, INS had no mechanism to prevent that visa from being counted against the cap.

I should add that the fact that INS devised this mechanism further tends to confirm my belief that it realized that at least up until December 9, it likely was systematically including at least a large number of change of employment visas in its initial overinclusive count and that it needed some device to address this overcount. Indeed, the emphasis INS has placed on this mechanism in staff briefings combined with the apparent absence of instructions before December 3, 1999 on what counts as a cap case suggests an even more disturbing possibility: that all approved applications, not only those marked "2a", were being included in the initial "count", with only this subtraction mechanism available to correct the erroneous inclusion not only of change of employer cases, but also of cases of concurrent employment, extensions on current visas, and changes in conditions of employment.

Thus while INS has not yet provided the most relevant documents (its instructions to those charged with counting or any documents those charged with counting have produced themselves), what it has provided demonstrates that even today the guidance it is giving its own employees, if followed by those charged with counting, will include one significant category of visas that should not be included in the count (individuals on H-1B status who change jobs and pick up their new visa abroad) and thereby inflate that count. The documents also strongly suggest that up until December 9, at best a significant number of change of employment visas, and at worst all kinds of other visas as well, were also being included in that count in error. Finally, it seems unlikely that INS's "duplicate check" mechanism could possibly eliminate the vast majority of these errors.

I should also note one other issue that has been raised in this connection. It is not uncommon that several employers may file petitions for the same employee, to whom they have all extended offers. My understanding is that the current system does not contain any mechanism to avoid including all of these petitions in the final count, other than the "duplicate check" performed by the computer, which is flawed for the reasons outlined above. This problem too is almost certainly inflating the count not insubstantially.

III. Other Evidence That INS's Count Is Likely Inflated

In addition, there are other good reasons for suspecting that INS has been including visas in the count that it should not have been including. First, after the H-1B cap was reached in 1999, attorneys have cited many cases to me that involved applications for renewal of an existing H-1B visa or a change of employer that INS erroneously treated as "cap cases" and accordingly denied. In fact, I am told that this even happened to individuals who were not applying for H-1B visas at all, but were instead applying for O or L visas, because INS uses the same application form for these visa categories and for H-1B visas.

Second, in response to inquiries from industry in FY 1996, and in response to my own inquiry in FY 1997, the INS went back and checked its H-1B counting procedures and concluded in both instances that it had double-counted thousands of visas and therefore had not reached the H-1B visa cap at various times when the agency had previously concluded that it had.

Third, earlier this year at least one service center reported that it was having a difficult time getting its contractor to classify only new cases as counting against the cap.

Finally, this past June, the INS released a fact sheet on H-1B visa usage that indicated the number of visas used by various companies in 1998. According to the companies, the INS inflated those firms' actual usage of H-1B visas by at least two to four times each. In one instance it claimed that a company with only about 5,000 employees had used 7,000 visas. INS issued a formal retraction of this fact sheet only a few days ago, on Friday, December 17.

For all these reasons, I believe it is critical to make sure that the KPMG audit that INS has contracted for extend beyond the computer glitches INS has indicated that it wants to see covered. Indeed, the one specific error that it has identified that resulted in a quantifiable counting error was not a computer glitch, but was the same kind of error that I am raising, only in the opposite direction: a "double subtraction" of visas issued but not used leading to an undercount of about 4,500 visas. (I would note for the record, by the way, that INS has also retracted its earlier estimate of a 10,000-20,000 "overage" in FY 1999.)

I apologize for the length and detail of this letter. The problem is that while the issues I am describing are both technical and tedious, if counting errors of the sort I have outlined above have been occurring and are allowed to persist uncorrected, the effect will be to negate a substantial portion of what Congress sought to do when it went to great efforts to raise the H-1B cap two years ago. This, in turn, will negate a substantial portion of the competitive advantage Congress was seeking to maintain for American companies by providing them access to highly skilled workers and thereby keeping the jobs that products created by these new workers generate in this country rather than sending them abroad. On a more mundane level, I thought it important to explain why an audit that does not address these various issues cannot possibly be treated as a serious product, and why it is therefore imperative that the KPMG audit focus not only on computer glitches but be broad enough to resolve these questions as well.

Thank you for your attention to this matter.

Sincerely,

Spencer Abraham

cc: The Honorable Doris Meissner
Thomas M. Conaway, Principal, KPMG
Doug Moll, Senior Manager, KPMG

Endnotes

1. This is one error that the predecessor document to the December 9, 1999 memo, a December 3, 1999 memo that INS also provided to me which I shall discuss shortly, at least tried to avoid, although, as I shall also explain, that memo probably did not succeed in avoiding it either.

2. The only additional instances beyond those listed in the December 9 memorandum that the December 3 memo expressly directs be treated as "cap cases" should in fact be a null set if employers read the instructions on the application form better than INS apparently was able to do on December 3. The December 3 memo states that applications where box 2c "change of employment" and 4b "change of employee status and extension of stay" are checked are "cap cases." The application form, however, directs employers only to check box 4b if they are also checking box 2a and to check only one of the options among 2a, 2b, 2c, or 2d. Thus there should never be a form filled out 2c, 4b. This instruction on the application form further tends ever so slightly to confirm the hypothesis suggested above: that until December 9, at least, INS viewed the checking of the "2a" box as the key indicator whether someone is a "new" H-1B and hence should be counted toward the cap.

Information & Documents Requested

1. Copy of contract(s) regarding H-1B count audit with KPMG.

2. All scope of work and audit design instructions and documents exchanged between INS and KPMG related to the H-1B count audit.

3. All contract provisions with, and instructions to, INS' EDS contractor relating to counting H-1Bs for the cap. This information should cover, at a minimum, the period from FY 1996 to the present.

4. Identification of:

A. Who has been charged with doing the count (in Headquarters and in each office, unit and contractor involved in counting and review for the cap);
B. What situations those persons have been instructed to include in the count;

C. What situations those persons have been instructed to subtract from the count.

This information should also cover, at a minimum, the period from FY 1996 to the present.

4. All instructions provided by Headquarters to the field regarding counting H-1Bs, as well as instructions provided within or between individual offices, units or contractors of INS regarding the counting of H-1Bs. This information should also cover, at a minimum, the period from FY 1996 to the present.

5. Any other documents that relate either to how H-1B's are supposed to be counted or how they were counted. This information should also cover, at a minimum, the period from FY 1996 to the present.

This document may be freely distributed, but please credit AILF as the source.