Increase in Medicaid under the ACA Reduces Uninsurance, According to Early Estimates
Lisa Clemans-Cope, Michael Karpman, Adam Weiss, and Nathaniel Anderson
June 25, 2014
One of the chief goals of the Affordable Care Act (ACA) is expanded health insurance coverage. An important strategy for reaching that goal is expanded enrollment in Medicaid, which provides free or very low cost health insurance to low-income people. In January 2014, the ACA set a nationwide standard for extending Medicaid eligibility to nearly all Americans under age 65 with family income up to 138 percent of the federal poverty level (FPL), about $16,100 annually for an individual. Before the ACA, Medicaid coverage options for low-income adults were limited in most states.
But enrollment in Medicaid under the ACA was dampened by two factors. First, the US Supreme Court ruled that the expansion of Medicaid eligibility under the ACA is a state option. So far, only about half of the states have decided to expand Medicaid1—putting many poor, uninsured individuals in the nonexpansion states into the so-called “coverage gap.”2 Second, early problems with the federal health insurance website, which famously stymied enrollment in private Marketplace plans, prevented many low-income people from signing up for Medicaid by hindering the transfer of enrollment information from HealthCare.gov to state Medicaid programs or by preventing the collection of application information altogether. Some state-run Marketplaces had similar glitches.
Despite these setbacks, the Centers for Medicare & Medicaid Services (CMS) reported that over 6 million additional individuals had enrolled in Medicaid or the Children’s Health Insurance Program (CHIP) between October 2013 and the end of April 2014—a 10.3 percent increase compared with the average monthly enrollment for July through September 2013.3 An earlier CMS report indicated that only 3 million had enrolled by the beginning of March 2014, a time period more consistent with the data presented in this brief.4
Yet simple enrollment counts do not fully answer important questions for those tracking how health insurance coverage has been changing under the ACA: How much of the increase in Medicaid coverage is a net gain in insurance coverage rather than a shift to Medicaid from other coverage? Are there differences in the patterns of Medicaid changes across states and among different population subgroups? This brief takes advantage of new data from the Health Reform Monitoring Survey (HRMS) to address these issues.
What We Did
We use the March 2014 (quarter 1) HRMS to examine recent changes in coverage in Medicaid and other state health insurance programs, referred to in this brief as “Medicaid/state” coverage, as well as uninsurance rates relative to data collected in 2013. This brief expands on an initial set of findings released on April 15, 2014 (Long et al. 2014), by including information on Medicaid/state coverage changes.
We report regression-adjusted trends that correct for the effects of observed shifts in the characteristics of the survey respondents across survey quarters. Our analysis compares changes in the share of nonelderly adults (age 18 to 64) who report coverage through Medicaid/state programs between 2013 and quarter 1 of 2014. We place these shifts in public enrollment in the context of the national decline in the uninsurance rate and show how these measures change over time for (1) all nonelderly adults, (2) adults in states that have and have not adopted the ACA’s Medicaid expansion option, and (3) key sociodemographic groups. Our analysis focuses on changes between September 2013, just before the October 1 launch of the first health insurance Marketplace open enrollment period, and March 2014, when the most recent round of the HRMS was fielded.
Estimating enrollment in Medicaid and other coverage types accurately based on survey data is challenging. Specifically, two potential estimation issues stand out. First, previous research has shown that survey estimates of enrollment in government assistance programs are generally significantly lower than enrollment counts based on administrative records, and that this enrollment “undercount” is particularly pronounced for Medicaid enrollment.5 Despite the overall undercount, the misreporting of Medicaid coverage in surveys actually goes both ways—some overreporting of Medicaid in surveys has been found to partially offset a larger Medicaid undercount (Davern et al 2009).
A second issue with estimating coverage type via survey data is that qualitative testing of responses to coverage questions based on research with the Current Population Survey (CPS) and other surveys reveals that many respondents are unable to identify what type of coverage they actually have—leading to reporting of the same plan in multiple coverage categories, misreporting one type of coverage for another, and failing to report coverage altogether (Pascale 2008; Pascale et al. 2013). In sensitivity analyses, we found evidence that respondents reported multiple coverage categories and inconsistently reported coverage across similar questions.6 Because the challenges in estimating enrollment in Medicaid in the HRMS survey data also affect federal survey data, definitive analyses of changes in Medicaid enrollment requires the analysis of detailed administrative data that is not yet available.
The measure of Medicaid/state coverage examined in this brief is based on a measure of any reported Medicaid or other state coverage program (referred to as “any Medicaid/state” coverage). It is constructed from responses to three survey questions probing coverage status and coverage type at the time of the survey. First, respondents who selected “Medicaid, Medical Assistance (MA), the Children’s Health Insurance Program (CHIP), or any kind of state or government-sponsored assistance plan based on income or a disability,” in response to the main survey question on their current health insurance coverage are counted as having Medicaid. For respondents in states with state-specific names for these health insurance programs, the question was adapted to include the program name. Respondents could select multiple coverage types in response to this survey question.7
Second, if no specific coverage type was selected, respondents were asked to verify whether they have any coverage and to provide the name of that coverage in a write-in response.8 Respondents who indicated coverage in a state Medicaid program (e.g., “MassHealth”) are counted as having Medicaid.9 Third, in the HRMS March 2014 survey, respondents who reported Medicaid or other nonspecified coverage were asked whether they enrolled in Medicaid through their state’s health insurance Marketplace; those who indicated such coverage are counted as having Medicaid. Because this question was not asked in earlier rounds of the survey, we tested the sensitivity of our results to the use of this additional question to classify coverage types.10
Although each round of the HRMS is weighted to be nationally representative, we use a regression model to control for changes in the characteristics of the sample over time that may not be fully captured by the weights. To produce estimated changes in Medicaid/state coverage and uninsurance rates, we estimate a weighted regression model that controls for sex, age, language, race or ethnicity, whether any children are present in the household, education, census region, marital status, homeownership status, an urban or rural indicator, household Internet access, household income, and family income. In presenting the regression-adjusted estimates, we use the predicted rate of uninsurance and Medicaid/state coverage in each quarter for the same nationally representative population. For this analysis, we base the nationally representative sample on survey respondents from the most recent 12-month period of the HRMS (i.e., quarters 2, 3, and 4 of 2013 and quarter 1 [March] of 2014). By controlling for the sociodemographic characteristics of the sample, we remove variation in insurance coverage that can be explained by the observed characteristics of people in each quarterly sample rather than by the change in health insurance options available under the ACA (which is the concept needed to estimate ACA-related change). See tables 3 and 4 for unadjusted estimates.
Finally, because about 80 percent of respondents completed the quarter 1 2014 survey by March 6, our estimates do not take into account the rapid increase in Medicaid/state coverage coinciding with the Marketplace enrollment surge that took place at the end of March.11 Our results also do not show increases in Medicaid reporting caused by expansions in Medicaid that occurred before October 2013.
What We Found
The share of nonelderly adults reporting any Medicaid/state coverage increased by an estimated 1.8 percentage points (95% CI[0.5, 3.1]) between September 2013 and early March 2014 (figure 1, table 1). Applying this increase to the national population of nonelderly adults from census projections yields an estimated increase of 3.6 million (95% CI [1.0 million, 6.2 million]) in the number of adults reporting any Medicaid/state coverage between September 2013 and early March 2014.
States that expanded Medicaid by March 2014 experienced a statistically significant, 4.0 percentage-point (95% CI [2.4, 5.7]) increase in the proportion of adults reporting any Medicaid/state coverage, which appears strongly associated with the 4.0 percentage-point (95% CI [2.7, 5.4]) decrease in the uninsurance rate in those states. The increase in any reported Medicaid/state coverage in expansion states is almost entirely among adults with family incomes at or below 138 percent of FPL (data not shown), and low-income adults also saw the largest decline in uninsurance rates of any income group (Long et al. 2014). In sharp contrast, there is no significant change in any Medicaid/state coverage in states that did not implement the Medicaid expansion by March 2014, suggesting that the 1.4 percentage point (95% CI [0.2, 2.7]) decline in uninsurance in those states may have been related to changes in other types of coverage.
The gap in any reported Medicaid/state coverage between states that expanded Medicaid under the ACA and states that did not widened from 2.0 percentage points in September 2013 to 6.3 percentage points in March 2014. The gap in uninsurance rates between the two groups of states increased from 3.1 percentage points to 5.7 percentage points over the same period.12 The difference in gains in any reported Medicaid/state coverage appears to be increasing the gap in uninsurance rates between expansion and nonexpansion states.13
Strong gains in any Medicaid/state coverage accompanied falling uninsurance rates among low-income adults, younger and older adults, white non-Hispanic adults, and women (figure 2, table 2). The share reporting any Medicaid/state coverage among low-income adults with family incomes at or below 138 percent of FPL increased by an estimated 5.1 percentage points (95% CI [0.3, 9.9]) between September 2013 and March 2014. These low-income adults saw a 4.7 percentage point (95% CI [1.5, 7.9]) decline in their uninsurance rate over the same period. These findings suggest that the decline in uninsurance for low-income adults was associated with an increase in Medicaid coverage. In contrast, the share reporting any Medicaid/state coverage in the middle-income group (between 138 and 400 percent of FPL) increased by only 1.6 percentage points (95% CI [0.1, 3.0]). In addition, the increase in any reported Medicaid/state coverage among low-income adults was large and significant in Medicaid expansion states, but the change was not significant in nonexpansion states (data not shown).
We found similar patterns for other population subgroups, though estimated gains in any Medicaid/state coverage were not always statistically significant (particularly for smaller population subgroups). The share of young adults (age 18 to 30) and older adults (age 50 to 64) who reported having any Medicaid/state coverage increased 3.0 percentage points (95% CI [0.7, 5.2]) and 2.2 percentage points (95% CI [0.7, 3.7]), respectively. This appears to account for a substantial portion of the change in uninsurance rates for these age groups. In contrast, increases in any Medicaid/state coverage were smaller and not statistically significant for middle-age adults (ages 31 to 49), even though their uninsurance rate fell 1.8 percentage points (95% CI [0.2, 3.4]).
Within racial and ethnic groups, the increase in any Medicaid/state coverage was only significant for white, non-Hispanic adults (1.5 percentage points, 95% CI [0.3, 2.6]). Nonwhite, non-Hispanic adults and Hispanic adults experienced estimated increases in insurance coverage of 3.8 percentage points (95% CI [1.2, 6.3]) and 3.9 percentage points (95% CI[-0.2, 8.0]), respectively, but neither change in any reported Medicaid/state coverage was significant.
Coverage rates increased among both men and women. Though the increase in the rate of insurance coverage for women was smaller (1.8 percentage points; 95% CI [0.2, 3.4]), it was driven by a 2.3 percentage-point (95% CI [0.8, 3.9]) increase in any reported Medicaid/state coverage. In contrast, men reported larger coverage gains (3.4 percentage points; 95% CI [1.9, 5.0]) but smaller and statistically insignificant gains in Medicaid.
What It Means
This early HRMS evidence suggests that significant gains in Medicaid/state coverage occurred following the first Marketplace open enrollment period, especially in the Medicaid-expansion states. Further, the gains in Medicaid/state coverage were a major component of the reduction in uninsurance. Though the evidence to date suggests that the magnitude of the estimated changes in Medicaid coverage and uninsurance varies across data sources, these HRMS findings are consistent overall with findings from Gallup14 and Rand15 surveys, as well as the CMS Medicaid report for enrollment consistent with the reporting period for this brief.16 We estimate that the number of adults insured through Medicaid/state coverage increased by 3.6 million between September 2013 and early March 2014, at the same time as the number of uninsured fell by 5.4 million.
Given the Marketplace and Medicaid enrollment surge that began in March (after this quarter of the HRMS was essentially completed) and the most recent CMS Medicaid enrollment report, we expect gains in Medicaid/state coverage to have continued over these past two months. Because there is no open enrollment period for the Medicaid program, enrollment among new eligibles as well as “woodwork effect” enrollment among those who were already eligible will continue throughout 2014. Further, in addition to these new applications for coverage, many states are working through a backlog of Medicaid applications, including Medicaid eligibility determination files that were or are currently being transferred from Healthcare.gov,17 that can be expected to bring total Medicaid enrollment yet higher. Findings from the quarter 2 2014 round of the HRMS (currently in the field) will capture additional coverage changes since early March.
These substantial early gains in Medicaid coverage under the ACA are concentrated in the Medicaid-expansion states, but despite corresponding large decreases in the uninsurance rate for low-income adults in these states, many low-income adults still remained uninsured. Previous HRMS research shows that over two-thirds (64.7 percent) of the adults who remained uninsured in March were in the income groups eligible for Medicaid (Shartzer et al. 2014). Thus the opportunity remains for substantial additional Medicaid coverage gains in the Medicaid-expansion states.
In contrast, there have been no significant gains in Medicaid coverage in states that did not expand Medicaid. The uninsurance rate for low-income adults in those states fell only slightly and not significantly (Long et al. 2014), and few opportunities for coverage exist for these low-income individuals without the Medicaid expansion.18 Taking into account the state’s Medicaid expansion decisions, recent research suggests that, overall, states that expanded Medicaid reached a higher percentage of their projected total increase in Medicaid by 2016—those states reached 53 percent of projected Medicaid enrollment by April 2014, compared with nonexpansion states who reached only 29 percent (Dubay et al. 2014).
Call, Kathleen T., Michael E. Davern, Jacob A. Klerman, and Victoria Lynch. 2013. “Comparing Errors in Medicaid Reporting across Surveys: Evidence to Date.” Health Services Research 48 (2 Pt 1): 652–64.
Carman, Katherine Grace, and Christine Eibner. 2014. “Changes in Health Insurance Enrollment Since 2013.” Santa Monica, CA: RAND Corporation.
Davern, Michael E., Jacob A. Klerman, David K. Baugh, Kathleen T. Call, and George D. Greenberg. 2009. “An Examination of the Medicaid Undercount in the Current Population Survey (CPS): Preliminary Results from Record Linking.” Health Services Research 44 (3): 965–987.
Dubay, Lisa, Genevieve M. Kenney, Matthew Buettgens, Jay Dev, Erik Wengle, and Nathaniel Anderson. 2014. “Measuring Medicaid/CHIP Enrollment Progress under the Affordable Care Act.” Washington, DC: Urban Institute.
Long, Sharon K., Genevieve M. Kenney, Stephen Zuckerman, Douglas Wissoker, Dana Goin, Katherine Hempstead, Michael Karpman, and Nathaniel Anderson. 2014. “Early Estimates Indicate Rapid Increase in Health Insurance Coverage under the ACA: A Promising Start.” Washington, DC: Urban Institute.
Pascale, Joanne. 2008. “Measurement Error in Health Insurance Reporting.” Inquiry 45 (4): 422–37.
Pascale, Joanne, Jonathan Rodean, Jennifer Leeman, Carol Cosenza, and Alisu Schoua-Glusberg. 2013. “Preparing to Measure Health Coverage in Federal Surveys Post-Reform: Lessons from Massachusetts.” Inquiry 50 (2): 106–23.
Shartzer, Adele, Sharon K. Long, and Stephen Zuckerman. 2014. “Who Are the Newly Insured as of Early March 2014?” Washington, DC: Urban Institute.
SNACC Phase IV. 2009. “Phase IV Research Results: Estimating the Medicaid Undercount in the National Health Interview Survey (NHIS) and Comparing False-Negative Medicaid Reporting in NHIS to the Current Population Survey (CPS).” Minneapolis: State Health Access Data Assistance Center.
SNACC Phase V. 2010. “Phase V Research Results: Extending the Phase II Analysis of Discrepancies between the National Medicaid Statistical Information System (MSIS) and the Current Population Survey (CPS) Annual Social and Economic Supplement (ASEC) from Calendar Years 2000-2001 to Calendar Years 2002-2005.” Minneapolis: State Health Access Data Assistance Center.
SNACC Phase VI. 2010. “Phase VI Research Results: Estimating the Medicaid Undercount in the Medical Expenditure Panel Survey Household Component (MEPS/HC).” Minneapolis: State Health Access Data Assistance Center.
About the Series
This brief is part of a series drawing on the Health Reform Monitoring Survey (HRMS), a quarterly survey of the nonelderly population that is exploring the value of cutting-edge Internet-based survey methods to monitor the Affordable Care Act (ACA) before data from federal government surveys are available. The briefs provide information on health insurance coverage, access to and use of health care, health care affordability, and self-reported health status, as well as timely data on important implementation issues under the ACA. Funding for the core HRMS is provided by the Robert Wood Johnson Foundation, the Ford Foundation, and the Urban Institute.
For more information on the HRMS and for other briefs in this series, visit www.urban.org/hrms.
About the Authors
Lisa Clemans-Cope is a senior research associate, Michael Karpman is a research associate, and Adam Weiss and Nathaniel Anderson are research assistants in the Urban Institute’s Health Policy Center.
The authors gratefully acknowledge the suggestions and assistance of Sharon Long, Stephen Zuckerman, and Katherine Hempstead.
1Michigan expanded Medicaid on April 1, 2014 and New Hampshire is slated to expand Medicaid eligibility in July 2014. See Centers for Medicare and Medicaid Services, “State Medicaid and CHIP Income Eligibility Standards.” Accessed June 18, 2014.
2Those with income below the federal poverty level who do not qualify for Medicaid are also not eligible for premium tax credits or cost-sharing subsides to reduce the cost of a private insurance plan in the newly created state-based insurance Marketplaces.
3Connecticut, Maine, and North Dakota are not included in this count. See Centers for Medicare and Medicaid Services, “Medicaid & CHIP: April 2014 Monthly Applications, Eligibility Determinations, and Enrollment Report,” June 4, 2014, accessed June 18, 2014.
4The CMS Medicaid report with the closest reporting period to the data presented in this brief shows a gain of 3 million in coverage through Medicaid or CHIP between October 2013 and the beginning of March 2014. Centers for Medicare and Medicaid Services, “Medicaid & CHIP: February 2014 Monthly Applications, Eligibility Determinations, and Enrollment Report,” April 4, 2014, accessed June 18, 2014.
5For example, a study comparing the 2005 Current Population Survey (CPS) to administrative Medicaid records showed that 41 percent of people who were enrolled in Medicaid according to administrative records were not reported to have Medicaid in the survey (SNACC Phase V 2010). Enrollment undercounts were also found in a study comparing administrative Medicaid records to the 2002 National Health Interview Survey (NHIS) (SNACC Phase IV 2009) and the 2003 Medical Expenditure Panel Survey Household Component (MEPS/HC) (SNACC Phase VI 2010). More recently, state-sponsored surveys were found to have an undercount of Medicaid of approximately 12 to 26 percent (Call et al. 2013).
6To explore the sensitivity of our results, we examined how reports of any Medicaid/state coverage interact with reports of nongroup coverage, including direct purchase of health insurance coverage either in the Marketplace or outside of the Marketplace. We compare those who report only Medicaid/state coverage and no other coverage types with those who may be less certain about their coverage and thus report multiple forms of insurance. We found that the increase in the share of adults who report only Medicaid/state coverage is less than half of the increase in the share reporting any Medicaid/state coverage (data not shown). Most of the discrepancy can be explained by an increased share of lower-income adults (those with incomes at or below 138 percent of FPL) reporting both Medicaid and nongroup coverage. A smaller share of the discrepancy comes from the same reporting concern among middle-income adults (those with incomes between 138 percent and 400 percent of FPL). For low-income adults, coverage options other than Medicaid are largely unavailable because Marketplace subsidies are not available to them. Thus, it appears reasonable to assume that insurance coverage is more likely to be Medicaid for low-income adults who report both Medicaid and nongroup coverage. For middle-income adults, who report both Medicaid and nongroup, their choice might reflect the fact that some have enrolled in Medicaid via the Marketplace.
9In addition, a process of making logical coverage edits is used to classify ambiguous write-in responses (e.g., Blue Cross) as Medicaid for respondents who are not employed and report that they do not pay a deductible.
10Out of 1,100 respondents in quarter 1 who report any Medicaid/state coverage, 34 are placed in this coverage category solely based on their response to the question on coverage in Medicaid through the Marketplace (i.e., they reported a nonspecified coverage type, not Medicaid/state, in the survey’s main question on current coverage and they did not indicate Medicaid in a write-in responses). Excluding this segment from the share of respondents with any Medicaid/state coverage yielded similar findings. It is likely that the new survey pathway to identifying Medicaid enrollment through the Marketplace exacerbates challenges in measuring Medicaid coverage in surveys, particularly as more states obtain waivers to expand Medicaid by directing funds toward premium assistance for the purchase of private health insurance plans through the Marketplace.
11In May 2014, the Centers for Medicaid and Medicare Services reported that, compared with July–September 2013, approximately 3 million additional individuals were enrolled in Medicaid or CHIP by the end of February 2014 and 4.8 million were enrolled by the end of March 2014. In June 2014, CMS reported that, by the end of April 2014, the enrollment gain had increased to more than 6 million from the July-September 2013 baseline. See Centers for Medicare and Medicaid Services, “Medicaid & CHIP: March 2014 Monthly Applications, Eligibility Determinations, and Enrollment Report,” May 1, 2014, accessed June 18, 2014, and Centers for Medicare and Medicaid Services, “Medicaid & CHIP: April 2014 Monthly Applications, Eligibility Determinations, and Enrollment Report,” June 4, 2014, accessed June 18, 2014.
12We used 2014 national population predictions available from the US Census Bureau. These files give population projections by race, ethnicity, and sex of all ages from 2012 to 2060 based on estimated birth rates, death rates, and net migration rates over the time period. Using the “Table 1, Middle Series” file (which has a 2014 projected population of 318,892,103), we summed the 2014 population projections for all 18–64 year-olds to arrive at 198,461,688 nonelderly adults in 2014. See US Census Bureau, “2012 National Population Projections: Downloadable Files,” US Department of Commerce, revised May 15, 2013, accessed June 18, 2014.
14We find similar outcomes if we use the coverage hierarchy applied in previous HRMS briefs: the gap in Medicaid/state coverage between expansion and nonexpansion states widens from 2.7 percentage points to 5.6 percentage points. Alternatively, if we focus on the share of respondents who only report Medicaid/state coverage and no other coverage types, the gap widens from 2.5 percentage points to 4.1 percentage points.
15Gallup reported that the share of Americans age 18 to 64 who report having health insurance through Medicaid rose to 9.0 percent in April 2014 from 6.9 percent in the fourth quarter of 2013. See Jenna Levy and Dan Witters, “More in U.S. Have Self-Funded Health Coverage, Medicaid,” Gallup Well-Being (blog), May 9, 2014.
16Rand estimated that Medicaid enrollment increased by 5.9 million (with a confidence interval spanning 3.1 million to 8.7 million) between September 2013 and March 2014. See Carman and Eibner (2014).
17Centers for Medicare and Medicaid Services, “Medicaid & CHIP: February 2014 Monthly Applications, Eligibility Determinations, and Enrollment Report,” April 4, 2014, accessed June 18, 2014.
18According to a survey by CQ Roll Call, “at least 2.9 million Americans who signed up for Medicaid coverage as part of the health care overhaul have not had their applications processed, with some paperwork sitting in queues since last fall.” These estimates include data from 41 states as of May 29, 2014. See Rebecca Adams, “The Hidden Failure of Obama's Health Care Overhaul,” CQ Roll Call, June 3, 2014.
19In states that did not choose to expand Medicaid, adults with incomes between 100 and 138 percent of FPL who do not have access to an affordable offer of employer coverage are eligible for Marketplace-based subsidies.