TV Advertising Volumes Were Associated With Insurance Marketplace Shopping And Enrollment In 2014

In September 2017, the administration of President Donald Trump announced that it was eliminating its broadcast television advertising budget for HealthCare.gov, as well as reducing its overall Affordable Care Act (ACA) enrollment marketing budget by 90 percent.¹ This constituted a reduction from $100 million during the 2017 open enrollment period to $10 million for the 2018 open enrollment period. The announcement alarmed researchers and advocates who have argued that robust marketing is essential to attract potential enrollees to the health insurance Marketplaces, encourage coverage, and facilitate a sustainable balance between healthy and sick enrollees in the Marketplaces.^2,3 However, the published empirical evidence base on the relationship between advertising and enrollment in Marketplace plans is still quite limited, particularly studies that isolate advertising effects from other media, demographic, and social factors that shape health insurance–related information seeking and enrollment.

The evidence indicates that marketing efforts are associated with enrollment in Medicare and Medicaid, although the magnitude of their effects appears to be small. For example, Bradley Shapiro studied the effects of television advertising on Medicare Advantage enrollment by analyzing differences in exposure to advertising across television media market borders and estimated that dropping TV ads entirely would reduce Medicare Advantage demand by 0.23 percentage points.⁴ Another study showed that television advertising of an enrollment assistance hotline was associated with an increase of 2.5–4.7 percent in enrollment in California’s Medicaid program.⁵ A more recent experiment also demonstrated small effects (3 percentage points) of low-cost outreach (direct mailings, not broadcast advertising) on enrollment in Medicaid in Oregon.⁶

However, extrapolating these findings from the long-established Medicare and Medicaid programs to the newer ACA Marketplaces might not be appropriate. In 2014, Marketplaces offered Americans an entirely new way to purchase coverage. Moreover, the targeted audience for Marketplace plans included segments of the population with historically less experience in purchasing health insurance, and the plans had complicated premium and cost-sharing supports.

Some evidence suggests that advertising has played an even more important role in attracting enrollees in this context. One study found that counties exposed to higher volumes of televised insurance advertisements during the Marketplaces’ first open enrollment period realized larger reductions in their uninsurance rates in 2014. State-sponsored advertisements and Medicaid enrollment appeared to drive this relationship.⁷ Another study found that in Kentucky, Gov. Matt Bevin’s (R) abrupt reduction of television advertising for the Kentucky Marketplace, Kynect, was associated with less traffic to the Kynect website, although there were no significant effects on phone calls to the state call center.³ Finally, the California Marketplace, Covered California, found that its marketing efforts delivered a three-to-one return on investment by attracting a healthier mix of enrollees into the Marketplace and reducing premiums.⁸ These core findings are further supported by research that the Department of Health and Human Services conducted during the administration of President Barack Obama attesting to the impact of advertising on Marketplace enrollment.⁹

Notably, the published studies described above^3,7–8 examined aggregate relationships between marketing investments and enrollment-related outcomes at the state or county level. To date, we know of no published evidence about the individual-level effects of ACA advertising on Marketplace shopping or enrollment behaviors.

To address this, we tested whether the volumes of three types of televised content (insurance advertisements, news media, and political advertisements referring to the ACA) were associated with the shopping and enrollment behavior of US adults in the new health insurance Marketplaces in 2014. Multiple types of televised messages about the ACA were aired during the first open enrollment period (October 1, 2013–March 31, 2014).¹⁰ Recent commentary has focused on the cuts to broadcast ads encouraging Marketplace enrollment sponsored by the Department of Health and Human Services.^1–2 However, ads were also aired by several other entities, including state governments; private insurers, brokers, and health systems; and nonprofit enrollment advocates. In fact, during the first open enrollment period, 42 percent of all insurance-related ads aired were from insurance companies and other private sponsors, 33 percent were from the federal government, and 25 percent were from state Marketplaces.¹¹ In addition, local television news frequently covered the launch of the first open enrollment period,¹² and political candidates and organizations gearing up for the 2014 midterm elections purchased political ads with anti-ACA messages.¹³ The current study examined associations between these three types of televised media messages and two individual-level outcomes related to the new health insurance Marketplaces: shopping for and enrollment in Marketplace plans.

Study Data And Methods

Survey Data

Our main data source was the 2014 National Health Interview Survey (NHIS). Volumes of televised messages aired in each county were added to the NHIS data based on respondents’ county of residence. To protect confidentiality, county of residence is not in the publicly available NHIS data. Because county is a restricted variable, the data were merged by the Research Data Center (RDC) of the National Center for Health Statistics (NCHS) and then accessed for analysis through the NCHS RDC at the University of Minnesota Federal Statistical Research Data Center.

The dependent variables were whether respondents reported having shopped for a Marketplace health insurance plan or having enrolled in such a plan. The following question about shopping was asked in 2014 of a randomly selected adult in each NHIS household (known as the “sample adult” subset): “Have you looked into purchasing health insurance coverage through HealthCare.gov or the Health Insurance Marketplace, such as [fill in state Marketplace name]?” We compared the characteristics of those who answered “Yes” to those who answered “No.” This measure is relevant to adults in all states, regardless of whether their state operated its own health insurance Marketplace or relied on the HealthCare.gov website. Respondents who refused to answer, didn’t know, or provided an answer that was “not ascertained” were excluded. Analyses using this dependent variable included 25,840 adults ages 18–64.

Marketplace enrollment was calculated based on the NHIS algorithm for Marketplace plan coverage. All adult respondents who reported being enrolled in a private health insurance plan were asked, “Was this plan obtained through HealthCare.gov or the Health Insurance Marketplace such as [fill in state Marketplace name]?” Any respondent living in any state who reported having a plan that they had obtained through either a federal or state-based Marketplace counted as “Yes” for this variable. Because the question about Marketplace enrollment was asked of all respondents rather than just the sampled adult subset, analyses using this dependent variable included 37,982 adults ages 18–65.

Data On Televised Content

The key independent variables were numbers of insurance ads, political ads, and keyword hits of closed captioning from local evening news broadcasts aired on television in the county in which each NHIS respondent lived.¹⁴ The Wesleyan Media Project compiled data on the three types of content examined in this study.¹⁵ Using data from Kantar Media’s Campaign Media Analysis Group (Kantar/CMAG), the project tracked every insurance-related ad aired on local television or national cable in each of 210 US media markets between October 1, 2013, and April 15, 2014 (to encompass the special enrollment period that followed the official end of the enrollment period on March 31). The data included the date and time of each ad, which we used to calculate the total number of ads that were aired in a given media market during the open enrollment period. We classified the advertisements into four sponsor types: federal (the Department of Health and Human Services or HealthCare.gov), state (state-based Marketplaces), private (including insurance agencies, brokers, and health care systems), and advocates or other sponsors (these ads include public service announcements by television stations or nonprofit groups). Online appendix exhibit A3 reports regression models that display results for all insurance ads aired, aggregated across all sponsors.¹⁶

The Wesleyan Media Project also used Kantar/CMAG data to track all political ads referring to the ACA that were aired during the same period. Since 95 percent of these ads contained messages opposed to the ACA,¹³ we included only the anti-ACA ads, to maintain a consistent valence within this type of televised content.

Finally, we included data on volumes of local broadcast news coverage. We conducted keyword searches of closed captioning from evening news broadcasts on the major networks¹⁷ to generate a measure of the volume of attention to the ACA in each media market across the six-month time period. To merge the data with information from the NHIS, we assigned each county to the volume of televised content that was aired in the media market in which the county was embedded.¹⁸

Analysis

We conducted separate logistic regression models of the likelihood of shopping and enrolling on the types of content that aired in each respondent’s county of residence. All models also included a robust set of control variables. At the individual level, the variables (drawn from NHIS data) included age, sex, race, Hispanic ethnicity, citizenship status, whether the respondent had been unemployed in the past year, educational attainment, income (in groups by percentage of the federal poverty level), self-rated health, and the quarter of 2014 in which the respondent was interviewed. Models predicting shopping also controlled for whether the respondent had employer-sponsored insurance and access to the internet (the latter was available only in the sample adult survey). All models also included a county-level variable for the percentage of voters in that county who voted for President Obama in 2012¹⁹ and the percentage of adults ages 18–64 who were uninsured in 2013, from the Census Bureau’s Small Area Health Insurance Estimates. Finally, all models incorporated state fixed effects to adjust for any other characteristics of the state’s political or health insurance environment. Regressions applied NHIS weights and adjusted the standard errors for the complex sampling design. We present the estimated odds ratios from our models and calculated the predicted probability of shopping or enrolling by the levels of televised content aired in a county, holding all other characteristics constant.

In additional exploratory analyses, we estimated models fitted with interaction terms between the media variables and a set of individual-level characteristics: income, age, and self-rated health (to assess whether sensitivity to ads varied by these characteristics). These models revealed no significant interaction terms (see appendix exhibits A5–A10).¹⁶ We also estimated models using the natural log of volumes of each type of televised content to explore whether there were nonlinear, diminishing returns at high advertising volumes (see appendix exhibit A4).¹⁶ The core results were robust to this specification; for simplicity we report the linear versions.

Limitations

Readers should consider a few limitations when interpreting our study results. First, we report on patterns of association, which limits our ability to draw strong causal inferences. Other (unmeasured) characteristics in counties with higher volumes of televised content could also explain the associations reported here, including nonbroadcast outreach and enrollment activities such as the efforts of in-person navigators. The reverse causal pathway is also plausible: Insurance advertisers might have chosen to place more ads in locations with viewers they thought would be particularly likely to shop for and buy Marketplace insurance plans. Indeed, previous research indicates that insurance ads were targeted at counties with higher uninsurance rates in 2013.⁷ However, by including a robust set of individual-level controls, county uninsurance rates, and state fixed effects, we adjusted for at least some of the competing explanations for the relationships observed.

Second, we measured only airings of ads and other content, not individual respondents’ actual exposure to them—which would be possible only with behavioral measures of media consumption or recall of advertisements.²⁰ Our “ecological” measure of volume, however, overcomes some limitations of exposure measurement that rely on self-reports. We also did not incorporate viewership information into our media measures, yet other research has estimated media effects using a simple advertising count measure as we did.^21–23

Third, our categorization of ads with private sponsors does not distinguish between those intended for the large proportion of the population with employer-sponsored insurance versus those that dealt with Marketplace plans. However, in a separate analysis of advertisement content, we found that 67 percent of the ads with private sponsors in 2013–14 included mentions of the ACA or state or federal Marketplaces.²⁴ This indicated that the majority of commercial insurance ads in this period made reference to the new plans available through the ACA.

Study Results

Descriptive Findings

Among respondents ages 18–64, 14.2 percent reported shopping the Marketplace in 2014, while 4.1 percent reported enrolling in a Marketplace plan (see appendix exhibit A1 for other respondent characteristics).¹⁶ Televised ad volumes were highly variable. Ads from private sponsors were most prevalent nationally, ranging from 74 aired for respondents living in the lowest-volume counties (at the tenth percentile) to more than 6,930 aired for respondents living in the highest-volume counties (at or above the ninetieth percentile) (exhibit 1). Some respondents lived in counties where no federally sponsored ads aired locally (for example, residents of states with state-based Marketplaces), while others lived in high-volume counties (ninetieth percentile or higher) where 5,723 or more federally sponsored ads aired. Political ads also had a skewed distribution, with no ads aired in counties in the tenth and twenty-fifth percentiles and 1,396 or more ads aired in high-volume counties (ninetieth percentile or higher).

Association Between Televised Ads And Marketplace Searching

Respondents living in counties with higher numbers of federal ads had higher odds of searching for a Marketplace insurance plan (exhibit 2). In contrast, respondents living in counties with higher numbers of political ads had lower odds. Based on the exhibit 2 model, we calculated that the probability of shopping for a plan on the Marketplace was 13.4 percent in counties in the tenth percentile of volume of federal ads and 16.2 percent in counties in the ninetieth percentile. The probability was 14.2 percent in counties in the tenth percentile of volume of political ads and 12.9 percent in counties in the ninetieth percentile (exhibit 3). In exhibit 3, the lines for ads from federal and private sponsors are upward sloping, while the line for political ads is downward sloping—particularly for counties at or above the fiftieth percentile.

SOURCE Authors’ analysis of data for 2014 from the National Health Interview Survey merged with data from Kantar Media’s Campaign Media Analysis Group. NOTES All models adjusted for all of the variables included in exhibit 2 as well as state fixed effects. The types of ads are explained in the notes to exhibit 1.

Association Between Televised Ads And Marketplace Enrollment

We saw similar relationships between the number of televised ads aired and enrollment. Respondents in counties with higher numbers of ads from private or federal sponsors had higher odds of enrolling in a Marketplace plan in 2014, while those in counties with higher volumes of political ads had lower odds of enrolling (exhibit 2). The predicted probability of enrolling was 3.7 percent for respondents in counties with median volumes of federal ads and 5.2 percent for respondents in counties in the ninetieth percentile (exhibit 4). Thus, the marginal effect of an increase of 5,645 airings of federal ads in a county—from the median volume of 78 ads to the ninetieth-percentile volume of 5,723 (exhibit 1)—would be an increase of about 1.5 percentage points in enrollment. In contrast, the calculated predicted probability of enrolling in a Marketplace plan was 4.5 percent in counties in the tenth percentile of volume for political ads and 3.3 percent in counties in the ninetieth percentile (exhibit 4).

SOURCE Authors’ analysis of data for 2014 from the National Health Interview Survey merged with data from Kantar Media’s Campaign Media Analysis Group. NOTES All models adjusted for all of the variables included in exhibit 2 as well as state fixed effects. The types of ads are explained in the notes to exhibit 1.

The control variables included in exhibit 2 also provide information on the relationships between other characteristics and Marketplace shopping and enrollment. Compared to their reference groups, men, people ages 18–25, Asians, noncitizens, people who had been unemployed in the past year, those with less than high school or only high school education, and those with employer-sponsored insurance had significantly lower odds of shopping for a Marketplace plan. In contrast, people with incomes of less than 100 percent or 100–400 percent of poverty and those in good health had significantly higher odds.

There were also systematic differences among groups in the likelihood of enrolling in a Marketplace plan in 2014. Compared to their reference groups, people ages 18–25 or 26–44 had significantly lower odds of enrolling. In contrast, Asians, people who had been unemployed in the past year, those with incomes of less than 100 percent or 100–400 percent of poverty, and those in fair or poor health had higher odds of enrolling. There was no relationship between the percentage of voters in a respondent’s county who voted for President Obama in 2012 and the odds of shopping for or enrolling in a Marketplace plan.

Discussion

This study demonstrates that the number of televised ads for health insurance, particularly those sponsored by the federal government, that aired in a respondent’s county was associated with the likelihood of shopping for insurance on the Marketplace and enrolling in Marketplace coverage, for adults ages 18–64 in 2014. These findings add to existing evidence that insurance Marketplace advertising is associated with information-seeking³ and insurance gains.⁷ We also found that the volume of anti-ACA political ads was associated with less shopping and enrolling, even after we adjusted for the political climate in respondents’ counties of residence. In additional exploratory analyses, we did not find evidence that marketing had a differential effect on people according to their age, income level, or health status. Finally, while previous research suggests that residents of counties exposed to more news about the ACA at the launch of the open enrollment period were more likely to indicate that they had sufficient ACA information,²⁵ we found no evidence of an association between the volume of local TV news coverage of the ACA and enrollment behavior. This could be because news media tended to devote attention to political issues and conflict, instead of providing policy-relevant information aimed to help consumers enroll.¹²

Policy Implications

The key implication of this study is that advertising efforts are likely an important component of health insurance enrollment campaigns.⁸ Specifically, the volume of televised advertisements for HealthCare.gov was associated with shopping for and enrolling in Marketplace plans in 2014. Funding for this advertising was cut for the 2018 open enrollment period. Our findings also show that advertisements sponsored by insurance companies and brokers were associated with enrollment, although the association was not as strong as that for federal advertising.

Our results also suggest that political ads attacking the ACA may have suppressed Marketplace information seeking in areas where they were aired in high volumes in 2013–14. This finding supports the results in other research on the negative consequences of extreme political polarization around the ACA. Benjamin Sommers and colleagues found that low-income adults living in Arkansas, Kentucky, and Texas who reported more exposure to negative ads about the ACA were more likely to say that the ACA had hurt them.²⁶ Amy Lerman and colleagues found that Republicans were more likely to forgo Marketplace coverage and more likely to purchase plans not on the Marketplace.²⁷ Similarly, Michael Tesler found that Democrats were more likely to gain health insurance during the first phase of the ACA.²⁸ Our study indirectly suggests one mechanism that could explain why this is so: proximity to negative messaging about the law.

It is tempting to extrapolate from this study to the 2018 enrollment period, when 11.8 million Americans enrolled in HealthCare.gov plans despite the federal government’s cutting all television advertising and reducing other outreach.²⁹ However, there are strong reasons to believe that the time frame of this study is not comparable to the 2017–18 environment. Our findings came from the period when the ACA was new, and thus advertising from all sponsors may have played a particularly important informational role. The messaging was likely different as well: We found that two-thirds of airings of insurance company–sponsored ads mentioned the ACA during the first open enrollment period, but that was so in only 22 percent and 25 percent of the airings in the second and third open enrollment periods, respectively.²⁴ In addition, 2017 news media attention to ACA repeal efforts and the threats of lost coverage may have contributed powerful reminders to enroll, alongside persuasive messages from advocacy organizations and state Marketplaces.

More research is needed to disentangle the influence of advertising and marketing on ACA enrollment beyond 2014. Such research could use experimental designs (see, for example, Lerman and colleagues)²⁷ to enhance causal inference, as well as leverage possible quasi-experiments to assess advertising influence, such as evaluating differences in impact for media markets whose borders span multiple states⁴ or assessing differences following policy changes that affect advertising.³ New insights may also be gained from closer partnerships between researchers and the policy makers responsible for ad buy decisions—for example, by randomizing the markets that receive different types of outreach. Given the political uncertainty since 2017 and the repeal of the individual mandate, consumers remain confused about whether the law still stands.³⁰ While much has changed since 2014, there is a continuing need for nonpartisan campaigns to inform the public about their health insurance options.

ACKNOWLEDGMENTS

The authors received funding from the Robert Wood Johnson Foundation State Health Access Reform Evaluation (Grant No. 72179, co–principal investigators Sarah Gollust and Erika Franklin Fowler), Wesleyan University (Laura Baum and Fowler), and the McKnight Land-Grant Professorship, University of Minnesota (Gollust). The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the Research Data Center, the National Center for Health Statistics, or the Centers for Disease Control and Prevention.

NOTES

• 1 Kliff S. Trump is slashing Obamacare’s advertising budget by 90%. Vox [serial on the Internet]. [Updated 2017 Aug 31; cited 2018 Apr 12]. Available from: https://www.vox.com/2017/8/31/16236280/trump-obamacare-outreach-ads Google Scholar
• 2 Lodes L. I ran ACA outreach under Obama. Trump’s funding cuts could ruin the health care law. Vox [serial on the Internet]. [Updated 2017 Sep 15; cited 2018 Apr 12]. Available from: https://www.vox.com/the-big-idea/2017/9/12/16294784/aca-outreach-advertising-sabotage-funding Google Scholar
• 3 Shafer P, Fowler EF, Baum L, Gollust S. Advertising cutbacks reduce Marketplace information-seeking behavior: lessons from Kentucky for 2018. Incidental Economist [blog on the Internet]. 2017 Sep 2 [cited 2018 Apr 12]. Available from: https://theincidentaleconomist.com/wordpress/advertising-cutbacks-reduce-marketplace-information-seeking-behavior-lessons-from-kentucky-for-2018/ Google Scholar
• 4 Shapiro BT (University of Chicago Booth School of Business, Chicago, IL). Advertising in health insurance markets [Internet]. Rochester (NY): SSRN; 2017 Jul 1 [cited 2018 Apr 12]. Available for download from: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2789904 Google Scholar
• 5 Aizer A. Public health insurance, program take-up, and child health. Rev Econ Stat. 2007;89(3):400–15. Crossref, Google Scholar
• 6 Wright BJ, Garcia-Alexander G, Weller MA, Baicker K. Low-cost behavioral nudges increase Medicaid take-up among eligible residents of Oregon. Health Aff (Millwood). 2017;36(5):838–45. Go to the article, Google Scholar
• 7 Karaca-Mandic P, Wilcock A, Baum L, Barry CL, Fowler EF, Niederdeppe Jet al. The volume of TV advertisements during the ACA’s first enrollment period was associated with increased insurance coverage. Health Aff (Millwood). 2017;36(4):747–54. Go to the article, Google Scholar
• 8 Lee PV, Pegany V, Scullary J, Stevens C. Marketing matters: lessons from California to promote stability and lower costs in national and state individual insurance markets [Internet]. Sacramento (CA): Covered California; 2017 Sep [cited 2018 Apr 12]. Available from: http://hbex.coveredca.com/data-research/library/CoveredCA_Marketing_Matters_9–17.pdf Google Scholar
• 9 Cohn J. Trump administration says Obamacare ads don’t work, but federal study says they do. Huffington Post [serial on the Internet]. 2017 Sep 20 [Updated 2017 Sep 27; cited 2018 May 10]. Available from: https://www.huffingtonpost.com/entry/obamacare-ads-trump-health-human-services_us_59c1dc7de4b0186c2206bdd3 Google Scholar
• 10 Gollust SE, Barry CL, Niederdeppe J, Baum L, Fowler EF. First impressions: geographic variation in media messages during the first phase of ACA implementation. J Health Polit Policy Law. 2014;39(6):1253–62. Crossref, Medline, Google Scholar
• 11 Barry CL, Bandara SN, Arnold KT, Pintor JK, Baum LM, Niederdeppe Jet al. Assessing the content of television health insurance advertising during three open enrollment periods of the ACA. J Health Polit Policy Law. [Forthcoming 2018]. Google Scholar
• 12 Gollust SE, Baum LM, Niederdeppe J, Barry CL, Fowler EF. Local television news coverage of the Affordable Care Act: emphasizing politics over consumer information. Am J Public Health. 2017;107(5):687–93. Crossref, Medline, Google Scholar
• 13 Fowler EF, Ridout TN. Political advertising in 2014: the year of the outside group. Forum. 2014;12(4):663–84. Google Scholar
• 14 Our analysis included all local broadcast, national network, and national cable television advertising. We lacked information on cable television advertising (sold by cable networks to local advertisers to reach much smaller geographic segments of audiences), as there were no systematic data available on that type of advertising. Local cable advertising is limited not only in audience size but also in quantity because of how it is purchased. For these reasons, we do not believe that the lack of local cable advertising in our data is a problem for our analysis.
• 15 While we refer to three types of “televised content” (insurance ads, political ads, and news closed-captioning keyword hits), in this study we did not actually conduct content analyses of these media types; we only measured their volumes. In other work, we have conducted content analyses of ads and of news to examine the specific messages conveyed to the public. See notes 11 and 12 for examples of content analyses.
• 16 To access the appendix, click on the Details tab of the article online.
• 17 Searches were conducted using the following terms: health care, healthcare, Obama care, Obamacare, and Affordable Care Act. For more details, see Gollust SE, et al. Local television news coverage of the Affordable Care Act (note 12).
• 18 Our media market calculations for this study came from resources in the public domain, not the proprietary Nielsen database. Our media market geographies were made up of unique collections of counties. Counties were not split across multiple media markets.
• 19 Leip D. Dave Leip’s atlas of U.S. presidential elections [Internet]. [Cited 2018 Apr 12]. Available from: http://uselectionatlas.org/ Google Scholar
• 20 Niederdeppe J. Conceptual, empirical, and practical issues in developing valid measures of public communication campaign exposure. Commun Methods Meas. 2014;8(2):138–61. Crossref, Google Scholar
• 21 Sides J, Karch A. Messages that mobilize? Issue publics and the content of campaign advertising. J Polit. 2008;70(2):466–76. Crossref, Google Scholar
• 22 Niederdeppe J, Avery R, Miller EN. Alcohol-control public service announcements (PSAs) and drunk-driving fatal accidents in the United States, 1996–2010. Prev Med. 2017;99:320–5. Crossref, Medline, Google Scholar
• 23 Ridout TN, Fowler EF, Franz MM, Goldstein K. The long-term and geographically constrained effects of campaign advertising on political polarization and sorting. American Politics Research. 2018;46(1):3–25. Crossref, Google Scholar
• 24 These are results of a subanalysis of data on the content of advertising that was conducted as an extension of another study. See Barry CL, et al. Assessing the content of television health insurance advertising during three open enrollment periods of the ACA (note 11).
• 25 Fowler EF, Baum LM, Barry CL, Niederdeppe J, Gollust SE. Media messages and perceptions of the Affordable Care Act during the early phase of implementation. J Health Polit Policy Law. 2017;42(1):167–95. Crossref, Medline, Google Scholar
• 26 Sommers BD, Maylone B, Nguyen KH, Blendon RJ, Epstein AM. The impact of state policies on ACA applications and enrollment among low-income adults in Arkansas, Kentucky, and Texas. Health Aff (Millwood). 2015;34(6):1010–18. Go to the article, Google Scholar
• 27 Lerman AE, Sadin ML, Trachtman S. Policy uptake as political behavior: evidence from the Affordable Care Act. Am Polit Sci Rev. 2017;111(4):755–70. Crossref, Google Scholar
• 28 Tesler M. More Democrats are getting health insurance, thanks to Obamacare. Why are Republicans staying away? Washington Post. 2015 Jul 16. Google Scholar
• 29 Keith K. Marketplace enrollment remained stable, increased in state-based Marketplaces, NASHP reports. Health Affairs Blog [blog on the Internet]. 2018 Feb 8 [cited 2018 Apr 12]. Available from: https://www.healthaffairs.org/do/10.1377/hblog20180208.778942/full/ Google Scholar
• 30 Kirzinger A, Wu B, Brodie M. Kaiser Health Tracking poll—January 2018: the public’s priorities and next steps for the Affordable Care Act [Internet]. San Francisco (CA): Henry J. Kaiser Family Foundation; 2018 Jan 26 [cited 2018 Apr 12]. Available from: https://www.kff.org/health-reform/poll-finding/kaiser-health-tracking-poll-january-2018-publics-priorities-next-steps-affordable-care-act/ Google Scholar