A Caltech Library Service

Sample selection and const underestimation bias in pioneer projects

Quirk, James P. and Terasawa, Katsuaki (1986) Sample selection and const underestimation bias in pioneer projects. Land Economics, 62 (2). pp. 192-200. ISSN 0023-7639. doi:10.2307/3146337.

Full text is not posted in this repository. Consult Related URLs below.

Use this Persistent URL to link to this item:


The recent growth of the defense budget has been accompanied by a heightened public awareness of the existence of massive cost overruns in defense procurement. The literature of defense economics has tended to center in on the lack of adequate incentives to keep defense contracting costs under control, given the frequent use of sole source contracts coupled with cost plus a fixed fee or renegotiable fixed price financing arrangements (Cummins 1977; Weitzman 1980; Peck and Scherer 1962; and Terasawa, Quirk, and Womar 1983). But it is not only in defense contracting that cost overruns have occurred. What has received less attention is the by now well-documented fact that cost overruns are almost as pervasive and almost as massive in many recent privately financed construction projects, particularly "pioneer" or "first of a kind" projects (Merrow et al. 1979, 1981; Montgomery and Quirk 1978; Burness, Montgomery, and Quirk 1980; and Quirk and Terasawa 1982). What pioneer projects have in common with many defense contracts is that there is a high degree of uncertainty as to the technological and economic parameters of the projects. Under such circumstances, it is understandable that cost estimates for such projects would be unreliable, but it remains to explain why the cost estimates are not only unreliable, but are also biased in a downward direction. The approach adopted in this paper is to abstract from the inefficiencies arising from principal-agent problems, and instead to examine the role of cost estimation in private profit-oriented project decision making as one possible source of the observed cost underestimation bias in pioneer projects. What is argued here is that a truly unbiased cost estimation procedure would generate data consistent with an observed cost underestimation bias. This arises because cost estimates are not only estimates of the costs of completed projects, but are also used by decision makers involved in the planning and overseeing of a project. A selection bias is introduced into comparisons between observed cost estimates and observed final costs of projects, because certain projects are rejected or abandoned on the basis of cost estimates. Under a monotonicity condition, we show that this pattern of sample selection leads to an observed cost underestimation bias in that decision makers choose to build precisely those projects with the best chance of exhibiting cost overruns. A strengthening of the monotonicity condition leads to the further conclusion that the observed underestimation bias increases with the riskiness of the project, again assuming an unbiased estimation procedure.

Item Type:Article
Related URLs:
URLURL TypeDescription ItemWorking Paper
Additional Information:© 1986 by the Board of Regents of the University of Wisconsin System. This work was performed by the Arroyo Center of the Jet Propulsion Laboratory, which conducts research for the United States Army through agreement with the National Aeronautics and Space Administration. The views expressed in this paper are those of the authors and do not necessarily represent those of the Arroyo Center JPL, NASA, or the U.S. Army. We would like to thank Kenneth Arrow, Dave Grether, Linda Cohen, Jennifer Reingaum, Louis Wilde, and an anonymous referee for comments on an earlier version of this paper. Earlier work on this topic was funded by a grant from Exxon at the Environmental Quality Laboratory at Caltech. Formerly SSWP 512.
Group:Environmental Quality Laboratory
Funding AgencyGrant Number
Environmental Quality LaboratoryUNSPECIFIED
Subject Keywords:Cost estimates, Estimation bias, Economic costs, Sampling methods, Estimated cost to complete, Cost efficiency, Sampling bias, Unit costs, Capital costs, Statistical estimation
Issue or Number:2
Record Number:CaltechAUTHORS:20171114-141649753
Persistent URL:
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:83196
Deposited By: Jacquelyn Bussone
Deposited On:15 Nov 2017 23:20
Last Modified:15 Nov 2021 19:56

Repository Staff Only: item control page