CRA Logo

About CRA



Government Affairs






What's New



The Supply of Information Technology Workers in the United States

Chapter 4: Worker Shortage

How Does One Determine Whether There Is a Labor Shortage?

Part of the controversy surrounding information technology (IT) workers stems from the fact that there is no definition of the term 'labor shortage' that is universally accepted. What one person might consider a labor shortage another might regard as only a "tightness" in the labor market, and a third person might dismiss the question entirely because of the way in which the supply and demand categories were defined. This report uses the Department of Labor's definition of shortage: a "market disequilibrium between supply and demand." A shortage requires two things: an occupational market in disequilibrium and a slow market response. Alternative definitions of labor shortage are provided in sources cited in the footnotes.32

When a worker shortage occurs, employees and workers take various actions, some of which can be tracked statistically as indicators of the presence and severity of a shortage—even when one cannot measure supply and demand directly. These indicators include: vacancy rates, amount of employment change (so-called 'churning') from firm to firm within the labor market, unemployment rates, changes in wage rates, predicted employment growth, and demand for foreign workers as indicated by labor certifications.33 In general, the presence of multiple indicators provides better assurance that a shortage exists. Perhaps the most commonly used indicator of a labor shortage is vacancy rates. Unfortunately, this indicator can be misleading. Vacancies occur in labor markets that have supply-demand equilibrium, and even in markets where there is a surplus of workers. Vacancies occur because of normal turnover and lags in filling open positions; indeed, they are a natural part of any business operation. One large company in the IT sector reported to the study group that it regards vacancy rates of five percent as normal—a management decision driven by the flexibility needed to fill positions quickly when the need arises in this rapidly changing field.

There are many kinds of shortages. They may be specific to one geographic region, or to a particular set of occupations or subspecialties within a labor market. They may be episodic (e.g., created by a one-time phenomenon of limited duration, such as the Y2K problem or the Euro conversion) or more enduring.

There may be non-statistical indicators of a labor shortage as well. If a large number of people say there is a labor shortage problem and act as though there is, then there may well be a problem. This is especially true if the action comes from companies ranging across multiple geographic regions and industry sectors, and if these companies begin to devote significant resources to correcting the problem by increasing recruiting, initiating studies, supplementing training, or even lobbying. A further indication of a shortage is when labor suppliers and professional societies representing the workers, not only the trade associations representing the companies, become concerned.

Is There a Shortage of IT Workers?

For some readers, this is the most important question addressed in this study. It is difficult to give a straightforward answer because of the imprecise definitions of supply and demand, the complexity of the IT workforce (many different kinds of jobs involving widely varying skill and knowledge sets, and distributed across many different sectors of the economy), and the lack of good data. There might be a shortage in one IT occupation (such as LAN administrators), but not in another (such as data entry clerks); or in one geographical region (such as Silicon Valley) and not in another (such as a purely agricultural area). The situation also changes rapidly over time. What is true at the time this report is written may not be true six months later.

In trying to answer this question about a shortage, it would probably be more insightful to do some labor market segmentation, rather than viewing the IT labor pool as one undifferentiated mass. There are barriers to geographical relocation of workers, which might make geographical segmentation useful. These include the cost of relocation, the cost of living in some of the areas where IT jobs are abundant (e.g., Silicon Valley), the social upheaval to the worker and the worker's family, and the difficulty of locating and securing a job outside the region where the worker is currently living. Distance may be a particular barrier in filling lower-skill, lower-pay jobs. Thus geographical segmentation of the IT workforce may be important to the understanding of shortages.

Because the skills and knowledge required vary so much from one IT job to the next, occupational segmentation may be even more useful than geographical segmentation. It would be a good test of the analytical value of this report's categorization scheme to segment the labor pool into conceptualizers, developers, modifiers/extenders, and supporters/tenders to segment the workforce when looking for labor shortages (see table 2-2). It is clear, however, that such a segmentation would not provide a complete breakdown. For example, computer security and networking are hot areas today, and there may well be shortages, or at least very tight labor markets, associated with these specialty areas. However, they cut across our four categories. The demand is uneven across even a single occupation, such as programmer. There is no known demand for APL programmers today; COBOL programmers are in short-term demand for fixing Y2K problems; and Java programmers appear to be in demand for the foreseeable future because of all the network applications being written in Java. Unfortunately, the data are inadequate to support a segmented analysis by either geography or occupational categories; thus this discussion of the shortage issue will consider the entire IT workforce.

At the outset of this study, it was clear that some trade associations believed there was a shortage of IT workers, while some labor unions disputed the claim. The H-1B debates showed that the worldviews and interests of these two kinds of organizations made it difficult to find common ground on which to evaluate the question of a worker shortage. The study group was somewhat surprised and dismayed to find a similar gulf between IT professionals and social scientists interested in this issue. The IT professionals have what they believe to be overwhelming local evidence of a shortage—personal experience of multiple unfilled jobs, multiple job offers for their graduating students, high salaries for graduates in hot technology areas, and more. The social scientists, trained to discount this kind of anecdotal evidence, demand strong statistical proof. As part of their scientific training, they tend to be skeptical about virtually all of the existing statistical data, which is in one way or another imperfect; and they tend to dwell on the methodological shortcomings of the data rather than using them as predictors—albeit imperfect ones—of a shortage or lack thereof.

Where (most) IT professionals see a shortage, (most) social scientists see a tight labor market. Is this just a semantic difference? This is unlikely. The distinction reflects an attitude about the nature of the problem and the need for a policy response. If there is simply a tightness in the labor market, there is a belief that the costs are bearable to the market participants, the market will eventually self-correct (or market participants will learn to live with the situation), and no government intervention is necessary. If there is a shortage, however, the costs to the market participants may be unbearably high, the ability of the market to self-correct is questioned, and government action may be warranted.

Unfortunately, no one kind of evidence currently available provides a clear and unambiguous answer as to whether there is a shortage—or even a tightness—in the IT labor market. Thus the strategy in this section is to look at six different kinds of evidence, each with its own strengths and weaknesses: 1) a direct counting of supply and demand; 2) secondary statistical indicators based on presumably solid federal data; 3) data from limited-scope studies that seem to be methodologically sound; 4) data from more general studies whose methodologies have been questioned; 5) anecdotal evidence of employer actions; and 6) other qualitative evidence. The limitations of each kind of evidence are discussed, as well as what it indicates about the existence of a shortage/tightness.

1. Direct counting. In order to demonstrate that there is a shortage of IT workers—that is, an excess of demand over supply—one would need a clear working definition of an information technology worker, as well as good statistical data about supply and demand. The problems of finding a good working definition of an IT worker have already been discussed. The supply system is complicated, as chapters 5 and 6 show; and anything close to adequate statistical data about the current potential supply, much less future supply, are nonexistent. Data about demand are even more elusive—they are out of date, the sample sizes are too small, and they rely on vague and inconsistent definitions of what kinds of demand to count as related to IT workers. Thus it is impossible at this time to provide meaningful quantitative information about a national shortage of IT workers. In particular, there is no adequate basis for quantifying the size of the shortage (if there is one). Collecting better data will be a major challenge, and the study group is not optimistic that reliable quantitative measures are likely to be available anytime soon.

2. Statistical indicators based on federal data. Even if the supply-demand match cannot be measured directly, it is useful to look at four secondary data indicators based on federal data. These data are likely to be as reliable as any data available on this issue because of the size and scope of federal data sets, the objectivity with which the government collects the data, and the methodological rigor with which they are analyzed. These secondary data indicators provide only inferential, not direct, evidence of a shortage; and for each indicator, there are some reasonable questions that might be raised about the validity of the inference. In general, these secondary indicators are also unable to distinguish between a shortage and a mere tightness in the labor market. While none of the indicators provide conclusive evidence, taken together they offer a preponderance of circumstantial evidence in support of tightness/shortage in the IT workforce.

There are several arguments against using these permanent labor certificates as a strong indicator of IT labor shortage/tightness. While the IT occupations rank high in the list of occupations for which permanent labor certificates are requested, the absolute numbers are small. Some even question how meaningful these statistics are. Anecdotal evidence from computer industry executives suggest that many of these certificates are applied for (and more than 90 percent are approved) not to fill an unfilled job, but at the request of a current employee who is working on a temporary visa (H-1B or one of the predecessor programs) who wants to work permanently in the United States. The Department of Labor's Inspector General has reported that the certification program is subject to extensive manipulation by employers and immigration lawyers.34

3. Limited-scope studies. Two studies are considered here. Both focus on software workers, which is the area of IT work with the greatest occupational growth. One study is national, the other regional. The national study was produced by the National Software Alliance, a consortium of industry, government, and academic leaders that was formed specifically to address their concerns that there is an IT worker shortage.40 The study gives many different kinds of evidence, some based on federal data, many based on private data. We did not evaluate the methodologies used in the collection and analysis of these private data sources, but presumably there is a range of quality and reliability across these sources. What is perhaps most remarkable about the National Software Alliance's study, however, is the large number of different statistical analyses, from various sources, that support one another in showing the existence of a shortage, or at least a tight labor market, for software workers. Here are a few examples from the report based on private data:

The second study, conducted by the Washington Software Alliance, looks at software workers in the State of Washington.46 As the Alliance itself notes, the situation in Washington is not entirely representative of the nation as a whole. The average software wage in the state is reported to be $66,752, the highest of any state in the nation. They also note that growth in the software industry in the state is outpacing the national average, with a growth from 1990 to 1996 of 17.8 percent compared with a national growth of 9.8 percent. The state has two major employers, Microsoft and Boeing, that may make its situation somewhat different from the nation as a whole. Given these caveats, it is useful to consider the report's statistics, which suggest a serious worker shortage.

The Alliance's survey includes all positions in the Washington State software industry, technical and non-technical. The industry includes 2,500 companies, with $20 billion in annual revenue and 47,000 employees. There are 7,300 current vacancies (15.5% of all positions), with 64,000 total desired hires in the next three to four years. A $12.8 billion three-year revenue gain is projected if these positions can be filled. Three-quarters of all jobs require a bachelor's degree or higher. The greatest need (largest number, hardest to hire, driving force for the industry) is for junior and senior developers—positions that require a BS or MS in computer science or computer engineering. There are eight jobs for every relevant in-state bachelor's graduate, and four jobs for every relevant in-state associate's (two-year) degree graduate. Unlike the views of a number of other industry representatives, the Washington State software employers are very satisfied with the technical, social, and English-language skills of those they hire. Their problem, they claim, is a numbers gap, not an educational quality gap.

In addition to considering how representative the Washington State situation is of the national scene, determining what the statistics actually mean also needs to be addressed. It is difficult to project accurately the number of new positions that will exist or be needed in the future, or the dollar value added by additional filled positions. While Microsoft and Boeing may hire many of their workers from in-state schools, they both recruit nationally, making it somewhat misleading to calculate state-graduate-to-open-job ratios. The problems of using vacancies as an indicator were described earlier.

4. Methodologically challenged national studies. The first ITAA report initiated the national debate over the IT worker shortage. This section will consider it and its follow-up report.47 Criticisms concerning the ITAA's methodology and results were discussed earlier, but are briefly reviewed here.

The main criticism was the direct count of supply and demand that ITAA tried to make. The first study used undergraduate graduation rates in computer science to measure supply, when this is only one of many supply sources of IT workers. The report also did not present available data that indicated the number of undergraduate majors in computer science was increasing after a long decline. On the demand side, a very broad definition of an IT worker was used, and the telephone survey used to collect the data had a very low response rate. Because of the low response rates, the results are statistically questionable. For example, companies that needed IT workers may have been more willing to respond to the survey than companies that did not. For these reasons, there are real questions about the reliability of the predictions in the two ITAA reports, respectively, that there are 190,000 and 346,000 unfilled IT positions in the United States. But even if the results do not have the weight of scientific evidence, they suggest that a number of companies are having difficulty filling positions.

Other kinds of evidence are also cited in the first ITAA study to support a worker shortage. Here are some samples:

5. Anecdotal evidence about employer actions. Table 4-4 provides a list taken from the Barnow, Trutko, and Lerman paper on what a company might do in reaction to a labor shortage. There is anecdotal evidence that every one of these strategies has been adopted in connection with IT workers.

6. Qualitative evidence. There is also an abundance of non-statistical evidence. This does not have the same persuasive force as methodologically sound quantitative data because it is hard to distinguish the special case from the general rule, but qualitative evidence can be both revealing and suggestive—especially if there is a lot of it, of various kinds, and from various sources. The preponderance of this anecdotal information supports either a shortage or a tightness in the IT labor market, although none of the evidence available can distinguish between the two. The third bullet below presents one of the few examples that speak against there being either a shortage or a tightness.50 Here are a few examples:


Where Are IT Shortages Occurring?

There is only anecdotal evidence to answer this question. The federal data categories are too coarse and the data too old to be of much help. The software area appears to be experiencing significant shortages at this time, especially for applications relating to networking, databases, and Internet-based applications. These shortages tend to be concentrated in niches and in jobs requiring higher skill sets. The problem is worsened by the fact that software workers are not always qualified to move from one software job to another. A programmer experienced in one programming language may not be effective in another. It is not simply a matter of learning the grammar and vocabulary of the language. More critical is learning the underlying methodology used with that programming language to attack programming projects, and this methodology varies considerably across programming languages. For example, programming in COBOL is very different from programming in Java; and it has been hard to retrain COBOL programmers (who received their experience on mainframe computers and are briefly in demand today because of the Y2K problem) to be Java programmers (who are likely to be in great demand well into the new millennium to program applications for use on the Internet).

Based on anecdotal evidence only, the following points represent the study group's consensus about occupations where either demand outstrips supply or many positions are filled with underskilled workers:


32 For a description and analysis of various definitions of 'labor shortage' given by economists over the past forty years, see Burt S. Barnow, John Trutko, and Robert Lerman, "Skill Mismatches and Worker Shortages: The Problem and Appropriate Responses," Draft Final Report, The Urban Institute, February 25, 1998, pp. 4-14. One definition in particular that we do not adopt is the so-called Social Demand Model of labor shortage. Under this definition, a shortage exists if there is not a sufficient number of workers of the type in question to meet a particular social goal. Thus some people might argue that there is a shortage of public school teachers because children seem to get a better education in smaller classes; or others might argue that there is a surplus of lawyers because our society would be better off if it were less litigious. The Social Demand Concept of a shortage depends on personal judgments about social welfare. The definition we adopt focuses on market equilibrium-the match between supply and demand as measured by economic indicators.

33 See Barnow, Trutko, Lerman, ibid., p. 55, for a discussion of this issue. Also see Timothy Bresnahan, "Information Technology, Workplace Organization and the Demand for Skilled Labor: Firm-level Evidence," Department of Economics, Stanford University (

34 See

35 However, some of the recent non-federal data show a flattening of salary increases. See the 12th annual salary survey by Computerworld and the related article by Leslie Goff, "Enough is Enough: The Joyride Is Over, As Corporate Managers Put the Brakes on Out-of-Control Salaries for IT Professionals," Computerworld, September 7, 1998 (

36 Lauren Brownstein, "Is There a Shortage of Information Technology Workers?" Symposium Proceedings, The Jerome Levy Economics Institute of Bard College, June 12, 1998, p. 5.

37 This example is not one discussed in the ITAA report. For an interesting overview of the occupation of systems administrator, including a description of the job and formal educational, commercial training, and independent study programs preparing a worker for this occupation, see David Kuncicky and Bruce Alan Wynn, "Educating and Training System Administrators: A Survey," published by the USENIX Association for SAGE, the System Administrators Guild, Berkeley, CA, 1998.

38 According to Stephen Johnson, the USENIX representative on the study group. USENIX pays close attention to the work of systems administrators.

39 For example, in 1984 BLS projected 520,000 computer science and systems analysts jobs in 1995, but there were actually 860,000. BLS predicted a 53-percent growth in electrical engineering jobs over this period, whereas there was actually a 9-percent decline. (See John H. Bishop's commentary in the Levy Symposium Proceedings, p. 8.)

40 National Software Alliance, "Software Workers for the New Millennium: Global Competitiveness Hangs in the Balance," Arlington, VA, 1998. This Alliance has a Department of Defense orientation. It has been argued that salaries and benefit packages offered to IT workers in the defense industry are relatively poor compared with those offered to IT workers in other sectors; and that this disparity creates a difficult recruiting and retention situation that is reflected in the National Software Alliance's view that an IT worker shortage exists, and the need to act on it for national security reasons (Michael Teitelbaum, Sloan Foundation, personal communication, March 1999). Whether one needs to question the reliability of the National Software Alliance's data, even if they are an interested party, is unclear.

41 "1997 Olsten Forum on Human Resource Issues and Trends: Staffing Strategies," William Olsten Center for Workforce Strategies, Melville, NJ, 1997.

42 "Ireland: The Software Capital of Europe," National Software Directorate, Forbairt, 1997; Christina Torode, "Closing the IT Skills Gap," Computer Reseller News, December 1, 1997, Issue 766.

43 See National Software Alliance, op. cit., page 2-21 for citations to the original studies.

44 Based on data from the National Association of Colleges and Employers, "Salary Survey," July 1997.

45 Based on 11th Annual Salary Survey, Computerworld, September 1, 1997.

46 See

47 "Help Wanted: The IT Workforce Gap at the Dawn of a New Century," Information Technology Association of America, Arlington, VA, 1997; "Help Wanted 1998: A Call for Collaborative Action for the New Millennium," Information Technology Association of America and Virginia Polytechnic Institute and State University, March 1998.

48 E. Cone, "Staffing: Short Supply," InformationWeek, November 1997 (

49 "Good Help Is Hard to Find," Computerworld, October 1997, and "Good Help Is Hard to Find," Computerworld, November 1997-both at; "Tech Corporate Culture Cozy, Creative," USA Today, November 1997 (; E. Cone, "Staffing: Short Supply," InformationWeek, November 1997 (

50 For some examples of articles questioning the existence of a shortage, see Margie Wylie, "The Skills Shortage That Isn't," CNET, February 4, 1998 (; also Dominique S. Black, "Taking Shots at the Labor Shortage," IT Careers, March 23, 1998.

51 For example, the Society for Information Management (SIM), a professional association of 2,700 managers of information technology and its applications to business, has drafted a position paper on the worker shortage. "SIM believes this labor shortage is the most severe in the 50-year history of computing, and will continue well into the next millennium...[in part because of] a fundamental shift in investment economics favoring increased use of IT." ("Addressing the Information Technology Workforce Shortage," Position Statement, Society for Information Management, October 1998, Chicago, IL.) A Web site maintained by the Department of Commerce profiles 170 IT worker-development programs throughout the country. See

52 A small survey on faculty hiring carried out by the ACM Special Interest Group on Computer Science Education, to which 64 institutions responded, suggests that faculty recruitment is more difficult in non-research institutions. When asked how difficult or easy faculty recruitment is, the survey received the following responses: at schools that grant the Ph.D. in computer science: easy 12.5%, moderate 31.3%, difficult 56.2%; at schools that grant the master's degree as their highest degree in computer science: easy 4.7%, moderate 28.6%, difficult 66.7%; for schools that grant the bachelor's degree in computer science as the highest degree; easy 3.7%, moderate 3.7%, difficult 92.6%. Preliminary results, July 27, 1998. For more information, contact Professor Paul Myers, Department of Computer Science, Trinity University, San Antonio, TX.


Copyright © 2004 Computing Research Association. All Rights Reserved. Questions? E-mail:

Document last modified on Wednesday, 04-Apr-2012 06:51:20 PDT.