Who decides?
Getty ImagesGetty ImagesGetty Images
· 18 min read
Europe justifies decarbonisation by its aggregate benefits: lower energy dependence, avoided damage, new industry. Those benefits are real but conditional, and the conditions are political. This article argues that the unresolved problem in European transition policy is not a shortage of decision-makers – the Commission, Parliament and Council decide constantly – nor a complete absence of standards. It is that the standards Europe has are too broad to show how competing considerations will be weighed, what evidence warrants revising a collectively agreed commitment, and how those who relied on it will be treated. The IMF’s 2026 modelling suggests that meeting the EU’s 2035 objectives requires additional investment of roughly one percent of GDP a year, with modest macroeconomic effects, provided carbon pricing and subsidies are combined. Policy credibility is one important condition for sustaining that investment. Using batteries and vehicle regulation as a worked example, and China and artificial intelligence as comparative tests, the article distinguishes two ways in which revision can erode the investment on which the transition’s benefits depend – weaker ambition and uncertainty about further change – and sets out a common framework of justification for revisions that preserves democratic judgment over the trade-offs themselves.
In the latest instalment of his CorriereTv series Oriente Occidente, Federico Rampini argued that China is the great absentee from the global debate on the dangers of artificial intelligence, and that its posture repeats the one it adopted on climate: unwilling to accept limits imposed from outside, Beijing reveals itself as a sovereigntist rather than a multilateralist power.1
The history is not quite right. China ratified the Paris Agreement alongside the United States in 2016, and it is the United States that has since withdrawn.2 China has signed the principal international AI declarations, and in July 2026 it launched its own World Artificial Intelligence Cooperation Organization in Shanghai.3 But Rampini has noticed something real. On climate and on AI, Beijing has preferred commitments it sets for itself to limits set by others, and has kept its international undertakings closely aligned with a domestic industrial strategy decided first.
The more useful question this raises is about Europe. It is not who decides in Europe: the Commission proposes, the Parliament and Council legislate, and they do so constantly. It is this:
By what standards do Europe’s authorised institutions weigh whose losses justify changing a collectively agreed commitment – and are those standards clear enough to be relied upon in advance?
The question matters for a practical reason, and that reason runs through this article. The benefits of the transition depend on investment. Investment depends, in part, on commitments that investors believe will hold. And commitments are more credible when the conditions under which they may change are known in advance. An obligation that repeatedly moves as its deadline approaches risks teaching every regulated party – carmaker, battery investor, AI developer – that deadlines are opening positions.
Three questions are routinely run together in European debate. Authority: who is entitled to decide? Standard: by what substantive criterion is the balance struck when legitimate interests conflict? Accountability: what justification must be given, and who may challenge it?
On authority, Europe is not deficient. Its institutions have clear legislative competence, and the transition’s main instruments – the Climate Law, the emissions trading systems, the vehicle CO2 standards, the AI Act – were adopted through ordinary procedures. Formal accountability exists too: proposals are published, assessed, debated and voted on. The weakness lies in the second question, and it is a weakness of specificity rather than of absence. Targets are set collectively and at a distance; when their distributional consequences become politically salient, they are reopened, and the standards that govern reopening are usually too broad to indicate how the balance will be struck. Rules of that kind are harder to scrutinise consistently, and harder for anyone to rely on in advance.
Changing a policy is not in itself evidence that the change lacked a governing standard; democratic institutions must be able to reconsider. Nor is European legislation silent. The vehicle CO2 regulation directs the Commission, when reviewing the targets, to take account of technological developments and the importance of an economically viable and socially fair transition, and connects that assessment expressly to possible revision.4 The amended Climate Law provides for a biennial review assessing competitiveness, energy prices, net removals and Member State flexibility.5 These are substantive standards. But they are broad. They name the considerations without indicating how they are to be weighed against one another, what evidence would warrant revision as distinct from ordinary adjustment cost, or what is owed to those who invested on the strength of the original commitment. The start of ETS2 has moved from 2027 to 2028, and in December 2025 the Commission proposed to lower the 2035 car target. Each decision is defensible on its own terms. What they share is that the standards in place left the decisive weighing to be done under pressure, once the costs had become salient.
The rest of this article applies that framework. The economic evidence shows why choices are unavoidable; the battery and automotive case shows what is at stake; China and AI test the argument from outside; and the final sections set out what a credible process for revising commitments would require.
Europe now presents decarbonisation as a growth plan: the Clean Industrial Deal promises cheaper energy, restored competitiveness and quality jobs.6 Three different questions hide inside that promise. Does decarbonisation increase welfare globally? Does Europe capture a net economic gain? And however Europe fares in aggregate, how are gains and losses distributed among its regions, workers, consumers and owners of capital? Evidence for one does not establish the others.
The most useful recent estimate of the European aggregate comes from the IMF’s March 2026 working paper, which bridges two literatures that had long disagreed: bottom-up sectoral estimates of two to three percent of GDP in additional investment, and top-down models that found the aggregate investment effect close to nothing.7 The IMF team estimates that meeting the EU’s 2035 objectives requires additional investment of roughly one percent of GDP a year. The wider macroeconomic effects are modest provided carbon pricing and green subsidies are combined, since they push prices and output in opposite directions. Public investment costs below half a percent of GDP annually, against emissions trading revenues that could eventually approach one percent.
These quantities need to be kept apart. Additional investment is not the transition’s net cost: it creates assets. Fiscal outlays are not output effects, and output effects are not welfare. What the modelling shows is that Europe can undertake a very large physical restructuring without a large macroeconomic penalty, if the policy mix is right. What it does not establish is a large, automatic growth dividend. Whether a growth strategy can be built on the transition – through innovation, productive investment and reduced vulnerability – depends on where the investment is made and whether it is sustained. That is a political question, not a modelling one.
At the global level, the welfare case rests on avoided climate damage, which most of the literature assessed by the IPCC finds to outweigh mitigation costs.8 The scale of investment – the IEA expects some $3.4 trillion in 2026, of which roughly $2.2 trillion goes to clean technologies9 – shows commitment and activity, not net welfare by itself.
For Europe specifically, the strongest gain is reduced exposure to imported energy. EU energy imports cost about €337 billion in 2025, roughly half the 2022 peak.10 Ember estimates that wind and solar capacity added after the invasion of Ukraine avoided some €72 billion of fossil imports between 2022 and 2025,11 and the Commission projects savings building toward €130 billion a year by 2030.12 The 2026 Hormuz shock, which led the Commission to cut its forecast for EU growth that year to 1.1 percent and raise projected inflation to 3.1 percent, showed both how far that exposure had fallen and how far it had not.13 This is less a growth effect than insurance.
Two further benefits are often merged and should not be. Europe is warming faster than any other continent, and weather and climate extremes cost the EU an estimated €822 billion between 1980 and 2024.14 But avoided climate damage in Europe depends on global mitigation, to which Europe’s own emissions cuts contribute only a share; it justifies European participation in a global effort, not a domestic return on European investment. Local co-benefits – cleaner air, lower mortality from particulate pollution – accrue to Europeans directly, whatever others do. Productivity spillovers, finally, are plausible but contested: the ECB notes possible medium-term gains alongside firm-level evidence that stringent environmental policy can slow productivity growth in high-emitting firms.15
Distribution is where aggregate figures say least. The OECD’s consistent finding is that the aggregate employment effects of climate policy are modest while the sectoral and regional effects are large.16 New jobs in cell manufacturing need not arise in the places, for the workers or at the time that jobs in powertrain manufacturing are lost. That is why a positive aggregate can coexist with concentrated and durable losses – and why the question of who bears them cannot be answered by the aggregate.
Suppose electrification creates a trillion-euro global industry. That proves an opportunity exists; it does not prove that Europe captures it. Importing equipment is not in itself a loss. Cheap imported solar modules have done more for European emissions than most European industrial policy, and imports sustain domestic installation, services and downstream production. The risk is conditional. It arises where the capabilities lost carry value that prices do not capture – technological learning, supply resilience, regional employment, defence relevance – and where Europe absorbs the adjustment costs of a transition whose industrial rent accrues elsewhere.
Batteries are where that risk is most visible. Northvolt, the flagship of European battery ambition and holder of the largest green loan ever raised in Europe, filed for bankruptcy in Sweden in March 2025; its Swedish plants passed to the American firm Lyten in February 2026.17 Its failure had several causes, among them its own production difficulties and Chinese price competition; it illustrates how hard European capability is to retain, not the effect of any particular regulation. In 2025 China accounted for over eighty percent of global battery cell production, about eighty-five percent of cathode active material and more than ninety percent of anode active material, and the share of Chinese producers in the EU market has almost doubled since 2023.18 Bruegel and Rhodium find clean-tech manufacturing investment now falling worldwide, after years of Chinese state-led expansion at prices competitors could not match.19
Against that background, two distinct things have happened to the demand signals that would anchor manufacturing in Europe. The first is that they have been weakened. The Commission’s Automotive Package of December 2025 proposes replacing the 2035 zero-emission requirement for new cars with a ninety percent fleet reduction, the remainder compensated through EU-made low-carbon steel and renewable fuels, together with three-year averaging of the 2030 target; at the time of writing the proposal remains in co-decision.4 Transport & Environment estimates that, under the scenario reflecting the industry’s own amendments, some 1,024 GWh of announced battery capacity would not be built by 2030 – the equivalent of thirty-four Northvolt-sized plants and up to 47,000 jobs – and that the EU would spend around €50 billion more on oil imports between 2026 and 2035. Even the Commission’s proposal, on T&E’s modelling, forfeits roughly twenty-one such plants.20 These figures measure the first effect: a lower target means less expected demand for batteries, and would do so even if the change had been announced years in advance. They are advocacy figures, not an official impact assessment, but the direction of the effect is not seriously in doubt.
The second effect is separate, and it is the one at the centre of this argument. When a commitment is revised under pressure, investors learn something beyond the new target: that targets may move again when costs next become salient. That uncertainty can discourage investment independently of the level of ambition, because battery and cathode plants are long-lived assets financed against demand that must be believed for much of their life. It also creates a feedback loop. A concession intended to moderate adjustment costs for existing industry can weaken the investment case for the new industry whose returns would have helped to pay for the adjustment. The revision need not be wrong to have this effect. It need only leave investors unable to judge when, and on what grounds, the next one will come.
Wolfgang Münchau’s account of Germany sounds the same warning. His argument is not that exporting and importing were mistakes, but that a contingent configuration – cheap Russian energy, Chinese demand, deferred domestic investment – hardened into an article of faith that could not be revised when its conditions changed.21 A settlement in which Europe writes the rules and others build the equipment may be rational at current prices. The question is whether Europe would notice, and be able to act, if it ceased to be.
The just transition began in labour politics, and the ILO’s framework of social dialogue, decent work and social protection is broader than compensation.22 In practice, though, European policy has operationalised justice mainly through compensation and regional support after the direction of travel has been set. The Just Transition Mechanism and the Social Climate Fund are necessary instruments. They are not a standard for deciding when a commitment should yield.
A just transition is not necessary because decarbonisation is inherently zero-sum. It is necessary precisely because a transition that is positive-sum in aggregate can still be profoundly distributive in practice.
Imagine a decade in which the transition delivers substantial net new value, lower climate damage and a smaller import bill, while eliminating tens of thousands of jobs concentrated in three regions, raising household energy bills for five years and transferring much of the manufacturing rent outside Europe. That is a positive-sum outcome, and every governance problem remains: whether the transition should proceed despite concentrated losses, under what standard, and what claim those bearing the losses have on gains created elsewhere.
Note what that formulation avoids. It does not assume that aggregate gain is itself the decision rule. The tempting move – winners could in principle compensate losers, therefore the change is justified – imports a particular welfare test as though it were a constitutional one. Whether Europe should adopt that test is exactly the sort of question a political community is entitled to answer for itself. It is also a question better answered in advance than under pressure.
China does not supply the answer to Europe’s problem, but it isolates one variable. Its industrial strategy and its international commitments have been more deliberately coordinated. Its formal climate targets have been cautious – emissions to peak before 2030, carbon neutrality before 2060 and, in the contribution announced in September 2025, net emissions seven to ten percent below peak by 2035 – while its deployment of solar, wind, batteries and electric vehicles has repeatedly run ahead of them.23 Commitments abroad have followed capabilities decided at home.
The comparison should not be overdrawn. China’s stated objectives – development and security, as the state defines them – identify priorities and the holder of power; they do not supply a predictable standard for resolving conflicts, and Chinese policy has shifted abruptly when priorities changed. Nor is its record free of the costs Europe fears: overcapacity, loss-making producers and trade friction are among its by-products. Its governance scores poorly on accountability, and those who bear its losses have little means of contesting the decisions that impose them. What the comparison shows is narrower. Investors respond to coordination they find credible, and in the sectors that matter here China has offered more of it.
Artificial intelligence offers a second test, inside Europe. The AI Act of 2024 was the world’s first comprehensive horizontal regulation of the technology, adopted while the frontier models it would govern were being built, with few exceptions, in the United States and China. Its obligations for high-risk systems listed in Annex III were due to apply from 2 August 2026, and those for AI used as a safety component of regulated products from 2 August 2027. The harmonised standards on which compliance depended were not ready – a respectable reason for delay. The Act already contained a fallback for missing standards; it contained no stated test for when the timetable itself could move. The remedy was a fresh legislative bargain deferring the two sets of obligations to December 2027 and August 2028. The change was foreseeable well before it became law: the Commission proposed it in November 2025, the co-legislators reached political agreement in May 2026 and the Council adopted it in June. But it was published in the Official Journal only on 24 July, nine days before the original date, so that for months regulated parties planned against obligations that were legally in force and politically expected to move.24
Anu Bradford describes Europe’s model of digital governance as rights-driven, set against an American model driven by markets and a Chinese one driven by the state.25 Europe’s regulatory power is real, but it is a power over what may be sold rather than over what is built, and it depends on others continuing to want to sell. The pattern is the battery case in another technology: rules adopted first, capability located elsewhere, and timetables revised under pressure against standards too broad to anticipate the outcome.
Predictability cannot mean that democratic institutions never reconsider. The distinction that matters is between justified adaptation and bargaining without a standard. A revision of a collectively agreed commitment – a target, a phase-out date, an application deadline – should at a minimum be required to explain five things.
These requirements do not amount to a substantive decision rule, and they should not pretend to. They tell institutions what to explain, not which outcome to choose. Elected institutions must still decide, and defend, the weight they give to affordability, emissions, employment and strategic capability. What the requirements provide is a common framework of justification, applied consistently, which makes those weightings explicit and comparable from one decision to the next.
None of this freezes policy. It changes the terms on which policy moves. A review clause that names these tests in advance tells investors what would have to be true for a commitment to change, and tells those bearing losses what they would have to show. A review clause without them is not flexibility. It is an invitation. That applies to the next review of the AI Act as much as to the biennial review of the 2040 target.
Two further commitments follow from the evidence. The case for the transition should be made in terms that can bear weight – energy security, industrial capability, avoided damage – rather than a GDP dividend that may not arrive. And the promise should be tied to benefits broadly shared within Europe: a gain that reaches households through cheaper electricity, and regions through construction, installation and manufacturing, sustains itself politically in a way that a gain concentrated elsewhere does not.
For thirty years Europe separated the benefits consumed inside its borders from the productive capabilities on which those benefits depended. Production migrated, energy was imported and security was largely provided by others. That arrangement was rational under the assumptions of the period, and those assumptions no longer hold. Decarbonisation becomes a programme of economic reconstruction to the extent that it reverses part of that separation. To the extent that it does not, Europe may decarbonise successfully while reproducing the dependencies of the globalisation era in new technological forms.
So who does decide? In Europe, its authorised institutions do, through procedures that are lawful and, in the formal sense, accountable, and under standards that the legislation itself states. What Europe needs are clearer ones: for weighing the trade-offs, for explaining revisions and for treating those who relied on its commitments. Where existing standards are too broad to show how the balance will be struck, revision tends to be prompted by salience, and practical influence over when commitments yield passes to whoever can make their losses most visible as a deadline approaches. Accountability then attaches to each revision as a political settlement rather than to a consistent standard against which settlements can be compared.
Rampini is right that China’s conduct reveals a power that prefers to be the author of its own limits. Europe’s answer cannot be to become sovereign in China’s way. It can be to become credible in its own: to decide, through institutions that can be held to account, what the transition is for, which capabilities it means to keep, how those bearing concentrated losses will be treated and on what grounds its commitments may change – and to say so before the conflict, not during it.
That is what “Who decides?” comes to in practice. It is not only a question of who holds authority. It is a question of how that authority is exercised, justified and made credible – of whether Europe can make the commitments on which its economic future depends, and be believed.
illuminem Voices is a democratic space presenting the opinions of leading Sustainability Thought Leaders, their views do not necessarily represent those of illuminem.
The world needs sustainability knowledge. At illuminem, no interest group or shareholder can influence our work. Thank you for supporting our mission to make high-quality and independent sustainability information free for all. Every contribution helps. Thank you for donating today.
Philip Corsano
Public Governance · Carbon
illuminem briefings
Carbon · Public Governance
illuminem briefings
Public Governance · Social Responsibility
Financial Times
Oil & Gas · Public Governance
The Guardian
Public Governance · Nature
Eco Business
Mitigation · Public Governance