Public health insurance programs in the U.S. are increasingly delivered through regulated markets of private insurers (Gruber, 2017). As part of the dynamic procurement process for these markets, contracts with incumbent insurers are not always renewed, forcing enrollees to switch to another health plan. Such transitions can disrupt patients’ utilization patterns, cause discontinuities of care, and result in adverse health outcomes, thereby giving rise to non-pecuniary switching costs between contracted insurers. For procurement officers, switching costs create a tradeoff between the potential benefits of reducing costs and/or improving quality by replacing an insurer, and the disruptions such changes may cause.

I examine switching costs in the procurement of health insurance within the setting of Medicaid managed care (MMC) — the regulated markets of private managed care plans that provide publicly financed health insurance to approximately 70% of Medicaid beneficiaries (see Layton et al. (2018) for a review). My focus is on incumbent plans that do not win a new contract in a state bid to serve a county or service area, forcing all their enrollees in these regions to switch out. To identify plan exits, I collect publicly available information about MMC bids, including details about bidders, winners and losers, and the bid milestone dates. I then use administrative data from the Medicaid Analytic eXtract (MAX) for the years 2006 to 2014 to examine enrollment and utilization patterns surrounding bid-induced plan exits in five states where such exits are identified in the data: Arizona, Minnesota, Missouri, Texas, and Washington. I focus on non-elderly beneficiaries who initially enrolled in Medicaid at least 18 months before the plan exit, ensuring a sufficiently long pre-period.

Within a stacked difference-in-differences framework, I compare about 320,000 (treated) enrollees, in 62 plans that exit 123 counties after state bids, to a control group of 2.4 million (never-treated) enrollees in plans that remain in the market. The stacked framework addresses potential bias due to heterogeneous treatment effects (Sun and Abraham, 2021, Goodman-Bacon, 2021). To mitigate bias due to anticipatory effects, the pre-period of the main analysis excludes any period after contracts are awarded and before the exit. I conduct event studies, controlling for individual and state-specific time fixed effects, to verify that no differential trends are apparent between the enrollees of exiting and remaining plans before contracts are awarded. I use placebo tests for services plausibly exogenous of plan influence (e.g. acute appendicitis hospital admissions), as evidence supporting no differential level of data reporting by these plans.1

I find that dropping a plan causes significant disruptions to the care of affected enrollees, and some suffer adverse health outcomes. Throughout the two years after a plan exit, enrollees from exiting plans use fewer prescription drugs, including those that treat chronic conditions such as diabetes and depression. Compared with the control group and relative to the baseline period, the number of days’ supply in filled prescriptions for these enrollees is lower by about 16% (1.9 days) during the first post-exit year, and only partially increases in the following year. Enrollees from exiting plans have close to 10% fewer visits to primary care physicians over the two post-exit years, and by the end of the first year, they are admitted to hospitals 8% more often. Using prices from Medicaid’s fee-for-service (FFS) program, I estimate that insurers’ spending on each new enrollee from exiting plans is lower by 7% ($234) in the first post-exit year than spending on enrollees who remain in their plans after the state bid.

Sicker enrollees, particularly children and non-white beneficiaries, are more sensitive to post-exit disruptions, experiencing higher rates of hospital admissions attributed to ambulatory-care-sensitive conditions, deemed preventable with appropriate community care. Enrollees from exiting plans that switch to newly entering plans after the state bid also suffer more severe disruptions.

Changes in provider networks and drug formularies may serve as mechanisms for the estimated disruptions. While during the pre-exit year, nearly two thirds of outpatient visits of enrollees from exiting plans were made to known providers (familiar from the previous year), this share drops by almost 40% after the exit. A substantial share (21%) of enrollees from exiting plans lose access to their pre-exit primary care physicians (PCPs) within the network of their new plan, compared to only 3% of beneficiaries in plans that remain in the market. Losing access to a PCP is correlated with more severe post-exit disruptions.

Exploring mechanisms that may contribute to the lower use of prescription drugs, I find that enrollees from exiting plans are more likely to fill prescriptions at unfamiliar pharmacies after the exit, suggesting that some pharmacies used during the pre-exit period are excluded from the new plan’s network. Additionally, the share of familiar drugs in these enrollees’ prescriptions decreases by approximately 8% immediately after the exit, indicating that new drug formularies and new providers prompt these enrollees to change their medication.

I rule out an alternative explanation that post-exit disruptions stem from changes in the mix of plans and their causal effects on utilization, showing that such disruptions occur among beneficiaries who switch to plans with both higher and lower plan effects on utilization (Appendix F).

This paper contributes to the literature in three main areas. First, it adds to the empirical literature on government procurement, and specifically switching costs in the procurement of services (see Farrell and Klemperer, 2007 for a general review of switching costs). Switching costs also arise in the dynamic procurement of other services, such as computer and IT systems (Greenstein, 1993, Lewis and Yildirim, 2006), and regulators may take them into account when contracting with insurers in regulated markets. Despite the large size of the Medicaid Managed Care program, economic research on states’ procurement within this program remains limited (see Layton et al. (2018) for a review). Closest to this paper, Frenier (2022) studies a single MMC bid in Minnesota in 2016, examining the effect of plan exits on enrollees from exiting plans. He finds substantially lower utilization across a wide range of health care services. Other papers in the area have examined the effects of payment rates in MMC contracts and their updating over time (Layton and Politzer, 2024) and the impacts of MMC automatic assignment rules of enrollees to plans (Marton et al., 2017).

Second, this paper extends the literature on the effects of disruptions in health care. Recent studies have primarily focused on disruptions to the relationships between patients and their primary care providers, mostly due to retirement or relocation (Schwab, 2018, Kwok, 2019, Sabety, 2021, Zhang, 2022, Staiger, 2022, Bischof and Kaiser, 2021, Simonsen et al., 2021). Only a few studies have examined disruptions at the insurer-level. Duggan et al. (2018) find that hospital utilization increased among Medicare Advantage beneficiaries who switched to traditional Medicare after their plan exited the market. Other papers provide observational evidence that changes in provider networks after a plan switch can harm relationships between patients and their physicians (Barnett et al., 2017, Lavarreda et al., 2008). This work is among the first studies (along with (Frenier, 2022)) to causally identify the impact of involuntary plan switching within a regulated market, demonstrating that plan switches can disrupt relationships not only with familiar primary care physicians, but also with specialists and pharmacies. Assessing the effects of plan switching is particularly important in the fragmented U.S. health care system, where no plan offers health insurance from cradle to grave and switching between health plans or types of health coverage is inevitable. A significant share of these switches are involuntary, both in employer-sponsored insurance (Cebul et al., 2011, Cunningham and Kohn, 2000), and in Medicaid’s and Medicare’s regulated markets (Ndumele et al., 2017, Jacobson et al., 2016).2

Third, the paper is related to the literature studying switching and plan choice in health insurance. This literature often uses structural choice models to estimate the cost of individuals’ switching frictions, that may include also inattention, information frictions, hassle costs. and other factors (e.g. Heiss et al., 2021, Polyakova, 2016 in Medicare Part D, and Handel, 2013, Handel and Kolstad, 2015 in employer-sponsored insurance). This paper does not estimate individuals’ implied disutility from switching plans, but instead examines the real-world effects of bid-induced switching on health care utilization. The disruptions and adverse health outcomes resulting from switching may support a rational explanation for the observed inertia of enrollees in their MMC plan (Marton et al., 2017).

The rest of the paper proceeds as follows. Section 2 presents the setting and the data. Section 3 outlines the empirical framework, and Section 4 presents the results. Heterogeneity in the results is explored in Sections 5 Heterogeneity, 6 Mechanisms examines possible mechanisms. Section 7 discusses the findings, and Section 8 concludes.