Health Insurance: Choices Guaranteed to Be Bad

In this excerpt from Nudge, Richard Thaler and Cass Sunstein examine insurance choices that cost more without providing a corresponding benefit. The studies and account arrangements are described in the context of the book.

Remember Carolyn and the cafeteria? When discussing that example, we mentioned that some people think choice architects should try to create what might be considered a neutral choice environment so as not to nudge people in any particular direction. We said this is generally impossible. Whenever people are choosing, there has to be some underlying structure. However, thanks to research by behavioral economists Saurabh Bhargava, George Loewenstein, and Justin Sydnor, we do have good evidence about the influence of that structure, and the difficulty of achieving neutrality. We also show that a seemingly sensible design can produce terrible choices and therefore costly outcomes.

The team studied what happened when a large U.S. company decided to change the design of the health care options it offered to its employees. It offered what we call the “salad bar” approach, meaning that participants would select exactly the ingredients they wanted in their health insurance. An important feature of this case study is that the only factors that differed among policies related to costs—all participants would have exactly the same set of options regarding where and from whom they could receive care, whether they needed permission for some procedure, and so forth. The salad-bar aspect was only about money.

Employees could choose their annual deductible from four options ($1,000, $750, $500, $350), their maximum out-of-pocket spending above the deductible from three options ($3,000, $2,500, $1,500), two coinsurance rates on coverage above the deductible and before the out-of-pocket threshold is reached (80 percent, 90 percent), and two office copayments ($15 for primary care/$40 for specialist, $25 for primary care/$35 for specialists). If you don’t understand what all those terms mean, don’t fret; it will not spoil the story. Since any of the possible combinations was permitted, there were 48 plans available (4 x 3 x 2 x 2), each with its own price.

This new health care plan was introduced with considerable internal fanfare, and employees were encouraged to build their own plan, much like the effort in Sweden to encourage people to build their own portfolio. However, one of the options, the one with the lowest monthly premium and highest levels of cost-sharing by the employee, was selected as a default plan for anyone who did not want to go through the choosing exercise. Okay, here is a quiz: What percentage of employees do you think made an active choice by designing their own plan?

There are plausible arguments for why one might think that many people would opt out of choosing. Just reading the paragraph listing the forty-eight different options is enough to give Sunstein a major headache. On the other hand, choosing the default seems to go against the grain. It is a bit like going to a costume party wearing your normal weekend outfit. Still, make your guess before going on.

Before we reveal the answer, we have to say that it is not quite possible to know exactly how many people just opted out because, as in the Sweden example, the default plan was also one of the options one could actively choose. Altogether, just 14 percent of the employees enrolled in the default plan, and the authors of the study believe that most of those actively chose that plan. They estimate that only 2 percent of the employees passively took the default. Regardless of whether this inference is precisely right, it does provide another example in which most people decline the default. And similar to the case in Sweden, the default plan was a pretty good option and, as we shall see, demonstrably better than many of the policies people chose instead.

Although choosing a plan may seem like a daunting task, the firm took some care in making the process user-friendly. As the authors put it: “Employees were asked to construct candidate plans by sequentially selecting levels within each of four cost-sharing attributes (for example, ‘Which annual deductible meets your needs?’) and were urged to consider the trade-offs between price and coverage (for example, ‘Remember, a lower deductible means higher annual paycheck costs’). After building a first plan, employees were shown the monthly plan premium and could either enroll in that plan or build a subsequent plan through the same sequential process.”

How, then, did the employees do as plan designers? Not so well. One might think this would be difficult to judge; after all, who is to say whether Sunstein’s salad is better than Thaler’s? Can’t we apply the dictum “De gustibus non est disputandum” (one can’t argue about tastes)? No! In this case, that dictum does not apply because many of the choices people made violated a fundamental principle of rationality, a word we rarely use in this book (not least because it creates a lot of trouble). The principle is dominance: if option A is better than option B on at least one dimension, and no worse on any of the others, then A must be preferred to B. If we give Sunstein a choice between two identical cans of Diet Coke, his drink of choice, and one is cheaper than the other, then the cheaper option dominates the more expensive one. No one who understands this principle would knowingly violate it. But remarkably, a majority of employees selected a plan that was dominated by another one.

Let this result sink in. The firm gave its employees total freedom of choice, and a well-functioning interface in which to compare options, but a majority of employees selected a plan that was unequivocally worse than at least one of the options they had rejected. Furthermore, these poor choices were costly. The employees who selected dominated plans ended up paying an average of 28 percent more for their health care than if they had switched to a better version of the plan they elected. The authors titled their paper “Choose to Lose.”

The poor choices all had something in common: low deductibles! In many cases, low-deductible plans were dominated by an identical plan with a higher deductible. For example, one plan had an annual deductible of $1,000 and an annual premium of $930. It dominated a plan with all the identical features except for a deductible of $500 and a premium of $1,568. Do the math. By choosing the latter an employee is paying $638 to reduce the deductible by $500, while keeping everything else the same. That has to cost at least $138 more, and if there are no claims, the entire $638 would be saved. In this comparison, choosing the higher deductible would have avoided the trap.

This deductible aversion extends well beyond health insurance. The same logic applies to home and automobile policies. In a study of fifty thousand homeowners insurance policyholders, economist Justin Sydnor documented the case for choosing a higher deductible. At that time the insurer offered deductibles of $100, $250, $500, and $1,000, with the $100 option being extremely expensive and almost never selected. Should consumers increase their deductible? On average, raising a deductible from $500 to $1,000 saved about $100 per year in premiums. If a policyholder then has a claim, they must pay up to an additional $500 out of pocket, so they would need a claim about once every five years just to break even. In reality, only about 5 percent of policyholders make a claim each year. That means increasing the deductible is a really good deal. If those same policyholders took our advice to raise the deductible and deposited the savings into their “On My Own” accounts, they would add roughly $100 per year and dip into the account only once every twenty years or so, on average accumulating over $1,500 (plus interest) after two decades. Similar savings opportunities exist on automobile insurance as well.

Returning to health coverage, the fact that many low-deductible plans at the original firm were dominated by higher-deductible alternatives is not unique to that one large company. A follow-up study by Chenyuan Liu and Sydnor used a survey of health insurance plans in a large sample of U.S. employers. They searched for companies that offered comparable insurance policies with both high and low deductibles. They found 331 firms that met these criteria. If workers in those plans had followed the “take the higher deductible” rule of thumb, they would have been better off at 62 percent of firms, even in years when they had lots of medical bills. At roughly half of the firms, the high-deductible plan virtually guaranteed lower costs. The expected savings from selecting the high-deductible plan are typically more than $500 per year, often with no increase in financial risk. That is a potential $500-a-year contribution to the families’ On My Own accounts.

Deductible aversion is not just an American affliction; the Dutch suffer from the same problem. In the Netherlands, everyone is required to buy health insurance and (as of 2020) all policies were required to have a deductible of at least €385 per year, but families can elect a deductible of as much as €500 more. Roughly speaking, increasing the deductible by €100 reduces the premium by about €50. Apparently, our deductible rule of thumb is not (yet) well known in the Netherlands, because only about 10 percent of the public choose a deductible above the minimum. Yet a team of economists from Berkeley and the London School of Economics find that most would gain from doing so. This applies even to those for whom the choice should be most obvious, those who are very unlikely to spend more than the minimum deductible, and so bear little risk that they would have to pay more. Even among this group only 15 percent choose a higher deductible, although if they were Econs, everyone would do it.

The book describes a U.S. option for households with qualifying high-deductible plans: a health savings account (HSA), with tax-sheltered contributions. Unlike the notional On My Own Account, the HSA holds actual savings. Many employers try to encourage employees to choose the high-deductible (HD) plan by making contributions to the HSA for any employee who does so. Any money contributed to the HSA that is not spent at the end of the year simply rolls over to the next year and eventually just turns into a retirement savings account. The existence of these accounts deepens the mystery as to why high-deductible plans get so little take-up, even in cases when they are a certain money saver.

Part of the explanation for the low take-up rate of HD plans may be that consumers do not realize that an HSA account is about as sludge-free as one could wish. The account is opened automatically by the employer, often with a starting balance already deposited on January 1. Policyholders receive a debit card that is linked to the account, and eligible medical expenses can be paid with the card. No paperwork or prior approvals are necessary. Workers might be confusing this with an older (but still existing) benefit called a flexible spending account (FSA) that we mentioned earlier in the book. These accounts are hassle nightmares with sludge lurking at every turn. Bills have to be submitted and are often rejected for no apparent reason. Worse, users have to guess how much medical spending they might need in the coming year and can lose money if they do not deplete the account by March 31 of the following year. This provision provides a boost to stores selling eyeglasses, which do a booming business to customers whose accounts are about to expire.

Deductibles now can be as high as $4,000 or more for a family plan, so one might think simple risk aversion could explain why more people don’t elect them, but in the case of dominating plans there is no risk to bear. Households pay less no matter how much health care they consume! Consider this (simplified) example: The employer offers two deductibles, $1,000 and $4,000, and contributes $1,000 to the HSA of any employee who elects the high deductible. Premiums are $300 a month lower for the HD plan, meaning $3,600 a year. Suppose a family puts all of the money saved on premiums into the HSA and considers it an On My Own Account. By the end of the year, $4,600 will have been deposited in the health savings account—enough to cover the deductible in this example. Whether it covers other out-of-pocket spending depends on the plan and the care received.

We suggest that the family adopt the following mental accounting plan: Whenever a medical expense is incurred, until they reach their deductible, they pay for it using the debit card that comes with their HSA. After that, insurance pays according to the plan’s remaining cost-sharing terms. The timing of expenses also matters: a large medical expense early in the year may exceed the money available in the account, especially before the family has accumulated savings from previous years. For families that are living on a tight budget, this can be a crisis. The authors propose a way employers could address that problem.

Remember that employers are contributing to the HSA because they want to encourage families to join the HD plan. If that is indeed their goal, we suggest they offer one additional benefit: interest-free loans to anyone who incurs a large medical expense early in the year. If properly communicated, this might increase take-up by low-income employees who may be most likely to worry about a liquidity crisis. Notice that because the HD plan is cheaper, these are families that could benefit most from the extra cash.

One important proviso. Although we are sticking to our rule of thumb for consumers in choosing deductibles, we remain agnostic about whether high deductibles actually make the health care system more efficient. It does seem that consumers spend less when they face a deductible, but whether that is a good thing is less clear. The usual argument for high deductibles is to mitigate what economists call “moral hazard,” which is the idea that if patients face no costs of medical care, they will spend too much. They want consumers to have “skin in the game.” Think of your experiences at open bars or all-you-can-eat restaurants to get the idea. And certainly, patients will think twice about going to the doctor or filling a prescription if they have to pay. The question is whether they cut back on the right things. Unfortunately, the evidence suggests that they do not.

Economists Katherine Baicker, Sendhil Mullainathan, and Joshua Schwartzstein argue that health care designers need to worry about what they call “behavioral hazard” at least as much as moral hazard. Recall that for many medical conditions, a major problem to overcome is what doctors call “adherence” or “compliance.” For diseases such as diabetes, hypertension, and high cholesterol, taking prescribed medication is essential to good health, and failing to do so can create costly medical emergencies. In a well-structured system, the price of these medications would be negative, meaning we would reward patients for adhering to their prescribed medications. The evidence suggests that when patients have to pay for some or all of the cost of their medications, they are as likely to cut back on high-value drugs such as insulin or beta-blockers as they are to cut back on nearly useless drugs such as those for the common cold.

One experiment conducted by a team of medical researchers led by Niteesh Choudhry is particularly telling. They randomly assigned patients discharged after having suffered a heart attack to either a control group with usual coverage under which prescription refills would usually cost between $12 and $20, or a treatment group for whom statins, beta-blockers, and ACE inhibitors (drugs of known efficacy for such patients) were provided for free. They then tracked adherence rates and clinical outcomes over the next year. Faced with lower prices, consumers used more drugs, just as economic theory would predict. But how does this behavior affect health outcomes? One recent experiment tracked new enrollees in Medicare’s prescription benefit program. Due to a quirk in the way eligibility works, those born earlier in the year (e.g., February) are more likely to face cost-sharing by the end of the year than those born later (e.g., September)—meaning they paid more for each drug out of pocket. While the change was small, about $10 per drug, patients cut back on their drugs, as in the previous study. But as medical science would predict, patients who cut back had a much higher risk of dying—about 33 percent higher! What’s worse, the people at the highest risk of problems like heart attack and stroke—exactly the ones that economic theory would predict should cut back the least—were the most likely to drop life-saving drugs like statins, beta-blockers, and ACE inhibitors when the price went up. Here, charging people for proven medications costs both money (in terms of higher health costs) and lives. So, requiring patients to share the cost of medical expenses may drive down spending in the short run, but it remains an open question whether that lowers costs in the long run, and whether it is a good idea, even if so.

In light of this, we think that our On My Own-HSA mental accounting may help on two fronts. First, it may help households choose financially superior plans. There is no point in paying $800 to reduce the deductible by $500. When we encourage using the On My Own to “pay” for the health care costs that come in before reaching the deductible, we are hoping to thread the following needle: we hope that families will not skimp on necessary medical services because of an out-of-pocket cost, but at the same time we want to discourage wasteful spending by reminding people that they get to keep and invest the money in their HSA for future medical expenses or retirement spending. If this worked, it would be helpful nudging via mental accounting.

While we have focused specifically on the American health-care system here, similar dynamics exist in many other countries. The useful comparison is between total costs under different claims scenarios, with attention to coverage and the cash available when bills fall due.