2
Data and Taxonomy: Computing Professionals Are Hard to Count

A number of methodological and taxonomic problems exist in attempting to produce data that accurately describe the supply, demand, and characteristics of computing professionals. This chapter summarizes the views and impressions of workshop participants about the strengths and limitations of available data for assessing the human resource base of the computer science and technology enterprise, focusing on inadequacies in the taxonomies used. It surveys the major available data sets, describes what the data show, and concludes with a discussion of the issues involved in improving on what is currently available.

INTRODUCTION

Accurate and timely information on the availability and utilization of computing professionals and their skills is vital to a wide variety of decisionmakers. Students choosing careers want to know about career opportunities and the nature of the work done by computing professionals. Policymakers need information to formulate proper programs of research, student support, and economic development. Academic employers want to know how easy it is to find people capable of teaching and doing fundamental research. Industrial and government employers are interested in knowing how easily they might recruit employees who can design, develop, or support computer systems.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 10
Computing Professionals: Changing Needs for the 1990s 2 Data and Taxonomy: Computing Professionals Are Hard to Count A number of methodological and taxonomic problems exist in attempting to produce data that accurately describe the supply, demand, and characteristics of computing professionals. This chapter summarizes the views and impressions of workshop participants about the strengths and limitations of available data for assessing the human resource base of the computer science and technology enterprise, focusing on inadequacies in the taxonomies used. It surveys the major available data sets, describes what the data show, and concludes with a discussion of the issues involved in improving on what is currently available. INTRODUCTION Accurate and timely information on the availability and utilization of computing professionals and their skills is vital to a wide variety of decisionmakers. Students choosing careers want to know about career opportunities and the nature of the work done by computing professionals. Policymakers need information to formulate proper programs of research, student support, and economic development. Academic employers want to know how easy it is to find people capable of teaching and doing fundamental research. Industrial and government employers are interested in knowing how easily they might recruit employees who can design, develop, or support computer systems.

OCR for page 10
Computing Professionals: Changing Needs for the 1990s Requirements for such information may be met in a variety of ways, for example, through the use of statistics generated in the many data collection efforts undertaken by governmental and nongovernmental organizations. While necessary, however, statistical information is not always sufficient. Many decisions involving human resources must be based on qualitative factors (e.g., personality, work experience, functional competencies) that are difficult to summarize in a data set. Moreover, no single set of information is able to satisfy the needs of all potential users. Recruiters for an individual company may need more detailed information than do those who formulate national policy on R&D or student support. Given the variety of data collection efforts and user needs, it is not surprising to find that the computing professions are one group for which, depending on one's perspective, there are no data, poor data, and/or conflicting data. Even in a very narrow category such as Ph.D.s in computer science, counts of the number of people differ. WHY ARE THE DATA UNSATISFYING? Current data on computing professionals reflect two classes of problems: (1) common difficulties encountered in gathering data and (2) uncommon disagreement about how to label and categorize computing professionals. Difficulties of the first type are largely methodological and include inconsistencies arising from differences in responding units (e.g., university registrars, personnel managers, individuals) as well as in how individuals might report when faced with alternative lists of fields, occupations, responsibilities, or skills.1 Third-party counts typically do not agree with self-reporting counts, and discrepancies are compounded by the fact that data-gathering activities differ in focus and purpose, geographic scope, how nonrespondents are handled, sampling, and aggregation of data. Data-gathering errors, although inevitable, are a special concern in analyzing occupations employing small numbers of people. In total, computing professional occupations, plus related technical occupations, appear to employ about 1 million workers, or less than 1 percent of the total U.S. work force—but also 7 percent of what the Bureau of Labor Statistics calls ''professional specialty'' occupations, a remarkable level for an occupational group that is less than 30 years old. Moreover, these occupations are changing dramatically, and this dynamism creates problems for classification and reporting. Commented Barbara Wamsley, deputy director, Federal Programs, National Academy of Public Administration, "By the time you get the standard written for the job, the job has changed."

OCR for page 10
Computing Professionals: Changing Needs for the 1990s Need for Agreement on Labeling However, the biggest problem plaguing data collection efforts relates to labeling: i.e., what is a computer scientist, computer engineer, or other computing professional? This question has not been answered with consensus within the computing professional community, nor translated into common practice among data gatherers. The essential problem is to define what different kinds of computing professionals do. There is little agreement even within segments of the community as to the tasks, the skills and training required, or other defining parameters of the various computing occupations. Absent a general consensus about what computing professionals do, it is difficult to attain agreement within the community on taxonomic labels. For example, academic researchers appear most likely to identify themselves as "computer scientists," but people in industry who have computer science degrees and are associated with systems design and development may have such job titles as "software engineer" or "systems integrator" and may report on themselves to data collectors according to these job titles. Members of the community cannot agree even on using the term computer in labeling computing professional occupations. Workshop participants involved in systems implementation in industry or government, for example, prefer "information systems'' and "management'' as elements of the overall label. As a result, there are serious differences among respondents, data sets, and analyses concerning what people would call individuals doing the same work, and a corresponding, almost overwhelming risk of comparing apples to oranges. The newness of the computing field, the dramatic changes that it has been experiencing, and community members' lack of consensus on appropriate labeling thus come together to complicate the survey process. Explained Eileen Collins, senior sciences resources analyst at the National Science Foundation, When you are dealing with an evolutionary field, a number of problems turn up . . . . It is . . . hard to come up with an instrument that the people you are surveying will recognize as making the distinctions you want them to make. As a field develops and becomes more distinctly defined, it becomes easier. This is a field clearly undergoing continuing change. . . . [D]escriptions of the actual tasks in the job are important, so that respondents can, with a reasonably high degree of confidence, feel that they understand your question and that the answer they give is interpreted by them in the same way that you interpret it. The greatest agreement on labeling seems to apply at both ends of the skill spectrum (taken most broadly). At the high end, "com-

OCR for page 10
Computing Professionals: Changing Needs for the 1990s puter scientists" are typically associated with advanced education and the conduct of research. At the other end are workers who may be grouped with higher-level computing professionals in some counts but, because of the tasks involved or skills required, appear better classified as clerical (e.g., data entry) personnel—who are users of computer systems—or administrative support personnel (e.g., computer operators).2 Box 2.1 lists the computer-related occupations for which the Bureau of Labor Statistics (BLS) and the National Science Foundation (NSF) report data. Note that in contrast to such broad categories as "physical scientists" and "mathematicians," the listing includes titles representing a wide array of skills—computer systems analysts, computer programmers, computer operators and data entry personnel, hardware and software engineers, and various types of computer scientists. However, it is possible to identify the narrower set of occupations that approximately encompass the computing professionals focused on by workshop participants. The wide array of skills displayed in computer-related occupations is not unlike the broad range of skills characterizing health care practitioners—ranging from doctors and nurses to medical technicians. The similarity in breadth between these two taxonomies may reflect the dependence on a range of practitioners of these sectors of the U.S. work force. Box 2.2 illustrates the commercial perspective on computing professional occupations, drawing from organizations that are major users of information technology. However, a different set of job titles might be composed for organizations focused on developing computer products (a list that might include developers of personal computer (PC) utilities, of distributed applications, and of local area network (LAN) and server software, as well as user interface designers, and so on). Further, noted Joe Kubat, vice president, Floor Systems, New York Stock Exchange, one would expect some differences in the set of job titles found in academic and other research environments as opposed to those found in business and other production organizations. Community consensus on labeling is necessary not only to enable good data gathering and analysis at aggregate, national levels, but also to meet recruiting and training needs at the level of the firm or the institution of higher education. This point was made by Betty Vetter, executive director of the Commission on Professionals in Science and Technology: . . . [Y]ou can't tell young people you want them to come into the field unless you have something to call it. . . . [I]f you don't have a name for it, I don't know what you are talking about and neither do they. . . .

OCR for page 10
Computing Professionals: Changing Needs for the 1990s BOX 2.1 Computer-related Occupational Groups BUREAU OF LABOR STATISTICS Current Population Survey [Bureau of the Census] Computer systems analyst Computer programmer Other computer specialist Dictionary of Occupational Titles Systems analysis and programming Software engineer Computer programmer (alternative titles: applications programmer; programmer, business) Programmer-analyst (alternative title: applications analyst-programmer) Programmer, engineering and scientific Systems programmer Data communications and networks Supervisor, network control operators (alternative title: data communications technician supervisor) Network control operator Data communications analyst Computer systems user support Supervisor, user support analyst (alternative title: help desk supervisor) User support analyst (alternative titles: customer service representative; end user consultant; help desk representative; information center specialist; office automation analyst) Computer systems technical support Computer security coordinator (alternative titles: data security coordinator; information security coordinator) Data recovery planner (alternative title: disaster recovery coordinator) Technical support specialist (alternative titles: project development coordinator; technical operations specialist) Computer systems hardware analyst (alternative titles: computer systems engineer; methods analyst, data processing; information processing engineer) Quality assurance analyst Computer security specialist Computer-related occupations, not elsewhere classified Database administrator Database design analyst Micro computer support specialist

OCR for page 10
Computing Professionals: Changing Needs for the 1990s Occupational Employment Survey Systems analyst, electronic data processing Computer programmer Computer programmer aide Programmers, numerical, tool, and process control Other computer scientists NATIONAL SCIENCE FOUNDATION Classification of Computer Specialists Computer engineer—hardware Computer engineer—software Computer operator, data key entry Computer programmer: Business/financial Scientific Industrial machines/process control Graphics/art/animation Other programmers Computer systems analyst Computer scientist except systems analyst (e.g., theorist, researcher, designer, inventor) Other computer specialists [W]e need a system that has enough names with which everybody generally agrees—[a system] that can say that an educational institution . . . has [educated] a person with these general competencies and I could count on that as an employer. It would say to the young person, "These are the requirements you will have to take in order to be able to say you have a degree in this field," whatever you call it. In computing, there appear to be too many names and no agreement about what those names mean. Need for Better Taxonomies Many differences in data sets and analyses arise from differences in how subgroups of computing professionals are broken out as elements of a taxonomy used in gathering data. As noted below, NSF collects data from a sample of individuals who have a high probabil-

OCR for page 10
Computing Professionals: Changing Needs for the 1990s BOX 2.2 Computing Professional Specialties Derived from a Sample of North American Leading-edge Information Systems Divisions, 1991–1992 Architect/Designer Network systems integrator Business integrator Network systems manager Business planning consultant Numerically intensive systems specialist Business process designer Office systems analyst Capacity planning analyst Operations support analyst Customer support analyst PC application development analyst Data administrator PC support analyst Database analyst/administrator Project manager Data intensive systems specialist Quality-improvement analyst Data security analyst Service-level analyst Emerging-technology specialist Skill resource manager Graphically intensive systems specialist Software engineer Help desk assistant/technician Systems consultant Information engineer Systems developer LAN system administrator Systems integrator Network engineer Systems maintainer Network software systems specialist Systems manager   Systems programmer   SOURCE: Ian Rose, IBR Consulting Services Ltd., Vancouver, B.C., Canada. ity of meeting the taxonomic criteria associated with the definition of a scientist or engineer; NSF reports estimates for the subset of individuals who actually meet these criteria. BLS collects data from a national survey of U.S. households—the Current Population Survey administered by the Bureau of the Census—and from a national survey of establishments—the Occupational Employment Survey. The surveys have different respondents and rely on different taxonomies in collecting information. It is not unreasonable for taxonomies to differ because they may support different goals in collecting data. For example, explained Alan Fechter, executive director of the National Research Council's (NRC) Office of Scientific and Engineering Personnel, when the focus is on supply of talent, the taxonomy will typically refer to embodied

OCR for page 10
Computing Professionals: Changing Needs for the 1990s skills such as educational attainment and experience. But when the focus is on demand, the taxonomy will typically refer to occupational characteristics, such as job titles or functional responsibilities. Workshop participants expressed concern that, regardless of the objectives of a particular data-gathering activity, the known taxonomies of computing professionals appear to have serious shortcomings. For example, confusion and error may arise when different kinds of jobs are grouped within the same category of a given taxonomy. Other difficulties include the fact that a single individual may do work associated with many of the job categories, possibly leading to arbitrary or inconsistent counting, and the tendencies for the taxonomy to assume that work is organized around centralized computing systems or to exclude managers. Overall, workshop participants agreed that job titles alone are poor indicators of what computing professionals do in their jobs, what the skill requirements are, and how the jobs relate to functions performed. Some of the problems arising from inappropriate taxonomies are reflected in BLS's Occupational Outlook Handbook, a guide for career counseling for high school students and others. The Handbook covers about 250 occupations but addresses only two categories of computing professionals—programmers and systems analysts. This aggregation into two gross categories was universally decried by workshop participants because it fails to differentiate among a large number of computer-related professional activities valued in the community—such as computer science research, systems management, and systems design—whose practitioners are lumped together with typically lower-skilled individuals. Further, the job titles of "programmer" and "systems analyst" have fallen out of favor in industry, particularly in large firms, with ''programmer" increasingly viewed as a relatively low-skill position. BLS recognizes this taxonomic shortcoming and is prepared to add up to three more occupational categories to the Handbook. The problem it faces is choosing and defining what those categories should be, a process for which it needs input (and presumably consensus) from the computing professional community.3 The level of aggregation, how grossly or finely occupations or jobs are differentiated, is a fundamental issue. As Fechter observed, How much you disaggregate these taxonomies to get to subfields or suboccupations or subtitles of jobs depends on how many of them you are dealing with. It is one thing to deal with 10 jobs. The kind of aggregation or disaggregation you talk about with 10 jobs is very different from the kind of aggregation you talk about for 500 jobs or

OCR for page 10
Computing Professionals: Changing Needs for the 1990s 5,000 jobs or 5 million jobs. . . . It makes no sense to have a very detailed taxonomy to describe three people. Jane Siegel, of the Empirical Methods Group at Carnegie Mellon University's Software Engineering Institute, underscored the value of even rough estimates of the numbers and characteristics of people engaged primarily in work relating to computer hardware development, computer software development, and systems integration, plus rough estimates of degree production. I would like to suggest that for that more senior audience and for the national-level work, keeping it simple is . . . essential. I would be thrilled if in the next major national surveys . . . they did nothing more than simply have a logical, simple structure that broke out people doing computer-related work . . . into those people who work primarily on the hardware side and those people who work primarily on the software side and say something about the folks who integrate it all and make it work together. If I could get even very rough estimates of the degree field and some simple demographics about who those people are and a little bit about their turnover rate and what they do in life, I would have a whole body of knowledge that I think would help a large set of our users. Toward this end workshop participants agreed on the need for a generally accepted, high-level set of computing professional categories. Ian Rose, president of IBR Consulting Services Ltd., commented on the challenge of determining how many categories would be ideal, at least for industry: "[M]ost [major] organizations are trying to put together some sort of skills assessment process that works for them. . . . [W]hat you have to do as an individual organization is to determine the number [of job categories] between 15 and infinity, and it will be different for every organization . . . ." Alan Fechter, echoing Jane Siegel, cautioned that although a moderate level of detail may be valuable for corporate planning, a greater level of aggregation may be appropriate for purposes of national planning and estimation. The steering committee concluded that this issue can be settled only after a thorough analysis has been made of the skills "portfolios" possessed by particular individuals at particular moments in time and those required to perform certain complexes of activities (jobs). Lack of consistency in available data, as well as the underlying heterogeneity of the group, makes it more difficult to analyze supply and demand for computing professions than for other occupations. Observed Betty Vetter, ". . . I have made comparisons in other fields,

OCR for page 10
Computing Professionals: Changing Needs for the 1990s and the differences are not significant. . . . [A]lmost everybody seems to agree on what is a Ph.D. in chemistry, what is a Ph.D. in chem[ical] engineering. . . ." Although other fields evolve, most have been more stationary than computer-related professional fields, and job descriptions and degree titles for them have been more consistent over time. Similarly, the greater acceptance of core competency in other fields allows for less variation in reported counts by survey respondents. Other professions also tend to import fewer people trained in other fields. Noted Jim Voytuk of the NRC, "In . . . for example, physics, one might not . . . go out and hire a biologist. . . ." Another complicating factor is change in industrial practice. Noted Rose, who specializes in skills assessment and planning, "Many [major corporations] are moving away from job titles and into roles within the organization. So, if you can imagine a skills inventory, it wouldn't call for a business process designer or a systems integrator. It would say, 'business process design function,' and, 'these are the skills you require to be in that function'." This is not to suggest that there is consistency in industry; Rose observed that one organization may have a position titled "systems integrator" while another may not, although both may require systems integration work. Further complicating the challenge of selecting realistic labels or categories is the need to allow for assessment of changes over time. Commented Robert Kraut, district manager at Bell Communications Research (Bellcore), "[If] you want to know whether the rise of microcomputers has led to a decrease in the need for programmers, you want to have the same definition of 'programmer'." That is, even though the nature and assignment of work may change, a constant metric may be needed in order to measure that change. Yet more confusion rises from the fact that some computing professional specialties necessarily straddle other fields. Hardware engineering, for example, draws on such disciplines as electrical engineering, mechanical engineering, and solid-state physics, as well as computer science and computer engineering. Software developers, by contrast, face a growing need to understand the area in which their systems will be applied, and they may be substantially educated or trained in that area. This heterogeneous character of computing professional occupations can play havoc with conventional taxonomies. If anything, drawing on people from mixed backgrounds may increase, as the growing complexity of computer systems and their applications increases the value of multidisciplinary collaboration (see discussion of demand in Chapter 3).

OCR for page 10
Computing Professionals: Changing Needs for the 1990s THE CURRENT SITUATION The taxonomic shortcomings described above warn us to treat with caution the available data on computing professionals. Despite such shortcomings, these data can be used to provide rough estimates of the existing situation. Generally, statistical analyses can be done using data on degrees awarded and/or counts of employed people, which are used to derive direct estimates of levels and trends in supply and demand. Such analyses also use salary levels and rates of employment and unemployment to help assess whether supply and demand are in balance. Other data helpful for assessing future supply and demand include population trends, including shifts in the numbers of entry-level and retirement-age people; trends in the number of high school graduates, undergraduate and graduate enrollments, and undergraduate degrees granted; and attitudes of college freshmen and their intentions to enter different fields. In addition, technology trends are an important factor to consider in examining occupations as technology-dependent as the computing professions. Sources of data in addition to the federal government's BLS, the NSF, and the National Center for Education Statistics (NCES) include such private and semiprivate organizations as the Engineering Manpower Commission, the Computing Research Association, the College Placement Council, and the American Council of Education/Higher Education Research Institute. A variety of data sets are discussed and compared by Betty Vetter in Appendix A. Federal data gathering is subject to various limitations. Any changes in field or occupational taxonomy proposed by BLS or NSF for its survey instruments must be reviewed and cleared by the Office of Management and Budget (OMB),4 the effect of which is to minimize changes. Eileen Collins explained, "What agencies can do is . . . pick a very aggregated, coarse-grained category to survey . . . . They can take a category that is disaggregated and aggregate it further, provided it is possible to add back up to basic categories [approved by OMB]. . . ." Collins also explained that new occupations tend to be added to a taxonomy only after some evidence of their existence has been accumulated: "When enough . . . anomalous people or jobs turn up on a survey . . . you expand . . . the number of categories you have . . . ." Federal data gathering is also limited with respect to frequency. While some surveys (e.g., the Current Population Survey by the Bureau of the Census) are conducted as often as monthly,5 most of NSF's surveys are biennial, and BLS's data gathering from the Occu-

OCR for page 10
Computing Professionals: Changing Needs for the 1990s pational Employment Survey is further limited by the fact that each group of industries represented is surveyed once every 3 years. Moreover, in each of these cases it may take an additional year and a half for the data to be reported. Data on Employment In an attempt to describe the employed human resource base for computer science and technology, data can be used to answer two types of question: Who is working actively in computer science and technology activities? and How are people being used who are formally trained in computer science or computer technology? Although the two groups of individuals overlap substantially, there is also a substantial area of mutual exclusivity. To a greater extent than in other fields, people actively working in computer science and technology are not necessarily formally trained in these fields; as in other fields, people formally trained in computer science and technology are not necessarily working in this area. The data in Table 2.1 suggest that there are sizable numbers in these mutually exclusive groups.6 Because available data are derived from surveys and may have been subject to delays in processing, the most recent data available are often somewhat dated. The BLS is the best source of information on who is working actively in computer science and technology. Using data collected monthly from the Current Population Survey, the BLS estimates employment in computer-related occupations. TABLE 2.1 Employed Scientists and Engineers (in Percentages) by Field of Degree and Occupation, 1988   Field of Degree     Occupation Computer Science Other Total Computer specialist 5.8 7.3 13.1 Other 1.7 85.2 86.9 Total 7.5 92.5 100.0   SOURCE: National Science Foundation, Division of Science Resources Studies.

OCR for page 10
Computing Professionals: Changing Needs for the 1990s TABLE 2.2 Estimates of the Number of Employed Computing Professionals and Technicians, 1978 and 1988   Number Employed (in Thousands) Source of Data and Occupational Category 1978 1988 National Science Foundation     Computer specialists, total 177 708 Computer specialists in science or engineering jobs 171 552 Bureau of Labor Statisticsa     Systems analysts and computer scientists 182 476 Systems analysts and computer scientists and programmers 337 919 a Estimates based on data from the Current Population Survey conducted by the Bureau of the Census. SOURCE: National Science Foundation, Division of Science Resources Studies, and U.S. Department of Labor, Bureau of Labor Statistics, unpublished tabulations. Table 2.2 summarizes BLS and NSF estimates of the number of employed computing professionals and technicians for the years 1978 and 1988, the latest year for which NSF data are available. Approximately I million people were employed in computer-related occupations that BLS considers professional or technical. Almost 500,000 of these were professionals.7 Employment in professional occupations overall more than doubled from 1978 to 1988, compared with a growth rate for the entire work force of 20 percent. The National Science Foundation also supplies information on this human resource pool, basing its estimates on surveys of individuals who meet NSF's definition of a scientist or an engineer. This definition uses as criteria (1) the field and level of degree, (2) occupation, and (3) professional self-identification. Because programmers are considered technicians rather than professionals, NSF generally excludes computer programmers from its estimates, even though computer programming is included as part of the occupational taxonomy of NSF's survey instrument. The NSF's estimates for non-programmer computing professionals are similar in order of magnitude to those generated by BLS—about 550,000 according to NSF and roughly 475,000 according to

OCR for page 10
Computing Professionals: Changing Needs for the 1990s BLS. The trends also match reasonably closely. NSF estimated a slightly faster rate of increase over the period from 1978 to 1988. The NSF publishes estimates for two sets of employment: total employment and employment in science and engineering activities. Its estimates of the former measure employment of all individuals whom NSF defines as scientists and engineers, regardless of the nature of their work. Its estimates of the latter are restricted to those individuals who meet NSF's definition and who also report that they are working in a science or engineering position. Although one can conclude from these estimates that approximately 500,000 to 600,000 workers were employed in computer-related professional jobs in 1988 and that employment in these jobs has been growing dramatically, one cannot take comfort from the apparent robustness of these estimates. The significant methodological differences underlying them and possible shortcomings in the taxonomies used provide ample reason to urge caution in their use. Employment in Academic Institutions Information on computer scientists and engineers employed in academic institutions is available from BLS, NSF, the Computing Research Association (CRA, which publishes the Taulbee survey), and the Conference Board of the Mathematical Sciences (CBMS). Data generated by CRA are based on information provided by a subset of all institutions: those with graduate programs in computer science leading to the doctorate.8 Similarly, estimates generated by CBMS are based on information provided by departments of mathematics and departments of computer science. Another source of information is the NRC, which collects information from individuals with doctorates through the Survey of Doctorate Recipients. These data sets are compared in Appendix A. The BLS data include all individuals who report such employment, regardless of their degree level or institutional affiliation. Thus, unlike other sources, BLS includes in its estimates non-Ph.D.s and individuals employed in community colleges. The NRC data are restricted to individuals who hold the doctorate. For 1989–1990, the BLS reported 22,000 computer scientists and engineers employed in academic institutions, while the NRC counted about 6,600 academically employed computer science Ph.D.s for the same year.9 This range in estimates provides a vivid example of the data problems that impede analysis of computing professional labor markets. Factors that may result in the higher BLS number include

OCR for page 10
Computing Professionals: Changing Needs for the 1990s counts from a broader range of institutions (such as community colleges), from a broader range of disciplines (i.e., individuals with degrees in other fields who teach computer-related courses), and from both Ph.D. and non-Ph.D. holders. Data on Degree Production Although degree awards should be among the most reliable and consistent types of data, counts of people with computer science and more or less similar degrees vary significantly (the spectrum of degree programs and their differences are discussed in Chapter 4). The higher the degree level—and the smaller the number of degrees awarded—the greater the agreement on the numbers of individuals with specialized education in computing disciplines. Vetter (Appendix A) presents three sources of data on bachelor's degrees awarded in computer science—CRA, NCES, and CBMS. The CBMS and CRA data are not examined here since they are derived from subsets of institutions. Estimates generated by NCES are based on information provided by all accredited institutions of higher education. They indicate that roughly 30,000 bachelor's degrees in computer science were granted in 1989 (Table 2.3). This number has TABLE 2.3 B.S. Computer Science and Computer Engineering Degrees Reported, 1986 to 1990 Year National Center for Education Statistics Engineering Manpower Commissiona Computer Science 1985     1986 42,195   1987 39,927   1988 34,896   1989 30,963   1990     Computer Engineering 1986 2,192 4,999 1987 2,021 5,012 1988 2,115 4,275 1989 2,198 4,398 1990   4,355 a EMC surveys only institutions with engineering programs. SOURCE: Table A.1, Appendix A, ''Comparison of Data Sources and Data," Betty M. Vetter.

OCR for page 10
Computing Professionals: Changing Needs for the 1990s declined dramatically from its peak of 42,000 in 1986. The dramatic downward trend parallels comparable trends in engineering and mathematics. (See "Encouraging Student Interest" in Chapter 4.) The Engineering Manpower Commission (EMC), which derives its data from institutions with engineering programs, reports consistently higher numbers of B.S. computer engineering degrees earned than does the NCES (Table 2.3), largely because the EMC numbers include some individuals with degrees in computer science.10 Taking the NCES estimate as the most comprehensive, B.S. degree production in computer engineering was about 2,200 in 1989, with no trend apparent for the period from 1986 to 1989. Estimates of the number of doctorates awarded in computer science are generated from data compiled by CRA, NCES, and the NRC. NRC estimates are derived from information provided by the doctoral candidate upon completion of his or her requirements for the degree, whereas CRA and NCES estimates are based on information provided by the institutions. Among sources reporting the number of Ph.D.s awarded, the absolute estimates agree relatively well (Table 2.4). In 1989, Ph.D. production in computer science ranged from 500 to 600, having shown a strong upward trend since 1986. The variation in the annual estimates of the three data sources was less than 100 (in absolute terms), except for 1988, when the CRA estimates showed a larger increase in Ph.D. production than did the NCES and the NRC estimates. TABLE 2.4 Ph.D. Computer Science and Computer Engineering Degrees Reported, 1986 to 1990   Computer Science Computer Engineering Year NCES NRC CRA NCES NRC EMC CRA 1986 344 399 412 56 77 176   1987 374 450 466 57 61 205 93 1988 428 514 577 77 92 262 167 1989 538 612 625 74 117 277 182 1990   704 734   132 339 173 NOTE: Data from annual and other surveys of the Computing Research Association (CRA), National Center for Education Statistics (NCES), National Research Council (NRC), and Engineering Manpower Commission (EMC). SOURCE: Table A.2, Appendix A, "Comparison of Data Sources and Data," Betty M. Vetter.

OCR for page 10
Computing Professionals: Changing Needs for the 1990s As is the case for its numbers on B.S. degree production, the EMC's numbers for doctorate production in computer engineering are consistently larger than the estimates generated from the other sources and add considerably to the dispersion (Table 2.4). Again, the EMC's numbers are larger because they include some computer science degrees awarded in the engineering schools covered by the EMC survey. Estimates of doctorates produced in computer engineering in 1989 ranged from 74 to 277 (or to 182 if the EMC estimates are excluded). ISSUES AND CONCLUSIONS Available data are inadequate to guide employers, students, educators, and policymakers. Both the kinds of data and delays in their publication pose problems. For example, data that are updated every 4 to 5 years will be inconsistent with technology and industry dynamics that change as quickly as those for computing. Because acquiring a comprehensive set of skills data could be prohibitively costly, modest insights might be obtained by collecting and analyzing skills assessments that, according to workshop participants, are currently being undertaken in industry and government. A robust high-level taxonomy of computing professionals is needed . Currently, data are collected using taxonomies or sets of job titles (see Box 2.1) that are too detailed and too prone to obsolescence as skills needed in real jobs change. For example, "hardware professionals," "software professionals," and "deployment professionals'' (responsible for supporting and facilitating the effective use of systems; see Chapter 3) could be developed as gross categories of computing professionals that could have more lasting meaning than do the current large, but inappropriate, sets of occupational titles. However, a high-level taxonomy alone will not provide information on shifting skill requirements; meeting this need requires developing better, more detailed occupational data that take into account shifting technology and industry dynamics. A major need is community agreement on how to label different types of computing professionals and whom to count in each category. But workshop participants differed on how to implement these changes: academic participants placed a high premium on degree and skill attainment, and industrial and government participants focused more on the nature of the work to be done and the skills applied. This disparity in perspective is not surprising; it reflects the greater challenge evident in industry and government to adapt job descriptions to evolving technology and applications. Thus a num-

OCR for page 10
Computing Professionals: Changing Needs for the 1990s ber of nonacademic workshop participants referred to skills assessments they have conducted for their employees and clients and emphasized the evolution of job titles and progressions. It is necessary to identify and evaluate changes in portfolios of skills associated with jobs, occupations, and individuals. Like portfolios containing stocks and bonds, portfolios of skills are subject to change over time in their individual components and in the volume of each component. A job taxonomy is needed that partitions jobs into sets that are equivalent in terms of the skills required. A similar taxonomy for individuals is needed to partition workers into groups with sets of embodied skills that are equivalent. Finally, one must also be able to convert functions required on the job to skills possessed by workers before one can meaningfully assess the strength of the fit between workers and jobs. To properly gauge shifting skill requirements, it is necessary to build on a greater understanding of shifts in technology and industry dynamics. The proposed high-level taxonomy should be related to portfolios of skills, providing a vehicle for tracking shifts in skill requirements that is independent of changing preferences in job titles. For example, the high-level taxonomy category "software professionals" might include skills that range from simple programming tasks to sophisticated database design or other software development skills. Over time, some skills will diminish in importance as others become more important (see Chapter 3); it is important to track both the details of change and the gross numbers of computing professionals employed in hardware, software, and deployment. NOTES 1.   For example, some workshop participants indicated that computing professionals may show more variability in self-identification than do people in other fields. 2.   Another class of workers involved in computer-related activities includes professional users, such as researchers and other professionals who rely heavily on computers to perform their work. For example, a growing number of computational scientists have expertise in both computing and other fields of science. Workshop participants distinguished these professionals from computing professionals by virtue of the fact that they use computers and computer programs as tools—means to an end other than advancing computer science and engineering. 3.   A representative of BLS present at the workshop requested such input from participants. As a second step, OSEP and CSTB convened a small meeting in March 1992, attended by representatives of BLS and NSF and individuals who had participated in the October 1991 workshop to explore further the kinds of assistance and input sought by federal statistical agencies. A second small meeting was convened in November 1992 by CRA and attended by representatives of BLS, the Association of Computing Machinery, the Institute of Electricial and Electronic Engineers, CSTB, OSEP, and CRA.

OCR for page 10
Computing Professionals: Changing Needs for the 1990s 4.   OMB is charged with administering the Paperwork Reduction Act, an objective of which is to constrain the volume of information that can be collected via government surveys. 5.   Data on the labor force are collected monthly for BLS by the Bureau of the Census through its Current Population Survey. The detailed occupational data derived from this survey are generally reported as annual averages. 6.   Because the first computer science degrees were awarded in the late 1960s, it is not surprising to find computer science faculty with degrees from other fields (e.g., mathematics), although their proportion is clearly diminishing. Moreover, according to both data and anecdotal accounts, significant proportions of people engaged in systems development and support do not have degrees in computing disciplines, a condition that is likely to continue (see discussion of employer perspectives in Chapter 3). Robert Kraut noted, "At Bellcore . . . there are 2,500 people who are involved in computer applications development of one sort or another. Half of them don't have a degree in anything related to computer science or computer engineering." 7.   Professionals included computer systems analysts and scientists and computer science teachers at the college and university level. Programmers are covered as technical workers (e.g., technicians). Although BLS data differentiate "professionals" from "managers," some managers will be counted as professionals, and certainly the popular conception of computing professionals includes some (technical) managers. 8.   CRA's Taulbee survey also includes Canadian schools in its sample. A recent comparison with NRC data indicates that only four U.S. universities awarding Ph.D.s were missing from the CRA data: Clarkson, Memphis State, Nova, and the University of Alabama, Huntsville. A total of eight doctorates were awarded by these institutions—a number so small that one can conclude that differences between CRA and NRC data should be minimal. 9.   The reader should be cautioned, however, that academically employed Ph.D.s include more than just tenured or tenure-track faculty. Also included are postdocs and nontenure-track employees supported by soft-money contracts. For example, CRA reports that 8 percent of its estimated academically employed computer scientists and engineers are in such nontenure-track appointments. See Appendix A. 10.   In 1991, for example, EMC estimates included 2,177 computer science degree recipients. These individuals were awarded degrees in the engineering schools that report to the EMC.