We have forgotten how to build
Swedish construction produces less per hour worked than it did 25 years ago. No other major sector has gone backwards. It is not the craft that disappeared; it is the paperwork that grew.
Swedish construction produces less per hour worked than it did 25 years ago. No other major sector has gone backwards. It is not the craft that disappeared; it is the paperwork that grew.
The evidence
It went backwards. Labour productivity in the Swedish construction sector fell 4 percent over the past 25 years, while the technical consultancies that design what construction builds raised theirs by about 40 percent. The figures come from economist Fredrik Bergström's report for Innovationsföretagen, 1 200 miljarder skäl att tänka nytt, written as Sweden prepared to spend 1,200 billion kronor on transport infrastructure through 2037.
Byggsektorns produktivitet har minskat med 4 procent på 25 år, medan teknikkonsultbranschen ökat sin produktivitet med cirka 40 procent.
One caveat belongs up front. Byggföretagen, the contractors' association, disputes the methodology behind sector-wide productivity measures, arguing they undercount quality improvements and stricter building regulation. Measure the whole sector differently and the minus sign can flip. But narrow the question to infrastructure and the dispute falls away. Rail investment prices have grown at nearly twice the rate of inflation every year since 1990, and a kilometre of new Stockholm metro costs 7.3 times more in real terms than in the 1970s, per the same authors' follow-up report Skenande kostnader. Whatever the right sector-wide number is, the part of construction that builds roads and railways has plainly become slower and more expensive per unit built.
The world
No. McKinsey Global Institute has tracked the same pattern worldwide: construction productivity grew about 0.4 percent a year between 2000 and 2022, roughly 10 percent in total, while the world economy grew about 50 percent and manufacturing managed around 3 percent a year. McKinsey estimates that if construction merely caught up with the economy as a whole, the sector would create an additional 1.6 trillion dollars of value per year. That is an estimate of forgone value, not money anyone has counted; the realised numbers are bad enough.
The output end confirms it. Bent Flyvbjerg and colleagues assembled a dataset of 258 transport infrastructure projects across 20 nations: nine out of ten came in over budget, and rail projects overran by 44.7 percent on average. The sample spans seventy years, from the 1920s to the 2000s, and the authors found no improvement over time. We are not learning.
| Measure | Figure | Source |
|---|---|---|
| Swedish construction productivity, 25 years | −4% | Innovationsföretagen (2025) |
| Swedish technical consultancies, same period | about +40% | Innovationsföretagen (2025) |
| Global construction productivity, 2000–2022 | ~0.4% per year | McKinsey (2024) |
| Global manufacturing productivity | ~3% per year | McKinsey (2017) |
| Transport projects over budget | 9 of 10 | Flyvbjerg et al., n = 258 |
| Average rail cost overrun | 44.7% | Flyvbjerg et al. (2004) |
The villain
No. Excavators dig faster than in 1999. Tunnel boring machines, prefabrication and GPS machine control have improved the physical toolkit the way every other industry's has improved. And the people on site and at the drawing board are better educated than any generation before them. The productivity decline did not happen in the concrete. It happened around it.
What changed is the volume of documents a project must produce, review and defend. Plans, investigations, impact assessments, technical descriptions, review rounds, permit applications: the written shell around a construction project has grown for thirty years, while the human capacity to read and write it has stayed where it was. The rulebook outgrew the humans.
The time cost is measured. FMI and PlanGrid surveyed 599 construction professionals in 2018 and found they spend 35 percent of their working time, more than 14 hours a week, on non-productive activities: 5.5 hours a week looking for project information, 4.7 hours resolving conflicts, 3.9 hours dealing with mistakes and rework. In the United States alone, that is 177.5 billion dollars a year in labour cost.
None of this is an argument against scrutiny. Environmental review, independent checking and public consultation exist for good reasons, and cutting them would be the wrong lesson. The problem is narrower and more fixable: every one of those pages is still produced and checked by hand, at human speed, and the pages keep multiplying while the hands do not.
The turn
The diagnosis has gone mainstream. Marc Andreessen's 2020 essay "It's Time to Build" made the inability to build a central failure of the West. Ezra Klein and Derek Thompson's Abundance (2025) turned it into a best-selling political programme, aimed squarely at permitting, review and proceduralism: the same forces documented above. Both end where this essay begins. The constraint on building is neither money nor engineering skill.
The constraint is documents, and that matters now for one reason: documents have become solvable. Machines can read, cross-reference and draft regulated text, which means the written shell around a project no longer has to move at human reading speed. The point is not less scrutiny. It is cheaper and faster iterations inside the same scrutiny, so that the years between idea and construction stop compounding into cost. Bergström's report projects that Sweden alone could free up 225 billion kronor for infrastructure over the coming twelve years if productivity recovered: a projection, and labelled as one. The direction, though, is not projected but recorded. Twenty-five years of minus, in the one sector everything else stands on. We did not forget how to build. We buried the knowledge under paper, and paper is no longer the hard part.
Sources
Yesper is the AI civil engineer for construction and infrastructure. AFRY, COWI, NRC Group and other Nordic firms use it to halve the time on a study, rerun calculations in minutes, and catch errors that would otherwise slip through. Get in touch if you'd like to see what it can do for you.
Book demo
