The views expressed here are my own, not those of people who provided feedback on the draft. I started this cost-effectiveness analysis in the context of Ambitious Impact’s (AIM’s) research program from September to December of 2024.

Summary

• I estimate the cost-effectiveness of:
  • Veganuary in 2024 was 1.20 % that of cage-free corporate campaigns helping hens.
  • School Plates in 2023 was 19.4 % that of the aforementioned campaigns, and 16.2 times that of Veganuary in 2024.
• I encourage Veganuary to include questions about the impact of their campaigns on the consumption of red meat, white meat, eggs, shrimp, and fish in the surveys they use to determine the number of people who reported participating in Veganuary.
  • These are sent to random people in the target countries, so selection bias would be mitigated. I do not trust the results of Veganuary’s 6 month survey in 2024, according to which 81 % of people decreased their consumption of animal products over the last 6 months by at least 50 %, because only 0.0237 % as many people responded as reported participating in Veganuary in 2024.
  • Asking questions about specific types of animal-based foods is relevant to determine how Veganuary affected the consumption of each of them. I assumed the same effect size for all, but I believe varying effects can easily make Veganuary very harmful or way more beneficial.
• I would also add a question to the survey used to determine the number of people who participated in Veganuary which would check whether the respondents are replying in good faith. This is important because only a small fraction of the population participates in Veganuary, such that small errors matter. 4 % of Americans believe (actually, report believing) lizardmen are running the Earth.
• I advise Veganuary to strongly emphasise that eating red meat and dairy products is much better for animals than eggs, white meat, fish, or other seafood. They can even experiment with alternative pledges like Beefuary, Reduary, Cheesuary, Milkuary, Dairyuary, Cattleuary, and Ruminanuary, where beef, red meat, cheese, milk, dairy products, cattle-, and ruminant-based foods are allowed. One Step for Animals asks people to stop eating chicken. A pledge for flexitarians, Flexuary, may work too, but I think it would have to be made concrete, such as by asking people to only eat plant-based foods 4 days per week.
• Wendy Matthews, international head of partnerships and expansion at Veganuary, thinks I greatly underestimated the impact of Veganuary by guessing total benefits to be 2 times as large as those linked to the people who reported participating in Veganuary. It would be great if they could quantitatively justify why the vast majority of their benefits come from influencing people who did not participate in Veganuary.
• I recommend School Plates to be transparent about how they estimate the number of meals they replace, and investigate their composition.
• My results suggest cage-free campaigns are way more cost-effective than Veganuary, and more cost-effective than School Plates. Nonetheless, I believe such campaigns are far from the most cost-effective intervention. I recommend people funding Veganuary, School Plates, and cage-free campaigns support the Arthropoda Foundation, Shrimp Welfare Project (SWP), or Wild Animal Initiative (WAI).

Cost-effectiveness

Veganuary

Veganuary is “a non-profit organisation that encourages people worldwide to try vegan for January and beyond”. I estimate Veganuary in 2024 averted 0.944 suffering-adjusted days (SADs) per $. I get this from the ratio between:

• 2.99 M SADs averted in 2024. 1 SAD is as bad as 24 h of disabling pain in humans.
• 3.17 M$ of spending from 1 March 2023 to 29 February 2024. This is the last financial year for which there is data, and covers the last Veganuary campaigns in January 2024.

I calculate the benefits of Veganuary in 2024 multiplying:

• 25 M people who reported participating in Veganuary in 2024.
• 0.0597 SADs averted per person among the people who reported participating in Veganuary in 2024. This is based on:
  • 231 SADs per person in the United Kingdom (UK) in 2021 accounting for all food, and values in other countries proportional to the meat supply per person in 2021 excluding eggs, fish and other seafood. My estimate for the SADs averted per person among the people who reported participating in Veganuary in 2024 is 0.0213 % of that.
  • A reduction in the meat supply per person in January 2024 among those people of 0.7 % (= 0.1*0.07) of the standard deviation of the monthly meat supply per person in 2024 those people would have had in 2024 without Veganuary. This is 10 % of the effect size of 0.07 of the meta-analysis of Green et al. (2024), which is expressed as Glass’s Delta, the difference between the means of the treatment and control group as a fraction of the standard deviation of the control group.
    • The meta-analysis covered “the most rigorous randomized controlled trials (RCTs) that aim to reduce consumption of meat and animal products (MAP)”, which I expect to have a greater effect size that a large scale intervention reaching 25 M people at a cost of 0.127 $/person (= 3.17*10^6/(25*10^6)).
    • I asked the main author of the meta-analysis, Seth Green, for a guess about the effect size of Veganuary. Seth replied as follows. “I don’t know, sorry. There would be a lot of additional assumptions needed to extrapolate from the RCTs we analyze to this”.
    • My effect size refers to people who reported participating in Veganuary. The number of people who actually participated may be significantly smaller. Veganuary concluded 25 M participated multiplying small fractions of participants in the surveyed countries by their populations. However, such fractions may be significantly explained by a few people mistakenly reporting their participation due to lack of attention or social desirability bias.
  • A coefficient of variation of the meat supply per person in January 2024 those people would have had without Veganuary of 9.41 % (= 0.289/3.07). I get this from an estimate of the coefficient of variation of the monthly meat consumption per person among the population of 2.98 k people studied in Struijk et al. (2018). In particular, from the ratio between:

    • A standard deviation of 0.289 kg (= ((34.0^2 + 28.2^2 + 28.3^2)*365.25/12)^0.5). This accounts for processed meat, and non-processed red meat and poultry, and assumes pairwise independence of the daily consumption across days and types of meat^[1]. At least the latter is approximately correct^[2].

    • A mean of 3.07 kg^[3] (= (35.0 + 31.9 + 33.8)*365.25/12).

  • The SADs averted among those people are 6.00 times (= 12*(1 + 0)/2) the SADs averted among those people in January 2024. This would be equivalent to the effect size decreasing linearly over 12 months, from 0.007 in January 2024 to 0 in January 2025.
• 2, which is my guess for the SADs averted in 2024 as a fraction of those linked to the people who reported participating in Veganuary in 2024. It accounts for all the benefits caused by Veganuary’s activities in 2024, including benefits in subsequent years, and ones respecting corporate engagement. My guess is informed by my impression that the benefits linked to the people who reported participating in Veganuary in 2024, and to the corporate engagement this year are highlighted roughly as strongly in Veganuary’s 2024 impact report.

School Plates

School Plates is a program from ProVeg UK aiming to increase the consumption of plant-based foods at schools and universities in the UK. I estimate School Plates in 2023 averted 15.3 SADs/$ multiplying:

• 64.5 lunches/dinners replaced per $. I infer this from the ratio between:
  • 12.4 M lunches/dinners replaced in 2023, as provided by Colette Fox, head of School Plates, in the context of my past cost-effectiveness analysis of School Plates.
  • 193 k$ spent in 2023, which corresponds to the 155 k£ shared by Colette in the context of my past analysis.
• 0.237 SADs averted per lunch/dinner replaced. I get this from the ratio between:
  • 173 SADs per person respecting lunches and dinners in the UK in 2021, guessing 75 % of the SADs respect lunches and dinners, and relying on my estimate of 231 SADs per person in the UK in 2021.
  • 731 lunches/dinners per person-year, assuming 2 per person-day.

Cage-free campaigns

I estimate cage-free corporate campaigns helping hens avert 78.8 SADs per $. I get this multiplying:

• 10.8 hen-years affected per $, which is the product between:
  • Saulius Šimčikas’ estimate of 54 hen-years per $.
  • An adjustment factor of 1/5, as Open Philanthropy thinks “the marginal FAW [farmed animal welfare] funding opportunity is ~1/5th as cost-effective as the average from Saulius’ analysis”.
• 7.30 SADs averted per hen-year affected, as implied by AIM’s estimates for furnished cages and cage-free aviaries.

Comparisons

I estimate the cost-effectiveness of:

• Veganuary in 2024 was 1.20 % that of cage-free corporate campaigns helping hens.
• School Plates in 2023 was 19.4 % that of the aforementioned campaigns, and 16.2 times that of Veganuary in 2024.

Discussion

Veganuary

According to Veganuary’s 6 month survey in 2024, 81 % of people decreased their consumption of animal products over the last 6 months by at least 50 %. In contrast, my assumptions imply a reduction in the meat supply per person over the 1st half of 2024 of 0.525 % (= (1 + 1 - 6/12)/2*0.007), which is less than 1.05 % (= 0.00525/0.5) of the reduction reported by 81 % of the people surveyed by Veganuary. I do not trust their survey’s data due to:

• Massive selection bias:
  • The survey “was sent to 277,000 participants who took part by receiving 31 days of support emails”. Only “5,931 people responded”, 2.14 % (= 5.931*10^3/(277*10^3)) of the people who received the survey.
  • Moreover, a random person among the 277 k who received support emails was arguably much more affected than a random person among the 25 M who reported participating in Veganuary in 2024.
  • The people who responded are only 0.0237 % (= 5.931*10^3/(25*10^6)) of the people who reported participating in Veganuary in 2024. So it can be that a random participant was less than 1.05 % as much affected as a random person who responded to the survey.
• Unreliability of self-reports. For example, due to social desirability bias.

I would also add a question to the survey used to determine the number of people who participated in Veganuary which would check whether the respondents are replying in good faith. This is important because only a small fraction of the population participates in Veganuary, such that small errors matter.

I encourage Veganuary to include questions about the impact of their campaigns on the consumption of red meat, white meat, eggs, shrimp, and fish in the surveys they use to determine the number of people who reported participating in Veganuary.

• These are sent to random people in the target countries, so selection bias would be mitigated.
• Asking questions about specific types of animal-based foods is relevant to determine how Veganuary affected the consumption of each of them. I assumed the same effect size for all, but I believe varying effects can easily make Veganuary very harmful or way more beneficial:
  • I estimate 56.8 % of the SADs in the UK in 2021 came from farmed shrimp, which in my model implies 56.8 % the benefits of Veganuary come from decreasing the consumption of farmed shrimp, although this only accounts for 0.810 % of the total consumption in terms of mass.
  • I worry about replacements of red meat with white meat, eggs, and farmed aquatic animals, which respect a greater suffering per kg. Veganuary started in the United Kingdom in 2014, and the production of broilers per person in there in January increased 4.27 times as fast from 2014 to 2024 as from 1994 to 2013. Data about consumption (production plus net imports) would be more informative, and the production of broilers per person could have increased faster without Veganuary, but the correlation is still concerning.

I advise Veganuary to strongly emphasise that eating red meat and dairy products is much better for animals than eggs, white meat, fish, or other seafood. They can even experiment with alternative pledges like Beefuary, Reduary, Cheesuary, Milkuary, Dairyuary, Cattleuary, and Ruminanuary, where beef, red meat, cheese, milk, dairy products, cattle-, and ruminant-based foods are allowed^[4]. One Step for Animals asks people to stop eating chicken. A pledge for flexitarians, Flexuary, may work too, but I think it would have to be made concrete, such as by asking people to only eat plant-based foods 4 days per week.

Wendy Matthews, international head of partnerships and expansion at Veganuary, thinks I greatly underestimated the impact of Veganuary by guessing total benefits to be 2 times as large as those linked to the people who reported participating in Veganuary. It would be great if they could quantitatively justify why the vast majority of their benefits come from influencing people who did not participate in Veganuary.

Vicky Cox, senior animal welfare researcher at AIM, also did a cost-effectiveness analysis of Veganuary. Vicky’s 6 estimates range from 0.722 to 97.3 SADs averted per $, i.e. 0.00916 (= 0.722/78.8) to 1.23 (= 97.3/78.8) times my estimate for cage-free campaigns. The highest estimate is 103 (= 97.3/0.944) times my estimate for the cost-effectiveness of Veganuary in 2024. In my model, it would correspond to assuming an effect size in terms of Glass’s Delta for January 2024 of 0.721 (= 0.007*103). I see this as unreasonably high even for a treatment group with just hundreds of people, so I do not think it is at all applicable to the 25 M people who reported participating in Veganuary in 2024. It is 6.01 (= 0.721/0.12) times the upper bound of the 95 % confidence interval of the meta-analytic effect size of Green et al. (2024).

School Plates

I estimate School Plates replaced 19.0 (= 64.5/3.4) times as many meals per $ in 2023 as Sinergia Animal’s analogous program, Nourishing Tomorrow. The cost-effectiveness of School Plates is proportional to the number of replaced lunches/dinners provided by Colette, but I have not checked how it was estimated, and organisations often inflate their impact. I asked Colette about it on February 13, but I have not heard back. I recommend School Plates to be transparent about how they estimate the number of meals they replace, and investigate their composition.

Comparisons

My estimate that School Plates in 2023 was 16.2 times as cost-effective as Veganuary in 2024 may be surprising. It implies a significant difference in their cost-effectiveness despite both focussing on decreasing the consumption of animal-based foods. I would not be surprised if the difference was smaller. Nevertheless, the difference is not large enough to raise red flags for me. The cost-effectiveness of human welfare interventions can vary much more than that. In the context of animal welfare, I estimate:

• Cage-free campaigns are 64.5 (= 1/0.0155) times as cost-effective as Fish Welfare Initiative’s (FWI’s) farm program.
• SWP has been 173 times as cost-effective as cage-free campaigns.
• Paying farmers to use more humane pesticides would be 51.4 times as cost-effective as cage-free campaigns, and guess that research on and advocacy for more humane pesticides would be way more cost-effective than paying farmers to use them more. WAI does research on pesticides.

My results suggest cage-free campaigns are way more cost-effective than Veganuary, and more cost-effective than School Plates. Nonetheless, as illustrated above, I believe such campaigns are far from the most cost-effective intervention. I recommend people funding Veganuary, School Plates, and cage-free campaigns support the Arthropoda Foundation, SWP, or WAI.

My estimates for SADs averted rely on AIM’s pain intensities, which I believe greatly underestimate the intensity of excruciating pain. They say this is only 56.7 times as intense as hurtful pain, and I guess this is as intense as fully healthy life. So I infer AIM’s pain intensities imply excruciating pain is only 56.7 times as intense as fully healthy life. In this case, 1 day of fully healthy life plus 25.4 min (= 24*60/56.7) of “scalding and severe burning events [in large parts of the body]”, or “dismemberment, or extreme torture” would be neutral, whereas I believe it would be clearly bad. Yet, I do not think AIM’s much lower intensity of excruciating impacts much the comparisons between Veganuary, School Plates, and cage-free campaigns, as none of these overwhelmingly focuses on decreasing excruciating pain. Feel free to ask Vicky for the sheet with AIM’s pain intensities, and the doc with my suggestions for improvement.

Acknowledgements

Thanks to Sagar Shah, Vicky Cox, and Wendy Matthews for feedback on the draft^[5].

1. ^{^}

    In this case, “variance of the monthly consumption of meat” = (“variance of the daily consumption of processed meat” + “variance of the daily consumption of non-processed red meat” + “variance of the daily consumption of non-processed poultry”)*“number of days per month”.

2. ^{^}

    “In this population, mean (standard deviation) consumption was 35.0 (34.0) g/day for processed meat, 31.9 (28.2) g/day for red meat, and 33.8 (28.3) g/day for poultry. The correlations between the intake of the different meat categories (in g/day) were: r = 0.05, p < 0.01 (between processed meat and red meat), r = -0.02, p = 0.21 (between processed meat and poultry), and r = 0.11, p < 0.01 (between red meat and poultry).”

3. ^{^}

    “Mean monthly consumption of meat” = (“mean daily consumption of processed meat” + “mean daily consumption of non-processed red meat” + “mean daily consumption of non-processed poultry”)*“number of days per month”.

4. ^{^}

    Thanks to ChatGPT for suggesting arguably catchy names.

5. ^{^}

    I listed the names alphabetically.