In a hearing before the House Energy and Commerce Committee on October 23, 2019, Seema Verma, the administrator of the Centers for Medicare and Medicaid Services, testified that new data show significant problems with Medicaid eligibility. This is not surprising; in a National Bureau of Economic Research (NBER) working paper issued in August, my academic colleagues and I used microdata from the Census Bureau’s American Community Survey (ACS) from 2012 to 2017 to examine eligibility of the Affordable Care Act’s (ACA) expansion of Medicaid. We found evidence of substantial improper Medicaid enrollment following the expansion.
Recent postings by Tricia Brooks and by Judith Solomon and Matt Broaddus have criticized both these findings and a recent Wall Street Journal op-ed I co-authored with Brian Blase, who until recently served the White House National Economic Council. As I explain below, much of the criticism is incorrect and was preemptively addressed in the paper. Moreover, numerous other pieces of evidence indicate there is a serious problem of ineligible enrollment in the ACA’s Medicaid expansion programs.
NBER Working Paper Findings
In the aforementioned article, my colleagues and I compared nine states that had not expanded Medicaid coverage to the childless, working-age population at all prior to 2014 and that adopted the expansion in 2014, to 12 non-expansion states. We found that nine expansion states had significantly higher Medicaid coverage rates after implementation compared to 12 non-expansion states. Most noteworthy, we found significant increases in enrollment among those whose incomes potentially render them ineligible.
The income cutoff for the Medicaid expansion is 138 percent of the federal poverty level, roughly $36,000 for a family of four in 2019. There are some reasons why adults with incomes above this threshold could qualify for Medicaid (see explanation below), but people with incomes above 138 percent of poverty are generally not included.
We found that when states expanded Medicaid, enrollment by working-age adults with incomes above 138 percent of poverty rose 3.0 percentage points (from 2.7 percent to 5.7 percent, an increase of 111 percent of the base rate). We view this as potentially improper enrollment. Potentially improper enrollment increased over time, to more than twice as much in 2017 (3.7 percentage points) as in 2014 (1.5 percentage points).
Given that approximately 17.4 million working-age adults had incomes exceeding the Medicaid threshold in these states, even seemingly modest numbers such as these could translate into many improperly enrolled individuals. For example, if 3 percent of all people with income above 138 percent of poverty improperly enrolled in Medicaid, that translates into more than 500,000 people in just those nine expansion states.
This analysis has limitations, which we acknowledged and addressed. Several critics mischaracterized the analysis and results in an apparent attempt to downplay the potential problem of improper enrollment. I address those here and cite numerous government audits on improper Medicaid expansion enrollment. My coauthors and I also conducted additional robustness checks in response to the criticism, and the conclusion is the same—there are sizable numbers of ineligible Medicaid expansion enrollees.
One reason that someone who appears income ineligible might be Medicaid eligible is income volatility. Someone with low income in the month they apply would likely have been properly enrolled at application if it turns out his or her annual income exceeds 138 percent of poverty.
To address this, we also conducted the analysis for individuals with annual incomes exceeding 250 percent of poverty (approximately $65,000 for a family of four). Far fewer people who have income above 250 percent of poverty for the year will have income in any month that would lead them to qualify for Medicaid.
Our substantive conclusions scarcely changed in response to this analysis. In expansion states, Medicaid coverage increased among this group from 1.4 percent to 3.1 percent—by 1.7 percentage points or 121 percent of the base rate. Potentially improper enrollment is much higher in 2017 than 2014. This increase in Medicaid coverage well above the eligibility thresholds suggests serious eligibility problems.
For ACS respondents with incomes at or above 250 percent of poverty, there was sizeable growth in Medicaid enrollment in expansion states relative to non-expansion states (Exhibit 1).
Exhibit 1: Average annual Medicaid enrollment among adults ages 19–64 with incomes above 250 percent of federal poverty level
Source: Courtemanche J, Marton J, and Yelowitz A. Medicaid coverage across the income distribution under the Affordable Care Act. Cambridge (MA): National Bureau of Economic Research; 2019 Aug.
Yet, the critics ignore this evidence. In their argument, they cite a Health Affairs study on income volatility that restricted its sample to individuals below 200 percent of poverty and states: “Most people with incomes of 200–400 percent of poverty receive insurance through their employers and are unlikely to participate in Medicaid or exchange plans in large numbers; therefore, they were not included in the sample.”
Since this paper acknowledges that people above 200 percent of poverty would likely not be participating in Medicaid at any time during the year, we used a more conservative robustness check by examining individuals with incomes exceeding 250 percent of poverty.
After the criticism, however, we conducted further checks. We created a subsample of respondents with income above 250 percent of poverty who report working both full year (50 or more weeks) and full time (40 or more hours per week). Income volatility should be less important for full-time, full-year workers, who are also more affluent than the full sample of respondents with incomes above 250 percent of poverty. Prior to the expansion, Medicaid coverage was 0.4 percent for full-time, full-year workers (compared with 1.4 percent for respondents above 250 percent of poverty), and the percentage with employer-sponsored health insurance was 77.6 percent (compared to 69.9 percent for respondents above 250 percent of poverty). The impact of the Medicaid expansion on coverage remains highly significant. The expansion raised Medicaid coverage for full-time, full-year workers with income above 250 percent of poverty by 0.8 percentage points (200 percent above the base rate). The results once again show much larger increases in potentially improper Medicaid enrollment in 2017 (1.0 percentage points) relative to 2014 (0.4 percentage points).
American Community Survey Data Issues
The critics raise some other issues with ACS data quality and our empirical judgements. Brooks notes that Medicaid coverage is not “based on actual administrative enrollment data” and is relying on “unadjusted self-reported survey data as a proxy for actual Medicaid income eligibility and enrollment.” Solomon and Broaddus note “a meaningful share of respondents appear to misreport their source of insurance coverage, as significant differences between survey-based estimates and administrative data show.” These comments are correct, but they ignore that public health insurance coverage tends to be underreported in such surveys.
The Department of Treasury’s Office of Tax Analysis compared health insurance sources from the Internal Revenue Service tax form 1095 to measures from various surveys. For individuals younger than age 65, administrative tax data revealed 75.6 million covered life years from all public insurance sources, while the point of interview measure in the ACS revealed 66.6 million individuals. Another study in Health Services Research found that “starting in 2014, there was a large undercount in expansion states that was absent in non-expansion states,” leading to “downwardly biased estimates of expansion on means-tested coverage in the ACS relative to administrative records.” The undercount exceeded 10 percent in expansion states for every year between 2014 and 2016, with ACS data missing approximately 3.9 million Medicaid enrollees. In contrast, non-expansion states had Medicaid enrollment counts far closer to administrative sources. Taken together, these findings almost certainly mean our results understate the magnitude of improper enrollment in expansion states.
Another issue raised by the critics is the complexity of household size for calculating Medicaid eligibility. Brooks notes: “The ACS household requests information for everyone in the household, including non-married partners, in-laws, roommates, and other individuals who should not be counted in determining the household size or income for Medicaid.” In response, we examined the results by restricting the sample to nuclear families, in which all individuals in the household are a household head or couple and their children. Once again, we found significant effects of the expansion on potentially improper enrollment. Overall, the expansions lead to a 2.2-percentage-point increase in Medicaid coverage for nuclear families with incomes at or above 138 percent of poverty (from a baseline rate of 1.8 percent), and 1.2-percentage-point increase for those at or above 250 percent of poverty (from a baseline rate of 0.8 percent).
Other Pathways To Medicaid
Our principal findings do not change when we exclude working-age adults who may have alternative pathways to qualify for Medicaid. In particular, we exclude survey respondents with the other lawful pathways to qualify for Medicaid—those who reported having a baby in the previous year; reported disability; or reported income from public assistance, Supplemental Security Income, or Social Security. The impact of the ACA’s Medicaid expansions on improper enrollment scarcely changes. Medicaid enrollment among this group increases by 2.7 percentage points after implementation of the expansion, compared to 3.0 percentage points for the overall population. In fact, as a percentage of the base rate (now 1.4 percent rather than 2.7 percent), the effect is now considerably larger (193 percent compared to 111 percent).
Brooks mischaracterizes this finding, asserting that “while the sensitivity analysis appears to substantially change their results, the authors still focus on the unadjusted numbers.” On the contrary, the results from the sensitivity analysis that exclude adults who may have alternative pathways support rather than undercut the main findings.
Our critics argue ineligible enrollment is not a significant issue because states have procedures for verifying eligibility. Brooks states, “All income must be verified through trusted electronic sources when possible, or through documentation provided by the enrollees.” Solomon and Broaddus write, “Medicaid Programs Have Stringent Verification Procedures.” Of course, such assertions are merely statements of what government is supposed to be doing. Second, several government audits corroborate our findings, showing that existing verification procedures are inadequate and that there are sizeable numbers of ineligible expansion enrollees.
As of September 2019, the Office of Inspector General (OIG) at the Department of Health and Human Services has published results from a series of audits covering part of the 2014–15 period in four states (California, Colorado, Kentucky, and New York) for beneficiaries enrolled in the expansion. The OIG has also released reports on similar determinations for non-newly eligible adults in California, Kentucky, and New York. All the reports show that there are serious problems with eligibility-verification systems, and policy experts are naïve to take it on faith that they are working.
Consistent errors include neglecting to verify income eligibility properly, misclassifying individuals into incorrect category classifications, failing to properly verify citizenship, and other issues. Some of these enrollment errors lead to incorrect and often higher federal reimbursement for individuals who would qualify for Medicaid under a category other than “new adults,” while others lead to complete ineligibility for Medicaid. As examples:
- California’s “eligibility determination systems lacked functionality or eligibility caseworkers made errors…the State agency did not properly input application information and verify income or lawful presence.” The OIG estimated more than 366,000 ineligible and 79,000 potentially ineligible beneficiaries.
- In Colorado, “contrary to the provisions of its own verification plan, [the state] relied on self-attestations rather than income verifications.” In addition, “lags in both the eligibility system and the State agency’s reasonable compatibility process…delayed disenrollment.” The OIG estimated more than 85,000 ineligible and 13,000 potentially ineligible beneficiaries.
- In New York, the OIG estimated more than 47,000 ineligible beneficiaries. The OIG points to an example in which one beneficiary was enrolled after attesting to an income of approximately $35,000 with a household size of one, despite the income threshold being $16,105 for a household size of one.
- Kentucky “did not always meet Federal and State requirements when making eligibility determinations because of human and system errors.” The OIG estimated nearly 35,000 potentially ineligible beneficiaries.
A state-level audit in Louisiana noted serious deficiencies in the eligibility process as well. Unfortunately, imprecise language in the Louisiana audit led many media outlets to report that the state had performed a random sample of 100 Medicaid expansion enrollees and found 82 of them were ineligible. It was actually a targeted audit. Regrettably, the imprecise language was reflected in the Wall Street Journal op-ed.
The critics of our NBER working paper and Wall Street Journal op-ed did not raise any substantive issues that we did not already acknowledge and address in the paper. The evidence—whether broad survey data from the Census Bureau’s ACS or highly detailed audits from the OIG—suggests serious problems with program integrity related to the ACA’s Medicaid expansion. Based on Verma’s recent testimony, it appears we will soon have more information about the extent of the issue.