How does project size affect cost estimation error? Statistical artifacts and methodological challenges

Empirical studies differ in what they report as the underlying relation between project size and percent cost overrun. As a consequence, the studies also differ in their project management recommendations. We show that studies with a project size measure based on the actual cost systematically report an increase in percent cost overrun with increased project size, whereas studies with a project size measure based on the estimated cost report a decrease or no change in percent cost overrun with increased project size. The observed pattern is, we argue, to some extent a statistical artifact caused by imperfect correlation between the estimated and the actual cost. We conclude that the previous observational studies cannot be considered to provide reliable evidence in favor of an underlying project size related cost estimation bias. We discuss the limited, statistically robust evidence and the need for other types of studies.