How to Lie With SUV Statistics

Shortly after I published my piece on the Mayor’s proposed SUV charge, London’s Walking & Cycling Commissioner posted this:

There's growing evidence of safety risks of bigger cars.

Heavier with higher bonnets & worse vision, bigger SUVs are:
⚠️14% more likely to kill you than if you're hit by a smaller car
⚠️increasing to 77% for children
⚠️and 200% more likely to kill a kid under 9#VisionZeroLDN pic.twitter.com/vIfEGmS2cA

— Will Norman (@willnorman) March 13, 2026

Let’s take the numbers seriously, because the source research – from the London School of Hygiene & Tropical Medicine – is legitimate. The findings come from a UK-specific study using STATS19 police crash data, and a broader international meta-analysis covering 680,000 collisions across 24 studies and 35 years.

The tweet presents three claims: SUVs are 14% more likely to kill an adult pedestrian, 77% more likely to kill a child, and 200% more likely to kill a child under 9 – all compared to a smaller car. Those are real findings, but the way they’re presented is a masterclass in how to make a modest statistical finding feel terrifying.

The biggest issue wit the tweet is that the base rate is invisible. Every one of those percentages is a relative risk – the increased likelihood of death given that you’ve already been hit by a car. That’s a conditional probability, and the condition is doing all the heavy lifting. Being struck by any vehicle as a pedestrian is rare. Being killed having been struck is much rarer. A child being killed by a vehicle in London is, mercifully, an extremely rare event – probably in the low single digits per year across the entire city.

London had around 100 road fatalities in 2024, of which roughly 40–50 were pedestrians. A subset of those involved children. A further subset involved vehicles classified as SUVs. A 77% increase applied to a number that is likely 2 or 3 per year means one or two additional deaths – each one a genuine tragedy¹, but a very different policy proposition from what the tweet’s formatting (warning emojis, bold claims, a child dwarfed by a Range Rover) is designed to make you feel.

Not to reiterate my previous post too heavily, but the category is too broad to be useful. The studies define “SUV” broadly enough to include everything from a Nissan Juke (4,210mm long, 1,595kg) to a Range Rover Sport (4,946mm long, 2,300kg+). The physics of being struck by a vehicle with a bonnet height of 1,100mm is fundamentally different from being struck by one at 750mm. Averaging them into a single risk ratio washes out the very gradient that matters. A Hyundai Kona has more in common with a Volkswagen Golf than with the Range Rover in the tweet’s image – but the statistic treats them as the same kind of vehicle.

It’s also worth noting that the international meta-analysis – which produced higher, scarier numbers (44% for adults, 82% for children, 130% for under-10s) – drew 16 of its 24 studies from the United States, where “SUV” means something physically much larger than what that term captures in the UK fleet. The UK-specific study produced notably lower figures, which is exactly what you’d expect when the typical British “SUV” is a crossover rather than a Ford Expedition.

Further, the image he posted is arguing against the statistics: look at the photo attached to the tweet. A small child walking past a full-size Range Rover, its grille towering over her head. The image is implying that “this is what we’re talking about.” The 14% statistic is measuring something much broader and less dramatic – an average across every vehicle a manufacturer chose to market as an SUV, most of which are small crossovers that would barely register as threatening in the same photograph. The emotional case is built on the largest vehicles; the statistical net catches the smallest ones. This is the same move as the BBC using a mud-splattered Range Rover to illustrate a story about a policy that would mostly affect Nissan Qashqais.

None of this means the underlying research is wrong: larger, heavier vehicles with taller front profiles do cause more severe injuries in pedestrian collisions. That’s physics, and it’s well established. But if the concern is really about bonnet height and front-end geometry – and the research suggests it is – then the policy response should target bonnet height and front-end geometry. Not a marketing label that captures a Peugeot 2008 and a Range Rover in the same regulatory net while letting a 5,289mm Mercedes S-Class sail through untouched.

I don’t want to go too crazy with this series of posts and turn this blog into a ranting about the mayor zone. It just always really grinds my gears when see politicians use statistics in such a lazy and shallow way to justify whatever policy they want. I also would like to add that whilst I have seen some violence against statistics from Will Norman on Twitter before, I have also seen a striking amount of abuse aimed at him on there as well. As ‘Walking & Cycling Commissioner’ his job is advocacy for walking and cycling, in what should be a relatively milquetoast public service job – not this. Part of me feels disturbed that I am sitting here writing a defence of SUVs of all things! Like I don’t even like SUVs: they are boxy, not fun to drive and overpriced for what they are.² I just wish we had better public policy and politicians who took what they are proposing and saying a lot more seriously.