Robert Hoffman photo
Step Right Up and See the Fancy (and Flawed) EPI Data

My college statistics professor was the source of a truly memorable warning:  “There are PT Barnums in data analysis who can easily influence people who’ve either never taken a statistics course or slept through one.”  That’s good advice for policymakers because in major policy debates statistical wars usually occur, and some PT Barnums do surface.

The Economic Policy Institute (EPI) is vying for the PT Barnum title.  A recent EPI report reached a dramatic conclusion:  The shortage of U.S. science, technology, engineering and math (STEM) workers is a myth.   This seriously flawed “STEM supply is greater than demand” conclusion just does not jive with the most important real world data out there:  thousands of unfilled STEM job openings.  

Here’s a snapshot of STEM demand today:

  • Microsoft’s General Counsel, Brad Smith, recently told the Senate Judiciary Committee that Microsoft currently has more than 6,300 U.S. job openings, and more than half are in “core research, engineering and development.” 
  • Mr. Smith also noted in his testimony that a group of 25 employers, ranging from AT&T to Accenture, together have more than 20,000 job openings in New York City alone.  
  • Last month, Business Insider reported that the ten leading U.S. tech firms together had more than 26,000 U.S. job openings.  
  •, a website that aggregates publicly posted listings of IT positions, today reports there are more than 83,000 tech job openings in the U.S.

These job openings track with a December 2012 analysis of Bureau of Labor Statistics (BLS) data we at ITI conducted with the Partnership for a New American Economy and the U.S. Chamber of Commerce.  Our analysis found that numerous sectors of the U.S. economy that depend on Master’s and Ph.D. graduates are experiencing underemployment -- i.e., a skilled workforce shortage.  

A national survey conducted by the Society for Human Resource Management reached similar conclusions:  The survey found that two-thirds (66 percent) of organizations currently hiring full time staff are having difficulty recruiting for specific jobs -- up from 52 percent in 2011 -- and scientists, engineers, and highly skilled technicians were cited as the top three most difficult positions to fill.

Absent changes in policy, BLS data also suggest that the mismatch of U.S. STEM demand in excess of supply will get worse.  The BLS estimated that approximately 122,000 new job openings will be created each year in computing occupations requiring at least a Bachelor’s Degree through the end of the decade.  However, U.S. universities currently produce annually roughly 51,000 Bachelor’s degrees in computer science.   And in one emerging field – Big Data – a recent McKinsey Report predicts we could see a possible shortfall of up to 190,000 experts with deep analytical skills and an astounding 1.5 million “data-savvy” managers and analysts during the next five years.

So, the EPI report is challenging some pretty tough real world data.  This would suggest, for EPI at least, that U.S. employers are engaged in a massive conspiracy to advertise for U.S. STEM jobs that do not exist.  The conspiracy also would have to include job forecasters from firms like McKinsey, who are suggesting that STEM demand over supply is bound to get worse.  Perhaps more shocking, the conspiracy would also have to involve federal government analysts at the BLS who have produced data that contradict EPI’s fundamental conclusion.  That’s quite a conspiracy.  I wonder if EPI thinks we faked the moon landing.

An initial review of the EPI report itself is just as disturbing.  To support its conclusion, EPI first points to a 2009 survey of STEM college majors one year after their graduation, and the data reveal that of those surveyed, just over half are hired in a STEM job. 

Time out.

The EPI is saying that STEM supply exceeds demand for the entire U.S. economy based on surveys of newly minted STEM college graduates in entry-level jobs in 2009 -- when the U.S. economy was mired in the biggest economic downturn since the Great Depression. 

[Read that last sentence again, but more slowly.]

Really?  Can we actually conclude that the perspectives of first-year college graduates during a massive economic recession is a good read on the present and future of the overall U.S. STEM economy?  Clearly not.   For those who think otherwise, give me a call.  I have some monuments here in D.C. for sale you might be interested in.

Interestingly, two of the EPI study’s authors made a similarly flawed conclusion in a 2007 report for the Urban Institute.  Stuart Anderson, who runs the National Foundation of the American Policy, noted significant flaws in their 2007 analysis, citing data from the National Science Foundation (NSF).  I’ll let Stuart do the talking: 


“To reach its conclusion that America is producing too much talent, the Urban Institute sought to match science-related jobs with U.S. degree production by using a definition of science and engineering (S&E) jobs that excluded 8 million employed U.S. professionals who use their math and science degrees in the place of work.”


Mr. Anderson, one of the more careful policy analysts in Washington, cautioned that “attempts to identify the precise number of workers needed in a given field [are]… impossible, given the vagaries of consumer demand, competition, future economic conditions and the potential impact of innovations in the marketplace.”

Apparently, Mr. Anderson’s advice did not stop the 2007 report’s authors from reviving this faulty thinking in 2013, and frankly, I am actually impressed they got the funding.  

The EPI report doesn’t provide a clear definition of a STEM job.  Certainly, a position at traditional information technology (IT) or computer firms, such as Oracle or HP, is easy to understand.  However, as NSF data make clear, millions of non-traditional STEM jobs require STEM skills, a fact that makes the competition for STEM talent even stronger. 

The EPI report also suggests that if STEM demand exceeded supply generally, we should see a general increase in wages.  However, the report uses or analyzes data across the entire range of computer and IT industries, which masks salary increases in specialized, highly skilled professions. 

Where there are short supplies of talent, we have seen increases in compensation. For example, since 1999, wages for computer research scientists have increased 54 percent, and for software engineers, the increase has been 52 percent over the same period -- well ahead of inflation.  And for high demand fields like mobile application developers, and wireless network engineers, salaries are expected to increase by nearly 10 percent in the near term.

The EPI report also suggests that temporary visas have created downward pressure on wages for U.S. workers. This finding contradicts a wide range of studies, including 2011 Government Accountability Office (GAO) data that made apples to apples comparisons of skilled talent and found that H-1B professionals earn comparable and, in some cases, higher wages than U.S. professionals of similar age and experience.  For example, the GAO found that the median salary of H-1B electrical engineers, ages 20-39, was $80,000, but for U.S. workers, the median salary was $75,000. 

I could go on, and I expect others will come forward with more perspectives on the EPI report, but the lesson is clear:  There are a few PT Barnums out there who will tell policymakers to look at their fancy data and ignore what’s going on in the real world.  That’s why it’s important to look carefully at analyses on all sides of an argument, and the sources themselves.

Public Policy Tags: Workforce