What Comes Out of the Tailpipe

I. The occasion

At the end of August 2026 it became known that the Technical University of Munich accuses the European Union’s climate strategy for transport of a substantial misdirection. On 1 September 2026 the work itself appeared: a 93-page white paper entitled Defossilization, Not Decarbonization: Toward a Life Cycle-Based Standard for Automotive Greenhouse Gas Regulation.

The charge in essence: the EU regulation classifies electric vehicles as zero-emission. An evaluation of 19 studies with 47 scenarios arrives instead at an average reduction of 41 per cent compared with the combustion engine. Less CO&sub2;, then, but not none — while the regulation reckons with none.

The reason lies in the point of measurement. Measurement takes place at the tailpipe. What arises in raw material extraction, in vehicle and battery manufacture, in operation and in recycling stays outside — in particular the CO&sub2; embodied in steel and aluminium, and the electricity mix with which the vehicle is built and charged.

Whether the figures hold up in detail is still open; more on that at the end. The matter itself is instructive even if one leaves the size of the reduction undecided.

II. Why measurement takes place at the tailpipe

Not out of negligence. The tailpipe is the only place at which a vehicle can be examined without knowing half the world economy.

One puts the car on a test bench, has it drive a prescribed cycle and measures what comes out at the back. The result is unambiguous, repeatable, verifiable by any testing institute and directly attributable to the manufacturer. It has every property a regulation demands of a metric.

At this point, however, a qualification is due, and a serious one.

For what is examined is not the vehicle but its behaviour in the test cycle. And a cycle is recognisable. In the diesel scandal, which came to light in the United States in 2015 and subsequently affected a considerable part of the European industry, the engine management detected that a test-bench run was under way and switched the exhaust treatment into a state it did not maintain on the road. The measurement was correct. Only what was measured did not occur in service.

The advantage of the measuring point therefore holds only on one condition: that nobody tunes the vehicle to pass the measurement. As soon as the test itself becomes the object of development, it no longer measures the thing but the adaptation to it.

The answer was to introduce, after the fact, measurements in real driving. That is the right direction — and at the same time an admission: the measuring point had to be corrected because it had become the goal. That is precisely what this text is about, one level down.

III. Defined rather than estimated

And here lies the actual break, for it would have been avoidable.

In engineering it is the normal case to estimate quantities that cannot be measured directly. One derives them from measurable values, gives a range rather than a figure, carries a safety factor and states the assumptions. The result is unsatisfying and nonetheless honest, because the uncertainty stays visible. Whoever goes on calculating with it knows what he is relying on.

Such an estimate would have been possible here too. One could have said: on present knowledge the reduction lies between so and so many per cent, depending on electricity mix, mileage and battery life, and this range will be reviewed every five years.

Instead it was defined. And therein lies the difference. An estimate carries its tolerance with it and invites improvement. A definition has no tolerance. It is not imprecise — it is valid until someone changes it. An open question became a fixed quantity, and by a legal act rather than a measurement.

One may suspect that this was not merely convenient but tactical. An honest balance sheet would have given electric mobility, at an early stage, a target its opponents would have used — and a technology that is not yet mature is judged by its present figures, not by its future ones. Whether that was the reason we do not know, and we do not assert it.

But if it was the reason, something uncomfortable follows. Then the zero was not a metric but a political instrument. Then one would have to call it that — and could change it without a worldview collapsing.

IV. Why the study comes only now

One circumstance of the matter deserves attention of its own, and it has nothing to do with the content of the study.

The question is old. Life-cycle assessments for vehicles have existed for decades, their methodology is standardised, and that considerable quantities of CO&sub2; are embodied in battery, steel and aluminium was never a secret. It was not the possibility of investigating this that was lacking.

What is new is the timing. And that can be described without imputing anything to anyone.

A few years ago the decision of direction was still open. A study of this kind would have been read as a contribution to a fundamental confrontation — and its author would have found himself in a camp, whatever he wrote. Today the decision has been taken, the investments have been made, and the dispute is only about review clauses and deadlines. The same study has thereby turned from an attack into a contribution.

Science too has a tempo. Not because anyone keeps silent, but because a study needs funding, a commissioner and a readership, and all three depend on whether the question is currently open. As long as it is not, the work finds no financing, no occasion and no readers.

For the finding of this text that means: the metric was not merely set. For a time it also made the examination of itself unattractive.

V. What a metric does to effort

A metric does not only measure. It steers.

As soon as measurement takes place at the tailpipe, every improvement at that point pays — and none anywhere else. A manufacturer who sources lower-emission steel gains nothing from it in any calculation. Whoever moves cell production to a place with clean electricity does not improve his balance, because that balance is not kept at this point. Whoever doubles the service life of a vehicle has halved the manufacturing burden — and gets no credit for it.

All three are effective contributions. All three are invisible.

The regulation is therefore not merely imprecise. It has changed the direction of effort: effort flows to where it is measured. That is neither evasion nor fraud; it is the reasonable response to the rule as written.

And from that follows the general proposition this text is about:

A metric that takes the place of the goal does not optimise the goal but itself.

VI. Why this is no isolated case

The same pattern is found wherever something has to be regulated that cannot be measured directly.

Machine tool engineering knows it at a point we have described elsewhere: the damping of a machine arises almost entirely in its joints, but only the damping of the whole system can be measured. Where the values are missing, one calculates with what one knows — the material — and overlooks what matters.

The rule of law knows it as well: it considers what is in the files. Whoever documents has the advantage over the one who was merely right.

In all three cases the same construction. The procedure captures what leaves traces — and what leaves none is treated as though it did not exist. The tailpipe leaves traces. The cell factory overseas does not.

VII. What the study actually says

This text arose initially from a report about a report. The work is now available, and three of its findings belong here.

The 41 per cent is a mean, and the range is the substance. In roughly 92 per cent of the 47 scenarios the electric vehicle performs better. But the range runs from minus 89 to plus 21 per cent — there are scenarios in which the electric vehicle performs worse. What decides is where the vehicle is built, how carbon-intensive its production inputs are, and what energy carrier it runs on.

And this is precisely the variance the tailpipe measurement cannot see. It certifies two vehicles as identical whose real footprints may differ by a factor that matters.

The term is more precise than ours. The work distinguishes decarbonization from defossilization — the point is not carbon as such, but fossil carbon drawn from geological storage. Whoever closes a loop draws none anew.

And its central sentence is this essay’s thesis, more sharply put: a regulation that cannot see a contribution cannot reward it, and what regulation does not reward, industry will not finance at scale.

From this follows the actual charge, and it is not the one the headlines carried. It is not that the electric vehicle is being talked down. It is that the entire regulatory incentive rests on a single lever — the drivetrain — while low-carbon steel, defossilized battery production, renewable energy carriers and circularity carry no regulatory value at all.

And a lever without a fallback has no answer if it alone proves insufficient. The work cites an analysis from 2023 according to which, at the stated pace, the passenger-car sector would deliver only around 80 of the roughly 188 million tonnes of reduction needed by 2030. That the Commission introduced a multi-annual compliance mechanism for 2030–2032 is itself an acknowledgement that this risk is real.

VIII. The counter-case

A text that stopped at the criticism would be incomplete. There are reasons for the rule as it stands, and they must be named.

Measurability is a value in itself. A rule that cannot be checked will not be followed. A life-cycle balance whose system boundaries the manufacturer helps to determine would be an invitation to calculate as required. Whoever doubts this need only think of the test cycle: even a comparatively simple measurement could be subverted. A balance over a worldwide supply chain that nobody can verify would be many times more vulnerable.

The electricity mix changes. A vehicle built today runs for ten or fifteen years. In that time the electricity with which it is charged becomes cleaner. Whoever calculates with today’s mix underestimates it systematically — a combustion engine, by contrast, does not improve; it burns the same on its last day as on its first.

And the level of regulation is a different one — this objection carries the most weight. It is not the case that everything arising outside the tailpipe goes unaccounted for. The fleet targets are one mechanism among several:

• Emissions trading raises the cost of carbon-intensive production of vehicles, of steel and of electricity.
• A carbon border levy applies to imported steel, so that production does not simply move abroad.
• The battery directive requires transparency about the origin of raw materials.
• The greenhouse gas reduction quota acts on oil companies and filling stations.
• A second emissions trading system acts indirectly on drivers.

And the principle behind this is not accident but construction: each measure addresses particular parties with exactly one task that they can actually carry out. A carmaker can change the drivetrain; it cannot change the electricity mix in Poland. To burden vehicle regulation with the entire supply chain would be to give it a task it cannot discharge — and to count the same emissions twice.

Whoever wants to change the attribution must therefore say where it should take place instead. Otherwise he merely moves the problem.

The study answers that life cycle assessment is long established in EU practice — in the Ecodesign Regulation and the Carbon Border Adjustment Mechanism. It therefore asks for nothing new, only for transfer. Whether that disposes of the objection remains contested.

IX. What follows

Not that electric mobility is a wrong turn — the study does not say that either; it certifies a considerable reduction.

Rather, two things.

Whoever sets a metric decides the direction of effort. That is a greater power than setting a goal, and it is usually exercised by people who take themselves for technicians. The choice of the measuring point is not a technical question but the properly political one.

And a metric that has become the goal can no longer be corrected without someone breaking his word. Whoever declared today that the zero-emission vehicle was never zero-emission would have to explain why an industrial policy was hung on it for fifteen years. That is why such quantities are seldom changed but rather supplemented, stretched and furnished with exceptions until they say nothing at all.

That is the point at which it is decided whether a body of rules is capable of learning: whether it still regards its own metric as provisional.