The conversation about construction's skills crisis is largely about supply: how to attract more young people into the trades, build more colleges, fund more apprenticeships, design new qualifications. This briefing argues the framing is wrong. The binding constraint is not the willingness of individuals to train. It is the capacity of the construction employer base to host and develop them.
Six findings, all from published sources:
The implication: more public spending on training individuals will not, on its own, close the skills gap. The constraint is one tier upstream. The system needs to invest in the conditions that allow employers, particularly SMEs, to participate.
of construction businesses employ apprentices. Only 31% offer any kind of work placement. The share funding any training at all has fallen from 57% to 49% since 2011.
DBT / CITB, March 2026; DfE, August 2025
Most of the public conversation about construction skills asks: how do we attract more people into the trades? That question has produced the £625m Construction Skills Package, foundation apprenticeships, technical excellence colleges, T-Levels, Skills Bootcamps, NEET trailblazers, sector-based work academy programmes, and a constant stream of careers campaigns.
This is necessary work. It is not sufficient. It rests on an unspoken assumption: that the binding constraint is the supply of people willing to train. The data does not support that assumption.
The right question is: who is going to host, supervise, and develop the people we train?
Construction needs approximately 250,000 entrants per year in England, around 200,000 to replace workers leaving and 47,860 net additional to meet baseline demand growth (CITB Construction Workforce Outlook 2025–29). On top of that, government ambitions for housing, retrofit and infrastructure imply an additional 110,000 per year by some estimates.
The formal training pipeline supplies a fraction of that. The most recent data points are:
Combined, the formal pipeline produces roughly 30,000–35,000 starts per year in England, about 12–14% of demand.
The remaining 86–88% comes through informal routes: direct hire, family connections, agency labour, self-employment, migration. Competence validation in these routes is partial. CITB's current published position is that c. 50% of new entrants come in via informal routes and only 11% via apprenticeships.
A finding that emerges as soon as you compare the three routes side by side: none of them publishes data on whether their graduates actually end up working in construction 12 or 24 months later.
| Route | Sector retention 12+ months published? |
|---|---|
| Apprenticeships (Construction) | No. Industry estimates only. CITB has stated "70% of those who start a construction apprenticeship never secure sustained employment in the industry" but the methodology is not transparent. |
| Skills Bootcamps (Construction) | No. Tracking ends at 6 months, and "positive outcome" includes destinations that are not new entries to the sector. |
| T-Levels (Construction) | No. Cohorts too small for robust analysis. |
The IDBR-linked sector-of-employment data already exists in the Department for Education's LEO dataset. It is not surfaced at the level needed to compare routes. This is itself a quiet structural issue: the public skills system funds training but does not robustly measure whether the people it trains stay in the sector that paid for them.
The headline figure in this briefing comes from the Department for Business and Trade's own consultation on Industry Training Board reform (March 2026), citing CITB:
Only 21% of construction businesses employ apprentices. Only 31% offer any kind of work placement.
A second figure from the Department for Education's August 2025 announcement on construction skills tells a parallel story: the share of construction firms funding or offering any training to their workers has fallen from 57% in 2011 to 49% in 2024.
In other words: roughly four-fifths of construction businesses are outside the apprenticeship system entirely, and engagement in training of any kind is going backwards, not forwards.
This is the bottleneck. Every pound of public investment in training individuals depends on someone, somewhere, hosting those individuals. Right now, that responsibility falls disproportionately on a shrinking minority of firms.
To understand why this matters, the structural facts of who builds Britain are worth stating plainly:
This is the population we are asking to absorb a 250,000-per-year recruitment requirement against rising insolvency, falling training engagement, and increasingly volatile workloads. Asked plainly: it is structurally unrealistic. A small subbie working contract-to-contract cannot reasonably commit to a 2–4 year apprentice contract when they cannot reliably predict whether they will be solvent in six months.
construction businesses have closed since 2020, the largest sectoral decline in the economy. Nearly 4,000 insolvencies in 2025/26 alone, the highest of any sector.
ONS; Insolvency Service quarterly statistics
Cost is one of four roughly equal barriers, and arguably not even the largest. The Federation of Master Builders' Small House Builders' Survey 2025 ranks the barriers SMEs themselves cite to taking on apprentices:
| Barrier | % of SME builders citing |
|---|---|
| Quality concerns about candidates | 51% |
| Cost | 39% |
| Dropout / retention risk | 39% |
| Admin burden | 34% |
But the financial dimension matters in a deeper way than the headline percentage suggests. It is what turns latent willingness into structural inability. The four drivers in detail:
Cash-flow fragility is structural, not chosen. Construction SMEs operate on margins of 1–3%, against payment behaviours imposed by upstream procurement. Three mechanisms compound:
A firm that cannot reliably predict whether it will be solvent in six months cannot reasonably commit to a 2–4 year apprentice contract. This is structural inability, not unwillingness, and it is largely imposed on SMEs by procurement and payment decisions made upstream.
The cost of hosting an apprentice is not the same as the cost of training one. Public funding (CITB grants, Apprenticeship Levy co-investment) covers most of the training cost. It does not cover:
Funded training is not the same as funded hosting. The two are routinely conflated in policy.
Capability gaps are real, not attitudinal. Most micro-firms have no HR function, no in-house training infrastructure, no compliance team. The 34% citing "admin burden" is rational allocation of scarce administrative capacity in a firm where the owner is also the supervisor, the estimator, and the lead operative.
System trust and pipeline confidence are weak. The 51% citing "quality concerns" and 39% citing "dropout risk" reflect SMEs' rational read of past experience: an apprentice who drops out at month 8 is a sunk cost the firm absorbs alone. SMEs that have been burned once are slower to engage again. This is the textbook market-failure pattern: the cost of dropout is privatised to the firm, while the social value of the training place (and the lost place when firms withdraw) is collective.
The Apprenticeship Levy was designed to drive employer training investment. It is a 0.5% payroll tax on employers with payroll over £3m, returnable as training spend. Construction's largest contractors pay both this and the CITB Levy.
Yet:
For construction specifically: most large Tier 1 contractors pay the levy in full but cannot fully spend it on apprentices, because the regulated apprenticeship products do not always match the labour they actually procure (much of which is sub-contracted, often through agencies or self-employment). Their SME supply chain, which would benefit from those funds via transfer, typically cannot absorb them. Three reasons:
The funding to train more people exists. It is being underspent, returned, or trapped in accounts that do not connect to the employers who could host the people it would train.
of Apprenticeship Levy returned unspent to the Treasury between 2019 and 2022, about £1.1bn per year, while construction has the highest share of "hard to fill" vacancies of any sector.
London Progression Collaboration / IPPR, FOI, 2023
The asymmetry between public investment in training individuals and public investment in employer participation is stark.
| Programme | Annual scale (England) |
|---|---|
| Apprenticeships budget | £3bn+ |
| Adult Skills Fund | c. £1.4bn |
| Skills Bootcamps | 60,000+ starts/year, all sectors |
| T-Levels (transitioning to V-Levels in 2027) | Growing toward 66,000 starts/year all subjects by 2029 |
| Construction Skills Package | £625m commitment (HMT, March 2025) |
| Construction Technical Excellence Colleges | 40,000 places |
| Sector-Based Work Academy Programmes | c. 30,000 starts/year |
| Foundation apprenticeships (under-25 SME training cost covered) | New, December 2025 |
The supply-side offer for individuals is rich, well-funded, and growing.
| Programme | Scale and conditionality |
|---|---|
| CITB apprenticeship grants | £8,500 over a 2-year apprenticeship. Grants and funding changing from January 2026; FMB has flagged the change as inadequately communicated and reducing support. |
| CITB qualification, work experience, short-course grants | 30–50% of cost, capped, means-rationed |
| CITB total annual levy income | c. £228m |
| Apprenticeship Levy 95% co-investment for non-levy SMEs | Available; 5% co-investment removed for new under-25 entrants from December 2025 |
| Apprenticeship Levy transfers (up to 50%) | Discretionary; brokerage thin |
| Build Academy pilot (CITB / FMB) | Small pilot, RMI sector |
| Shared Apprenticeship Services | CITB has invested £1m |
| CITB New Entrant Support Team (NEST) | Doubled in size; advice and brokerage |
The asymmetry is structural. Support for individuals is large, nationally branded, and visible. Support for employers to participate is thinner, more conditional, more bureaucratic, levy-recycled rather than additional, and currently being reduced.
This is not an accident. It is the logical outcome of a policy frame that treats skills as a problem of pipeline rather than a problem of host. If you believe the constraint is willingness of individuals, you spend on individuals. If the constraint is the employer base, you find the spend going to the wrong place.
A note on this section. The numbers below are a thought experiment, not a forecast. They chain several published baselines and estimates together to indicate the order of magnitude of what shifting the engagement constraint could deliver. They are intended to make the case that this lever is worth pulling, not to specify exactly what would happen.
If the share of construction businesses engaging in formal training rose by 20 percentage points, from 21% to 41%, supported by shared infrastructure that absorbs the capacity, capability and admin burden documented above, the modelled effect across the three routes is approximately:
| Today | Scenario | |
|---|---|---|
| Apprenticeship starts (England) | c. 19,600 | c. 39,600 |
| Skills Bootcamps construction starts | c. 3,250 | c. 9,000 |
| T-Level / V-Level construction starts | <2,000 | c. 5,000 |
| Total formal-route starts | c. 25,000 | c. 50,000–55,000 |
| Share of CITB net additional need (47,860/yr) | 52% | >100% |
Critically, the same intervention that drives the volume increase also addresses the attrition factors. Apprentices drop out when their employer fails. Bootcamp learners do best when employers are behind them. T-Level placements are limited by host availability. The volume gain and the retention gain are the same intervention from two angles. They compound.
Beyond the volume:
The point is not that the precise numbers are right. The point is that even on conservative assumptions, fixing the employer constraint is the largest single lever available. Continuing to spend on the supply side without addressing it is throwing learners at an absorption problem.
Five propositions follow from the evidence above.
The skills crisis is fundamentally an employer-engagement crisis, not a pipeline crisis. Public funding and training capacity for individuals significantly exceed the absorptive capacity of the construction employer base. The £1.1bn/year of unspent levy, the £625m Construction Skills Package, the foundation apprenticeships: none of it lands at scale without more employers willing and able to host.
Funded training is not the same as funded hosting. The cost of an apprentice's non-productive time, supervision, off-the-job training, and the risk of a 2–4 year commitment in a contract-by-contract business is not what CITB grants and Apprenticeship Levy co-investment cover. This gap, more than the cost of training itself, is where SMEs withdraw.
The financial pressure on SMEs is largely imposed from upstream. Payment retentions, late payment, contract instability, and insolvency exposure are conditions of operating in the supply chain, not choices. Reform of payment infrastructure (project bank accounts, prompt payment, retention reform) is therefore part of the skills agenda, not separate from it.
The shared infrastructure that would make hosting viable for small firms exists only in fragments. Pooled apprenticeship schemes, brokerage of levy transfers, shared HR and compliance, payroll and welfare support for new entrants: these are the levers that turn 21% engagement into 41%. They are mostly unfunded at scale.
The system needs to start measuring sector retention, not just starts. Apprenticeships, Bootcamps and T-Levels all stop measuring outcomes at 6 months and none distinguishes whether graduates remain in construction. The IDBR-linked LEO data exists. Surfacing it would let policy answer questions it currently cannot.
None of these propositions is novel in isolation. The construction industry, FMB, CITB, IPPR, the London Progression Collaboration and others have raised each of them. What is novel is the cumulative weight: when you read the data side by side, the conclusion that public investment has been mis-targeted at the supply side becomes hard to avoid.
The skills crisis will not be solved by training more individuals. It will be solved by building the employer base that can host them.