Should the animal advocacy movement change the amount of resources invested in digital media and mass media campaigns that aim to reduce people's consumption of meat and/or animal products?
Authors: Ren Ryba
Executive Summary
The animal advocacy movement regularly invests resources into campaigns that aim to reduce people's consumption of meat and/or animal products. Many of these campaigns are conducted using digital media (e.g. social media ads) and/or mass media (e.g. radio, TV, newspapers).
In this report, we summarise the evidence on these campaigns. We conduct a cost-effectiveness analysis, drawing on the high-quality scientific studies that have been published over the past few years. This report builds on previous analyses of meat-reduction campaigns, which have been published by various researchers over the past decade. Our cost-effectiveness analysis focuses on what additional funding could achieve on the margin, ignoring fixed costs that are already being paid by existing animal advocacy organisations.
We address three main questions. These questions, and our current view on them, are as follows:
What impact might these campaigns have on the lives of animals? Given the scientific evidence we have available, our best guess is that these campaigns spare 3.7 animals per dollar. There is significant variance in our estimate - it is very possible that the campaigns actually have 0 impact, and it is somewhat possible that the campaigns have an impact that is several times higher than our best-guess estimate. Our best-guess estimate is in rough agreement with the cost-effectiveness estimated by previous researchers. We erred towards the conservative side when producing our calculations, so there are several important choices we made with which reasonable researchers could disagree.
Should the animal advocacy movement change the amount of resources invested in these campaigns? Based on our calculations, this campaign does not appear supremely cost-effective. On the other hand, the absolute impact of this campaign (3.7 animals per dollar) is quite reasonable, and the evidence base supporting this campaign is quite strong.
Could mass media campaigns in developing countries be a particularly impactful opportunity for these campaigns? We believe that radio and mass media campaigns in developing countries can justifiably form part of the movement's overall portfolio of outreach campaigns for meat reduction. However, we do not think it would be justified to put a disproportionately high amount of the movement's resources into radio campaigns in developing countries.
Could we create a win-win by jointly funding meat-reduction campaigns with climate-motivated funders? We find that climate funders (at least the funders strictly interested in climate and motivated primarily by cost-effectiveness) would not be keen to put large amounts of resources towards meat-reduction campaigns, even as a joint initiative. The reason is that meat-reduction campaigns appear to be a less cost-effective way to reduce greenhouse gas emissions than other options available to the climate movement.
Our conclusions are based on our cost-effectiveness analysis in which we calculated the expected number of fewer animals farmed per dollar spent on advertising (animals/USD). We calculated this for each of four different scenarios. We also produce rough upper and lower bounds to accompany our point estimates. The results are as follows:
Best-guess. This is our main scenario, and we think this scenario is the most likely of our four scenarios (~45%). Under this scenario, we estimate that meat-reduction outreach campaigns spare 3.7 animals/USD (-0.7 to 8.7).
Optimistic (but still plausible). This scenario optimistically assumes that meat-reduction advertisements are slightly more frequent (than in our best-guess estimate) at convincing people to reduce their meat intake. We think this scenario is plausible, but somewhat unlikely (~10%). Under this scenario, we estimate that meat-reduction outreach campaigns spare 9.1 animals/USD (5.3 to 13.6).
Optimistic + cheap + long-lasting (but still plausible). This scenario assumes a longer duration of diet change and a cheaper cost-per-engagement. We think this scenario is plausible, but unlikely (~5%). Under this scenario, we estimate that meat-reduction outreach campaigns spare 84.8 animals/USD (50.1 to 127.3).
Pessimistic (but still plausible). This scenario simply assumes that the effect of the intervention is zero. As Mathur et al (1) argue, many of the diet-change experiments are subject to various weaknesses (e.g. desirability biases). This means that a true effect of zero remains a real possibility (2). We think this scenario is somewhat likely (~40%). Under this scenario, we estimate that meat-reduction outreach campaigns spare 0 animals/USD.
Overall, our best guess is that meat-reduction outreach campaigns spare roughly 3.7 animals/USD. But there is a really good chance (our gut feel says a ~40% chance) that these campaigns have 0 impact, and there is a small chance (our gut feel says a ~15% chance) that the impact is several times larger than our best-guess. About 90% of this benefit comes from chickens and farmed fish. If you assume a typical lifespan of ~6 weeks for chickens and perhaps a bit longer for farmed fish, our best-guess estimate would correspond to roughly ~5 animal-months spared per dollar.
Our best-guess estimate suggests that meat reduction campaigns are somewhat less effective (~one-sixth as effective) as what previous cost-effectiveness analyses have concluded. Our analysis draws on a stronger evidence base than some previous analyses. We also make some assumptions and choices that tend to err on the conservative side, so reasonable people could conclude that the point estimate should be higher than we have concluded. We emphasise that there is plenty of variance across our four scenarios that capture the possibility for a more optimistic view.
Based on this analysis, we do not think that there is much reason for the animal advocacy movement to invest a substantially higher amount of resources into these campaigns.
The movement will probably continue funding these campaigns at roughly the current level of investment for the foreseeable future. So, it is probably worthwhile for the movement to fund a few more studies following the recommendations made by Mathur et al in the conclusion to their recent experimental report (2). In that report, Mathur et al found that the existing studies may be systematically overestimating the effects of meat-reduction campaigns, and they recommended that experimental studies be conducted in a way that measures actual consumption with a sufficient sample size while minimising social desirability bias.
On the other hand, we have mainly discussed the evidence for one-off interventions. In these interventions, participants see a message or advertisement only one time. However, it is also possible for campaigns to use mass media with the intention of showing the same message to each person multiple times. Repeat exposure might have a larger effect size than one-off exposure.
Table of Contents
1. Introduction
The animal advocacy movement regularly invests resources into campaigns that aim to reduce people's consumption of meat and/or animal products. Many of these campaigns are conducted using digital media (e.g. social media ads) and/or mass media (e.g. radio, TV, newspapers).
In this report, we summarise the evidence on these campaigns. We ask the following questions:
What impact might these campaigns have on the lives of animals, given the scientific evidence that we have available?
Should the animal advocacy movement substantially change the amount of resources invested in these campaigns?
Could mass media campaigns in developing countries be a particularly impactful opportunity for these campaigns?
Could we create a win-win by jointly funding meat-reduction campaigns with environmentally motivated funders?
In this report, we use "meat-reduction outreach" and "outreach for meat reduction" as umbrella terms to capture any campaign that seeks to reduce or eliminate people's consumption of meat and/or animal products. These campaigns could ask people to reduce their intake of meat and/or animal products (reducetarian), to eliminate meat entirely (vegetarian), and/or to reduce meat and animal products entirely (vegan). For simplicity, we often use the term "meat" as shorthand for "meat and animal products".
The diagram below shows the theory of change for the main causal pathway of these campaigns. The animal advocacy movement creates and publishes meat-reduction advertisements on digital or mass media platforms. These advertisements attract engagement by meat-eaters, which (in principle) causes an increase in the number of meat-eaters who reduce or eliminate their consumption of meat and animal products. This causes the demand of meat and animal products to decrease, which (in principle) causes the supply to decrease too. As a result, fewer animals are farmed.
Note that these campaigns also have many indirect effects, which are detailed later in the report.
2. What is the effect of eating less meat?
Eating less meat has two types of effects: animals directly helped, and indirect effects.
Direct effects: How many animals are directly saved by adopting a vegan or vegetarian diet?
Using a simple calculation based on global production data, Animal Charity Evaluators (3) estimate that "about 105 vertebrates are spared per plant-based person per year, of which approximately 79 are wild-caught fishes, 14 are farmed fishes, and 12 are farmed land vertebrates (11.5 farmed birds and 0.5 farmed mammals)."
Using a simple calculation based on US consumption data, Hurford estimates that each year, one meat-eater consumes "0.09 cows for beef; 0.004 cows for dairy; 0.24 pigs; 13.2 chickens for meat; 0.95 chickens for egg; 0.6 turkeys; 1.3 aquacultured fish". Hurford subsequently accounts for elasticity, which we discuss separately below.
Taking these two sets of estimates as lower and upper bounds, it is reasonable to expect that adopting a vegan or vegetarian diet would prevent the farming of 16 and 26 farmed animals each year. By animal group, this corresponds to between 11.5 and 14.8 farmed birds; between 1.3 and 14 farmed fish; between 0.3 and 0.5 farmed mammals. We have focused on farmed animals in this estimate, ignoring wild-caught fish (due to the uncertainties around wild animals mentioned below).
One way to estimate the effects of reducing (rather than eliminating) meat intake is to make the following assumption: reducing meat intake by some proportion spares a number of animals equal to the number of animals who would be spared by adopting a vegan diet, multiplied by that proportion. This assumption means that if 10 people each reduce their consumption of meat and animal products by 10%, the effect is the same as if 1 person adopted a completely vegan diet. We adopt this assumption when conducting the cost-effectiveness analysis later in this report.
Indirect effects: What other effects does eating less meat have on animals?
Elasticity: economic dynamics cause the magnitude of the impact of eating less meat to be decreased by a modest proportion. The concern with elasticity is that if there are more vegans, then meat could become cheaper to buy, which could encourage other people to buy more meat, thus potentially cancelling out any benefits. The best exploration of this question that we have found was published by Lusk and Bailey Norwood (4) and summarised by Tomasik (5) and by Barrett and Raskoff (6). As Barrett and Raskoff express these results: "in the United States, for a sufficiently large number n, if n choices are made not to buy a chicken, 0.76n fewer chickens will be produced (6)." The corresponding numbers for other products are 0.91 for eggs, 0.74 for pork, 0.69 for veal, 0.68 for beef, and 0.56 for milk (5). As such, these economic effects might slightly reduce the impact of eating vegan, but only by a small amount (e.g. 24% for eating less chicken).
Threshold effects: we believe that threshold effects have a negligible effect on impact. The concern with threshold effects is that, sure, one person can go vegan, but will that one person be the one to cause the supermarket to purchase less meat from their supplier, and for farms to produce fewer animals (7)? Just how sensitive are supply chains? While there is some debate about threshold effects and the expected impact of veganism (7), such debates are less important for our case. In the present report, we are not thinking about one individual going vegan, but rather convincing many people to go vegan at the same time. What matters to us is not the effect of one additional vegan, but the effect of many additional vegans. Barrett and Raskoff (6) state that the argument "that vegans have a huge collective impact on reducing animal suffering" is "uncontroversial", and that "At least over the long run, the animal agriculture industry is sensitive to macrolevel market trends—for example, to large numbers of individuals refraining from chicken."
3. How effective are digital outreach campaigns at encouraging people to reduce meat consumption?
3.1 Key studies on effectiveness
Over the past couple of decades, there have been many experimental studies that aim to measure whether outreach campaigns cause people to reduce their intake of meat and animal products. These studies have also been examined in a couple of systematic reviews and meta-analyses.
The key study that guides our understanding is the systematic review by Mathur et al (1). The authors conducted a systematic review and meta-analysis of 100 studies that tested the effects of messages on meat consumption. This review did not focus specifically on digital advertising, and messages focused on animal welfare rather than environmental or health concerns, but the immense volume of data makes it a high-quality piece of evidence nevertheless. The authors found that "the interventions appeared consistently beneficial at least in the short term and with outcomes primarily based on self-reported behaviour or intended behaviour, on average increasing by 22% an individual’s probability of intending, self-reporting, or behaviorally demonstrating low versus high meat consumption." This estimate was lower when the authors included only studies with longer follow-up times (e.g. 7+ days) and/or when the authors included only studies that measured self-reported or observed consumption, rather than intended consumption. When both of these limitations are applied, the corresponding estimate is about 9% (i.e., outreach increases by 9% an individual’s probability of intending, self-reporting, or behaviorally demonstrating low versus high meat consumption). In our model (see below, "Our cost-effectiveness analysis"), this corresponds to the intervention causing around 4.5% of people to decrease their meat consumption. The danger of backfiring (people eating more meat after seeing messages) appeared rare. The authors point out that the studies in the review often used self-reported behavioural measures, which can lead to desirability bias, and that most studies only looked at meat consumption in the short-term (a month or less).
To complement our understanding, we also draw on the systematic review by Chang et al (8). This was a systematic review and meta-analysis of 31 studies on reducing meat consumption, specifically in university/college settings. The estimate from this meta-analysis is closely consistent with the estimate from Mathur et al (see below, "Reality checks"). The authors did not find significant differences in the effects of intervention in meat consumption between studies that used long vs. short follow-up times.
One of the key findings from Mathur et al's systematic review was that many experimental studies were at risk of desirability bias. This may be causing us to think that the ability for outreach to reduce people's meat intake (e.g. the 4.5% estimate we mentioned above) is larger than it actually is - we may be overestimating the effect size. So, Mathur et al (2) subsequently conducted an experimental study that explicitly aimed to understand this desirability bias. This study focused on documentaries, which may be less applicable for digital media campaigns. However, this study is particularly informative because the authors explicitly conducted different trials to see whether results differed when the experimental setup minimised desirability bias. The conclusion is particularly illuminating: "In a study designed to minimize social desirability bias [...] we found that the documentary did not affect participants’ reported consumption 12 days after random assignment [...] However, in a second study with less protection against social desirability bias [...] the documentary did substantially increase the percentage of participants who immediately intended to reduce consumption." This provides evidence that the academic literature may be overestimating the number of people who eat less meat after being exposed to information.
To summarise, there is strong evidence supporting the hypothesis that digital media campaigns may lead to people eating less meat. However, due to methodological limitations, such as social desirability bias, such campaigns may be somewhat less effective than previous research studies would suggest.
In terms of the cost for running online meat reduction campaigns, the most relevant piece of information is a recent experiment of animal rights video advertisements on Facebook (9). In a group of developed countries (Australia, Canada, the UK, New Zealand, and the USA), the costs per engagement were as follows: USD 0.0009 per impression (when the advertisement was shown to a Facebook user); USD 0.0029 per video play (when users played the video for at least 15 seconds); USD 0.0096 per video play (when users played the video for 50% of its duration, i.e. around 50 seconds); USD 0.03 per video play (95% of its duration, i.e. around 1 minute 30 seconds); and USD 0.56 per reaction (e.g. when the user makes a like, love, or sad reaction to the video on Facebook). In a group of developing countries (Angola, Ethiopia, Lesotho, Nigeria, Rwanda), the corresponding costs per engagement for each of those same categories were as follows: USD 0.00011, 0.00032, 0.0012, 0.0041, and 0.031 respectively. These numbers are roughly consistent with the costs per engagement described by Winchell for digital marketing in the broader context of effective altruism in general (10).
3.2 Our cost-effectiveness analysis
We conducted a cost-effectiveness analysis in which we estimate the number of animals helped per dollar by outreach for meat reduction, given the available scientific evidence. Our analysis is publicly available as a spreadsheet here, and is summarised in the table below.
Since we're highly uncertain about some parameters, we consider four scenarios:
Best-guess. This is our main scenario, and we think this scenario is the most likely of our four scenarios (~45%). This scenario assumes a risk ratio of 1.11, calculated by Mathur et al (1) using only studies that measure observed or self-reported consumption (rather than intended consumption). This scenario assumes a diet change of 1 year and a cost-per-engagement of USD 0.03.
Optimistic (but still plausible). This assumes a higher risk ratio of 1.22, calculated by Mathur et al (1) using all studies in their meta-analysis (even those that measure only intended consumption). We think this scenario is plausible, but somewhat unlikely (~10%).
Optimistic + cheap + long-lasting (but still plausible). This assumes a longer duration of diet change (3 years rather than 1 year) and a cheaper cost-per-engagement (USD 0.0096 rather than USD 0.03). We think this scenario is plausible, but unlikely (~5%).
Pessimistic (but still plausible). This scenario simply assumes that the effect of the intervention is zero. As Mathur et al (1) argue, many of the diet-change experiments are subject to various weaknesses (e.g. desirability biases). This means that a true effect of zero remains a real possibility (2). We think this scenario is somewhat likely (~40%).
Our model uses these key inputs (parameters):
Effect size for the relative increase in probability of reducing MAP (meat and animal product) consumption after intervention, relative to baseline (risk ratio). We use a risk ratio of 1.11, calculated by Mathur et al (1) using only studies that measure observed or self-reported consumption (rather than intended consumption). A risk ratio of 1.11 means that people who are exposed to the intervention, compared to people not exposed to the intervention, have an 11% higher probability of self-reporting or behaviourally demonstrating a lower meat consumption. In two scenarios, we increase the risk ratio to 1.22, which was calculated by Mathur et al (1) using all studies in their meta-analysis (even those that measure only intended consumption).
Adjustment for long time frames (7+ days) (effect modification risk ratio). Mathur et al calculated that a study’s having, versus not having, a follow-up length of at least 7 days is associated with an effect modification risk ratio of 0.81. In the best-guess scenario, this reduces our initial risk ratio from 1.11 to ~1.09. We ignore this in two scenarios.
Probability of net decrease in MAP (meat and animal product) consumption, for people not exposed to the intervention (percent). We assume that this parameter is 50%.
We need to include this parameter because we have a risk ratio, and we want to use our risk ratio to calculate the probability that the intervention will cause somebody to reduce their meat intake. To perform this calculation, we first need to make an assumption about the probability that somebody not exposed to the intervention will reduce their meat consumption. This probability equals the chance that if you ask somebody their meat consumption at two points in time, then their meat consumption will be lower at the second of the two points in time.
There are strong theoretical reasons to expect this number to be 50% - if you measure meat consumption as a point on a quantitative scale (i.e. you ignore the probability that the meat consumption of a meat-eater will be exactly identical at two points in time) and assume that there are no enormous trends in meat consumption by the population as a whole, then we would expect the baseline population (people not exposed to the intervention) to be divided into ~50% meat increases and ~50% decreases by chance alone. This is supported by our analysis of data from dietary change studies for people in the control groups: in the data from Schwitzgebel et al (11), the percentage of people in the control group who decreased their meat consumption was 49%; for Jalil et al (12), 45%; and for Macdonald et al (13), 49%. These are consistent with our theoretical estimate of 50%, subject to random sampling variation.
For reference, this baseline probability of 50% combined with our adjusted risk ratio of ~1.09 means that the intervention would have a 0.5 * 0.9 = ~4.5% chance of causing a person to reduce their meat consumption.
Mean magnitude of decrease in MAP (meat and animal product) consumption across all people who have a net decrease in MAP consumption after intervention (percent). If somebody indeed reduces their meat consumption due to the intervention, by what magnitude would we expect them to reduce their meat intake? This is a very specific parameter and has not been measured deliberately in previous studies. However, our analysis of data from Jalil et al (12) resulted in an estimate of ~16% for this parameter. A higher estimate is also plausible (see below, "Reality checks").
Duration of diet change (years). In our best-guess scenario, we assume that people change their diet for a mean of 1 year. We increase this to 3 years in one scenario. This choice is discussed further below ("Comparison to previous cost-effectiveness estimates").
Average number of farmed animals consumed by a non-vegetarian (animals). We assume that a non-vegetarian consumes, on average and each year, 12.4 farmed birds, 7.2 farmed fish, and 1.4 farmed mammals (3,14).
Cumulative elasticities. Due to economic effects, an X-unit decrease in demand for a particular animal product will reduce production by a bit less than X units (4). We assume that an X-unit decrease in demand for chicken will reduce production by 0.76X units. Our corresponding values for farmed fish and farmed mammals are 0.69X and 0.74X, respectively (4,5,15).
Cost per engagement (USD). We assume a cost per engagement of USD 0.03, based on the Facebook experiment described above (9). We decrease this to USD 0.0096 in one scenario. Notably, this is quite a bit lower than what previous analyses have assumed for the cost per click on an online ad (see below, "Comparison to previous cost-effectiveness estimates"). We also emphasise that this is the marginal cost per engagement - this excludes, for example, the costs to run an organisation and hire the staff who run the ads. The reason we opted to use the marginal cost per engagement is that there are many existing animal advocacy organisations with staff already being paid and online advertising campaigns already running. For these organisations, extra funding could be directed almost entirely to the marginal cost of additional ads.
As our key output (result), we calculated the expected number of fewer animals farmed, per dollar spent on advertising (animals/USD). We calculated this for each of our four scenarios, and we derived rough upper and lower bounds by using the upper and lower bounds of the risk ratio estimates. The results are as follows:
Best-guess. 3.7 animals/USD (-0.7 to 8.7).
Optimistic (but still plausible). 9.1 animals/USD (5.3 to 13.6).
Optimistic + cheap + long-lasting (but still plausible). 84.8 animals/USD (50.1 to 127.3).
Pessimistic (but still plausible). 0 animals/USD.
So, our current view of this intervention's effectiveness is something like this:
Our best guess is that we could prevent the existence of around 3.7 farmed animals for each USD spent on this intervention (scenario 1). Most of these animals would be farmed chickens or farmed fish, with the occasional farmed mammal.
There is a real chance that the intervention actually has 0 effectiveness (scenario 4).
There is also a decent chance that the intervention's effectiveness could be a bit higher than 0.5 animals/USD, but probably not much higher than ~20x that estimate (scenario 2-3).
We can also express this same view visually. The graph below shows one probability density curve that captures our current view of the true cost-effectiveness of this intervention. The horizontal axis shows the estimated cost-effectiveness (number of fewer animals farmed, per dollar spent on advertising). The vertical axis shows a rough, gut-feel probability for that estimated cost-effectiveness.
Most of the probability density (~45%) is centred on 3.7 animals/USD, plus or minus some variance.
There is also a large amount of probability density (~40%) centred on 0 animals/USD, representing the chance that the true cost-effectiveness is zero.
There is a small chance that the cost-effectiveness could be much higher than 3.7 animals/USD, but we think this is less likely.
There is a small chance that the intervention could be overall harmful (i.e. true cost-effectiveness is less than zero), but we think this is unlikely too. As Mathur et al write: "Although it seemed plausible [...] that in some settings, animal welfare appeals could be in danger of backfiring [...], our results suggested that this danger was rarely realised, as we estimated that the large majority of interventions (83%) had true population effects in the beneficial rather than detrimental direction."
3.3 Other important effects that are not captured by our calculations
Our cost-effectiveness analysis ignores some factors that are difficult to measure, but important to consider:
Dynamic engagement resulting from paid engagement. For example, it is possible that a paid advertisement will be shared by some social media users, magnifying the reach of the advertisement. Likewise, users who see the advertisement may follow the social media page of the organisation that was running the advertisement, leading to additional engagement in the future. This strategy is used by the advocacy organisation Kinder World.
Wild-caught fish and crustaceans. We did not include the effects of reducing meat intake on wild-caught fish and crustaceans. The economic and moral effects of reducing wild-caught fish intake are poorly understood (16).
Animals used for feed, etc. We did not include the effects of animals who are made to suffer and killed at earlier stages during the production of meat and animal products. For example, animals are killed to produce feed for farmed animals.
Room for more funding. We did not consider whether there are still markets that have room for additional outreach for meat reduction, or whether the movement has already saturated all possible markets.
Effects on friends and family. People who eat less meat may encourage their friends, families and colleagues to also eat less meat (17). This may be particularly relevant for people who cook for others (e.g. parents of young children).
Effects on policy support. People who eat less meat might (or might not) show greater support for further animal advocacy campaigns in the future, such as specific policy changes in their country (18). Likewise, people who become completely vegan or vegetarian might also become committed animal welfare advocates, meaning that there are more people working on reforming policy for animals in the future.
Systematically different effects on different demographics of people. We assume that the risk ratios provided by Mathur et al are roughly representative of the target audience of real-world meat-reduction campaigns, and we treat "people" as a single category rather than distinguishing between different demographics. However, there may be systematic differences in how different groups of people respond to meat-reduction campaigns. We might expect that people who are already seeking to reduce their meat consumption would be more likely to engage with meat-reduction advertisements. On the other hand, we might expect advertisements to cause a greater total reduction in individual meat consumption for people who are not already seeking to reduce their meat consumption.
Systematically different effects in different contexts. We use a cost-per-engagement comparable to a typical online ad. But we also use the risk ratios calculated by Mathur et al from a variety of different types of campaigns (e.g. online ads, hour-long lectures, full video documentaries). If lectures and documentaries are more impactful than online ads, then our cost-effectiveness estimate may be too optimistic.
Wild animals. If there are fewer animals farmed, then there might be increases and/or decreases in the amount of land used for crop farming and/or the amount of land left "wild". This could have enormous effects on the lives of wild insects and other wild animals (19), and there has not been much research into this question (20). Tentatively, researchers hold widely varying opinions as to whether these effects would be good or bad, important or unimportant, and so on.
Moral circle expansion. People who eat less meat might care more about animals. It is unclear whether eating less meat would cause people to care more about animals (21) and, if so, whether this "moral circle expansion" would be good or bad (22).
Effects on climate change and human health. Eating less meat would probably reduce greenhouse gas emissions (12). Eating less meat may also, in some contexts, improve human health (8). Most of humanity would probably regard both of these as good, though there is debate about the net effects of these on the global sum of moral value (23,24).
3.4 Reality checks
Where possible, we did our best to double-check our inputs and outputs against other sources of information. These checks aren't really independent, as there are overlaps between the evidence we used to build our model and the evidence we use to conduct these reality checks. However, they can help let us know if our model suffers from any glaring, obvious errors.
These reality checks are given in detail in the "Reality checks" section of the model spreadsheet. We give a brief summary here. For these reality checks, we only refer to the point estimates of our best-guess model scenario.
Effect size for the relative increase in probability of reducing MAP consumption after intervention, relative to baseline (input). We use a risk ratio of 1.11, based on the findings from the meta-analysis of Mathur et al (1). An alternative meta-analysis, that of Chang et al (8), found a corresponding odds ratio of 1.68. This odds ratio corresponds to a risk ratio of around 1.18. This is similar to our risk ratio of 1.11, especially since we adjusted our risk ratio downwards slightly based on additional findings in the meta-analysis of Mathur et al.
Mean magnitude of decrease in MAP consumption across all people who have a net decrease in MAP consumption after intervention (input). We use an estimate of 16%, based on our analysis of the data from the experimental study by Jalil et al (12) (one of the highest-quality experimental studies available to us). We produced two alternative estimates, based instead on our analysis of the data from two other experimental studies. The data from Macdonald et al (13) would suggest an estimate of 49%, and the data from Schwitzgebel et al (11) would suggest 52%. Our estimate is certainly on the conservative side, though all three estimates are within the same order of magnitude. There is no real reason to favour one estimate over the others, aside from intuitions about the quality and generalisability of each of these three experimental studies and the real-world implications of each estimate. In real-world terms, for somebody who eats meat two times every day, a reduction of 16% would mean eating meat at around two or three fewer meals per week. Assuming an average decrease larger than this strikes us as intuitively unrealistic.
Expected decrease in MAP consumption caused by the intervention, for any person who experiences the intervention (output). Our model produced an estimate of 1.11%. We produced two alternative estimates, based on our analysis of data from Jalil et al (12) and Schwitzgebel et al (11). The corresponding estimates would be ~9% and ~23%, respectively. This would suggest that we could be underestimating the true effect size by a factor of ~10x to ~20x. Nevertheless, there are many reasons to expect that the true effect would be smaller than what these experimental studies predict (e.g. our intervention of online ads or perhaps short videos being far less intensive than hour-long lectures; many reasons for caution that were expressed in Mathur et al 2021's meta-analysis and adopted in our model). So, we are comfortable with our more modest estimate.
Sum of expected number of fewer farmed animals consumed by one person for the entire duration of the effect, adjusted for elasticity (output). Our model produced an estimate of 0.172. We produced one alternative estimate, based on our analysis of data from Amiot et al (25). The corresponding estimate was between ~1.0 and ~100, depending on some assumptions (and probably somewhere in between those two extremes). This would suggest that we could be underestimating the true effect size by a factor of ~30x to ~300x. As in reality check # 3 above, we are comfortable with our more modest estimate.
3.5 Comparison to previous cost-effectiveness estimates
Where possible, we did our best to double-check our inputs and outputs against previous estimates of the cost-effectiveness of outreach on meat consumption. Again, these are not really independent, as many of these analyses were based on the results of studies that also form part of our evidence base. However, this can help us see whether our cost-effectiveness estimate is within the same ballpark as what has been calculated previously, or whether we came to a different conclusion.
We find agreement between our estimate and previous estimates for the key output: the overall estimate of the number of animals helped per dollar spent on the intervention. Our overall point estimate of ~3.7 animals helped per dollar is very similar to the magnitude of estimates that have been produced previously. The corresponding estimate from Animal Charity Evaluators (26) was between 3 and 10 animals per dollar; from Animal Charity Evaluators (27) was 3.4 animals per dollar; and from Wildeford (28) was 3.1 animals per dollar. Likewise, our estimate corresponds to ~166 animal-days per dollar, compared to the ~120 animal-days per dollar calculated by Tomasik (29). This means that we think that outreach for meat reduction is roughly as good as people who have analysed this campaign previously do.
We find disagreements between our estimate and previous estimates for three inputs:
Cost per engagement. The only estimate in the table below that corresponds to online advertisements comes from the study by ACE (~2016-2019) (27). The original source for the cost-per-engagement in that study was ACE's evaluation of The Humane League's (THL) online advertisements for diet change, factory farming, and other topics. The spreadsheet supporting this analysis is available here, though the spreadsheet has since been updated for the 2023 analysis. In that spreadsheet, ACE calculates that the cost per person reached by THL's online ads are, in fact, either USD 0.007 or USD 0.345. The costs vary between the THL's campaigns. The cost that we use in our model (USD 0.03, based on the Facebook experiment detailed above) is well within this range. Therefore, we are comfortable with using our estimate of USD 0.03, even despite the different estimates produced by ACE.
Duration of diet change. We use 1 year (increased to 3 years in one of our modelled scenarios). This parameter is subject to a large amount of uncertainty. The evidence for this input is pretty sparse - some people think this input should be lower (e.g. a few months), and some higher (e.g. 3-4 years or more). We guess ~1 year, and we think anything lower than this is probably too pessimistic. We suspect that effects could even last ~3 years, but the evidence is too tentative to have a super high amount of confidence in this. Jalil et al (12) show that the effects can persist even after 3 years, though that study involved a long lecture rather than a short video. The meta-analysis by Mathur et al (1) (see Figure S5 in that study) doesn't really show a decline in the effect over time, at least beyond the first 7 days. That said, the studies analysed by Mathur et al have, at most, follow-up times of a few months. Similarly, the meta-analysis by Chang et al (8) found no statistically significant effect of the follow-up time on the effect size. When thinking about vegetarianism specifically, Animal Charity Evaluators (30) concluded "that the authors' "final 90% [subjective confidence interval] of the mean length of vegetarianism period for use in cost-effectiveness estimates of veg*n advocacy is 2.5–6.5 years." The weakness of this estimate is that vegetarianism is different (and involves more of a conscious and public commitment) than simply reducing meat intake.
Percent of people who change their behaviour in response to the intervention. In our model, we find that the intervention causes about ~4.5% of people who engage with the intervention (e.g. people who click on the ad) to change their behaviour. In contrast, Kaufman (31) argued that expecting even ~1% of people to change their behaviour because of the intervention is unrealistic. Our figure of ~4.5% is derived from applying our risk ratio of 1.09 to a baseline probability of 50% (this is explained in detail above). The divergence in opinion here could arise from the fact that Kaufman was referring to ~1% of people who are handed a leaflet, while we're talking about ~4.5% of people who click on an ad. On the other hand, online ads could be less impactful than average when compared to all of the different interventions in the meta-analysis, many of which were quite involved (e.g. university lectures).
Chance of the intervention causing an increase, rather than decrease, in the consumption of meat and animal products. Animal Charity Evaluators (26) argued that advocating for people to reduce their meat intake has a decent chance of increasing chicken consumption - specifically, they wrote that "the estimated standardized mean difference from the meta-analysis was symmetrically distributed with most of its probability mass on leaflets causing an increase in poultry consumption." We believe that this position is no longer justified. For example, the experimental study by Jalil et al (12) found that when people ate less beef, they "switched from beef towards veg, rather than to poultry + fish". Similarly, the meta-analysis by Mathur et al (1) found that mean increases in meat consumption were quite rare, though this could mask instances where poultry increases but total meat consumption still decreases.
The rest of our comparisons are summarised in the table below.
4. Case study: The Human League
A leading organisation that conducts vegan/vegetarian outreach through digital media campaigns is The Humane League (THL).
This digital media campaign is only one campaign among THL's many. However, Animal Charity Evaluators (35) has conducted a detailed review on this campaign in particular:
"THL’s veg advocacy program focuses on encouraging individuals to reduce their consumption of animal products. They are primarily focused on using online ads to engage individuals who may be responsive to animal advocacy messaging and then provide them with animal advocacy literature. [...] After accounting for their budget, they do appear to have achieved a lot of reach with this program." (35)
Over an 18-month period, THL spent just over $1 million USD on its digital vegan/vegetarian outreach campaign. During this period, the digital media campaign reached over 30 million people on social media and resulted in 1.3 million downloads of vegan/vegetarian literature (35). This would indicate that reaching one person costs about 3 cents, or that getting one person to download literature costs about 80 cents.
The campaign also achieved some broader accomplishments, though they are less relevant for our purposes. We have not corrected for these other accomplishments in the quick description of their costs and reach above, which means that this analysis is actually slightly biassed against THL.
THL is a large organisation and is a well-established part of the animal advocacy movement. We chose THL for this case study because of their status as a Top Charity as judged by Animal Charity Evaluators. For these reasons, we would expect that a new animal advocacy organisation established to work on digital media campaigns may experience less reach than THL does currently, particularly as the new organisation scales. Nevertheless, if we assume that reaching people on social media does cause people to eat less meat, then the case of THL shows that investing in digital media may be a cheap way to have a large impact on the lives of animals.
5. The specific case of mass media campaigns in developing countries
One specific way to deliver meat-reduction outreach campaigns is to use mass media (including radio, TV, newspapers, and the internet) in developing countries. Mass media campaigns have been in a variety of contexts to achieve human behavioural change (36–43). These contexts include campaigns to prevent smoking, to encourage contraceptive use, to encourage healthier eating, among other uses. Such a campaign might be most promising in developing countries, as radio appears to be a particularly cost-effective delivery method for such a campaign (42). This raises the question: could radio and other forms of mass media in developing countries be a particularly cost-effective way to deliver outreach campaigns for meat reduction?
There is little evidence concerning the effectiveness of using radio and other forms of mass media for the specific task of reducing meat and animal production consumption in developing countries. The best evidence comes from the experiment, which we discussed above, that measured the cost per engagement for vegan advocacy videos in developed and developing countries (9). In the developing countries targeted in the experiment (Angola, Ethiopia, Lesotho, Nigeria, Rwanda), advertisements were about 7 - 9 times as effective as in developed countries. However, the author also points out that meat consumption per capita in the developing countries is around one-eighth that in the developed countries.
There is also evidence on two closely related questions.
Mass media campaigns can increase people's consumption of healthy food (36,37), including in developing countries (38). Numerous review studies have also concluded that mass media seems generally effective in changing people's health-related behaviours (36–43). This effect holds across a variety of contexts, including in numerous developing countries (41). Some studies have also found that this effect holds for mass media campaigns that seek to increase people's intake of fruits and vegetables (36–38). Notably, there is some suggestion that the influence of mass media has begun to decrease, as mobile phones overtake mass media as a key source of information in people's lives (41) - if true, this would also mean that the cost-effectiveness of mass media campaigns has also begun to decrease.
Campaigns in general can reduce people's consumption of animal products, including in developing countries. The meta-analysis by Mathur et al (1) included a handful of studies from developing countries (specifically China, Ecuador, and India). These studies mostly had positive effect sizes of a comparable magnitude to the other studies included in the meta-analysis. These studies used a few different interventions, which were pretty close in form and content to short ads shown on the internet or TV (e.g. images or short videos). None of these studies tested interventions delivered through radio specifically. Separately, a population-level modelling study found that vegetarian and vegan outreach campaigns in general can increase population-level rates of vegetarianism and veganism (44). However, this study focused on the UK, so it is unclear whether this finding would also apply to developing countries. And this study did not consider campaigns in general, but campaigns delivered via mass media specifically.
However, we do not have convincing evidence for whether a mass media campaign could successfully reduce the consumption of meat and animal products in developing countries. There may be barriers that make mass media campaigns less effective at the specific task of reducing meat and animal production consumption in developing countries, such as cultural or social factors. One important consideration is much of the existing literature on mass media campaigns cover campaigns that are clearly in the interest of the people who the outreach is aimed at. Meat-reduction outreach may have some health benefits for the people receiving the outreach, but the improvement is less clear than in some other cases, and the animal welfare and environmental benefits of eating less meat are not themselves realised by the people whose diet changes. This might mean that the evidence on mass media outreach for health-related behaviour change is not applicable to meat-reduction advocacy campaigns.
This evidence gap also means that we cannot reliably predict whether, or by how much, the cost-effectiveness of meat-reduction outreach campaigns would change relative to the baseline cost-effectiveness for meat-reduction outreach that we estimated in this report.
Furthermore, developing countries tend to have a) a lower consumption of meat per capita, and b) a lower level of intensification in meat production (45,46). This means that, all else being equal, we would expect developing countries to cause less suffering to farmed animals than developed countries in the first place.
We believe that radio and mass media campaigns in developing countries can justifiably form part of the movement's overall portfolio of outreach campaigns for meat reduction. However, we do not think it would be justified to put a disproportionately high amount of the movement's resources into radio campaigns in developing countries.
6. Teaming up with the climate movement?
Could the animal advocacy movement produce a win-win by teaming up with the climate movement to jointly fund meat reduction campaigns? Jalil et al (12) suggest, from a climate perspective, that reducing meat consumption is one way to reduce greenhouse gas emissions. This could mean that funders from the animal advocacy movement and the climate movement could jointly fund meat reduction campaigns, and such funding would have two effects: the funding would spare animals, and the funding would reduce greenhouse gas emissions. This would mean that jointly funded campaigns would, for each movement, appear more effective than campaigns that are only funded by one movement.
However, meat reduction probably won't look attractive to environmental funders who have a focus on cost-effectiveness:
Jalil et al estimate that switching from a poultry main course to a plant-based main course (each course being 5 oz or ~140 grams) would save ~1.82 kg CO2eq (carbon dioxide equivalent). Let's assume that $1 spares 0.5 of an animal (our best-guess estimate); for simplicity, the benefits are realised entirely by chickens; and that one chicken provides about 500 grams of cooked meat.
This would suggest that $1 prevents the farming and preparation of the equivalent of 250 grams of cooked chicken, which corresponds to about 1.8 of the 5 oz servings used in the calculations of Jalil et al. Thus, $1 would save around ~1.82 * 1.8 = ~3 kg CO2eq. Another way of saying this is that it would cost around ~$300 to save one tonne of CO2eq.
This is a quick back-of-the-envelope calculation. In practice, our model suggests that around two-thirds of the benefit would be realised by chickens and one-third by farmed fish. The greenhouse gas emissions from chicken farming and fish farming are very similar (12), though the amount of meat per animal may vary. Therefore, our estimate of greenhouse gas emission prevented per dollar may change slightly.
Giving Green writes: "we consider something to plausibly be within the range of cost-effectiveness we would consider for a top recommendation if its estimated cost-effectiveness is within an order of magnitude of $1/tCO2eq (i.e., less than $10/tCO2eq)."
Therefore, for climate funders who use a cost-effectiveness framework, meat reduction campaigns probably won't look attractive. This means that neither animal advocacy funders nor climate funders would be keen to put large amounts of resources towards meat-reduction campaigns, even as a joint initiative.
There may be climate funders who think differently to Giving Green and are therefore willing to spend money on campaigns that reduce greenhouse gas emissions for ~$300/tCO2eq. If such funders exist, then it could be worth teaming up with them to fund meat-reduction campaigns.
7. Conclusion
With our overall views on meat reduction campaigns in mind, we can now answer the questions we posed in the introduction to this report:
What impact might these campaigns have on the lives of animals? Given the scientific evidence we have available, our best guess is that these campaigns spare 3.7 animals per dollar. There is significant variance in our estimate - it is very possible that the campaigns actually have 0 impact, and it is somewhat possible that the campaigns have an impact that is several times higher than our best-guess estimate. Our best-guess estimate would indicate that these campaigns are roughly as impactful as previous researchers have concluded. We erred towards the conservative side when producing our calculations, so there are several important choices we made with which reasonable researchers could disagree.
Should the animal advocacy movement change the amount of resources invested in these campaigns? Based on our calculations, this campaign does not appear supremely cost-effective. That said, the evidence supporting this campaign is quite strong, and this campaign is easy to scale.
Could mass media campaigns in developing countries be a particularly impactful opportunity for these campaigns? We believe that radio and mass media campaigns in developing countries can justifiably form part of the movement's overall portfolio of outreach campaigns for meat reduction. However, we do not think it would be justified to put a disproportionately high amount of the movement's resources into radio campaigns in developing countries.
Could we create a win-win by jointly funding meat-reduction campaigns with environmentally motivated funders? We find that environmental funders (at least the funders motivated primarily by cost-effectiveness) would not be keen to put large amounts of resources towards meat-reduction campaigns, even as a joint initiative. The reason is that meat-reduction campaigns appear to be a less cost-effective way to reduce greenhouse gas emissions than other options available to the environmental movement.
Acknowledgements
We are particularly grateful to the author of this post for providing highly relevant data about the cost per engagement of advertising campaigns. We are also grateful to the authors of several key publications for sharing their data with us.
References
1. Mathur MB, Peacock J, Reichling DB, Nadler J, Bain PA, Gardner CD, et al. Interventions to reduce meat consumption by appealing to animal welfare: Meta-analysis and evidence-based recommendations. Appetite. 2021 Sep 1;164:105277.
2. Mathur MB, Peacock JR, Robinson TN, Gardner CD. Effectiveness of a Theory-Informed Documentary to Reduce Consumption of Meat and Animal Products: Three Randomized Controlled Experiments. Nutrients [Internet]. 2021 Dec 20;13(12). Available from: http://dx.doi.org/10.3390/nu13124555
3. Animal Charity Evaluators. Animal Charity Evaluators. 2021. THE EFFECTS OF DIET CHOICES. Available from: https://animalcharityevaluators.org/research/reports/dietary-impacts/effects-of-diet-choices/
4. Norwood FB, Lusk JL. Compassion, by the Pound: The Economics of Farm Animal Welfare. Oxford University Press; 2011. 432 p.
5. Tomasik B. Reducing Suffering. 2018. Comments on Compassion, by the Pound. Available from: https://reducing-suffering.org/comments-on-compassion-by-the-pound/#Elasticities
6. Barrett J, Raskoff S. Ethical veganism and free riding. Journal of Ethics and Social Philosophy [Internet]. 2023; Available from: https://jesp.org/index.php/jesp/article/view/2334
7. Tomasik B. Reducing Suffering. 2014. Does Vegetarianism Make a Difference? Available from: https://reducing-suffering.org/does-vegetarianism-make-a-difference/
8. Chang KB, Wooden A, Rosman L, Altema-Johnson D, Ramsing R. Strategies for reducing meat consumption within college and university settings: A systematic review and meta-analysis. Frontiers in Sustainable Food Systems [Internet]. 2023;7. Available from: https://www.frontiersin.org/articles/10.3389/fsufs.2023.1103060
9. PreciousPig. Effective Altruism Forum. 2023. Cost comparison of promoting Animal Rights content on social media in high income vs. low income countries. Available from: https://forum.effectivealtruism.org/posts/QQ7TfFLooTZEjzpgg/cost-comparison-of-promoting-animal-rights-content-on-social
10. Winchell JS. Effective Altruism Forum. 2021. Opinion: Digital marketing is under-utilized in EA. Available from: https://forum.effectivealtruism.org/posts/cPCra34aCGrYzKcDF/opinion-digital-marketing-is-under-utilized-in-ea
11. Schwitzgebel E, Cokelet B, Singer P. Students eat less meat after studying meat ethics. Rev Philos Psychol [Internet]. 2023; Available from: https://link.springer.com/article/10.1007/s13164-021-00583-0
12. Jalil AJ, Tasoff J, Bustamante AV. Low-cost climate-change informational intervention reduces meat consumption among students for 3 years. Nature Food. 2023 Mar 2;4(3):218–22.
13. Macdonald BNJ, Caldwell KD, Boese GD. The effects of “reduce” and “eliminate” appeals on individual meat consumption [Internet]. [cited 2023 Aug 10]. Available from: https://files.osf.io/v1/resources/44z7j/providers/osfstorage/5c3170927cf3f5001bb0e3c6?action=download&direct&version=1
14. Hurford P. Everyday Utilitarian. 2014. How Much Suffering is in the Standard American Diet? Available from: https://web.archive.org/web/20141022030549/everydayutilitarian.com/essays/how-much-suffering-is-in-the-standard-american-diet/
15. Ladak A, Garcia CS, Anderson J. Impact of replacing animal products: Methodology [Internet]. [cited 2023 Aug 17]. Available from: https://osf.io/rjp4m/download
16. Springlea R. Animal Ask. 2022. Subsidies: Which reforms can help animals? Available from: https://www.animalask.org/post/subsidies-which-reforms-can-help-animals
17. Salmivaara L, Niva M, Silfver M, Vainio A. How vegans and vegetarians negotiate eating-related social norm conflicts in their social networks. Appetite. 2022 May 13;175:106081.
18. Vestergren S, Uysal MS. Beyond the Choice of What You Put in Your Mouth: A Systematic Mapping Review of Veganism and Vegan Identity. Front Psychol. 2022 Jun 10;13:848434.
19. Tomasik B. Reducing Suffering. 2017. Habitat Loss, Not Preservation, Generally Reduces Wild-Animal Suffering. Available from: https://reducing-suffering.org/habitat-loss-not-preservation-generally-reduces-wild-animal-suffering/
20. Bruers S. Effective Altruism Forum. 2023. The harm cascade: why helping others is so hard. Available from: https://forum.effectivealtruism.org/posts/cypLFJNbsngcDgJqm/the-harm-cascade-why-helping-others-is-so-hard
21. Loughnan S, Haslam N, Bastian B. The role of meat consumption in the denial of moral status and mind to meat animals. Appetite. 2010 Aug;55(1):156–9.
22. Anonymous. Effective Altruism Forum. 2022. Why I’m skeptical of moral circle expansion as a cause area. Available from: https://forum.effectivealtruism.org/posts/J4uiugyKFBcsLC78x/why-i-m-skeptical-of-moral-circle-expansion-as-a-cause-area
23. Plant M. The meat eater problem. Controversial Ideas. 2022 Oct 31;2(2):1.
24. Tomasik B. Reducing Suffering. 2018. Climate Change and Wild Animals. Available from: https://reducing-suffering.org/climate-change-and-wild-animals/
25. Amiot CE, El Hajj Boutros G, Sukhanova K, Karelis AD. Testing a novel multicomponent intervention to reduce meat consumption in young men. PLoS One. 2018 Oct 11;13(10):e0204590.
26. Animal Charity Evaluators. Animal Charity Evaluators. 2017 [cited 2021 Oct 6]. Leafleting. Available from: https://animalcharityevaluators.org/advocacy-interventions/interventions/leafleting/#report
27. Animal Charity Evaluators. Google Sheets. Welcome to the ACE Leafleting / Online Ads Impact Spreadsheet. Available from: https://docs.google.com/spreadsheets/d/1YSkZDTWacpkmnZMdRsIMLuCOdIILNVmGTIQAomZDxD4/edit#gid=0
28. Wildeford P. Effective Altruism Forum. 2016. Thoughts on the Reducetarian Labs MTurk Study. Available from: https://forum.effectivealtruism.org/posts/vprRvgjz4uW5LhfEC/thoughts-on-the-reducetarian-labs-mturk-study
29. Tomasik B. Reducing Suffering. 2012. Donating toward Efficient Online Veg Ads. Available from: https://reducing-suffering.org/donating-toward-efficient-online-veg-ads/
30. Animal Charity Evaluators. Animal Charity Evaluators. 2017. LENGTH OF ADHERENCE TO VEGETARIANISM. Available from: https://animalcharityevaluators.org/research/reports/dietary-impacts/vegetarian-recidivism/#9
31. Kaufman J. Jeff Kaufman. 2011. Estimating Charity Effectiveness: Vegan Outreach. Available from: https://www.jefftk.com/p/estimating-charity-effectiveness-vegan-outreach
32. Utilitarian Essays. Reducing Suffering. 2006. How Much Is a Dollar Worth? The Case of Vegan Outreach. Available from: https://reducing-suffering.org/wp-content/uploads/2014/10/dollar-worth.pdf
33. Wildeford P, Davis MA. Effective Altruism Forum. 2018. Animal Equality showed that advocating for diet change works. But is it cost-effective? Available from: https://forum.effectivealtruism.org/posts/9ShnvD6Zprhj77zD8/animal-equality-showed-that-advocating-for-diet-change-works
34. Norris J, Hannan R. Vegan Outreach. Leafleting and Booklet Effectiveness. Available from: https://veganoutreach.org/leafleting-and-booklet-effectiveness/
35. Animal Charity Evaluators. Animal Charity Evaluators. 2021. The Humane League. Available from: https://animalcharityevaluators.org/charity-review/the-humane-league/#comprehensive-review
36. Afshin A, Penalvo J, Del Gobbo L, Kashaf M, Micha R, Morrish K, et al. CVD Prevention Through Policy: a Review of Mass Media, Food/Menu Labeling, Taxation/Subsidies, Built Environment, School Procurement, Worksite Wellness, and Marketing Standards to Improve Diet. Curr Cardiol Rep. 2015 Sep 14;17(11):98.
37. Afshin A, Abioye AI, Ajala ON, Nguyen AB, See KC, Mozaffarian D. Abstract P087: Effectiveness of Mass Media Campaigns for Improving Dietary Behaviors: A Systematic Review and Meta-analysis. Circulation. 2013 Mar 26;127(suppl_12):AP087–AP087.
38. Phulkerd S, Thapsuwan S, Chamratrithirong A, Gray RS, Pattaravanich U, Ungchusak C, et al. Implementing population-wide mass media campaigns: Key drivers to meet global recommendations on fruit and vegetable consumption. PLoS One. 2022 Aug 17;17(8):e0273232.
39. Stead M, Angus K, Langley T, Katikireddi SV, Hinds K, Hilton S, et al. Mass media to communicate public health messages in six health topic areas: a systematic review and other reviews of the evidence. Southampton (UK): NIHR Journals Library; 2019.
40. Diddi P, Kumble S, Shen F. Efficacy of Radio Entertainment Education in promoting health behavior: A Meta-analysis. Journal of Radio & Audio Media. 2023 Jan 2;30(1):387–407.
41. Iacoella F, Gassmann F, Tirivayi N. Which communication technology is effective for promoting reproductive health? Television, radio, and mobile phones in sub-Saharan Africa. PLoS One. 2022 Aug 17;17(8):e0272501.
42. Founders Pledge. Mass media interventions report [Internet]. Founders Pledge; 2022. Available from: https://www.founderspledge.com/research/mass-media-interventions
43. Wakefield MA, Loken B, Hornik RC. Use of mass media campaigns to change health behaviour. Lancet. 2010 Oct 9;376(9748):1261–71.
44. Waters J. A model of the dynamics of household vegetarian and vegan rates in the U.K. Appetite. 2018 Aug 1;127:364–72.
45. Blyth M, Springlea R. Can the trajectory of meat consumption and production in developing countries be influenced by policy? An agenda for animal advocacy researchers [Internet]. Animal Ask; 2023. Available from: https://www.animalask.org/research-database
46. Khire I. Working paper: How can countries develop economically without pursuing intensive animal agriculture? Animal Ask; 2023.