Infrastructure Megaproject Overruns: Per-Mode Data
Infrastructure is the most-studied corner of overrun research. Rail, road, bridge, and tunnel mean overruns from the Flyvbjerg dataset, with the worst cases per category.
Mean overrun by mode
| Mode | Mean cost overrun | Sample size | Worst case in sample |
|---|---|---|---|
| Rail | +39% | ~258 projects | Channel Tunnel (+80% real) |
| Bridges and tunnels | +34% | ~167 | Big Dig (+275%) |
| Roads | +24% | ~806 | Various +200%+ outliers |
| Dams | +96% | ~243 | Itaipu (+240%) |
| Power transmission | +8% | ~50 | +30-50% outliers |
| Oil and gas | +34% | ~318 | Kashagan (+400%) |
| Nuclear | +120% | ~180 | Olkiluoto-3 (+300%) |
| Solar | +1% | ~39 | +25% outliers |
| Wind | +13% | ~76 | Block Island (+40%) |
Source: Flyvbjerg and Gardner (2024). Sample sizes are approximate; new projects are added each year.
Why some modes overrun far more than others
Flyvbjerg's explanation, which he calls the "Iron Law of Megaprojects", is that overruns concentrate in modes where the technology is bespoke (nuclear, novel rail, dams), the political stakes are high (rail, Olympics), and the unit-of-work is large enough to make learning curves shallow. Modes with modular, repeatable units (solar, wind, transmission) overrun far less because each project is closer to the last.
The implication for planners: if you can shift a megaproject toward a modular, repeatable design, the cost-overrun risk drops by 4x to 10x. This is the argument behind small modular reactors (vs traditional nuclear), modular substation manufacturing, and pre-cast bridge segments.