The evidence supports both, in different places. Research finds the skills gap is largely an employer-side problem of narrow requirements and collapsed training, while wage premium data shows genuine scarcity in specific new-technology specialisms. The fix differs entirely depending on which you have.
This is not an academic distinction. The two diagnoses lead to opposite actions.
If it is a hiring problem, the answer is changing your requirements, your reach, and your willingness to develop people. If it is a skills problem, the answer is paying more, training internally, or accepting longer timelines.
Most companies never make the diagnosis. They assume scarcity, pay the premium, and wait. Here is what the evidence actually says, and a diagnostic for working out which problem you have. Rise92 operates at the scarce end of this market through curated sourcing.
The test economists actually use
Before the arguments, the method worth knowing.
A genuine shortage shows up in wages. If a skill is truly scarce relative to demand, its price rises. If employers complain about scarcity while pay stays flat, something other than scarcity is happening.
This is why the wage premium is the most informative single number in the debate, and it is what both sides of the argument keep returning to.
Why it is not conclusive on its own: wage data at the occupational level is too coarse. Broad categories do not correspond cleanly to actual skills, which is one reason the debate has run for over a decade without resolution.
The case that it is a hiring problem
The strongest version of this argument comes from Peter Cappelli, a management professor at Wharton, whose work has challenged the shortage narrative for more than a decade.
His central finding: in a National Bureau of Economic Research working paper, Cappelli reviewed the evidence behind employer complaints and concluded that academic research does not support the existence of serious skill gaps or shortages in the US workforce. What exists instead is a mismatch in the opposite direction, where the average worker has more education than their current role requires.
In his own summary for SHRM, published under the title The Skills Gap Is a Myth, Cappelli accepts that employers genuinely struggle to hire, then argues the solution sits in their hands rather than in the labour supply.
Four employer-side mechanisms he identifies:
One. Training investment collapsed. The figures are stark. In 1979, young workers received roughly two and a half weeks of training per year. By 1991, only 17% of young employees reported any training in the previous year, and in a later measure, only 21% reported receiving training during the previous five years. Cappelli’s framing is that this is a training gap rather than a skills gap.
Two. Hiring requirements became unrealistic. Cappelli’s characterisation is that employers increasingly look for someone currently doing exactly the same job somewhere else. That is a reasonable individual preference and a collectively impossible one, since if nobody will provide initial experience in a skill, nobody accumulates it.
Three. Screening tools filter out capable people. Automated systems search for experience that appears on a resume in a recognisable form, which removes people who could do the work but cannot evidence it in the expected format.
Four. Responsibility shifted. Cappelli describes a move away from employers assuming responsibility for technical skill training, toward expecting employees and educational institutions to carry it.
His recent position is unchanged and more pointed. In a 2026 conversation with UNSW Business School, Cappelli states that the mismatch sits primarily on the employer side rather than the university side, and that employers have decided training is not their job while paying a significant but largely unacknowledged cost for that decision.
The evidence point that lands hardest: in employer surveys he reviewed, only 12% reported difficulty hiring college graduates with STEM skills. That is difficult to reconcile with a general shortage of technical capability.
The case that it is a skills problem
The counter-argument deserves a fair hearing, because the dismissive version of the shortage-is-a-myth position is also wrong.
The strongest published rebuttal appeared in Harvard Business Review under the title Employers Aren’t Just Whining, and it makes three points that survive scrutiny.
One. The timing argument fails. Employers reported difficulty hiring skilled workers before the recession, which undermines the view that complaints are a cyclical artefact of a loose labour market.
Two. Cappelli’s own conclusion is narrower than its popular version. He concludes there are no major shortages of workers with basic literacy, numeracy, or engineering and technical training. That is not the same as concluding employers have no real difficulty hiring people who can work with new technologies, which the rebuttal argues they do.
Three. Pay for skills is rising within occupations. This is the economists’ own test, applied properly. Broad occupational wage data is too coarse to detect scarcity, but evidence of rising pay for specific skills inside the same occupation is exactly what a genuine shortage looks like.
That third point is where current data matters most. PwC’s 2026 Global AI Jobs Barometer, based on analysis of over a billion job advertisements across 27 countries, found a 62% wage premium for AI skills within the same occupation, up from 57% the previous year.
Within the same occupation is the operative phrase. That is not a composition effect or a category artefact. It is the signal the rebuttal said to look for, measured at scale, and it is rising.
The measurement problem both sides acknowledge: it is genuinely hard to measure skills relating to new technology, which is why this argument has never been settled cleanly and probably cannot be.
What the evidence actually supports
The honest synthesis is that both diagnoses are correct, in different parts of the market.
| Where the evidence points | Diagnosis |
| General technical capability, broad engineering skills | Hiring problem. Supply exists; requirements and reach are the constraint |
| Entry and mid-level roles | Hiring problem, compounded by collapsed training investment |
| Established specialisms with deep pools | Hiring problem |
| New-technology specialisms with rising within-occupation pay | Genuine scarcity |
| Production experience in fast-moving areas | Genuine scarcity, created by how recently the skill existed |
The mechanism that connects them: Cappelli’s own argument explains why genuine scarcity appears in new specialisms. If training has collapsed and employers only hire people already doing the job, then any skill that did not exist five years ago will have almost no supply, because no institution was producing it and no employer was developing it.
In that reading, the scarcity in emerging specialisms is real and is itself a consequence of the employer-side problem. The skills gap and the hiring problem are not competing explanations so much as the same failure observed at different time horizons.
Which problem do you have?
The practical payoff. Six diagnostic questions, answerable from information you already hold.
| Question | Points to a hiring problem | Points to a skills problem |
| Are you receiving applications? | Yes, in volume | Few, even after months |
| Are you rejecting most at screening? | Yes, on requirements | No, people are not applying |
| Has market pay for this skill risen sharply? | No, flat | Yes, steeply, within the occupation |
| Could a capable adjacent person learn this in a quarter? | Yes | No, it takes years |
| Does your posting require current experience in the exact role? | Yes | Requirement is genuinely irreducible |
| Where do your hires come from? | Almost all inbound | You have tried every channel |
How to read it: mostly left-hand column means your constraint is requirements and process. Mostly right-hand column means the scarcity is real and you are choosing between paying more, building internally, or waiting.
The mixed result is the most common, and it usually means one genuinely scarce component sits inside a role definition that is otherwise over-specified. Separating the two is where most of the value is.
If it is a hiring problem
Five changes, in order of how quickly they pay off.
Separate the must-have from the convenient. Most job specifications bundle one genuinely scarce capability with eight preferences. Write down what the person must be able to do in month one, and move everything else to desirable.
Stop requiring current experience in the identical role. This is the single requirement that creates artificial scarcity, because it means hiring only from the small pool already doing exactly that job rather than from the much larger pool who could.
Check what your screening is removing. If automated filtering is searching for a specific resume shape, it is removing capable people who present differently. Sample your rejections and read 20 of them properly.
Restore some training investment. This is slower and it is the structural fix. Cappelli’s point is that the cost of not training is real and simply goes unmeasured, appearing instead as vacancy, premium pay, and extended searches.
Check whether your reach is the actual constraint. If your pipeline is almost entirely inbound, you are only reaching people who are actively searching, which is a small and unrepresentative slice. Off-market sourcing through closed networks addresses that specifically.
If it is a skills problem
Four options, none of them free.
Pay the premium knowingly. If the within-occupation wage premium is real, paying it is rational. Pay it deliberately rather than discovering it at offer stage after three failed searches.
Build internally. Take a strong adjacent engineer and develop the specific capability. Slower, cheaper, and the only option that increases supply rather than redistributing it.
Widen the geography. Scarce specialisms are not evenly distributed, and a skill that is scarce in one market may be better supplied in another. Pakistan market data counts 300,000+ export-ready engineers across AI, cybersecurity, blockchain, and cloud, though senior production depth in AI specifically is thinner than headline figures suggest.
Reduce the requirement. Sometimes the honest answer is that the role as specified does not exist in sufficient quantity at any price, and the work needs splitting or sequencing differently.
On the cost comparison: all four options have a price, and the one most companies default to, extended vacancy, is usually the most expensive and the least visible. The breakdown sits in transparent pricing.
Frequently asked questions
Both descriptions capture something true. Academic research reviewed by Peter Cappelli found no evidence of serious general skill shortages in the US workforce, with the mismatch running in the opposite direction toward overeducation. At the same time, rising pay for specific skills within the same occupation is the recognised signal of genuine scarcity, and PwC’s 2026 data shows a 62% premium for AI skills on that measure.
The collapse in employer training investment, which is well documented. Young workers received around two and a half weeks of training annually in 1979, while later measures found only about a fifth reporting any training over a five-year period. Combined with requirements that specify current experience in an identical role, this produces scarcity that employers create and then describe as a labour supply failure.
That employers reported hiring difficulty before the recession, and that pay for skills is rising within occupations rather than just between them. The second point matters because it is the test sceptics themselves proposed, and it is now being measured at scale across more than a billion job advertisements.
Check whether you are receiving applications and rejecting them at screening, which points to requirements, or receiving almost none, which points to reach or genuine scarcity. Then check whether market pay for the specific skill has risen sharply within the occupation. Flat pay alongside hiring complaints is the clearest indicator of a self-created problem.
What to take from this
The choice is narrower than the marketing suggests. Establish whether the person is an employee or a genuine contractor, establish whether you own a local entity, and five of the six options usually eliminate themselves.
Then model total cost per country rather than comparing provider fees, because statutory variation dwarfs fee variation.
And keep the sourcing decision separate from the employment decision, because conflating them is how companies end up paying an ongoing percentage for something a one-time arrangement would have covered.
To work through which structure fits your situation, get in touch.
The useful conclusion
The question in the title has a frustrating answer, which is that it depends on the role, and that most companies have never checked.
The checking is cheap. Look at whether applications arrive and get rejected or never arrive at all. Look at whether pay for the skill has moved. Look at whether a capable adjacent person could learn it in a quarter. Those three questions separate the two problems in most cases.
What follows is genuinely different. One is solved by changing what you ask for and where you look. The other is solved by paying, building, or waiting.
Treating the second as the explanation when you have the first is the most expensive mistake in hiring, because it is invisible and it justifies doing nothing.
To work through which one your open roles actually represent, get in touch.
