“The Skills Gap”: Glut and Shortage at the Same Time

The Economic Policy Institute has done yeoman’s work in examining the truth beyond frequent assertions of a “skills shortage” in science, technology, engineering and mathematics (STEM) fields.  It is a common sentiment expressed by business executives that they cannot hire the right people with the skills that they need, or at least cannot hire them at the wage level at which they are willing to pay.  The EPI looked at actual employment trends and graduation rates in STEM, coming to the conclusion that there are twice as many STEM graduates as STEM jobs and that the import of guest workers is unnecessary.  Without delving too deep into the weeds of the study, I think they make three severe errors.

The first is simple – STEM grads working in non-STEM fields does not indicate low demand for STEM graduates.  The authors mention that many STEM graduates are in non-STEM fields for a variety of reasons, of which the first and most common is higher salaries elsewhere.  The authors interpret that as indicating low wages for STEM workers – I interpret that as a transferability of STEM skills.  Highly skilled statisticians, for example, can become professors and count as STEM workers – or they can work in finance and make multiples of that salary but not count as STEM workers.  Similarly, programmers and engineers are in high demand at consulting firms, in internal analytical roles at large firms, and generally anywhere in the business world where numerical competence and methodical thinking are useful.  That is, everywhere. 

The second issue is deep confusion over what counts as STEM employment.  Their STEM employment figures come from the NSF Science and Engineering Indicators Report 2012 – Table 3-3, to be specific.  Going to the source, STEM employment is narrowly defined as people actually working in science.  This is a fine definition for the NSF but completely inappropriate for EPI’s purposes – for example, it specifically exempts computer programmers, technicians, and managers!  Not to mention doctors, lawyers and teachers, fields which tend to draw many STEM graduates.  EPI seems to use “STEM employment”  interchangeably with “IT Employment”, which is both not measured by Table 3-3 and is not a great proxy for STEM employment.  Computer support workers (e.g., Dell’s call center) are “IT workers”.  EPI’s definitions seem to be fast and loose – which makes special sense given the issue I mentioned above, that there are many non-STEM jobs that draw on STEM competencies.

Finally, STEM personnel are not substitutable.  Gross numbers of STEM graduates aren’t particularly helpful if the American educational system is churning out large numbers of low-skilled graduates (which it is).  As anyone in the software business will tell you, a low-skilled development hire has an extremely high negative expected value to the enterprise – they not only make mistakes, but they make mistakes that cost a lot of money to fix.  This is even more true for firms that build bridges, boats, and skyscrapers.  A poor student in the humanities may do very well in the business world in sales or marketing, but a poor engineer is worse than useless.

As a corollary of this, the market for skilled STEM personnel is extremely thin and illiquid due to areas of specialization and varying competence.  If you need a highly-skilled developer proficient with a given software stack (e.g., an analytics engineer to redesign your entire reporting system), you may simply be unable to hire one at any price.  They are few and often will have idiosyncratic reasons for their employment choices.  On the other hand, a less-skilled developer does nothing for you.  So a real and meaningful skills shortage is not necessarily incompatible with a glut of STEM graduates.

These are critical methodological issues with the report that undercut their premise – the narrow definition of career paths leave much to be desired.  The question of substitutability, however, undercuts their entire premise – that there is no skills shortage because there is a STEM glut.  The thin and illiquid market in certain important skills mean that there can be skills shortages coexisting with a STEM glut.  Based on my experience in the software world, that is a truth that rings very true to me.

Tyler Cowen makes a pithy but even larger point:

Similarly, it could have been pointed out that, before the rise of Hyundai, South Korea had just the right number of auto workers (not many) for their factories (also not many).