A Meta-Analysis Of Convincing People To Go Vegan

Recent study shows that encouraging people to eat less meat makes them eat less meat-- with big caveats.

A very common animal-advocacy intervention is trying to convince people to eat less meat by telling them about the poor conditions that farmed animals experience. It’s difficult to know how well this works, with some people arguing that it doesn’t work at all. A recent meta-analysis and systematic review reviewed the available research on this subject (both peer-reviewed and conducted by advocacy organizations to inform their decision-making).

So what have we learned? Trying to convince people to eat less meat works—but with some very serious caveats.

The headline result is that, on average, the interventions studied increased by 22% the subject’s probability of being in a low-meat-consumption group rather than a high-meat-consumption group. Different studies tend to increase the subjects’ probability by very different amounts, probably because they’re testing a whole bunch of different interventions. You wouldn’t expect a fake news article with pictures of gestation crates to have the same effect as putting pictures of cows next to pictures of hamburgers in a cafeteria.

Almost all interventions reduce meat consumption. Nearly all studies show a definite reduction in meat consumption, and the interventions that possibly increase meat consumption do so by a very small amount that’s easily outweighed by the positive effects of other interventions. That’s good, because a lot of animal advocates are worried about the “backfire effect.” The backfire effect is when someone, say, reads an article about factory farming and eats more meat afterward—maybe it’s a really annoying article from PETA and they’re eating more meat out of spite, or maybe it reminded them of the existence of hamburgers and they think hamburgers are tasty. The backfire effect doesn’t seem to be very strong.

It’s important to note that, even though there weren’t any backfire effects in this study, there are definitely backfire effects for some interventions. If you do some incredibly tasteless intervention like comparing factory farming to the Holocaust, a bunch of people are probably going to be turned off veganism. It’s just that no one studied that. But if you do a broadly reasonable intervention, you shouldn’t be worried about backfire effects.

A related fact is that people asked to go vegan eat less meat than people asked to go vegetarian or to reduce their meat consumption. A lot of people (including me) have tended to ask people to reduce their meat consumption without necessarily going vegan, because we’re worried about backfire effects. Maybe asking people to go vegan will make them feel overwhelmed or like you’re a very unreasonable person, and then they’ll eat the same amount of meat or even more. But it seems like that’s not true, maybe because backfire effects aren’t really a thing. Instead, I think that people probably adjust their behavior based on what they’re asked to do—if they’re asked to go vegan, maybe they’ll compromise and reduce their meat consumption, but if they’re only asked to reduce their meat consumption, maybe they won’t change their behavior at all.

The studies included in the meta-analysis were mostly pretty high-quality. The effect of publication bias was low. Studies typically had randomized designs, and large sample sizes. There was a high rate of data-sharing and preregistration, especially among the studies conducted by nonprofits (the fact that nonprofits do better than actual scientists is incredibly sad). Unfortunately, there are two big limitations that limit how much we can learn.

The first is that most studies measured meat consumption by looking at intent or self-report—that is, they looked at either whether people thought that they were going to reduce their meat consumption going forward or whether they said they’d eaten meat after the intervention. Unfortunately, lots of people intend to do lots of good things, and most of them don’t end up doing them. We don’t know how many people’s intent to eat less meat lasted until they smelled the first delicious burger cooking on the grill.

The problem with self-report is more subtle. People don’t like admitting to doing bad things. If you successfully convince them that eating meat is bad, they’re less likely to eat meat—and they’re also less likely to admit to eating meat. So if there’s a difference between the treatment group and the control group, you can’t tell whether they’re actually eating less meat or they’re just doing some wishful thinking about those two or three hot dogs they had as a snack.

The second problem is followup. About half of studies measured outcomes immediately after the intervention. The rest, on average, measured outcomes after 39 days, with the longest follow-up period being 120 days. That’s not very long when we hope that people will be vegan for years or decades. Studies that followed up later also had much higher attrition rates. What’s worse, the longer the followup, the more meat the people in the treatment group eat. There aren’t very many studies, but we can guess that those relatively impressive effects disappear really quickly.

I’m not bashing the people who run the studies. Behavioral measures of meat consumption are expensive and often impossible outside of very limited contexts like college cafeterias. Following someone up months, much less years, later is expensive and makes it take much longer before your study can inform anyone’s decision-making. But we have to be aware of how little we know.

Overall, I think the meta-analysis suggests something like this. Almost no one is turned off by being told about how bad eating meat is, and many people do decide they want to eat less meat. Some of those people actually eat less meat, while others don’t but rationalize that they ate less when filling out surveys. But it’s hard to stick to any major dietary change, and within a few weeks or months they’re off the wagon for good.

Many interventions, like online ads and newspaper articles read by hundreds of thousands of people, are very cheap per view. So even if this model is true, you might end up saving a lot of animals through people’s few weeks of attempting to be vegetarian, which makes them cost-effective. On the other hand, the animal advocacy movement has spent decades trying to convince people to become vegan—and the number of vegans is basically flat. That suggests to me that, however short-term helpful trying to convince people to go vegan is, in the long term it’s not going to result in the kind of systematic change we need. Ultimately, the solution is good meat substitutes that make it easy for people to become vegetarian.