By Parker Rogers

Parker Rogers: You’ve spent much of your career studying policies intended to change unhealthy or risky behavior (like use of opioids, cannabis, and prescription drugs). After all of that work, what do you think economists most misunderstand about why people engage in behaviors that are bad for them?

David Bradford: I’ve spent probably the last 15 years doing work on that, honestly, because of my daughter, Ashley Bradford. She came to me as an undergraduate, knew she wanted to get her PhD, and wanted to work together. She said she wanted to study cannabis policy. I said, “I have no idea about cannabis policy, but let’s go for it.” It turned out to be a fascinating area. That naturally led to opioids, which linked back to a longtime research interest of mine: the pharmaceutical industry. What do people misunderstand? I think there are two things that economists probably understand, but that aren’t always top of mind when we do this work.

The first is that people don’t spring into life with an opioid use disorder (OUD) or an alcohol use disorder. They get there because these substances do, in fact, confer benefits. People don’t initially take heroin because they have an OUD. They take it because it has benefits for them. Economists often downplay the positive aspects of smoking, alcohol, or heroin use. Yes, the end can be bad, but people start because the beginning is presumably quite good. When we design restrictions intended to stop these behaviors, we sometimes neglect the fact that people are getting benefits from them—and people are going to seek things that yield benefits.

The second thing, particularly with something like heroin and OUD, is that we often misunderstand why people continue using once they have full-blown OUD. We tend to think they’re still taking heroin or fentanyl because they’re trying to get high, when in fact they may be desperately trying to avoid withdrawal symptoms. At that point, it’s no longer a model of pursuing pleasure. The behavior has changed from pleasure-seeking to pain avoidance. We tend not to model the pleasure at the beginning, and then we ignore the fact that the motivation can change fairly quickly from pleasure-seeking to pain avoidance. I don’t think our conceptual frameworks take that into consideration often enough.

Rogers: What does that imply for policy? Are there reforms that follow from that insight that economists—or policymakers more broadly—might resist?

Bradford: I don’t know that economists per se resist them, but one policy implication is harm reduction early on: giving people access to less harmful options for the kinds of benefits they’re seeking. That’s one reason I’ve continued working in the cannabis space. I have some hope—and I think there’s evidence suggesting it’s a reasonable hope—that cannabis can provide a safer alternative to opioids for some of the benefits people are trying to obtain. Cannabis is not a risk-free substance. No substance that alters your brain is. But it is less risky.

Harm reduction isn’t particularly controversial among economists, but we get slapped around so often when we bring it up in the popular debate that we sometimes stay away from it. On the front end, harm reduction is useful. On the back end, I think economists should be more aggressive about being honest about the benefits of things like safe injection sites and providing noncriminalized, safe places for people to use these substances in ways that allow them to interact with the medical and treatment community.

One of the great benefits of a safe injection site is not just that people are less likely to transmit disease or die from an overdose. Every time they come in, the treatment community has access to them. You can give them a cessation message. Eventually, that message may arrive at the moment when the person is receptive to it. You can’t force somebody to quit, and you don’t know when they’re going to become receptive. So give them that message every day if you can. One way to do that is to provide a venue where they are regularly interacting with the treatment community.

I think we need to be willing to be a little unpopular among people with ideological views on the subject. Our goal is to have people not misuse substances and die, and we think we have evidence about approaches that can help achieve that. On the right, there’s a kind of Puritanism that says you shouldn’t have pleasure. On the left, the criticism can be, “You’re just giving up on these people.” I think both sides are wrong. We’re not giving up on them. And at this stage, they’re not necessarily enjoying themselves either.

Rogers: Let me push that in a more philosophical direction. If these substances really do provide benefits or pleasure, do we simply want to maximize that? Imagine, at the extreme, that you could hook people up to a machine that produced continuous pleasure without killing them. Would that be desirable? Where do you draw the line?

Bradford: I think that takes us outside economics at that point. Philosophically, I can’t imagine that, as a species, we’ve evolved in a way that suggests that would be good for us. I probably have enough of a Puritanical streak in me from growing up in this country, with all the burdens associated with the Protestant work ethic, to think that struggle is valuable. I don’t know that pure happiness is, in fact, something people find particularly fulfilling. You need balance in life, and sometimes that implies pain. Think of an athlete. They’re pursuing that moment of victory, but there’s a lot of pain leading up to it—pain they willingly embrace. I don’t think hooking us up to some kind of brain stimulus that just leaves us sitting there happy is the answer. That’s kind of The Matrix, right? Everybody was basically happy, but they were stuck in a pod.

Rogers: Suppose tomorrow health economists lost access to claims data. In what ways do you think the field would be intellectually better off ten years from now?

Bradford: We get lazy as health economists. We always have, in part because we have access to much better data than almost any other kind of economist. Our data are extraordinary simply in terms of how much of it there is. So how might losing claims data make us better? We would have to think about things more carefully. We’d probably have to go out and create more of our own data and rely more heavily on surveys to reproduce some of the information we currently get from claims. But in doing that, we would also ask lots of other questions that would give us insights we simply don’t have from claims data.

When the claims data are sitting there with, say, 20 billion observations, we can estimate zero with such incredible precision that we get stars on every table we make. It makes us a little lazy. We look at what we can answer with claims data rather than asking questions about things that have actually been some of my longest-running interests and frustrations with the discipline.

How do time preferences form and evolve, and how does that affect choices? How do people perceive risk? How do risk and time preferences interact? And how does that unified thing affect the choices people make and the outcomes they have in their lives? I think I could motivate almost any economist to be interested in those questions in about 90 seconds. But then you stack that up against a great dataset that can’t answer those questions but can answer a whole bunch of other questions cheaply. It’s hard.

So I don’t want to lose access to claims data. But if we did, I think we would explore broader topics. There is another risk from relying so heavily on claims. It can almost turn health economists into health-services researchers or health-administration researchers. I left the Medical University of South Carolina, where I was for ten years, in part because I could see my CV becoming that of a health-services researcher. I was in a world where funding the work was very important and using all the claims data we had at our feet was very important. It can be hard to be an economist in that environment.

Rogers: What would be your dream research project if you had complete intellectual freedom and essentially unlimited resources?

Bradford: I would like to push economics beyond our traditional framework for thinking about human decision-making. That framework has been enormously productive, but it can also be narrowing. I’m interested in whether insights from neuroscience could give us a stronger theoretical foundation for things like risk preferences and time preferences.

Take time preferences. Economists have very well-developed ways of incorporating them into our models—we apply exponential or hyperbolic discounting to utility over time. But we have much less to say about where those preferences actually come from. Are time preferences static? Can they change? Can they be modified? How do they evolve over a person’s lifetime? We don’t have particularly satisfying answers to those questions.

I think a framework that takes more seriously how the brain actually processes information and makes decisions—Bayesian updating, for example, and the tendency of organisms to move toward homeostasis—might give us a better theoretical grounding for those preferences. Not because a model needs to be realistic for realism’s sake, but because it might help us answer questions that our current framework largely leaves open.

And yet virtually every single decision humans make in a modern economy is time-dependent. If I buy a can of Coca-Cola today, I’m not necessarily giving up something at that instant to get it. I put down my credit card, and sometime later money gets transferred out of my bank account. I’m trading present consumption against a future payment. There is a time dimension to essentially every economic decision we make.

And yet I think time is by far the weakest aspect of microeconomics. Full stop. Across the board. That has frustrated me since I was in graduate school. The empirical evidence also gives us reason to question the assumption that time preferences are static. Measured time preferences appear to change over the life cycle. If an 80-year-old displays a very high discount rate, are we supposed to believe that person had exactly the same discount rate when they were 20? That seems implausible. Yet we routinely treat these preferences as fixed.

So if you gave me a three-year sabbatical and all the assistance I can now get from AI for the subtle mathematical points I’ve forgotten since getting my PhD in 1991, I would spend my time working on that.

Rogers: What important question in health economics is surprisingly easy to make progress on, but people aren’t working on it?

Bradford: Theory. And it’s because we have all the data. This gets back to your question about what happens if claims data disappear. Look at our incentives. Our incentive is to publish as many really good papers as we can. There are a couple of paths to doing that. One is to sit and think really hard for a year and come up with a theoretical contribution that might actually change the way the discipline thinks about something. But that’s a risky investment—you still have to get it published. The other path is to get claims data and potentially produce several strong empirical papers.

So the incentives run against theory. But all you need for theory is a pad of paper, a few pencils, and some time. You can sit there, imagine these great worlds, and come up with some help for those of us who have to be empiricists because we’re just not qualified to be theorists. That’s why I think it is surprisingly easy in one sense. You don’t need grant funding. You don’t need complicated, big computers. You don’t need million-dollar datasets. You need five dollars’ worth of paper and pencils. We need more of it.

Rogers: In some ways, the relative costs have flipped. When data were difficult to obtain and computation was expensive, theory was comparatively cheap. Now we have enormous amounts of healthcare data, and empirical work has become much cheaper.

Bradford: Exactly right. Health economics has always had better data, and we’ve always been a much more empirical discipline than our fellow disciplines. And to some degree, if you’re in a pure economics department, people can look at health economists a little askance. I think some IO economists look at us and think, “You haven’t done the hard work to know what your theory ought to be”.

I think that criticism is probably correct.