Spare transport links

Why does a region maintain a road almost nobody uses until the day the main one closes?

Every region has one: the old road over the ridge, the single-track line through the hills, the second bridge two valleys along. It carries a trickle of traffic. It costs real money to grade, drain and clear of snow. Any honest count of vehicles per day says it is not earning its keep.

And then one morning the main road is shut — a slip, a fire, a collapsed span — and that road carries everything. So here is the question worth putting carefully. When we say the spare road is "hardly used", are we measuring the thing that actually determines its value?

Consider what happens the moment the main link closes. Traffic does not stop existing; it looks for another way. If a reasonable alternative exists, journeys get longer and dearer for a while and life continues. If none exists, the region is cut — and the cost is not a longer drive but a severed one: goods that cannot arrive, patients who cannot reach a hospital, workers who cannot reach work.

Notice that those two outcomes differ enormously in cost, and that the difference is created entirely by the road nobody was using. So its value is not what it carries on an ordinary Tuesday. Its value is the size of the disaster it prevents on the rare day the main link fails.

That means the value is a product of two numbers: how often the main link fails, and how bad the failure is without an alternative. Both are awkward. Failures are rare, so the frequency is estimated from thin data. And the severity is a counterfactual — the cost of something that did not happen — which never appears in any account.

Do you see why this is systematically undercounted? A traffic survey measures what happens; the spare road's benefit is defined by what does not. When Atlanta lost a section of Interstate 85 to a fire in 2017, or Genoa lost the Morandi bridge in 2018, the surrounding street and road network absorbed what it could — and what it absorbed was precisely the value of capacity that had looked surplus the week before. Neither event improved the case for that capacity in any spreadsheet, because the benefit arrived as an absence of catastrophe.

Now push on the reasoning, because there is a real cost too. The spare road is not free: it is graded, drained, patrolled, and cleared. Every pound spent on it is a pound not spent on the busy link. So the question is not "is redundancy good" — it is "how much, and where". And the answer turns on structure, not sentiment. A link whose removal disconnects a region entirely is worth far more as a backup than a link that merely offers a shortcut between two places already joined three other ways. Network people call the first kind a cut edge, and a region with many of them is fragile whatever its total road mileage says.

There is a failure mode worth naming honestly, and it is not the one people expect. The danger is rarely that a region builds too much spare capacity; it is that the spare link decays. Since it carries little traffic, it scores badly on every renewal list, so its bridges go unstrengthened and its surface unrepaired — and it may fail the weight or width test on exactly the day it is needed for lorries. A backup that has quietly degraded is worse than none, because it was counted on.

And an honest caveat to hold against all this: adding links is not always beneficial. Braess's paradox shows that in a congested network an extra road can, under particular conditions, make everyone's journey slower by changing how traffic distributes. Redundancy for resilience and capacity for congestion are different arguments, and conflating them produces bad decisions in both directions.

The refinement that ties this together is the shift from average to conditional value. Ordinary transport appraisal asks what a link contributes to typical journeys; resilience appraisal asks what it contributes given that something else has failed. Those two questions can rank the same set of roads in almost opposite orders, which is why a link that looks indefensible on the first can be essential on the second.

A misconception worth correcting gently is that resilience means duplication — a second road parallel to the first. Often it does not, and often that is the worst version, because two links crossing the same unstable slope or the same flood plain fail together. What matters is that the alternative fails independently: a different valley, a different mode, a different bridge over a different reach. Redundancy that shares a hazard with the thing it backs up is not redundancy at all.

And it follows that maintaining the spare road is part of the policy, not a separate budget line. The commitment is not to own an alternative but to own a usable one, which means periodic proof that it still takes the loads and widths the emergency would demand.

• How network analysis identifies cut edges and articulation points, so fragility can be located before it is demonstrated.
• Whether resilience is better bought as a spare road, a spare mode, or a stockpile.
• Braess's paradox, and why adding capacity to a congested network sometimes slows it down.

Every term the collection defines is gathered in the glossary.