A Food Label That Warned Us in 2012 Stopped Warning Us by 2018 — Here’s How We Know

Why I care about this. My work is about making sure health policies keep working for the people who need them most — not just at launch, but years down the line. This study looks at what happens to a well-meaning food label after the initial fanfare fades, and whether the people it’s supposed to protect are still being protected.

Randomised trials — the kind where you assign some people one thing and other people another, and compare the results — are the most reliable way to prove something works. But you can’t run one on a label that’s already been on every cereal box and ready meal in the country since 2013. You can’t randomly stop showing it to half the population just to see what happens. So how do you find out whether a label that’s been around for years is still doing its job?

Detective work, not a lab experiment

This is where a different kind of method comes in — one that works more like detective work than a lab experiment. I used six years of real survey data from nearly 9,000 UK adults, collected between 2012 and 2018, and asked a simple question: did people who found the label easier to read actually eat less of the foods it was meant to discourage?

The tricky part is proving the label caused any change, rather than just something else going on at the same time — general health trends, for instance. My solution was to compare foods the label was actually placed on and meant to influence (like pre-packaged sandwiches) against foods that either weren’t labelled at all, or were labelled but weren’t the label’s real target (like dairy products). If the label was genuinely doing something, its effect should only show up for the foods it was actually trying to influence — not everywhere. It’s a bit like testing whether a new road sign changes driving behaviour only near that sign, not on every road in the city.

A warning that quietly turned into the opposite of a warning

For most food categories, nothing changed — exactly what you’d expect if the effect wasn’t real. But for pre-packaged sandwiches, something notable happened. In 2012, people who found the label easier to read ate that food less often — the label was working exactly as intended, as a kind of warning sign. By 2018, that had flipped: people who found the label easier to read were now eating that food more often.

The label hadn’t changed. The product hadn’t changed. But its effect on people’s behaviour had reversed completely, and a single snapshot from any one year would never have caught it — you needed to watch it over time to see the drift.

Why would a warning sign stop warning people?

Think about a “wet floor” sign. The first few times you see it, you’re careful. After you’ve walked past it a thousand times in the same building, you stop noticing it at all — or worse, it becomes just part of the scenery that makes the floor feel normal, even reassuring. Something similar may be happening here: a label that once grabbed people’s attention as a caution may have simply become familiar wallpaper, or even started to feel like a stamp of approval rather than a warning.

And there’s a detail that matters even more: this shift was strongest among people who were less engaged with health information in general — exactly the group a nationwide label is most relied on to reach, because they’re the ones least likely to be actively seeking out nutrition advice elsewhere.

The lesson for policy

A policy that worked when it launched isn’t guaranteed to keep working. Public health labels tend to get evaluated once, when they’re introduced, and then left alone — treated as a solved problem. This research suggests that’s a mistake. Without checking back in, a label could be quietly losing its power, or even working against its original purpose, for years, without anyone noticing — and the people most affected are often the ones already hardest to reach.

That’s the real question this research is trying to get policymakers to ask more often: not just “did this work when we introduced it,” but “is it still working now?”

The full peer-reviewed study, along with my other published work, is available on my Google Scholar profile.