The processes of public megaproject cost estimation: The inaccuracy of reference class forecasting

• DOI: http://doi.org/10.1111/faam.12210
• Date: 01 November 2019
• Published date: 01 November 2019
• Author: Tim Neerup Themsen

Received: 9 March 2018 Revised: 4 January 2019 Accepted: 1 April 2019

DOI: 10.1111/faam.12210

ORIGINAL ARTICLE

The processes of public megaproject

cost estimation: The inaccuracy of reference

class forecasting

Tim Neerup Themsen

Department of Accounting, Copenhagen

Business School, Frederiksberg,Denmark

Correspondence

TimNeerup Themsen, Department of Accounting,

CopenhagenBusiness School, Solbjerg Plads 3,

DK-2000Frederiksberg, Denmark.

Email:tnt.acc@cbs.dk

Abstract

Governments worldwide are introducing “reference class forecast-

ing” to improve the accuracy of megaproject cost estimation and

thus ultimately the ability to delivermegaprojects on budget without

altering the project specifications and/or changing the time sched-

ule. In contrast to current findings, which show that reference class

forecasting leads to more accurate project cost estimates by coun-

teractinghuman cognitive and organizational biases, this article indi-

cates the contrary, that reference class forecasting does not lead to

more accurate cost estimates. The article theorizes that reference

class forecasting fails to produce more accurate project cost esti-

mates because estimates are always a relational network effect of

human and nonhuman actors’ “biased” efforts to establish them. This

finding challenges the existing literature by pointing to a more com-

plexunderstanding of project cost estimation and biases. The finding

is based on a longitudinal case study of a 23.6-billion-kroner Danish

public megaproject, which failed to meet its objectives despite the

application of reference class forecasting.

KEYWORDS

actor–network theory, megaproject, project cost estimation, public

budgeting, reference class forecasting

1INTRODUCTION

On 29 January 2009, the Danish Parliament granted Rail Net Denmark, the organization that manages the Dan-

ish railway infrastructure, 23.6 billion kroner to execute a project of replacing all existingrailway-signaling systems,

called the Signalling Programme. Rail Net Denmark had estimated the costs and established the budget of the project

using a newly developed technique called “reference class forecasting,” which was supposed to generateaccurate cost

estimates. In the beginning, Rail Net Denmark executed the project according to plan. In November 2015, however,

the organization announced that the project had been more challenging than anticipated. Rail Net Denmark’s chief

Financial Acc & Man. 2019;35:337–352. wileyonlinelibrary.com/journal/faam c

2019 John Wiley & Sons Ltd 337

338 THEMSEN

executiveofficer (CEO) and chief financial officer (CFO), along with the Signalling Programme’sdirector, resigned their

positions shortly thereafter. Rail Net Denmark assured the government that the project would still be delivered on

time, cost, and specification, but during the following 2 years,the new management announced that the project’s entire

cost reserve of 4.3 billion kroner would be spent, that the project would be a minimum of 9 years delayedand that the

project’s ambitions would have to be reduced. Something had gone terribly wrong.

One of the biggest challenges of delivering a successful project lies in accuratelyestimating project costs and estab-

lishing an accurate budget. Surveys carried out across countries, sectors, and organizations have pointed this out for

years (e.g., PMI, 2017; PwC, 2014). The most commonly used technique to estimate project costs is called taking the

“inside view” (Lovallo & Kahneman, 2003). This technique relies on the best judgment of human beings using the orga-

nization’s own capabilities, experiences, and expectations. In recent years, however, research has shown that most

projects that take this kind of approach end up incurring cost overruns (Flyvbjerg, Holm, & Buhl, 2002). An alter-

native view has been proposed, which is called taking the “outside view” or “reference class forecasting” (Flyvbjerg,

2006; Lovallo & Kahneman, 2003). The outside view, as its proponents argue, bypasses human judgment—and asso-

ciated notions of human cognitive and organizational biases—by requiring that costs be estimated by comparing the

project with a similar class (in other words, a “reference class”) of already-completed projects. In severalgovernments,

as discussed below,the outside view of forecasting is now being implemented with the explicit purpose of curbing cost

overruns.

In this article, I examine the public megaproject known as the Signalling Programme. I show that, for this specific

project,taking the outside view of cost estimation—that is, using reference class forecasting—did notcounteract human

cognitive and organizational biases and did not lead to an accurate cost estimate. In making sense of this finding, I rely

on actor–network theory. This theory approaches practices like cost estimation as a continually generatedeffect of

the web of relations that brings the practices about (Law, 2009). It further approaches biases as “indispensable medi-

ations without which no knowledge would be produced” (Latour, 2000, p. 371). In relying on actor–network theory,

I thus approach biases differently than most studies, which tend to view them solely as undesirable forms of human

delusional optimism and deliberate deceptiveness (e.g., Flyvbjerg, 2006; Lovallo & Kahneman, 2003). In this article, I

show that biases become an unavoidablecondition for the fruitful articulation of cost estimates and, as a consequence,

that attempts to outright counteract them are futile, because biases, fundamentally,are a way that allows the project

to enter into view and become relevant.

This article contributes to the current literature by providing the first empirical account of a budgeting processes

using reference class forecasting. The current studies on reference class forecasting have primarily been, on the one

hand, conceptual papers focused on developing the concept and a procedure for its application (Flyvbjerg, 2006;

Lovallo & Kahneman, 2003), and, on the other hand, empirical papers focused on comparing project budgets with

actual project cost outcomes (e.g., Batselier & Vanhoucke, 2016; Flyvbjerg, 2008; Flyvbjerg, Chi-Keung, & Wing,

2016). These papers have found that reference class forecasting either ought to or actually has led to accurate cost

estimates—a finding this article challenges. The article further contributes to the scarce literature on public project

budgeting (see Anessi-Pessina, Barbera, Sicilia, & Steccolini, 2016, for a thorough literature review of public budget-

ing). The literature on public budgeting has primarily looked into budgeting’s changing role and features related to

broader reform processes (e.g., Brun-Martos & Lapsley, 2017; Edwards, Ezzamel, & Robson, 2005; Ezzamel, Robson,

& Stapleton, 2012; Muniesa & Linhardt, 2011) and not on “understanding budgeting per se,” including “the dynam-

ics of the budgeting process” (Anessi-Pessina et al., 2016, pp. 509–510) and even less on public project budgeting

processes.

The remainder of the article is structured as follows. The second section describes in more detail how project

cost estimation has traditionally been carried out in megaprojects. It also introduces reference class forecasting.

The third section elaborates on actor–network theory, whereas the fourth covers the article’s methodology. The

fifth section contains the analysis, the sixth discusses the findings, and the seventh—and last—section concludes the

article.