The trend of increased degree production and course enrollment in computer science (CS) and related fields is prominent across a wide range of institutions, as discussed in Chapters 2 and 3. As with any academic discipline, students’ enrollment decisions are informed by a variety of factors, including planned career path, availability of jobs, salary expectations, intellectual interest, appeal of specific degree programs and courses within an institution, and other social or personal factors. In order to shed light on the specific drivers underlying the current growth and how these factors may affect future enrollments, this chapter examines the labor market in computing and discusses the changing role of computing in the economy, in higher education, and in society at large. While it is impossible to know what the future will bring, understanding the drivers of the recent increases in CS enrollment will inform expectations of future demand.
Undergraduate students’ choices about courses and major field of study may be directly and indirectly influenced by the current state of the labor market. The existence of dynamic adjustment in science and engineering labor markets was recognized as far back as the 1970s, when economist Richard Freeman wrote about the expansion and contraction of opportunities for physicists and engineers.1
1 Writing in 1975, Freeman described the labor market for physicists using this language of cyclical fluctuation: “Although the engineering profession has grown steadily over the past several decades, the labor market for engineers has exhibited marked cyclical fluctuation. A shortage of engineers in the late 1950s and early 1960s was followed by a surplus in the late 1960s and early 1970s” (see Freeman, 1975, 1976).
In essence, degree production in particular fields at colleges and universities can respond to opportunities in the labor market (especially wages) and how they are perceived by students, but the process is not instantaneous and involves lags of multiple years (depending on when students choose their area of specialization and how their expectations change).
Factors affecting the demand for computer scientists in the labor market include changes in technology (in particular, the price and availability of hardware and software), changes in macroeconomic conditions that affect the demand for computer services, and changes in trade or immigration policies (Bound et al., 2013).
Key historical changes in technology with an impact on the labor market and student interests include the rise of the computer hardware industry, concurrent with development of computer technologies for home and business use in the mid-1980s, and the rise of the Internet for commercial purposes starting in the mid-1990s. The latter launched what is often called the “dot-com boom” (Leiner et al., 1997), which increased the demand for computer scientists and included the growth of firms like Yahoo, Amazon, and eBay that helped sustain the boom in the IT sector. (These firms survived the bust that followed; many others did not.) More recent changes in both technology and market conditions include the trends toward data-driven discovery and decision making often referred to as “the Big Data Revolution,” fueled by advances in machine learning and computing power; the rise of the smartphone, mobile applications, and social networks; and the ubiquity of software-as-a-service, cloud computing, and the on-demand economy.
Forecasting challenges in science and engineering labor markets are well recognized: data deficiencies, incomplete models, and unanticipated events hinder prediction (Breneman and Freeman, 1974; NRC, 2000). Because of the variety of factors influencing student choice of major, it is not practical to determine causation of past fluctuations in CS and computer and information science and support services (CIS) enrollments, especially given limitations on available data that would help to distinguish between shifts in supply and demand; neither is it possible to predict future trends with certainty. In fact, the committee did not come to consensus on the extent to which economic factors were the decisive cause of past CS enrollment trends, though all agreed that economic trends are one major important factor. Accordingly, in the following, several important economic trends and principles that have played a role in past degree production and enrollment trends are discussed in the context of historical data. Current and projected conditions of the CS labor market are discussed in the context of the current wave of growth in CS undergraduate enrollments.
Historical Data on the Computing Labor Market
Employment in “computer occupations” (including those that do not require a bachelor’s degree) has grown significantly since the Bureau of Labor Statistics
(BLS) first introduced this category as one of its Standard Occupational Classifications (SOCs) in 1978, by roughly a factor of 20. According to the BLS Current Population Survey, more than 4.1 million individuals were employed in all computing jobs in 2015. While there are some discontinuities in this data set,2 it is evident that computing employment has increased significantly, especially during and since the 1990s.3 Due to discontinuities in data tracking before and after 2000-2002, one must consider a subset of relevant job titles whose definitions do not change (as do some under the unique group of “Computer Occupations”).
In a paper commissioned for the committee (Bound and Morales, 2016, p. 2),4 economists John Bound and Nicolas Morales tracked computing employment throughout this period by using the categories “Computer Systems Analysts,” “Computer Scientists,” and “Computer Software Developers.”5Figure 4.1 shows the overall employment profile in these jobs among those with a bachelor’s degree in any discipline. The cumulative number of CIS degrees produced over time is included (dashed line) as a reference. It is apparent that the total number of computer jobs has increased steadily over this time period, with the exception of dips between 2001 and 2002 (coincident with the dot-com bust) and between 2008 and 2010 (at the onset of the Great Recession). According to the plot the cumulative number of CS bachelor’s degrees ever produced at any given point in time was roughly half the number of current computer jobs. At the same time, it is unknown how many of the corresponding employers would prefer these positions to be filled by workers with a bachelor’s degree in computer science—or if a bachelor’s degree is required at all.
Bound and Morales (2016) also considered employment of young college graduates (age 23-29) in computer occupations. The share of young college grad-
2 The data over this time frame contain inherent limitations. First, the SOCs are periodically adjusted, meaning that some jobs may move between categories at certain points in time, and new job categories emerge with the evolution of the field; computer science in particular has seen significant growth from a single category into a multilevel hierarchy of occupations. In addition, data are unavailable for 2000-2001 on the public-facing site, and discontinuities in collection before and after these times make it difficult to compare trends directly across this boundary for the unique group of “computer occupations” as defined today (in the 2010 SOC revision).
3 Data from the 2015 Current Population Survey of the Bureau of Labor Statistics, using the SOC of 15-1100, corresponding to “computer occupations.” Additional SOCs outside the 15- series may also fall under computing occupations, but they were excluded to avoid overestimating the workforce by including less relevant occupational fields.
5 These categories correspond to 1990 SOCs of 171 (“Computer Scientists”), 229 (“Computer Software Developers”), 3921 (“Programmers, Business”), and 3972 (“Programmers, Scientific”); the data correspond to individuals with a bachelor’s degree in any field who are employed in one of these categories, which accounted for roughly two-thirds of all workers in “Computer Occupations” (15-1100 in the 2010 SOC revision) in 2015.
uates employed in computing fields is plotted in Figure 4.2. The share of new graduates with a bachelor’s degree in CS is also plotted for comparison. While these two data sets look at different but presumably overlapping populations, they should be related in two significant ways. First, some fraction of CIS graduates contribute to the population of young college graduates in computer occupations, with varying time delays between degree production and entering the workforce. Second, various aspects of the employment prospects in any given field, and how such conditions are perceived by undergraduates in the process of choosing a major, may influence their choice of major and affect the number of degrees produced at the corresponding time of graduation. Neither of these effects can be easily quantified; doing so would require primary research involving rigorous modeling and new mechanisms for further classifying these data, which is outside the scope of this study.
The plot illustrates that the fraction of young college graduates employed in computer occupations reached 7.4 percent in 2001 (coincident with the dot-com bust), after which they dropped slightly. CIS as a fraction of bachelor’s degrees produced in a given year peaked in 2004 at 4.2 percent, dropped to 2.4 percent in 2009, and steadily increased through 2015. While both employment and degree production seem to respond to the dot-com bust (the latter with a lag), no increase in employment in computing occupations among new college graduates is apparent in the years leading up to the recent CS enrollment surge.
In a white paper provided to the committee (see Appendix D), economist Jennifer Hunt examined similar trends, this time using a somewhat broader set of occupational fields for a time series analysis. Figure 4.3 illustrates time series of the share of workers in computer and related occupations. The blue line corresponds to the combination of all computer and mathematics occupations as identified in the Current Population Survey. While this is broader than the set of computer occupations, the trends are consistent with the data from Bound and Morales (2016) and similarly show that CIS and related fields have accounted
for an increasing share of occupational fields among all workers and among all college graduates, while those for other sciences and engineering have not.
Wages in a given occupational field are a dynamic component of the labor market, and their actual or perceived levels can also influence student choice of major. Median wages in computer occupations relative to all occupations held by college graduates are illustrated in Figure 4.4 for workers between 25-29 and 30-34 years of age. This figure shows that median wages for computer occupations in both age groups peaked around the dot-com boom, fell through 2006,
and have generally increased with some fluctuations between 2006 and 2012. At their peak CIS occupations offered on average a wage premium of 44 percent and 35 percent for the two age groups, respectively, compared to other fields. This boom and bust cycle does precede that of the fluctuation in CIS bachelor’s degree production by several years, which is consistent with the possibility that wages in computer occupations (or conditions connected to wages) may have contributed to student decisions to major in computer science and related fields. Post-2006 wage levels reset to higher than the pre-dot-com levels, similar to the trend observed for year-to-year CIS bachelor’s degree production.
Hunt conducted a similar analysis, looking at the absolute median wages for “computer/mathematics” and other occupations, using data from the American Community Survey (ACS). These results are illustrated in Figure 4.5. It is noteworthy that there are historical fluctuations in CS wage trends for young college graduates here as well, though not in recent years. In her white paper Hunt concluded that at least some of the increase in CS degree production since 2008 is in response to something other than wages for computer occupations.
As noted previously, a deeper exploration of the correlations and causalities of these trends is an area for research and beyond the scope of this study. Such
analysis would ideally involve a clear mapping between the training received in various degree programs with the requirements for a given occupational field, and assessment of whether computer science and related skills or degrees are a primary or preferred requirement for particular jobs. A full model of the complex computing labor market would decompose changes in wages and employment levels into shifts in supply and demand, and connect these to changes in degree production and course enrollments. To do so would require accounting for conditions underlying industry demand as well as factors underlying supply, such as student and parent perceptions of job and earnings opportunities, institutional conditions that affect student interest and the production of degrees and delivery of education leading to computing skills, and federal policies affecting the academic and industrial sectors. These underlying factors, and their relative role in the overall computing labor market, likely change over time.
The Labor Market for Computing in Recent Years
Labor market supply of computing workers includes computer science majors, students with bachelor’s degrees in other fields who have or can obtain sufficient computing skills and knowledge, and those with less than a bachelor’s degree. These workers can be domestic or foreign.
Table 4.1 illustrates the top field of degree for bachelor’s degree–holding computing workers from 2009 to 2014 according to the results of the ACS. It is noteworthy that only 37 percent of bachelor’s degree–holding computing workers age 23 to 39 identified in the ACS during this time reported a degree in a computing field (including the listed categories of computer science, computer engineering, computer and information, or information sciences). This is consistent with the evidence of significant levels of interest in computing among non-CS majors in recent years.
As noted earlier and in previous sections computer science as a field of study and an occupation has demonstrated monumental growth over the last four decades. A closer look reveals that the growth of computing jobs is not limited to employment in the computer services industries; computing occupations have emerged in a range of industries, as illustrated in Table 4.2. These statistics reflect CS degree holders working both in computing occupations and in other occu-
|Main Bachelor’s Degree||Age 23-29||Age 30-39||Foreign 23-39|
|Computer and information||6.7%||7.5%||5.9%|
|Business management and administration||4.3%||5.6%||1.3%|
|Number of observations||15,008||26,674||13,167|
|Computer and data processing services||21.3%|
|Professional and related services||17.7%|
|Finance, insurance, and real estate||11.1%|
|Colleges and universities||5.0%|
|Manufacturing—electrical and computing||4.8%|
pational fields, and illustrate that degrees in computer science and related fields provide skills and background relevant for a range of industries.
Current and Projected Demand for Computing Professionals
Although the committee notes that significant uncertainty must be attached to future employment predictions, the most authoritative forecasts of computing employment show sustained demand for computing workers over the next decade.
Every two years the BLS publishes projections of employment trends for the next decade. In the most recent projections, covering the decade from 2014 to 2024, the BLS predicted that the number of people employed in computing occupations will rise from 3,916,100 to 4,404,700. The addition of nearly half a million computing jobs to the economy corresponds to a growth of 12.5 percent over the decade, compared to an overall projected growth rate of 6.5 percent. This suggests that computing occupations are growing nearly twice as fast as the labor market as a whole.
The centrality of computing to the national science and engineering workforce is reflected in Figure 4.6, which compares the size and anticipated growth in the computing sector to those in other science, technology, engineering, and mathematics (STEM) fields. The left-hand chart in Figure 4.6 shows the current distribution of STEM jobs, of which computing accounts for 62 percent. The middle chart shows that the BLS expects that more than three-quarters of the new STEM positions created between 2014 and 2024 will be in the computing sector. However, as shown in the chart on the right, computing is expected to represent 58 percent of all STEM job openings, reflecting a greater need for replacement
of workers in existing positions in other STEM fields, presumably because the workforce in computing is, on average, younger than that in the other STEM disciplines and therefore less likely to lose workers through retirement.
Beyond the growth of the computing sector itself, analyses of job requirements show that employers now expect new hires to have significant levels of computing expertise, particularly at the high end of the labor market. A recent estimate from Burning Glass Technologies6 estimated that programming skills were important qualifications for nearly half of the job openings in the top quartile of the income distribution in 2015, as shown in Figure 4.7. Burning Glass’s analysis also projected that the number of coding job openings will grow faster (8.8 percent net growth for IT jobs requiring coding skills and 7.2 percent for all jobs requiring coding skills) than other career-track jobs7 (6.4 percent net growth) between 2016 and 2026. This suggests that those without the opportunity to acquire these skills may be less competitive for certain high-paying occupations, and thus potentially at an economic disadvantage. This could have a significant impact on certain underrepresented groups, who are less likely to have experience in computer science in K-12 and college, as discussed in Chapter 5; institutional structures that discourage the acquisition of these skills among some groups could perpetuate economic inequality.
The current starting salaries for bachelor’s degree graduates in computer science often influence the career choices of students. The 2017 National Association of Colleges and Employers (NACE) salary survey reported that computer science graduates received the second highest average salary ($65,540), right after engineering graduates ($66,097).8 NACE data show wide geographical differences in computer science, with a difference of $14,000 between jobs in the West and the South.9 For several above-average examples, one may look to the average reported starting salaries for CS graduates from Purdue University ($83,730 for 2015 graduates) (Purdue University, 2017), the University of Illinois at Urbana-Champaign ($85,000 for 2015 graduates) (University of Illinois at Urbana-Champaign, 2017), and the University of California at Berkeley ($99,700 for 2015 graduates and $103,963 for 2016 graduates) (University of California at Berkeley, 2016).
Both employment and degree production have risen substantially in the years since computing became an important component of the national economy. The question of whether a shortfall currently exists—that is, whether contemporaneous production of CS degrees is insufficient for employers to hire as many workers as and when they would like, at the going wage—in computing employ-
ment is a challenging one. Many estimates come primarily from sources tied to industry, which could be argued has some incentive to justify a more favorable hiring environment—to expand the pool of potential hires and create more competition, leading to more highly qualified talent and enabling them to offer lower wages.
As of 2014 the number of people employed in computer occupations was 3,916,100 (BLS, 2015). The total number of bachelor’s degrees in computer science awarded by U.S. institutions throughout the entire history of the field is only 1,313,034.10 The number of employees in computing occupations is therefore approximately three times larger than the number of bachelor’s degrees in computer science ever produced in the United States.
Another way to examine the relationship between employment levels and degree production is to plot the anticipated employment needs—calculated here by assuming that the projected growth from 2014 to 2024 is distributed equally over the decade—against the number of degrees produced each year in several STEM fields. The resulting graph appears in Figure 4.8, which is an updated version of a graph first presented by John Sargent, senior policy analyst, Department of Commerce, at the CRA Computing Research Summit on February 23, 2004. The leftmost bar in each cluster represents the projected growth in job openings for the most closely related occupational sector. The next four bars represent the number of degrees granted at the associate’s, bachelor’s, master’s, and doctoral levels.
The graphs in Figure 4.8 illustrate several ways in which computer science differs from other STEM fields. First, computer science is the only major discipline in which the projected number of job openings exceeds the rate of bachelor’s degree production (Freeman, 2016).11 Second, computer science is the only STEM discipline that produces a significant number of associate’s degrees, which currently account for nearly one-third of computer science degrees. Third, the number of doctoral degrees in computer science is proportionally smaller than it is in any other STEM field. Each of these differences has an effect on the balance between employment needs and degree production. While it is unclear how many of the new computer jobs would require bachelor’s degrees in CS (or any degree at all), they are worth observing; it may be that graduates in other disciplines simply go into occupations outside of their major fields, and that this is less common for CS.
In order to meet employment needs, one strategy is to hire employees with computer science degrees below the baccalaureate level. A second strategy is to hire university graduates with degrees in other areas and then provide those employees with whatever additional on-the-job training they require. A third is to hire computer science graduates from outside the United States. Driven by
11 In the discussion, Dr. Freeman noted that this was a high-level observation, so there may be other fields for which this same imbalance is true if one looks beyond the major disciplinary divisions.
its need to fill a rapidly expanding number of positions, industry is pursuing all three of these strategies.
That there is a market for workers with degrees below the baccalaureate level is supported by the unusually large number of associate’s degrees produced in the field. In 2014, U.S. institutions awarded 37,643 associate’s degrees in CIS, which comprises 31 percent of all CIS degrees. By contrast, associate’s degrees constitute an average of 4 percent of STEM degrees in other disciplines. Moreover, the number of associate’s degrees has been rising over time, as discussed in Chapter 2, though it has declined in recent years, in large part due to declines at the for-profits. It is worth noting in passing that computer science also produces a large number of certificates compared to other STEM fields, indicating that there is also a market in industry for these credentials.
Understanding the flow of employees both into and out of computing is complicated by the fact that graduates often take positions that would most naturally be categorized as outside their principal area of study. For example, many computer science graduates take positions in occupations other than computing, just as many graduates from other disciplines end up working in the computing field. Such migration is common among graduates in general.
As previously noted less than half of the people with CS bachelor’s degrees end up working in a core computer occupation. Conversely, computing occupa-
tions have an unusually high rate of in-migration, not only from other STEM fields but also from a wide range of academic disciplines: approximately 67 percent of the bachelor’s degree holders in the computing workforce have degrees from outside computer science. This is not necessarily because industry prefers to hire people without computer science degrees—it could also be due in part to their paucity in the labor pool.
The third strategy for meeting workforce needs—hiring highly skilled workers with technical degrees from universities outside the United States—has long been a matter of public controversy. Industry representatives have lobbied for increases in the number of visas available under the H-1B program (though many participants received their degrees at U.S. institutions), which allows companies to employ noncitizens in particular “specialty areas.” Under the law employment is on a temporary basis, although many employees who enter the United States under the H-1B program are later qualified to become permanent residents. Industry maintains that they need the ability to hire foreign workers because there are not enough people in the United States with the necessary skills. In their paper commissioned for the committee Bound and Morales report that the percentage of foreigners in the computer science workforce has grown over the last decade, rising from 10.6 percent in 1994 to 26.8 percent in 2015 (Bound and Morales, 2016), which is consistent with insufficient domestic supply to meet industry needs.
The Effects of Competition for Highly Skilled Software Developers
The expansion of the field of computer science has led to both specialization and heterogeneity among computer science graduates.
The intensity of the competition for the best software developers reflects a long-standing recognition that large variations in productivity exist among individual software developers. The original study in this area was published in 1968 by Sackman et al., who found that programmers with the same level of experience exhibited variations of more than 20 to 1 in the time required to solve particular programming problems. Beyond this high variability in coding time, the study revealed that individual programmers showed highly correlated differences in other metrics that contribute to overall productivity, in the sense that the best programmers were not only able to complete the problem in less time but in so doing also typically produced programs that had fewer errors and were more efficient in both running time and utilization of memory. Thus, the best programmers were in fact more than 20 times as effective.
Later studies have reaffirmed the results of the Sackman study and conclude that the best software developers are relatively few in number, but much more productive than the average.12 There is evidence that the variability in productivity has increased over time. In a Wall Street Journal story in November
2005, Alan Eustace, then Google’s vice president of engineering, said that the best software engineer at Google is 300 times more productive than the average Google software engineer—and this in a company that hires less than one-tenth of 1 percent of its applicant pool (Tam et al., 2005).
The variation in productivity is the critical dynamic that underlies employment policies in the software industry. Companies are under intense competitive pressure to identify and hire those extraordinarily talented individuals at the high end of the productivity curve. Particularly for startups employing a small team of implementers, hiring just one extremely productive person can make the difference between success and failure. After all, if the best software developer can do the work of 10, 100, or even 300 run-of-the-mill employees, a company that attracts such a superstar can compete effectively against a much larger enterprise. Thus, companies whose business depends on software production will try to hire applicants from pools in which the likelihood of finding the most talented individuals is high, such as graduates from top computer science departments, successful participants in collegiate programming contests, or entrepreneurs who have developed successful freeware and shareware systems on their own. Competition to attract employees from these populations is intense. Companies likewise have a strong incentive to avoid problem programmers and are unlikely to hire applicants whom they fear might fall at the low end of the scale.
In a column for the Los Angeles Times, Gary Chapman observed that “the software profession is beginning to resemble professional sports or the movie business much more than engineering. Exceptionally gifted programmers are referred to as ‘talent,’ like movie stars, and some have incomes that make those of film stars look puny” (Chapman, 2000). Images of career paths of computing superstars, which have been visible in contemporary news and media, may also contribute to increased interest among students in pursuing a career in computing. For example, wealthy technology magnates such as SpaceX and Tesla founder Elon Musk, or Facebook founder Mark Zuckerberg—including in films such as The Social Network—along with fictional characters on shows such as Silicon Valley and The Big Bang Theory, may create a glamorous or appealing image of computer science professionals among young people.
Workforce Demand in Key Areas
Another factor that complicates a full understanding of whether the supply of computing expertise is sufficient to meet workforce needs is that demand for particular specialties in computer science varies considerably over time. Some specialties have grown dramatically in importance in recent years. Three areas are commonly highlighted as facing acute demand: cybersecurity, data science, and machine learning.
One of the areas in which a need for skilled workers is most evident is computer security, which today is more often called cybersecurity or abbreviated even
further simply to “cyber.” Cybersecurity is a high priority for the U.S. government because of the need to protect classified data and other sensitive information. Ensuring that federal agencies are prepared against cyberattacks requires more workers trained in computer security. According to the Department of Homeland Security,
[a]s technology becomes increasingly sophisticated, the demand for an experienced and qualified workforce to protect our nation’s networks and information systems will only continue to grow. Information security and cybersecurity are rapidly growing industries with increased workforce needs (DHS, 2016).
Computer security is also increasingly important throughout the private sector, leading to significant growth in workforce needs in this area. In a report titled Job Market Intelligence: Cybersecurity Jobs 2015, Burning Glass Technologies (2015a) noted that
American employers have realized the vital importance of cybersecurity—but that realization has created a near-term shortage of workers that may require long-term solutions.
The report (Burning Glass Technologies, 2015a) listed the following findings about the cybersecurity job market:
- In 2014 there were 238,158 postings for cybersecurity-related jobs nationally. Cybersecurity jobs account for 11 percent of all IT jobs.
- Cybersecurity postings have grown 91 percent from 2010 to 2014. This is a faster growth rate than IT jobs generally.
- Cybersecurity postings advertise a 9 percent salary premium over IT jobs overall.
- Cybersecurity job postings took 8 percent longer to fill than IT job postings overall.
- The demand for certificated cybersecurity talent is outstripping supply. In the United States employers posted 49,493 jobs requesting a Certified Information Systems Security Professional (CISSP), recruiting from a pool of only 65,362 CISSP holders nationwide.
It is useful to put the first list item in context. American employers posted 238,158 jobs in the cybersecurity area, which is four times the entire production of bachelor’s degrees in computer science in that year, even though cybersecurity represents only 11 percent of IT jobs.
Data science is another area that has shown enormous growth in recent years. A 2014 issue brief from the Business Higher Education Forum outlines the growing importance of the field in the following terms:
The application of data science is pervasive in both the public and private sectors. Companies of all sizes rely on data science and analytics as key transformational components to their core operations. A 2011 report by the McKinsey Global Institute, Big Data: the Next Frontier for Innovation, Competition, and Productivity, noted that “big data” is growing at a rate of 40 percent each year and has the potential to add $300 billion of value to the nation’s health care industry alone, with broad application in virtually every sector, including scientific organizations and cultural institutions. Projections by Gartner, Inc., indicate that in less than 12 months, 4.4 million information technology (IT) jobs to support big data will be created globally. About 1.9 million of those jobs will be within the United States, and big data has the potential to create three times that number of jobs outside of IT (BHEF, 2014).
Additional recent studies have confirmed high levels of demand for “hybrid jobs” in data science that merge business and technology skills. A report from Burning Glass Technologies (2015b) offered the following assessment of employment demand in this area:
Data analytics, digital marketing, and mobile development are growing especially fast: Demand for data science skills has tripled over the past five years, while demand for digital marketing and mobile skills has more than doubled.
The Burning Glass report includes surveys of job postings indicating that “more than 250,000 positions were open in the last year for these hybrid technical roles”—again, four times the annual production rate of bachelor’s degrees in computer science. Those positions, moreover, offer high salaries and are, in many cases, open to students with less background in computer science than that of a computer science major.
The need for more qualified workers in such hybrid fields has led to an increase in the number of data-science courses, minors, and majors that colleges and universities offer to undergraduates. These trends are difficult to capture in systematically collected national data but are easy to illustrate by example. In recent years more than 25 universities—including Bowling Green State University, Columbia University, Drexel University, Florida State University, Marquette University, Portland State University, Purdue University, Rice University, the University of California at Irvine, University of California at San Diego, the University of Michigan, the University of Rochester, the University of San Francisco, and the University of Vermont—have created majors specifically titled “data science.” Other institutions offer similar courses of study under other names. At Stanford University, for example, the website for the major whose official title is Mathematics and Computational Science begins with a heading that describes the program as “the data science major.”
At the committee’s public workshop in August 2016 several speakers identified machine learning—particularly in applications that use multilayer neural networks
to create more sophisticated structures often referred to as deep learning—as an area of intense demand (Wing, 2016).
The emergence of machine learning as a critical technology also changes the prospects for employment of graduates in other fields. Just as new opportunities are created for graduates with the skills to apply machine-learning strategies to automate the analysis of massive collections of data, opportunities are closing in those fields in which such analysis was formerly conducted by people.
FINDING 4: Employment in computing fields has grown steadily since 1975, and the number of jobs in computing occupations far exceeds bachelor’s degree production in CS. The Bureau of Labor Statistics (BLS) projects that employment in computer occupations will rise more quickly than overall job growth for at least the next several years.
As discussed in previous sections demand for employees with computing expertise is high and has grown steadily over time. That need has also been projected to grow for the foreseeable future. The number of undergraduates interested in studying computer science—both as majors and as non-majors (as detailed in Chapter 3)—has been increasing rapidly, since about 2005. As computing becomes central to an ever-widening number of disciplines, student interest will presumably continue to grow. Several high-demand areas—including cybersecurity, data science, and machine learning—face labor shortages that academic institutions are currently unable to fill.
Each of these factors suggests a need to expand the nation’s supply of workers with sufficient understanding of computer science to meet the needs of the twenty-first-century workforce.
The emergence of new information technology industries, the increased reliance on computation in all parts of society and in research, and shifts in the demand for computing throughout the economy reflect changes in the field and its broad applications (NRC, 2012).
Two areas have been central in the last decade: the continued and increased need for information security, and data as a resource and driver for decision making. The protection of digital information and data; the protection of software and hardware systems and networks from unauthorized access, change, and destruction; and the education of users to follow best security practices are crucial to every organization. We rely upon a connected, networked, and complex cyberspace with vulnerabilities that is almost continuously under attack. Teaching safe information security practices to students and similar training of the workforce are increasingly expected. The National Science Foundation (NSF) Secure and Trustworthy Cyberspace program highlights many of the research challenges in
the area of cybersecurity involving hardware, software, networks, data, people, and the integration with the physical world.
During the last decade, computing has taken a new, more empirically driven path with the maturing of machine learning, the emergence of data science, and the “big data” revolution. Data science combines computing and statistical methods to identify trends in existing data and generate new knowledge, with significant applications throughout all sectors of the economy, including marketing, retail, finance, business, health care and medicine, agriculture, smart cities, and more. The availability and use of big data sets, combined with computation, simulation, and modeling, has created new academic areas including digital humanities and computational social sciences.
Software tools and systems for animation, visualization, virtual reality, and conceptualization have emerged as a medium for the arts (digital media and multimedia practices) and are driving advances in the entertainment industry (computer-generated graphics in films and video games, and digital methods in music recording), as well as training and education using virtual environments.
The digital age created by computing and information technologies has enabled a new sector of the economy with a wide range of temporary, short-term jobs for independent workers, mediated via the use of digital platforms, including through companies like Uber, Lyft, Airbnb, TaskRabbit, and DagVacay. While a recent study suggests that jobs in this area currently account for approximately 0.5 percent of the U.S. workforce (Katz and Krueger, 2016), they display significant potential for impact on jobs and the economy as a whole (NASEM, 2017).
Computing has become more pervasive among a host of academic disciplines, beyond just the practical use of ubiquitous software tools. New algorithmic approaches and discoveries are helping to drive advances across a range of fields, leading to new collaborations and an increased demand for deeper knowledge of computing among academics and researchers, challenging conventional disciplinary boundaries.
Important examples can be found in the area of computational science, where computational approaches are used to carry out scientific models or simulations to an extent that would be essentially impossible without the aid of a computer. Computational methods have contributed to significant advances and led to new specializations in domains such as chemistry and biology. As one measure of the impact of computing on the advancement of scientific research, the 2013 Nobel Prize in chemistry was awarded jointly to Martin Karplus, Michael Levitt, and Arieh Warshel for development of novel computational models for simulating chemical systems (Nobelprize.org, 2013).
As more and more software systems are used throughout the economy and become integral to daily life, it is becoming increasingly important for individuals to understand how to use them. The fraction of people who need to use software in more targeted ways, write scripts to process and prepare data, and integrate tools and systems is likely to increase. While it is not necessary for all individu-
als to be computer scientists or professionals in computing in order to succeed or function in society, it is clear that some level of computing skills and knowledge is increasingly necessary, and higher levels of competency can provide a significant practical and professional advantage in many contexts.
FINDING 5: Computing is pervasive, and its penetration is deep and growing in virtually all sectors of the economy, all academic disciplines, and all aspects of modern life. The broad opportunities in computing, both in the labor market and for enabling a host of intellectual pursuits, will continue to be drivers of increasing enrollments in undergraduate computer science, from both majors and non-majors. While there will probably be fluctuations in the demand for CS courses, demand is likely to continue to grow or remain high over the long term.
Increased Participation in Computing in Primary and Secondary Environments
Another potential driver of recent (and possible future) increases in CS undergraduate enrollments at the undergraduate level is increased engagement of students with computing in K-12. Throughout the history of the field, student exposure to computer science has been small compared to that of other STEM disciplines. Many high schools do not offer CS courses,13 and most do not require students to take a course in computing. Currently, 13 states do not allow CS courses to be counted toward high school graduation requirements.14 The changing role of CS in society, including the pervasive exposure of students to new technologies such as smartphones, along with the historically low and uneven access of K-12 students to computing (Google, Inc. and Gallup, Inc., 2016), has prompted a series of efforts to enhance the exposure of primary and secondary students to CS in recent years.
For example, in 2003, the Association of Computing Machinery established the K-12 Computer Science Task Force, and founded the Computer Science Teachers Association (CSTA) in 2004. CSTA produced the K-12 Computer Science Standards, which became the de facto national standards for the discipline, and produced and disseminated a host of classroom resources designed to develop student interest in computer science as an education and career pathway.
Universities have also participated in outreach to students in K-12 to help build interest in the field and break down stereotypical attitudes about computing. In particular, Carnegie Mellon University and Indiana University pioneered these efforts via a series of “roadshows.”15
13 Note that computing courses that teach computational thinking and/or coding are distinct from basic computer training courses that teach proficiency with word processing or other computer applications.
The NSF has sponsored such projects through its CS education programs. NSF’s Computing Education for the 21st Century was founded in 2011 with the aim of helping to reverse the decline in undergraduate computing majors that occurred after the dot-com bust “by engaging larger numbers of students, teachers, and educators in computing education and learning at earlier stages in the education pipeline” (NSF, 2012). Its Broadening Participation in Computing program track (building on alliances developed between 2006 and 2009) (NSF, 2016a) aims “to significantly increase the number of U.S. citizens and permanent residents receiving postsecondary degrees in the computing disciplines, with an emphasis on students from communities with longstanding underrepresentation in computing” (NSF, 2009). Another track, CS10K, “aims to have rigorous, academic computing courses taught in 10,000 high schools by 10,000 well-prepared teachers” (Cuny, 2015).
In 2016 the College Board introduced a new Advanced Placement (AP) course, Computer Science Principles (CSP), developed in part with NSF support, in 2,700 high school classrooms nationwide, “with the goal of creating leaders in computer science fields and attracting and engaging those who are traditionally underrepresented with essential computing tools and multidisciplinary opportunities” (College Board, 2017). This launch followed 10 years of development of course materials and other resources for high school teachers, and phased pilot efforts at dozens of high schools in the United States and Canada. In May 2017 it was announced that the implementation of this course has helped to increase high school student participation in AP computer science. In particular, the number of students sitting for any AP computing exam was approximately 111,000 in 2017, with approximately 51,000 of these students coming from the new CSP course, compared to approximately 57,000 in 2016 before CSP had launched (CRA, 2017b). Because participation in AP courses in general (and CS in particular) has been found to correlate to increased probability of majoring in the corresponding field (Mattern et al., 2011), the introduction of this new AP program is likely to newly contribute to an increase in the number of computing major enrollments in 2018 and computing degrees beginning in 2022.
In 2016 President Obama announced a national effort for “Computer Science for All,” to be led by NSF and the U.S. Department of Education in partnership with other federal agencies to boost K-12 students’ CS knowledge and skills. This persists as a non-profit, the CS for All Consortium.16
In addition, many industry-sponsored efforts have focused on supporting professional development and providing resources for CS teachers. For example, Google’s CS4HS aims “to provide CS teachers globally with an opportunity to improve their technical and pedagogical skills.” This program, launched in 2009 after several years of pilots, supports CS education professionals in providing professional development opportunities and educational resources to high school
CS teachers around the world.17 Google has also made a variety of learning resources available online. Another example is Oracle’s Oracle Academy, an initiative to assist teachers in preparing secondary students for college and careers, through the provision of educational resources online.18 Microsoft’s Technology Education and Literacy in Schools program works to connect U.S. high schools with technical expertise and support to help build and sustain computer science courses; specifically, the program connects volunteers to teach or support high school introductory or AP computer science courses and instructors.19
Several private-sector technology and engineering companies also have launched initiatives to boost exposure to CS courses and skills. For example, IBM’s Pathways in Technology Early College High Schools (P-Tech), launched in 2011, created grades 9-14 technology-focused early-college high schools that culminate in an associate’s degree. The program promotes postsecondary degree completion and career readiness and provides engaging academic courses and work-related experiences through partnerships with industry partners. Several states and cities have launched networks of P-Tech schools.20
Raytheon teamed with the Science and Innovation Center of Pinellas County Florida, SRI International, and St. Petersburg College to create an innovative cybersecurity educational program that prepares high school students for testing in industry-recognized certifications and earning credits toward a 2- or 4-year degree in network security from St. Petersburg College.
Combined, these and related activities indicate significant momentum toward earlier exposure of students to computer science. Impacts from efforts begun in 2009 could already be contributing to enrollment growth. Moving forward, continued increases in the level of exposure of students to computing before they enroll in college could have multiple possible effects on undergraduate enrollment. First, this earlier formal education in computing could provide some of the skills that certain students might otherwise seek in a postsecondary computing course. On the other hand, early exposure to computing could have the opposite effect of increasing course or major enrollments by stimulating student interest early and prompting further exploration in the postsecondary environment. Finally, the need for more teachers to administer these courses could also lead to increased demand on CS undergraduate education and training in order to meet it. These and related programs also have implications for student diversity, as discussed in Chapter 5.
This page intentionally left blank.