Megan McArdle is one of my favorite columnists, and she has a new gig over at the Washington Post, where she recently wrote an article about a fascinating just-released economics working paper that caught my eye. The authors, Jennifer Doleac and Anita Mukherjee, studied how the opioid addict population has responded to the drug naloxone becoming more readily available.
Naloxone (a.k.a. Narcan) shuts down the effects of opioids, so it is commonly used by EMTs to rescue people who have overdosed on opioids, including heroin. Out of concern that it can take a while for EMTs to arrive, states have passed laws making it easier for ordinary people to get naloxone, allowing addicts to be rescued by family, friends, and fellow opioid users.
The Doleac-Mukherjee study explores the possibility that this is creating what economists call a “moral hazard.” The basic idea is that naloxone reduces the risk of dying, and the risk of dying is part of the non-financial cost of using opiods, so naloxone effectively reduces the price addicts pay to get high. The law of demand predicts that when the cost of something goes down, people will buy more of it. So making naloxone more available could cause an increase in opioid usage.
If you’re not used to economic thinking, this might seem crazy. It’s hard to imagine addicts thinking, “Since people around me have naloxone, I can use more drugs!” It’s even harder to imagine that someone who has never used opiods will suddenly think, “Hey, with all this naloxone around, I think it’s time to try heroin!”
What makes this theory plausible, however, is that economic effects happen on the margins. Every day, some Americans decide whether or not they will use heroin that day. Some say “yes,” and some say “no.” And it’s reasonable to assume that some people who say “yes” are just barely saying “yes,” and some who say “no” are just barely saying “no.” Those are the people most affected by changes in incentives: Give them a slight nudge in the other direction and they might make the opposite choice. So while most heroin non-users will not be swayed by naloxone, there are likely to be some who change their mind.
Similar effects have been seen in other contexts. The most famous example is when seat belts were required for automobiles. Economist Sam Peltzman predicted that since seat belts made accidents less dangerous, people would start to have more accidents, and subsequent studies have confirmed this prediction. The effect is not huge, but it’s real. In economic terms, since seat belts reduced the “cost” of accidents, people “bought” more accidents.
That’s not to say that people literally make a conscious decision to have more accidents. But knowing that the seat belt will protect them from minor accidents, they may decide to drive faster, or they may feel it is safer to fiddle with the radio or talk to passengers more often.
If that still seems unlikely to you — if you’re sure that you would never drive recklessly just because your car protected you better — then consider a scenario suggested by economist Gordon Tullock: Disable your air bags, remove your seat belts, and mount a seven-inch steel spike in the center of your steering wheel, pointed straight at your chest. Now ask yourself how your driving would change. Given that even a moderate accident could drive that spike into your heart, you will probably drive a lot more carefully. But if you agree that making your car more dangerous would make you drive more carefully, then you must agree that the level of safety you currently enjoy is allowing you to drive less carefully. You are responding to an incentive by driving more recklessly.
(Use of the name “moral hazard” for this phenomenon is unfortunate, because it implies a judgement against some measure of morality, and that’s not usually how economics works. The term comes from the insurance industry, which has to worry that car owners who can’t make their loan payments will “sell it to the insurance company” by driving the car into a river and reporting it stolen, or that building owners who are upside down on their mortgage will burn the building down for the insurance money. But the problem also rears its head in more mundane ways, such as when people with theft insurance are more likely to leave their car windows cracked open on sunny summer days, and more likely to leave their car running in the driveway in winter with the keys inside. And yes, people with insurance also get into more accidents. The insurance industry takes a dim view of all this, thus the name “moral hazard.”)
Getting back to opiods and naloxone, there are no data sets that directly report opioid consumption, so the study looked at proxy variables, such as opioid-related crime and opioid-related emergency room visits. Because the dates on which naloxone became available vary from state to state, the study was somewhat able to randomize away confounding variables. E.g. The proxy measures in one state might change because law enforcement anti-drug activity changed at the same time that naloxone became available, but that coincidence would be unlikely to occur in every state. This creates a kind of “natural experiment” that allows for stronger results.
So, cutting to the chase, are addicts using more opioids because of naloxone?
Yes, and to a surprising degree. The Doleac-Mukherjee study observes that the introduction of easily-available naloxone is associated with a 17% increase in arrests for possession of opiods and a 27% increase in arrests for selling opioids. Opioid-related emergency room visits went up 15%.
(There are a few caveats, of course. The study is a working paper that is being released for review, so the results could be subject to revision. Also, as with any study of this kind, the authors make a lot of reasonable-seeming assumptions — which are extensively documented in the paper — some of which might turn out to be wrong in ways which substantially alter the conclusions.)
The opioid-related mortality figures are especially interesting, and I’ll get to them in a minute. But I want to address one part of the study that’s been upsetting people on Twitter:
Naloxone access may unintentionally increase opioid abuse through two channels: (1) saving the lives of active drug users, who survive to continue abusing opioids, and (2) reducing the risk of death per use, thereby making riskier opioid use more appealing.
To be very clear, contrary to what alarmed Twitter users are claiming, the study authors are not saying that saving the lives of drug users is a bad thing.
The reason they bring up this issue is because the study is intended to look at increasing opioid use due to the “moral hazard” behavioral change (channel 2), but opioid use could also increase simply because longer-lived drug users have more time to consume drugs (channel 1) and the study cannot directly distinguish between the two explanations. It would be dishonest not to mention this.
In any case, when we look at the mortality figures, Doleac and Mukherjee discovered something astounding:
On average across all urban areas, we find that these laws have no significant impact on the opioid-related death rate. Thus, while the risk per use has gone down due to Naloxone access, the number of uses increases enough that we find no net effect on opioid-related mortality.
In other words, in response to the availability of naloxone, opioid users appear to have increased their use of the drugs so much that it completely negates the life-saving benefits of naloxone.
The magnitude of this effect was a surprise to the researchers. In an email message, Dr. Doleac explained what it means for the first causal channel:
When we first started this paper, we expected to find a big decline in mortality, in which case that channel could be very important. Given that we don’t find a net decline in mortality, it is probably less important[.]
In her article about the Doleac-Mukherjee study, Megan McArdle discusses possible policy responses to this startling finding:
The coldly logical response to this would seem to be to discontinue naloxone use. But there’s something repulsive about that conclusion, and Doleac and Mukherjee can’t bring themselves to go there. “Our findings do not necessarily imply that we should stop making Naloxone available to individuals suffering from opioid addiction,” they write, “or those who are at risk of overdose. They do imply that the public health community should acknowledge and prepare for the behavioral effects we find here.”
Sally Satel echoes Doleac and Mukherjee, both on the moral hazard of naloxone and on whether access to it should continue. Satel, a psychiatrist who is also a drug policy scholar at the American Enterprise Institute, says the paper’s findings reinforce what she has heard from patients: “Patients occasionally tell me that having naloxone on hand has served as insurance against overdose. So, in some instances, it enhances risk taking.”
“That said,” she emphasizes, “we must use it to save people in the immediate term.”
So how can public policy prepare for those “behavioral effects” found by Doleac and Mukherjee? Satel suggests we look at civil commitment for patients who overdose multiple times in a short period. But she also notes that civil commitment can’t work without good treatment options — and in a lot of places, those aren’t available.
Which brings us back to something that’s easy to forget about the Peltzman Effect: It can be used to argue as much for more regulation as for less. Insurance companies, after all, have been fighting moral hazard for centuries, which is why they reward people who install burglar alarms or fill in their swimming pools (or punish people who don’t do those things). And so, too, can the government — for example, by aggressively ticketing speeders, passing tougher drunken driving laws, or using a combination of carrots and sticks to help addicts get clean. There are better policy responses to moral hazard than mounting a spike on the steering wheel — or depriving addicts of a second chance at life.
Much of that seems reasonable, but I wish McArdle had considered something else that’s easy to forget about the Peltzman Effect: It’s not necessarily a bad thing. When surgical anesthesia became a lot safer during the last century, we began to do a lot more surgery. When Elisha Otis invented the safety brake, people all over the world started riding elevators. As air travel becomes safer, more people flew through the sky. Doing more of what we want is usually considered a good thing.
That’s because economists usually make a formal assumption that people will act rationally to improve the quality of their lives. So if people decide to consume more of some goods — apples, oranges, antibiotics, haircuts, televisions, surgical anesthetics, elevators, airplanes — economists assume they do so because it makes their lives better. So why shouldn’t economists be happy that naloxone allows heroin users to consume more heroin?
There are two common answers to that question. One is to reject the idea that heroin is a “good.” This is a common position in public policy analysis, where certain actions, such as prostitution or consumption of illegal drugs, are deemed to be axiomatically bad. Doleac and Mukherjee kind of take this route by default, never really considering that increased opioid use may be a benefit. However, being good economists, they do allude to the possibility, noting that “welfare implications of drug and alcohol abuse themselves are unclear: Some argue that people can do whatever they want to their own bodies, no matter how harmful.”
McArdle takes the other route, arguing that because drug use is addictive, it’s not an actual choice:
It makes a certain amount of sense that the Peltzman Effect would show up particularly strongly in drug users; after all, drugs hijack the brain’s reward system, redirecting it toward drug-seeking even at high personal risk.
In other words, drug addicts fail to be rational in pursuing the improvement of their lives: Their addiction takes away their ability to make rational choices, so their apparent choices no longer represent consumption decisions that improve their lives. Somebody has to step in and make those choices for them.
McArdle’s argument is somewhat undermined by the Doleac-Mukherjee study itself:
It may seem surprising that drug users respond to incentives in a sophisticated way. One may think that drug users are poor decision-makers or that addiction makes rational choices impossible. Addiction surely clouds judgement and makes policy in this area difficult, but there is substantial evidence that even drug users respond to incentives. A large body of empirical evidence documents that the consumption of addictive substances is sensitive to prices. For example, increasing taxes on alcohol reduces alcohol consumption (Cook and Durrance, 2013). Alcohol abuse also responds favorably to increasing the likelihood of punishment, as seen in evaluations of the 24/7 Sobriety program (Kilmer et al., 2013). Hansen, Miller and Weber (2017) show that marijuana consumption is price inelastic in the short run, but quickly becomes price elastic, with consumers reducing their consumption in the face of higher marijuana taxes. And finally, Moore and Schnepel (2017) show that a massive reduction in the heroin supply in Australia resulted in a long-term reduction in heroin consumption among those using heroin at the time, due to a spike in the price of the drug. These findings suggest that, at least on the margin, drug abuse may be sensitive to non-monetary costs such as the risk of death.
If drug users are rational enough to alter their behavior to respond to these changes in incentives, that undermines the argument that their decision to consume drugs is irrational, leaving the way open for the possibility that increased opioid use is actually a benefit of naloxone.
That may seem a little insane, because, you know, it’s heroin. But suppose we could figure out a way to make opiates completely safe to use. This would likely cause a big increase in heroin use, but I hope you’ll agree that an increase in the use of completely safe drugs is not a health crisis. Nothing is completely safe, of course, but maybe between naloxone and some other policy changes, we can get close enough to greatly improve the lives of recreational opiate users.
Some nations have experimented with supervised injection centers which provide sterile equipment, trained staff, and of course a supply of naloxone.
Studies consistently show that supervised consumption facilities work. These kinds of sites have opened in Canada, Australia, and Europe, showing drops in drug overdoses, related emergency care calls, risky behaviors that lead to HIV or hepatitis C transmissions, and general public disorder and nuisance associated with drugs.
(I don’t know if there are studies on whether supervised injection centers lead to increased opioid use, but I wouldn’t be surprised if they do.)
One thing these centers do not provide are the drugs themselves, so addicts still face risks inherent to a drug supplied with uncertain concentrations, uncontrolled contamination, and unpredictable adulterants. At the risk of yet another increase in consumption, we could conceivably control all of those problems — and reduce the need for addicts to commit crimes to feed their habit — by permitting the development of a legal source of drugs that is inexpensive, uniform, sterile, and…hmm…I think I’ve just invented the friendly neighborhood Heroin Bar.