Wednesday, November 04, 2015

Some considerations for prioritization within animal agriculture

Summary: Conflicting rationales have been offered to prioritize reduction of particular sectors of factory farming and animal agriculture, and I review a selection of these. Cattle make the largest contribution to climate change, and cattle raised for meat create the greatest demand for agricultural land use. Smaller chickens and farmed fish are much more numerous. Taking still more numerous wild animals into account would suggest that cattle farming has the largest impact, positive or negative. Taking neural capacity into account favors attention to large farmed animals, but this measure is still dominated by wild animals. Mental abilities such as learning and social intelligence do not seem to have strong implications between chickens and cattle. An alternative perspective is that focus should be on changes in human attitudes, efforts, and organizations, as these contribute to further change.

Greenhouse gas emissions of factory farming are overwhelmingly due to ruminants, mostly cattle, with minimal contributions from chickens
Agriculture is a substantial source of greenhouse gas emissions, especially methane and nitrous oxide (these gases have shorter atmospheric lifetimes than carbon dioxide, so the relative contribution will be higher the shorter the time horizon for analysis). The U.S. Environmental Protection Agency attributes 9% of US emissions to agriculture given its analytical choices, about 600 million tons of CO2 equivalent.[1]:

Total U.S. Greenhouse Gas Emissions by Economic Sector in 2013
Pie chart of total U.S. greenhouse gas emissions by economic sector in 2013. 31 percent is from electricity, 27 percent is from transportation, 21 percent is from industry, 12 percent is from commercial and residential, and 9 percent is from agriculture.
Total Emissions in 2013 = 6,673 Million Metric Tons of CO2 equivalent 
*Land Use, Land-Use Change, and Forestry in the United States is a net sink and offsets approximately 13% of these greenhouse gas emissions.
All emission estimates from the Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1990-2013.
We can combine this with the EPA social cost of carbon estimates:

Social Cost of CO2, 2015-2050 a (in 2014 Dollars per metric ton CO2)
Source: Technical Support Document (PDF, 21 pp, 1 MB): Technical Update of the Social Cost of Carbon for Regulatory Impact Analysis Under Executive Order 12866 (May 2013, Revised July 2015)
Discount Rate and Statistic
Year5% Average3% Average2.5% Average3% 95th percentile
a The SC-CO2 values are dollar-year and emissions-year specific and have been rounded to two significant digits. The 2007$ estimates were adjusted to 2010$ using GDP implicit price deflator (108.289) from the National Income and Product Accounts Tables, Table 1.1.9.
Using the 2015 numbers and a 2.5% discount rate U.S. agriculture would be responsible for $37.2 billion in social costs from greenhouse gas emissions, $72 billion using the 3% discount rate and 95th percentile damage estimate.

For comparison, the agricultural sector makes up ~1% of U.S. GDP at $166 billion, while agriculture and food-related industries make up $789 billion. Farm beef sales were $44 billion in 2014, and cattle account for a disproportionately share of greenhouse gas emissions from their ruminant digestion and manure, as well as emissions associated with their large need for feed per kg of meat produced.

The Food and Agriculture Organization estimates that livestock contribute the bulk of global agricultural greenhouse gas emissions, 14.5% of the global total from human activity. Production of feed (including land-use change, which is a one-time cost) accounts for 45% of livestock emissions, 39% from flatulent cattle and other ruminants, 10% to manure processing, and the remainder to transportation. Ruminant emissions are higher when consuming less digestible feed such as grass.

By species and industry: beef cattle contribute 41% of global livestock emissions and dairy cattle 20%, pigs 9%, buffalo 8%, small ruminants 6%, and chicken meat and eggs 6%.

A focus on the impacts of factory farming on climate change would seem to clearly favor a focus on cattle (61% of livestock emissions) over chickens (6%).

Animal agriculture's impacts on price and supply of plant foods: impact on the global poor
Today humans consume approximately half of world grain production, while one third goes to livestock and much of the remainder to environmentally dubious biofuel production. The vast majority of world soybean production (with its high protein content) goes to feed livestock. As meeting this demand pushes towards food supply from more marginal agricultural land with worse conditions, the demand for animal products raises the price of plant crops generally.

Estimating the precise impact on food prices requires a careful examination of long-run elasticity (short-run price movements are a poor guide). There is a substantial economic literature on the question, and much recent work in the context of biofuels, as economists attempt to estimate the impacts of diverting a substantial portion of global crops to produce fuel.

One might hope that by reducing meat consumption grain and other crop prices would fall, disproportionately raising the incomes of the global poor for whom food makes up a larger portion of consumption. This seems plausible if one entirely eliminated all global meat consumption in one fell swoop, but for marginal changes (e.g. reducing factory farming by 10% in one country) leverage may be poor. People in rich countries consume many more calories per day than those in poor countries, and far more meat, as well as biofuels. So even if the benefits of lower grain prices are not as skewed as incomes, they are still distributed in quite unequal fashion.

Prioritization based on effects on crop prices would favor a focus on reducing farming of animals that are least efficient in converting feed into meat, milk or eggs. Statistics for feed conversion ratio (FCR) or feed conversion efficiency (FCE), i.e. kg of feed per kg of animal product, bear on this topic but are very easily misinterpreted. Feeds and products differ in their carbohydate, protein, fat, and thus caloric content, as well as digestibility.

Dairy production is relatively energy-efficient directly, although these figures exclude food consumption by calves:
Feed efficiency for cows
Both laying hens and broiler chickens are also highly efficient (although less so than plant foods):

Modern broilers weigh about 2.5 kg at 39 days, with a live-weight feed conversion ratio of 1.6 kg of feed per kilogram of body weight gain. Laying hens in modern commercial flocks typically produce about 330 eggs per year with a FCR of 2 kg of feed per kilogram of eggs produced.
Beef cattle have much lower efficiency of conversion for a given food type, partially offset by the ability to use pasture land that might otherwise be ill-suited to farming, although conversion efficiency for grass is very low (with high greenhouse gas emissions). One agricultural foundation source:
How efficient is our beef production system? On the surface, it does not look very efficient compared to other meat products. We can convert 2.5 pounds of feed concentrate into 1 pound of chicken and 3.5 pounds of feed concentrate into 1 pound of pork, but it takes about 6 pounds of concentrate to produce 1 pound of beef. However, this isn't the whole story. Table 1 shows the total amount of beef, pork and chicken produced in the U.S. in 2009 as well as the pounds of concentrate (grain) used to produce the meat, assuming the previously mentioned feed conversion ratios.

A pig or chicken spends its life consuming feed concentrates while most beef is raised on grass to a weight of 750 pounds prior to entering the feedyard. The average slaughter weight of a steer is 1,300 pounds, so only the last 550 pounds of beef is produced from grain. Thus, beef only uses 2.5 pounds of grain per pound of total product produced. When corn gets expensive, cattle are typically kept on grass to reach heavier weights. Conversely, when corn is cheap, we can afford to place cattle in the feedyard at lighter weights. The beef industry has the luxury of diversification between grass and grain, giving us an advantage when compared to pork and poultry. We also have the advantage of being able to use our grazingland resources without fear of competition for that resource from pork or poultry.
Nonetheless, overall cattle impose greater demands on agricultural land, which is reflected in the higher cost of beef. Pigs stand intermediate between cattle and chickens.

Quality of life

In Compassion by the Pound (page 229), Bailey Norwood gives subjective ratings (with detailed explanation) of the typical welfare of farmed animals.

A sample of n=1 is obviously inadequate, but the rough rank-ordering seems to be recapitulated by many different people thinking about animal welfare.

However, the assignation of cardinal scores, and especially relative to a threshold of 'lives worth living' is very controversial, and many animal advocates and experts strongly reject the idea that, e.g. broiler chickens in factory farmed conditions are better off alive than dead in light of the poor conditions.

To get estimates usable for prioritization I would want to see multiple surveys and panels, drawn from biology, philosophy, animal behavior, and other relevant fields. I would like to see welfare measurements carefully linked to animal willingness to pay (measured in behavioral experiments), and physiological measures of well-being such as cortisol. And I would want to test in humans how the types of behavioral and physiological evidence available for animals (unable to speak) translated to predicting human's subjective well-being and valuation of their own lives.

Some such studies have been conducted in the literature, including in the context of setting welfare standards.

Number of farmed animals killed

Compare to animal quality-adjusted life years, this measure places more weight on small and short-lived animals. USDA data, via the Humane Society:

U.S. Slaughter Totals, by Species (1950 - 2015*) : Totals

Updated June 25, 2015
YearCattleChickensDucksHogsSheep & LambsTurkeysTotal
(thousands of animals)(billions
of animals)

Numbers of fish and shellfish eaten are greater than for chickens.

This measure obviously favors a focus on broiler chickens and sea life. However, if one attends to all the animals killed by agriculture, including wild animals, the picture would change drastically, since the production of feed and other processes for the large land animals affect great numbers of wild animals.

Number of farmed animal-years
Different farmed animals have quite different average length of life before they are killed, from several years for cows raised for milk, to 7 weeks for broiler chickens. The sources from my previous post on animal life-years per kg in the United States give inventories for several kinds of farmed animal.

The inventory of laying hens (including about 1/6th mothers of broilers) in 2014 varied between 330-360 million , while the inventory of all cattle (including beef cattle, dairy cows, calves to be slaughtered for veal) was ~88 million. Multiplying the 8.66 billion broiler chickens killed by 47 days/365 days, and adjusting for 4.3% non-slaughter mortality (assuming death is evenly distributed), I estimate an average standing population of 1.14 billion broilers. Pig inventories were 61.5 million on March 1, 2014, 6 million kept for breeding and the remainder raised for slaughter after several months.

Farmed animalAverage population (millions)Percentage of listed farmed land animals
Layer chickens34521.11%
Broiler chickens114069.75%

Farmed fish would make another major contribution on at least the same order as chickens. Animal Charity Evaluators puts the number of farmed fish-years at about 20% greater than layer and broiler chickens combined, while ignoring farmed marine invertebrates.

Effects on wild animal populations, mortality, and welfare
If one is going to prioritize approximately by number of deaths or animal-years, without orders-of-magnitude adjustments for animals with different nervous systems or the like, then it seems that agricultural impacts on wild animals will loom larger than those on the farmed animals themselves, because wild animals are far more numerous. Matheny and Chan (2005) argue that growing crops to feed large livestock will reduce bird and mammal populations through land use change, and when considering reptiles, amphibians, fish, and invertebrates effects could be much larger.

Tomasik provides a table of estimates of wild animal populations:

Animal TypeWorld Population
Animals in Research Labs108 (underestimate)
Humans7 * 109
Livestock2.4 * 1010
Land Birds6 * 1010 to 4 * 1011
Land Mammals1011 to 1012
Land Reptiles1012 to 1013 (?)
Land Amphibians1012 to 1013 (?)
Fishat least 1013
Coral polyps1015 to 1018
Dust mitesmore than 1016
Insects1018 to 1019
Zooplankton1018 to 1021
Land use changes, crop residues, food waste, and greenhouse gas emissions from agriculture could in combination increase or decrease these wild animal populations, which one might view as positive or negative. Depending on views about the net sign of the population effect, and the desirability of larger or smaller wild animal populations this could favor a focus on cattle (which produce the greatest greenhouse gases and land use changes) or chicken and farmed fish/marine life (which produce the least). But in either case the actual farmed animals would be far outnumbered by the affected wild animals.

If one wishes to focus on killing, agriculture involves the killing of very large numbers of wild animals through pesticides, agricultural equipment, exclusion from food sources, and transport of finished goods.

However, once one is considering the well-being of wild animals, the best opportunities to assist might not involve factory farms at all.

In some ways, the move from a focus on farmed animals to more numerous but more socially distant wild animals is analogous to the move from the common focus of animal activists on charismatic and companion animals to more numerous farmed animals.

Entomophagy, and addition of smaller animals to farmed status
In recent years there have been some efforts by international bodies, scientists, and startups to spread the practice of entomophagy in the developed world. This is based on advantages in food conversion efficiency, greenhouse gas emissions, and other environmental advantages relative to some other animal products. However, perspectives that do not strongly adjust weights for different types of organism could be worried that this would greatly increase the population of farmed animals.

Similar issues could arise with aquaculture and small marine organisms.

Weighting by neural scale and number of computational processes
Many of the arguments for overwhelming focus on the smallest animals to be found involve a step along the lines of "this tiny nervous system might be less morally weighty than big and complex ones, but there is a nontrivial probability p that it is just as weighty, so 1/p of these tiny systems will equal a big one in importance." This reasoning is faulty along the lines of the two envelopes problem, since possibilities differ not only in relative values but also absolute values, and we need a prior over the latter.

If we avoid that confusion, there are questions about the valuation of different neural and computational processes when combined in the same brain or divided across many brains. Biological organisms differ by orders of magnitude in their quantity of neurons and brain tissue, and we might not assign far lower expected value to a brain process (e.g. positive reinforcement in response to a reward) because it is one of many in parallel in a brain. If so we might be interested in the scale of the nervous systems of different organisms. [See this thread, or a follow-up essay by Tomasik summarizing some of the discussion.]

In a previous post I found statistics for some farmed land animals:

AnimalGlobal populationBrain mass (g)Total grams of nervous systemPercentage of human and farmed land animal brain mass.

Fish tend to have much smaller brains relative to body size than birds or mammals, and the hyper-numerous insects have brains several orders of magnitude smaller than those of chickens. However, collectively wild invertebrates have many times the total neural capacity of all humans and farmed animals.

Weighting by particular mental abilities
Neural capacity might be a relevant measure when looking at processes with parallel elements (including positive and negative reinforcement), but many of our moral evaluations are based on overall capacities of the individual. Problem-solving, communication, empathy, and social reasoning (including reciprocity) show sharply diminishing returns to computational capacity across the animal kingdom. Even a honeybee brain with a few hundred thousand neurons can perform some associative learning, reciprocal behavior and recognition of allies or kin is widespread.

Differences in these abilities among farmed and wild animals might not provide much guidance unless they are used as multiplicative weights or a confidently held strict threshold.

Is it the thought that counts?
Some animal activists argue that the immediate priority to improve the well-being of nonhuman animals should be cultivating resources and building support to ultimately have large effects, even at the expense of immediate impact. A non-human animal which is prevented from existing in a factory farm will not be able to make any contributions to further reduce factory farming, but human supporters can.

From this perspective, key metrics for interventions would include changes in human attitudes about non-human animal status and treatment, establishment of legal and political precedents, increases in membership/funding/political victories for animal welfare groups, and similar human-focused measures but not animal welfare itself.

The idea would be that getting a law passed to seriously restrict cattle farming will generate almost as much momentum for the animal liberation movement as a law to restrict chicken farming, or at least much closer than the impact on farmed animal-years might suggest.

From this point of view it might make sense to ignore impacts on wild animals while arguing against factory farming, even if the changes demanded directly cause harm, simply because the public isn't thinking about those impacts. Or it might call for efforts to build momentum for improvements aimed at future work.

Movement building clearly can be extremely important, and more influential than early direct work. However, if one tries to build a movement around transparency in giving and giving wisely, recommending things that are directly harmful by one's transparent standards (and only helpful for movement-building) seems potentially too hypocritical to avoid self-undermining the message. So I think to be transparent one still needs to evaluate and address the rest of the issues, even if movement-building is the top priority.

[1] Some advocacy groups argue that agriculture emissions make up a much larger share of global emissions, over 50%, by including carbon dioxide released from farm animal respiration but excluding the carbon dioxide removed from the atmosphere by photosynthesis in the creation of feed. This misleads as to the net effect on the atmosphere, which can be seen clearly if we consider that applying the same analytical frame to wild animals would lead to the conclusion that they were responsible for far more than 100% of net carbon emissions.


Brian Tomasik said...

Great post!

Ants (and probably some other insects) have higher brain-to-body-mass ratios than most vertebrates, including humans, so even if one just counts neurons, farming insects plausibly involves more farmed-animal neurons per kilogram of food than farming cows and chickens. (Of course, this comparison ignores differences in life span, life quality, painfulness of slaughter, etc.) Since I care more about insects than their proportional number of neurons due in part to the "organism holism" considerations that you mention, I think entomophagy is probably worse in aggregate than farming of bigger animals for the farmed animals themselves.

One risk of ignoring the wild-animal question and promoting veg*ism for memetic reasons is that doing so may inhibit future concern for wild animals. In a similar way as omnivores sometimes downplay animal suffering to justify meat-eating, veg*ans might downplay wild-animal issues to justify (or at least quell doubts about the sign of) herbivorism. That said, I maintain high uncertainty about the sign of veg*ism for wild animals. Corn/grass/etc. crops fed to animals seem to involve relatively high amounts of biomass produced per hectare relative to other crop types, and more biomass generally means more wild-animal (especially insect) suffering.

Douglas Knight said...

Why is “factory” in the title?

You are comparing greenhouse gas emissions to suffering. People usually focus on factory farming because of suffering. But if you are looking at emissions, why are you focusing on factory farms? I doubt that any of the statistics you quote disaggregate them.

In fact, I have heard the suggestion that the efficiency optimization that leads to factory farms also reduces methane. In particular, it seems pretty plausible that speeding maturation reduces methane per calorie. You mention that grass-fed cattle produce more methane than corn-fed cattle. Choice of feed is not quite the same as factory farming, but it is suggestive. I don’t know if this is a separate phenomenon or an instance of speed of growth.

Carl said...

The title was was inspired by OpenPhil's new program in factory farming.

Yes, factory farming produces seems to produce less in the way of emissions than other ruminant farming systems per kg of output, e.g. emissions are much higher in traditional African cattle raising systems.

Anonymous said...

Is the photosynthesis density on the farmland used to grow feed for farmed animals significantly greater than whatever the photosynthesis density of that land would be if it was not used to grow feed for farmed animals?

Carl said...


That's complex, as well as the allocation of biomass produced between animals. Wikipedia claims that agriculture reduces net primary productivity:

"Extensive human land use results in various levels of impact on actual NPP (NPPact). In some regions, such as the Nile valley, irrigation has resulted in a considerable increase in primary production. However, these regions are exceptions to the rule, and in general there is a NPP reduction due to land changes (ΔNPPLC) of 9.6% across global land-mass.[23] In addition to this, end consumption by people raises the total human appropriation of net primary production (HANPP)[24] to 23.8% of potential vegetation (NPP0).[23] It is estimated that, in 2000, 34% of the Earth's ice-free land area (12% cropland; 22% pasture) was devoted to human agriculture.[25] This disproportionate amount reduces the energy available to other species, having a marked impact on biodiversity, flows of carbon, water and energy, and ecosystem services,[23] and scientists have questioned how large this fraction can be before these services begin to break down.[26] Reductions in NPP are also expected in the ocean as a result of ongoing climate change, potentially impacting marine ecosystems and goods and services that the oceans provide [5]"

Hedonic Treader said...

One point to consider is long-term supply elasticity of substitutes of farmland, e.g. hydroponics for animal feed and vertical animal farms. The energy requirements are not limited by direct sunlight if other sources of energy are available, e.g. nuclear.

Julian Simon ("The Ultimate Resource") pictures an innovation dynamic from increased demand to innovation incentives to higher supply to even lower prices.

(Full text of his book:

Now, if this innovation takes the form of substitutes like better meat alternatives or cultured meat, it would not increase animal suffering. But otherwise, the WAS displacement argument against veg*an consumption is weakened if he is right.

Howie said...

[Apologies in advance about length]

Hey Carl,

Liked this post but I thought your conclusion about focusing on movement building in "Is it the thought that counts?" was a bit too quickly dismissive.

You write:

"Some animal activists argue that the immediate priority to improve the well-being of nonhuman animals should be cultivating resources and building support to ultimately have large effects, even at the expense of immediate impact. . . .

[From this perspective you might focus on cows over chickens and ] it might make sense to ignore impacts on wild animals while arguing against factory farming. . .

However, this seems too hypocritical to be a primary metric for transparent giving and activism."

I agree that this strategy does place some constraints on extreme transparency but I think all activism has such constraints. So, unless you're never going to fund activism (or the word "transparent" is doing a lot of work here and you're going to do a follow up post recommending prioritization w/in nontransparent animal activism) I think it's worth looking at how big of a problem exactly the hypocrisy is here.

I'm not sure I have a great model of why this strikes you as so hypocritical so let me know if I end up neglecting your "true objection." You use two hypothetical examples where a movement-building-focused metric might lead to hypocrisy:

1) Treating the beef industry as though it is as big of a priority as the chicken industry;
2) Focusing on farmed animals instead of wild animals (and, in particular, insects).

There are versions of these tactics that would strike me as too hypocritical. My intuition is that the main determinant is the extent to which an activist is very explicitly arguing that their current target is "more important" than the alternative and the extent to which the second type of animal is actually being harmed by the campaign. So, let's say a group were to run a campaign to restrict cattle farming and explicitly call for sympathizers to switch to boycott beef, arguing that it will be easy because you can switch to chicken. If the primary goal of the campaign was to generate momentum so chicken agriculture could be attacked in the future, I'd agree that this was too hypocritical.

The more likely scenario, though, is that activists are always opportunistic. The fact that an activist is currently working on X campaign shouldn't be read as a signal that the activist believes X is the most important possible win. We expect activists to be incorporating tractability and momentum-building into their calculations so I don't think an activist who believes that insects are a more "important" priority is necessarily signaling otherwise by working on farmed animals. Even if asked directly why she is focusing on farmed animals over insects, I think she'd have an array of non-hypocritical responses available: "I think we ought to focus on expanding society's circle of caring gradually" seems like a fine response.

A possible objection would be that the activist is a hypocrite because she is pursuing a goal whose indirect effects (on insects) make the goal net negative. I think this charge is largely alleviated by an argument you've made elsewhere about deemphasizing second order effects. If refraining from advocating against factory farming is not among the most effect ways to reduce wild animal suffering then it doesn't seem hypocritical to advocate against factory farming (perhaps w/ offsets on behalf of insects).

Howie said...


One interesting complication is that an activist's hypocrisy seems to be sensitive to the level at which one's moral standards, beliefs, and actions are specified. Let's define hypocrisy as "the practice of claiming to have moral standards or beliefs to which one's own behavior does not conform" and let's hypothesize that veganism increases animal suffering on net in the short-run. A vegan who cares about wild animal suffering would be a hypocrite if her goal was "choose a diet with the smallest direct impact on animal suffering" but not if her goal was "choose a diet that pushes the boundaries of current circles of caring." Again, there's a certain lack of transparency here but I'm not convinced it's much greater than standard activism/persuasion. It's also very possible that such an activist is wrong on the merits but that's another discussion.

This issue reminds me of the debate over campaigns against the Keystone Pipeline - which activists tended to approach from a movement building perspective while policy analysts criticized campaigns because the pipeline itself was (arguably) not that big of a deal and ad hoc prevention of building pipelines is clearly not optimal policy. I'm not informed enough to have a net take on the Keystone issue but it at least doesn't strike me that activists were obviously too hypocritical.

All that said, this mode of activism does have some major weaknesses (which may have been what you were referring to w/r/t transparency). In discussions about effectiveness, I think there's a major temptation for activists to engage in a bait-and-switch. One dynamic that seemed to occur at times in the Keystone debate was:

*Activist campaign: "We need to shut down Keystone b/c of its contribution to climate change."
*Analyst: "Shutting down Keystone is bad policy for reasons Y and Z. Policy X would be better."
*Activist: "Please stop undermining the campaign. We know Keystone wouldn't directly contribute to climate change that much and the ultimate solution is a carbon tax (or whatever) but we need high resonance issues to organize around on the way. We need to be opportunistic and take the issues that come to us."

The activist doesn't necessarily strike me as unreasonable in this scenario but the danger is that it can be hard for the movement to receive feedback on its policy analysis and the tradeoffs its making if these are often being subordinated to opportunism.

Carl said...

"(or the word "transparent" is doing a lot of work here "

Basically this. I edited that section a bit in hopes of being a bit clearer.