When do nutri-score labels shift food choices? Evidence for moderation by health-goal activation and prior beliefs

Why they ran it

Front of package labels are meant to steer shoppers toward healthier products, but the evidence on whether they do has been mixed. The authors set out to work out when the label shifts choices, and whether that depends on health being on a shopper's mind at the moment of choice and on whether the rating agrees with what the shopper already believed.

Drawn from the paper's abstract.

Researchers ran a randomised online experiment with 792 Italian adults, who worked through 30 pairs of familiar grocery products and picked which one they would buy. Two things varied between participants: whether the Nutri-Score label appeared on the products, and whether people first saw a brief health goal prompt asking which product was healthier. The team also sorted the product pairs by whether the rating matched or clashed with each participant's own pre-label judgement of which item was healthier, which is the comparison the paper is really built around.

Overall, showing the label increased how often people picked the option with the better rating. The interesting part is where that effect concentrated. The shift was largest when the label confirmed what participants already believed and the health goal prompt was present, so the two supports stacked on each other. When the rating contradicted prior beliefs, it moved choices on its own, prompt or no prompt. The abstract reports these patterns as directions rather than as effect sizes, so how much choice shifted, and across how many of the 30 pairs, is not given there.

The authors add a wrinkle that cuts the other way. When a tempting product turns out to carry an unexpectedly favourable rating, that rating may be read as permission rather than as information. They flag this as something policymakers should weigh, since a label built to steer people away from one product can quietly wave them toward another. The design categorised pairs by belief alignment and did not test that licensing idea directly, so it stays a possibility the authors raise rather than a result.

The numbers

undefined	undefined
undefined	undefined
undefined	undefined
undefined	undefined

Why this might happen

Not available The full discussion is not openly available and the abstract offers no mechanism, so we do not know what the authors think caused this.

The authors do not offer an explanation we can point to, so we are not going to invent one.

Drawn from no reachable discussion, the abstract was the deepest source available.

What this does not show

It cannot show that the label changes what people actually buy. Every choice was made in a single online session across 30 fixed product pairs, with no price, no budget and no basket. Nothing in the design tracked a purchase.
It cannot tell you how big the label effect is. The abstract states the direction of the main effect and where it was strongest, but reports no effect sizes and no share of pairs that flipped.
It cannot confirm that good ratings on tempting foods create a sense of permission. The study measured choices, not the reasoning behind them. The authors offer that idea as an interpretation of the pattern, and the abstract describes no measure that tested it.
It cannot say anything about what people ate or how healthy they were. The only outcome was which product was selected in each pair. No intake, no follow up, no health measure.

Where this leaves us

It is now more likely that the label does move choices toward the better rated product, but conditionally rather than uniformly. That refines the mixed prior record: the effect was largest when the label confirmed prior beliefs and health had just been made salient, and it still moved choices without that prompt when it contradicted prior beliefs.

792 Italian adults choosing between pairs of familiar grocery products in a randomized online experiment, thirty pairs each.

The same two manipulations run on real grocery purchases rather than choices in an online task, so the outcome is what leaves the shop.

Caveats worth holding

These were hypothetical choices made on a screen, with no money spent and no food taken home.
The sample was Italian adults, and this label is a familiar sight in some European markets and unknown in others, so the starting knowledge of shoppers elsewhere may differ.
Thirty pairs of familiar products is a narrow slice of a real supermarket, and the abstract does not list which categories were used.
The abstract does not report funding or how participants were recruited and compensated.