One construction start before 2037

One of 26 new projects gets a construction start

One. In Trafikverket's proposal for the national plan 2026–2037, 26 new named objects enter the system, and a single one of them gets a construction start before 2037. The rest spend the plan period in planning. Trafikverket's own list runs to 27 additions, but one is an extension of an existing investment, a second icebreaker. The government has since fixed the plan and adjusted the list of objects; the figure describes the proposal.

The reason is not that anyone builds slowly. It is the chain every new road or railway must travel before construction: a route and measure study (åtgärdsvalsstudie), a formal road or railway plan, environmental permitting, public review, government approval (fastställelse), an appeal window, and only then construction documents (bygghandling). Each step carries its own consultations and its own openings for appeal. In the planning calendar, the chain has a standard name: years 7 to 12.

"Years 7–12" is a euphemism worth pausing on. It concedes, as a scheduling constant, that half of a twelve-year plan passes between the decision to build and the first machine arriving. It also explains where the new money goes: in the proposal, the lion's share of the roughly 300 billion kronor of new room is absorbed by maintenance debt and by cost increases in projects already under way.

Underestimating the cost is how a project gets in

Because it works. Riksrevisionen, Sweden's national audit office, described entry into the national plan as a nålsöga, the eye of a needle, as early as 2010: once a project is in the plan, it will in practice be built, whatever happens to its cost. The government almost never stops an object because it has grown more expensive.

The consequence is predictable, and the auditors spelled it out. Whoever wants a project built has a strong incentive to underestimate its cost early and let the realistic figure surface after the decision is taken. The optimism in early estimates is not an arithmetic error. In a system without re-examination, it is a rational strategy.

The same audits measured what accumulates while a project waits. The 2011 review of large railway investments found 24 percent average cost growth excluding price development: additions and standard creep, as requirements on environment, safety and design pile onto a project during its years in the plan, and no one holds a mandate to say no.

The cost grows in the planning years

In the planning years, measurably. Riksrevisionen's 2021 audit followed the objects that sat in both the 2014–2025 and the 2018–2029 plans: their estimated costs grew by 58 billion kronor in four years, 39 percent per project on average, before construction began. In the build phase, the auditors found no systematic deviations at all. The contractors deliver roughly on budget; it is the material they are handed that has already swollen. We traced that pattern, and where the money goes, in one metre of metro, four Volvos.

The drivers are concrete. Municipalities and regions hold real veto power over a project, through local plans, appeal rights and co-financing, and can delay or stop it. Trafikverket buys itself free through adaptations: new on- and off-ramps, stretches moved into tunnel to spare residents, wildlife crossings. The cost-adding extras are often conceded in advance, to avoid future conflict. Stockholms Handelskammare, working with Trafikanalys data, confirms the pattern for the 2022–2033 plan: most of the average 76 percent overrun arises in planning and design.

Trafikanalys, the state's transport analysis agency, follows the cost development of named objects across plan revisions and sees the same mechanics from another angle: the objects that have sat in the plan longest carry the largest increases. Each additional plan period triggers a round of omtag, do-overs: redesign against new requirements, new investigations, new rounds of consultation. Waiting is not a neutral state. It is a cost engine.

Trafikverket doubled its headcount, not its grip on cost

Not a starved agency, at least not of people. Trafikverket has roughly doubled since it was formed in 2010, from about 6,000 to about 11,000 full-time equivalents. Over the same period, Stockholms Handelskammare calculates, the investment and maintenance volume it delivers per employee fell from 11 to 9 million kronor in fixed prices: nearly a fifth less output per person, from twice the organisation. The agency's own internal review from 2024 points the same way: it recommends clearing out a large volume of internal governing documents that no law requires.

And the follow-up? Riksrevisionen notes that cost estimates were not even systematically documented before 2019. The routine since then: when a project closes, its calculations are gathered into a shared Excel document on a shared drive. No systematic evaluation is made of what causes the escalations, and no consolidated account is published to parliament or the public. The state's largest investment programme keeps its cost history roughly the way a small firm keeps petty cash. Roughly 450 consolidated impact assessments were produced ahead of the 2022 plan revision, about half by consultants. The analysis exists. The follow-up of it does not.

The fix is cheaper analysis, not less scrutiny

Not less scrutiny. The consultations, the environmental review and the right to appeal are the democratic content of the chain, and nothing above is an argument for trimming them. What the audit trail shows is different: the cost sits in the years around those decision points, in documents produced by hand, in calculations that drift out of sync with their underlying material, and in do-overs so expensive that the system avoids them by locking decisions in early — which is exactly how the underestimates survive. A chain like this is not slow because people deliberate too much. It is slow because every pass through it is produced, checked and reconciled by hand.

Follow the five mechanisms to their common root, and what remains is a single sentence: