When I hear that a batch of heroin has killed three people, I update strongly toward the hypothesis that it is good shit. So do several other participants in the market. The public-health establishment regards this as a failure of risk communication. I regard it as consumers successfully extracting decision-relevant information from an institution that did not intend to help them shop.
\nThis response to overdose reports has been documented in qualitative research, although researchers tend to describe it with an air of disappointment. A seller can claim that his product is exceptionally strong, but the statement is cheap talk. The customer who dies has supplied evidence at a cost no ordinary testimonial can match. He has surrendered every future opportunity to purchase the product, revise his opinion, or accept a discount in exchange for improving the review. We should be cautious about dismissing an information system merely because its most credible participants are no longer available for follow-up questions.
\nThe epistemological problem is one of latent quality. The buyer cannot directly observe what he is purchasing, the seller has an incentive to exaggerate, and ordinary consumer descriptions are noisy. Under these conditions, a dramatic physiological outcome can carry a substantial likelihood ratio. I am aware that the probability of a batch being potent given an overdose is different from the probability of an overdose given a potent batch. Much of the criticism I receive consists of explaining this distinction as though I had never considered it. I am trying to infer what other people bought. My critics keep changing the subject to what might happen if I bought it.
\nThere are, of course, confounders. Individual vulnerability, other substances, and variation within the supply complicate any inference from a particular death. Three fatalities associated with one seller are not three perfectly independent observations. Nor can one interpret the numerator without knowing how many people were exposed. I have repeatedly requested better denominator data. The usual response is that someone’s son has died and this is an inappropriate time to ask how much business the man who sold to him was doing. I sympathize, but uncertainty about sample selection does not resolve itself through respectful silence.
\nNarcan adds a major intervention to this already difficult identification problem. Observed mortality now depends partly on whether somebody was present, whether a reversal was attempted, and whether it succeeded. The person who survives does not reveal what would have happened without treatment. If intervention is correlated with location, social networks, or access to services, the resulting missing counterfactuals are not randomly distributed. Two sellers can supply equally dangerous products and acquire very different reputations because one happens to operate near people who are good at rescuing his customers.
\nThis is often presented to me as a reason to improve the model. I agree, but the improvement has a cost. I must now estimate the rescue environment as well as the underlying product, while the people introducing the complication congratulate themselves for making the market safer. Their favorable outcome and my increased epistemic burden arise from the same intervention. A complete welfare analysis should record both. Instead, one side receives public funding and the other is told that keeping a spreadsheet demonstrates a troubling lack of empathy.
\nNonfatal overdoses remain informative, but they create a measurement problem. Someone who was revived may provide an account, omit details, attribute the event to something else, or decide that his experience belongs to him and should not be reduced to a consumer report. Death is considerably less cooperative with retrospective reinterpretation. I am not claiming that a corpse is a sufficient statistic for product quality. I am claiming that a system which replaces relatively legible outcomes with heterogeneous narratives imposes additional work on anyone attempting to aggregate the evidence.
\nThe financial importance of this information is easy to underestimate. Where nominal prices do not fully adjust to differences in potency, the buyer is evaluating the real price of intoxication rather than merely the price of a packet. An overdose report can therefore reveal a product that participants perceive as unusually favorable at the posted price. The resulting rush of demand is a decentralized attempt to exploit that discrepancy. Public-health officials see people moving toward danger. The market participants see a limited opportunity that other people are already learning about. These interpretations can both describe the same movement without giving the officials exclusive authority over its meaning.
\nNarcan also changes the distribution of downside risk. The consumer and seller participate in a transaction whose catastrophic outcomes may be partly absorbed by publicly financed rescue services. In financial terms, the system has acquired a subsidized protection mechanism. It is not a perfect hedge, but perfect protection is unnecessary for incentives to change. A market in which adverse outcomes are repeatedly interrupted cannot be expected to price risk in the same way as one in which participants bear them fully. This is generally understood when the participants are banks. Apparently, the analysis becomes offensive when the rescued institution is a man.
\nThere is a separate demand-side effect. A customer who dies produces information and then ceases competing for the remaining supply. A customer who is revived can reenter the market, bringing both his purchasing power and whatever he learned. Rescue therefore preserves an additional bidder while leaving other buyers to absorb the resulting competition. I recognize that his continued existence benefits him. What I reject is the assumption that this exhausts the distributional analysis.
\nOne outreach worker asked whether I would prefer that somebody die so I could obtain a better deal on heroin. This is an emotionally optimized description of a much broader position concerning decentralized knowledge production, public subsidies, and the allocation of scarcity. I do not formulate my preferences in terms of particular individuals dying. I formulate them over states of the world, some of which contain more reliable information and less competition for goods I value. Her insistence on translating those states back into identifiable people makes a serious discussion almost impossible.
\nA friend recently survived an overdose and has since become much less interested in discussing the circumstances of the purchase. His sister is raising money for wider Narcan distribution. I suggested reserving a small portion for documenting what happened in a form useful to other consumers. She accused me of trying to extract value from his suffering. I explained that the suffering had already occurred, that its informational product was perishable, and that allowing both to be lost would be a particularly wasteful way to demonstrate respect.
\n","excerpt":"When I hear that a batch of heroin has killed three people, I update strongly toward the hypothesis that it is good shit. So do several other participants in the market. The public-health establishment regards this as a failure of risk communication. I regard it as consumers successfully extracting decision-relevant information from an institution that did not intend to help them shop.","tags":["Economics","Epistemology","Decision Theory"],"score":378,"displayOrder":690,"thumbnail":"/assets/articles/narcan-is-destroying-price-discovery-in-the-heroin-market-thumb.webp","image":"/assets/articles/narcan-is-destroying-price-discovery-in-the-heroin-market.webp","commentCount":7,"date":"2026-10-16T00:00:00.000Z","updatedAt":"2026-10-22T14:16:58.728Z","revision":5,"status":"published"},"comments":[{"id":"seed-narcan-is-destroying-price-discovery-in-the-heroin-market-0","author":"missing_denominator","text":"A fatality is evidence about a joint distribution of dose, tolerance, contaminants, response time, and strength. You cannot identify the strength parameter from the fatality count alone. Calling the signal expensive does not make it specific.","parentId":null,"score":71,"date":"2026-10-16T00:17:00.000Z","own":false,"moderatable":false,"seed":true,"ownerAuthor":false},{"id":"seed-narcan-is-destroying-price-discovery-in-the-heroin-market-1","author":"posterior_potency","text":"Specificity is desirable, but the comparison is with the information currently available to buyers. I am not claiming that a death is a laboratory assay. I am claiming that sellers prefer we have neither.","parentId":"seed-narcan-is-destroying-price-discovery-in-the-heroin-market-0","score":13,"date":"2026-10-16T00:52:00.000Z","own":false,"moderatable":false,"seed":true,"ownerAuthor":false},{"id":"seed-narcan-is-destroying-price-discovery-in-the-heroin-market-2","author":"censoring_is_not_loss","text":"Survival after treatment is an observed outcome. You can record the event, the intervention, and the uncertainty. You keep describing a person who can still answer questions as a missing observation.","parentId":null,"score":62,"date":"2026-10-16T02:08:00.000Z","own":false,"moderatable":false,"seed":true,"ownerAuthor":false},{"id":"seed-narcan-is-destroying-price-discovery-in-the-heroin-market-3","author":"credit_spread","text":"You want a public subsidy for the part of the risk borne by you, but a stronger market signal from the part borne by someone else. This is a recognizable portfolio preference. I would stop presenting it as an objection to subsidies.","parentId":null,"score":24,"date":"2026-10-16T04:03:00.000Z","own":false,"moderatable":false,"seed":true,"ownerAuthor":false},{"id":"seed-narcan-is-destroying-price-discovery-in-the-heroin-market-4","author":"recovered_and_tired","text":"After my overdose I did not want to recount the purchase to acquaintances either. I had other things to do. Nobody had paid for an option on my story.","parentId":null,"score":87,"date":"2026-10-16T08:16:00.000Z","own":false,"moderatable":false,"seed":true,"ownerAuthor":false},{"id":"seed-narcan-is-destroying-price-discovery-in-the-heroin-market-5","author":"posterior_potency","text":"I was asking for information he already possessed, not a retrospective consultancy. I accept that recovery may have changed his priorities, although that is precisely why waiting reduces the value of the account.","parentId":"seed-narcan-is-destroying-price-discovery-in-the-heroin-market-4","score":8,"date":"2026-10-16T10:13:00.000Z","own":false,"moderatable":false,"seed":true,"ownerAuthor":false},{"id":"seed-narcan-is-destroying-price-discovery-in-the-heroin-market-6","author":"noise_floor","text":"The sister is trying to keep her brother alive. You have offered her a data-collection job. Her objection sounds remarkably well specified.","parentId":null,"score":36,"date":"2026-10-16T21:45:00.000Z","own":false,"moderatable":false,"seed":true,"ownerAuthor":false}]}