The Fallacy of a Skills Gap
Over the last decade it has became a common complaint to say that US workers don’t have the skills necessary to succeed and there is a current or looming skills gap, creating problems as companies search for suitable workers. The line of thinking sometimes goes so far as to say that if we could just get workers with the right skills, then all the problems the US economy has would go away.
Most written or spoken commentary raising the above point is flawed. It either (1) misinterprets economic evidence to support the idea of a skills gap or (2) is poorly conducted research that uses non-representative surveys and little data.
Many of the people doing this research have a vested interest in promoting the idea of a skills gap (if a large company complains about problems finding skilled workers, they can offload the blame to others and push for favorable policy).
Below we will go through why there is no evidence for a skills gap at a high level, drawing upon a large body of research.
Let’s review the broad economic evidence that would need to be true if there were a skills gap to see that at a macro level, a skills gap does not exist.
A skills gap implies that firms cannot find workers with the right skills, which suggests hiring problems. If firms were truly having problems hiring one or more of these three things would be happening.
- Wages would be rising (I bolded this on purpose)
- Hours worked would be increasing
- Companies would replace workers with capital i.e. machines
So how do we know this? We can use our understanding of how labor markets work to elucidate the mechanisms at play.
As firms compete for a limited supply of skilled labor, wages would be rising because of the need to outcompete other companies for the small pool of skilled workers. This follows from basic supply and demand.
When firms cannot find more workers to do additional work, they need to work with the existing resources they have in the short term. That means increasing the hours worked by existing employees to be able to churn out more output and grow.
In the longer term since companies need to find more workers but can’t, they will shift production towards machines, substituting capital for labor in the production process to replace the work that people would normally do. An example of this capital substitution is robotic arms in manufacturing on the assembly line instead of workers. We should note that this capital substitution could happen for reasons other than a skills gap as well, such as it is cheaper or more efficient to use machines.
Increasing hours worked by existing employees or substituting capital for labor will also increase wages for employees at a firm eventually. People generally won’t work more hours consistently unless they are compensated (especially if skilled workers are in short supply and therefore likely have many other opportunities). Capital itself is also generally complementary to skilled workers (think of how computing power allows engineers to analyze data faster), increasing their productivity and requiring them to be compensated more as they output more.
Therefore (2) and (3) also drive (1), meaning that if wages are not rising there really is no evidence of a skills gap.
Now let’s examine a time series of each of these variables over the long term to see what the data says.
First, we can look at median wages.
We see that outside of an increase in the late 1990s, which can be explained by growth in wages in specific industries due to other things and not a skills gap, there has been no real wage increase except recently. See here for why wages rose in the 1990s but in brief wage increases seemed to be concentrated in sales and service occupations and in the retail and finance industries, which doesn’t really square with a narrative about an economy wide skills gap.
Specifically, in the early 2000s there was little to no systematic wage growth. Given that many complaints about a skills gap originated in the 2000s, there is no evidence in wages that firms were struggling to hire a limited pool of skilled workers.
And we know why wages are rising right now: the labor market is tight with low unemployment overall, so firms have to raise wages to attract candidates as workers are generally scarcer.
As an aside you may note that here we use the median wage instead of the average wage. This is because the distribution of wages is highly skewed with some people making lots of money, meaning the median is a much better descriptor of the regular worker.
Some articles claim that using the median is not the right measure and that the average is better because it shows those at the top end of skills are seeing a return to their ability as their wages are rising. This misses the point: if there is a large skills gap, then the median wage should still be changing as the majority of workers’ skills are inadequate. Furthermore the wages of those at the top end of the earnings distributions are easily explained by things other than their actual skills such as work experience and economic rents.
Turning to hours worked we again see no systematic increase, meaning there is no evidence that firms are forcing existing employees to work harder to compensate for a smaller staff.
In fact the number of hours worked has been trending down since 2000 mostly due to recessions with it only rising recently as the economy has recovered from the recent financial crisis.
Finally, we turn to examining whether firms have systematically substituted capital for labor at a macro level.
This graph is a bit nuanced because it shows the percent change in the capital to labor ratio from year to year. You will notice no general trend but rather a cyclical pattern over time with periods of rising and falling segmented by the gray shaded regions.
These gray regions are economic recessions. We would expect the capital labor ratio to rise prior to recessions because it is easier for firms to fire workers compared to getting rid of large machines when the business slows, meaning that as they lay off workers, the amount of capital relative to the amount of labor in companies rises. As the economy recovers, companies hire back workers, causing this ratio to reverse course and fall.
Since the trend follows the general business cycle as opposed to consistently decreasing, which would need to be true if firms were repeatedly substituting capital for labor, either there always was a skills gap that has never gotten better or worse (hard to believe) or it never really existed.
The economic evidence at a macro level is pretty clear: there is no skills gap. Wages have not risen in a way consistent with that story and hours worked and the rate of substitution of machines for people have not changed in a way that would suggest companies are having trouble finding skilled workers.
Let’s then examine the traditional narrative and why it is flawed.
The story usually goes that if the wages of college graduates relative to those without a college degree is increasing, then the demand for college graduates (and thus high skilled labor) outpaces supply. We must need more college graduates and therefore more skilled labor. So there is a skills gap.
What do we see when we look at the data? Well people typically measure the relative demand for skilled workers by looking at the median difference between annual earnings for a college graduate relative to a high school graduate.
This premium has risen over time but effectively stagnated since the early 2000s, which suggests that the demand for college graduates is not increasing (or at least is not changing faster than the demand for workers without college degrees) as the earnings gap is flat. This runs counter to a lot of reports on the skills gap that claim we have a shortage of college graduates.
While it has not changed, the college premium has maintained itself, which might leads to arguments that there is still a shortage for skilled labor because otherwise the wages of college graduates would fall.
This misses the point as if there were an emerging skills gap, the premium should widen as wages for high skilled college graduates increase above and beyond those without a college degree, but we can also identify the reason for the premium in the first place to moot this point.
The recent college premium is mostly due to a fall in the wages of those with just a high school degree.
Focusing on the left graph of men, we see that again since 2000 the wage growth of those with a bachelor degree has stagnated (outside of an increase in the 1990s that we can explain by specific circumstances). In the meantime the wage for those with just a high school degree in red has fallen.
We focus on male wages for this because wages for women have been rising more uniformly due to reforms in the workplace that have shrunk the gender gap in pay (although it still exists).
This information raises the question for why wages have fallen so dramatically for those with just a high school degree. There are numerous potential explanations that don’t relate to a skills gap: decline of unions, changes in market power and rise of monopolies, changes in the minimum wage, changes in the types of jobs available, occupational licensing, etc.
Some of these have more evidence for them than others, but let’s focus on a specific one that is relevant for our discussion of the skills gap: credential inflation.
This data comes from a Vox article that used data from Burning Glass. It compared the fraction of people in a job with a bachelor degree to the fraction of job postings requiring a bachelor degree. We see that new positions are looking for bachelor degrees at a much higher rate than the rate at which people in those jobs have one.
This is clearly credential inflation because the nature of skill for many of these jobs has not plausibly changed to require a degree. However, since there are so many people with bachelor degrees now looking for jobs, firms can afford to increase their hiring requirements across the board.
A large amount of the jobs that used to be there for people without a college degree now require one despite it not being necessary from a skills point of view, explaining the rise in importance of college degrees.
A common argument that people make next is that students are just studying the wrong thing, which is a complaint among people arguing for more STEM graduates. This is counter to the credential inflation data, but we can also look at the actual data and figure out in what subjects students are earning bachelor degrees.
Between 2004 and 2015 the fields that increased the most were business, health (think nursing), and STEM fields like biology and engineering. Students clearly listened to the many reports in the early 2000s about the shortage of nurses and STEM workers.
What about business, which is the most popular major and also grew a lot? Business majors actually have one of the highest starting salaries and good employment prospects after college as it is essentially a field that prepares you directly for a career, making it valuable and attractive to many students.
Given that students responded to news about the need for graduates in certain fields, you would imagine we now have many people with related degrees working in those fields, but that is not the case.
There is general underemployment where a significant fraction of college graduates report being in a job that does not require a college degree, and the vast majority report being in a field that doesn’t match their major.
The picture looks the same for STEM graduates where a majority do not work in a STEM occupation. When we dig into why a lot of survey respondents cite working in another field, they claim either the pay was too low relative to other opportunities or they could not find a STEM job.
There clearly cannot be a broad STEM shortage (or broadly college degree shortage) because people are working in jobs different from their major, having trouble finding a related job (either due to salary suggesting there is not enough demand to raise wages or even no jobs at all), or in a job that they are overqualified for.
Some will counter with the high salaries in Silicon Valley. But Silicon Valley is not representative of other places in the US, and it has its own problems such as very high housing costs that prevent people from moving there, partly explaining why high salaries are needed to attract people from a limited pool of workers in the area.
This standard narrative around a lack of college graduates/skilled labor, even in specific fields, is not supported by the evidence. A broader question though is whether a college degree is even what employers are looking for.
The answer is generally no. This data comes from a survey by the Chronicle of higher education that asked firms what factors were important in evaluating new graduates for hire. Most of the top factors relate to work experience.
Companies want employees who have relevant experience that can contribute right away even amongst college graduates. The situation is even worse for those out of college. Companies do not want to take time to train and develop their employees and rely on individuals to come in with all the necessary experience. This places the burden of training on universities and individuals.
Universities themselves are great at teaching generic skills but bad at providing job experience. That is clearly best suited to on the job learning.
For students and workers, it does not make sense to learn hyper specific skills that you would acquire on the job before joining a company because it is risky. You might not get a job offer at the firm that utilizes those skills in which case they become useless.
These in demand skills you get on the job themselves also constantly change as companies change systems or technologies they use, and there can be a long lag between when they are popular and when people acquire them.
Many graduates as we have seen then end up underemployed as companies do not value the actual experience and skills they have.
As an example here is the popularity of programming languages over time based on google trends search data. We see that python has risen dramatically in popularity over time since 2004.
Imagine you were a programmer who graduated in the early 2000s and learned Perl. You could be a naturally great engineer with knowledge of all the most efficient algorithms and best practices for any language, but firms nowadays may not hire you specifically because you cannot program in Python. Because you lack the specific skill they are looking for, regardless of your overall qualifications and ability, there is no chance.
Personally, this happened to me when I was looking for data science jobs after graduating. Most companies who interviewed me, including some of the very large tech firms, did not advance me in their interviews because I did not know SQL and others wanted only people with 3–5 years of experience. Luckily my current company overlooked both those things.
Intuitively, if data science is such a hot and emerging field, then naturally the pool of experienced people will be small because most of these jobs were recently created. To attract people then you can either hire for talent or offer really high salaries to lure experienced people who are really in demand (this premium for experience also explains some of the large salaries in Silicon Valley and other places).
In fact I was able to learn SQL in my first week on the job (SQL is actually not very hard). Imagine, most companies who were complaining about hiring data scientists, were not willing to invest a week in training a new hire.
This standard narrative we hear about the skills gap just doesn’t hold water when we take a broad look at the evidence. College graduates do not seem particularly needed for many occupations they are hired for, and the actual economic reasons for the trends we see today are a much more complicated confluence of other factors.
We are seeing credential inflation, underemployment, and people working in fields other than their training. Companies themselves seem to value experience over education and training, declining to invest in teaching these things to their employees and opting to not hire for potential. All this talk about a skills gap distracts us from the real problems out there.
We need to change the narrative we are using and broaden the discussion to come up with solutions that address the real challenges in the US job market today: wage inequality and barriers to access.
I want to close with a quote from Peter Capelli, an economist at Wharton, in his overview of the lack of a skills gap that really resonates with the story presented here:
“The real challenge we face is that if everyone is hiring for the ability to do a job, rather than for the potential to do it well, how does anyone get that initial experience?”
References (beyond images):
Skill Gaps, Skill Shortages and Skill Mismatches: Evidence for the US
