In New England, Partisan Differences In ACA Marketplace Participation And Potential Financial Harm

Abstract

Political orientation can be a powerful motivator of certain health care decisions. This study examines how political orientation was associated with decisions to use the Affordable Care Act Marketplaces to enroll in individual plans and whether it was also associated with adverse financial consequences. We used administrative records and surveys of nongroup marketplace enrollees from a large insurer in New England. Enrollees were categorized as Republican, Democrat, or independent on the basis of self-identification or were assigned to one of the political parties after responding to a political preference question. Republican enrollees of comparable subsidy eligibility were less likely than Democratic enrollees to enroll through the Marketplaces and receive subsidies. Among income-eligible enrollees, Republican subscribers received $66 a month less in premium subsidies than Democrat subscribers, equivalent to roughly $800 a year. However, this result varied by subgroups in the parties, and our results suggest that party effects on decision making may inversely relate to the magnitude of the financial consequence. Navigating the ongoing political polarization in the United States requires optimizing public policies, as well as the associated education and outreach, to ensure maximal efficacy regardless of political orientation.

The Affordable Care Act Marketplaces (also known as “exchanges”) are online insurance enrollment portals in the nongroup insurance market. In 2018 they distributed $52 billion in subsidies to income-eligible people who chose to use them, with projected distribution of $689 billion during the 2020–29 period.¹ Subsidies on the Marketplaces come in two forms: Advance Premium Tax Credits (APTCs), which are available to those with incomes between 100 percent and 400 percent of the federal poverty level, and within-plan cost-sharing reductions, which can be obtained by those with incomes between 100 percent and 250 percent of poverty. To an individual policyholder, these subsidies can be substantial. For instance, the average annual APTC value received among enrollees qualifying for subsidies was $5,010 in 2017 and $6,300 in 2018.² Despite the potentially large savings, however, estimates suggest that up to a third of nongroup market enrollees in some states have forgone subsidies for which they were eligible.^3,4 Hypothesized reasons for this forfeited assistance include limited public knowledge of subsidies, confusion about eligibility, and time burden of enrolling; growing evidence suggests that at least a portion of forfeited assistance may be related to a person’s political orientation.^5–7

Political orientation has been found to be a powerful motivator of behavior, including social, economic, and most recently, public health decisions such as those related to COVID-19 preventive measures and vaccine uptake.^8–10 The beliefs that drive political support have been found to be sometimes inconsistent with economic self-interest.^11–13 The extent to which partisanship factors into individual behavior with direct adverse financial consequences is less well-understood, but is plausible in instances when choices are stigmatized and are strongly identified with one political party and when political polarization is high.^14–16 Both conditions apply to the Affordable Care Act (ACA), which emerged from passage in 2010 as a political lightning rod and which remains divisive.¹⁷

A number of studies have investigated partisan dimensions to consumer participation in the ACA. One analysis found that partisan forces may have constrained gains in coverage through the individual insurance Marketplaces, which were created by the ACA, but that eligibility expansion of existing Medicaid programs increased insurance rates across political leanings.⁶ National survey data suggest that among those without group coverage, Republicans were significantly less likely to purchase plans on a Marketplace and were significantly more likely to purchase off-Marketplace plans.⁵ In addition, counties that voted for Barack Obama in 2012 had significantly higher Marketplace enrollment than those that did not.⁵ A major limitation of this prior work, however, is the reliance on self-reported data about insurance type. A recent validation study of survey reports of health insurance coverage by the census found that more than 40 percent of those who self-identify as having nongroup coverage could be misclassified.¹⁸ Moreover, self-reported measures of participation in the ACA Marketplaces or subsidy receipt may be politically biased and may misrepresent differences in true enrollment. Prior work is also unable to report on whether potential political avoidance of the Marketplaces has affected the receipt of subsidies.

This study addresses some of the limitations and unanswered questions of prior work, assessing both the strength and financial consequences of political affiliation for health insurance coverage decisions. We link survey data to the enrollment records of a sample of 2017 enrollees of on- and off-Marketplace nongroup plans from a large insurance carrier in New England. Using information on enrollee political orientation, income, subsidy receipt, and enrollment decisions we estimate the partisan differential in subsidy uptake and the financial cost associated with that differential in 2018.

Study Data And Methods

Data

This analysis was conducted within a larger study examining consumer preferences and behaviors in the nongroup insurance market.

Two surveys, one baseline and one follow-up, were sent to a stratified random sample of subscribers ages 18–63 from a single health insurer in New Hampshire, Maine, and Massachusetts. At the time of this study, potential subscribers in New Hampshire and Maine each used the federal platform (Healthcare.gov) to enroll in Marketplace plans, whereas potential subscribers in Massachusetts used a state platform. We targeted primary policyholders enrolled as individuals or with other family members in their plan. The survey population was stratified on the basis of state (New Hampshire, Maine, or Massachusetts) and enrollment source (on the Marketplace versus off the Marketplace). The baseline survey was sent by mail from March through September 2017, which follwed the 2017 enrollment period, to 1,300 randomly selected subscribers in each of the six strata, with the exception of off-Marketplace enrollees in Maine, where there were fewer than 1,300 enrollees total, so all enrollees in Maine (n = 706) were sent a survey (N = 7,206 subscribers total for all three states). A follow-up survey was sent by mail or email (according to respondent preference) to 1,859 baseline respondents from March through August 2018 after the completion of 2018 open enrollment. Both survey instruments underwent cognitive testing with nongroup enrollees before being fielded. For each survey, respondents received a $20 incentive for completion and up to three follow-up contacts were attempted to improve participation. A total of 1,223 subscribers completed both baseline and follow-up surveys for an overall response rate of 18 percent after accounting for ineligible subscribers (roughly 4 percent of the sample).

The baseline survey collected sociodemographic and clinical information including age, sex, education, income, family composition, and state. To assess the medical needs of the plan enrollees, respondents were asked whether they or any family members in their plan had a chronic health condition that has lasted or is expected to last a year or longer, may limit what one can do, and may require ongoing care. The follow-up survey included questions measuring political orientation and views of the ACA.

Survey data were merged with health insurance enrollment records including source of enrollment (Marketplace versus off Marketplace), enrollee demographics, federal subsidy amount (APTC) applied to the plan premium (if any), and health plan characteristics (including whether it was a cost-sharing reduction plan). Respondents from Massachusetts did not have administrative subsidy data available and did not include those enrolled through a separate state-specific program known as Connector Care, available to those with household incomes less than 300 percent of poverty.

Missing Data

Missing data were 10 percent for political party, 9 percent for household income, and less than 5 percent for all remaining covariates. Across all model covariates missing data were 18 percent. We implemented multiple imputation to estimate all missing item responses using the Amelia package in R, which is based on an expectation-maximization algorithm with bootstrapping. For each model in the study, results were computed and combined using appropriate formulas across twenty imputation data sets. In addition to model covariates, imputation models included several additional variables to improve predictive accuracy. To evaluate the influence of imputed data on our estimates, all analyses were repeated using only complete cases and our conclusions were unchanged (see online appendix for these results and further details on missing data, imputation).¹⁹

Measure Of Party Orientation

Each subscriber’s political orientation was measured by combining responses on two survey questions on the follow-up survey: one item that asked about party identification using categorical responses and a second item that collected feelings toward the Democratic and Republican parties, using a visual analog thermometer scale. For item one, respondents were asked whether they identify as a Democrat, Republican, independent, something else, or they would rather not say. The thermometer asked respondents to rate the Democratic and Republican parties on a 100-point scale ranging from “very cold” to “very warm.” Both are standardized measures that have been used in the American National Election Studies survey for many decades.²⁰ Respondents who did not identify as either a Democrat or a Republican (n = 692) were assigned a Democratic or Republican orientation on the basis of the party that they indicated warmer feelings toward in the thermometer question. Respondents who did not identify as either a Democrat or a Republican and who indicated equal feelings toward those parties, and respondents who identified as independent and who did not respond to the thermometer questions, were assigned an independent orientation (n = 296).

Subsidy Eligibility

To determine partisan differences in APTC receipt, we compared the amount of subsidy applied to premiums, available from administrative records, with the amount of subsidy for which subscribers were eligible, which we estimated. The amount of each subscriber’s 2017 federal premium subsidy (APTC) was determined by the amount by which that subscriber’s premium for the second-lowest-cost silver plan exceeded a statutorily defined level of affordability as a percentage of household income (levels of the percentage of income considered affordable varies by household income as a percentage of the federal poverty level). To estimate APTC eligibility amount, we first calculated the subscriber’s premium for the second-lowest-cost silver plan based within rating area and the number and ages of all plan members and then subtracted the subscriber’s affordable premium amount based on self-reported household income as a percentage of the federal poverty level in 2016. Household income was measured on the baseline survey, using a ten-category question ranging from less than $20,000 to more than $100,000 and set at the category midpoint; to deal with issues of top-coding, an income point estimate for the highest category was estimated on the basis of US census data. To evaluate the accuracy of our estimate of APTC eligibility, we compared it with the subsidy amount observed in our health plan administrative data for subscribers who received an APTC (n = 159), and a high degree of correlation was observed (Pearson r = 0.83).

Analysis

We report unadjusted descriptive statistics for all respondents and each party orientation, using chi-square tests of equivalence and t-tests to test for differences.

We analyzed the association between party orientation and source of enrollment (on- versus off-Marketplace) and whether the respondent claimed an APTC using multivariate logistic regression. We excluded Massachusetts enrollees from APTC analysis because of the lack of APTC data from this state. For these analyses, covariates included continuous measures of subsidy eligibility amount and household income, as well as categorical measures of age, sex, education, employment status, family composition, prior insurance coverage, and state. Covariates were selected on the basis of their association with either the outcome or covariate of interest and contribution to model fit based on Akaike’s Information Criterion. To check the robustness of covariate selection, the results were compared with those from a model with automated covariate selection, using a stepwise procedure.²¹ Analyses were stratified by income categories used to determine eligibility for premium and cost-sharing subsidies in marketplace plans (>400 percent of poverty, more than 250 percent to 400 percent or less of poverty, and 250 percent or less of poverty) and presence of an enrollee with a chronic condition in the plan. Models including interactions of political party with household income as a percentage of the federal poverty level and of political party with chronic condition were used to test for moderation effects. Results are reported as predicted probabilities from these models, standardized to the population from which the study sample was drawn. All unadjusted and adjusted analyses were weighted using inverse probability weights to account for sampling design and subject-level nonresponse.

Finally, we tested whether there were financial costs associated with politically associated enrollment behavior. We compared the amount of APTC received (drawn from health plan administrative records of APTC received in the last month the respondent was enrolled in 2017) among income-eligible Democrat subscribers with the amount of APTC received among income-eligible Republican subscribers. We used linear regression models to produce estimates of the amount of APTC received that were adjusted for the subscriber’s amount of subsidy eligibility and the other sociodemographic variables described above. For comparison, we produced the estimated APTC eligibility amount among income-eligible subscribers of each political orientation, adjusting for the same sociodemographics.

Limitations

Several limitations should be noted. First, findings from subscribers from a single insurance carrier in New England may not generalize to other nongroup market populations. However, given that New Hampshire and Maine use the federal Marketplace, findings may have a broader scope to the majority of US states that use HealthCare.gov. The populations of New England are also more politically liberal than the rest of the US, potentially affecting the nature of party effects; however, all three states have significant minorities of Republicans, and support of the ACA remains widely divergent among the parties. Second, our measurement of subsidy eligibility and amount are based on self-reported income and may not reflect data that would have been used for eligibility determination at the time of open enrollment. Third, the APTC amounts in health plan administrative records may not reflect final premium tax credit amounts, which become fully reconciled through a subscriber’s tax returns when filed. Differences among Democrats and Republicans observed in APTC administrative records may have resulted from differences in claiming credits in advance or estimating future income. To explore this issue, we estimated the financial consequences of political behavior using total APTC eligibility as an outcome rather than what was claimed in advance, which produced results consistent with our conclusions (see appendix).¹⁹ Fourth, this study used a measurement of political affiliation that differs from what is most commonly used in political polls; therefore, our categories of Democrat and Republican may vary from common categorizations. We chose to include the feelings thermometer as a validated way of gauging how people feel about the political parties without directly asking about party identity during a time of increasing partisan tensions. However, the proportions of party orientation in our data are similar to what other polls have found for the region. Robustness checks using alternative parameterizations of party yielded similar results (see appendix).¹⁹ Fifth, measurement of party occurred shortly after an attempt to partially repeal the ACA failed in the summer of 2017, and partisan sentiment toward the law may have been elevated. We believe that this is unlikely to have substantially affected party orientation, which research has shown to be highly durable over time.²² Finally, survey nonresponse is a limitation, although our response rate is consistent with other related work³ and we were able to weight observations for nonresponse, using administrative data.

Study Results

Among the sample of 1,223 enrollees, 538 were Democrats, 269 were Republicans, 296 were independent, and 120 did not report any party orientation or leaning (for example, data were missing) (exhibit 1). Among Democrats, 184 were in Massachusetts, 160 were in Maine, and 194 were in New Hampshire compared with 53, 94, and 122 among Republicans, respectively. Republican respondents were less likely to be female than Democrats and independent voters. Democrat respondents were more likely to have greater than four years of college education than Republicans or independent voters. Most Democrats in our sample had a very or somewhat favorable view of the ACA (86.1 percent), roughly half of independent voters had a very or somewhat favorable view (44.5 percent), and few Republicans had a very or somewhat favorable view of the law (11.4 percent) (data not shown).

Exhibit 1:

Sample characteristics bv party orientation

	Overall	Democrat	Republican	Independent	P value
Sample sizea (n)	1,223	538	269	296	–b
State (%)
Massachusetts	23.1	27.8	18.4	20.4	0.029
Maine	44.3	42.7	42.5	46.3
New Hampshire	32.6	29.5	39.1	33.2
Sex (%)
Female	51.8	55.0	43.5	51.4	0.041
Male	48.2	45.0	56.5	48.6
Age, years (%)
≤35	28.1	29.7	25.9	24	0.456
36–45	14.8	14.4	12.8	16.2
46–55	23	21.1	28.5	23.2
56–65	34.1	34.8	32.8	36.7
Non-White race (%)
No	94.1	94.2	92.9	95.3	0.596
Yes	5.9	5.8	7.1	4.7
Education (%)
High school graduate or less	15.8	8.3	18.0	21.0	<0.001
Some college	25.8	18.5	30.8	31.6
Four-year college degree	28	33.5	29.0	22.8
More than four-year college degree	30.4	39.7	22.1	24.6
Employment status (%)
Employed	38.5	36.3	37.8	42.2	0.480
Self-employed	36.1	36.2	40.4	32.6
Retired	10	11.5	7.7	10.4
Not working, other	15.4	16.1	14.1	14.9
Family with children <18 (%)
No	85.5	86.8	82.8	86.7	0.397
Yes	14.5	13.2	17.2	13.3
Primary language (%)
Not English	1.6	1.7	1.0	1.8	0.768
English	98.4	98.3	99.0	98.2
Household income (%)
≤250% FPL	40.5	38.1	35.4	43.8	0.202
251%–400% FPL	20.9	22.9	18	20.8
>400% FPL	38.5	39	46.6	35.5
Member in plan with chronic condition (%)
Yes	33.8	36.3	34.1	29.5	0.240
No	66.2	63.7	65.9	70.5
Plan purchased through Marketplace (%)
No	30.2	28.7	37.4	29.3	0.051
Yes	69.8	71.3	62.6	70.7
Previous insurance coverage (%)
Public or uninsured	10.9	8.5	11.8	12.5	0.756
Other	4.0	3.9	3.5	3.9
Nongroup	72.5	73.6	73.1	70.3
Employer sponsored	12.6	14	11.6	13.3

SOURCE Authors’ analysis of 2017 data on enrollees in individual market health insurance plans offered by Harvard Pilgrim Health Care in Massachusetts, Maine, and New Hampshire. NOTES Numbers show proportions of sociodemographic characteristics overall and by party orientation. Results are adjusted for nonresponse. FPL is federal poverty level.

^aParty sample sizes do not sum to full sample as a result of observations with missing party affiliation.

^bNot applicable.

We compared model-adjusted predicted probabilities of marketplace enrollment, which included adjustment for subsidy eligibility amount. Republican respondents had a 55 percent likelihood of enrolling in their insurance plans through the Marketplaces, compared with 67 percent among Democrats (exhibit 2). The association of Marketplace participation with party orientation was moderated by income (interaction term for coefficient of Republicans compared with Democrats and at least 400 percent of poverty relative to 250 percent or less of poverty, p = 0.04) and marginally by whether there was someone with a chronic condition in the plan (interaction term for coefficient of Republicans compared with Democrats and chronic condition in the plan, p = 0.06). No difference was detected in the likelihood of Marketplace enrollment among Democrat and Republican respondents reporting household incomes of 250 percent or less of poverty (87 percent versus 88 percent), whereas Democrat respondents reporting household incomes at least 400 percent of poverty were significantly more likely to enroll through the Marketplaces (54 percent) than Republican respondents with the same income (37 percent). No difference was detected in the likelihood of Marketplace enrollment among Democrat and Republican respondents who had someone with a chronic condition in the plan (66 percent versus 61 percent). In contrast, Democrat respondents without any chronic conditions who were in the plan were significantly more likely to enroll through the Marketplaces than Republicans (67 percent versus 51 percent).

Exhibit 2: Predicted probability of Marketplace enrollment by political orientation in 2017.

Source/Notes: SOURCE Authors’ analysis of 2017 data on enrollees in individual market health insurance plans offered by Harvard Pilgrim Health Care in Massachusetts, Maine, and New Hampshire. NOTES Numbers show adjusted predicted probabilities that Harvard Pilgrim Health Care members with individual market plans were enrolled through the Marketplace. Results are adjusted for Advance Premium Tax Credit subsidy eligibility amount, age, sex, education, employment status, family composition, prior insurance coverage, and state. Eligibility amount is determined by self-reported household income, rating area, and the age and number of members on the plan. FPL is federal poverty level. **p < 0.05 ***p < 0.01

Among subsidy-eligible respondents who enrolled through the Marketplaces, those who were Republican were less likely to take up subsidies than those who were Democrat (exhibit 3). Model-adjusted predicted probabilities estimate that 90 percent of Democrat respondents enrolling through the Marketplace who were eligible for a subsidy received one compared with 77 percent of Republican respondents. No moderation effects by income or chronic condition were detected. There was a similar pattern to Marketplace enrollment by income, with larger gaps across political parties observed among higher-income groups.

Exhibit 3: Predicted probability of receiving an Advance Premium Tax Credit (APTC) subsidy in 2017 among Marketplace enrollees, by political orientation.

Source/Notes: SOURCE Authors’ analysis of 2017 data on enrollees in individual market health insurance plans offered by Harvard Pilgrim Health Care in Massachusetts, Maine, and New Hampshire. NOTES Numbers show adjusted predicted probabilities that Harvard Pilgrim Health Care members who enrolled in individual market plans through a Marketplace received an APTC. Results are adjusted for APTC eligibility amount, age, sex, education, employment status, family composition, prior insurance coverage, and state. Eligibility amount is determined by self-reported household income, rating area, and the age and number of members on the plan. Massachusetts was excluded as a result of unavailable APTC data. FPL is federal poverty level. **p < 0.05

Exhibit 4 shows the average monthly amount of APTC received among respondents with incomes less than 400 percent of poverty by party orientation after adjusting for sociodemographics and differences in APTC eligibility amount. Republicans received an average monthly APTC of $238 compared with APTCs of $304 among Democrats and $289 among independent voters of comparable eligibility and sociodemographic characteristics. This was relative to an average APTC monthly eligibility level of around $380 a month among each of the political groups.

Exhibit 4:

Average monthly APTC subsidy eligibility and receipt in 2017 among APTC-eligible respondents, by political orientation

	Democrat	Republican	Independent
Average monthly APTC eligibility	$383	$380	$378
Average monthly APTC received	$304	$238**	$289

SOURCE Authors’ analysis of 2017 data on enrollees in individual market health insurance plans offered by Harvard Pilgrim Health Care in Massachusetts, New Hampshire, and Maine. NOTES Numbers show average monthly Advance Premium Tax Credit (APTC) subsidy eligibility in 2017 compared with APTC received in 2017 by party orientation. Results are adjusted for APTC eligibility amount, age, sex, education, employment status, family composition, prior insurance coverage, and state. Eligibility amount is determined by self-reported household income, rating area, and the age and number of members on the plan. Massachusetts was excluded as a result of unavailable APTC data. The p value is for a comparison of average monthly APTC received by Republicans versus Democrats.

^**p < 0.05

Discussion

Our results show that political orientation is associated with decision making about health insurance enrollment in the ACA Marketplaces. Overall, Republican subscribers in individual-market plans from a large insurer in New England were less likely to enroll through the ACA Marketplaces than Democrat subscribers with the same sociodemographic profile and APTC eligibility level. Republican subscribers who enrolled through the Marketplaces who appeared eligible for an APTC were less likely than Democrat subscribers to receive an APTC. Neither Democrat nor Republican subscribers fully took up APTCs for which they were eligible, but we found that the overall financial impact of not taking up subsidies was roughly double for Republicans than for Democrats. This result varied by subgroups in the parties, as the party effects on decision making were inversely related to the magnitude of the financial consequence. There was no difference in Marketplace participation or subsidy take-up by party among the lowest-income groups (<250 percent of poverty), which were eligible for both APTCs and cost-sharing reductions worth thousands of dollars. Similarly, no party differences were detected among subscribers with chronic conditions who typically experience higher total health costs.

That we do not observe a party effect in Marketplace participation among low-income subscribers or subscribers with chronic conditions may suggest two things: first, that subscribers put their financial and health interests above partisanship when the stakes are high, and second, that party effects may not be solely attributable to differential gaps among parties in access to information. The lack of party difference between subscribers with chronic conditions may be because subscribers with chronic conditions typically prefer more comprehensive health plans, which can be significantly more affordable with APTCs and cost-sharing reductions on the Marketplaces.²³ Our finding of lower participation in the Marketplaces among Republican-leaning subscribers without chronic conditions also supports recent evidence that politically motivated enrollment has contributed to adverse selection in the Marketplaces, leading to higher rates of premium growth in areas with more Republican voters.²⁴

Several reasons may underlie partisan differences in subsidy uptake. First, some degree of conscious attraction or avoidance of the Marketplaces may have occurred as a result of support of or opposition to a signature accomplishment of the Obama administration and the Democratic party. Examination of the write-in responses on our survey revealed expressions of both support and opposition to the ACA as the reasons for or against enrolling through the Marketplaces. Second, different underlying values or ideology may contribute to decisions about Marketplace and APTC take-up, as Republicans and Democrats differ in their views of social spending²⁵ and whether the federal government should play a major role in improving the health care system of the US.²⁶ Differential uptake of subsidies by party within the Marketplace suggests that other mechanisms in addition to support for the ACA may also play a role. For instance, trust in government is known to decrease among members of a political party when the opposition party holds political control.²⁷ If Republicans felt less comfortable providing income information to the federal government out of concern for privacy or data security, they may have forgone undergoing the necessary income verification to receive subsidies. Third, differences by political party in awareness of the availability of Marketplace subsidies among higher-income subscribers could contribute to differential uptake. Greater promotion of the Marketplace subsidies in Democrat organizations, regions, and media may contribute to increased uptake in this population.

Policy Implications

Navigating the developing political polarization in the United States requires optimizing public policies and the associated education and outreach to ensure maximal efficacy regardless of political orientation. The results of this study suggest that public assistance programs where participation is voluntary and benefits for some participants are modest may be less effective at delivering benefits to qualifying individuals and families when those programs are politicized and associated with one national political party, as has occurred with the ACA “Obamacare” Marketplaces.^6,28 When possible, policy makers might reduce politicized associations by avoiding program labeling that brings to mind contentious political battles at the national level. Such practice occurs in many state-run Marketplaces and state Medicaid programs, which are commonly known by state-specific program names. Researchers have cautioned, however, that although state-centered labeling of Medicaid may have improved enrollment, it may also have contributed to confusion about the provenance of the program and reduced support for “Medicaid” in recent years.²⁹ The ability of policy makers to prevent or avoid partisan labeling that becomes associated with a program may also be limited.

Another means of increasing participation may be through the expansion of enrollment and receipt of federal subsidies through alternatives to a government website. Research found that Republican enrollment in Marketplace plans increased significantly when offered through private exchanges such as third-party brokerage websites compared with Healthcare.gov.⁵ Greater coordination with individual carriers and brokerage networks or greater ability to enroll through private websites, such as through enhanced direct enrollment,³⁰ may expand opportunities for eligible subscribers to enroll in subsidized plans via avenues that feel more comfortable or less politicized. However, obscuring the source of benefits may distort the public’s ability to weigh the benefits and costs of the ACA,³¹ and enrollment through for-profit entities should be monitored to ensure enrollees are not diverted away from optimal plans or government-supported programs for which they may be eligible.

Conscious decisions to participate or not in public programs for politically or ideologically motivated reasons are not inherently flawed, and many valid reasons may exist for not claiming Marketplace subsides. However, it is important that these choices are made understanding the financial trade-off that may be involved and that people of all political persuasions have access to correct information about what is available to them. Politically associated differences in subsidy receipt in the ACA are a concern, then, to the extent that they are driven by a lack of awareness or misunderstanding of eligibility criteria. Outreach efforts to inform the public of the benefits of enrollment may be effective at closing any existing information gaps if placed in regions and media markets where political opposition to the ACA has been high.

Our findings also offer some insight into potential enrollment consequences of recent and upcoming policy changes by the Biden administration, which temporarily expand premium tax credits and roll back several prior administration executive orders. Expanded subsidies will substantially increase the benefits available to most Marketplace enrollees, including to those in households with incomes higher than 400 percent of poverty, where our analysis showed political avoidance of the Marketplaces was strongest. Restoration of restrictions on non-ACA-compliant plans may also reduce the number of low-premium options available off-Marketplace. As the financial cost of purchasing a plan off-Marketplace increases and becomes more widespread, these results suggest that the partisan gap in use of the Marketplaces and receipt of subsidies will narrow.

Conclusion

Under the current state of US politics, it perhaps comes as no surprise that we observe political influences in the health care choices of people and that these choices involve material trade-offs. Preliminary research on the divergent beliefs and social distancing behaviors of Democrats and Republicans during the COVID-19 pandemic and the economic consequences of these divisions underscore this point.^9,32 Yet the continued viability of the Marketplaces, both politically and financially, depends on robust participation by a range of consumers. The political gaps in Marketplace participation observed in this study may have grown even larger since 2018, with the elimination of the individual mandate penalty in 2019, promotion of non-ACA-compliant plans, and near elimination of federal enrollment outreach and assistance funds through 2021. Recent policy reversals by the Biden administration may change these possible trends in Marketplace participation. Future work should continue to monitor enrollment and subsidy receipt by party orientation to understand the political health of the Marketplaces and its potential impacts on enrollment and affordability for consumers.