Search the Site

Episode Transcript

In 2008, the state of Oregon decided to do something no state had done before. It expanded its Medicaid program through a lottery, selecting names from a waiting list to fill a limited number of available spots. Oregon did this because it couldn’t cover everyone and a lottery seemed like the fairest way to allocate resources. But in doing so, by chance, it also set the stage for a one-of-a-kind health insurance experiment.

BAICKER: My next-door office neighbor at the National Bureau of Economic Research was Amy Finkelstein. She came into my office and said, “Have you heard about this Oregon lottery? That sounds like a randomized controlled trial” So we were off to the races.

That’s health economist Kate Baicker. She’s Dean of the University of Chicago’s School of Public Policy. So, what was so special about the Oregon Medicaid lottery?

BAICKER: This was an unprecedented chance to look at the effect of Medicaid in a group of people where everything else about them was the same and some, by luck of the draw, get access to the health insurance plan and others don’t.

This experiment allowed Kate, Amy, and their colleagues to answer what may be the most fundamental question in the field of health policy; a question that had never before been tested in a randomized controlled trial; and which is more relevant now than ever, as the share of U.S. adults that are uninsured is at a record low. Does health insurance make us healthier?

BAICKER: It is an obvious question to ask, but it is a much harder question to answer than you might think.

From the Freakonomics Radio Network, this is Freakonomics, M.D. I’m Bapu Jena. Today on the show: I talk with Kate Baicker about why that’s such a hard question to answer.

BAICKER: “If you don’t take all of those factors into account, you’re going to think that health insurance has a very different effect.”

And what we should take away from the findings of the Oregon Health Insurance Experiment.

BAICKER: Everybody is trying to paint it as a black or white decision, and we live in a world of gray trade-offs.

*      *      *

BAICKER: I went straight to grad school from college, which I do not advise people to do. It worked out okay for me, but boy, I had no idea what I was doing.

Kate didn’t start off as a health economist. As a graduate student, she studied issues like taxation, public spending, and labor economics. But it wasn’t long before she shifted her focus to healthcare.

BAICKER: There was a lot of interesting work going on about geographic variation in healthcare spending, how we spend three times as much in some parts of the country as other parts of the country. And people don’t start off sicker. They don’t end up healthier. It’s just a sign that the system’s not working very well.

JENA: Can you give me an overview of how health insurance works in the United States? You know, just a one-minute elevator pitch?

BAICKER: One minute? I thought we were gonna devote the whole show to how health insurance works in the U.S. because it is a tangled patchwork.

Kate’s right; there’s nothing uncomplicated about health insurance in the U.S. First, there are public programs, like Medicare and Medicaid, that cover specific populations. Then, there are private insurance programs, which people can purchase on their own but usually receive through their employer. But there are also millions of people in the U.S. who are uninsured.

BAICKER: And they get access to some care through emergency rooms or social safety net hospitals or charity care, but they don’t get the same access to care that people with insurance do.

JENA: In health policy, you would then think that a fundamental question would be: how does health insurance impact health? And I think people would think that the answer is obvious, right? Healthcare should improve health and health insurance should give you access to healthcare. So why is it an important research question and why has it been difficult to study?

BAICKER: People who have health insurance look different from people who don’t have health insurance in lots of ways. Suppose you look at people who are on Medicaid and compare them to the uninsured. People who are on Medicaid have often much worse health outcomes. They have higher mortality rates. So if you just did that naive comparison and said, “What is Medicaid doing?” You think Medicaid is a terrible program. It is killing people, and we should end it right away. But of course, that’s not the right way to interpret that data because the people who are on Medicaid are also potentially low income. Being low income is hard on your health, independent of health insurance. Or they’re disabled. Being disabled is inherently hard on your health. And so if you don’t take all of those factors into account, you’re going to think that health insurance has a very different effect.

This is a straw man example; no serious researchers would design a study that way, for the reasons Kate describes. So, what would they do?

BAICKER: First thing you can do is throw in a bunch of controls. Like, if the income is different, we’ll control for income. If baseline health is different, we will take people with the same baseline health conditions. You can see it’s harder and harder to really measure all of those things and take them into account fully. So still a potentially weak design to really isolate cause and effect. Quasi experimental approaches take a more sophisticated strategy in trying to isolate those causal connections. You could compare the population in the state that expanded Medicaid to the one that didn’t. But you still have to ask: “Why did one state expand when the other one didn’t?” Maybe there was a health crisis in the state that expanded. If that were the case, health might be worse in that state and you would think, “Oh, the health insurance program isn’t working.” Or suppose a state that chooses to expand health insurance is also expanding food stamps and welfare and education programs. Well then if you attribute all of that to the health insurance program, you might think it’s doing much more than it’s actually doing.

Another strategy might be, you know, people get eligible for Medicare at age 65. Nothing about your body changes on your 65th birthday, so let’s compare 64-year-olds and 65-year-olds. One of the problems with that is maybe you also retire when you’re age 65 and that changes your physical health. So there are limitations to those quasi-experimental strategies. There’s some reasons you might not be really isolating cause and effect. And there are also limits to external generalizability. And those are the two things that we’re always looking for from really good studies: internal validity — have you really found cause and effect? And external validity — does the thing you found apply to anybody else of interest?

We talk a lot on this show about how randomized controlled trials are the gold standard of evidence. In medicine, for example, we would never accept a new drug without testing it in a trial where participants were randomly assigned to either the treatment or a control, which is usually standard of care, but sometimes a placebo; that’s how researchers can assess the actual effect of the drug itself. In the world of public policy though, setting up a randomized trial is more challenging.

BAICKER: If you wanna know the effect of education on kids, you don’t lock half the kids in the basement and not educate them and then wait 30 years to see what happens to their income and wellbeing. That would be wrong. And so we don’t often have randomized controlled trial evidence in public policy — but there are exceptions. One of the earliest true randomized controlled experiments in health economics was the RAND Health Insurance Experiment.

The RAND Health Insurance Experiment was conducted from 1971 to 1986 and directed by the economist Joe Newhouse. To this day, it’s the largest health policy study in U.S. history.

BAICKER: And what that studied was the effect of different patient copays or cost sharing on healthcare use and health outcomes. And they found that when patients have to pay more for healthcare, they use less of it. Now, for anyone who’s taken econ 101, higher prices mean lower demand. But it actually is surprising in healthcare because I don’t think that’s how most of us imagine consuming healthcare. It’s not like I, you know, come knocking at your door when we’re both working and say, “Hey Bapu, did you hear MRIs are on sale on floor two?”

JENA: Wait, wait, They’re on sale? How much?

BAICKER: “Let’s go get an MRI at lunch. This sale’s not gonna last forever.” Like, that’s not how people purchase healthcare. And so it was surprising to a lot of people that patients were sensitive to prices in that way — even for drugs that are potentially life saving! I have to say, cause rumor has it, you’re also a doctor, although I myself don’t believe it.

JENA: Rumor has it.

BAICKER: It turns out doctors are also sensitive to prices. When you pay doctors more for stuff, they do more stuff. Nobody thinks of physicians as making decisions based on those payments. But doctors are people too, and they also respond to the incentives that the system generates. So the Rand Health insurance experiment really was foundational to a whole line of economic inquiry But everybody in Rand had insurance. And so Rand gave us a lot of information about how to structure health insurance, but it didn’t tell us what the effect of having insurance versus not having insurance was.

There weren’t any randomized controlled trials that could do that … until that fateful day that Amy Finkelstein knocked on Kate Baicker’s door to talk about Oregon. In 2008, enough new revenue became available that the state could afford to cover 10,000 more low-income adults in its Medicaid program. But, there were a lot more than 10,000 people who qualified. So, they tried to think of the fairest way to select people.

BAICKER: The problem with first come first serve is that you’re likely to favor people who have internet connections or who are more tied into the social safety net or who have more education. And so in the end they decided they’d open a waiting list and they advertised really heavily and they got almost 90,000 people. And then they drew names from the waiting list by lottery.

You already know what happened next; Amy knocked on Kate’s door, and they were off to the races.

BAICKER: We called everybody we knew who’d ever been to Oregon and found our way to the Oregon Medicaid office.

Their research team collected data around two years after the lottery selection occurred, so they could see the effects of having insurance for two years versus not having insurance for two years. The experiment ended in 2010 when the Affordable Care Act was signed into law, since it expanded coverage to many more people on the waitlist, meaning the “control group” was no longer a control group.

BAICKER: One of the principles that drove our data collection is: can we get data on this outcome for both people who got Medicaid through the lottery and those who didn’t, because if you only get data on the people who have insurance, you’ve got nobody to compare them to.

JENA: So you surveyed people who had received Medicaid through the lottery and you surveyed people who by chance did not receive Medicaid through the lottery.

BAICKER: Exactly. We surveyed the whole lottery list — the “treatment and control group”. One of the pathways that we expect insurance to affect health outcomes is through healthcare use. So we got data on all of the hospitalizations and hospital discharges in the state of Oregon. We got data on emergency department use for the greater Portland metro area. We got data on physician outpatient use by asking people about how often they had been to the doctor. We got information on all of the prescription medications that people were taking. We also got information on people’s financial wellbeing. I think that gets lost in a lot of the debate about health insurance reform in the U.S. Yes, of course, it’s about access to healthcare, but insurance is also a financial protection product. Then of course, health outcomes. We collected information from mail surveys, but doing these in-person interviews let us collect physical data like blood pressure, obesity, cholesterol, diabetic blood sugar control, and also let us ask much more in-depth questions and follow up questions based on what people said. We also were voracious consumers of available data. Anything that could get matched to our data set, we matched to our data set so that we could look at the effect of insurance on interactions with the criminal justice system, on voting, on anything that we could measure. So we have a really wide range of outcomes to look at.

After the break, what did Kate and her colleagues find?

BAICKER: It is a much more complicated story than simply saying, “Yes, insurance worked,” or, “No, it didn’t work.”

I’m Bapu Jena, and this is Freakonomics, M.D.

*      *      *

BAICKER: You would expect when you take something that was expensive and make it affordable, people will use more of it. And that is exactly what we found.

When Kate Baicker and her collaborators conducted the Oregon Health Insurance Experiment back in 2008, they were more surprised by some findings than others.

BAICKER: People went to the doctor more, they used more prescription medications. They reported that they had better access to care, that the care was of higher quality. They went to the hospital more. And these hospital admissions were largely of the kind recommended by their doctor. So all of that I think was as expected by most people, although we didn’t know what the magnitudes would be.

The Oregon experiment is important for a few reasons, most notably because it’s the only randomized controlled trial to date testing how health insurance impacts people’s lives. As Kate just said, giving people insurance led them to go to the doctor more, go to the hospital more, and to use more prescription medications. But it also led to some other, less predictable outcomes.

BAICKER: The thing that was surprising to a lot of people is that they also went to the emergency room on average, 40 percent more when they had insurance than when they were uninsured. And this caused a lot of head scratching and hand ringing among those who were advocating for expanding insurance, partly on the grounds that it would make healthcare so much more efficient because we would get people out of the emergency room. Where we saw the increase in use was for the more discretionary types of visits. If you get hit by a bus, you’re going to the emergency room, insured or not. If you have a really, really, really bad headache and you’re not sure, is it just a headache or is something going on? If you don’t have insurance, you may be more hesitant and you may wait to see what happens.

You might think, “Well, maybe some people went to the doctor more when they got Medicaid and others went to the E.R. more because not everyone could find a doctor.” But you’d be wrong.

BAICKER: What we found was that when you got insurance, you were more likely to go to the doctor, you were more likely to go to the emergency room, and you were more likely to go to the doctor and to the emergency room. If anything, insurance was making these two kinds of care more complimentary, more likely to go together, versus more substitutable. The people who were least surprised by this were physicians and healthcare providers. When we talked to them, they said, “If I have a patient who has a problem that I just can’t deal with in the timeframe it needs to be addressed, and I tell that patient to go to the emergency room, my insured patients go, and my uninsured patients don’t.”

JENA: This finding is really emblematic of a broader issue that we often face in medicine and health policy, which is whether preventive care can be cost saving. Preventive care can improve health outcomes, but that doesn’t mean that it’s gonna save you money. It also doesn’t mean that you shouldn’t do it, if it doesn’t save you money. If you give people primary care through insurance, it doesn’t mean it’ll keep them out of the E.D. But the fundamental thing is you’ve given them access to better primary care. You’ve given them the opportunity to go to the emergency department if they need that acute medical care.

BAICKER: I agree and I think a lot of people hung their hopes for expanding insurance on the idea that it would somehow pay for itself because you would cut out emergency department use and save so much money that way, that that money would then pay for all the other healthcare. It’s a much easier case to make if you can say it’ll also save money, so then you don’t force people to decide how much they value healthcare for other people. The only problem is that it’s just not true. When you expand insurance, people use more healthcare and that costs money. But think about something like food stamps. Nobody says, “We expand food stamps and people will be so much better nourished that they’ll be much more productive at work and they’ll pay more in taxes and those taxes will pay for the food stamps.” Well, no. People won’t be as hungry. That’s the whole point. And so I think holding up the idea that it should save money, in my view, is not the right social metric. The right metric is, is the cost greater than the benefits?

Kate and her collaborators also looked at how having health insurance impacted people’s financial security. And there, the evidence was clear. People who got insurance were much better off financially. They were less likely to have to borrow money from friends and family, or to skip paying other bills, because of their healthcare expenses. They were also 25 percent less likely to have a bill sent to collection. What about health outcomes though? How did those change with access to insurance?

BAICKER: There we have kind of a nuanced story. When we asked people about their health, they reported that their health was much better, those who had insurance relative to the control group. They reported they’re much more likely to have their healthcare needs met. They reported they missed fewer days of their usual activities because of health. We also did a clinical assessment of depression, and we found that Medicaid substantially reduced the incidence of depression. There was something like a 9 percentage point or 30 percent drop. So this is an enormous mental health benefit, and that’s a huge unmet health need in this population. But when we measured things like blood pressure, cholesterol, diabetes, obesity, we didn’t detect any improvements in those chronic physical health outcomes and this again, caused some consternation among proponents of expanding health insurance. It is a much more complicated story than simply saying, “Yes, insurance worked,” or, “No, it didn’t work.” It had some really beneficial effects, but it didn’t solve our problems with chronic physical health conditions.

JENA: Why is it that the access to insurance, the access to more healthcare, in your view, didn’t move the needle on things like obesity, high cholesterol, high blood pressure?

BAICKER: I don’t think that it’s because we didn’t study it for long enough. That was one of the questions that people raised, but for things like blood pressure, there are medications available that if they’re used as prescribed, they’re gonna work in less than two years. And we didn’t even find people having those medications at a higher rate when they were insured. People who got Medicaid got care in the same system and it does not seem like we adequately manage those chronic physical health conditions in most segments of the U.S. population. So in some ways it is not so surprising that Medicaid didn’t cure diabetes because neither has Medicare or private health insurance. We have a serious problem with these chronic physical health conditions in the U.S., but Medicaid alone was not sufficient to address it.

Earlier, Kate said that two things we look for in really strong research studies are internal validity —

BAICKER: Have you really found cause and effect?

— and external validity.

BAICKER: Does the thing you found apply to anybody else of interest?

So, how did the Oregon Health Insurance Experiment hold up?

BAICKER: The people who got coverage in Oregon look a lot on many dimensions, like the populations who would get coverage under Medicaid expansion under the A.C.A., except Oregon is a lot whiter than many of the other states. And so if you’re interested in, for example, racial disparities and access to care, Oregon’s not a good place to study that. So there’s some limits to external generalizability, although some strengths as well. Whereas I think we have really robust internal validity. We really nailed down cause and effect. It is not the end of the story and complimentary studies help fill in the picture and also give information from other contexts that I think it’s important for policy makers to put together in thinking about what to do now in the country or for their state.

Kate’s expertise in healthcare policy extends beyond her academic work; she spent two years serving on the Council of Economic Advisors. That’s a group of three economists who are charged with advising the president on all economic policy decisions. In that group, Kate was specifically responsible for government spending programs like unemployment insurance, public education, food stamps, welfare, and — of course — healthcare. So, I asked her, as someone who’s been in the room where it happens, what she thinks her findings mean for health insurance policy in the United States.

BAICKER: Everybody is trying to paint it as a black or white decision, and we live in a world of gray trade-offs. Policy makers and journalists and even podcast hosts sometimes ask me like, okay, so does this mean we should expand Medicaid or not? And my answer is always, “It depends on what your policy priorities are.” The evidence to me seems decisive that being insured is substantially better for you than being uninsured. But it also seems clear to me that expanding public insurance costs money. If we were in a simple world where Medicaid didn’t seem to help anybody, it’s easy to vote no. And if we were in a simple world where Medicaid seemed to pay for itself, it’s simple to vote yes, but we’re not in that world. We’re in a world where there are costs and benefits and policy makers and voters have to take a stand about what they care about — and that doesn’t seem to go over very well on a bumper sticker.

JENA: We often don’t connect the issue of waste in U.S. healthcare to the issue of health insurance for the poor. But they’re related in a really important way. If it’s expensive to be altruistic, then you’re going to be less altruistic. If we were to eliminate or reduce waste in healthcare, what could that do for our ability to expand health insurance coverage to people?

BAICKER: If we can provide better health to people with a more effective healthcare system that uses fewer resources, that’s all the more people who can be covered without crowding out food stamps or housing assistance or roads or public education or all of those other things that we also care about. There’s a lot of public rhetoric about whether healthcare is a right or not. I think that’s the wrong question because healthcare’s not one thing. It’s not like you either have healthcare or you don’t. The real question I think is: how much healthcare is a right? And I think almost everyone would agree that the answer is more than zero. My own view is that it’s really important that everyone have access to care that’s producing substantial health benefit. And to make that financially sustainable, we have to limit public resources going to care that has really questionable health benefit. And that means defining an entitlement that’s substantially more than zero, but less than infinity.

JENA: And then, what is the evidence that’s required to help you make that determination? Because it’s really hard to know what works and doesn’t work. You know, vaccines is a no-brainer, but an MRI for someone who’s had back pain for six months as opposed to two weeks — then it’s a little bit or more of a gray area. You have to have the evidence to be able to figure out what is high value, what is really of marginal value.

BAICKER: Most care is not uniformly high value or uniformly low value. It’s high value in some circumstances and low value in others. Antibiotics, when you have a bacterial infection, really high value. When you have a viral ear infection, probably negative value. And so it requires not only evidence on a specific procedure or drug or mode of care, but also for which patients, and then insurance coverage needs to be tailored accordingly. That is a sticky wicket indeed, but it all starts with the evidence. No one study, no matter how fantastic it is, is gonna be the final answer.

The good news is there’s plenty of innovative research chipping away at these questions. A new study in the American Economic Review found evidence that the health benefits associated with early life Medicaid exposure extend to the next generation of offspring. Kate and some colleagues also have a piece coming out in the winter issue of the Journal of Economic Perspectives, in which they propose a universal health insurance coverage strategy that would reduce inefficiencies in our current system.

That’s it for today’s show. A big thank you to my guest, Dr. Kate Baicker, for joining me. And, as always, thank you for listening. If you have thoughts about this episode, or about health policy in general, I’d love to hear them.

Also, each episode, I try to come up with an idea for a new study based on my conversations. This week, let’s see what you come up with. Imagine that you could link data from the Oregon Health insurance experiment to any other dataset and study whatever question you wanted to. What would it be? Kate and her colleagues have already merged in employment data, to study the impact of health insurance on labor market outcomes, and even voter registration data, to study how insurance affects civic participation, like voting.

Email me your ideas. I’m at That’s B-A-P-U at

Coming up next week: my friend, the master clinician, Dr. Gurpreet Dhaliwal, will try to figure out why a 60-year-old man suddenly collapsed at the supermarket.

DHALIWAL: This is one of the most serious medical emergencies you can have.

Will Gurpreet get it right? Or will I manage to stump him?

DHALIWAL: You can almost take each one of those things and, form a differential diagnosis or a list of possibilities around them

That’s all coming up next week on Freakonomics, M.D. Thanks again for listening.

*      *      *

Freakonomics, M.D. is part of the Freakonomics Radio Network, which also includes Freakonomics Radio, No Stupid Questions, and People I (Mostly) Admire. All our shows are produced by Stitcher and Renbud Radio. You can find us on Twitter at @drbapupod. This episode was produced by Lyric Bowditch and mixed by Eleanor Osborne. Julie Kanfer is our senior producer. Our executive team is Neal Carruth, Gabriel Roth, and Stephen Dubner. Original music composed by Luis Guerra. If you like this show, or any other show in the Freakonomics Radio Network, please recommend it to your family and friends. That’s the best way to support the podcasts you love. As always, thanks for listening.

JENA: I think what you’re saying is that if you were to have a podcast, having a podcast about health and economics is just a no-brainer.

BAICKER: Why would anyone have a podcast?

Read full Transcript


  • Katherine Baicker, dean of the University of Chicago Harris School of Public Policy.



Episode Video