What a Car's CO2 Figure Actually Measures

Nobody put a probe in your exhaust pipe

In the type-approval laboratory, the exhaust is diluted, collected in bags and read by analysers, so the CO2 really is measured there — and the fuel consumption printed on the brochure is then computed from it. On the road it runs the other way: nobody measures your car's CO2, so the figure is computed from the fuel you burned. Both directions use the same identity, called the carbon balance, and the identity is exact.

This matters because it kills a whole category of wishful thinking. There is no CO2 dial on an engine that is separate from the fuel dial. A catalytic converter cleans up carbon monoxide, unburned hydrocarbons and nitrogen oxides; it cannot remove CO2, because CO2 is what complete combustion is for. Any advert claiming lower CO2 without lower fuel use is claiming that carbon has gone missing.

The derivation, on one line

Start with the molecule. Carbon has an atomic mass of 12.011, oxygen 15.999, so CO2 weighs 12.011 + 2 × 15.999 = 44.009. Every carbon atom that goes into the cylinder comes out inside exactly one CO2 molecule, so every gram of carbon burned becomes 44.009 ÷ 12.011 = 3.664 grams of CO2. Two-thirds of that mass was pulled out of the air, which is why a 6-pound gallon of fuel produces 20 pounds of exhaust gas.

Now the fuel. Petrol is close enough to iso-octane, C8H18, whose molar mass is 8 × 12.011 + 18 × 1.008 = 114.232 — so 96.088 ÷ 114.232 = 84.1% of it is carbon by mass. The US Department of Energy does the same sum with the real blend and rounds it: a gallon of gasoline "weighs about 6.3 pounds", carbon is about 87% of that, giving 5.5 lb of carbon, and 5.5 × 3.7 ≈ 20 lb of CO2. Real petrol has more carbon per hydrogen than pure iso-octane because it contains aromatics, which is why the published carbon fraction sits a little above the textbook 84%.

EPA's published numbers are 8,887 grams of CO2 per gallon of gasoline and 10,180 grams per gallon of diesel. The gap between a back-of-envelope 8,700 and EPA's 8,887 is about two per cent, and all of it lives in the assumed density and carbon fraction of "average" fuel. Nobody disagrees about the chemistry; they only disagree about what is in the tank.

Which is why the two numbers are one number

Once the grams-per-gallon constant is fixed, converting between fuel economy and CO2 is a single division. Grams of CO2 per mile = 8,887 ÷ miles per gallon. A 30 mpg car emits 296 g/mile; a 40 mpg car emits 222; a 20 mpg truck emits 444. Nothing about the engine, the transmission or the aerodynamics enters the conversion — they all act on the fuel figure, and the CO2 figure follows.

This also explains a figure that gets quoted as if it were a separate discovery. EPA says the average American passenger vehicle emits about 400 grams of CO2 per mile. Divide 8,887 by 400 and you get 22.2 mpg — which is precisely the average fuel economy EPA assumes. The 400 g/mile is not an emissions measurement at all; it is the fuel economy of the US fleet, written in different clothes.

One trap for anyone comparing figures across the Atlantic: a US gallon is 3.785 litres and an imperial gallon is 4.546, a difference of 20.1%. A car quoted at 40 mpg in the United States and 48 mpg in the United Kingdom is the same car with the same emissions.

Layer one: the fuel had a life before your tank

Everything above is tank-to-wheel: the carbon that was in the fuel when you bought it. Getting it there took energy too — extraction, refining, shipping, trucking to the forecourt. That upstream share is called well-to-tank, and the UK government's published factor for gasoline adds about 26% on top of the tailpipe figure — roughly 2,280 grams of CO2-equivalent per gallon burned, before a drop of it reaches your engine.

Applied to a 30 mpg car, the tailpipe 296 g/mile becomes about 372 g/mile well-to-wheel. The correction is close to uniform across petrol vehicles, which is why it changes almost no comparison between two combustion cars — and why it changes every comparison between a combustion car and something that does not use liquid fuel at all.

Layer two: which test, and how far it is from your commute

The United States and Europe do not use the same test. EPA runs a five-cycle procedure on a dynamometer, covering city, highway, high-speed, air-conditioning and cold-start conditions, and applies adjustments before the number reaches the window sticker. Europe uses WLTP, the Worldwide Harmonised Light Vehicles Test Procedure, which replaced the old NEDC. A figure from one system cannot be dropped into a sentence about the other.

Whatever the test, a laboratory is not a road. Since 2021 European cars have carried on-board fuel consumption meters, and the Commission and the European Environment Agency have published what roughly eight million of them reported. For cars first registered in 2021 the real-world gap averaged 19.8% for petrol and 18.2% for diesel; for 2022 registrations, 21.1% and 17.1%. Applied to the 2025 fleet target of 93.6 g/km, a 21% gap turns it into about 113 g/km on the road.

The same data set contains a much larger discrepancy. Plug-in hybrids registered in 2021 emitted around 3.5 times their laboratory value in real use, and the gap widened towards five times for 2023 registrations. The physics is not mysterious: the laboratory assumes a plug-in is plugged in, and a great many of them are not.

Layer three: zero at the tailpipe is a definition, not a result

An electric car has no tailpipe, so its tank-to-wheel CO2 is zero — not approximately zero, exactly zero, by construction of the metric. That is not a claim about the world; it is a statement that the metric stops measuring at the vehicle boundary. Move the boundary to the power station and the answer becomes a question about geography.

Take a car drawing 0.32 kWh per mile at the wall, charging losses included. EPA's national average grid factor, 823.1 lb of CO2 per MWh, is 373 grams per kWh, so that car emits about 119 g/mile well-to-wheel — against 400 g/mile for the average vehicle on the road and 296 for a 30 mpg one. The same car on a coal-heavy regional grid would be worse; on a hydro or nuclear one, close to nothing.

None of this counts manufacturing, and battery production is genuinely front-loaded carbon. A complete comparison is a life-cycle assessment, which is a much bigger exercise than a division. But the point stands: the tailpipe figure is a boundary condition, not a verdict, and reading it as a verdict is exactly the mistake the number invites.

What to do with all this

Three rules follow. First, when comparing two combustion cars, compare fuel economy and ignore the CO2 column entirely — it carries no extra information. Second, when a manufacturer quotes a CO2 improvement, convert it back to fuel and see whether the saving is one you would notice. Third, when comparing across technologies, insist on the same boundary for both, and say which grid you assumed.

The calculator on this page does the first step for you in both directions, so you can check any figure you are shown against the chemistry that produced it — which, as it turns out, is the only part of the whole business that is not negotiable.

Frequently asked questions

Is the CO2 figure measured or calculated?

Both, depending on where you stand. In the type-approval laboratory the diluted exhaust is analysed directly, so CO2 is measured and the published fuel economy is derived from it by carbon balance. On the road nobody analyses your exhaust, so the CO2 is derived from the fuel you bought. The link between them is chemistry, and it runs in either direction with equal accuracy.

Why does diesel emit more CO2 per gallon but sometimes less per mile?

Diesel is denser and slightly richer in carbon, so a gallon of it contains about 14.5% more carbon than a gallon of gasoline: 10,180 grams of CO2 against 8,887. But a diesel engine, running unthrottled at a high compression ratio, converts more of the fuel's energy into work, so it burns fewer gallons for the same distance. Whether the per-mile figure ends up lower depends on whether the efficiency advantage beats the 14.5% penalty per gallon.

Does a catalytic converter or a particulate filter reduce CO2?

No — they slightly increase it. A three-way catalyst oxidises carbon monoxide and unburned hydrocarbons, and both of those reactions produce CO2. A particulate filter traps soot, which is almost pure carbon, and burns it off during regeneration, again into CO2. Aftertreatment cleans up what harms lungs and smog chemistry; it has no mechanism for removing the end product of combustion.

How much CO2 does an average car emit in a year?

EPA's arithmetic for the United States: 22.2 mpg, around 11,500 miles a year, 8,887 grams of CO2 per gallon, giving about 4.6 metric tons a year. You can redo it for your own car in one line — annual miles ÷ mpg × 8,887 grams — and the answer will be right to the accuracy of your mileage estimate, because every other term is a constant of chemistry.

Are WLTP and EPA figures interchangeable?

No, and there is no official conversion factor between them. They use different drive cycles, different ambient conditions, different treatment of options and different adjustments before publication. EPA's procedure is generally regarded as the more conservative of the two, but a rule of thumb is not a conversion. If you are comparing a European brochure with a US window sticker, you are comparing two different experiments.

Is an electric car always cleaner?

On the well-to-wheel measure used here, an electric car beats a comparable gasoline one on almost any grid in current use, because a power station plus a motor converts fuel to motion more efficiently than a small engine does. The margin, though, ranges from overwhelming to modest depending on the generating mix. Manufacturing is a separate account, and settling it properly needs a life-cycle study rather than a division.