Duking It Out: Bentham, Amos, and Elusive Chicken Utils

It might be okay to eat meat after all

A fellow university student substacker Amos Wollen recently posted an interesting blog, titled The Causal Impotence Objection to Veganism. After a little bit of research, I found out that another fellow undergrad substacker, Bentham’s Bulldog, wrote a refutation of the causal impotence objection to veganism too. Although Nick and I love both of their blogs, I’m afraid it’s time to cross swords with these philosophical birds of a feather.

Okay, first, a little bit of context. The classic argument for veganism, simply put: Eating meat doesn’t pass the cost-benefit test. The costs are incredible amounts of suffering and pain, and benefit is, what? The difference between chicken and vegetable stock in your gravy?

Okay, seems sensible. High costs for negligible benefits. However, there’s a hidden assumption. When I buy that chicken stock, as opposed to vegetable stock, I’m causing chicken suffering.

You might say, well obviously you are! Bentham writes:

Let’s imagine that you know that, on average, each of ten people who consume chicken will result in one extra chicken being produced. You don’t know if there’s some specific threshold—maybe it’s the case that at particular points you trigger a threshold, maybe it’s just a smooth continuum. The claim is that if you have no extra information beyond what the average impact will be from one extra chicken consumed, you should expect to have the average impact.

If you think about 10 people, the producer probably knows everyone by their first name. The producer probably notices, definitely notices, if Bentham doesn’t pick up his weekly roast chicken, and adjusts by dropping production by 1. This is the average person’s intuition on the matter but it becomes more inaccurate the larger the numbers get.

The logic reminds me of a scene in East of Eden, where a character named Samuel came down to a ranch every so often, and when he did, Lee would go around to the back of the ranch and kill a chicken, fry it, and they would eat it for supper. Often Samuel would even decline Lee’s invitations, purely on the grounds that he felt bad killing chickens every time.

I find this example helpful because it illustrates how easy causal logic can be when the numbers are closer to what our brains evolved to understand. But let’s get out of econ 101 world or John Steinbeck’s imagination, and into the real world of the chicken industry.

9.22 billion chickens are produced in the U.S. each year, for a market of around 100 million people. What is your marginal effect on production? Amos writes:

There are features of chicken supply chains besides demand that affect how many chickens get produced: limitations on crate space, the farmer’s option of switching to a different retailer if their current retailer orders less than usual (as opposed to reducing supply), the tendency for chicken farmers to produce more chickens than they need since they’re so easy to stuff into tiny cages in large numbers, etc.

Thus, even if some number of people going vegan will have some reducing effect on the number of chickens produced, the expected value of going vegan in terms of reducing chicken supply may be less than a reduction in chicken supply by one consumer’s worth.

So Amos partly concludes that we should be empirically skeptical about our causal influence on chicken production. But this doesn’t ruin the argument for Amos. Amos thinks that the pain caused is so great that, even in the face of uncertainty, you should still be cautious, and gives a Pascal’s Wager-type conclusion: the chance that you could cause extra chicken production still outweighs that small benefit of say, going to Popeyes instead of finally learning how to make tofu taste good.

But wait a minute. I’ve seen an argument like this before. It’s similar to the ‘paradox of voting’ argument that economists like Bryan Caplan give all the time.

Imagine: The chance your vote is the decisive one in an election is 1 in 500,000. There is a bill that, if passed, would give a benefit of $25,000 to all poor people. You, kind reader, happen to be poor! 25k seems pretty good, however, you are faced with another option. You could go down to the car dealership and buy a car for 25k. What are the expected values of both decisions? The benefit of getting the car is simple, there is basically a 100% probability that if you go, you will drive back home in a new car. So the benefit is just the price of the car that is now yours (25k). Now take the policy proposal. The only way you could have a causal influence is if you are the deciding factor in the vote, so your expected value is 1/500,000 X 25,000 = 0.05. A nickel. Hence, only about half of America votes! These simple calculations help unveil a fallacy: Even if the effect of your decision could be huge, the low probability makes the expected value vanishingly small.

But of course, the two examples aren’t the same. If they were it would be a slam dunk argument. The expected value of eating a chicken would be much higher than the absolute negative expected value of eating it. Say the probability that you caused a chicken to be produced was again 1 in 500,000, and the expected value was 1 util, and the expected value of a chicken being produced was 10,000 times worse, or -10,000 utils. Well, then it would easily pass the cost-benefit test. Cost would be 1/500,000 X -10,000 = -0.02 utils vs 1 util.

But this is bad economics. We know, as Amos said, the probability isn’t 100% that you cause 1 chicken to be produced if you buy 1, but how do we know it’s 1 in 500,000? I just made the number up. In fact, it’s very likely it’s higher than that, considering that, unlike ballot voting, voting with your wallet has marginal effects.

So what’s the true number? What’s the probability? Let’s get back to Bentham. Before that though, a funny anecdote.

In writing this essay, I was really wanting a professional opinion. So a couple of days before my macroeconomics midterm I went to my professors office and asked him about these damn elusive chicken utils. “Mike, I’ve been debating with myself for days on my effect on chicken production, please give me the answers,” I said. At first, he told me you would have none. Standard argument, consumers are price takers, one person isn’t squat. I said okay, that’s what I’m thinking, but here’s this guy Bentham’s argument against that. And then he said, oh well yes marginally you would have a small effect. So I of course gave him my argument a 2nd time. Getting annoyed, he simply told me “Who cares?”. I guess I should have expected a macroeconomist (over 70 years old as well) to not care about unit-by-unit changes in demand, but I persisted. I wanted to get my answers on these elusive chicken utils! So I told him why he should care. We then went in circles like this for about 5 more minutes until a couple of girls showed up with actual important mid-term questions, and I just said to hell with it and let them have their time with him. My takeaway: Academics get startingly incurious the minute they step even an inch out of their field.

Back to Bentham: Bentham claims that your marginal effect is actually higher than 1, because every purchase not only increases chicken production, but also increases the production of chickens that die before being sold, and the portion of meat that gets wasted before being sold. So Bentham then goes on to argue that consumers should expect to increase chicken production by more then 1, as the producers wouldn’t make chickens to sell to no one.

Seems pretty good, but there’s a key assumption in his argument. Bentham repeats the phrase: “if a consumer has no special reason to think that their action will not trigger any thresholds” before he concludes his argument, and this is doing a lot of work, considering that, we do have a lot of information about thresholds which greatly lower the probability that your marginal effect is anywhere close to 1.

It’s important, because someone could forward this type of argument: “In a competitive market with millions of buyers, each individual is kind of like an atom, whose decision may be far, far too small to shift the market outcome. So yes, collectively consumers determine total production, but any one extra chicken in a market of tens or hundreds of millions of purchases has an almost zero chance of being the marginal decision that tips the producer into increasing output.”

Bentham answers this argument by arguing that, as said earlier, unless you have information about thresholds, you should in an a priori sense, think your marginal effect is more than 1, because firms react marginally, and your effect has to trickle down to the firms somehow. This is a pretty good argument if you accept the threshold assumption. I don’t think I do.

Yes, the firms react to changes marginally, but marginally in a lumpy sense, certainly not in a unit-by-unit sense. Prices are sticky. Chicken production is planned on production cycles (often seasonal or quarterly) and relies on forecasts, contracts, and economies of scale. Production adjustments occur over time and are based on aggregate demand trends—not on day-to-day or chicken-by-chicken variations in individual consumer choices. Price signals and inventory decisions are set by processors who aggregate demand over millions of units. The “pivotal” consumer who could tip the production decision is not your single purchase but rather the collective shift in demand.

So Bentham’s assumption about threshold effects might not be so relevant. It seems highly probable that a consumer’s impact on production is relative to whether they impact the threshold or not, and because producers only respond to large changes, the probability of being the reason for a shift in production (being the last to meet the threshold) would be exceedingly low. If there’s a threshold at 10,000 chickens, that doesn’t mean the market is sitting right on the edge of 10,000 chickens at all times. It could be at 1,050, 9,332, or 27. Your choice to buy or not is a rounding error, very likely going to be lost in noise, entirely due to these market frictions.

But think marginally you say! With chickens, let's say the ‘lump’ is 1,000, and you buy a chicken. Aren't you responsible for 0.001 of that lump, or 1 chicken? Doesn’t your effect come out of the woodwork somehow?

This reasoning is intuitive, but it incorrectly assumes deterministic causation, when in reality, the market operates probabilistically.

Think about it like this:

When you decide not to buy a chicken, it doesn’t automatically deduct a fraction of a chicken from future production. What’s actually going on is your decision slightly increases the probability that at some point in the future, demand will drop low enough to trigger a batch reduction. This means your expected effect is much smaller than 1 chicken because the probability of you actually being the "pivotal" buyer (the one who tips the balance) is tiny.

Let’s say the market operates in units of 1,000 chickens. If 1,000 fewer sales occur, production decreases by 1,000. Your decision is one of those 1,000, meaning that in only some cases, your choice is what actually tips the balance. But demand fluctuates constantly! Maybe 500 people stopped buying today, but tomorrow demand rises by 200. The firm doesn’t respond to each individual sale, it responds to aggregate trends. If you stop buying a chicken, but 999 other people do not, nothing happens. Mathematically, the probability that you specifically were the pivotal consumer who pushed demand below the threshold is not 1/1,000. It's much, much smaller.

But you might say, even if my affect is tiny, it’s still there.

Sure, but the same is true for every random fluctuation in demand. A chicken farmer might decide to produce 10,000 more chickens because of weather conditions, or because they misinterpreted market signals. Again, your effect is drowned in noise.

Bentham gives another, similar argument, this time from Huemer later in his essay.

Huemer’s argument is the following.

1. If 1 million people gave up meat, meat production will decrease by roughly 1 million.

2. The average impact of one person giving up meat is a reduction in meat production by 1 million portions/1 million people=1 portion per person.

3. You’re not special—you’re not more or less likely to trigger a change in meat production than any other person.

4. Therefore, (from 2 and 3), your expected impact from giving up meat is a reduction of 1 meat portion produced.

But this assumes that same claim, that markets are perfectly marginal. Meat producers don’t adjust supply for every individual sale. The impact of one person is not “1 portion”. Huemer’s math only works if firms adjust supply continuously with each purchase. But firms adjust in large batches! In short: If you stop buying one chicken, nothing changes unless it contributes to a threshold drop.

So, to conclude, should Bentham and Amos start live streaming Lousiana seafood boils on substack? I don’t feel confident enough to recommend it morally. (Count me in for the live stream on entertainment value though) However, I do feel confident claiming some kind of moral agnosticism on eating meat in a large, non-econ 101 world economy. If you go out in the woods and kill a deer just to eat it, that’s morally wrong if you accept the traditional arguments for veganism But in an economy of this size, with this much uncertainty about your true causal effect, I think eating meat is, at the very least, highly defensible and possibly even right on expected value grounds. (conditional on you getting utility out of eating it of course) I think this is part of the reason veganism isn’t more popular, but people still feel terrible when they run over an animal on the highway. People’s intuition is: “I could buy this ground beef and get immediate and unmistakable benefits, or I could not buy it and comfort myself with, what? The uncertain mental gymnastics that is my causal influence on the meat industry? I think I’ll buy the ground beef, and I think I’ll buy that car for 25k and stay home on election night.” Is this reasoning really so off the mark?