A state imposes a new emissions standard, applicable to all, prompted by a serious health study and adopted after public debate. Bringing the plant into compliance costs more than its residual value, the period allowed is eighteen months, and operations stop. No bad faith is detectable, and that is precisely what makes the file difficult: the investor instinctively looks for hostile intent, will not find any, and will lose months on that ground.
The principle standing against it has a name in treaty jurisprudence and must be known in order to stop fighting it: good faith, non discriminatory regulation taken in the exercise of police powers for a public health or environmental objective gives no right to compensation even where it destroys value. It is the same wall as fiscal sovereignty, transposed to industrial ground, and it rests on the same reason: if every new standard opened a claim, no state could regulate without buying the consent of every existing installation.
Two routes cross that wall and neither runs through intent. The first is proportionality, which does not contest the objective but the burden: a standard whose compliance cost exceeds the value of the asset, imposed within a period making amortisation impossible, produces in substance a closure, and a closure effect is assessed even where the aim is legitimate. The second is legitimate expectations, and it is stronger where it exists: what the state specifically promised this investor when it invested. A general regulatory framework creates no expectation; a permit granted for a term, a stabilisation clause, a letter of undertaking create one.
That distinction must be pressed because it separates winnable files from the rest. An investor arguing that it could not foresee environmental tightening argues against the obvious: the global regulatory trajectory has been public for decades, and an underwriter as much as an arbitrator knows it. An investor producing an operating consent granted for twenty years, or ministerial correspondence promising stability of the regime applicable to its installation, is no longer arguing foreseeability: it is arguing an undertaking, which is an entirely different matter.
The transition period deserves treating as a variable in its own right, because it transforms the nature of the measure. The same standard with ten years allows amortisation, planning a compliance investment, or organising an orderly exit; with eighteen months it does not. An arbitrator will be alive to this because the period reveals what the state actually weighed, and an administration that granted five years to a domestic sector and eighteen months to another has produced, without saying so, the comparison that decides the file.
Then comes what the policy does with all this, and the answer depends on an exclusion to be read precisely. Many wordings exclude general regulatory measures taken in good faith, in terms close to the tax exclusion, which returns to exactly the same argument. Others respond to deprivation of control without concerning themselves with the characterisation of the measure, and are then markedly more favourable: what counts there is not the nature of the act but its effect on the operation. Those two constructions are told apart in one reading, and a buyer operating heavy industrial installations has an obvious interest in knowing which it signed.
One last point concerns conduct during the making of the standard, and it is almost always neglected because it happens before there is a problem. A standard adopted after a public consultation in which the investor filed a reasoned, costed observation proposing an alternative trajectory is not the same standard as one it never contested. The consultation file becomes evidence: it establishes that the cost was known to the administration, that a proportionate solution had been proposed, and that it was set aside. Without it, the state will argue that nobody had flagged any difficulty.
This module closes on a shift of ground a professional must be able to make in front of an angry client. The question is not whether the state acted in good faith, because it probably did and proving otherwise is lost in advance. It is what the state had promised this particular investor, and whether the burden imposed is proportionate to the objective pursued. Two questions of fact, documentable, handled with evidence rather than with indignation.
A European group has operated a secondary lead smelter in an Eastern European country since 2011, under an operating consent granted for twenty years, that is until 2031. In March 2026 a decree lowers the lead emission limit to a level requiring complete replacement of the filtration system. The cost is put at 41 million euros, the installation's residual value at 26 million, and the period allowed is eighteen months. The decree covers all installations in the category; there are four in the country, three of which belong to a state group that obtained, by a separate order made the same day, a five year period under a state funded modernisation plan. The group filed no observation during the 2025 public consultation. What should be argued?
One must first tell the client what not to argue, because that is where it will want to go. The health objective is real, the study exists, public debate took place: alleging the state's bad faith is lost in advance and would consume the file. Three serious grounds remain and must be taken in this order. The first is the operating consent granted for twenty years, running to 2031: that is not a general regulatory framework but an individual act addressed to this operator, carrying a term, and it is the soundest foundation for a legitimate expectation. It must be produced first and the case built around it. The second is disproportion, and the figures carry it: 41 million of compliance on an installation worth 26 is a cost exceeding the asset, which produces in substance a closure, and eighteen months forbid any amortisation of the outlay. The demonstration does not contest the standard, it contests the burden and the period, which is the only way to contest a legitimate measure. The third is the strongest of all and was nearly invisible: three of the four installations concerned belong to a state group and obtained five years by an order made the same day, with state funding. The standard is general, its application is not, and that difference of treatment between comparable situations is exactly what is called discrimination. It is the evidence to obtain first, with the public order and its funding plan, before they become hard to consult. One weakness must be acknowledged rather than concealed: the absence of any observation at the 2025 consultation will be raised, and the state will argue that nobody flagged the cost. The answer is to establish that the draft put out for consultation did not yet contain the limit finally adopted, if that is so, or failing that to concede the point and press the other two grounds, which suffice. Finally, the policy's regulatory exclusion must be read before notifying: if it covers general good faith measures, the whole file runs through the discrimination showing; if the cover responds to deprivation of control without characterising the measure, the closure suffices and the rest becomes incidental.
- 01Good faith, non discriminatory regulation in the exercise of police powers gives no right to compensation even where it destroys value: that is the wall, and it is the same one as fiscal sovereignty.
- 02Two routes cross it and neither runs through intent: proportionality of the burden, and legitimate expectations arising from an undertaking addressed to this investor.
- 03A general regulatory framework creates no expectation; a permit granted for a term or a letter of undertaking create one, and that is the difference between arguing foreseeability and arguing an undertaking.
- 04The transition period transforms the nature of the measure and reveals what the state weighed: different periods granted to comparable situations produce the comparison that decides.
- 05A standard never contested in consultation is not the same as one against which a reasoned, costed observation was filed: the consultation file is evidence.