Do My Meat Purchases Cause Additional Factory Farming?

Bryan Frances is the world’s only intellectual wisdom coach. He’s a former professor of philosophy & logic, doing research & teaching at universities in the US, UK, Europe, Asia, and the Middle East. He teaches you how to become the wisest thinker in the room—which is different from being the most knowledgeable or having the highest IQ. Contact for a free session.

I read shocking things like this from time to time, from several authors:

I think eating meat is morally much worse than being rude to a waiter—if you’re choosing whether to be rude to a thousand waiters or to eat one chicken sandwich, you should choose to be rude to the waiters.

I quoted Matthew Adelstein here, the maestro behind Bentham’s Bulldog, because it’s his posts that prompted me to (finally) write about this topic.

I could hardly disagree with him more. To be precise: I think being rude to one waiter is morally much worse, when it comes to consequences, than eating not just one but 1000 chicken sandwiches. I would have that view even if I thought that factory farming is the worst thing in the world.

But set aside my moral views. This post is NOT about ethics. It’s about causality.

Consider this remark, which is frequently cited as part of the reason for remarks like the one about waiters and chicken sandwiches:

• CAUSE: If you purchase a pound of beef, you’ll cause very roughly one extra day of factory farming.

Adelstein’s idea seems to be this:

If you insult a waiter, his feelings will be slightly hurt for a short period of time, but if you buy a pound of beef, you cause an additional day of torture for a cow (similar facts for chickens and pigs). Clearly, the second thing is way, way worse than the first thing. That’s why insulting the waiter is much, much less bad than buying the meat.

You can see the role CAUSE has in his reasoning for the waiter-chicken sandwich remark. You can find remarks similar to CAUSE in the writings of lots of ethicists.

CAUSE is an empirical claim—not an ethical one. It’s about causality, not ethics. In this post, I investigate it. Just to be absolutely, stunningly clear:

I am addressing a non-moral, causal claim. It doesn’t say anything about ethics or morality. It’s just about causes and effects.

When people read or hear CAUSE, they often think it means that if they buy meat, then their purchases will cause more factory farming to happen than if they didn’t buy that meat. In order to have a specific example to treat in detail, let’s consider a typical beef-eater and examine this CAUSE-inspired thesis:

CAUSE*. There are two possible futures in front of me. In the beef future, B, I buy a pound of beef every day on average. In the non-beef future, NB, I never buy beef again and replace it with food that doesn’t use any factory farming at all. What we want to know is the likely difference between those two future scenarios. That is, if the two scenarios B and NB proceed under the same usual assumptions (we assume that pretty much everything is the same in both scenarios with the exception of my beef purchasing decision and its consequences), but in one scenario I buy one pound of beef every day while in the other scenario I never buy any factory farming items, what are the likely differences in consequences from those two choices?

Here’s the answer: if I proceed with B instead of NB, then over a period of time of X days, there will be—as a causal result of my beef purchases—very roughly X more days of factory farming than if I did NB.

This interpretation of CAUSE is pretty natural. After all, CAUSE literally says one thing—purchasing a pound of beef—will cause another thing—one more day of factory farming. I mean, just look at CAUSE: the word “cause” is right there in it.

I think CAUSE* is obviously false. I claim: if in one scenario I buy a pound of beef per day, going forward, the odds are absolutely overwhelming that zero additional days of factory farming will happen, compared to the scenario in which I never buy any beef at all. I’ll argue for that thesis in the next section.

Unlike other philosophers who have argued for this thesis, I will try to be as realistic and detailed as possible. We must be, in order to see why the argument is so strong and certain objections to the argument fail.

We need to have a vivid sense of what our task is, the task of figuring out if my regularly buying beef will cause more factory farming. Otherwise, our reasoning will run off the rails.

As I wrote above, we have two futures to examine, NB and B. We want to figure out if B will have more factory farming than NB—as a result of my buying beef. Clearly, B could have a lot more factory farming than NB due to events that have nothing whatsoever to do with my buying beef. If the president of the USA makes beef practically free in B but not NB, then we’d expect a lot more factory farming in B than in NB. But of course that’s irrelevant.

So, we want B and NB to be as similar as possible, in order to isolate the causal consequences of my buying beef. That similarity in B and NB will come up a couple times in my argument.

Let’s try to be realistic. Accordingly, suppose that I buy beef from two grocery stores regularly. I also eat out four times a week, for breakfast, lunch, dinner, or even just a snack (buying a snack counts as “eating out” here). So, my beef purchases are spread out amongst two grocery stores, restaurants, fast-food places, cafeterias, street vendors, vending machines, and so on. Let’s say I get 2.5 lbs of beef from each grocery store, per week on average. That amounts to 5 lbs from grocery stores and 2 lbs from the other sources I listed.

Suppose I made my B/NB choice a few weeks ago. Obviously, there aren’t going to be any differences in factory farming right away. No one who is sane thinks that the instant you buy a 1-lb package of beef a cow undergoes another day of torture. That’s why we fast forward a few weeks.

Today the grocer looks at a spreadsheet of his beef sales over the last few weeks, in order to figure out what adjustments he might want to make for his future purchases. In both B and NB everyone’s beef-buying behavior is identical—except for me, since I buy beef in B but not NB. So, the beef purchases made by everyone except me are exactly the same in NB and B. Again, we’re keeping B and NB as similar as possible in order to “isolate the variable X”, where X = my beef buying habits.

Although there is a significant range, a typical mid-sized grocery store in the USA sells about 1400 lbs of beef per week. When looking at his recent sales figures, my grocer sees something like 1429 lbs/week in the B scenario and 1426.5 lbs/week in the NB scenario, since the only difference in the two comes from my beef purchases. Just imagine the scene:

• In each of B and NB he is in his office, seated at his desk, looking at the spreadsheets on his laptop. Everything in the two scenarios is exactly the same—with the exception that in NB the number in the relevant box on the spreadsheet reads “1426.5” whereas in B the number in that very same box is “1429”.

The odds that he’s going to react differently to these numbers is ridiculously small. He is exactly the same in each scenario, and he’s merely reading numbers that are ever so slightly different. 1429 is 0.18% more than 1426.5.

Just to be clear, that’s not a difference of 18%. It’s a difference of much less than 1%.

So, it’s extremely unlikely that he will react to the differing figures 1429 and 1426.5 differently, buying less beef in NB than in B.

But for the sake of really, vividly, seeing how implausible CAUSE* is, let’s assume otherwise. Let’s suppose the grocery store owner in NB looks at his spreadsheets and orders less beef from his beef distributor-warehouse compared to how much he orders in B. For some reason, he looks at the 1429/1426.5 figures and reacts differently.

He can’t call up the beef warehouse and say “Hey, starting next month I want 2.5 fewer lbs of beef per week”. According to the LLM I used, the minimal decrease in a purchase order would be 10 pounds. So, let’s assume that in both scenarios, NB and B, the grocer looks at his spreadsheets and then orders 10 fewer lbs/week in NB than in B (it won’t matter what the figure is: 10 lbs, 25 lbs, 100 lbs).

What will happen next? In particular, will there eventually be any difference in factory farming? Although we’ve assumed, against expectations, that my beef purchases have altered my grocer’s purchases from the beef warehouse, we haven’t seen anything about its effects on factory farming.

Grocery store owners get their beef delivered from a beef warehouse, which gets its inventory from a slaughterhouse. There’s a range of sizes of beef warehouses, but a typical one will distribute about 350,000 lbs of beef/week. For the sake of using realistic numbers instead of abstract reasoning, let’s go with 350,000 lbs/week.

We are currently assuming that in the NB scenario, the grocer orders 10 fewer lbs of beef from the warehouse than he orders in the B scenario. So, now we have the warehouse owner looking at his spreadsheets in both scenarios. He’s just like the grocer:

• He’s in his office, looking at his laptop with its spreadsheets, and everything is exactly the same in NB and B with one exception. In NB, he sees that the recent sales have been at something like 344,444 lbs/week. In B, he sees that the recent sales have been at 344,434 lbs/week (the difference coming from my grocer).

The odds that he’s going to react differently to these numbers in the two scenarios is ridiculously small. In particular, he’s not going to want to order less beef from the slaughterhouses in NB vs B. And that means there will be no difference in factory farming.

One really needs an adequate sense of the numbers in order to see this. Let’s suppose your net income is about $75,000 a year. In one future, you’re in your office gazing at the spreadsheet on your laptop and see that your average net income over the last few months was $6252/month. In another future, everything is exactly the same but you see that it was $6251.82/month. (That’s the percentage difference for the warehouse owner, 344,434 vs 344,444.) The idea that you’re going to do anything different in reacting to these different numbers is just plain stupid. It’s 18 cents for goodness sakes.

Hence, even if the grocer does different warehouse orders in NB and B—which as we saw is incredibly unlikely—it’s almost absurdly unlikely that the warehouse owner is going to want to do anything different in NB and B. In particular, he’s not going to want to order less beef from the slaughterhouse in NB compared to B, as a causal result of my weekly beef purchases.

If the meat industry were incredibly different, then the warehouse owner might well react differently to 344,444 and 344,434. If the entire industry from start to finish were almost supernaturally precise, efficient, predictable, and had zero waste, then the warehouse owner’s purchases from the slaughterhouses might be incredibly sensitive to miniscule differences, such as that between 344,444 and 344,434. But we don’t live in that world.

Note that there are two tiny probabilities here. First, there’s a tiny probability that the grocer is going to order less beef in NB than in B. Second, even in the scenarios in which the grocer orders less beef in NB than in B, there’s only a tiny probability that the beef warehouse owner wants to order, from the slaughterhouse, less beef in NB than in B. To figure out the probability that the warehouse owner is going to want to order less beef from the slaughterhouse, you have to multiply those two miniscule numbers together. It’s like multiplying 0.0001 and 0.000001: you get 0.000000001.

Guess what? We’re not finished with the miniscule probabilities.

Imagine that the warehouse owner really does want to react differently to the numbers in the box on his spreadsheet: 344,444 vs 344,434. He wants to order more beef from the slaughterhouse in B than in NB. Will he do it?

I don’t think so. According the LLM I used, if a beef warehouse that usually sells around 350,000 lbs/week wanted to increase its purchases from the slaughterhouse the minimal amount, that amount would be anywhere from 1,000 to 10,000 pounds, depending on the slaughterhouse. So even if the warehouse owner wanted to give different orders in the two scenarios, based on the differing numbers 344,444 vs 344,434, I doubt that he would actually do it, since a difference of 10 lbs hardly seems to justify a difference in orders to the slaughterhouse of 1,000-10,000 lbs. Even the lowest amount, 1,000, is too much.

Think about it:

• The warehouse owner is in his office, looking at his laptop with its spreadsheets, and everything is exactly the same in NB and B with one exception. In NB, he sees that the recent sales have been at something like 344,444 lbs/week. In B, he sees that the recent sales have been at 344,434 lbs/week. Is he going to order an extra 1,000-10,000 lbs/week in B?

Hell no!

Hence, even if the grocer ordered differently in NB vs B, and even if the warehouse owner wanted to order differently in NB vs B, I really doubt he actually would do so, given the amounts involved in the real world.

In sum, it’s virtually guaranteed that your buying beef will not lead to any causal differences in the slaughterhouse’s purchases from the ranchers.

Here’s a continuation of the argument, one that I’m not as sure about but seems worth articulating:

Suppose, against all odds, that the warehouse owner orders more beef from the slaughterhouse in B than in NB. So, yeah: he reacted differently from seeing 344,444 and seeing 344,434. Since the only differences in B and NB come from my beef purchases, surely he is doing this in response to my beef purchases.

Maybe, but I don’t think so. It seems more like a randomly caused difference, rather than anything caused by the differences in the two numbers. It’s not as though he was thinking, before seeing the figures, “If the recent sales are more than 344,439, then I’ll order 2000 lbs extra from the slaughterhouse”. The odds that he had that thought before seeing the recent sales figures are ridiculous.

Whether this mini argument is right raises issues I can’t deal with properly, so I’ll ignore the argument.

Call the argument of the previous section the zero-days argument, since it’s saying that your weekly beef purchases will cause zero days more factory farming.

The conclusion of the zero-days argument is not that buying beef is morally okay. It’s this: what you do with your ordinary meat purchases isn’t going to make any causal difference to the amount of factory farming.

Some philosophers agree with my thesis. I’m not surprised, as the zero-days argument is not exactly earth-shaking.

Maybe it should have been fairly obvious all along? To see why someone might think so, temporarily set aside CAUSE* and focus on CAUSE**.

CAUSE**. There are two possible futures in front of me. In the beef future, B, I buy a pound of beef JUST ONCE in the rest of my life. In the non-beef future, NB, I don’t buy any pound of beef or other item that uses factory farming. What we want to know is the likely difference between those two future scenarios. That is, if the two scenarios B and NB proceed under the same usual assumptions (we assume that pretty much everything is the same in both scenarios with the exception of my beef purchasing decision and its consequences), but in one scenario I buy JUST ONE pound of beef while in the other scenario I never buy any factory farming items, what are the likely differences in consequences from those two choices?

Here’s the answer: if I proceed with B instead of NB, then there will be—as a causal result of my SINGLE 1-lb beef purchase—very roughly one more day of factory farming than if I did NB.

I take it as obvious that CAUSE** is false. No one thinks that my buying just 1 lb of beef is going to lead to 1 more day of factory farming. The reason no one thinks that is we all know that the purchase is too small to cause any differences in the ranches. You would have to believe in magic to think the grocer, warehouse owner, and slaughterhouse owner are going to do anything different because you bought one freakin’ pound of beef.

And yet, as we have seen, modest weekly purchases are really not that different from the single 1-lb purchase. They’re still way too small to make a difference.

Even so, some people have objected to the thesis that your buying beef will not lead to any causal differences in the amount of factory farming. I’ll look at three of them.

OBJECTION 1: Probabilities

For this first objection, I’m going to rely on Amos Wollen’s post from Feb 04, 2025. He quotes Dustin Crummett as follows:

If we are uncertain whether the causal impotence objection succeeds, we still need to make decisions about whether to eat factory-farmed products. [...] it might seem [...] plausible to employ some sort of precautionary principle here. Such a principle would tell us that we should tend to err on the side of caution in cases where a course of action might [my emphasis] turn out to be seriously wrong. In the meat-eating case, nothing of great moral value will be lost if we refrain from eating meat: just a bit of gustatory pleasure. But it may [my emphasis] be that something very bad will be brought about if we eat meat and thereby trigger a production threshold, namely, the suffering and death of many animals, along with the other negative consequences of factory farming. Here, we might suppose that the burden is on the person who wants to run the unnecessary moral risk to show that their action is sufficiently safe.

I’m sorry but this is not an impressive argument. The main problem lies in the two emphases I provided: he uses “might” and “may”, which suggest that we’re dealing with probabilities that aren’t stupidly small. When you say you might or may buy a car soon, you don’t mean that there’s a one in a zillion chance you’ll buy a car soon. We already know that the odds that factory farming in NB is going to be less than it is in B are stupidly small. You might as well worry that your buying beef will cause WWIII.

OBJECTION 2: Adding up Zeros

Suppose that over the next few weeks, 100,000,000 pretty ordinary people in the USA increase their beef purchases by 1 lb/week. After a few weeks, there’s an extra demand of 100,000,000 lbs of beef per week. Over a year’s time, that’s about 5 billion lbs extra demand. Since the current total is about 27.5 billion/year, an extra 5 billion 1-lb purchases/year will definitely cause more factory farming.

But consider the zero-days argument I gave. I can trot that argument out for each of those 100,000,000 people. Each person fits the argument perfectly. If my zero-days argument works when applied to me, then it must work when applied to each of those people—since they’re not relevantly different. Hence, I can give 100,000,000 of my zero-days arguments to allegedly prove that there will be zero extra days of factory farming: 0 + 0 + 0 + …. But that’s clearly false. So, my zero-days argument must have been flawed.

The problem with this objection is the claim that the zero-days argument will apply to each of those people. Remember, in the zero-days argument both of these happen:

• The grocer looks at his spreadsheets and sees something like 1429 lbs/week in the B scenario and 1426.5 lbs/week in the NB scenario, since the only difference in the two comes from my beef purchases.

• The warehouse owner looks at his spreadsheets and sees something like 344,444 lbs/week in the B scenario and 344,434 lbs/week in the NB scenario, the difference coming from my grocer.

Neither of these happen in the vegan-nightmare scenario in which 100,000,000 people increase their beef purchases. So, no: the zero-days argument can’t be multiplied 100,000,000 times.

A causal claim like “If a huge group of people purchase an additional 5 billion lbs of beef per year, they’ll cause very roughly an extra 5 billion days of factory farming per year” is probably true. The grocery stores, beef warehouses, slaughterhouses, and ranches will definitely notice the differences in their spreadsheets and act quite a bit differently. But that doesn’t happen in the realistic scenario we have been examining. That’s one of the reasons I’ve been trying for realism in my zero-days argument.

OBJECTION 3: Average Impact & Not Being “Special”

Here’s Matthew Adelstein again, with one of his key premises in his argument for claims like CAUSE*:

This is a pretty general logical principle—if you have no specific information beyond the average impact, and no reason to think either your contribution will be more or less than average [my emphasis], you should expect to have the average impact. If you know that the average person rides 10 trains in their life, and you are assigned a random person who you have no information about beyond that they exist, on average, you should expect them to ride 10 trains.

Roughly put, he’s arguing for the idea that you should expect that the causal impact from your buying a pound of beef is one day of factory farming, given that (a) in order to produce X pounds of beef, we typically use X days of factory farming, and (b) you have no reason to think you’re special here.

There are a couple problems with his argument.

Problem 1. His “pretty general logical principle” includes an assumption which is usually false. I’ve put it in bold italics. In many cases, I have overwhelming, virtually conclusive evidence against the claim in bold italics.

In fact, his own train example demonstrates this quite well. If there are 8 billion people, and there are 80 billion train rides amongst those people, then of course the average number of train rides per person’s life is ten. But you’d be grossly mistaken to expect that a randomly chosen, non-special, person rides the train exactly ten times in their life. Contrary to Adelstein, that’s extremely improbable. A huge number of people have zero train rides in their life. A great many others have many hundreds or even thousands of train rides in their lives (one trip to work, one trip back home, every working day for a period of years). There’s an enormous range, and the odds that a person has exactly ten—or even, say, 7-13, is quite small.

For another example, suppose you’ve got a million people, and their combined net worth is $100,000,000,000. Then the average net worth is $100,000. The odds that a randomly chosen, non-special, person in that group has a net worth of $100,000 is extremely small. In fact, the odds are extremely small that their net worth will be even near that figure.

The problem lies in the use of “you have no reason to think you’re special here”. The person who rides the train 1234 times in their life isn’t special at all, since zillions of people ride trains very roughly that many times in their lives. The person who rides the train only 4 times in their life isn’t special either, since zillions of people ride trains about that many times in their lives. Even so, the two people are very different in their train life and far away from the statistical average.

Problem 2. Terms such as “average impact”, “expected impact”, and “expected effect” have to be interpreted carefully, and when we do it right, I don’t see that they give good reason to accept CAUSE*.

For some research modeling purposes, one can use the term “expected effect” so that the “average person” does have exactly 10 train rides in their life, to use the train example. You just do the calculation: 80,000,000,000 rides/8,000,000,000 riders = expected train rides is 10/rider. And then you use that figure in subsequent calculations.

Similarly, you do the beef calculation: 27,500,000,000 lbs/27,500,000,000 days of factory farming = expected effect of purchasing 1 lb is 1 additional day of factory farming. (The actual numbers are different, but not by much. If the “expected effect” turns out to be 0.5 days or 2.0 days, it doesn’t matter here.)

I have no objection to any of that! If using that model is helpful for some purposes, go for it. When doing science, we use idealized models all the time—and the results can be awesome. For instance, when faced with certain problems in physics, we treat electrons as point particles—and that’s a useful idealization for some but not all calculations. In formal (mathematical) epistemology, researchers often assume that each person knows all the logical consequences of each truth they know. This is a false but useful assumption. Science is littered with this kind of thing, and it’s turned out to be enormously useful. Ya gotta love scientific modeling.

But that’s a technical sense of “expected effect”. Like I already showed, the overwhelmingly likely effect is wildly different:

• Expected effect (theoretical modeling) = 1 lb beef purchase —> 1 additional day of factory farming.

• Overwhelmingly likely effect (causality) = 1 lb beef purchase —> zero additional days of factory farming.

Here’s why this matters so much:

When people worry about the ethics of their meat purchases, they are often picturing themselves in a grocery store, looking at a package of beef. They’re wondering “If I keep buying beef regularly, will it really cause about an extra day of factory farming for cows, for every pound I buy???” They are worried about causality, about the causal consequences of continuing to pick up packages of beef and buying them. They aren’t wondering about theoretical modeling.

It’s a mistake to respond to a person’s worries about the causal consequences of their meat purchases by ignoring the causal consequences of their meat purchases in order to focus on theoretical notions such as “expected effect”.

Again, I’m not saying it’s morally okay to buy meat in grocery stores. I’m not saying that at all. But I am saying this: if there is a good moral case against buying meat from your grocery store, it won’t be because buying it will increase factory farming; it will have to be based on something else.