Analysis: Which animal products cause the most animal suffering and deaths in the United States?

post by Casey Riordan · 2020-09-16T18:38:52.146Z · EA · GW · 8 comments


  The Results
      Total Weight Of Product Consumed Per Day
      Product Per Life
      Loss-Adjusted Edible Proportion 
      Elasticity Factor
      Indirect Impact Factor
      USDA Weighting Factor
      Weight Per Individual Serving


Faunalytics has released a new set of Animal Product Impact Scales, identifying which animal product formats (out of 98 categories) cause the most days of farmed animal suffering and cost the most animal lives to feed the United States.

This analysis takes into account both animals who are directly consumed and those who are fed to another farmed animal, die before they can produce a consumable product, or are slaughtered because they are considered “useless” to the agriculture industry.

The Results

We created the following scales, ranking each animal product by its impact on animals in terms of total U.S. daily consumption. Full methodological details are presented in the subsequent section.

1) The products that cause the most days of animal suffering (first image) and animal deaths (second image) to feed the U.S. population per day. Top 10s pictured; full lists in this spreadsheet.

2) The products per category (i.e. chicken products, fish products, etc.) that cause the most days of animal suffering (first image) and animal deaths (second image) to feed the U.S. population per day. Top 3s pictured; full lists in this spreadsheet.

3) The most impactful products for consumers to reduce animal suffering (first image) and lives taken (second image) on a per-serving basis. Top 10s pictured; full lists in this spreadsheet.

A reminder that each of these scales is also available at


Author: Ali Ladak, MA

This is a copy of our methodology document, which is also available on the OSF.

This section outlines Faunalytics’ methodology for estimating the number of animal lives that would be saved and the days of suffering that would be avoided if various animal food products were replaced. This was estimated for two different use cases:

  1. If individual animal food products were replaced in the U.S. for one day, and
  2. If an individual replaced a single serving of various animal food products in their diet with non-animal alternatives.

Data from the National Health and Nutrition Examination Survey (NHANES) were used to create total daily consumption estimates using R. Calculations estimating the impact of consumption on animals can be found in an Excel workbook entitled Animal Product Impacts - Calculations Workbook. These were incorporated with the total consumption estimates using R. All steps of the analysis can be found on the OSF.

Impact Equations

We estimated the impacts using the following equations. Equations 1 and 2 were used to estimate the U.S. total impacts of replacing the animal food products; equations 3 and 4 were used to estimate the individual impacts.

Each component used in the following equations is described below.

Equation 1: Total animal lives saved in the U.S. = [Total weight of product consumed per day / (Product per life x Loss adjusted edible proportion)] x Elasticity factor x Indirect impact factor x USDA weighting factor

Equation 2: Total days of suffering avoided in the U.S. = Total animal lives saved in the U.S. x Lifespan

Equation 3: Individual animal lives saved = [Weight per individual serving / (Product per life x Loss adjusted edible proportion)] x Elasticity factor x Indirect impact factor x USDA weighting factor

Equation 4: Individual days of animal suffering avoided = Individual animal lives saved x Lifespan

Equation Components

Total Weight Of Product Consumed Per Day

Used in Equations 1 and 2 for U.S. totals

To estimate the total weight of different animal products consumed per day, we used the 2015-16 National Health and Nutrition Examination Survey (NHANES), a nationally representative, in-person survey of over 8,000 people in the U.S. [1]. The survey includes a dietary recall interview, which asks respondents to report their food consumption over a 24 hour period [2]. It also estimates the weight of the animal product component of each food item consumed. For simplicity, we included only one animal product per food code in our estimates (e.g., the beef patty from a cheeseburger but not the cheese slice or mayonnaise). This means that our estimates from the NHANES data, if used alone, would probably slightly underestimate total consumption. However, as described in the ‘USDA Weighting Factor’ section below, we applied a correction to our overall estimates.

The NHANES data allows for very specific categorization of animal products consumed in the U.S.. We categorized respondents’ reported consumption based on the type of animal products consumed (e.g. beef, pork, dairy) and the product formats (e.g. steak, bacon, cheese). We then applied sample weights included in NHANES to ensure the estimates are representative of the U.S. population on an average day.

Product Per Life

Used in Equations 1 to 4

The product per life refers to the initial weight of product per animal life that we use in our analysis. For livestock meat products (beef, pork, chicken, turkey), eggs, and dairy, we used data from the Food and Agriculture Organisation of the United Nations (FAO) for the U.S. in 2018 [3]. For livestock meat products the data are carcass weights -- the weight after slaughter with some parts, such as the head and inedible organs, removed. For eggs and dairy, the data are the yield per animal in the year. To estimate the weights of eggs and dairy produced over the animals’ whole lives, we multiplied these yields by how long (in years) cows and hens typically produce these products [4] [5].

For fish and shellfish we estimated the weighted average weights of the species most commonly consumed in the U.S. using two key sources of data: (1) The per-person edible weights of the most commonly consumed seafood in the U.S. estimated by the National Fisheries Institute [6]; and (2) harvest weights of each of these most commonly consumed fish and shellfish from Fishcount [7]. We applied the proportion of each product consumed to the harvest weights to estimate weighted average weights for fish and shellfish. Fishcount does not estimate weights of wild-caught fish, so we estimated a weighted average for crabs using several alternative sources, as most crabs in the U.S. are wild-caught [8] [9] [10] [11] [12].

Loss-Adjusted Edible Proportion

Used in Equations 1 to 4

There are several forms of losses in the weight of an animal from the farm to the point of consumption. These include removal of inedible parts of the animal, loss at the retail and consumer levels due to waste or damaged products, and shrinkage during cooking. The United States Department of Agriculture (USDA) estimates these losses in its ‘Food Availability (Per Capita) Data System’ [13].

For livestock, the USDA estimates the proportions of carcass weights lost, so we used these estimates directly in our analysis. We also used the USDA estimates for eggs and dairy directly in our analysis. For eggs this should be accurate, as the USDA proportion is an estimate of the farm weight. For dairy, the estimate is a proportion of the retail weight, so our estimate for dairy products will likely be a slight underestimate.

For fish and shellfish, the USDA estimates are proportions of the edible weight of the products. Because our estimates for fish and shellfish start at harvest weights, we applied an additional adjustment for the edible proportions of fish and shellfish using FAO data [14], in addition to the loss adjusted proportions estimated by USDA.

Elasticity Factor

Used in Equations 1 to 4

If an individual (or group of individuals) reduces their consumption of an animal product by a certain amount, this will likely result in a less than proportional reduction in society-wide consumption of that product. This is because the lower demand for the animal product will likely reduce the market price of the product, which will, in turn, slightly increase the quantity demanded by consumers. The size of this effect depends on the price elasticity of demand and supply for the product; that is, how responsive demand and supply are to changes in the price of the product.

We made adjustments to our estimates to account for this effect. We followed the approach of Norwood and Lusk (2011), which uses the following equation to estimate the net impact on quantity consumed of a unit change in consumption [15]:

Equation 5: Impact of a unit change in consumption = Price elasticity of supply / (Price elasticity of supply - Price elasticity of demand)

Full details of how this equation is derived can be found in the appendix of chapter 8 of Norwood and Lusk (2011), and also in chapter 3 of ‘Agricultural Marketing and Price Analysis’ [16].

We used estimates directly from Norwood and Lusk (2011) for impacts of reducing consumption of beef, pork, chicken, dairy, and eggs. The book does not contain estimates for turkey, finfish, and shellfish. We therefore estimated these as consistently as possible with the method of Norwood and Lusk. We used demand elasticities for turkey from Huang (1986) and fish from Huang and Lin (2000); the same papers used for the demand elasticities in Norwood and Lusk [17] [18]. For turkey, we estimated the same supply elasticity as used in Norwood and Lusk for chicken. For fish, data on long run supply elasticities are quite scarce. We used estimates of the long run supply elasticity for wild caught fish from Pascoe and Mardle (1999) and aquaculture from Andersen, Roll and Tveteras (2008) [19] [20]. Due to a lack of data about price elasticities for shellfish, we applied the overall elasticity estimate from finfish.

Indirect Impact Factor

Used in Equations 1 to 4

We applied adjustments to account for the fact that consuming the various animal products also has indirect effects on animal lives. We aimed to account for the largest impacts: pre-production mortality of farmed animals, male chicks in the egg industry, and feed fish used in the diets of farmed animals.**

**Note: We did not include bycatch of wild-caught fish, as some bycatch is used for feed fish [21], which could lead to double-counting in our estimates, and while a large proportion of bycatch is discarded back to sea, some of these discards survive [22]. Reliable estimates of the breakdown were unavailable. We also did not account for breeder animals to avoid double-counting, as the meat of these animals are sometimes used for human consumption (e.g. [23]) and reliable estimates were not available. With these exclusions, our indirect impact estimates are likely to be somewhat conservative.

To account for the additional number of lives lost due to pre-production mortality of farmed animals, we used estimates of mortality rates largely from USDA sources [24] [25] [26] [27] [28]. For fish, we estimated the weighted average mortality rate of the species consumed in the U.S. from Open Philanthropy data [29]. For shellfish, we use the average of the mortality rates in the six largest producing countries in the world [30].

Our favored approach to estimate the number of additional days of suffering due to pre-production mortality was to use data on mortality at different phases of the animals’ life cycles. For example, for pigs, the data suggests a 10% mortality rate at the preweaning stage, 4% at the nursery stage, and 4% at the grower to finisher stage [26]. Based on an average lifespan of 168 days [31], we estimated the average age of pre-production mortality to be 33 days. We used the same general approach for other livestock, though in some cases we relied on weaker mortality data. For fish and shellfish, data on the average age of mortality are particularly weak, so to estimate the additional days of suffering we assumed the point at which fish and shellfish die prematurely in proportion to their expected lifespan is the same as the average across the livestock species in our analysis.

To account for the indirect impact on male chicks, we assumed one male chick life is lost for each layer hen. This is in line with estimates of the number of layer hens and male chicks killed in the egg industry every year, at around 7 billion each [3] [32]. We assumed one additional day of suffering is caused per layer hen, as the male chicks are typically one day old when they are killed [33].

To account for feed fish used in the diets of aquacultured fish and shellfish, pigs, and chickens, we first applied global estimates of the proportion of feed fish per species to Fishcount data on the total number of feed fish caught globally [34] [35]. We then converted these estimates to the number of feed fish per individual animal based on estimates of the number of farmed fish, shellfish, pigs, and chickens consumed globally every year [3] [36] [37]. For farmed fish and shellfish we applied an additional weight to account for the difference between the average weight of fish and shellfish in the global Fishcount data and the average weight of those consumed in the U.S.

Since feed fish are generally wild caught [34], we assumed that each fish used causes the equivalent of one life lost and one additional day of suffering (see ‘Lifespan’ section for more details on this assumption).

In the final infographics we present the impacts of the various animal food products including the indirect impacts. However, the totals with indirect impacts excluded are available in the full data spreadsheet.

USDA Weighting Factor

Used in Equations 1 to 4

We also estimated the overall number of lives and the days of suffering for each animal based on the USDA data on food availability, which is often taken as a proxy for total consumption in the U.S. [13]. Summing the number of lives and days of suffering avoided for each of our animal product categories showed that our total estimates based on the NHANES data were, in most cases, slightly below our estimates based on the USDA data. This was as we had expected, given that our procedure with the NHANES data excluded some animal products.

We consider the USDA estimates to be the most reliable source of information on the total amount of consumption in the U.S. Therefore, we applied multiplicative factors to our NHANES estimates using the ratio of our estimates and the numbers expected from the USDA data.

This adjustment has the impact of increasing the number of lives saved or days of suffering avoided (where we underestimated compared to what we expected from the USDA data) by replacing a certain food product, but retaining the relative impacts of the different food products reported in NHANES.


Used in Equations 1 and 2 for days of suffering avoided

To estimate the days of suffering avoided for each animal food product, we multiplied the number of lives saved for each product by the average lifespan of each animal. For livestock, we used estimates of typical lifespans directly in our analysis from several sources [4] [5] [31] [38] [39][40].

For fish and shellfish, we estimated weighted average lifespans based on the proportion of consumption that is wild caught versus farm raised. These proportions were largely taken from Seafood Health Facts, a joint project of several U.S. universities [41]. We used lifespans for farm raised finfish from Open Philanthropy, and used an FAO estimate for farmed shellfish [29] [42]. For wild caught fish and shellfish, we assumed a “lifespan” of one day, since consumers are only responsible for the period of capture and slaughter in the animals’ lives.

Weight Per Individual Serving

Used in Equations 3 and 4 for individual impacts

The weight per individual serving refers to the weight of the animal product in an average serving of a particular food product (e.g. the weight of the meat patty in a burger). The NHANES dietary recall interview includes this data for each of the food products consumed. We estimated the average of all the servings in the dietary recall data for each of our product categories (e.g. beef steak, pork bacon, etc.) to give us an average weight per individual serving for each category. This allows individuals to directly compare products that they may eat more or less often than the general population.


We would like to extend a huge thank you to Ali Ladak for his hard work on these analyses, the Food System Research Fund for funding the project, and the many people who provided suggestions on the analyses: Tom Billington, Marco Cerqueira, and Haven King-Nobles of the Fish Welfare Initiative, Galina Hale and David Meyer of the Food System Research Fund, Lewis Bollard of Open Philanthropy, and Saulius Šimčikas of Rethink Priorities.


[1] Centers for Disease Control and Prevention. (2020, August 31). National Health and Nutrition Examination Survey.

[2] Centers for Disease Control and Prevention. (2018). National Health and Nutrition Examination Survey 2015-2016 Data Documentation, Codebook, and Frequencies

[3] Food and Agricultural Organization of the United Nations. (2020, August 26). FAOSTAT.

[4] Compassion in World Farming. (2012, September 1). The Life of Dairy Cows.

[5] Food and Agricultural Organization of the United Nations. (2003). Egg marketing–a guide for the production and sale of eggs. FAO Agricultural Services Bulletin, 150, 29-51.

[6] About Seafood. (2018, December 13). Top 10 List Shows Significant Increase in Seafood Consumption.

[7] Fishcount. (2016). Fishcount estimates of numbers of individuals killed in (FAO) reported fishery production.

[8] Seafood Health Facts. (n.d.). Crab.

[9] Sweat, D. (2017). Preparing Blue Crab: A Seafood Delicacy

[10] Siegel, J. (n.d.). Chionoecetes opilio (snow crab)

[11] Sea Grant California. (n.d.). Dungeness Crab.

[12] Alaskan King Crab Co. (2017, September 22). The Anatomy of Alaskan King Crab.

[13] United States Department of Agriculture Economic Research Service. (2020, July 24). Food Availability (Per Capita) Data System.

[14] Torry Research Station. (1989). Yield and nutritional value of the commercially more important fish species (No. 309). Food & Agriculture Org.

[15] Norwood, F. B., & Lusk, J. L. (2011). Compassion, by the pound: the economics of farm animal welfare. Oxford University Press.

[16] Norwood, F. B., & Lusk, J. L. (2018). Agricultural marketing and price analysis. Waveland Press.

[17] Huang, S. K., & Lin, B. H. (2000). Estimation of Food Demand and Nutrient Elasticities from Household Survey Data. US Dept. of Agriculture, Economic Research Service.

[18] Huang, K. S. (1985). US demand for food: A complete system of price and income effects (No. 1714). US Dept. of Agriculture, Economic Research Service.

[19] Pascoe, S., & Mardle, S. (1999). Supply response in fisheries-the North Sea. University of Portsmouth, Centre for the Economics and Management of Aquatic Resources.

[20] Andersen, T. B., Roll, K. H., & Tveteras, S. (2008). The price responsiveness of salmon supply in the short and long run. Marine Resource Economics, 23(4), 425-437.

[21] Wijkström, U. N. (2009). The use of wild fish as aquaculture feed and its effects on income and food for the poor and the undernourished. In Fish as feed inputs for aquaculture: practices, sustainability and implications (Vol. 518, pp. 371-407). FAO Rome.

[22] Roda, M. A. P., Gilman, E., Huntington, T., Kennelly, S. J., Suuronen, P., Chaloupka, M., & Medley, P. A. (2019). A third assessment of global marine fisheries discards. Food and Agriculture Organization of the United Nations.

[23] Compassion in World Farming. (2019, December 16). The Life of Broiler Chickens.

[24] United States Department of Agriculture, Center for Food Security and Public Health. (2013). Foreign Animal Disease Preparedness and Response Plan Poultry Industry Manual.

[25] United States Department of Agriculture. (2014). Layers 2013.

[26] United States Department of Agriculture. (2015). Swine 2012.

[27] United States Department of Agriculture. (2017). Death Loss in U.S. Cattle and Calves Due to Predator and Nonpredator Causes, 2015.

[28] United States Department of Agriculture. (2018). Dairy 2014.

[29] Open Philanthropy Project. (n.d.). Finfish numbers.

[30] HATCH Aquaculture Accelerator. (n.d.). HATCH Global Shrimp Farm Report.

[31] United States Department of Agriculture Economic Research Service. (2019, August 20). USDA ERS - Hogs & Pork - Sector at a Glance.

[32] Chick Culling. (2020, June 7). In Wikipedia.

[33] Krautwald-Junghanns, M. E., Cramer, K., Fischer, B., Förster, A., Galli, R., Kremer, F., Mapesa, E., Meissner, S., Preisinger, R., Preusse, G. & Schnabel, C. (2018). Current approaches to avoid the culling of day-old male chicks in the layer industry, with special reference to spectroscopic methods. Poultry science, 97(3), 749-757.

[34] European Market Observatory for Fisheries and Aquaculture Products. (2019). Case study – Fishmeal and fish oil.

[35] Mood, A., Brooke, P. (2019). Fish caught for reduction to fish oil and fishmeal.

[36] Mood, A., & Brooke, P. (2012). Estimating the number of farmed fish killed in global aquaculture each year., 1.

[37] Mood, A., Brooke, P. (2019). Numbers of farmed decapod crustaceans.

[38] United States Department of Agriculture. (n.d.). Slaughter Cattle.

[39] United States Department of Agriculture. (2019, July 17). How old are chickens used for meat?

[40] United States Department of Agriculture. (2013, August 5). Turkey from Farm to Table.

[41] Seafood Health Facts. (n.d.). Description of Top Commercial Seafood Items | Seafood Health Facts.

[42] Food and Agricultural Organization of the United Nations. (2009). Penaeus vannamei


Comments sorted by top scores.

comment by sawyer · 2020-10-21T19:06:06.266Z · EA(p) · GW(p)

Very cool analysis! I wonder how these orderings would change if the suffering levels were adjusted by brain complexity / capacity for suffering. In other words, most people think that killing or hurting a cow is worse than killing or hurting a salmon, but your analysis seems to treat them as equivalent. I realize that this would introduce a lot more uncertainty into the calculations, and that there's potential for significant moral disagreement on this topic, but  has there been any effort to do something like this with your data?

Replies from: jo
comment by jo · 2020-10-22T20:22:26.184Z · EA(p) · GW(p)

Thanks, sawyer! We haven't done that ourselves for exactly the reasons you mentioned--there were already many degrees of uncertainty in the analysis just based on spotty data availability, so we felt that adding something with that amount of subjectivity would reduce the overall utility of the analysis because some people would agree with our decisions and some wouldn't. But the lists within species may work as a simple proxy, and the data and code are available to anyone with more investment who would like to make that adjustment themselves. If you're thinking about doing it and would like to chat, feel free to reach out! It's my first name @


Replies from: sawyer
comment by sawyer · 2020-10-22T21:14:56.483Z · EA(p) · GW(p)

Totally, that makes sense. I think unfortunately this is beyond my expertise at this point, but I appreciate you offering! For anyone else particular, I'm envisioning something like the Excel tool created for this SSC adversarial collaboration, where a user can input their subjective "human life-year equivalents" for various animals, then connect that to your data and analysis, and output...something. Maybe new charts, but more realistically something more basic.

Replies from: jo
comment by jo · 2020-10-23T13:03:31.459Z · EA(p) · GW(p)

That's a good idea and actually not too hard to implement in the grand scheme of things. It's not something that will get done right away, but I can add it to the list! And if anyone reading would like to collaborate to  produce that, please get in touch!

comment by Zonulet · 2020-10-21T21:05:15.952Z · EA(p) · GW(p)

For me the most surprising result in the infographics is that raw eggs should have so much suffering per serving. I tried to dig into the data to work out why this might be, and either I'm misunderstanding it – very possible! – or this result is a bit misleading.

The full "Per-Serving Impact on Lives and Days of Suffering" data gives raw eggs a 10.3 for total suffering per serving, and boiled/fried/scrambled only around 4-5. So that means the serving size for raw eggs must be twice as big. 

But this doesn't make sense. If I want to eat three eggs in scrambled form, I simply need to eat a plate of scrambled eggs. But if I want to eat three eggs in "baked good" form, I need to eat an entire cake – which hopefully I'm not doing too often even during a pandemic.

I can't find any reference to raw eggs in baked goods in the "Merged Impact & Consumption Data" spreadsheet. However, they are mentioned in the original "2015-2016 FNDDS At A Glance - FNDDS Ingredients" spreadsheet: under various types of "Cakes and pies", this gives an ingredient weight for eggs of 150g i.e. the amount you'd put in a whole cake. Is it possible this has mistakenly filtered through as a serving size, resulting in the elevated figure for suffering per serving, when in fact 150g of raw eggs represents maybe 10 servings of cake?

Replies from: jo
comment by jo · 2020-10-22T20:48:31.403Z · EA(p) · GW(p)

I think you're right, that one does seem to be a bit misleading... thank you for calling my attention to that. It looks like an eccentricity in the NHANES data. While it has things like cakes and pies with eggs in them as one type of food respondents could report, it also allowed them to report foods in terms of their constituent ingredients. So you could report a ham and cheese sandwich as a ham and cheese sandwich OR as sliced ham, sliced cheese, mayo, and bread. Because we restricted our analysis to just the primary animal product in each report, raw eggs are a bit of an odd case...the specific foods in that category are "Egg, yolk only, raw",  "Egg, white only, raw",  and "Egg drop soup." It's actually the soup that's the problem, because the ingredient list has it categorized as its own sole ingredient (egg drop soup, made up 100% of egg drop soup!). That's definitely a problem since it ended up assigning the entire weight of a serving of soup to the egg. I think the best option is probably just to remove raw egg as a category entirely, but I'll double-check and consult the rest of the team first.

Thanks for the feedback!



comment by Equanimitivity · 2020-10-22T01:01:06.992Z · EA(p) · GW(p)

The equations used to calculate “suffering” do not seem to take into account actual harm done to animals in the process of producing these products. Rather it converts animal lives lost into lost years of an animals life.

These graphics are misleading when they refer to this as suffering, when what is meant is life-years.

I am glad, however, with the groupings by animal type. As Sawyer commented below, some people might wish to assign different moral value based on the neurobiology of the life taken. Perhaps a year of a clam’s life is not directly comparable to a year of a pig’s life.

Replies from: jo
comment by jo · 2020-10-22T20:29:57.203Z · EA(p) · GW(p)

Thanks for your comment. If you want to use the objective term, "days of life" is most accurate -- we used the number of animals affected,  and  multiplied by their lifespans, accounting for different causes of mortality. You could certainly argue we editorialized a bit by referring to any day of life for a farmed animal as a day of suffering. I agree that there are differences in the extent of that suffering between species and circumstances, but in the interest of keeping the degrees of uncertainty as low as possible (see my comment to sawyer), we chose to say a day is a day is a day. Given that the numbers are massively dominated by factory farmed chickens and aquacultured fish, I feel pretty comfortable referring to it as days of suffering.