Theodore R. Marmor, Emeritus Professor of Public Policy and Management, Yale School of Management,
Joseph White, Luxenberg Family Professor of Public Policy, Case Western Reserve University
This interesting article by Anna Sinaiko and Richard Zeckhauser provides both useful data and some persuasive analysis, but its lessons may not be quite what Dr. Sinaiko and Professor Zeckhauser conclude.
On the one hand, the co-authors show how some common tropes in behavioural economics can be taken beyond their appropriate bounds. On the other, the article illustrates the persistent effort to see the donkey of insurer choice as a beautiful unicorn: a splendid policy, if only the right choice architecture (here called “nudges”) could be created. In this case the behavior is actually better than the authors claim, while the entire discussion fails to confront a basic paradox.
This study involves Medicare beneficiaries whose previous Medicare Advantage (MA) choices have been terminated (TCs). They could choose a new MA plan, a separate Part D plan, or default to neither. This latter default is very different from other health care choices for which the authors hope to draw lessons. Defaulting to only traditional Medicare involves far less severe consequences than for example, not signing up for ACA coverage. In the latter case people will have no insurance at all, and possibly pay a fine.
The pattern of visible and actual costs and benefits in this case is also quite different from the classic situations for which nudges have been recommended, such as default contributions to pension plans. Default contribution is meant to reduce visibility of immediate loss of consumption, and compensate for the fact that if a person does not sign up for a 401(k), the future loss of retirement income is less visible. In the situation studied in this article, however, beneficiaries chose an MA plan; its benefits were then taken away; so the issue is how they would respond to a visible loss of something they already made an effort to gain. We doubt they should need an extra nudge in this situation.
In fact, they almost all do respond, but the article does not see that because it defines “not choosing” in a highly puzzling way: as not choosing an MA plan, instead of not choosing pharmaceutical coverage.
Beneficiaries may choose among MA or Part D plans with varying premiums, network restrictions, and benefit levels. Table 2 shows that 15.4% of TCs from 2009 were in traditional Medicare in 2010. But that does not mean they lost benefits. Table 3 shows that only 20.6% of those did not choose a Part D plan. Therefore, barely three percent of TCs in 2009 emerged with no drug benefit. Why, then, should we worry about defaults in this case?
The article presumes that people who wanted an MA plan before must want one now. Yet the experience of enrolling in a plan with enough problems to be cancelled may have changed enrollees’ preferences. Table 2 provides some evidence that TCs did have negative experiences with HMOs. In 2006 65% of MA enrollees were in HMOs (Gold et al. 2010). The table shows that only 52.4% of enrollees who had chosen an HMO before chose a new HMO for 2010. In short, leaving MA could be a choice based on experience. If so, imposing a version of the previous choice as a default would be paternalistic even by the usual standards of nudging.
The authors deserve praise for including network concerns in their effort to craft a better default. Considering the importance of networks to enrollees (especially the elderly), a “most comparable” plan should have a comparable network. We have to wonder, however, if the authorities will ever have sufficiently reliable network data for accurate comparison. In our experience, the plans themselves often lack up-to-date information. The needed database may not be another policy unicorn, but we would not count on that.
In any event, with only 3 percent actually “doing nothing,” we think the study shows that default is a minor concern compared to other MA problems. The theoretical implication for behavioural economics is that there are situations in which the arguments for directive nudges do not apply.
For the choice among Part D and MA plans, there is better reason to worry about risk-selection, risk-exaggeration, and network adequacy. We are struck, however, by the article’s unexamined assumptions about choice of insurers.
Arguments for choice presume that consumers should choose plans that offer them the most value. If everyone does that, each plan will get the most expensive possible enrollees. That pattern is normally called adverse selection, and a threat to stable insurance markets.
How, then, can insurance markets survive informed choice?
In practice, insurers make choice difficult, so less informed, by proliferating options – as in Part D. Even high income and presumably sophisticated choosers then often make objectively wrong (“dominated”) choices (Bhargava et al. 2015). Limiting adverse selection, in short, may depend on poor choices and deception.
Perhaps behavioural economists could study the conditions that yield more or less stable equilibria of deceptions on one side and mistakes on the other.
The full article by Anna Sinaiko and Richard Zeckhauser in the third issue of the journal can be read online here
Bhargava S, Loewenstein G and Sydnor J. “Do Individuals Make Sensible Health Insurance Decisions? Evidence From a Menu With Dominated Options.” National Bureau of Economic Research Working Paper 21160, May 2015.
Gold M, Phelps D, Jacobson G and Neuman G. “Medicare Advantage 2010 Data Spotlight: Plan Enrollment Patterns and Trends.” Kaiser Family Foundation, June 2010. https://kaiserfamilyfoundation.files.wordpress.com/2013/01/8080.pdf