Anyone who has sat through a workforce strategy presentation in the last three years has seen the same slide.
Sixty-three percent of employers say skills gaps are the single biggest barrier to transforming their business. It comes from the World Economic Forum's Future of Jobs Report 2025, which surveyed over a thousand employers representing more than fourteen million workers, and the figure is real.
What the slide never includes is the twenty years of labour economics arguing that employers systematically misdiagnose what they are looking at.
We sell training. It would be straightforwardly in our interest to skip that part. But a cluster of articles that opened by repeating the vendor line would be worth nothing to the person who has to sign off the budget, so here is the argument against our own product first.
The number everyone quotes
Three figures from the same report travel together.
Thirty-nine percent of key skills required in the job market are expected to change by 2030.
Eighty-five percent of employers plan to prioritise upskilling their existing workforce.
Sixty-three percent name skills gaps as the primary barrier to business transformation, ranking above organisational culture, outdated regulation and shortage of capital.
That is a survey of what employers believe. It is good evidence of employer belief. It is not, on its own, evidence about the labour market, and the distinction matters more than it sounds.
A survey of employers tells you what employers think the problem is. It cannot tell you whether they are right.
The critique that number has never answered
The clearest version of the argument comes from Peter Cappelli at Wharton, who has spent much of his career examining employer claims of skills shortages and generally not finding them.
The core observation is an economic one. In a genuine shortage, a scarce input gets more expensive. If employers cannot find machinists, the price of machinists rises until the market clears. When employers report that they cannot fill roles while the wages for those roles are flat or falling in real terms, something other than scarcity is going on.
The candidates for that something else are unglamorous and well documented.
The pay is below the market. The role is fillable at a higher number and the budget was set at a lower one.
The requirements are inflated. Degree requirements added to roles that did not previously carry them, or job specifications listing a decade of experience in a technology that has existed for six years. A substantial body of work on degree inflation has documented employers requiring bachelor's degrees for roles where most current jobholders do not have one.
The hiring process loses people. Automated screening that rejects qualified applicants on keyword mismatches, six-stage interview loops, and a time-to-decision long enough that candidates take other offers.
The working conditions are the problem. Shift patterns, location, or a reputation that circulates among people who would otherwise apply.
None of these is fixed by training, and all of them present to an employer as "we cannot find people with the right skills".
Four things that get called a skills gap
It is worth separating them, because each has a different fix and only one of them is a training problem.
| What it actually is | How it presents | What fixes it |
|---|---|---|
| A price problem | Vacancies stay open; comparable employers fill the same role | Pay, or a changed scope |
| A pipeline problem | Nobody anywhere has the skill because the field is new or training capacity is too small | Training, and usually years of it |
| A process problem | Applications arrive and do not convert; offers get declined | Fixing hiring, not training |
| An internal mobility problem | The capability exists inside the company and nobody can find or move it | Skills visibility and redeployment |
The second row is the one where a training budget does work, and it is not rare. It is simply much rarer than the way the phrase gets used.
The diagnostic, which takes about a week
Before commissioning a program, five questions separate the rows above. None of them requires a consultant.
What has happened to real pay for this role over three years? Rising sharply and still unfilled points to a genuine supply constraint. Flat in real terms points at the price problem.
Are competitors filling it? If three companies in the same city and the same sector are hiring the identical role successfully, the constraint is not the labour market.
Where in the funnel do candidates disappear? Applications to screen, screen to interview, interview to offer, offer to acceptance. A large drop at the last step is not a skills problem at all. A large drop at the first is often the screening rules rather than the applicants.
Does the job description describe the job? Take the requirements list to three people currently doing the role well and ask which ones they did not have on their first day. The answers are usually uncomfortable.
How many people already inside the company are one training step away? This is the question that most often goes unasked, and the one where the answer is most often surprising.
Why we ask partners this before scoping anything
Astra Trainer builds training programs with industry, government and institutional partners across ten domains, from AI and biotechnology to semiconductors, energy and advanced materials. The scoping conversation starts with which roles are unfilled and why, because a program aimed at a price problem will fail no matter how good the content is, and it will fail slowly and expensively.
Where the constraint really is capability, the shape of the answer is usually internal: people already employed, already holding the context, one structured track away from the role you cannot hire for. You can see the ten domains and how programs are scoped here.
Where the shortage is real
The critique above is not a claim that shortages never exist. In several fields the supply constraint is measurable and specific, and the semiconductor industry is the cleanest example available.
The Semiconductor Industry Association and Oxford Economics projected the US semiconductor workforce growing by roughly 115,000 jobs by 2030, from about 345,000 to about 460,000. At current degree completion rates, roughly 67,000 of those, or 58 percent, risk going unfilled.
That estimate is built from graduation pipelines rather than from asking employers how they feel, which is what makes it a different kind of number.
The composition matters more than the headline. Of the roles at risk, 39 percent are technician positions requiring a two-year degree or less. Thirty-five percent need a four-year degree. Twenty-six percent need a master's or a doctorate.
So the largest single block of the gap is the part that does not require a four-year degree at all, which is precisely the part that training can address on a timescale shorter than a decade.
Similar structural evidence exists elsewhere. In US construction, the Associated General Contractors of America and NCCER surveyed around 1,400 firms in 2025 and found 92 percent of those hiring reported difficulty finding qualified workers, with 45 percent already running projects late as a result. Notably, 57 percent named a lack of skills and qualifications, while 48 percent named new hires failing to show up or quickly quitting, which is a retention and conditions problem sitting inside the same survey.
That survey is honest about carrying both kinds of problem at once, and most do.
The figure that went down
Here is the part that gets left out of nearly every deck built on this report.
The 39 percent skills-disruption estimate for 2025 to 2030 is down from 44 percent in the 2023 edition.
The number describing how much of the skill base is about to be upended fell by five percentage points between editions. That is not nothing, and it does not fit a narrative of accelerating crisis.
Several readings are available. Employers may have absorbed more change than they expected. Estimates made at the peak of a technology cycle may have been high. Survey composition changes between editions. The honest position is that the direction of travel is downward and the reason is not settled.
A test worth applying to anyone selling you workforce training, including us. Ask whether they know the 2023 figure. A vendor quoting 39 percent as evidence of an accelerating crisis, without knowing it came down from 44, has not read the report they are citing. That tells you something about the rest of their claims.
What this means for a training budget
Four practical consequences.
Run the diagnostic before the procurement. A week spent separating price, pipeline, process and mobility will save a year of a program aimed at the wrong thing. Training bought to solve a pay problem produces trained employees who leave for the employer paying market rate.
Look inside first. Where the constraint is genuinely capability, the existing workforce is usually the cheapest and fastest source, and they arrive already holding the domain context, the relationships and the tacit knowledge a new hire spends a year acquiring. Eighty-five percent of employers say they intend to prioritise this. Far fewer have a mechanism for it.
Do not let training paper over hiring. If the funnel loses good candidates at the offer stage, no program fixes that, and the training budget becomes a way of not having the harder conversation.
Where the pipeline problem is real, commit on the right timescale. Structural shortages of the semiconductor kind are not closed in a quarter. They are closed by sustained programs, and by starting with the technician layer where the gap is largest and the training distance is shortest.
What to take from this
The 63 percent figure is real and it measures employer belief rather than labour supply.
Four different problems share the name "skills gap" and only one of them is solved by training. Separating them takes about a week and it is the highest-return week in the whole process.
Real shortages do exist, and the good evidence for them comes from pipeline arithmetic rather than from employer sentiment. Where it exists, the gap is often largest in the roles that need the least formal education.
And the number that describes the coming disruption fell between the last two editions of the report everyone cites. Anyone who tells you the crisis is accelerating, while quoting that report, has not finished reading it.
Is the skills gap real or not?
Both, in different places. Employer-reported shortages frequently describe pay, hiring-process or working-condition problems. Structural shortages measured from training pipelines rather than employer sentiment, as in semiconductors, are real and specific. The work is telling which one you have.
How do I tell a pay problem from a skills problem?
Look at what real pay for the role has done over three years and whether comparable employers are filling it. Unfilled roles alongside flat real wages, while competitors hire successfully, point away from scarcity.
Is it cheaper to train existing staff or hire?
It depends on the gap size and the role, and the comparison is covered in detail in a companion article. The factor most often left out is that internal candidates already hold domain context a new hire spends months acquiring.
Why does the 39 percent figure matter?
Because it is down from 44 percent in the 2023 edition of the same report. Quoting the current figure as evidence of accelerating disruption misrepresents its own source.
Where do I start if the shortage is genuine?
With the layer where the gap is largest and the training distance is shortest, which is usually technicians rather than doctorates. Astra Trainer scopes programs with partners across ten industrial domains, and you can see how that works here.