Burdens of proof for dynamic norms interventions

Are they visible? Memorable? Plausible? Persuasive? Well-targeted?

Peer Pressure 4 Good

If you’re a researcher who wants to get people to do something — drink less alcohol, report incidents of domestic violence, use less electricity, etc. — it’s common to appeal to social norms. Per Miller & Prentice (2016), such interventions reflect “a growing disillusionment with the capacity of factual information and economic inducements”¹ for changing attitudes, behaviors, and beliefs.

Using social norms to reduce meat consumption has proven challenging because eating meat is extremely normal in both the ‘common’ and ‘approved of’ senses of the word. Meat consumption is rising almost everywhere and has been for decades. In the United States, eating meat is sacrosanct, and opposition to it is about half as prevalent as opposition to contraceptives:

Evans, J., & Corichi, M. (2026, March 19). What do Americans consider immoral? Pew Research Center. https://www.pewresearch.org/religion/2026/03/19/what-do-americans-consider-immoral/

In Germany, telling university students that a majority of their peers are vegetarian and that their next meal will be observed by those peers — again, activating both the ‘common’ and ‘approved of’ senses of normal — has no overall effect on behavior, and in fact causes men to eat more meat than they would otherwise (Dannenberg, Klatt & Weingärtner 2024).

Eating vegetarian is not like voting

To get around this, Sparkman & Walton (2017) employ what they call dynamic norms: “drawing attention to the change of a norm over time.” They propose that dynamic norms may work through two mechanisms. The first is preconformity: “people may anticipate ongoing change and a future world in which that behavior is normative and then conform” to where they think everyone is going. The second is encouragement to “reconsider specific barriers that they had assumed would prevent change…People may see a change as, among other things, more important, more possible, or more appropriate than they had thought.”

Here’s an example dynamic norm message from Sparkman et al. (2020):

Figure 1 from Sparkman, G., Weitz, E., Robinson, T. N., Malhotra, N., & Walton, G. M. (2020). Developing a scalable dynamic norm menu-based intervention to reduce meat consumption. Sustainability, 12(6), 2453. https://doi.org/10.3390/su12062453

I think this is a really nice theoretical framing, and it’s inspired a small but growing literature of applications to meat reduction.

Does it work though?

Dynamic norms and meat consumption: a mixed bag of small, mostly null effects

For our 2025 meta, we coded these four dynamic norms papers:

• Aldoh, Sparks, & Harris (2024) run a Prolific survey where some participants see this message: “More and more people in the UK are changing. In 2020, 33% of British people—a figure increasing every year over the previous 5 years—successfully engaged in one or more of the following behaviours to eat less meat,” along with a list of those behaviors, compared to a static norm. Attitudes and intentions were measured immediately and consumption was measured a week later. The study “found no positive effect of dynamic norms (versus static norms) on any outcome at time 1, and no positive effect on changes in outcomes from time 1 to time 2.”

• Çoker et al. (2022) test a dynamic social norm message at an in-store restaurant at a major chain in the UK: “More and more [retail store name] customers are choosing our veggie options.” The study found “no evidence of an effect of the

  intervention…on the percentage sales of meat- vs plant-based dishes.” However — once again! — there were implementation snafus: “adherence to the intervention procedure was often low, with inconsistencies in the placement and display of the intervention message.”

• Sparkman et al. (2020) run four field experiments that test “dynamic-norm messages by incorporating them into restaurant and web-based menus.” The interventions increased vegetarian sales in the first three studies by “1–2.5 percentage points" — ahem — “although this effect does not always reach statistical significance and varies across populations and analytic models.” The fourth study, however, “significantly reduced vegetarian orders compared to the control condition” by 3.7 percentage points.²

• Finally, Sparkman et al. (2021) have American participants on mTurk read op-eds that discussed “a rise in people curbing meat consumption, and the beneficial consequences of this dietary trend on personal and public health, the environment and mitigating climate change, and animal welfare.” Some participants saw op-eds encouraging then to cut back and others saw text about eliminating all meat consumption. Overall, the ‘reduce’ op-eds motivated “people to eat less meat over a substantial time frame—at least 5 months without diminished effect. These changes in diet are fairly substantial, representing a 7–9% reduction in total servings of meat consumed. By contrast, an eliminate appeal has a smaller, non-significant effect (2–4% reduction).”

  • By the way, all these data are self-reported, but I tend to take that at face value. YMMV.

Since our review period ended, a few more papers have come out:

• Weikertová et al. (2024) conduct web-based experiments on “trends in the adoption of a meatless diet” and “found that dynamic norm information influenced perceived dynamic norms but not the preference for a meatless diet. Our Bayesian meta-analysis revealed a negligible effect size, d = 0.07, 90% CrI [0.01, 0.12], of the dynamic norm interventions on the preference for vegetarian meals.” (Our meta’s overall effect was also d = 0.07 across intervention categories.)

• Biggs et al. (2025) put up dynamic norm signs in UK university cafeterias, where four cafeterias got messages “incorporating a socially ‘close’ referent group and three cafeterias displayed a message incorporating a socially ‘distant’ referent group.” The study is not an RCT, but “both messages decreased odds of cafeteria diners purchasing vegetarian meals.”

• (04-21-2026 update: the next two papers came to my attention via Elif Naz Çoker) Wolfswinkel et al. (2025) run a quasi-experiment in six supermarkets in the Netherlands where treatment supermarkets got “both placement nudges (e.g. increased product salience, accessibility, or availability) as well as property nudges…[that] included textual communications of dynamic norms throughout the supermarket” to promote “meat substitutes and legumes.” The “retailer was in charge of the supermarket selection” and there were large baseline differences between treatment and control supermarkets in meat alternative consumption. No overall effect was found.

• Edwards et al. (2026) put images of plant-based items, along with a few different messages promoting them (including dynamic norms appeals), at the self-service kiosks of 22 outposts of a burger-brand quick service restaurant chain in Germany, compared to 23 restaurants in a control group. Treatment was not randomized, but rather assigned at the levels of geographic clusters as determined by the company. Dynamic norms were associated with a small but statistically significant increase of plant-based sales of ~1 pp.

• Çoker, Pechey, & Jebb (2025) put up dynamic norm messages in worksite cafeterias “for 8 weeks on free-standing banners, posters, and floor stickers.” In intervention cafeterias, vegetarian sales declined by 2.22 percentage points, but the effect wasn’t significant (95% CIs [− 7.33, 2.90], p = 0.378). The study’s real payoff is that they interviewed people afterwards to see if the messages made any impression whatsoever:

Across 10 sites, 155 cafeteria customers participated in opportunistic interviews. Most (n = 88, 57%) customers did not recall seeing the intervention materials. Of those who could recall seeing the materials (N = 67, 43%), only three people were able to recall the dynamic norm message correctly. Around a quarter (N = 19, 28.4%) of those who saw the materials recalled it contained the message “Spotted the star?” and a similar proportion (N = 18, 26.9%) recalled it contained information about vegetarian options. Only five people recalled that the message both referred to a star and vegetarian options.

To summarize: most dynamic norms interventions do not detect an effect on meat consumption. Extant effects are typically in the 1-2 percentage points range. A few have positive effects and a few have negative effects, which is what you’d expect if the true effect is a null and the outcome is noisy.

Burdens of proof: what if we’re just not doing it right?

I discussed these issues with a colleague at Reducetarian Summit last year. If memory serves, he said that it’s a lot easier to put up a dynamic norms message than to convince people that more and more people are cutting back on meat. I think this is correct, and also one of several burdens of proof that dynamic norms interventions need to meet. I personally have no idea how to meet these. But if you can manage it, I’d be very interested to read your study.

1. Are they literally visible? Of the 57% of people in Coker et al. (2025) who couldn’t remember anything meaningful about the message — how many saw it at all? Can we make the signs bigger?³

2. Are they memorable? Are they well-crafted rhetorically? Are they “Infesting in your…ears and nesting”?

3. Are they plausible? If you tell people that vegetarianism is rising, do they believe it?

4. Are they persuasive? Do people care? If vegetarianism rises from 4% to 10% over the next two decades, does that matter for what I eat? Would it matter if those 10% are especially cool or influential?

5. Are they well-targeted? Harkening back to a previous post touching on identity (and a follow-up conversation with Jennifer Channin of the Better Food Foundation) — are we talking to the people who would most be interested in hearing about changing rates of vegetarianism?

Taking dynamic norms and identity seriously

Were I a dynamic norms researcher, I’d be thinking about specific groups where plant-based eating is trending up, e.g., African Americans in major urban areas. Black veganism actually is on the rise, and we’ve got some exceptional black-owned vegan businesses to show for it: in NYC, RAS plant based, aunt et uncles, Next Stop Vegan, and Uptown Veg, for example. Maybe collaborate with someone like Jenné Clairborne or KRS-One to get folks into these great places for a conversation about food, identity, and the liberation of a plant-based diet?

I think there’s something promising here.

References

Bibliography here.

1. By the way, if financial incentives don’t work to change behavior, I’d generally ask, in the spirit of Bryan Caplan, have you tried doing ten times as much?

2. They also encountered a true wild card snafu when the “virtual restaurant” setting of study 2 “abruptly closed after 10 days of the intervention.” The experiment still worked, just with much less statistical power than they planned for.

3. Again, have you tried doing ten times as much?