Breaking the Sisyphean loop: reconceptualizing the treatment of risk and uncertainty in transport projects

Abstract

Worldwide transport projects repeatedly exceed their budgets. This problem has received widespread attention over several decades, with little progress made to improve project cost performance. Consequently, this has led to the emergence of the following recurring question: How can public agencies responsible for delivering transport projects improve the estimation accuracy of their actual construction costs? We use the metaphor of the Sisyphean task to assist us in suggesting that the conflation of risk and uncertainty results in an inability to estimate actual construction costs accurately. Ambiguity aversion, an illusion of certainty, and the repeated occurrence of cost deviations, as demonstrated by our illustrative example and analysis of the elemental cost models of 39 transport projects, collectively create a Sisyphean loop that is impossible to break. Thus, we question the prevailing conceptualisation and treatment of risk and uncertainty used to estimate the actual construction costs of projects. We offer a new theoretical lens framed around ecologically rational heuristics and provide suggestions to reconceptualize and recalibrate the treatment and assessment of uncertainty in cost estimates. We hope this paper provides a much-needed circuit breaker to public agencies’ prevailing treatment of risk and uncertainty and stimulates new lines of inquiry to improve the cost performance of transport projects.

Keywords:

• Cost performance
• estimate
• heuristics
• risk
• transport projects
• uncertainty

1. Introduction

Worldwide transport projects regularly exceed their budgets and are delivered late – a phenomenon yet to be adequately explained, though suggestions have been put forward over several decades (Merewitz 1973; Odeck 2004; Eliasson and Fosgerau 2013; Edwards and Kaeding 2015; Terrill and Danks 2016; Catalão, Cruz, and Sarmento 2019; Cavalieri, Cristaudo, and Guccio 2019; Love, Pinto, and Ika 2022; Welde and Klakegg 2024). A debate has unfolded on why transport projects experience cost overruns and benefit shortfalls, yet a consensus about how to address the problem remains a point of contention (Terrill and Danks 2016; Ika 2018; Flyvbjerg et al. 2018; Love, Ika, and Ahiaga-Dagbui 2019a; Kreiner 2020; Ika, Love, and Pinto 2022; Pinto 2023). However, with the increasing postponement or axing of projects due to labour shortages and the surging cost of materials, many governments have embarked on a period of self-reflection to determine and prioritise their infrastructure needs.

Examples include the delay of the Metro West in Sydney, Australia, which is impacting the construction of high-density apartments that are planned around most stations (Ludlow 2023; Tillet 2023) and the cancellation of High Speed 2 (HS2) in the United Kingdom (UK) (Hakimian 2023). At the centre of this existential introspection are governments’ concerns that their projects will exceed their cost and time forecasts and fail to deliver expected benefits. In addition, there is a fear of political fallout, the destabilisation of their term in office, and the disenfranchisement of the electorate, if not their distrust.

When projects exceed their budgets and planned completion dates, governments tend to question the nature of their procurement practices and toil by initiating inquiries and reviews to determine why this is the case. A case in point is the Ottawa Light Rail Transit project, which experienced cost increases, delays, and safety issues, resulting in an independent commission to investigate what went wrong and to identify ways to avoid similar problems in the future (Hourigan 2022).

Emerging from such inquiries and reviews are changes to policy and practice, which have tended to focus on the appraisal and the design and use of monitoring and evaluation before, during, and after the implementation of projects (Meunier and Welde 2017; Caffieri et al. 2018). The new or revised policies (e.g., the UK’s HM Treasury’s ‘The Greenbook 2022’) enacted under prevailing legislative and institutional frameworks (e.g., the requirement to achieve net zero carbon emissions), for example, have done little to mitigate the incidence of cost and time overruns. Project misperformance remains an ever-present reality despite the efforts to de-risk estimates using methods such as Estimating Using Risk Analysis, Real Options, Monte Carlo Simulation, and Reference Class Forecasting¹ (RCF) to improve the accuracy of a transport project’s actual construction costs (Mak, Wong, and Picken 1998; Flyvbjerg and COWI 2004; Tseng, Zhao, and Fu 2009; Batselier and Vanhoucke 2016; Traynor and Mahmoodian 2019; HM Treasury 2022).

Burdened by the need to be accountable to taxpayers, most public sector agencies charged with procuring infrastructure assets would no doubt, at some point during the project planning and design process, have questioned their ability to establish an accurate estimate of their actual construction costs unconsciously knowing it is—what the absurdist philosopher Albert Camus (1942) refers to in his writings—a Sisyphean Task; that is, an accurate assessment of actual construction costs cannot be achieved considering current practice and its tendency to conflate risk and uncertainty.

Against this backdrop, this paper addresses a question that perennially confronts the public sector: How can public agencies responsible for delivering transport projects improve the estimation accuracy of their actual construction costs? Indeed, this question has stumped many researchers and practitioners over several decades as they have sought to improve the cost performance of their projects with little avail (Love, Pinto, and Ika 2022). We by no means aim to provide a definitive answer to this question. Instead, we seek to identify a critical gap in knowledge contributing to and stymieing progress to procure cost-effective projects.

The conflation of risk and uncertainty – whereby uncertainty is reduced to risk – is a practice, we argue, that has contributed to inaccurate estimates of construction costs. We define risk as having a full range of potential outcomes that can be known and assigned known probabilities, whereas uncertainty can take varying forms. Uncertainty can be categorised by type (i.e. aleatory and epistemic), take various guises (e.g., ‘factual’, ‘probabilistic’, ‘meta-doxastic’, ‘value’ and ‘linguistic’) and be described using several dimensions (e.g., resolvable to radical) (Meder, Le Lec, and Osman 2013; Hansson 2022; Vis 2024). However, for simplicity’s sake, we adopt a ‘Knightian’ lens, whereby some of the outcomes and probabilities are unknown (Knight 1921).

In contrast to most previous studies examining the risk and uncertainty of cost overruns (e.g., Flyvbjerg 2008; Flyvbjerg 2012; Kim and Kwak 2018; Fu and Gil 2024), we consider these concepts as mutually exclusive concepts and, in doing so, we suggest that more accurate assessments of actual construction costs may be achieved. While our paper is not the first to distinguish between the concepts of risk and uncertainty (e.g., Ika, Love, and Pinto 2022: Love, Ika, and Pinto 2024), research has yet to couple this conceptual conflation with the depressive cycle where many projects tend to overrun their costsand the remedies are not on par with expectations, which is akin to a Sisyphean task from a metaphorical standpoint. We are cognisant that accurately determining actual construction costs is akin to an unrealistic goal. Still, strides can be made to mitigate the extent to which actual costs exceed planned budgets by applying subtle but distinct changes to estimating a project’s cost contingency and formulating its risk and uncertainty components (Love et al. 2021a,2021b).

The central argument in our paper is that public agencies are trapped in a Sisyphean loop, repeatedly de-risking estimates using probabilistic approaches (e.g., overly relying on reporting ‘P10, ‘P50’ and ‘P90’). Treating uncertainty with the same optic for decision-making will not, as prevailing evidence for transport projects has manifestly demonstrated, result in improved assessments of actual construction costs. Thus, different rules for dealing with cost estimation under uncertainty in transport projects are required (Love, Ika, and Pinto 2023a).

Typically, the decisions to formulate cost estimates are based on a ‘risk characterization of an uncertain world’ (Mousavi and Gigerenzer 2014; Gigerenzer, 2014). However, structural differences between risk and uncertainty exist. Yet public agencies do not acknowledge or appear to understand these differences, as evidenced by the absence of strategies to deal with uncertainty in their published estimation guidelines². So, by merely reducing uncertainty to risk, we argue that public agencies unfittingly assume they have control of project costs. Consequently, they unintentionally underestimate the uncertainty that pervades the construction process, falling naïvely prey to optimism bias (Sharot 2011), which they, however, take to be the casus belli of cost underperformance, according to dominant theory and practice (Flyvbjerg et al. 2018; Love, Ika, and Pinto 2023a).

We commence our paper by using The Myth of Sisyphus as a framing to describe the public sector’s impossible task of accurately determining a transport project’s actual construction costs before briefly reviewing the cost performance literature (Section 2). Next, to illustrate the recurring mis-performance problem, the cost profile of transport projects delivered in New Zealand is analysed to reinforce and provide further context to assist us with addressing our research question (Section 3). In conjunction with the literature and the empirical analysis, we suggest a new theoretical lens for treating cost uncertainty and offer changes to practice to break the Sisyphus loop (Section 4). We conclude our paper by identifying future avenues for research and its contributions (Section 5).

2. The public sector’s Sisyphean task

We now draw on previous literature to examine how the public sector produces cost estimates for transport projects and identify key areas where it falls short in determining actual construction costs. Simultaneously, we refer to Camus’s (1942) philosophical essay, The Myth of Sisyphus, as a metaphorical backdrop to provide deeper theoretical insights and a nuanced understanding of why inaccurate cost estimates of actual construction costs recurrently occur. Camus (1942) tells the story of the mortal Sisyphus, condemned by the gods to roll a stone up a mountain for all eternity. Every time Sisyphus reaches the summit of a mountain, the stone rolls back down, forcing him to begin the task again, reinforcing the cycle that his existence is meaningless.

Producing an accurate estimate of expected actual construction costs, years or even decades in advance, for a transport project is a hopeless task for the public sector as the environment within which it is delivered is fraught with volatility, complexity, uncertainty, ambiguity (VUCA) (Flyvbjerg et al. 2018; Love, Pinto, and Ika 2022). Yet incumbent governments overwhelmingly exude confidence in their business case and decision-making processes, presenting a commitment to taxpayers that a project’s costs and delivery dates are absolute (Caffieri et al., 2018).

Beneath the façade of self-confidence– stemming from political aspirations of the government – the public sector agency responsible for delivering a project and its design team are confronted with a seemingly endless cycle of producing a scope and design, accommodating evolving stakeholder needs and demands, undertaking countless engineering iterations, and revising cost estimates (Caffieri et al., 2018; Fu and Gil 2024). Unquestionably, a public sector agency can relate to Sisyphus’s plight.

2.1. Naïveté: emergence of ambiguity aversion

Completing each stage of a transport project’s ‘gateway process’ provides a brief reprieve for a public agency and its design team until the proverbial stone rolls back down and work needs to begin on the next stage with more design-related information requiring estimates to be revised.

As more information becomes available, risks become known, and the pressure and stress to produce an accurate estimate build up (Love, Ika, and Ahiaga-Dagbui 2019a; Love et al. 2019b). Generating another estimate can seem as meaningless and frustrating as Sisyphus pushing a stone up the mountain. Cost delimiters are invariably identified, and the estimate is fine-tuned – sceptics may argue it is fudged – so as not to deviate too much from that affirmed from the previous stage (Flyvbjerg 2012). But, uncertainties can be downplayed due to ‘ignorance in knowledge’, which can take the following forms (Kutsch and Hall 2010; Nottelman 2015; Gigerenzer and Garcia-Retamero 2017): (1) factual ignorance (absence of knowledge of some fact); (2) object ignorance (unacquaintance with some object), (3) technical ignorance (absence of knowledge of how to do something); and (4) deliberate ignorance (a wilful decision not to know). Indeed, it is counterintuitive and irrational to forego any curiosity about the future and anticipate what could happen under varying conditions.

Without empirical research examining how ignorance influences decision-making during the formulation of cost estimates – bar Flyvbjerg (2016) and Ika (2018), who question whether ignorance is good or bad for projects – we take a charitable position and thus assume the technical know-how to assess uncertainties is hard to come by. In this instance, we view public agencies as consciously possessing an ‘ambiguity aversion’,³ unable to reconcile risk and uncertainty, instead focusing on the known risks impacting a project’s actual construction costs (Gigerenzer and Garcia-Retamero, 2017, 124).

It has been well-documented, however, that a transport project’s scope can change significantly from the date its business case is approved until it is handed over for operation (Merewitz 1973; Morris 1990; Hinze, Selstead, and Mahoney 1992; Love et al. 2017a,2017b;2017c; Love and Ika 2021; Welde and Odeck 2017; Welde and Dahl 2021; Welde and Volden 2022). The reasons for scope changes naturally vary, but the most common are overlooking or shifting stakeholder requirements and needs, price inflation, and resource constraints.

In 2015, the business case for the 1,700km Melbourne to Brisbane Inland Freight Rail project, for example, was estimated to cost AU$9.9 billion, and in 2021, AU$14.5 billion, with design changes being the main reason attributable for this sizeable increase (Terrill and Emslie 2021). Dwarfing this case is the California High-Speed Rail project, which was approved in 2008 and to be completed by 2022 at a cost of US$33 billion, but in 2022, this figure was revised to US$113 billion, with it likely to balloon to US$128 billion (Vartabedian 2022; Yost and Schelpp 2024). Design issues (e.g., tunnelling) and scope changes, inflation, delays in acquiring environmental approvals, and shortfalls in funding have contributed to increased costs and delays (Pettitt, Boyo, and Evers 2023).

Terrill, Emslie, and Moran (2020) study of the cost performance of transport megaprojects revealed that 10% of those sampled final construction costs exceeded their approved business case estimates by more than 49%. This is a common situation, which is to be expected as the time between the approval of the business case and the completion of a project can be years or even decades; scope changes are bound to occur as the design is not finalised at a business case approval and costs will increase with inflation (Love et al. 2015a,2015b).

Governments are certainly cognisant of the dangers of scope changes and creep, unplanned events, and the likely need for change orders, and accordingly, as we have mentioned, they have developed robust processes and procedures and utilised various tools and techniques to mitigate their impact. Calling for greater rigour in the estimation process, Terrill and Emslie (2021) have called for the public sector to embrace RCF (i.e. adding an uplift to a project’s cost estimate and contingency) so that unknown risks and optimism bias can be avoided.

Regardless of the application of RCF, the risk of cost and time overruns is a chronic problem for public agencies. While they are invariably optimistic that they can control, manage and mitigate such risk, a forlorn hope prevails as there is an awareness that all their hard work to revise their cost and time estimates will likely be undone by unexpected events. A case in point occurred during the formulation of HS2’s original budget, which was determined using simple calculations that did not consider the project’s complexities and uncertainties (e.g., ground conditions⁴, backfilling and stabilising mines, gas storage caverns, disposal of hazardous waste and high-cost elements)⁵.

Indeed, Camus (1942) saw the fate of a workforce doing the same repetitive tasks daily as no less absurd than that of Sisyphus. We can draw parallels here with an epigram from Wilczek (2015) – recipient of the 2004 Nobel Prize in physics – who cogently remarks, ‘naïveté is doing the same thing over and over, and always expecting the same result’. So, if public agencies continue to moderate the impact that uncertainties can have on an estimate’s formulation, transport projects will continue to mis-perform (Love, Ika, and Ahiaga-Dagbui 2019a).

2.2. Acceptance: an ‘illusion of certainty’

For the industrious Greeks, Sisyphus’s punishment was the ultimate torment—a futile task with no hope of completion; we equate this ineffectuality to the unlikelihood of transport projects being delivered within cost and time targets. As part of their procurement strategy, we often see public agencies offloading unidentified risks and uncertainties to the private sector and expecting them to absorb these associated costs (Love et al. 2021a). Traditionally, the practice of risk aversion – low-risk tolerance – has been the public sector’s typical default when modifying/extending existing and constructing new transport infrastructure assets. As such, it has provided a breeding ground for adversarial and opportunistic behaviours to manifest, resulting in contractual claims, delays, and, ultimately, a project’s mis-performance (Love, Pinto, and Ika 2022).

Proponents of relational (i.e. collaborative) procurement where risks are typically shared, such as Walker and Lloyd-Walker (2015), Australian Constructors Association (2017, 2021), Love et al. (2021a), Walker, Vaz-Serra, and Love (2022), and Haaskjold, Andersen, and Langlo (2023) have been critical of public agencies’ low-risk tolerance approaches. Instead, such proponents have suggested that the public sector make greater use of relational procurement approaches such as alliancing – where collaboration, trust, and best-for-project outcomes thrive – to deliver large-scale transport projects as target cost⁶ and time figures can be reliably estimated. The collaborative team of an alliance comprising the public agency, the design/engineering team, and a constructor jointly formulate a project’s target outturn cost and target adjustment events, which equally consider the allocation of risks and uncertainties (Walker, Love, and Matthews 2023).

Camus (1942), like Kafka (1916), Ionesco (1954), and Pinter (1959), to name a few, shares a similar absurdist and pessimistic vision of humanity where people vainly struggle to find a sense of purpose and control their fate—feelings of hopelessness, bewilderment, and anxiousness take centre stage. Camus (1942) believed that people seek meaning in existence, but the universe is inherently meaningless, indifferent to this need. In Camus’ (1942) view, Sisyphus was the archetypal absurdist hero—burdened by this meaningless task and facing an uncaring universe. Sisyphus acknowledges and accepts the absurdity of his fate. However, this is in stark contrast to the public sector, as it is unable to admit it has a mutual responsibility with the private sector to divide risks and uncertainties into two moieties based on who is best to manage them and deliver shared value (Love et al. 2021a).

For Camus (1942), acceptance is crucial and equally so for the public sector as it must acknowledge that risks have to be shared to make headway towards ameliorating the performance of its transport projects. It is only natural for taxpayers to be vexed when a transport project’s performance goes astray from its planned cost and time targets and for incumbent governments to feel resentful when the media seeks to apportion blame on them for what was always going to be an inevitable outcome. Flyvbjerg (2011) describes this inevitability with an ‘Iron Law’ where transport projects will be ‘over budget, over time, over and over again’ (321). An analogy can be drawn here with Sisyphus’ lived experience whereby he surrenders to the fate bestowed upon him.

When Sisyphus makes the long walk down the slope to retrieve the stone, he is not sad; quite the contrary. Sisyphus is at peace. Left by fate with no other options, Sisyphus revolts in the only way he can—by accepting the absurd situation, joyfully shouldering the burden, and making an ascent once again. Here, the accomplishment of the task provides contentment for Sisyphus, much like a public agency that goes through the motions to produce estimates, giving no meaningful consideration to the uncertainties that will derail a transport project’s performance. This point may be considered unfair, considering the due diligence afforded to preparing an estimate and implementing gates to improve project outcomes and mitigate risks. However, public agencies are cognisant that if a realistic assessment of uncertainties—even though they tend to implement any formal approaches to do so—were incorporated into an estimate, then a project’s cost-benefit analysis could be adversely impacted, rendering its viability questionable (Love et al. 2021b).

An inquiry by the New South Wales (NSW) Auditor General into Sydney’s Light Transit Rail (LRT) project, Australia, for example, revealed that the original capital budget of AU$1.6 billion established in 2013 increased by AU$549 million mainly due to mispricing and omissions in the business case (NSW Auditor-General 2016). Due to ensuing scope changes, the cost-benefit ratio decreased from 2.4 to 1.4. The project might not have proceeded if a cost-benefit ratio of 1.4 had been originally put forward in the business case.

According to Flyvbjerg and COWI (2004), estimates are often prone to optimism bias and strategic misrepresentation⁷. What is more, estimates are often purposefully manipulated or fudged to enable a project to appear as a viable proposition. As a matter of fact, Flyvbjerg et al. (2018) make a bold claim that ‘the root cause of cost overrun [for all transport projects] is behavioral bias’ (174). To cater to such biases and strategic behaviours, RCF is drawn upon to de-risk estimates and provide an additional up-lift to a contingency to ensure actual construction costs do not exceed the project’s planned budget (Flyvbjerg 2008).

In essence, RCF acts as a ‘contingency on a contingency’ or ‘artificial inflator device’ and thus does not provide a rationale for accommodating cost uncertainties that can manifest during construction and derail a project’s cost and time performance (Love, Pinto, and Ika 2022, 3). Before proceeding any further with our paper, we must address the elephant in the room, which the UK public sector, for example, ignores when justifying uplifts derived from RCF for de-risking estimates. Optimism bias and strategic behaviour have not been shown to manifest empirically in the cost estimation process (Mak and Raftery 1992; Love, Pinto, and Ika 2022; Love, Ika, and Pinto 2024). Optimism bias has been only assumed to exist and while strategic behaviours may have occurred, they are difficult to detect and prove (Pinto 2023). Examples of strategic misbehaviour are few and far between but prominent cases receiving media attention are:

• Brisbane’s Clem 7 – a 4.8-kilometre motorway tunnel, where a consultant provided misleading and deceptive estimates for a project’s toll revenue as it forecasted based 100,000 vehicles per day when the actual traffic volumes were approximately 22,000. Unable to collect enough toll revenue to pay interest on an AU$1.3 billion debt, the operator went into receivership (Tam 2015); and

• HS2, where the intentional cost underestimation⁸ occurred and their subsequent escalation, was deliberately covered up (The Committee of Public Accounts, 2020; Calvert and Artbuthnott 2023). The project has been cancelled due to spiralling costs and mismanagement (Oakervee 2020; Norris and Tetlow 2023; Public Accounts Committee 2023).

The assumption that optimism bias and strategic behaviour are the root causes of cost overruns and, therefore, all estimates produced for transport projects need to be de-risked for their possible presence is devoid of ‘objective meaning’ (Love, Ika, and Ahiaga-Dagbui 2019a; Ika, Love, and Pinto 2022; Pinto 2023). Bolstering this position, Jørgenson et al. (2023) suggest we should not ‘reward’ misrepresentation and bias by accommodating them in a cost estimate as a risk when they are an unknown entity (3481). Nonetheless, statistical modelling can only accommodate what is known, not what is unknown. Frankly, assigning objective probabilities with future consequences, where a decision-maker ‘hasn’t got a clue’, is ‘highly questionable if not invalid in principle’ (Davidson 1991, 130). Moreover, ‘Bayesian decision models are not necessarily valid in the natural environment and appear to be primarily driven by mathematical convenience—and less so by Brunswik’s argument (2001) that cognition and environment are indivisible like a married couple who have come to terms with each other through mutual adaptation’ (Hertwig et al. 2022, 468). Yet, methods relying upon statistical analysis such as RCF, for example, conflate risk and uncertainty, treating them to be ‘blindly synonymous’ and providing an ‘illusion of certainty’ for determining the actual construction costs for transport projects (Love, Pinto, and Ika 2022, 3; Pinto 2023).

We will now briefly look at previous transport cost performance studies where the recurring problem experienced by the public sector is identified before presenting what is repeatedly known using a new dataset. The absurdity here is that the cost overruns are normalised, becoming a putative feature of transport infrastructure procurement. This tragic situation resembles Sisyphus’s hopeless struggle, where he understands his fate but has no hope for reprieve. However, the lucidity he achieves with this understanding places him above his fate. This is a position the public sector seemingly takes after audits, reviews, and inquiries have been undertaken to determine why project costs exceeded their budget. Indeed, a degree of learning may occur, and changes to procurement practices may be implemented. Despite this, the public sector continues to misunderstand ‘how projects work’ in VUCA environments. Flyvbjerg (2011) uses his ‘Iron Law’ to describe this state of affairs that the public sector confronts when procuring transport projects.

2.3. The recurring problem: over budget, over and over again

In Table 1, we present examples of studies that have sought to quantify the cost performance of transport projects in a range of countries. A cursory glance at Table 1 reveals that the reported cost deviations vary significantly due to the delimiters that quantify their impact on a project’s performance (i.e. reference point, X $\to$ Final Construction Cost, FCC). For example, the calculated cost deviation is larger in the studies where the delimiters are ‘the decision to build’ and the ‘initial estimate’ to FCC. Studies using different delimiters to determine a project’s performance make it impossible to compare findings (Love, Ika, and Ahiaga-Dagbui 2019a; Love et al. 2019b).

Table 1. Examples of transport cost performance studies (in chronological order).

Author(s)	Location(s)	Reference point (X $\to$ FCC)	Sample size	Key cost descriptors	Mean cost deviation (%)
Hinze, Selstad, and Mahoney (1991)	USA	Contract award	433	Type
				New road (n = 110)	9.23
				Resurfacing (n = 181)	3.90
				Bridge (n = 58)	4.24
				Safety improvement (n = 74)	2.98
Joint Legislative Audit and Review Commission (JLARC) (2001)	USA	Contract award	211	Road	11.10
Flyvbjerg, Holm, and Buhl (2002)	Worldwide	Decision to build	258	Type
				Road (n = 167)	20.00
				Rail (n = 58)	44.70
				Fixed link (n = 33)	33.80
Vidalis and Najafi (2002)	Canada	Contract award	708	Road	10.52
Bordat et al. (2004)	USA	Contract award	2668	Road	4.5
Odeck (2004)	Norway	Detailed planning level	620	Road	7.88
				Very small $\le$15 million NOK (n = 420)	7.55
				Small > 15 $\le$100 million NOK (n = 156)	25.16
				Medium >100 $\le$350 million NOK (n = 33)	5.32
				Large > 350 million NOK (n = 11)	−2.50
National Audit Office (2007)	England	Business case	36	Roads including bypasses, junction upgrades, safety improvements, road widening	17.00
Lee (2008)	South Korea	Budgeted costs at project approval	161	Type
				Road (n = 138)	24.58
				Rail (n = 16)	47.64
				Airport (n = 2)	60.38
				Port (n = 5)	36.27
Makovšek, Tominc, and Logožar (2012)	Slovenia	Decision to build		Road
				20 projects as part of a 374 km National Motorway Construction Programme (costs for construction, design, land acquisition, and others)	130.86
				36 projects (construction costs only)	119.19
Cantarelli et al. (2012)	Netherlands	Decision to build	78	Type
				Road (n = 37)	18.80
				Rail (n = 26)	10.60
				Tunnel (n = 8)	34.90
				Bridge (n = 7)	6.60
Love et al. (2014)	Australia	Contract award	58	Type
				Road (upgrade) (n = 44)	22.91
				Road (elevated) (n = 5)	11.89
				Bridge (n = 7)	11.36
				Tunnels and subways (n = 2)
Verweji, Van Meerkrek, and Korthagan (2015)	Netherlands	Contract award	45	Road	22.37
Kostka and Anzinger (2016)	Germany	15		Type
				Road (n = 4)	28
				Rail (n = 6)	33
				Airport (n = 2)	48
				Tunnel (i.e. fixed link)	364
				Waterway	91
				Port	80
Terrill and Danks (2016)	Australia	Budget announcement	836	Road	29.00
				Rail	16.00
Chong and Hopkins (2016)	Burkina Faso, Cape Verde, El Salvador, Georgia, Ghana, Honduras, Mali, Mongolia, Mozambique, Nicaragua, Senegal, Tanzania, and Vanuatu	Project life - costs examined: Funding authorisation (FA); engineer’s estimate; contract award (CA); and final cost (FC)	48	Roads
				FA $\to$ FC	235%
				FA $\to$ EE	100%
				CA $\to$ FC	21%
Love et al. (2017a)*	USA	Full funding grant agreement	32	Light transit rail.	42
Love et al. (2017b)	Australia	Contract award	16	Rail	23.00
Catalão, Cruz, and Sarmento (2019)	Portugal	Contract award	1091	Project types are not specified.
				Deviation from initial estimated costs	17.80
				Projects where only the final cost exceeded the initial cost – (positive cost deviation/overrun),	26.20
Cavalieri, Cristaudo, and Guccio (2019)	Italy	Project-life – planning, contractor selection, and execution stages	1083	Roads
				Planning	−1.34
				Contractor selection	−20.31
				Execution	10.92
Andric et al. (2019)	Asia	Decision to build	74	Type
				Roads (n = 34)	24.05
				Rail (n = 10)	48.01
				Roads (n = 21)	−12.15 −8.77
				Rail (n = 9)
Love et al. (2019b)	Hong Kong	Project life –	85	AB $\to$ PTE	−28.31
		Costs at Approved budget (AB), pre-tender estimate (PTE), contract award (CA), and final contract (FC)		PTE $\to$ CA	−4.00
				CA $\to$ FC Other Road improvement/noise barriers (n = 18) Fixed linked (i.e. tunnels/bridges/expressway) (n = 42) Universal accessibility (e.g., crossings and lifts) (n = 7) Footway/subway (n = 10) Ancillary highway facilities (n = 8)	12.16
Welde and Dahl (2021)	Norway	Contract award	712	Road	17.00
Molinari, Haezendonck, and Mabillard (2023)	Belgium	Initial estimate	36	Road (n = 11)	22.82
				Rail (n = 15)	4.42
				Inland waterways (n = 10)	4.62

* Data was obtained from Pickrell (1990) and O’Toole (2015).

If the decision to build, for example, is used as a reference point, then a more significant cost deviation is expected as the project’s level of definition⁹ would typically be incomplete. As the project’s scope is defined and its design and engineering are finalised, estimated costs will tend to increase (Cavalieri, Cristaudo, and Guccio 2019).

However, as shown by Love et al. (2019b) in Table 1, this is not always the case as estimated costs may decrease from a project’s ‘Approved Budget (AB) $\to$ Pre-Tender Estimate (PTE)’ (Table 1). Bids are then called for by potential contractors (or consortiums) and negotiations about a project’s price will ensue. Additions or deletions to the project’s scope can occur during these negotiations (Table 1). Once a price for the works is agreed, the contract is awarded. The contract award provides the public agency price certainty for the agreed scope of work specified in the contract documentation; hence, it is used as a delimiter for determining cost deviations. Any changes sanctioned by the public agency (or their representative) after the contract award and/or unexpected events occur, which the estimated contingency cannot accommodate, will result in additional costs to a project.

In Table 1, scope and change orders from a contract’s award are the main contributors to cost deviations writ large (e.g., Hinze, Selstad, and Mahoney 1991; Vidailis and Najafi 2002; Makovšek, Tominc, and Logožar 2012; Love et al. 2017c; Love et al. 2019b). However, it has been vehemently argued by Flyvbjerg, Holm, and Buhl (2002), Cantarelli et al. (2012), and Flyvbjerg et al. (2018) that optimism bias and strategic misrepresentation are the root causes of cost overruns in transport projects despite the lack of causal evidence to substantiate their assertions.

Explanations of why cost deviations manifest and result in transport projects repeatedly mis-performing abound in the literature. Three prominent theories have been relied upon to explain a project’s cost deviation, these being the: (1) Hiding Hand (Hirschman 1967); (2) Planning Fallacy (Malevolent Hand) (Flyvbjerg 2016); and Fifth Hand (Ika, Love, and Pinto 2022). It is outside the scope of this paper to examine these theoretical perspectives, as reviews of them can be found in the works of Kriener (2020) and Ika, Love, and Pinto (2022). We hasten to note that no consensus exists on which theory provides the ‘best’ explanation for the causes of deviations in project costs. While Love, Pinto, and Ika (2022) and Ika et al. (2023) acknowledge the explanations of project behaviour suggested by the Hiding Hand and Planning Fallacy, they argue the Fifth Hand provides a balanced and nuanced account. However, empirical research is needed to test the validity of the Fifth Hand in real-life project settings.

Within a cost estimation context, we see the Hiding Hand acknowledging the role of ignorance and underestimating task complexity in projects, providing a setting for considering uncertainty but without clear-cut strategies that can enable rather than limit decision-making. In stark contrast, the Planning Fallacy does not consider the presence of ignorance but instead assumes a wilfulness to underestimate the complexity and overlook risks due to innate behavioural biases or deliberate strategic misbehaviours. As a result, uncertainty is ignored and emphasis is placed on risk settings, which gives rise to an illusion of certainty.

While the Hiding Hand and Planning Fallacy adopt monistic views to deal with uncertainty and risk or conflate these notions under their respective positions, the Fifth Hand treats them as distinct entities, giving equal credence to how they will impact a project’s performance (Ika, Love, and Pinto 2022). Nevertheless, the conflation of risk and uncertainty by much of the literature and public agencies results in them having to shoulder the burden of trying to produce an accurate assessment of actual construction costs.

The consequence of conflating risk and uncertainty is examined using an illustrative example and abductive reasoning. The dataset we present in the next section of this paper reinforces the studies identified in Table 1, adding to the body of work that examines project cost performance. In addition, we show that a transport authority conflates risk and uncertainty, like many others worldwide, thereby reducing uncertainty to risk.

3. Illustrating a leitmotiv of practice

Determining a project’s cost deviation appears to be a straightforward task and, as mentioned above, it is typically expressed as a ratio of the cost estimate at the decision to build or contract award, for example, over FCC (Makovšek, Tominc, and Logožar 2012). However, a fundamental difficulty is obtaining primary data due to political and commercial sensitivities and in a format that can be effectively used for analysis (Makovšek, Tominc, and Logožar 2012).

Surprisingly and without solicitation, a consultancy organisation involved with providing pre-and-post-contract cost management services to a public agency voluntarily provided us with primary cost data for a sample of 39 transport projects constructed in New Zealand as they could see parallels with the findings reported in Love et al.'s (2019b) Hong Kong study (Table 1). Due to confidentiality issues, the names of the projects, the contractors, consultants, and the like have been withheld. We use procedures associated with formulating the estimates and the analysis of the project’s actual costs to provide an empirical backdrop to help us address our proposed research question.

3.1. Dataset

Under the public agency’s cost estimation guidelines, the consultancy organisation was required to complete an ‘elemental cost questionnaire’ after a project’s final account had been completed so that data could be incorporated into their ‘Elemental Cost Database’ and used to support the production of future estimates of actual costs. The questionnaire requests a considerable amount of detail about a project, such as its type, the project’s terrain, length, the nature of earthworks, details about structures and pavements, traffic management, procurement method duration, delays, and an elemental breakdown of the implementation estimate (i.e. contract award) compared with actual costs. Examples of the elemental cost data provided to the public agency at a project’s completion can be seen in Figures 1 and 2.

Figure 1. Example of elemental cost model.

Figure 2. Example of elemental cost model.

The cost estimates prepared before the contract award were unavailable – they were confidential. However, it is the point at which a degree of cost certainty can be provided to the public sector as they sign a contract with a construction organisation (or consortium) to deliver an asset for an agreed price. Specific details of the cost estimation processes for transport projects can be found in the Waka Kotahi New Zealand (NZ) Transport Agency (2023) ‘Cost Estimation Manual’. Within this cost estimation manual, risk and uncertainty coalesce under the auspices of an estimate’s ‘contingency’ and ‘funding risk contingency’ components, with the explicit instruction to calculate their values using statistical models. Thus, uncertainty is characterised and determined with a risk-focused lens, as noted in Section 9 of the NZ Transport Agency’s ‘Cost Estimation Manual’ (2023, 35). The data used for analysis and derived from the 39-transport project’s elemental cost models are available as a supplementary file, and all monetary values presented in our study are in NZ dollars. The analysis demonstrates that even though the price for delivering a project’s scope of work is agreed upon, cost still increases over and above the contingency allowance.

We were not permitted access to the cost contingency allowance as it was confidential for each project. However, we were advised by the consultancy organisation that an average of 5% would be a reasonable assumption to make across the sample as a significant proportion of projects were re-measurement contracts. The cost contingency assumption aligns with Clark and Lorenzoni's (1985) recommendation of 3% to 5% and the empirical observations of Love et al. (2015a). In the case of traditional lump sum contracts, for example, the cost contingency would be lower as the design is supposed to be complete, with the likelihood of changes being minimal. Similarly, this situation would apply to the other procurement methods adopted (e.g., D&C and ECI).

3.2. Analysis

Figure 3 presents a series of histograms to represent our sample project characteristics visually. We can see that 74% (n = 29) of the projects were < NZ$30 million, and only three were > NZ$100 million. Table 2 provides a breakdown of the costs by project size. The sample’s costs per kilometre (km) were x̅ = NZ$9,252,306, $\sigma$ = NZ$11,896,312.

Figure 3. Histograms of project characteristics.

Table 2. Breakdown of projects by size.

Project Size (NZ$ million)	Number (N)	Mean (NZ$) (x̅)	Standard Deviation (NZ$) ($\mathit{\sigma }$)
< $10	16	5,792,967	865, 0572
>$10 $\le$ $30	13	5,563,118	5,944,765
>$30 $\le$ $50	4	14,771,703	9,017,011
>$50 $\le$ $100	3	19,710,711	21,423,772
>$100	3	25,870,995	22,610,454.

Several procurement methods were used to deliver projects, with traditional lump sum and re-measurement¹⁰ forms accounting for 69% (n = 27). Traditional re-measurement was required for projects where the nature of the critical work could be detailed at the time of tender. Still, quantities would have been unknown for several elements, such as earthworks and remedial works. Markedly, re-measurement methods were utilised across the spectrum of project types sampled, except for motorways (new) and carriage improvement (safety). However, most project types were upgrades and additions to existing infrastructure, 82% (n = 32), with new motorways accounting for 18% (n = 7).

The descriptive statistics for the sample’s cost and time deviations are summarised in Figures 4 and 5, with Figure 6 providing a breakdown by procurement method and project type. Here, cost deviations are x̅ = 18.28% and $\sigma$ = 24.73% of the contract value, and time deviations from the expected completion, x̅ = 1.74 months and $\sigma$ = 5.53 months. As noted in Figure 6, the mean cost contingency at contract award of 5% was insufficient, falling short, on average, by 13.28%.

Figure 4. Distribution and summary of time deviations.

Figure 5. Distribution and summary of time deviations.

Figure 6. Mean cost and time deviations by procurement method and project type.

We also observe a project’s contract period (i.e. duration) and reveal the cost is not a predictor of deviations (Figure 7). A breakdown of cost and deviations by project type, procurement method, and size are presented in Tables 3–5. The total cost for the 39 projects analysed is NZ$1,308,321,641, with a x̅ = NZ$33,546,708, $\sigma$ = NZ$58,793, 985, $\wedge$ (min) = NZ$2,623,250, and $\vee$ (max) =NZ $342,089,198.

Figure 7. Scatter plots. (a) Contract period and cost deviations. (b) Project cost and time deviations

Table 3. Project type: Cost and time characteristics.

	Descriptive
		x̅		$\vee$		$\wedge$		$\mathit{\sigma }$
Project type	N	Cost (NZ$)	Time (months)	Cost (NZ$)	Time (months)	Cost (NZ$)	Time (months)	Cost (NZ$)	Time (months)
Motorway (upgrade)	2	13,994,745	21.00	18,828,180	32.00	9,161,311	10	6,835,508	15.55
Motorway (new)	7	84,971,719	34.12	342,089,198	58.00	10,640,300	18.00	120,072,734	12.45
Interchange (upgrade)	8	30,068,437	28.25	136,519,788	67.00	5,630,358	12.00	44,4095,07	14.00
Rural realignment	9	15,027,343	31.44	57,181,723	60.00	2,623,250	13.00	17,901,795	16.24
Bypass	3	3,093,1276	43.00	48,082,848	60.00	8,172,882	33.00	20,533,7251	14.79
Bridge	6	21,622,223	20.08	69,486,721	36.00	4,926,190	5.50	24,383,846	10.48
Carriageway improvement (safety)	1	45,198,344	20.00	–	–	–	–	–	–
Rural expressway	3	14,003,693	26.66	15,077,447	33.00	12,470,688	17.00	1,362,711	8.50
Total	39	33,546,708	29.21	342,089,198	67.00	2,623,250	5.50	5,8793,985	14.54

Table 4. Procurement method: cost and time characteristics.

	Descriptive
		x̅		$\vee$		$\wedge$		$\mathit{\sigma }$
Procurement method	N	Cost (NZ$)	Time (months)	Cost (NZ$)	Time (months)	Cost (NZ$)	Time (months)	Cost (NZ$)	Time (months)
Early Contractor Involvement (ECI)	4	49,143,343	37.50	136,519,788	67.00	5,935,439	24.00	59,404,486	20.33
Design & Construct (D&C)	5	98,292,831	32.20	342,089,198	58.00	4,926,190	12.00	144,201,825	18.11
Traditional lump sum	14	18,469,655	32.30	69,486,721	60.00	2,623,250	5.50	20,209,071	15.78
Traditional (remeasurement)	13	19,829,803	21.76	69,719,252	35.00	2,693,970	10.00	17,911,088	8.61
Alliance	2	27,919,322	28.00	45,198,344	36.00	10,640,300	20.00	24,436,227	11.31
Mixed (Traditional, D&C)	1	48,082,848	36.00	–	–	–	–	–	–
Total	39	33,546,708	29.21	342,089,198	67.00	2,623,250	5.50	58,793,985	14.54

Table 5. Project size: cost and time deviations.

	Descriptive
		x̅		$\vee$		$\wedge$		$\mathit{\sigma }$
Project size (Millions)	N	Cost (%)	Time (months)	Cost (%)	Time (months)	Cost (%)	Time (months)	Cost (%)	Time (months)
$\le$$10	16	22.18	2.4	137.00	16.00	−15.00	−3.00	34.15	5.27
> $10 $\le$ $30	13	14.07	0.00	65.00	6.00	0.00	−4.00	17.51	2.48
> $30$\le$$50	4	20.25	−0.75	36.00	2.00	4.00	−3.00	13.57	2.21
> $50 $\le$$100	3	21.66	9.33	35.00	24.00	0.00	0.00	18.92	12.85
>$100	3	18.28	1.74	13.00	3.00	4.00	0.00	4.93	1.73
Total	39	18.28	1.74	137.00	24.00	−15.00	−4.00	24.73	5.53

To measure the strength of the linear relationship between cost deviations and project characteristics, we performed a correlation analysis with the results presented in Table 6. The results show a significant but weak correlation between procurement method and project type, which can be explained by the public agency predominately drawing on the use of traditional methods as they sought to provide cost and time certainty as the design is complete at the time of contract award. Also, a significant but weak correlation exists between project cost and the cost per km, with a strong correlation with contract duration. Thus, larger projects will tend to have longer contract durations, and therefore, a contractor’s preliminaries will rise, and labour and materials prices are prone to escalation, for example, increasing the cost per km.

Table 6. Correlation matrix.

Variables	Project type	Procurement method	Cost per km	Contract duration	Project cost	Time deviation	Cost deviation
Project type	1.00
Procurement method	.322*	1.00
Cost per km	−0.09	−.37*	1.00
Contract duration	−0.13	−0.25	0.16	1.00
Project cost	−0.21	−0.23	.37*	.43**	1.00
Time deviation (%)	0.05	−0.06	0.20	−0.30	−0.72	1.00
Cost deviation (%)	0.03	0.29	−0.179	−0.08	−0.79	.51**	1.00

* . Correlation is significant at the 0.05 level (2-tailed). **. Correlation is significant at the 0.01 level (2-tailed).

We performed a one-way Analysis of Variance (ANOVA) test to determine if there were any statistical differences between the means for various project types and procurement methods for time and cost deviations. In doing so, we can ascertain the characteristics of projects impacting their performance. It can be seen in Figure 8 that the one-way ANOVA reveals that there was not a statistically significant difference in cost deviations between project types F (23,15) = 0.968, p > 0.05, and procurement methods F (23,15) = 1.304, p > 0.05. Similarly, in Figure 9, the ANOVA shows that there was not a statistically significant difference in time deviations between project types F (20,18) = 3.777, p > 0.05, and procurement methods F (18, 38) = 0.909, p > 0.05.

Figure 8. Mean plot and one-way ANOVA for project type and procurement method and cost deviations.

Figure 9. Mean plot and one-way ANOVA for project type and procurement method and time deviations.

As we categorised project size as a nominal variable comprising five classifications, we used a Kruskal-Wallis Test to determine if there were differences in mean cost and time deviations between them. However, no significant differences exist for cost H (4) = 2.43, p = 0.65 and time deviations H (4) = 4.84, p = 0.30. These findings accord with previous studies, and, as a result, the notion that projects of the same reference class display similar statistical cost and time profiles is an inaccurate assumption to make, which renders the method of RCF as an estimating method highly questionable (Love et al. 2014; Love et al. 2019b).

Figure 10 presents extracts from a sample of a project’s elemental cost model’s explanation for cost and time deviations, with scope and design changes identified as the main contributors. We also note that escalation and inclement weather had an adverse impact on the cost and time of several projects. In the case of the rural alignment project that experienced a 137% and 133% cost and time deviation, respectively, a considerable amount of redesign occurred due to the uncertainty surrounding geotechnical conditions and the difficulty with acquiring Resource Consents¹¹. Due to the re-design, the project’s completion date had to be extended beyond what was initially planned, and preliminary costs increased.

Figure 10. Examples of contributors to cost and time deviations.

To recap, the data we have presented demonstrates the recurring problem of projects exceeding their expected costs and that estimating guidelines treat the concepts of risk and uncertainty as the same, ignoring their distinct differences. Despite the project cost data being rich in content (Figures 1 and 2), the sample is small. The differences between procurement methods, project types, and cost deviations thus should be treated with caution as the number of projects in each grouping was insufficient for definitive conclusions to be made. Additionally, we did not have access to the initial estimates and the cost contingencies at contract award. Governments and consultants rarely provide access to the actual cost data from inception to completion, so we have had to do with what was provided. In doing so, we have had to rely on abductive reasoning to make inferences based on our knowledge of the cost estimation process and the literature, contributing to answering our research question.

4. Improving the estimation of construction costs

Despite the wealth of research and the tireless efforts of practitioners, the public sector still unequivocally struggles to deliver their projects on budget and time, metaphorically resembling Sisyphus’ journey of pushing the rock up the hill. So, how can public agencies improve the cost estimation accuracy of their transport projects? We answer this question by drawing on the broader literature and the context behind the production and analysis of elemental cost models. In doing so, we hope to help break the Sisyphean loop by suggesting reconceptualizing and recalibrating the treatment and determination of a project’s cost uncertainty. Thus, changes to practice need to be engendered and enacted if strides are to be made in producing more realistic assessments of a project’s actual construction costs.

We now focus on how public agencies can engender meaning into the estimation process and counter the absurdity of accepting the fate that their transport projects will naturally succumb to the ‘Iron Law’. In doing so, we aim to provoke debate and contribute to an alternative way of thinking about decision-making uncertainty over and above the classic view of (unbounded) rationality that pervades cost estimation practice¹².

4.1. Implications for theory

We suggest that public agencies embrace the theoretical lens of ecological rationality, which extends Simon’s (1955, 1956, 1990) concept of bounded rationality, in their decision-making process under uncertainty when it comes to estimating a transport project’s actual construction costs. Equally, this applies to the procurement of all types of infrastructure projects. In stark contrast to unbounded rationality, ecological rationality considers human reasoning and behaviour to be ecologically rational when adapted to the environment in which they act (Todd and Brighton 2016). Accordingly, it aims to identify the environmental properties that permit the simplification of a decision strategy and determine the properties that enable its ease of use and accuracy when information is limited (Hertwig et al. 2022). The decision strategy we refer to here is smart heuristics (Gigerenzer and Brighton 2009; Gigerenzer and Gaissmaier 2011; Gigerenzer, 2014).

Smart heuristics are simple, task-specific decision strategies that form part of a decision maker’s repertoire of cognitive strategies for solving judgement and decision tasks. Such heuristics, relying on limited information, can be applied consciously or unconsciously to make fast and frugal judgments under conditions of uncertainty and limited learning opportunities with high accuracy (Todd and Gigerenzer 2000; Gigerenzer and Gaissmaier 2011; Luan et al., 2019). Even if managers have unbounded cognitive capacities, it has been shown that smart heuristics can make more accurate, efficient, and effective decisions in the real world (Marewski and Gigerenzer 2012; Luan et al., 2019). The accuracy that smart heuristics can provide renders them the decision-making strategy of choice for determining the cost uncertainty of a project’s estimate (Gigerenzer and Goldstein 1996; Gigerenzer et al. 1999; Love et al. 2021b).

As heuristics enable quick decision-making, they have become associated with making errors and an accuracy-speed trade-off is deemed to prevail (Kahneman 2011). Be this as it may, when heuristics are used to inform strategic decision-making, for example, organisations have been shown to experience improvements in profitability and growth (Baum and Wally 2003). Moreover, it has been shown that when managers rely on heuristics for analytical decision-making in project environments, accurate outcomes ensue (West, Acar, and Caruana 2020). In theory, environments subject to VUCA, such as the procurement of transport projects, provide prime conditions to utilise smart heuristics for decision-making under uncertainty, with probability theory only appropriate in risk settings (Gigerenzer, 2014; Gigerenzer, Reb, and Luan 2022).

As shown in Figure 11, we suggest that the risk component of an estimate should be dealt with separately using probability theory. Thus, it follows that smart heuristics would require public agencies to separate their treatment of risk and uncertainty formally and, in doing so, improve an estimate’s transparency. Both the traditional and RCF methods denoted in Figure 11 fall short in estimating a project’s actual construction costs – empirical evidence manifestly and repeatedly demonstrates this to be the case, as we have put forward in this paper. Our ‘balanced approach’ of cost estimate production identified in Figure 11, which we suggest public agencies should consider, embraces the optics of unbounded and ecological rationality.

Figure 11. Unbounded and ecological rationality approaches for cost estimation.

Our proposed approach is theoretically sound and, while practically intractable at this point, it offers a way for public agencies to escape the Sisyphean loop of repeatedly underestimating project costs, conflating uncertainty and risk. Indeed, we have limited understanding and knowledge about the environments where a given heuristic will succeed and in which it will fail (Ika, Love, and Pinto 2022). Knowing what decision strategies are the most effective within varying contexts and situations will facilitate the determination of uncertainty and re-evaluation of risk (Love et al. 2023a, 2023b, 2023c).

A post-completion database and piggybacking on the vicarious experiences of contractors can provide public agencies with knowledge about how smart heuristics can be coupled with an environmental structure to make decisions to help contribute to producing a realistic and accurate assessment of a project’s actual construction costs. As knowledge about ‘what goes right and wrong’ in transport projects accumulates, suitable smart heuristics can be unearthed.

To ensure success and move our understanding of how smart heuristics can be tailored to different epistemic and pragmatic tasks that confront practice in construction, an adaptive toolbox will be required to ensure their generalisability in varying contexts. An adaptive toolbox is the collection of heuristics and building blocks an individual and team have at their disposal for constructing heuristics, together with the core mental capacities that building blocks exploit (Gigerenzer et al. 1999; Gigerenzer, Reb, and Luan 2022).

Within a given adaptive toolbox there would typically be several classes of smart heuristics that enable individuals and teams to make judgments, which include (Gigerenzer and Gaissmair 2011): (1) recognition-based decision-making used to evaluate options based on their identification (e.g., recognition and fluency heuristics); (2) one-reason (sequential) decision-making (e.g., one-clever-cue heuristics, and take-the-best); (3) trade-off (equal weighting) decision-making where simple 0/1 or equal-eights are forgone to reduce estimating error (e.g., tallying, and 1/N rule); and (4) satisficing, where the first option is chosen to meet an aspiration level (i.e. information consideration does not follow an ordering sequence).

As mentioned, uncertainty ranges from ‘resolvable’ to ‘radical’, whereby the former can be removed with more or better information, but the latter cannot. Multiple unknowns, obscurity, ambiguity, and vagueness with unknown and unknowable outcomes characterise radical uncertainty (Vis 2024). So, as smart heuristics are identified and developed from actions and situations arising in projects and applied to accommodate uncertainty, we can equally see information being accumulated, enabling some uncertainties to be resolved and unknown outcomes to become knowable. Resolvable uncertainty can, therefore, be mitigated as information and knowledge become available, allowing the elicitation of subjective (or frequencies) rather than discrete probabilities for phenomena to be garnered from experts (Blass, Lach, and Manski 2010). Undoubtedly, practitioners will hold some subjective beliefs about cost estimation. Still, with incomplete information, there is a proclivity for them to make things up and thus overlook the reality of practice (Fischoff et al. 1999).

4.2. Implications for practice

Fields such as business, finance, medicine, law, and psychology have benefitted from utilising smart heuristics when making decisions under uncertainty (Gigerenzer 2008; Mousavi and Gigerenzer 2014; Neth, Meder, and Gigerenzer 2014; Luan et al., 2019). However, heuristics have received limited attention within the context of procuring transport projects and, more broadly, infrastructure asset delivery, but their potential use in practice is boundless (Love, Ika, and Pinto 2023a;2023b).

Transport agencies often outsource the production of cost estimates (including the contingency) for their projects to an external consultant. A team of quantity surveyors or cost engineers with the expertise and knowledge of estimating practice will prepare a cost estimate based on the design and specification provided by a transport agency. There is an explicit expectation that the consultancy organisation adheres to its procedures and guidelines for producing, reviewing, and submitting cost estimates. Moreover, cost estimates are subjected to independent peer review (e.g., Waka Kotahi NZ Transport Agency. 2023) making improbable the likelihood of optimism bias manifesting in a cost estimate. In the 39 projects we analysed, the cost estimates were prepared per the transport agency’s cost estimation manual; however, uncertainty was reduced to risk, with the contingency being insufficient to accommodate subsequent changes that materialised in the projects. Apart from assigning arbitrary probabilities to possible events, no attention was given to formally assessing uncertainty.

A cogitative and inquisitive mindfulness is typically required to assess uncertainty to ‘anticipate what might go wrong’ during construction and cause cost increases and delays. Determining the unknown requires understanding a project’s context and attention to past situations where cost and time deviations from those of a similar reference class occurred (Love et al. 2023c). Likewise, an understanding of ‘what went right’ and the best practices used to deliver projects successfully is de rigueur. Thus, practitioners’ experience and knowledge from previous projects are needed to garner such an understanding and identify best practices. If strides are to be made to adopt smart heuristics, then the public sector needs to: (1) better learn from failure and success to manage uncertainty so that an adaptive toolbox of heuristics can be established; and (2) embrace relational contracting so that risks and uncertainties can be shared, better identified, and given greater consideration.

4.2.1. Learning from failure and success to manage uncertainty

Learning from failure and success provides a basis to unearth the intricacies, fragilities, and nuances of practice that may be taken for granted when making decisions though they can significantly impact project outcomes (Ika and Pinto 2022). A case in point is the repeated conflation of risk and uncertainty, a leitmotiv of practice that is theoretically questionable in our view. A forensic examination of a project’s cost data will invariably reveal that its contingency is underestimated (Baccarini and Love 2014) not because of optimism bias¹³ but rather due to its uncertainty being ignored or merely reduced to risk (Love, Ika, and Ahiaga-Dagbui 2019a).

Cynically, it can be argued that a project’s viability will be questioned when its uncertainty is realistically assessed and added to an estimate. However, this is a myopic view, as the actual cost for work will not be known until a contractor submits a price based on the information provided in the tender documentation and it is agreed upon. Requesting changes after a contract has been agreed upon is costly for the public sector and increases costs; it is not unusual for a contractor to submit a low bid and over-charge for scope and change orders to increase their profit margin. Conventional procurement methods, such as traditional and D&C forms, provide no incentive for project participants to minimise the cost impacts of subsequent changes. Instead, ‘they allow the incumbent participants to charge ‘monopoly’ prices for the additional work’ (Hayford 2018, 3).

Feedback from completed project costs, such as the elemental cost models produced by cost consultants and used in this research, provides a point of reference for public agencies to learn and formulate new estimates and simultaneously improve their ability to assess the risks that can influence actual construction costs. The creation of a post-completion database, as suggested previously, accessible by all transport agencies, containing details about scope and change orders and their justification, agreed actual start and finish dates, actual costs, and progress against performance indicators would, for example, also help improve the accuracy of estimates, though, at the moment, it does not exist (Terrill and Danks 2016; Liu et al. 2021; Love et al. 2023c). Should a database be created, collaboration with the private sector (e.g., contractors and consortiums) is needed to populate and maintain its currency. Still, the absence of incentives (e.g., tax relief) and mechanisms (e.g., regulatory reform) to facilitate information sharing between the public and private sectors hinders progress towards its development (Love et al. 2021a).

4.2.2. Embrace relational contracting

Changing tack, we now suggest public agencies actively pursue relational contracting principles to improve the cost performance of their transport projects. This suggestion has been repeatedly made for decades as a result of an array of independent reviews and calls from industry bodies worldwide. Still, relational procurement has yet to form the mainstay of the public sector’s procurement strategy for transport projects (Love, Pinto, and Ika 2022). Conventional procurement methods are preferred for the cost certainty, simplicity, and risk transfer they provide to public agencies (Hayford 2018). Yet, as reiterated in our results, we know cost certainty is simply a fallacy, yet these continue to be the procurement methods of choice by many public agencies worldwide.

As we mentioned above (Section 2.2), the reliability of a project’s target outturn cost can generally be better guaranteed under an alliancing arrangement than other procurement forms (Walker, Harley, and Mills 2013; Love, Pinto, and Ika 2022). But why and how? We must reiterate that our study found no statistical difference between procurement methods and mean cost and time deviations, though the sample is small (Figures 8 and 9). Thus, we are not concerned with the procurement method per se but the process of generating a robust, transparent, and reliable target outturn cost within an alliance, as it significantly differs from the conventional approach.

Like an alliance, ECI allows for collaborative input into the design from a contractor, enabling them to provide suggestions to de-risk a project’s scope, but their influence on determining actual construction cost estimates is limited. In the four projects analysed, scope and design changes (e.g., specifications, such as the example in Figure 10) were required, increasing costs and causing delays.

In short, an alliance is underpinned by a Project Alliance Agreement (PAA) and a governance system that enables collaborative behaviours (e.g., open communication, no blame, respecting diversity, and ensuring a ‘best-for-project’¹⁴ mindset) to be maintained and managed. Moreover, including a financial reward mechanism to support ‘best-for-project’ decision-making provides an incentive for both the public and private sectors to work in unison to share knowledge and produce realistic assessments of target outturn cost and target adjustment events (i.e. risks and uncertainties) (Walker, Vaz-Serra, and Love 2022). The collective learning environment of an alliance creates an environment where the project team focuses on sharing their knowledge and experiences to reduce uncertainty. The collaborative environment inter alia facilitates dialogue and trust and confidence in applying smart heuristics. The process of generating a target outturn cost and the willingness to treat uncertainty as a distinct entity can help an alliance better plan and manage a project.

Despite embracing a process of anticipation, it was surprising to uncover that the two alliance projects in our sample experienced 25% and 65% cost deviations, significantly larger than the samples x̅ = 18.28% (Figure 6). They were, however, delivered early and on time. Regardless, there will always be exceptions to the rule, even though the prevailing evidence indicates that alliances offer superior cost performance outcomes than other procurement methods (Weigle and Garber 2010; Walker, Harley, and Mills 2013; Walker and Lloyd-Walker 2015; Almaaz, Zakaria, and Odimegwumas 2020; Walker and McCann 2020; Walker, Vaz-Serra, and Love 2022).

If we take, at face value, the cost deviations for the alliance projects, for example, it would be rational to assume that they had mis-performed. However, alliances are typically used for complex projects with an ill-defined scope (Department of Infrastructure & Regional Development 2015a); thus, vicissitudes are expected to arise. In both projects, significant changes to the scope were required due to complex ground conditions. One project required increased motorway length and additional in-situ road bridges (including bored piles), and the other required a level crossing and road bridges. Comprehending the context behind the number is important before a conclusion can be drawn.

Accordingly, we need to take heed and seek to understand and learn from the environment, situations, and set of actions that cause projects to mis-perform (Ika and Pinto 2022; Love, Ika, and Pinto 2023a)—and in doing so, building resilience to the unexpected. However, there is a mismatch between the unbounded view of rationality utilised in alliances and, more generally, by public sector agencies for decision-making in the VUCA environment within which they are made. The acceptance of unbounded rationality by public agencies, like Sisyphus recognising his fate, has a certain formality, as the eternal struggle against the fate of the ‘Iron Law’ encourages the adoption of statistical methods to deal with unknowns when they are seemingly inevitable.

Unlike Sisyphus, who acquiesced his fate and remained content with never reaching his goal, we believe the public sector cannot continue to accept the absurdity of normalising transport project mis-performance. Treating uncertainty in cost estimates with a probabilistic lens has contributed to this situation. As noted earlier, an alternative view to unbounded rationality is required to deal with uncertainty, which we hope public agencies will consider in their guidelines for cost estimation.

5. Conclusion

Despite the sheer volume of research that has sought to explain why and how transport projects experience cost deviations, they remain an ever-present reality for the public sector and taxpayers. Tools such as RCF – based upon applying probabilistic-based uplifts to a cost estimate and contingency – have been developed and adopted by some public agencies to de-risk projects as there is a belief that optimism bias and strategic misrepresentation are the root causes of cost overruns. However, optimism bias has never been empirically shown to exist, and only a few instances of strategic misrepresentation have been anecdotally identified. Thus, the justification for adding an uplift to a project cost estimate and contingency and thus reducing uncertainty to risk is flawed.

Reconceptualizing the treatment of risk and uncertainty is needed if strides are to be made to better estimate a project’s actual construction costs. In doing so, we have taken an absurdist view of transport cost performance in this paper using the Sisyphean loop as a metaphor to describe the public sector’s tireless effort to deliver their projects within budget while knowing that this may be a meaningless task and accepting the inevitability of a cost overrun. Repeatedly conflating risk and uncertainty and using probabilistic approaches to estimate a project’s construction costs contribute to trapping public agencies into hoping they can deliver their projects on budget.

We have suggested that the combination of ambiguity aversion, an illusion of certainty, and the repeated occurrence of cost deviations, as noted in Table 1 and supported with empirical data from a small sample of 39 projects procured by a transport agency, reinforces the Sisyphean loop, rendering it impossible to break it, considering the prevailing conceptualisation and treatment of risk and uncertainty. Changes in design and scope were the leading causes of cost deviations, not optimism bias and strategic misbehaviour. In fact, the cost estimates and final elemental cost models for the examined transport projects were produced by external consultants and vetted by independent reviewers. However, the allocated cost contingency was insufficient at the contract award by an average of 13.28%. Thus, we have suggested that changes to practice are required as circuit breakers to enable public agencies to: (1) harvest knowledge from developing a post-completion database, (2) acquire vicarious experiences of contractors through collaboration and relational procurement; and (3) reconceptualize and recalibrate their treatment of uncertainty through ecological rationality lens, where smart heuristics become the apex strategy for decision-making.

We acknowledge that our practice recommendations are speculative and that research is required to develop a post-completion database’s format, structure, and content and determine how it can be utilised effectively during the cost estimation process. Similarly, knowledge about the type and nature of smart heuristics that can inform decision-making within varying environmental structures is lacking. As a result, future work is required to fill this void.

Using the Sisyphean task as a metaphor, we have put forward a different and untapped line of inquiry that may explain why projects experience cost overruns. To this end, the main contributions of our paper are twofold: (1) the analysis demonstrates that the cost profiles of projects do not vary and reconfirms that vicissitudes contribute to cost deviations, not optimism bias and strategic behaviour (2) a new conceptual framing by taking a balanced approach to examine risk and uncertainty in a cost estimate by combining unbounded and ecological rationality is presented.

Data availability statement

All data supporting this study’s findings are available in the attached supplementary file and from the corresponding author upon request.

Disclosure statement

The authors report that there are no competing interests to declare.

Additional information

Notes on contributors

Peter E. D. Love

Peter E. D. Love received a Ph.D. in operations management from Monash University, Melbourne, Australia, in 2002. He is a John Curtin Distinguished Professor with the School of Civil and Mechanical Engineering, Curtin University. His research interests include operations and production management, resilience engineering, infrastructure development and digitisation in construction. His research has been published in leading scholarly journals such as the European Journal of Operations Research, Journal of Management Information Systems, Journal of Management Studies, IEEE Transactions on Engineering Management, International Journal of Operations and Production Management, Production Planning and Control, and Transportation Research A: Policy and Practice. Professor Love holds a Higher Doctor of Science (Curtin 2012) for his contributions to the field of civil and construction engineering.

Jane Matthews

Jane Matthews received a Ph.D. in architecture from the University of East London, London, U.K., in 2001. She is a Professor of Digital Construction at the School of Architecture and Built Environment, Deakin University, Victoria, Australia. She has ten years of industry experience as a Software Design and Development Manager for the Royal Institute of British Architects. Professor Matthews has published extensively in leading scholarly journals, which include Automation in Construction, ASCE Journal of Construction Engineering and Management, Production Planning and Control, Transportation Research A: Policy and Practice, and IEEE Transactions on Engineering Management and Reliability Engineering and System Safety. Her research interests include the management and visualisation of information in construction.

Lavagnon A. Ika

Lavagnon A. Ika received a Ph.D. in project management from the Université du Quebec, QC, Canada, in 2011. He is a professor of project management and the program director for the M.Sc. in management at the Telfer School of Management, University of Ottawa, Ottawa, ON, USA. He is affiliated with the School of International Development and Global Studies, University of Ottawa. He has a keen interest in international development projects. His work has appeared in leading journals such as World Development, the IEEE Transactions on Engineering Management, Production Planning and Control, Transportation Research Part A: Policy and Practice, the International Journal of Project Management, the Project Management Journal, the European Management Review, and the Journal of African Business. His research interests include: what makes projects complex; what makes projects successful; why do projects fail and what can be done about it; why projects experience cost overruns and benefit shortfalls; how do projects really “behave” or work; and what is the role of strategy, supervision and management in project success/failure.

Notes

1 Reference class forecasting is a method of predicting the future by looking at similar past situations and their outcomes (Flyvbjerg 2008).

2 For example, refer to the UK’s HM Treasury’s (2022) “Greenbook” and Waka Kotahi NZ Transport Agency’s (2023) “Cost Estimation Manual (SM014)”, to name a few.

3 Ambiguity aversion is the tendency to favor the known over the unknown, including known risks over unknown risks. So, in this case, public sector agencies will be uncomfortable with situations in which there is incomplete or vague information about the probabilities associated with events impacting a project’s actual construction costs.

4 For example, drift deposits and the disintegrated brecciated [sedimentary rock type] bedrock.

5 Refer to “HS2: The £100BN Railway Dividing a Nation” Available at: https://www.youtube.com/watch?v=FSD5ps9bLQ0.

6 At target outturn cost is “a specific sum estimated as being the pre-estimate of Reimbursable Costs, Risk & Contingency Provisions, Corporate Overhead and Profit to achieve the minimum conditions of satisfaction, perform the Works and bring the Works to final completion (i.e., following completion of the defects liability period)” (Department of Infrastructure & Regional Development 2015a: p.12).

7 Strategic misrepresentation “is the planned, systematic distortion or misstatement of fact-lying-in response to incentives in the budget process” (Jones and Euske 1991: p.174).

8 For further information refer to “Secret recordings expose multibillion-pound HS2 cover-up”, Times Radio, 24^th October 2023, available at “https://www.youtube.com/watch?v=O77wvGRSdSM”.

9 Several cost estimate classifications exist such as produced by the: (1) American Society of Professional Estimators (2021) Level 1 – Order (Range) of Magnitude, Level 2 – Schematic/Conceptual Design, Level 3- Design Development, Level 4 – Construction Document, and Level 5 – Bid; and (2) Association for the Advancement of Cost Engineering (2020) which uses the following definitions to express the level of project’s definition– Class 5 (0-2%), Class 4 (1 to 15%), Class 3 (10 to 40%), Class 2 (30-70%) and Class 1 (50-100%).

10 A contract where the amount to be paid for the work carried out is based on the remeasurement of work after it has been carried out, usually against a schedule of rates contained in the contract.

11 In New Zealand a Resource Consent is permission from the local council for an activity that might affect the environment.

12 The classic view of rationality views reasoning and behaviour as rational when they conform to the norms of logic, statistics, and probability theory (Marewski and Gigerenzer 2012).

13 We do not discount the presence of optimism bias, quite the contrary. However, there is a lack of empirical evidence of its presence within the formulation of cost estimates developed by a group of people and checked by independent organizations.

14 Best-for-project refers to the “determination, decision, outcome, solution, or resolution which is consistent with the aim of ensuring that the works are fit for their intended purpose and constitute value for money to the State, consistent with the stated objectives of an alliance rather than the achievement of self-serving goals (such as maximising a Non-Owner Participants’ (NOP) profit)” (www.lawinsider.com/dictionary/best-for-project) (Department of Infrastructure& Regional Development, 2015b).