Consequences of animal product consumption (combined model)

(this is mostly old con­tent—I’m just com­bin­ing this and this into a sin­gle post with slight up­dates)

This spread­sheet es­ti­mates the con­se­quences of an­i­mal product con­sump­tion. I start by dis­count­ing 1kg pur­chased of each an­i­mal product by the cu­mu­la­tive elas­tic­ity fac­tor. But when other peo­ple stop buy­ing an ex­pen­sive an­i­mal product, they may start buy­ing a differ­ent one. There­fore, I make rather ar­bi­trary guesses for these effects—how much more chicken will peo­ple buy when they give up beef, etc. So for each ad­di­tional 1kg of beef you buy, I es­ti­mate that the ad­di­tional pro­duc­tion will be 0.68kg of beef, 0.05kg of chicken, 0.01kg of salmon and so on.

I use data on lifes­pan and meat yields, as­sump­tions for in­trin­sic moral weight rel­a­tive to hu­mans in­formed by in­tu­ition and neu­ron counts (pigs = 0.17, cows = 0.15, turkey = 0.07, chicken = 0.06, fish = 0.03), and as­sump­tions for qual­ity of life (com­bined rat­ings from Char­ity En­trepreneur­ship and from Com­pas­sion by the Pound) to then es­ti­mate the net welfare lost per kg pur­chased. The figures are in hu­man-equiv­a­lent welfare-days on a −100 to +100 scale, where −100 means none of your needs are met (note, still much bet­ter than tor­ture). So when the spread­sheet says 1kg of chicken causes −37.7 welfare (cell AI3), that means that buy­ing 1kg of chicken causes an equiv­a­lent amount of an­i­mal suffer­ing to the feel­ing of be­ing fully hun­gry, mis­er­able, in pain, etc for about 6.5 hours out of a 17 hour wak­ing day, in­stead of feel­ing com­pletely neu­tral dur­ing that time.

Then I use data on CO2e emis­sions per kg of food pro­duced. Note that this doesn’t ac­count for the coun­ter­fac­tual emis­sions from eat­ing plants—but these are gen­er­ally much lower than the emis­sions from an­i­mal prod­ucts. Us­ing both short-run and long-run es­ti­mates of the costs of cli­mate change and other en­vi­ron­men­tal costs upon hu­mans, I get a mea­sure­ment of the hu­man cost of each kg of CO2e.

With some typ­i­cal calorie/​kg num­bers, I get com­bined figures for the welfare costs of 750 calories which is ap­prox­i­mately one meal.

I then use es­ti­mates of effec­tive char­ity im­pacts to see how much money it would cost to Pareto offset these im­pacts (i.e.: donate to cli­mate change char­i­ties to offset the CO2e, donate to an­i­mal char­i­ties to offset the an­i­mal suffer­ing).

Fi­nally, I look at the amount of ad­di­tional farm an­i­mal suffer­ing that is pre­dicted when peo­ple gain in­come in a set of Afri­can coun­tries sup­ported by GiveWell aid pro­grams—the “meat eater prob­lem.”

Note: this model ig­nores wild an­i­mal suffer­ing, leav­ing it all to be con­sid­ered un­clear. The calcu­la­tions for wild fish are just about the pos­si­ble in­di­rect effects on farm an­i­mal prod­ucts.


  • Farmed fish is the worst product, poul­try/​eggs/​pork/​beef/​wild fish are similar in the mid­dle, and cheese/​milk are the most benign

  • For some­one with ac­cess to ad­e­quate ve­gan food, all typ­i­cal an­i­mal prod­ucts have enough costs to out­weigh the di­rect plea­sure and con­ve­nience of con­sump­tion, with the pos­si­ble ex­cep­tion of small amounts of dairy

  • Us­ing de­cently con­ser­va­tive as­sump­tions for char­ity im­pacts, the eth­i­cal costs of a high-meat diet can be offset with ap­prox­i­mately a nickel a day donated to rele­vant effec­tive char­i­ties (or less, if you se­lect a more effec­tive cause area in­stead of a cli­mate & an­i­mals bun­dled dona­tion)

  • In­creas­ing an­nual in­come for a poor per­son by $1,000/​year has a farm an­i­mal welfare cost equiv­a­lent to adding sev­eral weeks of mis­ery. To me, given the di­rect noneco­nomic benefits of efforts such as malaria pre­ven­tion and de­worm­ing, such efforts seem ro­bustly good enough to out­weigh the an­i­mal costs in the short run. Things like in­dus­tri­al­iza­tion and GiveDirectly seem to have un­clear short run welfare im­pact, de­pend­ing on the char­ac­ter­is­tics of the re­cip­i­ent populations