The Choice Architecture of Automatic Enrollment in Health Insurance

As discussion advances on healthcare reform and expansion of coverage in the U.S., the health insurance enrollment process takes on heightened importance. These notes look at the issues through the lens of behavioral economics – using both psychology and economics to understand individual decision-making. One leading example of ways in which conditions where choices are made – the choice architectures – have had substantial impact is automatic enrollment of employees in 401(k) plans on plan participation. Thaler and Sunstein, in their book Nudge, describe six attributes of good design of choice architecture: pay attention to incentives, make important information salient, choose smart defaults, give feedback, expect error, and structure complex choices.

Looking at enrollment in health insurance through the lens of behavioral economics and considering the choice architecture provides several insights. My own thinking has also been advanced considerably through numerous discussions with Stan Dorn at the Urban Institute, and especially his Automatic Enrollment Strategies work. These notes make five main points:

  • Establishing automatic enrollment is valuable for any type of health insurance coverage expansion, regardless of whether there is an individual mandate
  • Allowing states to augment a national base of health care reforms would enable refinement of various approaches
  • Determining eligibility based on data collected through the tax system, either on tax forms or through data matching, would greatly facilitate automatic enrollment
  • Implementing collections of individual contributions to premiums through the tax withholding system may facilitate continued enrollment and reduce administrative costs
  • Creating a system in which third-parties provide enrollment advice to individuals and are rewarded for their performance may be preferable to legislating how to select a default plan

Automatic enrollment and coverage expansion. Any health insurance coverage expansion will need to get people newly enrolled. Automatic enrollment is a term describing various ways in which enrollment can occur as automatically as possible, with little need for individuals to exert effort in the process unless they want to. Even if there were an individual mandate legally requiring health insurance coverage, it will be valuable to make enrollment as easy as possible to assist compliance with that mandate. If there is no mandate, the role of incentives and other aspects of choice architecture will be even more critical in attaining enrollment goals. In particular, the findings from behavioral economics about how losses loom larger than gains suggest that taxes on non-enrollment may affect behavior more than economically equivalent subsidies for enrollment.

State augmentation. Health care reform will be enormously complicated. It will likely make sense to have an initial set of national reforms that are relatively simple. There is good reason to believe that the designs specified in the choice architecture of the reforms will have substantial effects on behavior, but the magnitude of response is uncertain and new information on differential impacts of variations could have high returns. States across the U.S. also vary tremendously in their capacity and political willingness to engage in reform. For states that wish to do more, there are also great benefits both directly for the citizens of those states and indirectly to others from spillovers in learning if these states can augment basic national policy. For example, national policy could provide information from the National Directory of New Hires to state health exchanges, to enable the exchanges to contact new employees about health insurance options. Alternatively, firms nationwide could be offered a payment in return for connecting new hires directly with an exchange. A single state, however, may wish to try something more ambitious, such as requiring that all new employees file W4 forms indicating their insurance status, and automatically enrolling those without insurance such that the individual would have to take action in order to not be enrolled. Everyone will learn about what works if states are allowed to experiment with augmentation.

Tax system eligibility determination. A major barrier to health insurance coverage expansions is likely to be the complexity of an arduous application process – particularly when government subsidies for insurance are involved. To the extent that eligibility determination can be made on the basis of data that is already collected for other reasons, this complexity could be substantially reduced. The most straightforward system for means-testing would be to base eligibility on the previous year’s tax return. This would increase simplicity but would reduce targeting efficiency, as some individuals would have high assets or substantially increased income by the time of enrollment such that the available dollars may not be as efficiently targeted to those most in need. Better systems integration would be needed to merge data on immigration status to tax data in order to automatically use the information in eligibility determination. There could be an additional process to apply for a higher subsidy level based on income changes or other factors.

In the short-run, it will likely be useful to collect data on insurance status from individuals on their tax forms, so that eligibility for subsidies can be determined for the uninsured. Over a longer time horizon, systems can be developed to support data-matching, so that insurance status can be determined based on electronic reports from insurers about who they are covering (while perhaps at the same time, return-free filing can be implemented for many taxpayers). Once subsidy amounts are determined, the Internal Revenue Service (IRS) can convey this information to health insurance exchanges and to individuals.

Collections. If the funds for an individual’s share of health insurance premiums can be automatically collected, becoming enrolled in health insurance is much easier. If collections are automatic, the biggest single barrier to getting people to start coverage and to maintain coverage may be overcome. One approach that states could use would be to collect deposits in a new Insurance Purchase Fund (IPF) account administered by a state health exchange. Most people would use the IPF as an account into which their employer could automatically deposit a portion of their paycheck in order to pay for their share of their health insurance premium.

While a decision about whether to purchase insurance could be voluntary, states could require that everyone without health insurance open an IPF account. This would help create a social norm where everyone is putting aside funding to pay for health insurance. Each uninsured individual filing a tax return could then be required to have a minimum balance in the account at the end of the year, where the minimum would be related to the number of months uninsured during the year. These funds would: belong to the individual, be taxable income, accrue any investment returns tax-free, be withdrawn to pay health insurance premiums or other health care costs, and be withdrawn with a notarized letter for other purposes if above the minimum balance. If an individual’s IPF were below the minimum balance, then a penalty would be assessed, justified on the grounds that the uninsured impose costs on other citizens when they have insufficient savings (or insurance) and then receive uncompensated care – with the penalty helping the government cover the costs of assuming this risk.

Ways to direct funds into the IPF would include automatic bank account debit, credit card payment, monthly billing, or paycheck withholding. While withholding would be the most automatic, this variety of methods would provide options for the self-employed and for very small firms without electronic payroll that may be exempt from any withholding requirements. Firms making employer contributions to health insurance could make these to the IPF, reducing necessary employee contributions. Firms without employer sponsored insurance could be offered a small tax credit for setting up paycheck withholding. W4 withholding forms at these firms could collect any additional information necessary for firms to determine the approximately appropriate IPF withholding amounts. The IRS could also mail pre-filled W4 forms to individuals after exact subsidy determinations are made, so that individuals could adjust withholding to be more precise. The IRS could also send instructions for adjusting the W4 if there is more than one employer, if new household members are to be covered by insurance, etc.

Regardless of whether IPF contributions are mandatory or voluntary, the IPF would be structured to gather approximately the right amount needed for premiums during the year and to get these funds to the insurer. Exact amounts would be reconciled at the end of the year as part of the tax return filing process, with excess deposits returned and any balances due being collected by the IRS. Using this existing collections infrastructure should reduce the administrative costs of collection relative to a system where individuals receive subsidies from the government and make payments directly to insurers. Making collections as automatic as possible will facilitate enrollment. An IPF account system could be designed to serve only people who do not work at firms that offer employer-sponsored health insurance, or to be broader. Employers who do sponsor insurance could potentially be compensated for continuing to provide both collections and enrollment services, and thereby lessening the costs to the government of having a health exchange do so (or this could be explicitly considered part of an employer contribution to health insurance coverage).

Enrollment. Once information about uninsured individuals and their subsidy level has been collected – and conveyed to a health exchange, say – then automatic enrollment would involve proactive outreach to those individuals. In some states, enrollment could automatically start coverage with the lowest cost plan, a public plan, or some other default plan. There could be an initial period of no premiums for three months in order to allow transitional issues to be sorted out before premiums were collected.

An alternative version would require one active step before coverage began. For example, a health insurance exchange would send an insurance card to an individual with a sticker on it, as with most credit cards, that required calling 1-800-INSURE-US to activate the card. For states using an IPF, they could send the card when the IPF opened, and each subsequent year that individual remained uninsured. When an individual called in, she could indicate whether she would like additional individuals covered by her insurance. The individual could indicate which plan she wanted to have cover them. An active choice of plan could be required in order to receive coverage, or default plan could be assigned if no choice is made.

One particularly attractive way to implement 1-800-INSURE-US card activation would be to route calls to third party advice providers on a rotational basis. The advisors would be responsible for assisting with plan choice, and providers could structure the choice process in different ways, in terms of what information they used in deciding which plans they recommended and how to set defaults (if any) for specific individuals. There would be a firewall between advice providers and insurers, with regulations to ensure that advisors provided advice free of financial conflicts of interest. If an advice provider did not have a representative available, a new caller would be routed to another provider that did – providing an incentive to have adequate staffing. More generally, a set of performance metrics would be gathered on each advice provider, with information such as: the duration that individuals continue with the selected plan, satisfaction with the plan, satisfaction with the advisor, health status of the individual, etc. It would be important to have multiple factors in the assessment, so that firms do not focus their efforts on a narrow set of performance targets that are not necessarily in the broader interest of their clients. The pool of individuals being advised by each provider would be similar on average due to the rotational assignment of calls, so differences in performance between providers could be directly attributed to the providers themselves in a simple and transparent way. Advisors with the best performance would receive a greater share of new callers in the future, and poor performers would have their share reduced or eliminated.

To conclude, ensuring enrollment is easy is an important part of any plan to extend health insurance coverage to the 45 million Americans who currently are uninsured. Enrolling forty-five million new enrollees would be a truly major administrative task. Good choice architecture for automatic enrollment can help attain coverage goals under most any type of health care reform seriously being discussed. We will learn a great deal if states can experiment. Eligibility determination, collections, and enrollment can be structured to both facilitate enrollment and reduce costs.