The Pathology of Infrastructure Megaprojects and the Contribution of Pre-Decision Governance: Designing Institutional Architecture to Correct Institutionalized Epistemic Distortions
ABSTRACT
Infrastructure megaprojects across various countries consistently exhibit similar patterns of failure: cost overruns, delays, and benefits falling short of predictions. Cross-national studies show that between 80-90% of megaprojects experience cost overruns (Flyvbjerg, 2017; Ansar et al., 2014). Existing literature has long documented this phenomenon and identified its causes: optimism bias and strategic misrepresentation. However, this article argues that this literature has yet to answer a more fundamental question: why do these distortions persist despite being widely known and despite various correction efforts having been undertaken?
The novelty of this article lies not in documenting cost overruns, but in theorizing their persistence as a stable institutional equilibrium resistant to moral or technical correction. Drawing on game theory frameworks, we show that epistemic distortions persist because it is strategically rational for each actor within the existing incentive structure—project proponents, bureaucrats, and politicians—to maintain strategies that produce overly optimistic estimates, even though collective outcomes are suboptimal. This model does not assume complete information or fixed preferences. Even under conditions of incomplete information, if actors observe short-term reward patterns (project approval, secure careers) and weak long-term sanctions (weak audits, passive media, leadership turnover), then optimistic strategies remain the best response. Thus, this equilibrium can be understood as a coordination equilibrium reinforced by career incentives and political pressures, not merely as a static payoff matrix.
The Pre-Decision Governance (PDG) framework is proposed as a middle-range theory with clear scope conditions. This article compares PDG with existing governance instruments, demonstrating its uniqueness in creating credible temporal coupling and epistemic contestability—the institutionalized capacity to effectively challenge epistemic claims. Epistemic contestability differs from ordinary transparency, adversarial legalism, devil's advocate procedures, institutional review boards, and internal audit challenge processes because it explicitly combines: (1) mandatory dissent documentation at the pre-decision phase, (2) mandatory response mechanisms that create accountability traces, and (3) traceable future consequences through temporal coupling with ex post review. It is not merely an ad hoc challenge procedure, but an institutional infrastructure designed to transform actors' incentive structures.
PDG operates through three causal pathways. The detection effect increases the probability of error identification through assumption testing and dissent documentation. The behavior modification effect transforms incentive structures by creating documented reputational costs: if an official ignores valid dissent and the project subsequently fails, the documentation trail becomes grounds for career evaluation, media investigation, or even legal prosecution. The selection effect is positioned as a consolidation mechanism following initial change, differing from ordinary bureaucratic turnover by explicitly linking recruitment and promotion criteria to documented responsiveness to dissent.
This article also integrates the political settlements framework (Khan, 2018) to explain that PDG's effectiveness depends on power structures, with the proposition that PDG is more likely to succeed in competitive clientelism regimes compared to dominant patronage regimes. PDG does not assume voluntary adoption by dominant elites; it becomes viable when competing factions perceive contestability as a strategic tool to weaken political opponents or build reformist reputations. Two case reinterpretations—Big Dig and Flamanville—are presented as analytical illustrations, with additional support from similar patterns in Crossrail and Olkiluoto 3 to demonstrate pattern consistency. These cases were not selected for variance testing, but for pattern illustration across institutional contexts. These illustrations are not intended as counterfactual proof that PDG will necessarily succeed, but as heuristic demonstrations of intervention points that could theoretically shift equilibrium. This article explicitly acknowledges potential endogeneity problems and emphasizes that the proposed propositions are conditional, requiring further empirical testing with designs that account for selection. Three research propositions are advanced, accompanied by potential operational indicators. The article concludes that PDG is not an automatic solution; its effectiveness depends on context and requires gradual adoption strategies through layering mechanisms triggered by exogenous factors such as crisis, the entry of reform champions, or strategic use by one faction to weaken another in political competition.
Keywords: megaprojects, infrastructure, Pre-Decision Governance, epistemic distortion, institutional equilibrium, Nash equilibrium, coordination equilibrium, cost overrun, political settlements, temporal coupling, epistemic contestability, polycentric governance, layering, endogeneity, repeated game, bounded rationality, incomplete information, adversarial legalism, devil's advocate
1. INTRODUCTION
Infrastructure megaprojects have become a primary development instrument in many countries. Yet empirical evidence shows that these projects systematically fail to meet promised cost, time, and benefit targets. Flyvbjerg's (2017) comprehensive study of over 2,000 projects across 104 countries found that 90% of megaprojects experience cost overruns. Other studies confirm similar patterns: the Channel Tunnel (80%), Sydney Opera House (1,400%), Boston's Big Dig (220%), and London's Crossrail (from £14.8 billion to over £18 billion) (Flyvbjerg et al., 2003). Ansar et al. (2014) add that large dam projects in India and China show different success rates due to variations in institutional capacity. This pattern is also evident in European nuclear projects: Flamanville 3 in France ballooned from €3.3 billion to over €19 billion, while Olkiluoto 3 in Finland increased from €3 billion to €8.5 billion (Grall, 2020).
Existing literature has long documented this phenomenon. Flyvbjerg (2008) identifies two primary causes: optimism bias (psychological bias) and strategic misrepresentation (strategic manipulation). Kahneman and Tversky (1979) explain the cognitive biases underlying optimism, while Merrow (2011) shows how feasibility studies often become tools to justify already-made decisions. Heikkila and Gerlak (2019) add that early warnings are frequently ignored in decision-making processes.
However, this article argues that this literature has yet to answer a more fundamental question: why do these distortions persist despite being widely known and despite various correction efforts having been undertaken? Why have interventions such as cost-benefit analysis guidelines, stage-gate reviews, or even reference class forecasting failed to alter patterns that have persisted for decades?
The novelty of this article lies not in documenting cost overruns, but in theorizing their persistence as a stable institutional equilibrium resistant to moral or technical correction. Drawing on game theory frameworks, we show that epistemic distortions persist because it is strategically rational for each actor within the existing incentive structure—project proponents, bureaucrats, and politicians—to maintain strategies that produce overly optimistic estimates, even though collective outcomes are suboptimal.
With this diagnosis, we propose the Pre-Decision Governance (PDG) framework as a middle-range theory with clear scope conditions: (1) strategic public sector decisions (not routine operational decisions), (2) high technical complexity and uncertainty, (3) potential for large financial and social impacts, and (4) the existence of institutional structures that can be modified gradually.
The article is structured as follows. Section 2 elaborates the diagnosis of epistemic distortion as institutional equilibrium. Section 3 compares PDG with existing governance instruments. Section 4 develops the PDG framework and its causal mechanisms. Section 5 presents case reinterpretations as analytical illustrations. Section 6 discusses theoretical contributions, limitations, and research propositions. Section 7 contains conclusions.
2. ROOT CAUSE: EPISTEMIC DISTORTION AS INSTITUTIONAL EQUILIBRIUM
2.1 Four Well-Known Contributing Factors
Cognitive psychology research shows that humans tend to be overly optimistic (Kahneman & Tversky, 1979). In project planning, this bias manifests as planning fallacy (Kahneman, 2011). More problematic is strategic misrepresentation (Flyvbjerg, 2008). Critical assumptions are rarely independently tested; feasibility studies often justify already-made decisions (Merrow, 2011). Early warnings are frequently ignored (Heikkila & Gerlak, 2019).
These four factors—optimism bias, strategic misrepresentation, absence of assumption testing, and disregard of dissent—are widely recognized in the literature. Yet what remains unexplained is: why do these four factors persist despite our awareness of their existence?
2.2 Equilibrium Logic: From Nash to Coordination Equilibrium
The answer lies in the incentive structure that makes these distortions individually rational. Consider the position of each actor:
| Actor | Strategy | Payoff if Compliant | Payoff if Deviating | Incentive |
|---|---|---|---|---|
| Project proponent | Optimistic estimates | Project approved, profit | Conservative estimate → project rejected | Compliant dominant |
| Bureaucrat | Not questioning assumptions | Secure career, avoiding conflict | Questioning → isolation/conflict risk | Compliant dominant |
| Politician | Promoting large projects | Short-term political achievement | Delaying → losing political momentum | Compliant dominant |
The following payoff matrix illustrates this logic (numbers are illustrative only):
In this matrix, (O,T) is a stable outcome. The project proponent will not switch to conservative estimates because that would lower their payoff (from 10 to 2) if the bureaucrat remains non-testing. The bureaucrat will not switch to testing because that would lower their payoff (from 5 to 0 or -2) if the proponent remains optimistic.
It is important to clarify the assumptions underlying this model:
First, this model does not assume complete information or fixed preferences. Even under conditions of incomplete information, if actors observe short-term reward patterns (project approval, secure careers, political achievements) and weak long-term sanctions (weak audits, passive media, leadership turnover), then optimistic strategies remain the best response. They need not know payoffs precisely; it suffices to observe that "those who play it safe" (with conservative estimates or daring to question) tend to be marginalized, while optimists receive projects and promotions.
Second, interactions are repeated games, not one-shot. These actors will interact repeatedly across various projects, so long-term reputation should theoretically matter. However, the future discount rate is very high due to short-term political pressures (election cycles, annual targets) and career uncertainty. Consequently, the reputational penalty for optimistic behavior remains weak—short-term benefits from approving projects outweigh long-term reputational risks.
Third, actors have bounded rationality (Simon, 1947)—they cannot process all relevant information, but they understand reward patterns based on experience and prevailing norms in their environment.
With these assumptions, the resulting equilibrium can be understood as a coordination equilibrium reinforced by career incentives and political pressures. All actors "know" that the established way of playing is individually safest, and no one has an incentive to deviate from this pattern alone. If one actor attempts to deviate (for example, a bureaucrat questioning assumptions), they face isolation risk with no guarantee that others will follow. In simpler terms: epistemic distortions persist not because actors are evil or stupid, but because they respond rationally to incentives within the existing structure, and that structure itself results from all actors' implicit coordination.
Crucial implication: Changing this equilibrium requires simultaneous transformation of the incentive structure, not merely moral appeals or technical corrections targeting only one aspect.
3. PDG IN RELATION TO EXISTING GOVERNANCE INSTRUMENTS
3.1 Existing Governance Instruments
Before elaborating PDG, it is important to systematically compare it with widely used governance instruments:
| Instrument | Primary Function | Area of Improvement from Epistemic Perspective |
|---|---|---|
| Stage-gate review | Divides projects into stages with formal decision points | Focuses on financial and technical feasibility, does not explicitly target assumption testing or dissent documentation |
| Independent review board | Provides independent assessment at critical stages | Quality highly dependent on member selection; no standard mechanism for temporal coupling |
| UK Treasury Green Book | Cost-benefit analysis guidelines for public policy | Emphasizes options analysis but does not institutionalize structured dissent or epistemic contestability |
| Infrastructure Australia Gateway Review | Structured review process for infrastructure projects | Similar to stage-gate, focuses more on risk management than pre-decision reasoning quality |
| Reference class forecasting (Flyvbjerg) | Statistical technique to correct optimistic estimates using historical data | Valuable as a correction tool, but does not endogenously transform actor incentive structures |
PDG does not replace reference class forecasting. Rather, it complements it: reference class forecasting provides statistical ex post correction, while PDG creates institutional incentives for more accurate ex ante estimates. Both can work synergistically.
3.2 Positioning Epistemic Contestability: Sharpening Boundaries
To avoid criticism that epistemic contestability is merely "old wine in new bottles," it must be explicitly positioned in relation to existing challenge mechanisms:
| Mechanism | Characteristics | Difference from Epistemic Contestability |
|---|---|---|
| Adversarial legalism (Kagan, 2001) | Contestation through litigation, focus on legal rules, reactive, costly | Epistemic contestability is proactive, integrated into planning cycles, targets assumptions not legal violations |
| Devil's advocate procedures | Ad hoc roles to challenge proposals, often without mandatory documentation | PDG requires documentation, written responses, and creates traces for future evaluation |
| Institutional review boards (IRBs) | Focus on research ethics, human subject protection, does not target policy assumptions | PDG targets public policy assumptions, integrated with procurement cycles, has temporal coupling mechanisms |
| Internal audit challenge processes | Focus on compliance and internal control, often reactive, limited documentation | PDG is proactive at the pre-decision phase, creates traces traceable by external audit |
Epistemic contestability differs from all these mechanisms because it explicitly combines: (1) mandatory dissent documentation at the pre-decision phase, (2) mandatory response mechanisms that create accountability traces, and (3) traceable future consequences through temporal coupling with ex post review. It is not merely an ad hoc challenge procedure, but an institutional infrastructure designed to transform actors' incentive structures.
3.3 PDG's Unique Position
PDG differs from the above instruments in three fundamental respects that simultaneously affirm this article's novelty:
First, PDG institutionalizes epistemic contestation at the pre-decision phase as micro-institutional design. Unlike stage-gate or independent reviews focusing on technical-financial feasibility, PDG explicitly targets assumptions, framing, and dissent as governance objects. This shifts the question from "is the project feasible?" to "how do we know this project is feasible and what assumptions underlie it?"
Second, PDG creates credible temporal coupling between ex ante reasoning and ex post evaluation. Assumption and dissent documentation is not merely archived but designed to be traceable by future review mechanisms (audit institutions, media, parliaments). This temporal coupling is absent in other instruments.
Third, PDG explicitly distinguishes itself from conventional deliberative governance. Deliberative governance focuses on public forums and consensus-building through argumentation (Habermas, 1996). PDG, by contrast, is embedded in the technical design pipeline, producing technical artifacts (assumption sheets, model portfolios) that become procurement requirements, and operates through institutional incentives created by reputational shadow, not through public participation.
With these three distinctions, PDG offers something absent in both literature and practice: a micro-institutional framework to correct epistemic distortions at the most upstream phase, with explicit causal mechanisms accounting for variations in power structures.
4. THE PRE-DECISION GOVERNANCE FRAMEWORK
4.1 Definition and Scope Conditions
PDG is a set of institutional protocols to correct epistemic distortions by strengthening pre-decision reasoning quality. As a middle-range theory (Merton, 1968), PDG applies under the following scope conditions:
- Strategic public sector decisions (not routine operational decisions)
- High technical complexity and uncertainty
- Potential for large financial and social impacts
- Existence of institutional structures that can be modified gradually
4.2 Four Protocols
4.2.1 Assumption Testing
- Explicit documentation of assumptions (cost, time, benefits, risks)
- Sensitivity testing with historical data (Flyvbjerg, 2008)
- Independent expert validation
- Contingency scenarios
4.2.2 Problem Reframing
- Is the project the right response?
- Compare non-project alternatives
- Engage stakeholders with different framings (Rein & Schön, 1993)
4.2.3 Multi-Option Mandate
- At least three options with different risks and costs
- Explicit trade-off analysis (Keeney & Raiffa, 1993)
4.2.4 Structured Dissent
- Independent team (red team)
- Documentation of all dissent
- Mandatory written responses
- Protection for dissenters (Sunstein & Hastie, 2015)
4.3 Causal Mechanisms and Incentive Transformation
PDG operates through three interrelated causal pathways. Most importantly, PDG is explicitly designed to transform the incentive structure that has sustained epistemic distortions.
4.3.1 Detection Effect
Assumption testing and dissent documentation protocols increase the probability that errors are identified before resource commitment. Assumptions are forced through external verification thresholds, making overly optimistic assumptions more likely to be detected. This is the most direct and readily achievable effect.
4.3.2 Behavior Modification Effect: Transforming the Payoff Matrix
This is the core of incentive transformation. PDG transforms the payoff matrix through two pathways:
First, PDG increases the expected cost of "ignoring dissent" or "not testing assumptions." When an official ignores valid dissent and the project subsequently fails, the documentation trail generated by PDG becomes grounds for:
- Career evaluation by superiors or oversight institutions
- Media investigation that can damage reputation
- Legal prosecution or audit by state institutions
- Political attacks from opposing factions in competitive regimes
In other words, the payoff for the "ignore dissent" strategy is no longer (5) as in the initial matrix, but becomes (5 - documented reputational risk). The more credible the ex post review mechanisms (independent audit institutions, active media, responsive parliaments, factional competition), the greater this risk.
Second, PDG reduces the risk of "testing assumptions" or "voicing dissent" strategies. By providing formal protection and ensuring that dissent must receive written responses, PDG reduces the isolation or conflict risk previously faced by bureaucrats questioning assumptions. The payoff for "testing" strategies increases from (-2) to (0 or even positive) by providing reputational protection: if the project subsequently fails, those who voiced dissent have evidence that they warned.
With these changes, the payoff matrix shifts. The (O,T) strategy is no longer the only stable equilibrium. Actors begin to have incentives to consider long-term consequences.
However, this behavior modification effect is not automatic. It will only occur if the following conditions are met:
- The existence of credible (independent, consistent, fair) and visible (results published with consequences) audit institutions
- Active investigative media covering major projects
- Responsive parliaments or oversight bodies to audit findings
- In competitive clientelism regimes, factional competition creating incentives for mutual oversight
Without these conditions, the reputational shadow remains weak and behavior unchanged. Lofthouse and Schaefer (2024) emphasize that expert knowledge reliability depends on a competitive epistemic ecosystem; PDG creates such an ecosystem at the micro-level but requires an accountability ecosystem at the macro-level.
4.3.3 Selection Effect: Consolidating Change
The selection effect is positioned as a consolidation mechanism following initial change. Initial change is typically triggered by exogenous factors such as crisis, the entry of reform champions into the dominant coalition, or strategic use of PDG by one faction to weaken another. Once PDG is adopted (for example, through layering), and detection and behavior modification begin showing results, actor selection gradually changes through:
- Career effect: Officials proven to have ignored valid dissent begin facing career consequences.
- Reputational screening: Actors with reputations for courage in voicing dissent become more likely to be recruited or promoted.
- New entry barriers: Dissent protection mechanisms lower barriers for actors with alternative perspectives.
PDG-induced selection differs from ordinary bureaucratic turnover because it explicitly links recruitment and promotion criteria to documented responsiveness to dissent. In ordinary turnover, personnel changes occur due to retirement or routine rotation, without changes in epistemic quality. In PDG, selection is directed to gradually replace actors unresponsive to dissent with actors more open to epistemic contestation.
In other words, the selection effect results from PDG's early success, not its cause. It is a positive feedback loop reinforcing change after equilibrium has shifted.
4.4 Paths Toward Empirical Operationalization
To enable future empirical testing, epistemic contestability requires operationalization. Potential indicators include:
- Number of documented dissents per project, reflecting the extent to which contestation space is utilized.
- Rate of design revisions before resource commitment, indicating the extent to which contestation generates substantive change.
- Variance between initial and final estimates after undergoing assumption testing and dissent processes.
- Actor perception surveys on psychological safety in voicing dissent and credibility of ex post review mechanisms.
- Frequency of red team use and proportion of recommendations adopted.
- Dissent responsiveness rate: proportion of dissent receiving substantive (not merely administrative) responses.
It should be acknowledged that these indicators have not been empirically validated and require further research to test construct validity and inter-rater reliability. However, they are offered as reasonable starting points for future instrument development.
5. ANALYTICAL ILLUSTRATIONS: REINTERPRETING MEGAPROJECT CASES
This section presents reinterpretations of two famous megaproject cases through the lens of the proposed framework. It is important to emphasize that these are analytical illustrations, not empirical tests. These cases were not selected for variance testing, but for pattern illustration across institutional contexts. The aim is to demonstrate how the PDG framework can explain failure dynamics and potential interventions, not to prove its effectiveness.
5.1 Case 1: Big Dig (Boston, USA)
Context: Boston's underground highway widening project, with initial estimate of $2.8 billion (1985) and final cost of $22 billion (Greiman, 2013). Primary causes: continuous design changes, 118 separate contracts difficult to coordinate, and weak oversight.
PDG Reinterpretation:
- Epistemic distortion: Assumptions about technical complexity and coordination capacity were highly optimistic. No mechanism existed to independently test these assumptions.
- Detection effect: Within the PDG framework, an independent red team would likely have identified coordination risks and design changes from the outset.
- Behavior modification effect: If credible temporal coupling existed (for example, GAO or state audit tracking decision traces), project managers might have been more cautious. However, at that time, accountability was weak.
- Selection effect: Initially absent. PDG could have created selection after several years by promoting managers more responsive to risk.
- PDG failure: Elite override might have occurred if the governor at the time pushed the project despite warnings. However, if PDG were institutionalized with protected independent red teams, warnings would be documented and become material for future evaluation.
5.2 Case 2: Flamanville 3 (France)
Context: EPR nuclear reactor project in France, initial estimate €3.3 billion (2007), final cost >€19 billion (Grall, 2020). Main problems: welding defects and substandard valves, disregard of early warnings.
PDG Reinterpretation:
- Epistemic distortion: Assumptions about component quality and industrial capacity were highly optimistic.
- Detection effect: Independent assumption testing would have revealed supply chain weaknesses.
- Behavior modification effect: Weak, because France's hierarchical culture tends to punish bearers of bad news. Temporal coupling requires strong audit institutions (like the Cour des Comptes) genuinely capable of tracing decision footprints.
- Selection effect: Did not occur. PDG could create selection by protecting whistleblowers, but in a patronage culture, resistance is strong.
- PDG failure: This case illustrates elite override (political pressure to complete the project) and selection manipulation (choosing non-critical red team members).
5.3 Similar Patterns in Other Projects
The same pattern appears across various megaprojects, demonstrating phenomenon consistency:
- Crossrail (London, UK): Europe's largest underground railway project experienced cost escalation from £14.8 billion to over £18 billion and 3.5 years delay. A parliamentary investigation revealed that early warnings from technical managers were ignored by the board, and assumptions about London's geological complexity were inadequately tested (National Audit Office, 2019).
- Olkiluoto 3 (Finland): EPR nuclear reactor built by the same consortium as Flamanville experienced cost escalation from €3 billion to €8.5 billion. Audit reports found that assumptions about manufacturing capacity and quality control were highly optimistic, and dissent from field engineers went undocumented (Grall, 2020).
5.4 Reflection: What Do These Illustrations Show?
These illustrations are not intended as counterfactual proof that PDG will necessarily succeed, but as heuristic demonstrations of intervention points that could theoretically shift equilibrium. These cases were selected to show institutional variation (USA, France, UK, Finland) while maintaining the same core pattern: untested assumptions, ignored warnings, undocumented dissent. By varying contexts, we can see that epistemic distortion is a trans-national phenomenon, not a local anomaly.
In each case, we can identify where epistemic distortion occurred and where PDG protocols could be placed to create correction mechanisms. Big Dig demonstrates the importance of testing assumptions about technical complexity. Flamanville shows how dissent protection could prevent disregard of early warnings. Crossrail illustrates how temporal coupling with active parliaments can create accountability. Olkiluoto exhibits the same pattern as Flamanville, confirming the systemic nature of the problem.
In other words, these illustrations serve to demonstrate framework plausibility, not to prove it empirically. Empirical proof requires more rigorous research designs, as elaborated in the propositions and research agenda in the following section.
6. DISCUSSION
6.1 Theoretical Contributions
This article makes four main theoretical contributions:
First, it redefines cost overrun not as an anomaly or ordinary forecasting failure, but as an equilibrium outcome of distorted epistemic institutions. This diagnosis goes beyond psychological (optimism bias) or strategic (misrepresentation) explanations by showing that epistemic distortions persist because they are rational for actors within the existing incentive structure. The novelty of this article lies not in documenting cost overruns, but in theorizing their persistence as a stable institutional equilibrium resistant to moral or technical correction. By formalizing assumptions of repeated games, bounded rationality, incomplete information, and high future discount rates, the coordination equilibrium model becomes analytically defensible.
Second, it introduces the concept of epistemic contestability as the institutionalized capacity to challenge epistemic claims, explicitly distinguishing it from existing challenge mechanisms such as adversarial legalism, devil's advocate procedures, institutional review boards, and internal audit challenge processes. Epistemic contestability is unique because it combines mandatory dissent documentation, response mechanisms with accountability traces, and temporal coupling with ex post review.
Third, it formalizes three causal mechanisms with emphasis on how PDG explicitly transforms incentive structures. The behavior modification effect is explained through increasing the expected cost of ignoring dissent strategies and decreasing the risk of testing assumption strategies—an explicit transformation of the payoff matrix that has sustained equilibrium. This effect depends on explicit conditions: credibility of ex post review mechanisms, which may include independent audit institutions, investigative media, responsive parliaments, or factional competition. The selection effect is positioned as a consolidation mechanism differing from ordinary bureaucratic turnover by linking recruitment and promotion criteria to documented dissent responsiveness.
Fourth, it integrates variation in power structures (political settlements) to explain PDG's conditional effectiveness, while explicitly acknowledging potential endogeneity problems. PDG does not assume voluntary adoption by dominant elites; it becomes viable when competing factions perceive contestability as a strategic tool to weaken political opponents or build reformist reputations.
6.2 Political Settlements and Typology
Following Khan (2018), PDG effectiveness strongly depends on political settlement type. Khan distinguishes at least two relevant ideal types:
- Dominant patronage regime: Power concentrated in one dominant coalition. Rent distribution concentrated, opposition weak. In such regimes, resistance to PDG will be very strong as it threatens elite rent bases. Elite override and actor selection manipulation are highly likely. PDG is unlikely to succeed without fundamental change in the dominant coalition.
- Competitive clientelism: Multiple factions compete for power, rent distribution more dispersed. In such regimes, reformist actors have maneuvering room and can build supporting coalitions. Natural checks and balances emerge from inter-faction competition. PDG is more likely to succeed because it can be strategically used by one faction to criticize another, creating transparency incentives.
This typology clarifies Proposition 3: PDG effectiveness is mediated by political settlement structure, and is more likely to succeed in competitive clientelism compared to dominant patronage regimes.
6.3 Over-Institutionalization Risks and Mitigation
A common criticism of procedural reforms is that they can add veto points and cause paralysis by analysis. PDG also faces this risk. Strategic actors can manipulate the system through:
- Formal compliance: Meeting protocols literally but ignoring substance.
- Procedural capture: Filling key positions (e.g., red team) with people who will not challenge the status quo.
- Delay tactics: Using protocols to slow down undesired projects.
However, PDG is designed to mitigate these risks through:
- Layering, not overhaul: PDG does not add new veto points but shifts power balances within existing forums. Layering is chosen because it is more easily adopted in institutional contexts resistant to radical change; it adds new procedures without necessarily removing old structures protected by powerful actors.
- Structured dissent, not veto: The red team lacks authority to stop projects, only the obligation to be heard and receive responses.
- Transparency and traces: Public documentation makes manipulation more difficult.
Nevertheless, in practice, these mitigations do not always succeed. Therefore, PDG effectiveness remains conditional on context.
6.4 Acknowledging Endogeneity: A Value-Added
One serious technical criticism of causal propositions is the endogeneity problem: PDG may be more likely adopted by institutions already possessing better governance capacity. In other words, project success might result from pre-existing capacity, not from PDG itself.
Explicit acknowledgment of this problem is rare in conceptual papers and constitutes a value-added:
"Propositions P1–P3 do not claim simple causal relationships. In practice, PDG may be more likely adopted by institutions already possessing higher governance capacity (selection bias). Therefore, future empirical testing must control for variables such as initial institutional capacity, previous project experience, and bureaucratic quality. Ideal research designs would employ difference-in-differences approaches or instrumental variables to address endogeneity. However, for theory development purposes, these propositions are offered as hypotheses requiring further testing, not as established causal claims."
6.5 Research Propositions
P1: Projects with independent assumption testing produce more conservative estimates and lower cost overruns, after controlling for complexity, institutional context, and initial institutional capacity.
P2: Structured dissent increases the probability of design revisions before resource commitment, with stronger effects in systems with credible ex post review mechanisms (independent audit institutions, active investigative media, responsive parliaments, or factional competition within political settlements).
P3: PDG effectiveness is mediated by political settlement structure, with stronger effects in competitive clientelism regimes compared to dominant patronage regimes, as well as the presence of reform actors and credible audit institutions.
7. CONCLUSION
Infrastructure megaprojects continue to fail, not because we are ignorant of the causes, but because those causes are embedded in incentive structures that make them rational to sustain. Cost overrun is not an anomaly; it is an equilibrium outcome of distorted epistemic institutions.
The novelty of this article lies not in documenting cost overruns, but in theorizing their persistence as a stable institutional equilibrium resistant to moral or technical correction. Drawing on coordination equilibrium frameworks in repeated games, with assumptions of bounded rationality, incomplete information, and high future discount rates, we show that epistemic distortions persist because they are strategically rational for each actor within existing incentive structures. This diagnosis opens pathways for more precisely targeted interventions: not merely technical corrections, but transformation of incentive structures through micro-institutional design.
PDG offers a framework to correct these distortions through three interrelated causal pathways. The detection effect increases error identification probability. The behavior modification effect transforms incentive structures by increasing expected costs of ignoring dissent strategies and decreasing risks of testing assumption strategies—an explicit transformation of the payoff matrix that has sustained equilibrium. This effect depends on explicit conditions: credibility of ex post review mechanisms, which may include independent audit institutions, investigative media, responsive parliaments, or factional competition. The selection effect acts as a consolidation mechanism following initial change, differing from ordinary bureaucratic turnover by linking recruitment and promotion criteria to documented dissent responsiveness.
Epistemic contestability—the institutionalized capacity to challenge epistemic claims with mandatory response obligations and traceable consequences—is the core of this framework. It is explicitly distinguished from adversarial legalism, devil's advocate procedures, institutional review boards, and internal audit challenge processes through its integration of mandatory documentation, response mechanisms with accountability traces, and temporal coupling with ex post review.
Reinterpretations of the Big Dig, Flamanville, Crossrail, and Olkiluoto cases demonstrate consistent patterns of epistemic distortion across various institutional contexts. These cases were not selected for variance testing, but for pattern illustration across institutional contexts. These illustrations are not intended as counterfactual proof that PDG will necessarily succeed, but as heuristic demonstrations of intervention points that could theoretically shift equilibrium. PDG is vulnerable to elite override and manipulation in unsupportive political settlement contexts. The political settlements typology clarifies that PDG is more likely to succeed in competitive clientelism compared to dominant patronage regimes. PDG does not assume voluntary adoption by dominant elites; it becomes viable when competing factions perceive contestability as a strategic tool to weaken political opponents or build reformist reputations.
This article explicitly acknowledges potential endogeneity problems and emphasizes that the proposed propositions are conditional, requiring further empirical testing with designs accounting for selection. This acknowledgment, rare in conceptual papers, actually strengthens argument credibility.
With awareness of over-institutionalization risks and its conditional nature, this article reframes megaproject failure as an equilibrium phenomenon and identifies institutional design levers requiring systematic empirical investigation.
Transition from old to new equilibrium is enabled by exogenous factors opening windows of opportunity—crisis, entry of reform champions, or strategic use by one faction to weaken another—followed by PDG adoption through layering, and reinforcement through positive feedback from its internal mechanisms. PDG is not an automatic solution, but it provides institutional infrastructure that can, under appropriate conditions, shift equilibrium from epistemic distortion toward healthier epistemic contestation.
REFERENCES
Acemoglu, D., & Robinson, J. A. (2012). Why nations fail: The origins of power, prosperity, and poverty. Crown.
Ansar, A., Flyvbjerg, B., Budzier, A., & Lunn, D. (2014). Should we build more large dams? The actual costs of hydropower megaproject development. Energy Policy, 69, 43-56.
Flyvbjerg, B. (2008). Curbing optimism bias and strategic misrepresentation in planning: Reference class forecasting in practice. European Planning Studies, 16(1), 3-21.
Flyvbjerg, B. (2017). Introduction: The iron law of megaproject management. In B. Flyvbjerg (Ed.), The Oxford handbook of megaproject management (pp. 1-18). Oxford University Press.
Flyvbjerg, B., Bruzelius, N., & Rothengatter, W. (2003). Megaprojects and risk: An anatomy of ambition. Cambridge University Press.
Grall, J. (2020). The Flamanville EPR: Anatomy of a megaproject failure. Revue Générale Nucléaire, 2020(3), 42-48.
Greiman, V. A. (2013). Megaproject management: Lessons on risk and project management from the Big Dig. Wiley.
Habermas, J. (1996). Between facts and norms: Contributions to a discourse theory of law and democracy. MIT Press.
Heikkila, T., & Gerlak, A. K. (2019). Working on learning: How the institutional rules of environmental governance matter. Journal of Environmental Policy & Planning, 21(1), 92-107.
Kagan, R. A. (2001). Adversarial legalism: The American way of law. Harvard University Press.
Kahneman, D. (2011). Thinking, fast and slow. Farrar, Straus and Giroux.
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263-291.
Keeney, R. L., & Raiffa, H. (1993). Decisions with multiple objectives: Preferences and value trade-offs. Cambridge University Press.
Khan, M. (2018). Political settlements and the analysis of institutions. African Affairs, 117(469), 636-655.
Lofthouse, J. K., & Schaefer, A. (2024). Expert knowledge and the administrative state: An epistemic ecosystem approach. Public Choice, 202, 433-454.
Mahoney, J., & Thelen, K. (2010). A theory of gradual institutional change. In J. Mahoney & K. Thelen (Eds.), Explaining institutional change: Ambiguity, agency, and power (pp. 1-37). Cambridge University Press.
Merrow, E. W. (2011). Industrial megaprojects: Concepts, strategies, and practices for success. Wiley.
Merton, R. K. (1968). Social theory and social structure. Free Press.
National Audit Office. (2019). Crossrail: A progress update. National Audit Office, London.
North, D. C. (1990). Institutions, institutional change and economic performance. Cambridge University Press.
Ostrom, E. (2010). Beyond markets and states: Polycentric governance of complex economic systems. American Economic Review, 100(3), 641-672.
Rein, M., & Schön, D. (1993). Reframing policy discourse. In F. Fischer & J. Forester (Eds.), The argumentative turn in policy analysis and planning (pp. 145-166). Duke University Press.
Sunstein, C. R., & Hastie, R. (2015). Wiser: Getting beyond groupthink to make groups smarter. Harvard Business Review Press.
