Cost Estimates as Predictors of Actual Weapon Costs: A Study of Major Hardware Articles

摘要

This is a statistical study of military cost estimates aimed at decreasing the uncertainties about their interpretation and use. Based on a sample of 68 cost estimates of major hardware articles in 22 weapon systems, it assesses their accuracy as predictors of actual costs. In section 1 the aims and scope of the study are given along with a detailed summary of methods and results. Section 2 presents the statistical analysis of estimating errors, and section 3 shows quantitatively that the observed errors can significantly affect the choice between alternative systems. The grosser differences between estimated and actual costs disappear when the estimated costs are adjusted (1) to refer to the actual quantities of the items procured, and (2) to take account of the secular change in the level of prices. But even these 'adjusted' estimates exhibit great variability; in the samples studied they range from 15 per cent to about 150 per cent of actual costs. They are also systematically biased, about fourfifths of the adjusted estimates being below actual costs. The study identifies the situations in which variability is likely to be large, and presents a numerical method for 'debiasing' the estimates so that (although still variable) they are no more likely to be low than high. The important parameters are (1) the time the estimate is made in relation to the development program, (2) the degree of technological advance required, and (3) the length of the development period. (Author)