The Studies Behind the Budget-Overrun Numbers
Almost every cost-overrun statistic in circulation traces back to a handful of studies. This is the directory: each one summarised with its headline figure, its sample, and a link to the full citation page so you can quote the number with the source attached.
IT and software project studies
The evidence base for "most IT projects overrun", from the Standish CHAOS Report to the McKinsey-Oxford large-project study.
Standish CHAOS Report
The most-cited source on IT project outcomes, published since 1994. CHAOS 2020: 31% successful, 50% challenged, 19% failed. Success falls sharply with project size (6% for the largest vs 61% for small).
Standish Group CHAOS 2020 / 2015
McKinsey-Oxford: Large IT Projects
5,400-plus large IT projects (budgets above $15M), studied with the University of Oxford in 2012. On average 45% over budget, 7% over time, and delivering 56% less value than predicted.
McKinsey & University of Oxford, 2012
Flyvbjerg & Budzier: IT Black Swans
1,471 IT projects in the Harvard Business Review, 2011. The 27% average overrun hides a fat tail: one in six projects is a black swan with a 200% cost overrun and a near-70% schedule overrun.
Flyvbjerg & Budzier, HBR, 2011
NASA: Requirements vs Overrun Curve
NASA project data links requirements effort to outcome: projects that spend only 5-10% of effort on requirements run about 50% over budget, while those spending 20% run only about 10% over.
NASA project studies
Megaproject and infrastructure studies
Bent Flyvbjerg's work at Oxford: the 16,000-project database that underpins the Iron Law of Megaprojects, and the original 258-project transport study that started it.
Flyvbjerg Project Database
The largest dataset of large-project outcomes in the world, 16,000-plus projects assembled at Oxford. 47.9% come in on budget, only 8.5% on budget and on time, only 0.5% on all three. The empirical basis for reference class forecasting.
Flyvbjerg & Gardner, How Big Things Get Done, 2023
Flyvbjerg, Holm & Buhl 2002
The original 258-project transport study across 20 nations, 1927-1998: cost was underestimated in 86% of projects. Rail overran by 44.7% on average, fixed links 33.8%, and roads 20.4%.
Flyvbjerg, Holm & Buhl, 2002
Government, defense and survey studies
The public-sector benchmarks: PMI's cross-industry survey and the US GAO's annual assessment of major weapon programmes.
PMI Pulse of the Profession
The source for the widely-quoted 43% over-budget figure: a global, cross-industry, self-reported survey of project professionals across all project sizes and sectors.
PMI Pulse of the Profession, 2018
GAO Weapon Systems Annual Assessment
GAO-26-108457 (2026): 104 major US defense acquisition programmes worth over $2.4 trillion, with average time to initial capability now over 12 years. The prior-year GAO-25-107569 assessment is covered separately.
US GAO, GAO-26-108457, 2026
Related on this site
- Budget overrun statistics hub (all figures side by side, by sector)
- The Iron Law of Megaprojects
- Reference class forecasting (Flyvbjerg's remedy)
- The “53,000 tech projects” statistic, traced