The numbers: why Europe can't build on budget

How often do infrastructure projects run over budget?

Nine times out of ten. The largest dataset in the field, Bent Flyvbjerg's study of 258 transport megaprojects across 20 countries and 70 years, found cost overruns in 9 of 10 projects: rail overran by 44.7 percent on average, bridges and tunnels by 34 percent, roads by 20 percent. The rate did not improve over the seven decades studied.

Swedish contract data says the same. VTI, the national road and transport research institute, analysed 776 Trafikverket contracts above 10 million kronor and put the probability of a cost overrun at 94 percent for rail contracts and 86 percent for road contracts.

This page compiles the load-bearing statistics on why European infrastructure runs over budget: how often, by how much, where the growth happens, what errors in the documents cost, and what has happened to productivity. Every figure carries its primary source and a label. Confirmed means a primary source with a published method. Contested means the methodology is disputed. Projection means a forecast, not an outcome.

How large are the overruns?

Size depends on where you place the measuring stick. Flyvbjerg's figures measure from the decision to build. Move the baseline earlier, to a project's first appearance in a national plan, and the numbers grow far larger, as the next section shows. Both measurements are honest; they answer different questions.

One figure deserves its label. McKinsey's widely quoted claim that 98 percent of megaprojects overrun by more than 30 percent comes from consulting research whose underlying data was never published. We list it because it circulates, not because it meets the bar the rest of this page sets.

When does the cost actually grow?

Before construction. Riksrevisionen, Sweden's national audit office, examined the investment objects in the national transport plans and found that projects grew 39 percent in estimated cost over four years of planning, 58 billion kronor across 142 objects, while completed projects landed close to their build-start estimates. Its follow-up audit found that projects still in planning grew 51 percent between the 2018 and 2022 plans, against 16 percent for projects under construction. The auditors' own summary:

De systematiska kostnadsökningarna förefaller främst inträffa under planeringsfasen; i byggfasen hittar Riksrevisionen inga systematiska kostnadsavvikelser.

Independent research in three countries agrees. Jonas Eliasson tracked 276 Swedish projects finished between 2004 and 2022: planning-stage escalation averaged 25 percent for road and 40 percent for rail, while the step from build decision to completion added 2 and 10 percent. In the Netherlands, Cantarelli and colleagues measured 5 percentage points of additional overrun per extra pre-construction year. In Norway, Welde and Odeck found 40 percent front-end escalation in road projects that then finished about 2 percent under their construction budgets.

A caution the sources themselves insist on: these are correlations, and the mechanism is lock-in, the early commitment to a single design that removes the incentive to keep searching for cheaper alternatives. Nothing here shows that rushing planning saves money. What it shows is where the growth is born. We walk through the mechanism in Cost overruns are born in planning.

What do errors in the documents cost?

The UK's Get It Right Initiative, an industry body founded to measure exactly this, puts the direct cost of avoidable error at around 5 percent of project value, and at 10 to 25 percent once indirect and latent costs are counted, roughly £21 billion per year in the UK alone. Its most quoted line:

The annual spend due to error is estimated to be around 7 times the total annual profit of the UK Construction Industry.

Survey research points at the documents as the origin. FMI and PlanGrid's 2018 study of about 600 US construction professionals attributed 48 percent of all rework to poor data and miscommunication, 31.3 billion dollars per year, and measured 35 percent of professionals' time, more than 14 hours per week, going to non-productive activities, 5.5 of those hours spent looking for information. Autodesk and FMI's 2021 follow-up estimated the global cost of bad data at 1.85 trillion dollars in 2020. The dispute statistics close the loop: Arcadis has named errors and omissions in contract documents the number one cause of construction disputes three years running, and HKA reports disputed sums averaging 33.4 percent of contract budgets.

An honest counterpoint: what you count decides the number. A 2026 study by Peter Love in the Journal of Construction Engineering and Management found formally logged field rework at only 0.38 percent of contract value. GIRI's figures count latent, indirect and unrecorded costs; Love's count what reaches the rework log. The two are consistent with each other, and with a large iceberg.

What does an RFI cost?

When a contractor cannot resolve an ambiguity in the documents, the formal remedy is a request for information. The reference dataset is the Navigant Construction Forum's 2013 study of 1,362 projects and 1.1 million RFIs: 796 RFIs on the average project, at 1,080 dollars each, roughly 860,000 dollars per project spent clarifying what the owner's own documents should have said. The median question waited 9.7 days for an answer. More than a fifth never received one.

Industry estimates today run higher, 2,000 to 3,000 dollars per RFI before delay and rework, but the 2013 figures remain the bulletproof anchor. A larger Autodesk sample published in 2024 confirmed the underlying rate of 9.9 RFIs per million dollars of construction: a decade of digitalisation has not moved the defect rate in the documents.

Has construction productivity really fallen?

In Swedish infrastructure, the direction is not seriously contested; the sector-wide series is. Innovationsföretagen's 2025 analysis has construction-sector productivity down 4 percent over 25 years while the engineering-consultancy branch improved by roughly 40 percent. Byggföretagen, the contractors' association, disputes the sector-wide methodology and cites a series showing construction up 62 percent between 1998 and 2021. We therefore label the sector-wide series contested and scope our own claims to infrastructure, where the underperformance is documented by the audits above.

Globally, McKinsey's series shows construction labour productivity growing about 0.4 percent per year since 2000, against roughly 2 percent for the whole economy and 3 percent for manufacturing, with 1.6 trillion dollars per year of unrealised value. Its 40 trillion dollar shortfall by 2040 is a projection, and labelled as one.

Prices tell the same story from the other side. Swedish rail investment prices have grown 4.0 percent per year since 1990 against CPI's 2.1, and a krona spent on infrastructure today buys a fifth less than it did in 2009. The cost of a metre of Stockholm metro has risen 7.3 times in real terms; we unpack that number in One metre of metro, four Volvos.

What do the numbers say about Sweden's 2026–2037 plan?

Sweden's national transport infrastructure plan for 2026–2037, fixed by the government on 24 April 2026, allocates 1,171 billion kronor. For the named investment objects in the plan proposal, Innovationsföretagen computed a weighted net present value ratio (nettonuvärdeskvot) of −0.3, built from Trafikverket's own published project appraisals. In plain terms: each invested krona is calculated to return monetized societal benefit worth 70 öre.

Read the figure precisely. It is monetized benefit, not cash flow to the treasury. It covers the named new-build objects only; maintenance and smaller trimming measures, roughly half the plan, are generally profitable and sit outside the ratio. The defensible reading is that the state's selection of megaprojects destroys calculated value, not that infrastructure investment does. The sharpest detail: the candidate projects that were rejected carry the same average ratio, −0.3, as those selected. We unpack the calculation in 70 öre på kronan.

The pattern across every section of this page is the same. The money disappears in the years of planning, review and documentation before construction starts, and the measured error costs sit in the documents themselves. That is why these numbers are worth watching: they will move when the way documents are produced and checked moves.