There is an element of chance to any individual project finishing on budget. It would therefore be fair to argue that imperfect cost estimates simply reflect the difficulty of the cost estimation task - if it were true that cost underruns were anywhere near as common or as large as cost overruns.
But overruns are much more common and much larger than underruns. The reasons are not a mystery - as this report has shown, large projects are more prone to overruns and to larger overruns; also prematurely announced projects are more prone to overruns, which are often larger too.
The following sections point to two other predictors of overruns: broader market conditions, and contract type.