The cold is not what empties the battery.

AAA put five electric cars on a dynamometer and ran each of them twice at 20 °F: once with the heater off, once with it on. Cold alone cost 12 per cent of combined range. With the cabin heater running it was 41 per cent. So driving in the cold costs an eighth and heating the cabin costs the other 29 points — and only one of those is something you can choose differently when you buy.

The cold takes an eighth. The heater takes the rest

The 2019 report is the one worth reading, because it is the only one that separates the two things people mean when they say “cold kills range”. Same cars, same drive cycle, same test cell — the only variable is whether the HVAC system is switched on.

Read the first two bars against each other. Between them is the entire winter problem: 29 points of range, spent on keeping the cabin at a temperature you chose. A petrol car heats its cabin with waste heat it cannot avoid making, and has it to spare. An electric drivetrain is efficient enough that there is far less of it, so most of each degree comes out of the same battery that moves the car.

Three cars, one freezer, seven years later

AAA came back to the question in April 2026 with three current electric cars — not a repeat, because the heater-off run is gone: the cabin is held at 72 °F in every scenario, which is how anybody actually drives in winter. What that buys instead is a comparison between cars measured identically, and they are 21.7 points apart.

The car with the most range at 75 °F finished behind the car with the least. The Chevrolet Equinox EV has the longest range of the three at 320.8 miles and ends the cold test on 177.2, behind the Ford Mustang Mach-E, which started 33.5 miles short of it and finished on 214.6. A range figure quoted at room temperature does not rank these cars in the weather you bought them for.

Thirty thousand cars in the actual weather

A dynamometer holds everything still except the thing being measured, which is its strength and its limit. Recurrent takes the other approach: telemetry from more than 30,000 cars in the United States, driven by their owners in whatever weather arrived. Their figures are gentler than the laboratory’s and they should be — a real trip includes preconditioning on the plug, short journeys in a garage-warm car, and drivers who turn the heat down.

• At 32 °F, cars kept 78 per cent of their maximum range on average. At 20 °F, 70 per cent.
• The best model held 88 per cent at 32 °F and the worst 69 per cent — a 19-point spread between cars in the same freezing weather.

The part everybody names, and what it does not explain

A heat pump is the part everybody names, and it is worth naming. A resistive element turns one unit of electricity into one unit of heat; a heat pump moves heat rather than making it, so the same unit delivers three or four. Across Recurrent’s fleet that is worth about 10 per cent more range at 32 °F.

That 10 points is a fleet average, not a ceiling for any one car. It is what Recurrent reports from a fleet of every kind of car, so it says what a heat pump is typically worth — not the most one can be worth, and not what any particular implementation achieves. It cannot be subtracted from the 21.7 points between the three cars above to leave a remainder.

AAA is careful about this where the summaries are not. Its explanation of the Ford Mustang Mach-E’s result names two things, not one: a patented vapour-injection heat-pump architecture in the 2025 model year and revised thermal control algorithms. How the car decides to warm the battery, when it draws on waste heat, and how much it heats before you have driven anywhere are software questions, and they do not appear on a specification sheet at all.

What this page cannot tell you is how much of the 21.7-point spread above is the heat pump and how much is everything else. That would need to know which of the three cars has one and what each implementation achieves, and neither AAA nor Recurrent publishes it car by car. Anyone who tells you the spread proves a point about heat pumps is reasoning past their evidence, and so was an earlier draft of this page.

What to do about it

1. Precondition on the plug. Warming the cabin and the battery while still connected spends grid electricity rather than range. It buys the initial warm-up only — the heater keeps running once you unplug, so this is a bite out of the 29 points above and not a way round them.
2. Heat the seat before the air. Seat and wheel heaters warm a person directly; the cabin heater warms several kilos of air first.
3. Plan the winter number, not the sticker. A Tesla Model Y that returned 297.7 calculated miles at 75 °F returned 157.7 at 20 °F on the same dynamometer. That is not an EPA rating and it is not a sticker; it is one car on one cycle. Recurrent’s fleet sees a gentler 70 per cent at 20 °F, but against a different baseline and across every model at once, so the two numbers do not bracket a single car.

What nobody has measured

Nobody has published a controlled comparison of preconditioning against not preconditioning on the same car, over the same cycle, at the same temperature — which is the number every one of these reports implies and none of them measures. The advice above is mechanism and arithmetic, not a measurement, and it is worth exactly that.

The two AAA reports are also seven years apart on different cars, so the 12/41 split from 2019 is not a property of the three cars in the 2026 table. It is the reason to look at the split at all.