I think it’s helpful to just quote the buffer example in Budolfson (2018):
Richard makes paper T-shirts in his basement that say ‘HOORAY FOR CONSEQUENTIALISM!’, which he then sells online. The T-shirts are incredibly cheap to produce and very profitable to sell and Richard doesn’t care about waste per se, and so he produces far more T-shirts than he is likely to need each month, and then sells the excess at a nearly break-even amount at the end of each month to his hippie neighbor, who burns them in his woodburning stove.10 For many years Richard has always sold between 14,000 and 16,000 T-shirts each month, and he’s always printed 20,000 T-shirts at the beginning of each month. Nonetheless, there is a conceivable increase in sales that would cause him to produce more T-shirts—in particular, if he sells over 18,000 this month, he’ll produce 25,000 T-shirts at the beginning of next month; otherwise he’ll produce 20,000 like he always does. So, the system is genuinely sensitive to a precise tipping point—in particular, the difference between 18,000 purchases and the ‘magic number’ of 18,001.
Suppose that a consumer knows all of these facts about Richard’s business, and is considering buying a T-shirt for himself. What is the expected effect on the number of T-shirts produced of that consumer purchasing a T-shirt? The correct answer is essentially zero, because given what is known about the history of demand for Richard’s T-shirts and how production quantities are determined, there is virtually no chance that exactly 18,001 people are going to buy Richard’s T-shirts this month and trigger a dramatic threshold effect—which, of course, is not to claim that there is zero chance of that happening, but rather that the odds of that happening—of exactly 18,001 of Richard’s T-shirts being sold—is certainly dramatically lower than 1⁄5,000 or any other number that would drive the expected effect of an individual buying one T-shirt anywhere near the consequence that 1 additional T-shirt is produced. This shows that the reasoning behind the Singer/Norcross/ Kagan Response is invalid, because insofar as that response is taken to show that consuming meat should be expected to have significant bad effects for animal welfare (e.g., equal to the average effect of those individual actions), similar reasoning would show that buying one T-shirt in the story above should be expected to result in approximately 1 additional T-shirt being produced, which is the wrong result.11 The problem with the reasoning is that it overlooks the fact that we can know enough about the supply chains in both cases to know that threshold effects are not sufficiently likely and are not of sufficient magnitude to drive the expected effect of consumption anywhere close to the average effect
I think the response in McMullen & Halteman (2018) to buffers is good. My understanding is basically that buffers either a) exactly reduce to the unknown threshold case where the expected value works out, or b) the amount of product to fill the buffer is set based on anticipated prices, which are informed by previous sales, so the buffer moves with anticipated prices.
First, it is possible that some amount of these buffers has to be “used up” or get very large before production will increase or decrease. For example, an egg producer might find that it is cheaper to stock a local warehouse all at once each week, and only restock when the warehouse is almost empty. In this case, consumers would have to purchase a warehouse full of eggs before a new restocking order was made. If this is the case, then the buffers are just functioning as part of the threshold framework described by Singer and Kagan. The number of eggs in a warehouse shipment would be the size of the threshold that needed to be overcome.
Alternatively, it is also possible that some buffer is a constant part of the system, and will not change in response to consumer choices. This is particularly likely with food waste. There could be some imprecision in production and shipping that is just too expensive to eliminate. For example, there might be a potential electronic system that would track chicken age, location, and quality, and thus minimize waste, but if that system cost the firm more money than it would save, firms might choose to dispose of some chickens rather than track them more precisely. This kind of waste would be present in the system whether demand was high or low, and might exist at all levels of the supply chain. Note that a consumer decision to consume one more or one less chicken will not have any effect on this kind of waste – the waste would be exactly the same regardless of consumer choices.
I’m less sure about cases where there’s coordination and quotas in the industry, like supply management, or if the industry is too non-competitive, concentrated in a few companies (oligopoly, oligopsony) at some points along the supply chain. This depends on the country/region.
I think it’s helpful to just quote the buffer example in Budolfson (2018):
I think the response in McMullen & Halteman (2018) to buffers is good. My understanding is basically that buffers either a) exactly reduce to the unknown threshold case where the expected value works out, or b) the amount of product to fill the buffer is set based on anticipated prices, which are informed by previous sales, so the buffer moves with anticipated prices.
I’m less sure about cases where there’s coordination and quotas in the industry, like supply management, or if the industry is too non-competitive, concentrated in a few companies (oligopoly, oligopsony) at some points along the supply chain. This depends on the country/region.