TL;DR
On greenhouse emissions per 100g of protein, beef from beef herds runs at about 49.9 kg CO₂-equivalent and peas at 0.44 — a gap of about 113 times.1 So no, foods are not equal, and anyone claiming otherwise is wrong. But the ranking that falls out of the measurements is not the ranking people hold: cheese (10.8) is nearly double poultry (5.7), and rice (6.3) sits above chicken too.1 Meanwhile the psychology of moralised eating is organised around disgust rather than harm — moral vegetarians score significantly higher on disgust toward meat than health vegetarians (9.9 vs 6.9) while showing no difference in how much they like its taste, smell or texture.2 In India, 39% of adults call themselves vegetarian, with a caste gap of 53% to 40% among Hindus and a much larger regional gap of 71% to 18%.3 Both rankings are real. They are not the same ranking, and only one of them has an arithmetic.
Everything here sits somewhere on a measurable harm ranking. Almost nobody buying it is using that ranking. Photo: Abdullah Azeez on Unsplash.4
Two Questions Wearing One Costume
“Are all foods morally equal?” is usually asked as a rhetorical question, and it gets answered as if it were one. But there are two entirely separate questions inside it, and almost every argument about diet conflates them.
The first is whether foods differ in measurable consequences — emissions, land, water, animal suffering, health outcomes. That question has answers, and they are lopsided.
The second is whether the moral rankings people actually hold and enforce correspond to those measurements. That question also has an answer, and it is largely no.
Holding both is uncomfortable, because each side of the usual argument owns one of them. The relativist (“who are you to judge my food”) is wrong on the first question. The moralist (“my diet is the ethical one”) is usually wrong on the second. Being right about one does not get you the other for free.
What a Real Ranking Looks Like
The reference dataset here is Poore and Nemecek’s meta-analysis, built from roughly 38,700 farms across 119 countries covering 40 products — about 90% of global protein and calorie consumption. Our World in Data publishes the per-protein figures, in kilograms of CO₂-equivalent per 100 grams of protein:1
| Food | kg CO₂eq / 100g protein |
|---|---|
| Beef (beef herd) | 49.9 |
| Lamb & mutton | 19.9 |
| Prawns (farmed) | 18.2 |
| Beef (dairy herd) | 16.9 |
| Cheese | 10.8 |
| Milk | 9.5 |
| Pig meat | 7.6 |
| Rice | 6.3 |
| Fish (farmed) | 6.0 |
| Poultry | 5.7 |
| Eggs | 4.2 |
| Tofu | 2.0 |
| Groundnuts | 1.2 |
| Other pulses | 0.84 |
| Peas | 0.44 |
| Nuts | 0.26 |
Read the shape rather than the individual rows. There is an enormous spread — two orders of magnitude from top to bottom — and it is not a plants-versus-animals binary. Eggs and poultry sit below cheese and milk. Rice, the most moralised-as-innocent staple on earth, sits above chicken.
One honest caveat about the metric before anyone builds a diet on it. Per-100g-of-protein is the right denominator for comparing protein sources, and it is doing real work — it stops you concluding that beans are worse than beef because you’d need to eat more of them. But it distorts badly for foods nobody eats for protein. The same dataset puts dark chocolate at 93.3 and coffee at 35.7, which sounds damning until you notice that a person’s annual protein intake from chocolate is approximately zero.1 Emissions per kilogram, per calorie and per protein gram give three different league tables, and choosing among them is itself a judgement. I am using per-protein because the interesting comparisons in this piece are between things people eat as their protein.
What survives all of that is the shape: foods differ hugely, the differences are measurable, and the boundary between the high-impact and low-impact groups does not fall where dietary moral categories fall.
The Ranking People Actually Hold
Now the other side, and this is where the psychology gets specific.
Paul Rozin, Maureen Markwith and Caryn Stoess studied 104 meat-avoiders and split them by their original reason: 36 who became vegetarian for moral reasons with no health reason, and 26 who became vegetarian for health reasons with no moral reason.2 Their subject was what they called moralization — the process by which something morally neutral acquires a moral component, converting a preference into a value.2
The moral-origin group differed on almost everything you’d predict. They gave more total reasons for their vegetarianism (67.7 vs 57.7 on the composite). They rejected a wider range of animal products (23.8 vs 19.9). And they scored substantially higher on the composite disgust measure — 9.9 against 6.9, a difference the authors report as significant at p < .001.2
Then the finding that makes the whole thing interesting. On the sensory measures — how much they disliked meat’s taste, smell, texture and appearance — there was no significant difference between the two groups, and if anything the moral vegetarians scored slightly lower on dislike.2 Moralising a food does not change whether you would enjoy it. It adds disgust on top of unchanged liking.
Rozin’s group also probed a set of “personality” beliefs: that eating meat makes people more violent, more aggressive, more animal-like. The correlation between that score and moral-ecological motivation was .77 — among the highest in the paper.2 Which is a contamination belief, in the technical sense: the idea that a substance transmits its properties to whoever consumes it. Not a claim about consequences in the world. A claim about what the food does to the eater’s nature.
That is a different logic from the emissions table, and the difference is structural rather than a matter of degree. Harm is quantitative and continuous — you can eat 30% less beef and cause 30% less harm. Purity is categorical. You cannot be 30% contaminated. Which is why the moralised version of a diet is so much more likely to be all-or-nothing than the measured version has any reason to be.
The Indian Case, Carefully
India is where this argument is least abstract, so it’s worth using real numbers rather than impressions.
Pew’s 2021 survey found 39% of Indian adults describing themselves as vegetarian, with 81% restricting meat in some way — avoiding certain meats, or avoiding meat on certain days, or both.3 By community: Jains 92%, Sikhs 59%, Hindus 44%, Buddhists 25%, Christians 10%, Muslims 8%.3
Two gradients within that are worth putting side by side, because the comparison disciplines the argument.
Among Hindus, members of Scheduled Castes, Scheduled Tribes and other lower castes were less likely than General Category Hindus to describe themselves as vegetarian: 40% versus 53%.3 That gap is real and consistent with vegetarianism functioning partly as a status marker.
But the regional gradient is far larger. Hindu vegetarianism ran at 71% in the North and 61% in Central India, against 30% in the South, 19% in the Northeast and 18% in the East.3 A 53-point spread, against a 13-point caste spread.
I think that comparison is the honest way to hold this. The caste story is not nothing — 13 points is a real gap and it points the way the critics say it does. But geography moves the number four times as much, which means any account treating Indian vegetarianism as essentially a caste-purity practice is over-reading the data. Regional food cultures, agricultural history and what was locally cheap are doing more work than a purity gradient alone can carry.
What the moral-ranking question does cut against is narrower and sharper. A lacto-vegetarian diet built on paneer, ghee and milk is, on the measured ranking, worse than one containing eggs and chicken — cheese at 10.8 and milk at 9.5 against eggs at 4.2 and poultry at 5.7.1 Whatever the vegetarian/non-vegetarian line is tracking in Indian food morality, it is demonstrably not tracking measured harm, because it puts the higher-impact food on the permitted side.
What This Settles and What It Doesn’t
It does not settle whether vegetarianism is right. That question turns on animal moral status, which emissions data has nothing to say about, and a person who thinks killing animals is wrong is not refuted by being shown that cheese emits more than chicken. Their argument was never about emissions.
What it settles is narrower: the moral vocabulary of food and the measurable consequences of food are two different orderings, and the people using the vocabulary mostly aren’t consulting the ordering. A person who eats no meat and a great deal of dairy, and who experiences the meat-eater as unclean, is running a purity system. That system may correlate with reduced harm in some places and increase it in others, and its correlation with harm is not what makes it feel compelling — the disgust data says the feeling is doing its own work.
There is a practical implication I’d defend. If you actually want less harm from your eating, the intervention with the best measured return is not adopting a category. It is reducing the two or three highest-impact items, which the table identifies precisely. Someone who eats beef twice a month instead of twice a week has, on these numbers, done more than someone who switched to a paneer-heavy vegetarian diet. Categories are legible and satisfying. Quantities are what the emissions respond to.
Where I’d Hold This Loosely
Three limits.
The Rozin study is from 1997, with 104 self-selected American respondents recruited by campus advertisement, split into subgroups of 36 and 26. That is a small, unrepresentative sample, and the framework it proposes — moralization as a general process — has been far more influential than any single one of its numbers deserves. I’d treat the direction as well supported and the specific figures as illustrative.
Second, emissions are one axis among several. Land use, water, biodiversity, antibiotic resistance and animal welfare all produce partially different rankings, and a food can score well on carbon and badly elsewhere. I’ve used one metric because it’s the best-measured one, not because it’s the only thing that matters, and a full accounting would move some rows.
Third — the caveat against my own framing — “your moral ranking doesn’t match the measurements” is not automatically a criticism. Some purity-based food rules have had defensible functions: coordination, group cohesion, historically sound heuristics about spoilage and disease. The critique lands only against the specific move of presenting a purity ranking as a harm ranking, which is a claim about the world and can therefore be wrong. Eating according to your tradition is not a mistake. Telling someone their food is worse for the world, when the numbers say yours is, is a mistake — and it’s a very easy one to make, because the feeling that accompanies both is identical.
Footnotes
-
https://ourworldindata.org/grapher/ghg-per-protein-poore ↩ ↩2 ↩3 ↩4 ↩5
-
https://foodethics.univie.ac.at/fileadmin/user_upload/inst_ethik_wiss_dialog/Rozin__.._1997._Moralization_and_becoming_a_vegetarian_in__psychological_science.pdf ↩ ↩2 ↩3 ↩4 ↩5 ↩6
-
https://www.pewresearch.org/religion/2021/06/29/religion-and-food/ ↩ ↩2 ↩3 ↩4 ↩5
-
https://unsplash.com/photos/a-person-sorting-fresh-green-leafy-vegetables-at-a-market-yDDRJS9hdwM ↩
