The Bureau of Labor Statistics projects that by 2020 there will be 1.4 million computer-science-related jobs available and only 400,000 computer science graduates with the skills to apply for those jobs.
Then about a year later, the White House web site posted a factsheet dated December 8, 2014 that contains the following statement:
By 2020, more than 50 percent of STEM jobs are projected to be in computer science-related fields. If current trends continue, 1.4 million computer science-related jobs will be available over the next ten years, but only 400,000 computer science graduates will be added with the skills to apply for those jobs.
Hence, the blog post gives the source as the Bureau of Labor Statistics (but gives no link) and the fact sheet gives no source at all. However, the Hillary Clinton web site contains a factsheet dated June 27, 2016 (according to the URL) that repeats the same basic claim but does link to a source:
—and by 2020, there will be 1.4 million computer-science related jobs in America, with only 400,000 computer science graduates to fill them.[5]
The footnote [5] references an ITIF paper which itself references a Code.org chart. That chart does not give a source but text with the same chart at code.org lists the sources "College Board, Bureau of Labor Statistics, NSF". Likewise, an earlier report from Microsoft which has similar numbers gives the source for the jobs number to be the Bureau of Labor Statistics and the source for the graduates number to be the National Center for Education Statistics (NCES).
Analysis of the Claim
Overview of the Claim using the Code.org Chart
Figure 1 below basically replicates the original Code.org chart. As can be seen, it shows the projection of 400,000 computer science students (actually, graduates with bachelor degrees in computer science) and 1,000,000 unfilled computing jobs in 2020. Figure 2 is a slightly revised version of the chart. The main difference is the addition of the blue area that shows a revised estimate of computer science students that is explained in the next section. Figure 2 also starts in 2010 to make clearer that these are new jobs since 2010 and not all jobs. It also changes the label "CS Graduates" to "US CS Grads" to clarify that the projection is looking only at computer science graduates from U.S. colleges. Finally, Figure 2 changes the label of "Unfilled Jobs" to "Not US CS Grads" since the assumption that computing jobs can only be filled by graduates with computer science degrees from U.S. colleges is false (as explained below).
Now if one wanted to estimate the number of unfilled computer science jobs that would exist in 2020, the following would seem to be a reasonable approach:
Post- Post- Less than
Doctorate Master's Bachelor's Associate's Master's Baccalaureate 4 Year
Degrees Degrees Degrees Degrees Certificates Certificates Certificates Total
-------------------------- ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------
Computer Science 1,999 31,552 60,309 36,421 109 1,573 29,228 161,191
Mathematics and Statistics 1,802 8,269 23,136 2,697 92 331 168 36,495
As can be seen, there are 60,309 bachelor degrees in 2015, over 50 percent more than the 40,000 per year estimated by Microsoft. Following are the number of bachelor's degree from 2000 to 2015 for both computer sciences and mathematics and statistics:
S&E bachelor’s degrees, by field: 2000–15 (Thousands)
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Field
------ ----- ----- ----- ----- ----- ----- ----- ----- ----- ----- ----- ----- ----- ----- ----- --------------------------
37,519 43,597 49,706 57,926 59,968 54,588 48,000 42,596 38,922 38,496 40,107 43,586 47,960 51,586 56,130 60,309 Computer sciences
11,714 11,437 12,254 12,863 13,735 14,816 15,310 15,551 15,841 16,208 16,832 18,021 19,819 21,567 22,226 23,136 Mathematics and statistics
Sources: TABLE 33. Computer sciences degrees awarded, by degree level and sex of recipient: 1966–2012
TABLE 34. Mathematics degrees awarded, by degree level and sex of recipient: 1966–2012
TABLE 5-1. Bachelor's degrees awarded, by sex and field: 2004–14
The above table shows that the estimate of 40,000 CS bachelor degrees per year appears to be based on the level in 2010. This is close to the low that was reached in 2009 during the financial crisis. This can also be seen in the graph on the second page of a powerpoint presented by Michael Wolf of the Bureau of Labor Statistics that was presented at a workshop on August 15, 2016. As can be seen, the number of such degrees reached about 60,000 in 2004 and has reached that level again. Hence, 40,000 has proved to be a very bad estimate, at least until now. The column labelled "Actual w/ 2015 ext" in the following table shows what the estimate would be if the actual numbers from 2011 to 2015 were used and the value for 2015 is extended to 2015 through 2020.
Computer Science Bachelor Deg. Unfilled CS Jobs
New ------------------------------- --------------------
Computing Code.org Actual w/ Code.org Revised
Year Jobs Estimate Actual 2015 ext Estimate Estimate
---- --------- --------- --------- --------- --------- ---------
2010 0 0 40.107 0 0 0
2011 140 40 43.586 43.586 100 96.414
2012 280 80 47.96 91.546 200 188.454
2013 420 120 51.586 143.132 300 276.868
2014 560 160 56.13 199.262 400 360.738
2015 700 200 60.309 259.571 500 440.429
2016 840 240 319.88 600 520.12
2017 980 280 380.189 700 599.811
2018 1120 320 440.498 800 679.502
2019 1260 360 500.807 900 759.193
2020 1400 400 561.116 1000 838.884
This difference between the old and new estimate is shown as the blue area in Figure 2. As can be seen, it decreases the number of "unfilled jobs" by 161,116. This is a fairly conservative estimate since it seems likely that the number of computer science bachelor degrees will continue to increase after 2014 rather than remain at the 2014 level as assumed by the estimate.
Table 1.7 Occupational employment and job openings data, 10-year projected, and worker characteristics, first year
(Numbers in thousands)
Employment Employment change Percent Job openings Median
First Last -------------- ----------------- self due to growth annual First Date
Title Code Year Year First Last Number Percent employed and replacements wage Archived
-------------------- ------- ----- ---- ------- ------- ------- ------- -------- ---------------- ------- -----------
Computer Occupations 15-1100 2010 2020 3,426.0 4,184.7 758.8 22.1 4.5 1,366.2 $73,710 04-13-2012
Computer occupations 15-1100 2012 2022 3,682.3 4,333.6 651.3 17.7 3.0 1,240.1 $76,270 12-20-2013
Computer occupations 15-1100 2014 2024 3,916.1 4,404.6 488.5 12.5 2.6 1,083.8 $81,430 01-01-2015
As can be seen above, the BLS did project nearly 1.4 million job openings in computer occupations from 2010 to 2020. However, this dropped to about 1.2 million for 2012 to 2022 and 1.1 million for 2014 to 2024. It would seem wise to use updated projections than to continue to quote projections that were made over 4 year ago. Unfortunately, it's not easy to update the 2010-2020 projections since the later projections are for different 10-year periods. Current projections for 2010-2020 should likely be lower to more closely match the later 10-year periods but it is possible that the higher number is due to the hiring back of computer workers who were laid off in the financial crisis. In that case, however, the additional jobs may be "recovered jobs", not "new jobs".
Many citations incorrectly say that BLS projects the shortage when, in actuality, BLS economist Jennifer Chi clarified in an email that all of these attempts are made by third parties. “[It is] an incorrect comparison of the total employment and total labor force projections, which are two separate and fundamentally different measures.”
Hence, an economist from the Bureau of Labor Statistics states that the comparison of the 1.4 million and 400,000 number is an incorrect comparison.
The first significant problem with Microsoft’s report is the assumption that job openings “in computing” not filled by college graduates with computer science (CS) degrees will go unfilled. It is a well-known fact that computer science graduates are not the only source of new hires in computing. In the late 1990s, the Department of Commerce (DOC) published a report warning of looming labor shortages in the information technology (IT) sector, citing the lack of college graduates with CS degrees as a principal reason (DOC 1997). The Government Accounting Office (GAO)1 later published a report chastising the DOC for its faulty methodologies, noting that “IT workers come from a variety of educational backgrounds and have a variety of educational credentials such as master’s degrees, associate degrees, or special certifications” (GAO 1998, 8). National Science Foundation (NSF) data at the time indicated that “only about 25 percent of those employed in computer and information science jobs in 1993 actually had degrees in computer and information science” (GAO 1998, 5).
The same is true in the main computer-related occupations, according to the most recent comparable NSF data, presented in Table 1. As the table shows, even in the occupation where one would most expect workers to hold a CS degree—computer and information science researchers—only 43 percent (approximately two out of every five) hold a CS degree. A similar percentage (41 percent) of computer software engineers have CS degrees, as do computer system analysts (39 percent); and 17 percent to 29 percent of workers in the remaining listed computer-related occupations earned a CS degree. Professor Norman Matloff of the University of California, Davis, recently conducted a similar analysis of data from the NSF’s National Survey of College Graduates, which showed that “only 40.2% of those with Software Engineer, Programmer or Computer Scientist titles came to the profession from a CS degree” (Matloff 2012, 5). It is important to note that neither of these data sets includes workers with less than a bachelor’s-level education; as Matloff mentions in his paper, many titans of the tech world, including Bill Gates, Larry Ellison, and Mark Zuckerberg, do not have a college degree in any field.
The above arguments and others are also covered in a paper titled The STEM Anxiety Business published in March of 2016.
Often references do name the source but do not link to or otherwise provide a specific source by which anyone can verify the number(s) in question. That is the case with the first source given above which states that the numbers are Bureau of Labor Statistics projections. With no link, however, it's not possible to tell that only one of the numbers (the 1.4 million jobs) actually came from the Bureau of Labor Statistics. As can be seen in the references listed below, a large number of them wrongly give the Bureau of Labor Statistics as the source for both of the numbers. This would imply that the BLS made some effort to ensure that the numbers were comparable. Since they are the source of only one of the numbers, however, this cannot be the case.
Finally, some other references simply source or provide a link to another reference that suffers from one of the previously described problems. This is the case with the third source given above. As a result of all of these problems, it's impossible for anyone to verify the claim. It's also possible for the claim to change over time. In this case, the claim morphed from Microsoft's claim that there could be a shortfall of 80,000 computer science graduates per year to the claim that there would be a million unfilled jobs by 2020.
Still, the biggest error is the assumption that the only people who can hold computer jobs are computer science graduates from U.S. colleges. As stated above, the assumption that only computer science graduates can hold computer science-related jobs has never held in the past and there is no evidence that it will hold in the future. And, stunningly, this projection makes no accounting for all of the H-1B Visa workers and recent immigrants who graduated from non-U.S. colleges.
All of these errors were likely made more possible by the bad or missing sourcing by nearly all of the references. Of the many references listed below, only the 2012 Microsoft source correctly attributes the two numbers in the claim to two different sources. Of course, this is because it seems to be the chief originator of the claim.
Footnotes:
Many citations incorrectly say that BLS projects the shortage when, in actuality, BLS economist Jennifer Chi clarified in an email that all of these attempts are made by third parties. “[It is] an incorrect comparison of the total employment and total labor force projections, which are two separate and fundamentally different measures.”
Summary
The 400,000 projection of computer science graduates appears to have come from extrapolating the 2010 figure and has proven to be significantly low. It appears that the BLS did project nearly 1.4 million computer science-related jobs for 2010-2020 but that projection has dropped for later 10-year periods and may need to be revised. In any event, these two numbers come from different sources and a BLS economist states that comparing the two is an "incorrect comparison".
References to the Claim
Following are references to the claim that, by 2020, there will be 1.4 million computer-science related jobs in America, with only 400,000 computer science graduates to fill them. Some of the early references in 2012 refer to 2018 instead of 2020 and some later ones into early 2013 refer to 1.2 million instead of 1.4 million jobs. They all refer to 400,000 computer science graduates over ten years (or 40,000 annually).
2012 References
[4] This estimate is based on the U.S. Bureau of Labor Statistics’ occupational employment and job openings data, projected for 2010–2020. Available at http://www.bls.gov/emp/.
[5] Integrated Postsecondary Education Data System from the U.S. Department of Education’s National Center for Education Statistics (NCES). Available at https://webcaspar.nsf.gov.
2013 References
2014 References
Sources: College Board, Bureau of Labor Statistics, NSF
2015 References
2016 References
[9] https://code.org/images/cs-stats/more-jobs-than-students.png
Note: Surprisingly, the BLS report linked to by the reference above contradicts its own numbers, projecting just 778.3 new computer jobs and making no mention of the number of graduates.
[5] Nager, Adams and Atkinson, Robert, “The Case for Improving U.S. Computer Science Education,” The Information Technology & Innovation Foundation, May 2016. http://www2.itif.org/2016-computer-science-education.pdf.
2017 References
2018 References
2019 References
2020 References
2021 References
Information on STEM Workers
Search of Google on shortage of STEM-related workers
Commentary on the Skills Gap
Composition of STEM Workers in Selected Locations: 2014
Ages by Occupation in the Selected Locations: 2014
Computer Workforce by Age: 2012
Information on H-1B Visas
H-1B Data
H-1B Information
H-1B Labor Condition Applications: 2001-2013
Analyses of Studies
Analysis of "Immigration and American Jobs"
Analysis of the claim that each H-1B worker creates 1.83 jobs
Analysis of the claim that each STEM worker with an advanced U.S. degrees creates 2.62 jobs
References to Claims that Foreign-born Workers Create Jobs
Claims regarding computer science jobs on White House website
Claim that there were over half a million good-paying tech jobs unfilled last year
Claim that there could be 2.4 million unfilled STEM jobs by 2018
Analysis of "Foreign STEM Workers and Native Wages and Employment in U.S. Cities"
Analysis of "STEM Workers, H-1B Visas, and Productivity in US Cities"
A Look At Mariel Using R
To contact me, use the contact box at the bottom of http://usbudget.blogspot.com/.
