📈 economics

A Coordination Theory of NHS Whistleblowing Failure

This paper argues that persistent whistleblowing failures in the NHS stem from structural coordination problems among whistleblowers, management, and colleagues—rather than just legal or incentive gaps—which are revealed through a global games model to require three new mechanisms beyond current policy.

Original authors: Ari Ercole

Published 2026-07-22

📖 9 min read🧠 Deep dive

CC BY 4.0

Original authors: Ari Ercole

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a world where the rules of the game change not because someone broke the law, but because the players are stuck in a trap of their own making. This paper lives in the corner of science called Game Theory, which is basically the study of how people make decisions when their choices depend on what everyone else decides to do. Think of it like a massive, invisible chessboard where every move you make is a reaction to a move someone else might make. The paper also leans on Global Games, a fancy way of saying "what happens when everyone has a little bit of private information and has to guess what the others know." Finally, it looks at Coordination Failure, which is the frustrating moment when everyone wants to do the right thing, but nobody dares to go first because they're terrified of being the only one left standing. Why does this matter? Because in places like hospitals, where lives are on the line, getting everyone to coordinate on telling the truth is the difference between a safe system and a disaster. If the players are stuck in a bad equilibrium, no amount of new laws might be enough to get them unstuck.

The Great Hospital Silence: A Game of Whistleblowers, Managers, and the "Bystander Trap"

For thirty years, the National Health Service (NHS) in the UK has been stuck in a loop. Every decade, a massive public inquiry digs into a hospital scandal, finds that nurses and junior doctors tried to warn people about terrible mistakes, and discovers that those warnings were ignored, punished, or buried. The usual story goes like this: "The whistleblower needs more legal protection," or "The managers need to be scared into investigating." But this paper argues that this story is missing the most important player in the room: the colleagues.

The author, Ari Ercole, sets up a mathematical model—a sort of "what-if" simulation—to see what happens when a whistleblower, a hospital manager, and a group of colleagues play a game. The goal? To see why good people stay silent even when they know something is wrong.

The Three Players and the Trap

Imagine a hospital as a stage with three groups of actors:

The Whistleblower (W): The person who sees a problem (like a dangerous error) and has to decide whether to shout about it or stay quiet.
The Manager (T): The boss who decides whether to investigate the shout or sweep it under the rug.
The Colleagues (C): The big group of other staff members who hear the shout. They have to decide: Do I back the whistleblower? Do I ignore them? Or do I actively try to make the whistleblower look bad?

The paper finds that the standard fix—giving the whistleblower a shield against getting fired—doesn't actually solve the problem. Why? Because the real bottleneck isn't the whistleblower's fear; it's the Colleagues' Trap.

The "Self-Incrimination" Trap (Why Colleagues Stay Silent)

Here is the first big twist. In the model, for a colleague to support a whistleblower, they have to give evidence. But giving evidence is risky. If the whistleblower turns out to be wrong (or if the manager decides to ignore the report), the colleague who spoke up gets dragged into the mess too. Their own past actions, their own mistakes, and their own proximity to the problem get put under a microscope.

The paper calls this Collective Implication. It's like being in a group project where if you speak up to say "Hey, the teacher is grading us wrong," you also have to admit, "And I might have messed up my part of the project too."

Because of this, even if a colleague privately thinks the whistleblower is right, they rationally decide to stay silent. They align with the manager to avoid getting burned. The paper shows that employment protection for the whistleblower doesn't fix this. You can protect the person who shouts, but you can't protect the person who has to stand up and say, "I agree with them," without risking their own career.

The "Active Elimination" Trap (When Colleagues Fight Back)

The model gets even darker. It shows that colleagues don't just stay silent; sometimes they actively work against the whistleblower. They build a "counter-narrative"—a documented campaign of complaints and negative reports designed to make the whistleblower look crazy or incompetent.

This is called Active Elimination. It's different from just ignoring the problem. It's like a group of people in a meeting deciding to write a fake report saying the person who raised the alarm is actually the one who caused the fire. The paper finds that this is a "discrete" failure. If the environment is scary enough (meaning, if there's a high chance someone will find out the truth later), this active fighting can be stopped completely. But if the environment is quiet, colleagues will happily build these fake cases to protect themselves.

The "Turnover Ratchet" (Why the Record Gets Worse Over Time)

Here is the third mechanism, and it's a bit like a game of "Telephone" played with a broken phone. The paper introduces the Turnover Ratchet.

In the NHS, managers often don't stay in their jobs for long. When a new manager arrives, they look at the file left by the old manager. The file says "Unsubstantiated" (meaning the complaint was dismissed). But the file doesn't say why it was dismissed. Did the new manager investigate and find the whistleblower was lying? Or did the old manager just ignore it?

Because the new manager doesn't know the history, they assume the worst: "If the last guy said it was unsubstantiated, maybe this system is full of false alarms." So, they become even more skeptical of the next report. Every time a complaint is dismissed without a real investigation, the "reputation" of the reporting system gets a little worse. Over time, the system becomes so skeptical that no one believes any report, even the real ones. The paper argues that this degradation happens faster when managers are fired and replaced quickly, because they leave behind a messy, confusing record that the next person has to guess at.

The "Coalition" Problem (When Bad Guys Team Up)

The paper also looks at what happens when colleagues don't just act alone, but form a Coalition. This is like a secret club of managers and staff who agree to work together to silence a whistleblower.

The model shows that a small, tight-knit group can be much more dangerous than a bunch of people acting alone. They can trade their individual uncertainty for "certainty of success." They know they will win because they are working together. And because they are working together, they are harder to catch. The paper suggests that the only way to break this is to make it so that if one member of the secret club snitches, they get a "get out of jail free" card (an amnesty). This turns the secret club against itself, making it too risky to stay silent.

What the Paper Says We Should Do

The paper argues that we've been trying to fix this by just protecting the whistleblower, but that's like trying to stop a flood by putting a bucket under the leak while ignoring the hole in the dam. The authors suggest four new tools:

Colleague Amnesty: Just like in the law where a criminal cartel member gets a deal if they snitch on the others, the paper suggests giving colleagues immunity if they support a whistleblower. This removes the fear of "self-incrimination" and lets them speak up without worrying about their own past mistakes.
Named-Trust Publication: Instead of hiding the results of investigations, hospitals should publish exactly what happened, including the name of the trust. This makes it impossible to hide the fact that a complaint was ignored. If a hospital ignores a warning and then a patient dies, the public record will show it. This raises the risk for the "bad actors" and makes them think twice.
Investigate the Silence, Not Just the Crime: If there is a suspicion that a group of people is conspiring to silence someone, the paper says we should investigate that conspiracy immediately, even before we know if the original complaint was true. Waiting to investigate the conspiracy until after the main case is solved just lets the damage compound.
Graduated Investigation Design: This is a crucial change to how investigations are run. Currently, when a report comes in, it often immediately implicates the whole team, making every colleague feel like they are on trial for their own past mistakes. The paper suggests a "graduated" approach: first, determine if the problem is a one-off error or a systemic issue. Only after that is clear should colleagues be asked to testify, and they should be assured that their testimony won't be used to prosecute their own past conduct. This lowers the fear of "self-incrimination" right at the moment they have to decide whether to speak up, making it much more likely they will choose to support the truth.

The Bottom Line

The paper concludes that the failure of whistleblowing in the NHS isn't just about bad laws or scared individuals. It's a structural trap. The way the game is set up—where speaking up risks your own career, where managers come and go leaving a messy record, and where colleagues can team up to hide the truth—makes it rational for everyone to stay silent.

The authors are careful to say this is a model based on logic and simulation, not a final proof of every single case. But they argue that the pattern fits the real-world data from decades of inquiries. The solution isn't just to protect the person who shouts; it's to change the rules so that the people standing next to them feel safe enough to shout, too. Without fixing the "coordination failure" among the colleagues, the paper suggests, we will keep seeing the same scandals, over and over again.

Technical Summary: A Coordination Theory of NHS Whistleblowing Failure

1. Problem Statement

The paper addresses the persistent failure of whistleblowing mechanisms within hierarchical healthcare organizations, specifically the UK's National Health Service (NHS). Despite thirty years of legislative reform (e.g., the Public Interest Disclosure Act 1998) and repeated public inquiries (Francis, Ockenden, Messenger) documenting a pattern of declined warnings and punished reporters, the standard diagnosis—that whistleblowers lack sufficient legal protection or that management lacks incentives to investigate—is argued to be structurally incomplete.

The central problem identified is the omission of the colleague group as a strategic player. The paper posits that the decision to validate a report often hinges not on the whistleblower or management alone, but on whether colleagues choose to corroborate the report. The failure to account for the strategic incentives of this "colleague layer" explains why standard interventions have failed to produce stable improvement.

2. Methodology

The paper employs a game-theoretic model involving three types of players in a repeated, infinite-horizon game:

Whistleblower ( $W$ ): Observes a violation of severity $\theta$ and decides whether to report.
Trust Management ( $T$ ): Receives reports and decides whether to validate or decline them.
Colleague Group ( $C$ ): A large set of $n$ agents who observe the report and choose between Support, Defect (silence), or Eliminate (actively constructing a counter-narrative).

Key Analytical Techniques:

Global Games: The model utilizes global games techniques (Carlsson & van Damme, 1993; Morris & Shin, 1998) to solve for a unique equilibrium in the colleague coordination subgame. Colleagues receive private signals about the severity of the violation and coordinate via threshold strategies.
Intertemporal Dynamics: The model incorporates a "turnover ratchet" where short management tenure ( $\delta_T$ ) leads to the inheritance of coarse institutional records (dispositions without process), causing the credibility of future reports to degrade geometrically over time.
Strategic Elimination: The model extends the colleague strategy space to include "Eliminate," where colleagues generate a documented persuasion signal ( $\sigma$ ) to manipulate management's posterior belief, distinguishing this from passive silence.

3. Key Contributions and Mechanisms

The paper identifies three structural mechanisms that existing policy neither recognizes nor addresses:

A. Collective Implication (The Support Margin)

Corroborating a report requires giving evidence, which exposes the witness's own conduct to scrutiny. In systemic violations, this "self-incrimination cost" ( $\phi H_C$ ) aligns colleagues with management against validation.

Result: Even if colleagues privately believe a report is genuine, the risk of their own conduct being scrutinized deters support.
Policy Failure: Standard employment protection for the whistleblower ( $P_W$ ) does not alter this cost for colleagues. Similarly, statutory liability for management ( $L_T$ ) is ineffective when the probability of exposure ( $\lambda$ ) is low. The paper proves that employment protection can increase the volume of reports but leaves validation rates unchanged, shifting the equilibrium from "Silence" to "Decline."

B. Active Elimination (The Independent Action Margin)

Colleagues may actively construct a documented counter-narrative to manipulate management's belief that a report is vexatious.

Result: Unlike passive silence, active elimination leaves an attributable trail. The paper demonstrates that if the exposure environment is sufficiently credible (high $\lambda$ ), active elimination becomes strictly dominated for independent actors, regardless of their confidence in success. This offers a "discrete deterrence" result impossible for passive defection.
Distinction: This mechanism is distinct from passive silence; it admits a "kill condition" where participation becomes irrational for all belief levels.

C. Collusive Elimination (The Organized Coalition Margin)

Colleagues may form a coalition to trade uncertainty for the certainty of success.

Result: A coalition removes the aggregation uncertainty of independent action but bears a distinct risk: if the conspiracy is confirmed, members face aggravated exposure ( $\Psi_i$ ).
Policy Failure: Instruments that deter independent elimination (raising $\lambda$ ) do not deter collusion, as coalitions trade the uncertainty of independent failure for the certainty of success. The paper argues that collusion is best deterred not by general exposure, but by prompt investigation and confirmation of suspected coalitions, which destabilizes the coalition from within by raising the probability of internal disclosure.

D. The Turnover Ratchet

Short management tenure creates a "ratchet" effect. Successors inherit files recording outcomes (e.g., "unsubstantiated") without the process behind them.

Result: Rational successors interpret a history of "unsubstantiated" dispositions as evidence that the reporting culture is poor, causing the prior belief in report credibility to degrade geometrically. This degradation is accelerated by active elimination, as fabricated documents enter the record as "evidence" against the reporter.
Implication: Without external shocks (like public inquiries) or structural repairs, the institutional record permanently degrades, making future validation impossible.

4. Results and Equilibrium Characterization

The paper characterizes four equilibrium outcomes: Support, Decline, Elimination, and Silence.

Under a parameter configuration consistent with recurrent NHS failures (short tenure, low enforcement visibility, high systemicity, high career leverage), the model predicts that for all but the most severe violations, the outcome will be Decline, Elimination, or Silence.
Complementarity: The paper proves that interventions are supermodular. Single-instrument reforms (e.g., only increasing whistleblower protection) fail because multiple constraints bind simultaneously. Effective reform requires a bundle of interventions targeting the specific binding constraints of colleagues and management.

5. Policy Implications

The paper proposes three specific interventions, each targeting a distinct structural failure:

Colleague Amnesty: Adapted from cartel leniency programs, this grants immunity from investigative consequences to colleagues who voluntarily support a report. This removes the self-incrimination cost ( $\phi H_C$ ), enabling colleagues to self-organize and corroborate without fear, effectively breaking the alignment between colleagues and management.
Mandatory Named-Trust Outcome Publication: Requiring the public publication of whistleblowing outcomes at the level of specific trusts, including the subsequent clinical/regulatory status of the concern. This repairs the institutional record by separating "decline" from "investigation," halts the turnover ratchet, and raises the effective exposure probability ( $\lambda$ ) for independent eliminators, pushing the system toward the "kill condition" for active elimination.
Prompt Investigation of Suspected Coalitions: Suspected collusion should trigger an immediate, parallel investigation prioritized over the underlying report. A confirmed conspiracy acts as positive evidence for the reporting culture (repairing the record faster than waiting for the underlying case to resolve) and prevents the compounding costs of delay (continued harm, record degradation, and intact incentives for reoffending).

6. Significance and Claims

The paper claims that whistleblowing failure in the NHS is not an anomaly or a compliance gap, but the equilibrium outcome of a game defined by the institutional structure of publicly funded hierarchical healthcare.

Structural Diagnosis: The failure is driven by the strategic incentives of the colleague group, specifically the costs of self-incrimination and the ability to coordinate active elimination.
Limitations of Current Policy: Thirty years of reform have failed because they target the whistleblower ( $P_W$ ) and management liability ( $L_T$ ), neither of which reaches the binding constraints on the colleague group ( $\phi, b_C, \beta_i$ ).
The Inquiry-Reform Cycle: The paper explains the recurring nature of NHS scandals: public inquiries create a local "common knowledge" event that temporarily destroys the elimination equilibrium at the specific trust investigated. However, because these inquiries are local and do not alter system-wide parameters (like tenure or anonymity), the structural failure recurs at other trusts.
Conclusion: The proposed interventions (amnesty, named publication, prompt coalition investigation) are designed as standing, system-wide instruments to break this cycle by addressing the root coordination failures rather than treating symptoms after the fact.