The Opera House was 1,357 percent over budget for a reason that isn't about honesty
Sydney's estimate wasn't a lie. It was built the way almost everyone builds forecasts — and almost everyone gets it wrong the same way.
In 1957, the projected cost of the Sydney Opera House was £3.5 million, with a completion date set for Australia Day, January 26, 1963. It opened in 1973 — a decade late — having cost roughly $102 million, more than thirteen times the original estimate in real terms.
The usual story blames the architect, Jørn Utzon, who resigned partway through amid political pressure, or blames the famous shells of the roof, which required years of engineering just to figure out how to build at all. Both are true and neither is the interesting part. Daniel Kahneman later pointed to a more boring, more universal failure mode: planners almost always estimate a project by imagining how this specific project will go, drawing on their own optimism, their own hopes for a clean execution, and their own blindness to the problems they haven't hit yet. He called this the inside view.
The alternative he proposed, reference class forecasting, ignores your specific plan entirely and instead asks a colder question: how did every past project that looked roughly like this one actually turn out? Studied broadly, the answer is almost always worse and slower than anyone building the estimate expected — cost engineer Bent Flyvbjerg's research on large infrastructure projects across decades and countries found overruns to be closer to the rule than the exception.
The estimate for the Opera House wasn't a con. It was one team confidently imagining their own best case, with nothing forcing them to check it against everyone else's.