Independent research in inspiring and scaling collective intelligence
Previously Head of Procurement at Anthropic starting 2021
Executive Director at CFAR, starting 2018; Productivity Coach in 2017; Senior Research Analyst and Operations Associate at GiveWell 2013-2017
My “not buying a chicken” seems like it would look very similar to anyone else’s “not buying a chicken”.
Thanks for acknowledging that.
I think standard economics agrees with your vision of “~always positively-sloping finite supply curves” in the short term, but not necessarily the long term. Here’s a quote from AmosWEB (OK, never heard of them before, but they had the quote I wanted)
As a perfectly competitive industry reacts to changes in demand, it traces out positive, negative, or horizontal long-run supply curve due to increasing, decreasing, or constant cost.
Oops, I meant to edit that rather than retract. Since I don’t believe there’s a way to un-retract I’ll re-paste it here with my correction (Changing “Supply Elasticity is 1” to “Supply Elasticity is finite”):
Cumulative elasticity = Supply Elasticity/(Supply Elasticity—Demand Elasticity). A cumulative elasticity factor of one means a demand elasticity of 0.
I believe your math skipped a step; it seems like you’re assuming that Supply Elasticity is finite. I actually claim in the original article that “the ‘price elasticity of supply’ in the arbitrarily long term becomes arbitrarily high”. In other words, as “length of ‘term’” goes to infinity, the Supply Elasticity also goes to infinity and the cumulative elasticity factor approaches 1 for any finite Demand Elasticity.
Thanks for the math demonstrating my point.
Stepping back, I worry from your sarcastic tone and the reactive nature of your suggestions that you assume that I am trying to ‘beat you’ in a debate, and that by sharing information that helps your argument more than it helps mine, I have made a mistake worthy of mockery.
Instead, I am trying to share an insight that I believe is being overlooked by the ‘conventional wisdom’ of this community and is affecting multiple public recommendations for rational behavior (of cost/benefit magnitude ~2x).
If I am wrong, I would like to be shown to be so, and if you are wrong, I hope you also want to be corrected. If instead you’re just debating for the sake of victory, then I don’t expect you to ever be convinced, and I don’t want to waste my effort.
Cumulative elasticity = Supply Elasticity/(Supply Elasticity—Demand Elasticity). A cumulative elasticity factor of one means a demand elasticity of 0.
I believe your math skipped a step; it seems like you’re assuming that Supply Elasticity is 1. I actually claim in the original article that “the ‘price elasticity of supply’ in the arbitrarily long term becomes arbitrarily high”. In other words, as “length of ‘term’” goes to infinity, the Supply Elasticity also goes to infinity and the cumulative elasticity factor approaches 1 for any finite Demand Elasticity.
Thanks for the math demonstrating my point.
Stepping back, I worry from your sarcastic tone and the reactive nature of your suggestions that you assume that I am trying to ‘beat you’ in a debate, and that by sharing information that helps your argument more than it helps mine, I have made a mistake worthy of mockery.
Instead, I am trying to share an insight that I believe is being overlooked by the ‘conventional wisdom’ of this community and is affecting multiple public recommendations for rational behavior (of cost/benefit magnitude ~2x).
If I am wrong, I would like to be shown to be so, and if you are wrong, I hope you also want to be corrected. If instead you’re just debating for the sake of victory, then I don’t expect you to ever be convinced, and I don’t want to waste my effort.
That question seems to have a simple answer: your decision will not affect the long-term production of chicken.
OK, so I argue option A, you state option B, and the articles I link argue option C.
That is a much more complicated question
I agree it’s a complicated question (in that it requires lots of information to answer precisely and accurately). If you had no empirical data to work with, what would be your best guess/expectation? Also if your answer is proportionally different than in the ‘single chicken’ case, I’d be curious to know why.
Right. In this case, to answer the question, “If I decide to reduce my lifetime consumption of chicken by one, should I expect the long term production of chicken to drop by ~1, ~0, or something in between?” Which is of demonstrated interest to the authors I am critiquing.
What you are effectively claiming is that there are no suboptimal producers of chickens. Unless every producer of chickens is ideally located, ideally managed, ideally staffed, and working with ideal capital there are differences in production costs.
It’s not that this will ever actually be the case, but the argument is that, in the long term, the market approaches what you would expect with such assumptions (and continues to have short term fluctuations away from that). But yes, even this assumption is clearly not actually true in all cases (as with all assumptions in neoclassical economics); the better question is whether it’s a good simplification (enough to form a reasonable prior) or whether there is a better simplification we can consider (either simpler or more accurate).
The estimates I’m critiquing in the original post assume “short term elasticities are the best prior for long term elasticities” and I am advocating that “a better prior for the long term cumulative elasticity factor is 1″.
There is a reason, that economics assumes that the amount of a good supplied changes as price changes, and I haven’t seen any argument that exempts the case of chickens. Also, how does the market create less chickens as demand falls? If there are differences in cost, the highest cost producers leave the market as price falls. Easy to answer with the standard assumptions, but almost impossible with your nonstandard prior.
The explanation of both of these issues is the short term supply curve (which is not flat). In the short term, if people stop eating chicken, the price drops, and the producers that are (in the short term) able to improve their (expected long term) profits by scaling or shutting down do so.
Empirical evidence is nice and often more convincing than theory, but I don’t think it’s necessary for an argument to be convincing (to believe otherwise would be quite… burdensome).
In this case, the original articles I am critiquing used purely theoretical arguments to claim that there will be long term price elasticity of supply, and I think that a theoretical critique is sufficient to show that the strength of their arguments is currently too weak to support the complexity of their theory.
I’m certainly open to any empirical evidence that may exist. Would you find a quick analysis of Big Macs moving (or if not, do you have a suggestion for a different empirical analysis)?
No, I haven’t looked at the empirical evidence because I didn’t think it would be as convincing as the 2 theoretical arguments I made in the original post; let me know if you are aware of any such analysis.
Would you accept the results we find from an analysis of Big Macs as relevant?
Since Big Macs aren’t generally transported across national boundaries, we can think of the market for Big Macs in each country as largely independent.
While we would both expect various factors such as the price of labor to affect the price of Big Macs differently in each country, you would expect the price of Big Macs to positively correlate with # sold in that country (or possibly # sold per person), right? I would not expect such a correlation. (I think looking across countries is better than in one country across time, since then technology or other time-dependent factors would bias the results.)
If we had time we could control for all the other factors we don’t want to bias the results like price of labor; but maybe even without these we might see some interesting initial patterns.
In the specific example, they could be cloned by expanding in the good locations.
More generally, if you’re claiming that there’s a limited supply of good locations from which to produce chickens, then that reduces to a “finite inputs” argument I discuss in the last section of the OP. (For further discussion see responses to this comment .)
In short, I agree that such effects can create a sloping long term supply curve in some cases, but I also believe that there are other effects that can lead it to slope the opposite direction, and it’s not immediately obvious which wins out. My prior is that the long term supply curve for an arbitrary product is virtually flat.
Said another way, if you’re going to argue that the long term cost-per-widget is higher when producing 2X widgets than X widgets, then you have to argue that the effect of finite inputs outweighs gains to scale and other factors. I haven’t seen such an argument generally or in the case of chickens.
This is true in the short term, but in the long term, the dynamic changes for producers:
The producers that know how to make chickens for $8 scale up or their production strategy is replicated by others.
The marginal cost of production (and hence price) keeps falling until all producers are making no profit (relative to opportunity cost of capital)
The industry can scale up/down (in the long term) to meet changing demand, but it can’t drive prices any lower. If prices were any higher the industry would scale up in the short term and keep expanding until the price fell back to the Cost in the long term.
The elasticity of the demand curve changes less than the supply curve in the super long term, but if you agree with me that the supply curve is virtually flat at that point, then the elasticity of the demand curve is negligible (because as the supply curve shifts left and right, the only point on the demand curve that matters is quantity @ price = Cost/Supply price).
I think your scenario is a good illustration of “finite inputs”, which I listed as one of five example ways in which the long term supply curve may not actually be flat (at the end of the original article).
While I think that finite supply is a very real force (that, if strong enough, would create significant long term price elasticity of supply as you claim), the other four examples I mentioned also seem very real to me, and it’s not obvious which ones win out for any particular industry.
If Cost always grew with industry size, products in big industries would always cost more than the same product from equivalent but smaller industries (where both supply and demand is reduced). Intuitively/anecdotally this doesn’t seem to be true; I think the most common reasons it’s not true are “gains to scale”.
So what you are basically saying is that in the long run the price will be driven close to the lowest average price—right?
Yes; in the long term the producers that have higher-than-average Costs will be driven out of the market.
Still not quite, as once you recognize that “perturbations” will happen, you need to engage in some risk management (zero mean does not imply zero volatility). In your scenario the chicken producer seems to be fine with the 50% chance of going bankrupt at the delivery time which isn’t a good assumption to make.
I’m modeling risk management as part of the typical Cost of doing business, along with things like interest rates, opportunity costs of capital, chicken feed, other inputs, etc. Separating out risk management as a stand-alone variable doesn’t seem to change anything.
If the received wisdom is that a larger chicken industry will increase the price of chicken feed, then my prior is that it’s true in the short term, but not the long term. Chicken feed might be in finite supply, in which case Cost might grow with chicken industry size, but there are other reasons I can imagine Cost might shrink with increasing chicken industry size (listed in original article) and I don’t have enough confidence about any of these factors to break my prior that Cost at industry size 2X is ~Cost at industry size X.
Huh? A flat supply curve means that the producers will produce the same number of chicken regardless of the price at which they can sell them. I don’t see why this should be true in long term.
I mean “horizontal” rather than “vertical”. In that sense, a flat supply curve means a constant price, not a constant quantity.
Not quite. You are implicitly assuming that the Cost is fixed in stone and it isn’t. The chicken producer should accept all 10-year forwards on chicken if and only if he can buy matching forwards on his production inputs, otherwise he is exposed to the risk of, say, the price of feed going up.
I agree, but this is the type of pressure on Cost that I have no expectation of being in any particular direction. As a result, on expectation these perturbations average to zero, and the argument holds. It would be interesting if we had a reason to expect that Cost would go up or down depending on the amount of production. These are the issues my last section in the original article was intended to address.
It’s important to note that supply and demand aren’t perfectly linear. If you reduce your demand for meat, the suppliers will react by lowering the price of meat a little bit, making it so more people can buy it. Since chickens dominate the meat market, we’ll adjust by the supply elasticity of chickens, which is 0.22 and the demand elasticity of chickens, which is −0.52, and calculate the change in supply, which is 0.3. Taking this multiplier, it’s more accurate to say you’re saving 7.8 land animals a year or more. Though, there are a lot of complex considerations in calculating elasticity, so we should take this figure to have some uncertainty.
I think the calculations would be simpler and more accurate to assume that long term supply is in fact flat, so that eating one fewer animal causes ~one fewer to be produced in the long term. A more complete argument here.
If true, this would strengthen your overall point and make people even more empowered to reduce suffering!
One can be very modern and say that supply and demand curves are not real, but they are.
I’m not arguing that supply curves aren’t real; I’m arguing that the super-long term supply curve is virtually flat (the price elasticity of supply is arbitrarily high). I find it compelling to imagine a chicken producer accepting contracts to produce chickens 10 years from now. At what prices and quantities would the producer accept the contracts? I would say it would accept all contracts at or above the Cost, at as high a quantity as possible. With lead time and certainty, the issues that create short term elasticity don’t apply. (And yes, short term market shocks and uncertainties are real, but they’re just built into the profitability model of Cost.)
the lower price will cause a switch towards chicken consumption among other people who were on the margin of demand.
Under the simplified assumptions we’ve been using, the starting price is the Cost of producing chicken. When the price drops in your scenario, that means chicken producers are operating at a loss. True they might not throw out any chickens if they appropriately adjust price, but they will not be satisfied with the new equilibrium (because the price is too low). Instead they will reduce production, let the price rise back to the starting point, and let the new chicken consumer (who values chicken at slightly less than Cost) drop back out of the market.
If you don’t believe that the producers will keep reducing consumption until the price rises back to the original level and the “new consumers” (who value chicken < Cost) stop buying again, then you have to explain why producers aren’t already producing more chickens than they are (in the absence of any dietary changes). In other words, if “new consumer” demand exists, and producers are still profitable with slightly lower prices, why haven’t they scaled up production to larger than what it actually is today to sell chickens to “new consumers”?
Yes; another way to think of this is, “How do you model waste?”
If you think waste is best modeled by a fixed percentage of all production, then our best guess about the waste is that it changes proportionally with consumption. We don’t get to magically assign our consumption to the ‘waste’ category without highly specific information (such as, “I found it in a dumpster”).
If you expect the percentage of waste to grow/shrink with industry size, that could be an argument for slightly less/more than 1:1 effect (I’d put it in the “Gains to scale” category, even if it were negative). But I’ve never seen someone make that argument or attempt to model it.
Richard Chappell discusses a very simple model here.
Thanks for sending this; the ‘chunky fallacy’ comes up frequently when discussing this issue. Unfortunately, he explicitly endorses using short term elasticities at the end of his article.
OK so you have no prior for large cases, you have no prior about the relationship between large cases and small cases, and your guess for small cases is “zero impact”.
My prior for large cases is 1:1 impact, my prior is that the impact in large cases is proportionally similar to the impact in small cases, and therefore my prior for small cases is 1:1 impact.