A randomized natural experiment finds that privatizing managed care can reduce drug costs largely by steering enrollees to cheaper substitutes rather than by denying them medication—while its medical-side savings are small and come with potential tradeoffs, including lower quality and lower satisfaction.

These findings are the result of research conducted by the SCALE Lab at the Yale School of Public Health with additional support from Yale’s Tobin Center for Economic Policy. The lab’s directors are Professors Chima Ndumele and Jacob Wallace.

This policy brief is based on:

Abstract

This paper examines the effects of privatizing social health insurance. We exploit a natural experiment in Medicaid, wherein nearly 100,000 enrollees were randomly assigned between a publicly operated fee-for-service system and private managed care. Managed care reduced costs by 5.6 percent via cost-effective substitutions among prescription drugs and via lower prices for outpatient services. We present evidence that pharmacy utilization management was the key mechanism reducing overuse and encouraging substitution to lower-cost drugs without decreasing observed quality. In contrast, privatizing medical benefits led to only modest savings and was associated with decreased health care quality and consumer satisfaction.

More than three-quarters of the roughly 80 million Americans on Medicaid now get their coverage through private managed care plans (Medicaid managed care, “MMC”), arrangements in which the state pays a private insurer a fixed amount per enrollee to organize and pay for that person's care. This represents a shift from the program's original design, fee-for-service (FFS), in which the government pays doctors and hospitals directly for each service they provide. The most commonly stated motivation for transitioning away from state-managed FFS toward privately administered MMC is the potential for cost savings. That potential, however, has been challenging to test, because beneficiaries with different needs may select different models, leaving uncertainty about whether privatizing Medicaid has reduced spending in Medicaid.

New research from Danil Agafiev Macambira, Anthony Lollo, Chima Ndumele, and Jacob Wallace (Yale University) and Michael Geruso (University of Texas at Austin) measures the causal effect of private versus public provision of health insurance. The researchers use a natural experiment in Louisiana Medicaid, where nearly 100,000 enrollees were randomly assigned between FFS and MMC. Because assignment was random and not based on beneficiary selection, the researchers are able to identify how privatization impacts costs, the care people receive, and whether they stay with their plan.

What We Learned

Managed care reduced health care spending. Enrollees randomly assigned to private managed care spent 5.6% less on health care, about $82 less per person per year—a significant reduction compared to FFS.
These savings came almost entirely from prescription drugs, not medical care. Pharmacy spending fell 18% over the study period, and 25% once MMC plans took full responsibility for drug coverage. By contrast, outpatient spending fell about 3%, and inpatient spending did not fall at all.
The main driver of prescription drug cost savings is substitution toward cheaper drugs that treat the same condition. MMC shifted prescriptions roughly one-for-one from brand-name drugs to generics, and toward lower-cost therapeutics within the same drug class.
On the medical side, the picture is more troubling. Assignment to MMC reduced access to primary care, modestly lowered use of recommended preventive services, and was associated with a 14% increase in avoidable emergency department visits—suggesting some enrollees may have used the ER due to insufficient access to primary care.
Enrollees appeared dissatisfied with managed care. Those randomly assigned to MMC were more than three times as likely to switch out of their plan as those assigned to the public option. Strikingly, most of that exit happened before the plans took over drug coverage, suggesting dissatisfaction was driven mainly by how MMC handled medical benefits, the domain where it saved the least, rather than by the drug substitutions that generated most of the savings.
In the absence of a randomized experiment, the cost savings of privatization are overstated. Outside of this experiment, the average difference in spending between people who enrolled in MMC versus FFS is more than three times larger ($266 per year). This is largely because sicker enrollees are more likely to opt for FFS, making the public option look more expensive by comparison.

Impact of Assignment to Managed Care on Spending, by Category

Four quarters before managed care began, quarterly inpatient (hospital) spending was $1 lower per enrollee for people assigned to managed care compared to fee-for-service.

Range of uncertainty: $5 lower to $2 higher (a range that includes zero).

Data and Approach

The study's central advantage is randomization. In 2012, when eastern Louisiana shifted to MMC, enrollees who did not choose a plan within 30 days were automatically assigned to one, either public (FFS) or private (MMC). This created a randomized comparison among roughly 95,000 enrollees, whose spending, health care use, drug prescriptions, quality of care, and satisfaction the researchers are able to observe from 2012 to 2014 in order to estimate the causal effect of the private versus public provision of social health insurance. The researchers also use detailed administrative claims data to uncover precisely how the plans cut costs.

In addition, the researchers use a second natural experiment to confirm their results. In 2015, Louisiana discontinued its public option entirely, forcing the last remaining public plan's enrollees to convert to private managed care. The reductions in spending—concentrated in prescription drugs—along with the same drop in primary care access and rise in avoidable emergency department visits, closely tracked those from the random-assignment analysis, increasing confidence that the results generalize beyond the initial study. The researchers do note, however, that Louisiana's pre-expansion Medicaid population in this period was particularly young (the average enrollee was about nine years old), so their study may have limited power to speak to inpatient and adult-heavy spending, where managed care's effects could differ.