Government contracting of public services to private firms is highly prevalent in the United States, accounting for 10% of GDP. In particular, private provision is increasingly common in the context of public health insurance in the US. Under Medicaid, the largest government-funded means-tested health insurance program, beneficiaries increasingly receive coverage through private managed care plans that have contracts with the government, rather than getting it directly through the public program. In 2016, over 80% of beneficiaries received Medicaid coverage through private managed care plans, a pronounced increase from only 10% in the early 1990s (Duggan and Hayford, 2013, Centers for Medicare & Medicaid Services, 2018). Government fiscal spending on private managed care plans also increased substantially, from $61 billion in 2007 to $269 billion in 2016, a figure that represents almost half of total Medicaid expenditures.

Despite this increased reliance on private Medicaid Managed Care (MMC) plans, the effects of Medicaid privatization on beneficiaries are not fully understood. First, the theoretical consequences of the two key features of Medicaid privatization—competition and capitation—are ambiguous. Competition among plans for beneficiaries could lead to quality improvement, but at the same time plans’ cost saving incentives under capitation could lead to quality deterioration (Hart et al., 1997). While this is consequently an empirical question, limited empirical work has been done on care quality or beneficiary satisfaction, with most work instead focusing on Medicaid privatization’s effects on government health care spending and coarse health outcomes, documenting mixed effects.1,2 The limited empirical work on this question arises from two longstanding challenges. First, private Medicaid plan data has been difficult to obtain, making private Medicaid plans a black-box for researchers. Second, administrative claims data used in past studies has limited information on care quality, outcomes, or beneficiary satisfaction.

In this paper, we overcome these past challenges through a number of innovations, to ultimately identify the effects of Medicaid privatization on beneficiary well-being. Given that cost-sharing is effectively non-existent under both private and public Medicaid, our focus notably is on non-financial dimensions of well-being.

To start, we develop a novel approach to measuring beneficiary satisfaction, using a ‘revealed preference’ framework that is based off beneficiaries’ relative likelihood of Medicaid attrition depending on whether they are initially assigned to a private Medicaid plan or the public fee-for-service option. Underlying our framework is the assumption that beneficiaries are ‘voting with their feet,’ with higher attrition reflecting lower degree of satisfaction (Geruso et al., 2020). Given the high rate of program attrition observed for Medicaid, this constitutes an important policy issue. For example, historically, about half of all Medicaid children who left the program in a given year did so despite still being eligible (of those, about 45% obtained new private insurance, while 55% became uninsured; Sommers, 2005). The largely voluntary nature of attrition, as well as the significant fraction shifting to alternative sources of coverage, lends support to our argument that attrition could indicate dissatisfaction with Medicaid. Furthermore, this is of important policy relevance given low Medicaid take-up rates of around 70% among those eligible nationwide. These fall below 50% for certain states, and below 35% for certain high-need populations (Wright et al., 2017).3

Second, we combine this framework with plausibly exogenous variation in public versus private Medicaid coverage assignment in New York: automatic enrollment in the public option for infants weighing below 1,200 grams, alongside mandated private Medicaid enrollment for those above that threshold. Finally, we leverage unique enrollment and claims data covering all New York infants in public as well as private Medicaid, and comprehensively tracking their claims across both programs. These data are individually linked to beneficiaries’ original birth weights, and longitudinally track beneficiaries’ Medicaid enrollment and utilization for up to seven years.

We empirically confirm the presence of a sharp discontinuity in private versus public Medicaid assignment at birth, around the 1,200-gram threshold; those born above the threshold appear 49 percentage points more likely to be in private Medicaid in the month of birth than infants born below the threshold. The enrollment discontinuity around the threshold persists through the first few months after birth, and then attenuates subsequently, in line with policy rules that lift the exclusion from private Medicaid at the age of six months.

Leveraging this quasi-random variation, we proceed to examine the implications of initial assignment to private versus public Medicaid on subsequent Medicaid program participation. We find strikingly large effects: getting assigned to private instead of public Medicaid at birth increases the probability of still being in Medicaid at age 4 by 37.7 percentage points, while increasing the cumulative time in Medicaid by 12.5 months over the first 48 months. The substantially higher Medicaid retention following the initial enrollment in private Medicaid has important policy implications. Understanding the differences between private and public Medicaid will shed light on relatively low observed Medicaid take-up rates nationwide, uncertainty around the root causes of this low take-up, as well as policy debates around how to best increase take-up. The stark difference in Medicaid retention is also relevant, given the potential adverse health consequences from lack of health insurance coverage, as well as from high churn in coverage.

We further investigate potential mechanisms underlying the large Medicaid retention effects of private Medicaid. Lower attrition rates among those initially assigned to private as opposed to public Medicaid could plausibly be beneficiary-driven, and reflect greater beneficiary satisfaction under private Medicaid, with beneficiaries ‘voting with their feet.’ However, lower attrition could alternatively be driven by plans, thereby not necessarily reflecting increased beneficiary satisfaction with private Medicaid. We find evidence in support of beneficiary-driven explanations, such as observably better access to care and increased use of routine care in private Medicaid in comparison to public Medicaid. As would be expected under this mechanism, we also find more pronounced attrition effects among families with greater expected sensitivity to care quality, such as those with relatively sicker infants or with mothers historically more proactive about health care (based on proxies such as those enrolling in Medicaid earlier in their pregnancies).

We then examine alternative plan-driven mechanisms, under which attrition would not necessarily be reflective of beneficiary satisfaction. One potential plan-driven mechanism could be reduced administrative costs of renewing Medicaid coverage, in comparison to public Medicaid. For example, private Medicaid plans may provide added reminders for beneficiaries to renew or offer better assistance with the re-enrollment process for beneficiaries than the public Medicaid system. In reducing administrative burdens or pushing to retain beneficiaries, private health plans may focus in particular on lower cost beneficiaries, who could be more profitable for the plans to cover. To test these explanations, we begin by looking at the degree of advantageous selection, and whether the attrition reduction is more pronounced for beneficiaries expected to be lower cost and more profitable. Inconsistent with this explanation, we find attrition reductions are actually greater among higher cost beneficiaries, that is, those with high costs of health care relative to capitation payments received. Moreover, in the analysis described below, we do not find evidence of an untargeted, general increase in re-enrollment under private Medicaid plans.

Specifically, to further rule out these alternative, plan-driven explanations for our effect, we leverage a difference-in-differences (DD) approach using the staggered roll out of private managed care enrollment mandates, on a county-by-county basis over time. With this approach, we find no evidence of advantageous selection by plans, as we again see more pronounced attrition reductions among lower birth weight infants who are on average less profitable.4 At the same time, we find that private versus public Medicaid exposure has no effect on future Medicaid enrollment among infants with average birth weight ranges. These results are inconsistent with systematically lower administrative costs of Medicaid re-enrollment under private plans, relative to public Medicaid. Given we would operationally expect lower administrative costs of re-enrollment to extend to all newborns in a plan, we would likewise expect to see a corresponding attrition reduction for all newborns, and not just ones with lower birth weight as we actually observe. In addition, we find greater utilization of preventive care among lower birth weight infants, further suggesting that our results are consistent with sicker enrollees being more responsive to coverage quality, resulting in a greater enrollment response. Taken together, our results suggest that the attrition effects are likely to reflect beneficiaries ‘voting with their feet’ and preferring private to public Medicaid.

Our study builds on existing literature on Medicaid privatization’s impact on enrollment, notably Currie and Fahr (2005), which looks at how privatization impacts selection into Medicaid among low-income children. By contrast, we focus on Medicaid privatization’s impact on selection out of Medicaid among existing beneficiaries, in part because of the arguably closer link between program attrition and beneficiary satisfaction. Focusing on attrition out of Medicaid additionally allows us to examine potential mechanisms, which is less feasible when looking at selection into Medicaid as Currie and Fahr (2005) did.

More broadly, our paper contributes to the literature on efficiency trade-offs in privatization of public health insurance programs, particularly Medicaid and Medicare in the US. Past work on privatization of these programs has focused primarily on its impact on health care usage or government fiscal spending, with much more limited focus around care quality or beneficiary well-being (Brown et al., 2014, Cabral et al., 2018, Curto et al., 2019, Duggan et al., 2018).5 This gap is unfortunate, given that understanding privatization’s impact on beneficiary well-being is pre-requisite to understanding its overall welfare implications, and is thereby a gap that our paper looks to fill.

In doing so, we contribute to a small but growing literature that provide insights on the impacts of Medicaid privatization on quality of care. Most recently, Layton et al. (2019) examine county-level MMC mandates for adults with disabilities in Texas. They find improvements in health care under privatization, as measured by increased consumption of high-value drug treatments and reduced avoidable hospitalizations. Their results are contrary to previous works that suggest worse care under privatization. For example, Aizer et al. (2007) study mandatory enrollment of pregnant women in California and find lower quality of prenatal care and worse birth outcomes under MMC. Kuziemko et al. (2018) show that cream-skimming incentives of MMC plans lead to widening infant health disparities in Texas, exploiting the county-by-county transition from FFS to MMC. While most papers use a staggered implementation of MMC for identification, ours take a novel approach by using a birth weight based RD. Our results add to this literature by providing evidence of improved quality under privatization for a high-cost/disabled population, consistent with Layton et al. (2019). We also introduce a new revealed-preference approach to measuring quality, which has not previously been applied in the Medicaid context, giving our findings on quality here additional weight.

This finding of improved quality also adds to the recent literature that documents potential advantages of managed care, such as cost savings without harming patients (Dranove et al., 2021, Brot-Goldberg et al., 2023). This is contrary to beliefs that led to the public backlash against managed care in 1990s (Blendon et al., 1998, Mechanic, 2001, Cooper et al., 2006). This may be particularly true in the context of Medicaid, where the alternative non-managed care system is also subject to unfavorable budgetary and political constraints (Layton et al., 2019).

Further, our paper is related to the literature on take-up of public social programs (Currie, 2006), and take-up of Medicaid specifically, for which take-up rates have been estimated at around 70 percent (Swartz et al., 2015, Wright et al., 2017). The drivers of this low take-up are not well understood, even though such understanding could substantially inform policy efforts around increasing program enrollment. Our findings point to program quality as an important driver of take-up, and indicate that improved program quality could ultimately increase participation in Medicaid as well as other social programs, especially for the most vulnerable subgroups who may be sensitive to the quality of services. Finally, our paper is related to the recent literature on the value of health insurance and Medicaid (Finkelstein, Hendren, and Shepard, 2019b; Finkelstein, Hendren, and Luttmer, 2019a). We build on this literature by showing the value of health insurance coverage to not just be a function of financial risk protection levels, but also of coverage quality broadly defined. Unlike the Finkelstein et al. framework, ours allows for beneficiaries to differentially value program quality, given we find higher spending beneficiaries to ultimately be most quality sensitive. In doing so, our paper also contributes to the literature on cost-quality tradeoffs in the design of public programs.

The remainder of the paper is organized as follows. Section 2 provides background on Medicaid contracting and the exclusion policy in New York State. Section 3 describes our data. Section 4 describes our empirical strategy. Section 5 presents our results, and Section 6 discusses potential mechanisms. Section 7 presents several robustness tests including the difference-in-differences estimation. Section 8 concludes.