John Raftery (1993), in his Inaugural Lecture at the University of Greenwich, suggested that project cost estimates be presented in the form of cumulative probability functions (termed here 'Raftery curves') rather than the current practice of single-point estimates. This paper describes a method for the empirical construction of Raftery curves for tender price forecasts, which then is applied to ten previously published data sets gathered throughout the world. In comparing the resulting curves, the most consistent feature is shown to be the shift associated with to the number of bidders entering bids for contracts. This is examined both in terms of bias and consistency. Contrary to some previous studies, no evidence is found of any trends related to the value size of projects.