NewsletterOther FoodsRed Meat

Estimating the environmental impacts of 57,000 ‘foods’

Executive summary

* This week’s paper is one of the worst I have reviewed.

* It described a modelling exercise, which was undertaken to estimate the environmental impact of 57,000 ‘foods.’

* There are a few hundred real foods, so this exercise was primarily estimating the impact of processed foods.

* The four measures estimated were: greenhouse gas emissions; land use; water stress; and eutrophication potential. Travel miles, processing, packaging, factory building, worker emissions etc., were not taken into account.

* Cattle (beef) was assumed to be the worst food by such a margin that virtually nothing else mattered. The best foods were sugary drinks, chips, onion rings and teacakes.

* The researchers had the gall to claim that the more environmentally sustainable foods tended to be the more nutritious.

* This model used another model ( to make its claims. This is increasingly happening. Models with no transparency or reasonableness of assumptions are being used in models with no transparency or reasonableness of assumptions. A body of literature is being developed to condemn animal foods. It has no substance and will do great harm.


Many thanks to Nick from Switzerland for this week’s paper. I’m not sure if ‘thanks’ is the right word, as this is up there with one of the worst papers that I’ve ever had to spend a week reviewing.

The paper was called “Estimating the environmental impacts of 57,000 food products” and it was written by Clark et al (Ref 1). I spotted a vegan in the author list (Marco Springmann) and there may be others.

As the title suggests, the aim of this study was to estimate the environmental impact of 57,000 food products. First, there are not 57,000 foods. There may be 57,000 varieties of processed food items, but the basic foods are meat, fish, eggs, dairy products, grains, legumes, nuts, seeds, fruits and vegetables. There are a number of different types of each of these, but real foods will number a few hundred at most, not 57,000.

The rationale for the study was given as “While previous analyses compared the impacts of food commodities such as fruits, wheat, and beef, most food products contain numerous ingredients.” This confirmed that the study looked at mostly processed foods and not what people should be eating – real food.

The study

The study was a modelling study and the key word in the title was “estimating”. The paper used the word “estimated” 85 times, the word “estimates” 19 times, the word “estimate” 14 times and the word “estimating” 7 times – all in a 12 page paper.

This was not a trial. It was not a study of populations. It was a modelling exercise, and most models are wildly inaccurate. Modelling studies are only as good as their assumptions. Do your retirement spreadsheet with an assumption of 1% inflation and then repeat it with an assumption of 10% inflation and see how that goes.

This study set out to “derive estimates of a food product’s environmental impact across four indicators: greenhouse gas emissions, land use, water stress, and eutrophication potential.” (Eutrophication potential is defined as the potential to cause over-fertilisation of water and soil, which can result in increased growth of biomass.)

The unique contribution of this study was that the researchers developed a way to model fake foods. The paper reported that composition information was only available for approximately 3% of 57,185 food products. The paper went into some detail about how the food label ingredients list could be broken down to estimate the individual ingredients in each product. Food labels list ingredients from the largest first and the researchers used this fact to estimate the component parts of each fake food. The paper gave an example of a green pesto product, where the label specified that 47% of the product was basil, then sunflower oil (no percentage given), then cheese (5%), then yogurt (no percentage given), then cashew nuts (5%). Other ingredients included sugar, bamboo fibre, salt, acidity regulator and garlic powder. The researchers used this to estimate that sunflower oil was 24%, yoghurt was 5%, sugar was 3.7% and so on.

A massive database was built, breaking down 57,000 products in this way. Then each basic ingredient (basil, sunflower oil, salt, sugar etc.,) was assumed to have an environmental impact when scored against the four measures that the researchers decided were the relevant ones: greenhouse gas emissions, land use, water stress, and eutrophication potential.

The environmental scores came from a website called Take a quick look and you can see diagrams showing the environmental impact of different foods. I’ve screen grabbed the image on the home page (below). There’s an overall global warming potential score (GWP100). There’s a eutrophication score, an acidification score, a water use score and a biodiversity loss score. As you can see, at a glance, cattle (beef) is truly evil and everything else pales into insignificance compared to this.

There is no explanation on the site for how these scores were derived. There is no explanation in the Clark et al paper for how these scores were derived. They just were, so don’t ask questions.

The methodology of the paper was thus – put 57,000 products, mostly processed foods, into a database. Estimate the different ingredients in each product from the ordered list on the food label. Use to determine the environmental impact of those separate ingredients and then come up with a score per 100g of product, where 0 is no impact and 100 is highest impact.

The final twist was that the paper reported, “The estimated environmental impact score and the estimated impact for each indicator are right skewed. The median average estimated environmental impact score is 1.6, the 75th percentile score is 4.1, and the 95th percentile score is 14.1, with a similar skew observed for each environmental indicator.” Note the word “estimated” three times in those two sentences alone. Median is the middle of an ordered list. In lay terms, this sentence means, when we put all products in an ordered list, the middle one scored 1.6 (which is really low on a scale of 1 to 100). The product sitting in 75th place out of all of them scored 4.1 and the product sitting in 95th place out of all of them scored 14.1. The scores to the right of the median (1.6) were therefore much higher than the scores to the left of the median.


The findings were presented in a number of different diagrams in the paper. One of the clearest summaries was Figure 3. I’ve copied it below.

You can see the food items judged ‘best’ from Figure 3. These are the ones closest to the horizontal line i.e., lowest down on the Y axis. The ‘best’ items are sports and energy drinks, fizzy drinks and cola, kids and lunchbox drinks, squashes and cordials, roasted potatoes, chips, onion rings and rice, teacakes, fruit loaves and scones, table sauces, sugar and sweeteners. The ‘very good’ products are breads, muffins, sugar cereals, crisps, snacks and popcorn, desserts, cakes, sweets, chewing gum, biscuits, doughnuts, cookies, pizza etc. These are very close to the horizontal line. The worst products by a margin are the ones grazing in the fields from Britain to New Zealand – beef and lamb. You could not make this up.

The abstract reported “Using the approach on 57,000 products in the United Kingdom and Ireland shows food types have low (e.g., sugary beverages, fruits, breads), to intermediate (e.g., many desserts, pastries), to high environmental impacts (e.g., meat, fish, cheese).”

You would think that the discovery that Coca-Cola is best and Welsh lamb is worst would send researchers back to the drawing board. But no, they doubled down. They claimed that matching these findings to the nutritional value of products revealed “Pairing it with a measure of nutrition shows a tendency for more nutritious foods to be more environmentally sustainable…”

I’m all out of words.

What they didn’t do

They didn’t take into account where products were made or travel miles. “We did not identify ingredient sourcing, such as country of origin, as this was not available for most products, and this is needed to fully understand the impacts of different foods.” So, cows grazing in the field next to me (who end up in the farm shop down the road) score far worse than lentils grown in Canada and shipped across the Atlantic.

The researchers confirmed, “The estimated environmental impacts account for the processing and transportation of commodities to retail stores, but do not incorporate postproduction processing, packaging, and transportation of, for example, converting sugar into a sugar-sweetened beverage or flour and butter into a croissant.”

The rationale for this was given as “This is unlikely to have a large influence on the estimated environmental impact scores as the large majority of food-related environmental impacts result from agricultural production.” i.e., cattle is so far and away the issue, we don’t really need to look at other products in detail.

We’ll see something else they didn’t do in the next section.

Cattle & the environment

I asked one of my favourite farmers (I know a few!) for a picture of an animal grazing on land, which couldn’t be used for other food production. North Wales farmer, Gareth Wyn Jones, replied with the images below. One of Ermentrude the cow and one of Dolly the sheep. Gareth tweets as @1GarethWynJones and is well worth a follow for those of you on twitter.

If you go back to the home page and look at the image I extracted from it, you can click on the different tabs to see how the overall claim for GWP was comprised for cattle. No claims whatsoever were made for cattle and biodiversity loss. Interestingly, the worst foods for that were peaches, wheat and avocado. Barely any impact was claimed for cattle and water use. Again, the worst food for that was wheat. Cattle was slammed for eutrophication potential in that graphic. I assume that it was also slammed for greenhouse gas emissions (cows burp) and land use. Zero credit was given for the rejuvenation of topsoil, which is a unique environmental attribute of ruminants and without which the planet cannot feed itself naturally.

The eutrophication claim against cattle is that the nutrients from livestock manure accumulate on agricultural land and the excess runs into streams causing eutrophication. Researching cattle and eutrophication confirms one’s suspicions. The word “intensive” crops up consistently, which is ‘over-farming’, not natural farming. The opening sentence of this paper states, “Nitrogen (N) and phosphorus (P) are fundamental nutrients in dairy cattle nutrition, but the majority of N and P fed to dairy cattle is excreted in urine and faeces, which may cause water eutrophication” (Ref 2). It has thus been assumed that cattle are being fed chemically to enhance the output (meat and dairy) from them. This is always the case when meat and dairy are condemned. Real farming, as undertaken by Gareth, is never compared. It’s always intensive farming and then cattle are condemned for the chemicals and the feed, which is unnecessary if only cows and sheep were left to graze the land. Cattle also gets penalised for land use – again – not taking into account the hillsides where Gareth farms, which produce food where no other food could be produced.

Turn to the best ‘food’ – sugary drinks. Hilarious! Given that comparisons were made per 100g (weight), output from a factory will be multiples higher than output from a grazing animal. According to Coca-Cola, 1.7 billion products are consumed daily (Ref 3). Hence, we can assume that factories are churning out products at a rate that makes a cow’s output look pitiful. Factory waste into rivers has probably not been considered as eutrophication. Land use will be small for the number of 100gs pumped out. The water use factor had a caveat of scarce water and so a factory built where water is plentiful won’t score on this measure. It would not surprise me if greenhouse gas emissions were zero (or even negative), as they might try to claim that CO2 (which is what makes drinks fizzy) is sequestered in the process. Only to be burped out by humans once consumed but let not the facts get in the way of a good story. The following will not have been taken into account: the building of the factory; the fuel used to power the factory; the daily emissions created by many people travelling to the factory for work etc. And, as was admitted above, the estimates did not “incorporate postproduction processing, packaging, and transportation of, for example, converting sugar into a sugar-sweetened beverage.”


In summary, this team of ‘researchers’ would have us believe that the worst food on earth is beef jerky and the best are sugary drinks. In their own words, “In this scoring system, many of the highest impact products (e.g., having a score close to 100) were dried beef products such as biltong and beef jerky, which contain more than 100 g of uncooked beef per 100 g of final product, while many of the lowest impact products were composed mainly of water, such as sugary drinks.”

Papers such as this are dangerous, and they are becoming more frequent. A team of plant-biased researchers does a modelling exercise. Assumptions are made, which are rarely shared, let alone examinable. These papers make their way into the academic literature and then become ‘fact’. Other modelling papers use these previous models for their own assumptions and a house of cards is built.

We are rapidly developing a body of literature which is being taken as undeniable evidence that animal foods are bad for the environment (we already have a mass of epidemiological literature claiming that animal foods are bad for human health). The opposite is true. The most nutritious foods are meat, fish, eggs and dairy – not sugary drinks, chips, onion rings and teacakes, for goodness’ sake. The only foods that give back to the planet are ruminants. All others take nutrients from the land. The truth is being turned upside down because there is a global agenda to force a plant-based diet on the population. Shame on all ‘researchers’ taking part in this delusion.

As Nick spotted, the normally sensible Marion Nestle must have lost her mind in her blog about this paper (Ref 4). She opened by saying “If I wore a hat it would be off to the authors of this astonishing paper.” I can think of something to do with this paper and it’s not tipping my hat to it! ;-)


Ref 1: Clark et al. Estimating the environmental impacts of 57,000 food products. PNAS. June 2022.
Ref 2: Biagini & Lazzaroni. Eutrophication risk arising from intensive dairy cattle rearing systems and assessment of the potential effect of mitigation strategies. Agriculture, Ecosystems & Environment. 2018.
Ref 3:
Ref 4:

9 thoughts on “Estimating the environmental impacts of 57,000 ‘foods’

  • Springmann, Scarborough, Harrington and Clark, + several others are also responsible for this article
    which I suspect is equally appalling. Even worse, it is the basis for an article on the BBC Good Food
    which swallows the message whole and, inter alia, claims that a meat diet is sadly lacking in nutrients.

  • Wow, just wow. I don’t know whether to laugh or cry when I see things like this. The thing that worries me most is that this will likely be cited over an over again, particularly on blogs, social media, and the next generation of documentary* films geared towards vegans, such as The Game Changers and Cowspiracy.

    *Calling these documentaries is a rather sick joke, as they are overwhelmingly nothing more more than propaganda pieces filled to the brim with disinformation. Sadly, such films have an impact. The truth, facts, evidence, and cogent analysis/dissection of studies (which Zoë does so brilliantly) often get drowned out by the unrelenting onslaught of garbage like this.

  • Excellent review of an appalling paper. While reading it I was baking a gluten and sugar free cake containing 6 eggs, 250g of butter and various nut meals in the rice cooking (powered by our PV system). Oh no, home baking is the worst of “Kitchen Accessories”. Anyone with any reasoning ability would not believe that it would be better for the environment if I purchased Danish pastries, which funnily are actually made in Denmark (I’m in Australia), from the supermarket.

  • Please Zoë, keep shining your statistical intelligence on these types of ‘research papers’, so we are not bamboozled by an overwhelming amount of very shaky ‘studies’ seemingly demonstrating that plant-based highly processed foods are so much better to consume than old-fashioned, real, unprocessed foods from both animal and plant sources…

  • 57,000 different foods seems like a reference to Heinz and their 57 varieties. And this paper is going to bring joy to multinational food processing companies. Time for the sane majority to shout louder.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.