Minnesotans For Sustainability©
Sustainable: A society that balances the environment, other life forms, and human interactions over an indefinite time period.
Debunking the Myth of a Desperate Software Labor Shortage
Testimony to the U.S. House Judiciary
Dr. Norman Matloff
April 26, 1998
2. There Is No Desperate Shortage of Computer Programmers
3. Rampant Age Discrimination-at Age 35
4. Trends in University Computer Science Enrollments
6. On Skills Requirements
7. On the General Unwillingness of Employers to Retrain Mid-career Programmers
8. The Role of Imported Workers
9. Employers Are Shooting Themselves in the Foot with Their Hiring Policies
10. Author's Background and Further Reading
That is the term used in the software industry when a firm announces a new product which actually does not exist. Extending the term a bit, one can say that the industry's latest vaporware is the claim of a desperate software labor shortage.
The fact is that there is no such shortage.
Due to an extensive public relations campaign orchestrated by an industry trade organization, the Information Technology Association of America (ITAA), a rash of newspaper articles have been appearing, claiming desperate labor shortages in the information-technology field. Frantic employers complain that they cannot fill many open positions for computer programmers.
Note on terminology: I use the term programmer to include software developers having various job titles, including software engineers, system analysts and so on. Similarly I use the word industry to mean all employers of software developers, not just those in the high-tech field. Also, the reason for restricting attention to computer software is that that is where the jobs are; computer specialists outnumber electrical engineers among the foreign nationals hired under the H-1B work visa program by almost 15 to 1.
Yet readers of the articles proclaiming a shortage would be perplexed if they also knew that Microsoft only hires 2% of its applicants for software positions, and that this rate is typical in the industry.1
Call any employer, and they will concede that they receive huge numbers of résumé's but reject most of them without even an interview. One does not have to be a "techie" to see the blatant contradiction here. If employers were that desperate, they would certainly not be hiring just a minuscule fraction of their job applicants.
The major points I wish to make are:
Following is an outline of the details which will be presented in the sections below, concerning the above points:
Age discrimination is rampant in the industry:
Our educational system is producing sufficiently many graduates trained in computer science:
By demanding that Congress increase the yearly quota for the H-1B work visa for importing foreign-national programmers, the computer industry is asking for an expansion of a program that is already overused and badly abused4:
2. There Is No Desperate Shortage of Computer
There are a number of open positions,2 but employers are not willing to fill them with most programmers who apply for them, as can be seen from the extremely low percentages of applicants whom they actually hire, such as3:
There is simply no room for argument here —these low hiring rates (and low offer rates; see below) flatly contradict the industry claims of a desperate labor shortage. On the contrary, the fact that employers can be so picky in their hiring demonstrates an oversupply of labor.
Indeed, when asked about the author's citing of a low 2% hiring rate, Microsoft admitted that it is "very, very selective."5
It also should be noted that it is not just Microsoft that is hiring only a tiny proportion of the applicants. The above companies comprise a broad range of employers, from giants to the tiny five-programmer startups, from the software vendors to the applications firms that write software for their own internal use, and so on.
The situation is typified by the fussy John Otroba (Washington Post, November 30, 1997), who ... has no shortage of incoming résumé's. When he logs onto his office computer every day, he has at least 50 in his electronic mailbox...But only about one in 12 résumé's leads him to pick up the telephone to call the job seeker. Some don't pass that screening step. Of those who come in for an interview, fewer than a quarter are offered jobs [making an overall rate hiring rate of under 2%].
In other words, there is no shortage of "bodies," i.e. there is no shortage of experienced computer programmers. The problem is that employers are not willing to hire them.
Employers are only willing to hire from three narrow categories of programmers:
The companies' résumé-scanning machines search for key words corresponding to currently —"hot" skills desired by the employer. Any résumé lacking these words is rejected, untouched by human hands.
The same is true for the employment Websites set up by most companies in the industry, which filter responses based on skill sets and reject any applicant who lacks the given skills.
Moreover, the industry claims of a labor shortage are even more strongly contradicted by the fact that even among applicants who have the skills demanded by "picky" employers, few are made offers.
Patrick Schmidt of Deltanet notes that the programmer employment agencies he uses will only refer an applicant to an employer if the applicant is an exact match to the skill set defined by the employer-and yet even then Schmidt says he hires well under 10% of such applicants, due to the large number of agencies which send him applicants.
In this light, it is very instructive to look at offer rates, meaning the proportion of those made offers among those who are interviewed (in person, not just on the telephone). Those who are interviewed have already been prescreened for skills criteria; the employer will have chosen the applicant's résumé because of specific skills listed, and will typically performed a mini-interview with the applicant by telephone.
Here are some typical offer rates6:
Note that these low rates are for offers, not hires. Thus the low rates cannot be explained away, for instance, by postulating that an applicant gets multiple offers but can only accept one, or by suggesting that many résumé's are casually submitted via e-mail by programmers who may not really be in the job market but are merely "testing the waters." So we do indeed see that employers are very picky in their hiring. Again, note that Microsoft admitted this, and we saw that its hiring rates are typical for the industry.
Instead, it is clear that even if one grants the employers' claim that they must hire someone with a given skill set (which I strongly disagree with), they still are being very selective in their hiring-contradicting their claims to be "desperate" to hire.
Much has been made of dizzying claims of numbers of open positions made by the ITAA and its partners (Dept. of Commerce and the Virginia Polytechnic Institute). The methodology underlying these claims has been criticized by the General Accounting Office7 and the Bureau of Labor Statistics. However, as Urban Institute/American University economist Robert Lerman pointed out in his testimony before the Senate Judiciary Committee on February 25, 1998, the size of the demand for labor is irrelevant anyway; what matters is the difference between the supply and demand, and these studies do not address that question. Instead, Dr. Lerman points to wages, whose mild rate of increase does not indicate a massive labor shortage.8
Similarly, the ITAA claims a 10% vacancy rate for IT positions —but does not mention that the industry always has had high vacancy rates.
ECbridges' Raymond Lim even considers 10% low, saying that rates of 20% were typical a few years ago.9
3. Rampant Age Discrimination —at Age 35
Mid-career programmers actually have a very difficult time finding programming work, so much so that large numbers of them leave the field.
There are two underlying factors here:
Intel lobbyist Eva Jack conceded to Computerworld magazine (IEEE Computer, February 1996) that the firm often focuses its hiring policy on new graduates. Tim Jackson's book, Inside Intel (Dutton, 1997), provides a number of disturbing details:
Intel's focus on new or recent college graduates is so intense that it even has a special acronym for the term, RCG (Recent College Graduate), which dominates its employment recruiting literature. In a television debate between the author and Intel representative Coeta Chambers,11 the author offered to give Intel a list of mid-career unemployed or underemployed software specialists who are seeking jobs. Intel declined the offer.
There is no better way to show the industry's emphasis on youth (and its general disinterest in workers who are in their 40s or even 30s) than to look at how firms in this field define "senior" workers.
Consider for example the employment Web page of Sun Microsystems,12 one of the most vocal firms claiming a labor shortage. One of the first questions asked of the job seeker there is "Experience Level," which of course is a proxy for age, and thus possibly an illegal question. But even more interesting is the choices the user is given for answers to this experience question:
[ ] Entry Level (0 - 2 years )
[ ] Intermediate (3 - 5 years )
[ ] Senior (6+ years )
Leiden was one of the drafters of legislation proposing an increase in the yearly cap on temporary work visas for foreign professionals. During the debate Leiden defined "mid-level" programmers to be those having only "a year or two of experience."
Employment agents tell the story clearly. Andrew Gaynor notes15
that anyone with 10 or more years of experience without currently —"hot"
skills "is at a complete disconnect" in finding work. Susan Miller
says16 that former defense industry programmers "are usually
shunned by the industry.
A July 14, 1997 article in the Washington Times reported:
The article then quotes NeuroSystems CEO Ed Robertson as to why the new graduates are so attractive: "If we go out to the marketplace and find a 40-year-old software engineer, we'll have to pay that individual more."
One CEO (Bob Forman of IMI Systems) who heard me speak on this subject approached me after my presentation and said, quite angrily, "You are wrong that the industry does not hire mid-career people who don't have hot skills. My company is anxious to hire as many programmers as it can get."19
I then suggested that my wife, a software engineer with 15 years of experience, send his company her résumé, and that if the résumé were rejected without an interview, I use his firm as an example in my future writings on this topic. He quickly backpedaled, saying that my wife's résumé might be rejected because the firm perceives that her salary is too high.
My reply to Forman was that this indeed is the problem. The industry claims there is a shortage of workers when what they really mean is that there is a shortage of cheap workers, in the form of new college graduates and imported foreign nationals. I discovered later that Forman is Chair of the ITAA's Immigration Policy Committee.
As a result, most careers in the programming field are short-lived.
Below are data, extracted from the National Survey of College Graduates in 1993, showing the percentage of computer science graduates working in software development various numbers of years after they finish school:
Five years after finishing college, about 60 percent of computer science graduates are working as programmers; at 15 years the figure drops to 34 percent, and at 20 years —when most are still only age 42 or so— it is down to 19 percent. Clearly part of this attrition is voluntary, but most are forced to seek other work when they see the handwriting on the cubicle wall: Employers do not want to hire mid-career programmers.
As noted earlier, because people who cannot find programming work leave the field, unemployment statistics for programmers are meaningless. The former programmer who becomes an insurance agent counts in government statistics as an employed insurance seller, not an unemployed programmer, so unemployment rates do not give an accurate picture of the employers' general refusal to hire the mid-career workers. Nevertheless, it is significant that there was a high 17% unemployment rate for programmers over age 50 as of August 1997.20
This was questioned by reporter Miranda Ewell, San Jose Mercury News, April 5, 1998, who said that such information is not available, and who gave other statistics which, though not exactly measuring the same quantity as in the Computerworld article, seemed to be at odds with it.
However, I have checked with the author of the Computerworld article, Laura DiDio, who explained that she doggedly went through call after call to the BLS to get the exact information she wanted, and she finally did find someone who was able to provide it.
As a stroll around any high-tech company will show you, most people who work in this field are young. It should be noted that other technical fields do not show this rapid decline of work in their field. For example, consider civil engineering majors. Six years after graduation, 61% of them are working as civil engineers, and 20 years after graduation, the rate is still 52%. True, Some computer science majors eventually go into management, start their own consulting firm and so on, but the same is true for civil engineers. Careers in programming are far shorter than in civil engineering, even though both fields are technical and require attention to detail. The difference is that skill sets change rapidly in programming, but not in civil engineering. And again, it is not that programmers are incapable of acquiring the new skills, but rather that the employers won't give them that chance.
Such policies have spawned a cottage industry in self-help books for programmers who find that they are no longer desired by employers, such as The Computer Professional's Survival Guide, Downsized But Not Out: How to Get Your Next Computer Job and The Programmer's Job Handbook: The Skills You Need for Long-Term Job Security and Programming Success.
After seeing me quoted in the press on this topic, many mid-career programmers have sent me laudatory e-mail, saying my description of the plight of such workers fit them perfectly.
Here are a few geographically-diverse samples:
(From a man in Tennessee):
(From a man in the San Francisco Bay Area):
(From a woman in Seattle):
(From a man in New York City):
(From a 47-year-old man in the San Francisco Bay Area):
(From a man in the Southwest):
(From a 27-year-old man on the West Coast who graduated three years ago):
(Similar themes were published in letters to the editor in the San Jose Mercury News, January 24, 1998, as well as in articles in US News and World Report, March 16, 1998, the San Diego Union-Tribune, March 7, 1998, the San Francisco Chronicle, March 9, 1998 and Wired Online News, February 25, 1998.)
University students are beginning to be aware of this problem, and though computer science enrollment trends are currently on the upswing (more details below), in the future this may deter many of them from pursuing computer science majors. An article in the January 13, 1998 edition of the New York Times says that [current Stanford computer science student Graham Miller] is already thinking about an exit strategy [from the computer field].
"Programmers only last up to 10 years or so," Miller said. "After that, you need to find something else to do."
It is crucial to keep in mind that the plight of the mid-career programmer cannot be solved simply by the programmer taking some refresher courses in the new software skills. Even if a programmer takes a course in, say, the new Java programming language and then applies for a job requiring Java, employers will still not hire him or her. As noted by software employment agent Maryann Rousseau, "Taking a course is just not going to work for a senior person, given his salary." Why hire a newly-retrained but more expensive 40-year-old when a newly-trained cheap new graduate is available?
The ITAA claimed throughout 1997 computer science enrollments are declining, and is calling on the federal government to fund programs to attract more college students to the field. But ITAA's assertion was so misleading as to border on fraud.
The ITAA report lists declining computer science enrollments
from the late 1980s to 1994. But computer science enrollment reversed its
declining trend in 1995, increasing by 5% in 1995-1996, and by a whopping 40%
nationwide in 1996-1997, and then by another 39% in 1997-1998, according to the
Computing Research Association (CRA), a national consortium of university
computer science departments.21
This information (the original 40% increase) was conveyed to ITAA's Harris Miller and Tony Vickers by a CRA official when ITAA distributed a preliminary draft of their report at a roundtable discussion organized by the Stanford Computer Industry Project on February 19, 1997. Though ITAA stated at the time that they were soliciting comments and suggestions for improving their report, they did not include this information about the sharp increase in computer science enrollment in the final version of the report, apparently because it undermined their argument. In fact, ITAA continued to claim enrollment is declining,22 even after ITAA's suppression of the 1995 reversal trend was brought up in an interview with ITAA by the Electronic Engineering Times (September 29, 1997), until forced to stop when even the Department of Commerce found ITAA's claim to be untrue. Since that time, the ITAA statements have avoided using the present tense in the word decline, but they continue to obfuscate the issue by discussing what happened in 1994, leading the listener to believe the situation still holds today.
Moreover, even industry representatives (including ITAA) interviewed by Business Week (March 10, 1997) blamed the earlier decline in enrollment partly on "a glut of programmers in the mid-1980s." The recent increase in enrollment is due to the rapid expansion of the industry which began around 1994. In other words, market forces are working quite well here; the supply of computer science students has been quite elastic to demand. So, quite contrary to ITAA's assertion that students do not want to study computers (due to claimed "nerd" images) or do not have the background to do so (see discussion on mathematics below), the fact is that computer science enrollment has responded quite well to labor markets demands.
Those new computer science students are at the advanced stage of the pipeline, and are graduating now. In 1996-1997, the year of the 40% increase in enrollment, there was already a 10% increase in the number of computer science graduates, according to the Computing Research Association.
The industry lobbyists also say that college computer science curricula are only producing 25% of the nation's needs for programmers, again claiming this is due to a decline in enrollment.
But it has always been the case that programmers have always come from many different fields, not just computer science. For instance, according to the National Science foundation's SEASTAT data, only 26% of all those working as programmers in 1993 had computer science degrees. In other words, the situation today is no different from the past, quite contrary to the alarmist tract written by the ITAA.
In this light, it should be mentioned that Clifford Adelman of the federal Department of Education has found that large numbers of non-computer science majors take at least mid-level courses in computer science.23
It should also be noted that most firms recruit at only a few colleges. For example, at San Francisco State University, right in Silicon Valley's back yard, only a handful of computer industry employers do on-campus recruiting.24 Even Intel, with its heavy emphasis on new or recent college graduates, does not recruit there.25 At the University of Nebraska, Microsoft does engage in on-campus recruiting, but very few other high-tech firms do.26
TJ Rodgers of Cypress Semiconductor, testifying in support of claims of an engineering labor shortage,27 stated his firm recruits at only 26 colleges nationwide, out of the 320 which have engineering programs accredited by ABET; Cypress was not listed in the yearly recruiting schedule posted at the Massachusetts Institute of Technology, the nation's premier engineering school.
Indeed, even the ITAA's Harris Miller has admitted this to Tech Tech Week' Umberto Tosi28:
The ITAA also claims that American students do not study computer science because (a) they think it is "nerdy," and (b) the lack math skills. But the fact that computer science enrollment has been skyrocketing in the last few years shows that there are plenty of students with the interest and background to study this subject. By the way, one does not use math in software development in the first place.29
In very narrow segments of the programmer labor market, some salaries have indeed risen substantially. As stated earlier, employers are overdefining job requirements, with ads like "Must have experience writing C++ code for TCP/IP applications on SPARC platforms." The pool of programmers satisfying such conditions is of course small, thus raising salaries for those within that narrow pool.
However, outside of these subsegments, programmer salaries are
not rising rapidly. Bureau of Labor Statistics data show that salaries of
programmers overall —i.e., both the ones who have currently —"hot"
skills and the "ordinary" programmers— rose about 7% from 1996 to
1997, compared to about 3% for all American workers.
The ITAA criticizes the BLS data as being inaccurate, yet TJ Rodgers, CEO of Cypress Semiconductor, testified before the Senate Judiciary Committee on February 25, 1998, said he is raising salaries by 6%, quite in line with the BLS figures.30
Comparisons of 1996 and 1997 salaries in Silicon Valley by the employment agency Heuristics Search Inc., were presented in Tech Week, March 9, 1998. Salaries were tabulated for five areas of skill sets (software engineers, client/server, communications, database and graphical user interfaces), over four levels of experience (0-2 years, 3-8, 8-9, 10 or more).
Once again, the differentials between 1996 and 1997 were in most cases in the 6-9% range.
Similar (and somewhat lower) figures were found for other regions of the country in the Datamasters survey, cited in the same article. This survey also features detailed salary information yearly from 1990-1998, extremely useful. It is available at < http://www.datamasters.com/dm/survey.html >.
Salaries for new college graduates in computer science rose 3.9% during 1996-1997.31 Qualcomm, another firm which insists there is a high-tech labor shortage, admits that its starting salaries have only be rising about 4% per year.32 Starting wages for new computer science graduates of UC Berkeley have been rising at the rate of 5% yearly.33 And in spite of wild newspaper stories about new Bachelor's graduates getting salaries approaching six figures, the going rate is in the low $40,000 range.34
The ITAA says that salaries are less relevant because of nonsalary compensation such as bonuses. Yet the ITAA made no claims that nonsalary compensation is rising at more than the 7% rate we see in salaries. The key point is not that there exist nonsalary forms of
Compensation —this has always been true in this industry, since 1980 or earlier— but rather whether the amount of such compensation has been rising. If overall compensation, both salary and nonsalary, is rising only at a mild rate like 7%, then employers are obviously not desperate to hire.
Moreover, the size of nonsalary compensation in this field is usually small in proportion to salaries. Most programmers do not get bonuses, and those that do average about 5% of their salaries. (See the Datamasters surveys cited above.) Except for a few cases of "miracle" companies, the profit made through stock options in established firms is on average also a small portion of salary, typically about 1%.
Of particular interest is a special kind of bonus, the "bounty" paid to an employee who introduces a friend to the firm and who is hired.
At Oracle, for instance, the size of this bonus is currently $1,000 for a job paying more than $40,000, and $500 for one paying less than that. The size of this bonus has been constant at Oracle for several years. In fact, the figure ($1,000) was common in the industry even in the 1980s, so we once again do not see evidence of escalating levels of desperation among employers.
6. On Skills Requirements
Employers claim, "We cannot hire `just any programmer.' It is crucial that the programmer have experience in a specific software technology." But this is quite incorrect. What counts is general programming talent, not experience with specific software technologies.
Bill Gates has described Microsoft hiring criteria thusly: "We're not looking for any specific knowledge because things change so fast, and it's easy to learn stuff. You've got to have an excitement about software, a certain intelligence...It's not the specific knowledge that counts." (Wall Street Journal, November 8, 1994.)
Jim McCarthy, one of Gates' software development managers at Microsoft, points out in his book, Dynamics of Software Development (Microsoft Press, 1995, p.168), The biggest mistake I see managers make as they hire people for software development teams is that they overvalue a particular technology. To verify this tendency, all you have to do is look at the want ads: `Wanted: foobar programmers. Experience with whatsit required.' Obviously, conversance with a given technology is a wonderful attribute in a candidate, but in the final analysis it's an extra, not mandatory. After all, most software development technologies have a half-life of about one year.
Ironically, Microsoft has grown so large that Gates' and McCarthy's philosophies don't reach down to the shop floor, and managers there are now just as obsessed with skills as the rest of the industry. The first demand made of users accessing Microsoft's employment opportunities Web page (and those of most other software firms) is "State your skill set."
Programmers can become productive in a new software technology in a month or so. As Garrent Bechler, a recruiter with RHI Consulting in Walnut Creek, California put it, "Any programmer who already knows C [the industry standard for the last 15 years], needs only a week, maybe two, to reach proficiency in Java."35
This point on the quickness with which new software technologies are learned can be seen in data on factors affecting completion time for software development projects, cited in one of the central works on software engineering, Software Engineering Economics, by Barry Boehm (Prentice-Hall, 1981, p.530). Those data indicate that programmers reach perhaps 80% of their full productivity level by one month, and full productivity by the next time period studied, four months.
Amit Kamra, head of Information Systems Transition Services, said that his company could not afford to hire someone who would have to learn the given technology on the job, say Microsoft Windows programming.36
But when I asked him how long it would take for an experienced programmer to become productive in Windows if he/she did not know this technology beforehand, he answered, "[Up] to two weeks, maybe all the way up to a month and a half to become truly productive."
I asked why they did not hire such people, given the shortness of such time periods, to which Kamra replied, "Well, we could, and we did so once with good results [he then gave the details]...But well, during those two weeks [of learning] the project is slowed down a bit, especially since others on the project would have to help the new person."
Though Kamra's remarks show that learning on the job is of course not ideal, the point is that they certainly show that the industry is wrong in claiming that possession of specific skills is an absolute necessity.
Note that the Boehm reference and the Kamra quote above, as well as numerous others in my long Web document referenced earlier, demonstrate that programmers in general learn quickly, not just the programmer "superstars."
7. On the General Unwillingness of Employers to Retrain
However, the fact is that most employers are not willing to retrain their programmers and engineers. As mentioned earlier, Kim Lee, of the Network Connections employment agency in the Silicon Valley has noted that37 Pointing out frankly that her own high income as an employment agent depends largely on the fact that the industry is not providing retraining for existing employees, she nevertheless feels that It's a very closed industry in that respect [retraining]. The trap the industry falls into is that they don't spend time retraining.
It would be much more cost-effective for them to retrain the employees they already have; by not retraining they are driving salaries way up, since so few people have the ``right'' skill sets.
The employers haven't been smart. They have been very closed-minded, with blinders. If I could change one thing about the industry, that would be it.
Andrew Gaynor, another employment agent, called the industry "very short-sighted" in this regard.39
The industry has been a mass of contradictions on the retraining issue. Intel has said that it does retrain (Bay TV, San Francisco, March 3, 1998) and that it is not willing to retrain (IEEE Computer, February 1996; also CNN, March 1998). Cypress Semiconductor also said yes (California Report, NPR, March 27, 1998) and no (The NewsHour with Jim Lehrer, April 3, 1998) to the retraining question.
The industry often says it cannot afford the time to retrain, because they need to hire personnel for a project immediately. But this is disingenuous, as the industry is also claiming that as many as 30% of software positions take six months or more to fill.40
During this time the same programmers employers are unwilling to hire now because of lack of a hot skill could learn the given technology and be productive.
The industry also says that if they retrain their programmers in a "hot" skill, the newly-enfranchised programmers will leave them for higher pay elsewhere. This is correct, but the employers are missing the point: If the industry did not pay a premium for these skills in the first place (a consequence of refusing to hire programmers who lack the skills), this frequent job-hopping would not occur.
As pointed out earlier, programmers seeking work cannot remedy the problem by retraining on their own. Employers will not hire a mid-career programmer for a Java project on the basis of the applicant's having taken a Java course.
For this reason, retraining programs for engineers and programmers are largely a waste of money, not helping employers cope with their claimed shortage of such professionals. Such programs often have a high placement rate, but the problem is that the engineers and programmers who "graduate" from them don't get jobs as engineers and programmers. Then tend to get jobs as technicians, customer support personnel and so on.41
Even the Massachusetts Software Council, widely viewed as the best of the software retraining programs, only places 20% of its participants in software development positions. These don't help employers fill the engineering and programming positions they say they are desperate to fill. The retraining programs may not be so beneficial to the engineers and programmers either, since the jobs are often lower-paying and of a nature that the participants in the programs could have obtained on their own, without the programs. Gene Nelson, for example, had been making $46,600 as a programmer before being laid off, and now makes only $23,000 staffing a software help desk. These programs are analogous to what would occur if there were unemployment among doctors, and they were "retrained" for paramedic positions.
Anyone who thinks that "education is the answer" should consider the case cited in the Sacramento Bee, March 14, 1998:
One such prospective high-tech employee [with an advanced degree in computer science who cannot find work] is Peter Van Horn, 31, who is looking for a job in computer graphics. He has an undergraduate degree in aeronautical engineering and a master's in computer science from California Polytechnic State University, San Luis Obispo.
In nearly four months of looking for a job, he has applied to more than 38 companies and has, so far, talked to only two. "At Cal Poly, I always heard how great the market was, how if you have a degree in computer science you could get a job," said Van Horn, now a Bay Area resident.
"My credentials are good...Companies are constantly talking about a shortage of workers, but if that were the case, you would think I'd have more than two interviews."
Van Horn has done "all the right things," everything society told him to do, and yet he could even get an interview, right in the middle of Silicon Valley, for four months. In addition to sending out résumé's on his own, he also was working through employment agents, but with no results.
After four months, at the end of March 1998, Van Horn did finally get a programming job, but his experience shows quite graphically that these supposedly "desperate" employers are not so desperate after all.
Van Horn's experience is shared by Bard-Alan Finlan, age 43, who also went back to school and was shunned by employers after he graduated. (San Diego Union Tribune, March 7, 1998.) Armed with a new computer engineering degree from UC San Diego, he applied four times to Qualcomm, a large San Diego firm which claimed to be desperate to hire engineers, and yet Qualcomm did not even give him an interview. He had no luck with all the other firms he applied to either; at the time the newspaper article appeared, Finlan had had only one interview in a year and a half. He finally did secure a job, but even then it was only as a technician, a job typically paying only half what an engineer makes.42
8. The Role of Imported Workers
As we have seen, industry employers tend to shun mid-career programmers. One of the major factors underlying this is that employers have another labor source to turn to, in the form of foreign nationals whom they sponsor for immigration or work visas. For example, about one-third of Silicon Valley programmers and engineers are foreign-born, most of them sponsored for immigration originally by employers.43
Without the foreign-national labor supply, employers would be forced to use the existing domestic labor pool of mid-career people. For this reason, the foreign-labor issue is central to our theme here, and will be addressed briefly. (See the author's analysis at < http://heather.cs.ucdavis.edu/svreport.html > for extensive details.)
It is easy to see that there is something very wrong with the H-1B program, in that its growth rate is far out of balance with the growth rate of jobs: The number of H-1B work visas requested by industry for computer programmers increased by 352% from 1990-1995, during which time the number of programming jobs increased by only 35%.44
The "country of choice" among employers importing this labor is currently India. Many employers find foreign-national programmers and engineers attractive because they will accept lower salaries and poor working conditions. The Department of Labor has found widespread abuse of the work-visa program.45
Among other things, they found that 19% of the employers were not even paying the salaries they had promised in their H-1B applications, even more remarkable because the salaries in the applications tend to be low to begin with. The employer requesting an H-1B is supposed to pay the prevailing wage, but there is such a large variation in wages anyway that it is easy to mask an offer of an unfairly low salary.
DOL also found that though state employment departments are supposed to refer domestic candidates for jobs for which an employer-sponsored green card is pending, "Of the 28,682 applicants referred on 10,631 job orders during the period, only 5 (0.02 percent) were hired."
General Dynamics, the aerospace giant, even admitted in federal court that the imported workers from England were presented as attractive due to their "indentured'' status; the pitch made by the British employment agency to General Dynamics said that its clients were "prepared to work here in the United States for as much as a 40% reduction in current United States salary levels."
A researcher with the pro-immigration Carnegie Endowment for International Peace put it this way46:
Papademetriou and Stephen Yale-Loehr —who is an immigration lawyer and thus would be expected to oppose reform of the H-1B process— reported in their book, Balancing Interests: Rethinking U.S. Selection of Skilled Immigrants (Carnegie Endowment for International Peace, 1996), on their study of wages paid to foreign nationals in various professions.
In data from the labor certification applications in the process of sponsoring the foreign workers for green cards, the foreign applicants in Computer Programmer positions in New Jersey were being offered salaries which were on average 21% below the mean for that profession, with an 11% figure in Texas. In the Computer Systems Analysts and Scientists category, gaps of 30% and 21% were found in New Jersey and New York, respectively. By law the gap is supposed to be no more than 5%.
Moreover, Department of Labor regulations allow the employer to provide his/her own data on prevailing wages, such as listing typical salaries in his/her own firm, rather than being determined by the DOL, clearly producing enormous potential for abuse.
Even if an H-1B employer pays a prevailing wage determined by a government survey, that wage will usually be lower than the market rate for the job's skill requirement, as follows. As explained earlier, the only programmers who are enjoying large increases in salary as those with "hot" skills, say Java. H-1Bs are brought to this country ostensibly for those skills. Yet the employer need only pay the prevailing wage for programmers in general, rather than the prevailing wage for, say, Java programmers. Thus the employer gets a Java programmer for the price of a generic programmer —all while complying with the prevailing-wage requirement of the law. As noted by immigration attorney Donna Fujioka of Oakland, California,47
"[The prevailing wage system] takes a meataxe approach...It doesn't appreciate how hot a skill is [such as SAP]...This is great if you are an attorney representing an SAP programmer."48
But as seen in my "short-lived career" data above, almost no one lasts 25 years in this field, or even close to it, so the point is moot. Moreover, the salary curve is steepest in the first few years of experience.
Asian-American Studies Professor Paul Ong of UCLA, after correcting for a host of important variables-including English proficiency-found that immigrant engineers were paid up to one-third less than their native counterparts, and that the gap took 20 years to close.49 And though Ong hypothesized various factors, he noted that "Companies took advantage of immigrants."50 An industry analyst in Bangalore, India quoted by MSNBC News in August 1997 also says that Indian programmers imported to the U.S. under the H-1B program make 30% less than their American peers.51
The author's own analyses of the 1990 Census data on programmers and electrical engineers in Silicon Valley found that the immigrants were paid on average 15-20% less than natives of comparable age and education. In one of the analyses, for instance the author tabulated salaries in Silicon Valley, for workers who had Master's degrees (and not a Ph.D.), and at most 32 years old. For the foreign-born, the worker was included if his entry to the U.S. had been no more than eight years earlier. The author then simply computed mean salaries for all native and all foreign-born.
The results were:
The native figure is 20.2% higher than the foreign-born one.
Abuses also occur abroad; Sun Microsystems, a firm often cited by ITAA analyst (and now Senate Immigration Subcommittee staffer) Stuart Anderson as paying fair wages to foreign nationals, has boasted of hiring programmers in Russia at "bargain prices."52
It is ironic that even a Wall Street Journal article claiming that American firms recruit abroad because of a labor shortage stated that "recruiting foreign talent is cheaper than hiring Americans."53 The article quotes an American recruiter of foreign programmers as saying that he pays them $20,000 to $25,000 less than Americans with the same skills.
Moreover, the simple law of supply and demand tells us that, again, even the sincere employers who hire H-1Bs bring down the price of labor, by increasing the supply. Industry officials have admitted this, such as this comment on CNN on February 9, 1998:
Robert Walley, executive vice president of Gemini, says that unless his company and others are able to find a new source of workers, "it would increase the prices of the resource pool. The people out there looking for jobs, they're demanding premium salaries now, and it will just drive that higher."
Amazingly, policy makers in a federal agency, the National Science Foundation (NSF), actually planned for a glut of labor in science and engineering, at least at the postgraduate degree level. In early 1998, Dr. Eric Weinstein, a mathematician at the Massachusetts Institute of Technology, uncovered internal NSF documents which expressed concern that science and engineering salaries were getting too high, and proposed as a solution to this "problem" bringing in a glut of foreign labor. It is amazing that a federal agency would actually plot to keep U.S. citizens' salaries down. The NSF, by the way, is one of the chief architects of the modern H-1B program. Dr. Weinstein's paper on this scandal is available at < http://www.mit.edu/afs/athena.mit.edu/user/e/r/erw/Public/SG/NSF.html >.
The industry lobbyists say it actually costs them more to hire the foreign nationals because of the legal fees involved. This is one of their most misleading arguments. First of all, filing for an H-1B is quite simple, and the typical legal fee for it is only about $1,500 for small employers who hire only a few H-1Bs, and down to about $700 for large employers who file many H-1B applications.54
Here are some examples.
An employer in Washington DC told the author in December 1997, "Most attorneys around the U.S. charge $1,000 to $1,500 for an H-1B petition." An attorney in the San Francisco Bay Area put the typical figure at $1,200 to $1,500 in a discussion with the author on March 5, 1998. Robert Baizer, a San Francisco immigration attorney, in an interview with the author on March 5, 1998, gave $1,500 to $2,000, and also is the source of the $700 figure. See also David North, Soothing the Establishment: the Impact of Foreign-Born Scientists and Engineers on America, (University Press of America, 1995, p.52.
This is far less than the $10,000-$20,000 ITAA claims for the H-1B.55
Second, many employers have the foreign employees pay the legal fees (for both kinds of processes) themselves, and even when employers foot the bill, the cost is often less than they save in salary.
Note that an H-1B employee is essentially immobile during the two or three years while the green card is pending, thus refuting ITAA's argument that H-1Bs who are exploited in terms of salary can simply move to another job. The workers certainly do not want to start the green card process all over again. The anonymous author of an op-ed piece in TechWeb News, March 16, 1998, wrote, I am an immigrant from India...the H-1B visa allows someone to work only temporarily at a high-tech job for a few years. An employer has to sponsor one for an H-1B visa. These engineers cannot switch jobs at will. To do so requires a new H-1B visa.
Companies love these H-1B workers, as they are eager to please their sponsors [in the hope] that they can be sponsored for green cards. These engineers are virtually "indentured slaves" of their sponsors.
Once a company initiates the process of sponsoring a candidate to green card, it can currently take three to four years. Companies love this and frequently delay the process on purpose. Some big companies have this delay built into their sponsoring process.
During this period, candidates are virtual slaves. They are forced to work long hours at low wages. And usually they do not get good raises or promotions...I myself left my company when I got a green card and I got a raise of 40 percent.
Industry lobbyists have threatened that if the yearly cap on H-1B work visas is not raised, employers will ship software work to foreign countries. Yet such extortionary language is not backed by action, and in fact will not be in the future. While it is true that some companies are experimenting with having work done abroad, this will not become the major mode of operation of the industry. The misunderstandings caused by long-distance communication, the problems of highly-disparate time zones and so on result in major headaches, unmet deadlines and a general loss of productivity. See the author's analysis at < http://heather.cs.ucdavis.edu/svreport.html > for extensive details on this point, including many quotes from industry figures.
For example, Bill Gates says,56
It is my opinion that in the case of foreign nationals of
extraordinary talent, our immigration law should indeed facilitate the ability
of employers to hire such workers.
However, workers of extraordinary talent comprise only a small fraction of the overall population of H-1Bs and employer-sponsored green cards.
The industry lobbyists say that the H-1Bs are needed to retain the industry's technological edge, but the fact is that the vast majority of technological advances in the computer field have been made by U.S. natives. This can be seen in rough form, for example, in the fact that of the 56 awards given for industrial innovation by the Association for Computing Machinery, only one recipient has been an immigrant. Of 115 U.S. recipients of computer-related awards given by the Institute of Electrical and Electronic Engineers, only nine of the recipients have been immigrants.
Mary Dumont, a Palo Alto attorney representing Californians for Population Stabilization in a lawsuit against Hewlett-Packard's hiring of Indian engineers via the Tata Corporation.
Dumont describes the judge's questioning of a Hewlett-Packard representative. When the judge asked about the quality of the imported Indian workers relative to natives from, say, the nearby University of California at Berkeley, the Hewlett-Packard executive conceded that the UC graduates were better. Recall that Sun Microsystems, which claims to scour the globe for "the best and the brightest," seems to be also interested in the cheapest; it boasted to the Los Angeles Times that it had employed programmers in Russia "at bargain prices."
Lobbyists also point to the fact that about 40% of U.S. Ph.D.'s granted in engineering go to foreign students. (Note: Some newspaper reports have erroneously stated that large numbers of U.S. undergraduates in computer science are foreign students. This is incorrect; only 6% of the undergrads nationwide are foreign students. (Computing Research News, March 1998.))
But this ignores the fact that we are overproducing Ph.D.'s in the first place. A report by William F. Massy of Stanford University and Charles A. Goldman of the RAND Corp., The Production and Utilization of Science and Engineering Doctorates in the United States, studies the problem in great detail,57 finding for example that we are overproducing Ph.D.'s in electrical engineering by 44%.58
Again, we do not need to produce so many Ph.D.'s in the first place.
However, it is interesting that the National Science Foundation actually promoted policies which they knew would result in low enrollments of domestic students in Ph.D. programs. Recall our earlier discussion of NSF's plan to hold down wages by bringing in a glut of foreign scientists and engineers. The NSF documents acquired by Dr. Weinstein reveal that NSF realized that by holding down Ph.D. salaries they would cause domestic students to lose interest in Ph.D. programs, while foreign students would still enroll in those programs as steppingstones to immigration. Since the lobby for increased H-1B quotas has often made use of data provided by allies in the NSF, Dr. Weinstein's discoveries take on special significance.
Another class of foreign nationals sponsored for H-1Bs consists of university instructors and researchers. While this is outside the scope of our report here, I wish to point out that there is no shortage here either-our Computer Science Department has been receiving approximately 400 applicants each year for faculty positions (for typically two or three openings).
During the 1995-1996 Congress, Senator Alan Simpson (R-Wyo.) proposed that a fee be imposed on employers who hire H-1B workers. Originally set at $10,000 and then lowered to $5,000, the fee would go to retraining domestic workers. The industry lobbyists furiously opposed the proposal (which was then dropped), and expressed the same fierce opposition in February 1998, even for a tiny fee of $250. Clearly, this shows that the insincerity in the lobbyists' claims that they are desperate to hire people, that they are not seeking H-1Bs as a source of cheap labor, etc.
The fact is that the industry lobbyists are not doing right even by their industry constituents, because under current hiring policies the employers are shooting themselves in the foot59:
Studies have shown that programmers who are twice as productive are paid only 10% more. (Peopleware: Productive Projects and Teams, by Tom DeMarco and Timothy Lister (Dorset House Publishing Co., 1987.)
Employer obsession with skills is resulting in sharp increases in salaries within the very narrow segments of the software labor market corresponding to those highly-specific skills. (Again, programmers outside of those narrow segments are not experiencing sharp increases in wages.)
It is simply not cost effective to pay someone $10,000-15,000 more in salary simply because he/she knows Java, given that any competent programmer can learn Java and be productive in it within a couple of weeks. So employers are shooting themselves in the foot under their current policies.
The fact that the industry pays a premium for certain skills is resulting in frequent job-hopping by programmers who are out to maximize their salaries. Employers say they place high value in finishing projects under deadline. Yet if a programmer who knows a project inside out suddenly leaves the employer in the lurch by jumping to another company, clearly this has a sharply adverse effect on the first employer's ability to complete the project on time.
So here too, employers are shooting themselves in the foot under their current policies.
By using unimportant skills as their résumé-screening criteria, employers are not using the criterion which far outweighs any other: General programming talent. The best way to ensure success of a software project —finishing under a short deadline, minimizing the number of program bugs, maximizing innovation and so on— is to hire talented programmers, not people with specific software skills.
So again employers are shooting themselves in the foot under their current policies.
The fact is that although employers shun the mid-career programmers in favor of new college graduates and foreign nationals in an attempt to reduce personnel costs, if the employers were to utilize a more broad-based hiring policy, then their overall costs (not to mention headaches) would actually decrease.
Dr. Norman Matloff is a professor of computer science at University of California at Davis, and was formerly a statistics professor at that institution. He is also a former software developer in Silicon Valley. For his bio, see < http://heather.cs.ucdavis.edu/pub/MyBio.html >.
1. Sources for all statistics cited in this Executive Summary
are available in the body of this paper.
Please send mail to
email@example.com with questions or comments about this web site. Minnesotans For Sustainability
(MFS) is not
affiliated with any government body, private, or corporate entity.
Copyright © 2002 Minnesotans For Sustainability