The empty rack was the expensive part.

Consumer Reports measured two 2019 cars at a steady 65 mph with a bare roof, then with a roof rack fitted and nothing on it, then with a carrier on top. On the sedan the empty rack alone cost 11 per cent — more than half the penalty of the whole loaded set-up.

The rack alone, carrying nothing

Of the 19 points the loaded set-up cost, 11 were the bars by themselves. The box you bought is the visible half of the bill. The rack you left on after the holiday is the other half, and it goes on charging you every motorway mile until you take it off.

Why the sedan lost several times as much

The same test on a RAV4 came out very differently. The empty rack cost the SUV about 2 per cent against the sedan’s 11 — roughly about 5.5× less — and since CR describe the SUV figure as “nearly 2 per cent”, read that ratio as approximate rather than exact.

It is not that the SUV is efficient. It is that the SUV was already pushing a large flat front through the air, so a pair of bars adds little to what it was doing anyway. A saloon has a smooth roofline and the air was staying attached to it; the bars break that up, and the car loses something it actually had.

The practical version: the sleeker your car, the more a roof rack costs you, which is the opposite of how people usually assume it works.

The same car, the same route, two carriers

Car and Driver ran the comparison that decides the purchase, on one long-term Toyota Grand Highlander, back to back on the highway with one variable changed:

4 mpg, on the same car, for putting the same luggage behind the car instead of on top of it. A hitch carrier sits in air the vehicle has already disturbed. A roof box sits in clean fast air and has to push it aside.

The same team tested a Kia Carnival, 2022 with its factory crossbars fitted and then removed: 25 mpg against 28. 3 mpg for bars that came with the car and were carrying nothing at all — which is the same finding as the Altima’s, arrived at from a different direction.

Where the published numbers disagree

Consumer Reports’ own figures do not fully reconcile, and we are not going to paper over it. They give the RAV4 as 41 mpg bare and 35 loaded, which on the face of it is a 14.6 per cent drop and 6 mpg — while the text alongside says 13 per cent and “just over 5 mpg”.

The likely explanation is unremarkable: the mpg are published rounded to whole numbers and the percentages were computed before rounding. We print CR’s percentages rather than our own arithmetic on their rounded figures, because they had the unrounded data and we do not — and we would rather show you the seam than hide it.

What to buy

The measured advice is awkward for a page that earns a commission on roof boxes: if your car can take a hitch, the hitch carrier measured 4 mpg better. A roof box is the right answer when there is no hitch, when the load is long and light, or when you need the tailgate to open — not because it is the efficient option, because it is not.

• Hitch-mounted cargo carriers on Amazon The option that measured 4 mpg better on the same car — check your hitch class and tongue weight
• Rooftop cargo boxes on Amazon If there is no hitch. Budget the fuel as part of the price, and take the bars off afterwards

Affiliate links — we earn a commission at no extra cost to you and it does not change what we recommend. The single most profitable action on this page costs nothing: take the empty rack off. How this site makes money.

What nobody has measured

1. Which box. Both tests measure the idea of a rooftop carrier, not brands against each other. No independent test we found puts several boxes on one car and measures them, so we are not ranking them.
2. Speed. Drag rises with the square of speed, so these penalties are specific to the speeds tested. Nobody has published the curve for a loaded box across a range of speeds on one car.
3. Crosswind. Every figure here is from still-air testing on a road. A side wind changes the drag on a roof box considerably and is not in any of these numbers.
4. Your car. Four vehicles across two publications. The direction is consistent and the magnitude is not: 2 per cent to 11 for an empty rack, on two cars alone.