Research Corner: Picking Plans
Season 1: Episode 76
October 13, 2020
We return to the Research Corner where health economist Sayeh Nikpay discusses two research papers about picking health insurance plans: why it’s so difficult to select the best plan and why so many uninsured Americans who are eligible for low-cost plans don’t sign up.
Dan Gorenstein: Hey, it’s Dan. I want to welcome you back to our Research Corner segment where we discuss health policy’s newest, most interesting studies, and what they mean for us. Today, we catch up with our occasional co-host University of Minnesota health economist Sayeh Nikpay.
From the Annenberg Studio at the University of Pennsylvania, I’m Dan Gorenstein, and this is Tradeoffs.
DG: Hey Sayeh, how are you?
Sayeh Nikpay: Hey Dan.
DG: Sayeh, great to see you. So it’s fall. That means it’s the season to sign up for health insurance! I even saw an ad on TV before the VP debate trying to get people to sign up for health insurance. And in that spirit, you’ve brought us two papers about picking health insurance plans.
SN: So to some people, fall is pumpkin spice lattes, it’s decorative gourd season, but to health economists it is open enrollment time.
SN: And it’s a very exciting part of the year.
DG: Are you serious? Do economists get excited about open enrollment?
SN: Oh yeah, definitely.
News clip montage
SN: I mean, it’s all the drama of health economics. It’s about choice. Are you going to choose a new plan? Are you going to stick with what you had? It’s really exciting.
DG: Ok, so I can believe as an economist you and Bacon, also an economist and your husband, geek out over open enrollment.
But as a regular person, is it easy for you to actually shop for plans?
SN: Well, even for a married couple where both people have PhDs and training in economics, it’s really hard. We just moved here to Minneapolis to take new jobs and between the two of us, we’ve got about 13 different health insurance plans that we’re picking between. So we can do the easy thing, which is looking at the premium — that’s how much I pay per month — and we can look at deductibles — that’s how much I have to pay out of pocket before health insurance actually kicks in. But it’s way more complicated than just looking at those two factors. If I want to take a good health insurance plan, I really have to think about how much money I’m going to spend over the year, how much health care I’m actually going to use. And research tells us over and over and over again that we, including highly skilled workers, are really bad at picking health insurance.
DG: And that’s where this paper comes in, right? The question is what happens if people get help from an expert — in this case an insurance broker — and what happens if that insurance broker expert gets help from artificial intelligence.
SN: Yeah, so it turns out, a finding from wider economics literature is that experts also don’t do a good job at picking insurance. And so what this paper is doing is pairing these experts — health insurance brokers — with a tool. This tool is artificial intelligence, so it’s a sophisticated algorithm that helps them pick insurance plans. And they’re trying to figure out with a randomized controlled trial: do the brokers with the AI tool do better at picking insurance than those who don’t?
DG: Sayeh, so what did they find?
SN: So they found that when brokers didn’t use this tool, they left about $1,200 on the table. And that just means that consumers were buying plans that left them spending more money than they had to or that they needed to. But when those brokers were paired with AI, they helped consumers make better choices. So that meant they got most of the money they were leaving on the table back.
DG: That’s awesome, right. Essentially, the researchers found that pairing a broker and AI helped consumers avoid an almost instinctual impulse to overbuy. What’s your big takeaway here? What were consumers sort of better positioned to do?
SN: What they found is that people spent less time focused on the really easy things to compare: the premium, the deductible. They spent more time thinking about things like expected out-of-pocket spending from coinsurance, co-pays, things like that. The other really cool thing is they found that people were less attached to brands.
DG: When you say people were less attached to brands, the AI would flag a plan with a name the consumer had never heard of, but once the broker endorsed this as potentially a good fit, the consumer was like, ‘oh, cool I don’t care if I ever heard of this plan, if it’s going to save me money, if it’s a better deal, I’m happy to do it.’
SN: Exactly. I love this paper because it sets up this man versus machine kind of thing. But instead of man versus machine, it’s like man plus machine will conquer the earth. Even though this paper is looking at Medicare products, it really could be kind of moved over into other types of situations where consumers are buying health insurance to help people make better choices.
DG: And anything that’s going to make it easier to shop for health insurance people are going to be comfortable with because it’s such a pain.
SN: Yes, it is a huge pain.
DG: OK, on to the next paper, Sayeh.
SN: There’s one more thing about this paper, Dan. I had the pleasure of hearing Jonathan Gruber actually give this presentation in the spring in the Electronic Health Economics Colloquium, which is this kind of post-COVID web seminar series. But the coolest thing is that he gave the entire presentation with his bird chicken on his shoulder. And I think that this may be the only time in human history that an economist has given a seminar with an avian copresenter.
DG: You really are a dork. But thank you.
SN: It was incredible. It was spectacular.
DG: On that foul note, let’s go to our second paper, Sayeh.
SN: You can see all my tweets about it.
DG: OK, Sayeh, so enough about birds.
SN: Let’s do this on a wing and a prayer.
DG: Oh my let’s stay grounded and let’s focus on paper two, which has got the name “The role of behavioral frictions in health insurance marketplace enrollment and risk: evidence from a field experiment.” What’s this paper all about?
SN: Ok, let me just hit you with some information that really floors me every time I think about it. So, 60% of people who are uninsured and eligible for either Medicaid or the ACA premium tax subsidies, which is what this paper is about, don’t pick up coverage. They remain uninsured. That’s 16 million Americans who are the targets of these policies, but they’re not benefiting from the financial protection, from insurance. These authors are exploring a possibility that, you know, people don’t sign up for coverage because they kind of don’t know how to sign up for coverage. They face barriers, which they call frictions, in figuring out “how do I sign up?”
DG: So let me jump in here for a second, Sayeh. Why do the authors assume information figuring out how to actually sign up and enroll, which arguably is a cumbersome, annoying process is actually the real chief barrier here?
SN: Well, so they don’t know that it’s actually the chief barrier. It turns out that the prevailing theory about why people don’t sign up for coverage may be you just don’t value insurance that much because if you get sick, you can always go to the ED and get some free care, either through charity care or through the hospital writing off your bad debts. But there may be some other reasons why that that theory can’t explain all of the lack of takeup. I think what these authors are identifying is that, you know, this population may have some kind of health insurance literacy barriers.
DG: Ultimately what was the experiment here?
SN: They did another randomized controlled trial, same as the other paper. They focused on a group of uninsured, subsidy-eligible people who showed up to California’s health insurance exchange, Covered California. They sent out five different types of letters. And these letters had escalating amounts of information in them. So there is a control letter and then there’s a letter that has basic information, just sort of saying that like “guess what? open enrollment season is coming and, you know, here’s how you can enroll,” escalating all the way up to a letter that had information about exactly how much subsidy you would get if you if you applied for a plan in your area and also giving you quality information about the different plans that are available to you.
DG: So all of the options which letter was the most effective?
SN: What they find is that, on average, across all the types of letters take-up or enrollment increased by about 16%, they actually didn’t notice any statistically significant differences between the types of letters. The basic reminder seemed to be doing a lot of the work. However, if you zero in on lower income households, they found that the letters that had information about the amount of subsidy actually were more effective in increasing enrollment than the basic letter.
DG: In other words, when people were told, “hey, you’re getting a great deal here,” that prodded people into action?
SN: Yeah. Especially people who have less money to spread around to different things, who have a tighter budget constraint.
DG: And so what’s the take home from this study?
SN: The thing that makes me really excited about this is that letter is such a low cost intervention. If it’s truly the case that you can boost enrollment by sending people a letter, that’s fantastic! And one thing that the study found that I thought was really cool from a kind of wonk perspective is that they found that the people who weren’t enrolling but were kind of nudged to enroll by getting these letters, those people were healthier, on average, than the people that had already enrolled. And that’s actually really good for risk pools. So if you can sort of nudge more of these people into the ACA marketplaces, that’s going to make risk pools more stable over time because you’re adding healthier people to that pool.
DG: Can you put the findings from this paper into a broader context, Sayeh? What do we know now as a result of this paper that we didn’t know the day before?
SN: Yeah, in general our kind of health insurance literacy is so low. Even when you have a subsidy to purchase this really great coverage, you just might not know how to do it.
DG: So to tie these two papers together, there’s more evidence now to suggest that if we’re interested in helping consumers make better choices and making sure that people sign up for public health insurance programs, more information is needed.
SN: Yeah, definitely. We can’t expect people to bear the psychic burden of figuring out all of this information. Giving them a little bit of help to figure out how to actually access these benefits is beneficial.
DG: Sayeh, it’s always great to talk with you. I feel you are the wind beneath my wings.
SN: Ah man, I’m trying to think of a bird pun. This has been a real feather in my cap.
SN: I know it’s not very good.
DG: Thank you, Sayeh Nikpay.
SN: Thanks, Dan. It’s been fun.
DG: Great talking to you as always.
Jonathan Gruber, Benjamin R. Handel, Samuel H. Kina and Jonathan T. Kolstad wrote the NBER working paper on Artificial Intelligence.
Richard Domurat, Isaac Menashe and Wesley Yin wrote about behavioral frictions. That research is scheduled to appear in American Economic Review early next year.
Links to both papers are on our website tradeoffs.org. Sign up for our newsletter for more top research picks.
I’m Dan Gorenstein, and this is Tradeoffs.
Want more Tradeoffs? Sign up for our weekly newsletter!
Articles discussed in the episode:
Managing Intelligence: Skilled Experts and AI in Markets for Complex Products (Jonathan Gruber, Benjamin R. Handel, Samuel H. Kina, Jonathan T. Kolstad; NBER; 05/2020)
The Role of Behavioral Frictions in Health Insurance Marketplace Enrollment and Risk: Evidence from a Field Experiment (Richard Domurat, Isaac Menashe, Wesley Yin; American Economic Review; scheduled to appear early 2021)