Thames Water and the Cost of
Being the One Who Decides
Thames Water had a rescue mechanism available. The difficulty was that activating it required a specific actor to publicly own the decision.
The Mechanism That Exists
Thames Water has spent more than a year suspended between collapse and recovery. Liquidity has been extended repeatedly since 2024. Each extension was presented as sufficient. None resolved the underlying uncertainty.
A statutory rescue framework has existed in UK law since 1991, and the mechanism remains unused.
Every crisis eventually reaches a point where contingency becomes reality: preparation ends and someone must act. Thames Water has already reached that point.
A statutory route exists through which the government can place a failing water company into Special Administration.¹ Once triggered, an administrator can preserve essential services while the company is restructured. The framework is not theoretical. A version of it was used in the 2021 collapse of Bulb Energy.
The existence of the mechanism is therefore not the puzzle. Its activation is.
The Single Point of Initiation
The mechanism does not lack an entry point. It has a single point of initiation.
A Special Administration cannot be triggered by the parties most directly exposed to the financial outcome. Creditors, shareholders, and company management cannot unilaterally open the process. The authority to initiate the process sits with public institutions.¹
This structure reflects a deliberate trade-off. Restricting initiation prevents private creditors from using administration as a tool to maximise their own recovery at the expense of public infrastructure. But it also creates a different constraint: the actor capable of moving the system forward is not an external referee, but a political actor subject to its own set of pressures.
The actor capable of moving the system forward is not an external referee, but a political actor subject to its own set of pressures.
The Price of Initiation
There is another route through the same crisis, but it follows a different attribution structure.
A company can propose its own rescue plan, creditors can vote on it, and a court can approve it over the objections of a minority.² A version of this route was used in February 2025, when Thames Water’s interim restructuring plan was approved by the High Court and upheld on appeal the following month.
Notice what the government did not have to do during that process: initiate the restructuring. The company authored the plan. The court reviewed it. The public authorship of the decision rested elsewhere.
A Special Administration follows a different path. The Secretary of State would have to initiate a process involving a company that serves around 16 million people — one that publicly places the government at the origin of the decision. Whatever the eventual outcome, the political framing begins with the same association: the government placed Thames Water into special administration.
The two routes distribute authorship differently.
Why Delay Is Not a Mistake
It is tempting to read the extended period of uncertainty as dithering: a government that cannot make up its mind. That interpretation assumes delay is a failure of decision-making. Under the existing incentive structure, it may be the rational outcome.
Consider what each option costs a minister, not in the abstract, but in the form political actors experience most directly: attention, blame, and the shape of next week’s headlines.
Acting - triggering a Special Administration - converts a slow, technical financial problem into a directly attributable political event. Waiting does not remove costs. It changes their timing and distribution. Responsibility disperses into a system with no single address.
Thames Water has disclosed that a Special Administration could leave taxpayers exposed to as much as £2 billion within a year. That figure is one of the costs any government decision would need to account for. But it exists alongside another asymmetry: one cost is immediate, visible, and attached to a specific decision; the other emerges gradually, is distributed across multiple actors, and is harder to trace to a single moment.
Waiting is what the incentive structure produces, not necessarily a failure to decide.
A Live Test Case
In a select committee hearing, a minister for water read out a list of thresholds the department uses internally to think about when a Special Administration might apply.
In a separate legal proceeding, around the same time, the government’s stated position was that there was nothing to publish.
Both statements came from the same department, in public, within weeks of each other.
The proceeding is a judicial review filed in July 2025 by an environmental group, River Action, against the Environment Secretary. Their claim is narrow: not a request for the court to force a Special Administration, but a request that the government state, in public, what conditions would trigger one.
A threshold that exists only in internal conversation costs nothing to keep private. A threshold that exists on paper invites a question every time it is met and nothing follows: why hasn’t the government acted?
Making inaction checkable may be enough on its own - turning an unexplained absence of action into a decision that must be explained.
The significance of the case does not depend on a Special Administration occurring. A published threshold only functions this way if it is specific enough to test against. Language like “when appropriate” would satisfy an order to publish something without giving anyone a fact to hold the government to. Whether the eventual wording clears that bar is still open.
The Reframe
Many proposed fixes for situations like this aim at the same target: making it politically safer to act. Reduce the personal exposure attached to the decision. Route it through a court, an independent panel, or a pre-agreed formula: anything that puts distance between the person and the moment.
That approach is worth exploring, but its limits matter. Legal liability can often be reduced through institutional design. Political attribution operates under a different logic. A statute can protect a minister from legal consequences for a good-faith decision. It cannot prevent tomorrow’s headline.
This suggests a different target. Instead of only trying to reduce the cost of action, a system may need to increase the cost of unexplained inaction. Not by forcing a decision, but by making the absence of one visible.
A robust crisis mechanism does not need to eliminate the political cost of emergency action. It can instead ensure that silence carries its own cost: publish the criteria, publish the relevant state of the system against those criteria, and make delay something that must be explained rather than merely endured.
That is a smaller intervention than it first appears. It does not require removing discretion.
It requires making discretion observable.
The evidence from River Action points toward this direction. The objective is not necessarily to force a Special Administration, but to change the conditions under which delay occurs.
layer
Beyond Thames
Thames is not interesting because the mechanism failed. It is interesting because the mechanism existed.
The case reveals a recurring tension in crisis governance: the authority to act and the incentive to use it do not always align. That tension becomes sharper when authority and attribution are concentrated in the same place. The actor capable of moving the system forward is also the actor who becomes most visible if the outcome is politically costly.
The failure was not simply a lack of emergency mechanisms. It emerged from the coupling of emergency authority and attribution exposure. Systems designed to concentrate responsibility during crises may unintentionally create incentives to avoid exercising that responsibility.
Whether this pattern extends beyond Thames Water remains an open question. What Thames provides is a clear example of a broader tension: systems need someone to own crisis decisions, but concentrating ownership can also make those decisions harder to make.
The hardest part of a crisis mechanism is often not designing the exit. It is designing who can open the door without becoming the crisis itself.
¹ Section 24 of the Water Industry Act 1991. The petition can be brought by the Secretary of State, or by Ofwat with the Secretary of State’s consent — no other party has standing to bring it.
² Part 26A of the Companies Act 2006 — the restructuring plan procedure used across UK corporate insolvency, not specific to utilities.