Do narrow networks save money?

According to a recent paper by Wallace (2023), the answer is ‘yes’, but it does so in a highly inefficient manner. Using 2008-2012 Medicaid data from the New York State Department of Health, the author find that:

Leveraging the random assignment of over 50,000 Medicaid enrollees in New York, I present causal evidence that narrower networks are a blunt instrument for reducing health care spending. While narrower networks constrain spending, they do so by generating hassle costs that reduce quantity, with modest effects on prices paid to providers. Enrollees assigned to narrower networks use fewer of both needed and unneeded services and are less satisfied with their plans. Using my causal estimates to construct counterfactuals, I identify an alternative assignment policy that reduces spending without harming satisfaction by matching consumers with narrower networks that include their providers.

If using narrow networks is not efficient, why do plans use this tool? According to Wallace, one of the key reasons is adverse selection.

…plan choices of sicker enrollees are more responsive to network breadth, suggesting that—absent regulation—plans may construct networks that are narrower than is socially optimal in an effort to select healthier patients

Wallace recommends using a “smart defaults” approach whereby Medicaid enrollees are auto-assigned to low-cost plans that they rae likely to prefer.

Intuitively, this is achieved by matching enrollees with narrower networks (to reduce spending) that nevertheless include their usual source of care (to increase satisfaction). These simulations have clear policy implications for New York but offer a broader lesson to the more than 30 states that operate mandatory Medicaid managed care programs: auto assignment can be a powerful tool for achieving program goals (e.g., reducing cost and increasing satisfaction) without unnecessarily restricting enrollee choice of plans.

The degree to which these “smart defaults” would actual save cost likely depends on a variety of factors including, default stickiness, the degree to which State Medicaid Agencies can predict usual source of care, and the persistence of usual source of care across time. Nevertheless, having smart defaults with more choice would be preferable for patients as compared to very narrow networks with limited choice.