Accuracy issues in FAO animal numbers

Some strategic decisions in animal advocacy are informed by the numbers of animals in various countries. The most widely used statistics about the numbers of farmed animals come from the Food and Agriculture Organization of the United Nations (FAO) website FAOSTAT. In this article, I provide some examples of inconsistencies in FAOSTAT animal data, and further reasons to think that the data may sometimes be inaccurate. The main point of this article is that it’s probably worth trying to verify FAO numbers before using them to make important decisions. In the appendix, I also explain some minor caveats that should be understood when interpreting FAO numbers.

What relevant statistics FAO provides

For land animals, FAO provides these statistics for each country and each year:

• How many farm animals of each species are alive at any time
• Livestock primary (element “Producing Animals/Slaughtered”)
  • How many animals are slaughtered for meat
  • How many egg-laying and milk animals were alive on average at any time during the selected year
• Live imports and exports

Some of these statistics are visualized by Our World in Data and Faunalytics.

Statistics about fish farming and wild fish catch for each country, year, and species can be seen by downloading the FishStatJ program. Some cumulative statistics are presented and visualized in The State of World Fisheries and Aquaculture (SOFIA) 2018 report.

Known issues

FAO claims that “it is not possible to assess the overall accuracy of the dataset, as the source data is largely collected by member countries.”[1] I’d guess that some countries don’t have good statistics and that there is significant activity that is not captured by government statistics.

For example, page 93 in FAO’s SOFIA 2018 report claims that “it is recognized that the FAO capture database does not include all fish caught in the wild, as it omits the portion of the catch that is discarded at sea and catches from illegal, unreported or unregulated (IUU) fisheries.” Pauly and Zeller (2016) estimate that because of this issue, global catches between 1950 and 2010 were roughly 50% higher than data reported to FAO suggest. Since websites like fishcount.org.uk use FAO data for their estimates, they are probably underestimating wild-caught fish numbers as well.

Another potential problem is that country officials may not always have incentives to report accurate data. For example, according to Yu and Abler (2014), China has been overreporting pork production because local officials have been inflating production figures to improve their prospects for promotion. However, Xiao et al. (2015) claims that over the last decades, the accuracy of China’s meat production statistics have been significantly improved as the incentives to over-report agricultural production have gradually disappeared. I’m unsure if there are similar problems in other countries. Note that when mistakes or misreporting is noticed, in some cases FAO may correct its statistics. E.g., see FAO (2001).

Finally, FAO data for some animals may be incomplete. For example, according to Waldhorn (forthcoming), FAO does not include snail production figures for any European country, despite other sources indicating that snails are farmed in the continent. Similarly, FAO data on rodent farming only covers Peru and Bolivia, even though rodents are also farmed in Africa (see Maass et al. (2014)).[2]

Examples of inconsistencies

When analyzing FAO statistics for land animals, I noticed some inconsistencies. Hence, I started comparing FAO statistics with official statistics from various countries to understand what is going on and noticed further potential issues. I reported some of these issues on the FAOSTAT feedback forum and they promised to revise the data at the end of this year. Nevertheless, there are probably many other issues with the data that will remain uncorrected. I wanted to document my reasoning to showcase what kind of problems can be expected. Note that the origin of these issues is unclear. They could be due to misreporting by countries, misestimating by FAO, or miscommunication between countries and FAO. For brevity, I use ‘_M’_ for a million and ‘_B’_ for a billion.

Egg-laying hens

Bangladesh

According to FAO statistics, there are 275M chickens in Bangladesh (based on official data) and 301M egg-laying hens (FAO estimate) alive at any time. It makes no sense that there are more egg-laying hens than chickens in total because the total number of chickens should include egg-laying hens.[3]

This inconsistency may be explained by the following FAO claim:

“Estimates have been made for non-reporting countries as well as for countries reporting incomplete data. However, in certain countries, data for chickens, ducks and turkeys do not yet seem to represent the total number of these birds. Certain other countries give a single figure for all poultry; data for these countries are shown under “Chickens”.

It’s possible that the FAO estimate for egg-laying hens includes egg-laying birds other than chickens, while the estimate for chickens alive at any time does not. Another possibility is that either Bangladesh official statistics underestimate the number of chickens in the country, or FAO overestimate the number of egg-laying hens.

Other countries that according to FAO data have more egg-laying hens than chickens in general are North Korea, French Guiana, Latvia, and Iceland. For Latvia and Iceland both numbers come from the official data. For North Korea and French Guiana the number of chickens come from official data, and the numbers of hens are FAO estimates. I reported these inconsistencies for all these countries to the FAO and they promised to correct it by the end of 2019.

The U.S. and the UK

FAO statistics for the U.K. seem to include not only layers, but also pullets (hens who are too young to lay eggs) and roosters (who are used for breeding hens).[4] FAO statistics for the U.S. include layers, but not pullets or roosters.[5] Pullets and roosters make up about 27% of all chickens involved in the U.S. egg industry[6] so the difference is non-negligible. It could be that FAO simply does not specify whether hen numbers should include pullets and roosters. Hence, different countries report statistics in different ways which makes comparisons problematic. I reported this issue to the FAO and they promised to revise the data by the end of 2019.

Broilers (meat chicken)

Indonesia

According to FAO, in Indonesia, in 2017 there were 2.18B chickens alive at any time (official data), 167M chickens in the egg-laying hen industry (official data), and 2.85B chickens slaughtered for meat (“FAO data based on imputation methodology”). It would follow that the average slaughter age of meat chicken is about 365 × (2.18B-167M) / 2.85B = 257 days,[7] which is much higher than any figure for broiler lifespan I’ve seen. According to Wright and Darmawan (2017), broilers in Indonesia rarely live beyond 35 days. Statistical Yearbook of Indonesia 2017 (page 280) seems to claim that there are 1.6B broilers alive at any time, which is lower than the figure I estimated from FAOSTAT (2.18B-167M = 2B) but still much higher than we should expect by looking at the slaughter total. I reported this issue to the FAO and they promised to revise the data by the end of 2019.

U.S.

According to FAOSTAT, there were 1.97B chickens alive at any time in the U.S. in 2017. Earlier in this article I explained that 505.8M of these chickens were used in the U.S. egg-laying industry. It would follow that 1.97B - 505.8M = 1.47B broilers are alive at any time. Similarly, table 30 of the 2017 U.S. Agriculture census claims that there are 1.62B broilers alive at any time. FAOSTAT and USDA stats also claim that there were 9B chickens slaughtered in the U.S. in 2017. It would then follow that the average slaughter age of broilers is 365 days × 1.47B / 9B = 59.6 days. However, according to the National Chicken Council (2019), the average broiler slaughter age in the U.S. is 47 days. According to the same source, preslaughter mortality of broilers is 5% which means that it can’t explain the 24% difference in slaughter age estimates. Broiler mothers who live longer are already excluded from the estimation, so this can’t explain it either. Live exports and imports of chickens are also not big enough to explain the difference (69M live chickens imported, and 5M live chickens exported according to FAO).

However, there is one possible explanation I’m aware of that could explain the discrepancy. Between periods of raising broilers, farms have downtime periods. These last about two weeks, during which facilities are cleaned and prepared. Farms that are in a downtime when USDA is collecting data could be reporting the number of broilers that are in farm during the broiler raising period instead of reporting that the farm has no broilers. This would inflate the number of broilers who are alive at any time, but not slaughter totals. However, I’m far from certain that this explanation is correct.

High variance in estimated slaughter age

In this section, I will show that there is a very high variance in estimated slaughter age of meat animals in various countries. This makes me suspect that there are similar inconsistencies in FAO’s data for some other countries and animals.

I made a spreadsheet in which I very roughly estimated the slaughter age of meat chickens, turkeys and pigs[8] in most of the countries. I assumed that all pigs and turkeys were raised primarily for their meat. I estimated their mean slaughter age in each country by dividing the number of animals alive in 2017 by the number of animals slaughtered in 2017. For chickens, the estimation was very similar except that I also had to estimate the number of meat chickens alive by subtracting the number of egg-laying hens from all the chickens alive in each country. All the numbers were taken from FAO.

Such an estimation gives a very imprecise measure of slaughter age for multiple reasons:

• Animals are sometimes raised in one country and slaughtered in another one.[9]

• It ignores pre-slaughter mortality.

• If the number of animals farmed in a country is growing or shrinking in a given year, it may distort the estimate.

• As explained before, egg-laying hen numbers for many countries don’t capture all the chickens used in egg-production. Hence, if we subtract FAO’s number for egg-laying hens alive from FAO’s number for all chickens alive in a given country, we may overestimate the number of broilers.

• Chicken slaughter statistics for at least some countries include slaughters of egg-laying hens (see the Appendix), which should ideally be excluded for the purposes of this estimation.

Nevertheless, if all countries submitted data to FAO in a consistent manner, I’d expect this estimated ratio of meat animals alive in 2017 to slaughtered in 2017 to be roughly similar in most countries. However, as it can be seen in Table 1, there is a very high variance in my estimated slaughter ages in various countries.

Table 1: Roughly estimated mean slaughter ages for selected countries

First 15 rows in the table are the biggest countries by chicken slaughters. The remaining countries are selected because of some particularly suspicious values.

The same species of animal is undoubtedly slaughtered at different ages in different countries but I think the differences are not nearly as drastic as the table suggests. Furthermore, I see some nonsensical values in the table. In some cases, they can be explained by live imports and exports[10] or a sudden change in the industry size.[11] However, in at least some cases they seem to indicate inconsistencies in FAO data.[12]

Conclusions

The only actionable advice here is that if you are basing important decisions (e.g. the countries in which in which countries to expand your charity) on FAO numbers, it may be worth verifying whether or not those numbers are correct. For example, it could be useful to check my spreadsheet to see if slaughter age estimates for the country of interest seem to be suspiciously high or low. If they are, it may be worth looking deeper into the statistics or asking about it on the FAOSTAT feedback forum.[13]

Appendix: caveats about FAO’s animal statistics

Some egg-laying hens are in the meat industry

FAO egg-laying hen statistics for the U.K., the U.S., and probably other countries, include broiler mothers. They are a part of the meat industry rather than the egg industry because they lay eggs that hatch into chickens, who are raised for meat. According to my extrapolations, there are 350M-600M broiler mothers in the world.[14]

Not all slaughtered chickens were raised for meat

According to FAOSTAT, there were 9,050M chickens slaughtered in the U.S. in 2017. Looking at the USDA statistics, it can be seen[15] that 135M of these slaughters were of mature chickens, who are not broilers. Most of them are “spent” egg-laying hens who are slaughtered when their rate of egg-laying slows down. Others are roosters used for chicken breeding. I’m unsure if “spent” hens and roosters are included in FAOSTAT statistics for other countries.

In the egg industry, male chicks are killed at a very young age (e.g. one day) because they can not lay eggs, and are less suitable for meat production than broiler breeds. It seems that they are excluded from the FAOSTAT chicken slaughter totals.

Animals slaughtered within the dairy industry

In the dairy industry, most male calves are slaughtered at a young age because they can’t produce milk, and older female cows are slaughtered when they no longer produce enough milk. By comparing FAO and official U.S. statistics, I can see that slaughters of both of these groups of animals are included in the FAO’s slaughter totals for the U.S.[16] I haven’t checked the numbers for other countries and other species used for milk (sheep, goats, buffaloes, camels, etc.).

References

Brienen, M., Cavenagh, B., Van Vliet, W., Copier, M. (2014). Meeting the challenge of Indonesia’s growing demand for poultry

FAO Fisheries Department (2001). Fishery Statistics: Reliability and Policy Implications

Mass, B. L., Metre, T. K., Tsongo, F., Mugisho, A. B., Kampemba, F. M., Ayagirwe, R. B. B., Azine, P. C., & Chiuri, W. L. (2014). From taboo to commodity: history and current situation of cavy culture in the Democratic Republic of the Congo. Livestock Research for Rural Development, 26, 151.

Mass, B. (2019). Africa: Why More People in Africa Should Farm Guinea Pigs for Food

Nathan Associates Inc. (2013). Indonesia’s Poultry Value: Chain Costs, Margins, Prices, and Other Issues

National Chicken Council. (2019). U.S. Broiler Performance.

Pauly, D., & Zeller, D. (2016). Catch reconstructions reveal that global marine fisheries catches are higher than reported and declining. Nature communications, 7, 10244.

Waldhorn D. (Forthcoming). Snails used for human consumption. The case of meat and slime

Wright, T., Darmawan, B. (2017). Indonesia Voluntary Poultry Report. USDA Foreign Agriculture Service, Global Agricultural Information Network.

XIAO, H. B., Qiong, C. H. E. N., WANG, J. M., & Oxley, L. (2015). The puzzle of the missing meat: Food away from home and China's meat statistics. Journal of integrative agriculture, 14(6), 1033-1044.

Yu, X., & Abler, D. (2014). Where have all the pigs gone? Inconsistencies in pork statistics in China. China Economic Review, 30, 469-484.

This essay is a project of Rethink Priorities. It was written by Saulius Šimčikas. Thanks to Cash Callaghan, David Moss, Jason Schukraft, Marcus A. Davis, Persis Eskander, and Sabrina Ahmed for reviewing drafts of this post and making valuable comments.

Notes

---

1. There is a longer explanation about how data is collected in the methodology document:

   In general, figures have been supplied by governments throught national publications and FAO questionnaires (both paper or electronic). To make the coverage of this data collection as complete as possible, official data have sometimes been supplemented with data from unofficial sources. Use has also been made of information supplied by other national or international agencies or organizations.

   ↩︎

2. Maass (2019) claims that in Africa rodents are often kept in houses or kitchens and are not included in most national statistics. Furthermore, the scale of rodent farming in Africa could be relatively small. This makes their exclusion from FAO statistics understandable. However, it still follows that FAO statistics provide an incomplete picture on the topic of rodent farming. ↩︎

3. I see that hens are included in chicken numbers for at least some countries by comparing FAO’s live animal numbers for chickens in the UK government website, Livestock numbers in UK spreadsheet, Poultry tab. For all years since 2006 (except 2009), we can get the FAO number for chickens in the U.K. by adding Hens and pullets laying eggs for eating, Breeding flock, and Table chickens (broilers) numbers from the UK government spreadsheet. For example, 158,202 thousand for 2006. For years after 2009, FAO numbers are rounded to the nearest million. The year 2009 seems to be inputted incorrectly and FAO’s number for chickens seems to include not only chickens, but also other poultry (ducks, geese, turkey, etc.). ↩︎

4. The numbers in the UK government website, Livestock numbers in UK spreadsheet, Poultry tab, Total laying and breeding fowl row seem to be very similar to numbers in FAO’s livestock primary (“United Kingdom”, “Producing Animals/Slaughtered”, “Eggs, hen, in shell”). For years until 2013 they are exactly the same. For example, 47,024,000 for 2013. From the UK government's spreadsheet it can be seen that these numbers include pullets and breeders. ↩︎

5. The layer numbers for the U.S. in the table “Annual Average Number of Layers, Eggs per Layer, and Total Egg Production” from USDA’s Chickens and Eggs 2018 Summary are identical to the ones in FAO’s livestock primary (“United States of America”, “Producing Animals/Slaughtered”, “Eggs, hen, in shell”). For example, 375,845,000 for 2017. According to my correspondence with USDA “Any layer number in the summary report does not include pullets. NASS’s definitions of the two are mutually exclusive: layers are females of any age laying marketable eggs while pullets are females who are not yet laying marketable eggs.” Since the table in the USDA summary says “layers”, we know that it and FAO’s numbers exclude pullets. ↩︎

6. According to the table 30 in the Census of Agriculture, in 2017 in the U.S. there were 368.2M layers, 130.5M pullets, and 7M roosters. So the total number of chickens involved in egg production in the U.S. is about 505.8M. (130.5M + 7M) / 505.8M = 27.2%. ↩︎

7. This might be a slight overestimate because it ignores pre-slaughter mortality. Some chickens are alive for some time but are not slaughtered because they die of disease or other causes before their slaughter age. However, the broiler pre-slaughter mortality rate in Indonesia is not nearly high enough to explain these discrepancies. According to Nathan Associates Inc. (2013), it’s 6-7 percent." A similar rate can be inferred from Brienen et al. (2014). According to FAO, in 2017, Indonesia imported 5.3M live chickens and exported 7,000. If this was taken into account, the estimated slaughter age would increase even more (although very slightly). I didn’t take this into account for the sake of simplicity. Other possible problems with this estimate are explained in the High variance in estimated slaughter age section. ↩︎

8. I didn’t include sheep in the spreadsheet because sheep raised primarily for wool are probably slaughtered when they are older than sheep raised primarily for meat. Consequently, there might be big differences in sheep slaughter age in different countries because it depends on whether sheep in that country are mostly raised for meat or wool. Unfortunately, FAO doesn’t collect statistics for how many animals are raised for wool since 2013. Similarly, I excluded cattle from the table because in many countries they are used as working animals and FAO does not collect statistics about it. ↩︎

9. Stats about live imports and exports are also provided by FAO. In total, about 1.8B chickens, 45M pigs, and 86M turkeys were transported live between countries globally in 2017. Hence, it may create significant distortions. However, I see no easy way to incorporate them in the estimations because animals are transported live to other countries not just for slaughter, but also for breeding and production. More detailed statistics would be needed to do it (e.g. like the data for the EU). ↩︎

10. E.g., according to Table 1, meat chickens are slaughtered when they are just 7 days old in Singapore, an implausible figure. Looking closer, in 2017 Singapore had 3.6M chickens alive at any time, of which 2.8M were egg-laying hens, slaughtered 46.8M chickens, and imported 46.8M chickens alive. The most plausible explanation is that chickens are imported to Singapore just to be slaughtered, hence there is no inconsistency. ↩︎

11. E.g., according to my estimation, pigs in Sweden are slaughtered when they are just 20 days old. This is because according to FAO statistics, in 2017 Sweden slaughtered 2.6M pigs but had only 138K pigs alive at any time on average. However, in 2016 Sweden had 1.5M pigs alive at any time on average. If many of these pigs were slaughtered early in 2017 and not replaced with new pigs, it could at least partly explain the discrepancy (live imports and exports of pigs were very low in both years). ↩︎

12. E.g., according to Table 1, meat chickens are slaughtered in Ethiopia when they are almost 3 years old. According to FAO, in 2017 in Ethiopia, 17.1M chickens were slaughtered for meat and there were 59.5M meat chickens alive at any time of which 11M were egg-laying hens, 0.4M chickens were imported live and there is no entry for chicken exports. Numbers for previous years are quite similar so that cannot explain the inconsistency. Hence, it could be that at least one of these statistics is inaccurate. Another example: according to Table 1, turkeys in Denmark are slaughtered when they are more than 300 years old. Looking closer, according to FAO, in 2017 there were 304,000 turkeys in Denmark at any time, 1,000 turkeys slaughtered, 2.7M live turkeys imported and 872,000 live turkeys exported. I can’t make sense of these numbers, hence it’s possible that at least one of them is inaccurate. ↩︎

13. Note that when you ask on the FAOSTAT feedback forum, questions need to be approved which can take weeks. Furthermore, all questions in the forum are regularly deleted for some reason. Hence, after posting a question it is advisable to check if it’s answered periodically. ↩︎

14. I haven’t found a dependable estimate of how many broiler mothers there are in the world. I tried to extrapolate based on statistics for various years and regions (U.S., UK, EU, South Africa, Brazil, Pakistan). Extrapolations assumed that the ratio between chicken slaughters and broiler breeders alive at any time is similar in different countries and years. I’m unsure to what degree this assumption is correct. If broiler mothers in different countries lay a similar amount of eggs per month, these estimations should be roughly correct. They might still be a bit inaccurate because:

    • I don’t take broiler mortality rates into account and they differ by country
    • Not all chicken slaughters that are included in the FAO statistics are broilers. E.g., U.S. statistics include “spent” hens, who are slaughtered when their egg production slows down. For some countries, chicken slaughter totals also include other slaughters of other species of birds.

    ↩︎

15. U.S. statistics seem to include some of the slaughtered spent hens and none of culled day-old male chicks. USDA’s Poultry Slaughter report and FAO report the exact same number of chickens slaughtered in the U.S. in 2017: 9,050,716,000. According to USDA, 8,916,097,000 of these chicken were “young”. “Young” is defined as “Commercially grown broilers-fryers and other young immature birds such as roasters and capons.” Since culled male chicks are not mentioned, I assume that they are excluded. “Mature” chickens are defined as “fowl from breeder and market egg flocks and stags and cocks”. This seems to include spent hens from the egg-laying industry. However, according to the USDA’s report, there were 134,619,000 mature chickens slaughtered in 2017. I would have expected that there would be a higher number of egg-laying hens slaughtered because there were 375,845,000 egg-laying hens in the U.S. on average and they live 1.5 to 2 years. This makes me think that not all spent hens were included. Note that all of this is about the U.S., and other countries could be reporting the data differently. ↩︎

16. See the table “Commercial and Farm Slaughter by Species – United States: 2017 and 2018” in Livestock Slaughter 2018 Summary. If we add 32,280.6 thousand cattle slaughtered in 2017 and 536.8 calves, we get that 32,817.4 thousand cattle were slaughtered in 2017 in the U.S. in total. That is the exact figure we see in FAO’s Livestock Primary statistics if we choose “United States of America”, “Producing Animals/Slaughtered”, “Meat, Cattle”, and “2017”. From the table Federally Inspected Slaughter and Percent by Classification and Month – United States: 2018 and 2017 Total of Livestock Slaughter 2018 Summary we can see that dairy cows are included in the statistics. ↩︎