First Impressions: Geographic Variation in Media Messages During the First Phase of ACA Implementation

Abstract

The fall 2013 implementation of one major component of the Patient Protection and Affordable Care Act (ACA) — the new health insurance marketplace — was challenging to say the least, with frequent national news reports documenting problems. Public exposure to media coverage of ACA implementation has been high, with 55 percent of Americans reporting following news coverage of the ACA closely and 61 percent reporting having seen advertisements about the law in November 2013 (Kaiser Family Foundation 2013). National estimates of the public's exposure to media, however, overlook one critical aspect of ACA implementation — dramatic geographic differences in the political environments and insurance options available and potentially accompanying differences in media portrayals of the marketplace. This is consequential because media portrayals can influence two distinct audiences: the roughly 25 million who are estimated to gain health insurance coverage under the ACA (CBO 2014) and the much larger group of currently insured people who are forming and solidifying impressions of the law in part on the basis of media. These impressions may contribute to the public's perceptions of health insurance coverage options and support for the law (Levey 2013; Soroka, Maioni, and Martin 2013). As such, policy makers and researchers concerned with evaluating the impact of the ACA on enrollment in health insurance — or on other longer-term outcomes, such as support for subsequent reforms — should be attune to the potential role of media.

We sought to measure the variation in messaging during the early phase of implementation as a first step toward assessing whether media messaging might provide a distinct contribution to consumers' perspectives and their health insurance enrollment behaviors. The goal of our research was to describe the volume of advertising and local news media messages and to examine the geographic variation in both volume and tone of messages across the United States during the two-week period following the launch of the health insurance marketplace (October 1–17, 2013). To do so, we analyzed two forms of broadcast media in the 210 local media markets: advertising (including insurance product marketing and political advertising) and local television news coverage. Advertisements have two distinct persuasive goals: to encourage eligible members of the public to purchase new insurance products and to promote competing political visions in support of or opposition to the ACA. Local television news coverage is the news source most widely viewed by the American public (Pew Research Center for the People and the Press 2012), and media messages transmitted via local news often reflect and shape their local political environment (Gentzkow and Shapiro 2010). In addition, political commentators suggest that the Obama administration's strategy for promoting enrollment has targeted local TV news, heightening its importance as a potential influence on the public's impressions of the ACA (Brown and Epstein 2013).

Collecting and Analyzing Media Data

We collected two sources of data through the Wesleyan Media Project: (1) Kantar Media / CMAG data on health insurance–sponsored advertisements and political advertisements mentioning the ACA and (2) local television coverage of the ACA. The advertising data from Kantar Media / CMAG include volume of airings, cost estimates (calculated by Kantar), and video files of each unique advertisement that aired on local television or national cable in each of the 210 US local media markets (not including local cable buys). We classified each advertisement as either a product or a political advertisement. Product advertisements provided marketing information about the newly available marketplace options, and we classified each sponsor as federal, state, or other/private. Political advertisements focused on a politician or a political stance related to the ACA.

For local news coverage, we conducted closed-captioned searches for ACA-related words (i.e., “health care,” “healthcare,” “Obamacare,” “Obama care,” and “Affordable Care Act”) appearing in up to two news broadcasts during the highest-rated half hour of local news in each of the 210 markets. A trained team of coders determined whether the story mentioned or focused on the ACA and whether it pertained primarily to the new insurance products, the politics of the issue, or both. Coders also assessed whether the news story conveyed a negative, positive, balanced, or neutral impression of the marketplace. Coder reliability, as measured by kappa statistics (a measure of interrater reliability that adjusts for chance), was high, with all kappas exceeding 0.78 (Landis and Koch 1977). We also compared the content of coverage between states running their own exchanges and those that are not, using a market classification that considered a local television market as part of adjacent states if the number of residents within a state makes up at least 20 percent of the total market population. We excluded news coverage focused primarily on the US government shutdown that occurred October 1–16, 2013. Given logistical challenges acquiring a very high volume of video segments across 210 markets, we did not successfully capture video for 28 percent of segments identified through closed-captioned searches. As a result, we weighted results for the capture rate in each market, to more accurately reflect the volume of coverage. (See the online appendix for information about weighting and sensitivity analyses of market categorizations.)¹

Volume and Content of Media Messages

During the period October 1–17, 2013, 44,656 airings of 309 unique ads marketing insurance products appeared across the 210 media markets at an estimated cost of $16.2 million, including 34 percent sponsored by state marketplaces or federal agencies and the remainder by nonpublic sponsors, including insurance companies and brokers. Of the fourteen states running their own marketplaces, two did not air advertisements observable in our data set (Massachusetts and Connecticut), nor did the District of Columbia. We observed substantial variation in advertising volume across the media markets (fig. 1). For example, television audiences in Kansas City were exposed to over one thousand airings promoting health insurance options in this two-week period, while audiences in Nashville, with a similar population size, were exposed to only fifty airings.

Seventeen unique political ads mentioning the ACA aired 1,978 times during the same two-week period, at an estimated cost of $1.1 million. Many of these ads aired in jurisdictions where November 2013 elections were held (fig. 1), including New Jersey and Virginia; however, a large number aired in Arkansas, where the 2014 Republican Senate contender Tom Cotton and interest group allies used the rollout as an early opportunity to attack Senator Mark Pryor ahead of the 2014 elections. All but two of the seventeen unique political ads aired were negative in their content about the ACA.

During this time, 1,286 local news stories mentioning the ACA aired. The content of local news coverage varied, with 42 percent focusing solely on the politics of the ACA, including demands that it be delayed or defunded in exchange for reopening government, 46 percent focusing on the new health insurance products, and the remaining 12 percent mentioning some aspects of both. Within the subset of local news stories containing product information (n=744), 50 percent conveyed an optimistic or encouraging impression of the new insurance products, 26 percent conveyed a pessimistic or discouraging impression, 19 percent conveyed both encouraging and discouraging information, and 5 percent were neutral.

Here again, however, we observed substantial geographic variation in both volume and content of local news coverage of the ACA (fig. 2). For example, only four news stories mentioning the ACA aired in Syracuse during this period, compared to fourteen news stories airing in Buffalo (top panel). The balance of encouraging versus discouraging content also was highly variable, even across small geographic areas (bottom panel). For example, local television audiences in Tyler and Lubbock, Texas, saw similar amounts of ACA marketplace–focused coverage, but the impression of that coverage was very different. Three-quarters of Tyler coverage was discouraging, while 60 percent of Lubbock coverage was encouraging. Finally, we found moderate evidence of coverage differences based on the state-level context of ACA marketplace politics. News airing in markets located primarily in states operating federal/partnership marketplaces was significantly more discouraging than news airing in markets located primarily in states running their own marketplaces (30 percent discouraging versus 19 percent, p<0.05).

Implications of Variation in Media Messaging for ACA Implementation

The politics surrounding ACA implementation are dynamic, and media messages will evolve in ways that are difficult to predict. Yet our snapshot of the first two weeks of media coverage following the rollout of the insurance marketplace suggests a few important themes. First, where you live matters greatly in the types of messages to which you are exposed. While geographic variation in health care delivery has been well documented (Welch et al. 2011, our findings related to geographic variation in media messages could enhance regional differences in the number and health status of new insurance enrollees, in the public's political attitudes toward the ACA, or both. Second, the media environment offers both positive and negative visions of the insurance marketplace, with insurers marketing a product against a backdrop of negative political advertising and widely varying tone and content of local news coverage. While political advertisements were a minor component of the media environment in early October 2013, their volume is likely to grow in 2014 heading into the midterm elections (Cunningham 2013). In fact, an estimated $7.4 million was spent on 18,415 political ad airings mentioning health care from October 1, 2013, through the end of 2013, targeting candidates in nine 2014 Senate races and a few House races; an estimated $100 million was spent on 250,195 insurance ad airings during this period.

Moving forward, carefully evaluating the extent to which geographic variation in media exposure shapes individuals' attitudes and choices in the marketplace will be critical. Research efforts to better understand the influence of media on public attitudes and enrollment outcomes are needed, although we acknowledge that such efforts are challenging given the multiplicity of available media sources and difficulty disentangling influences from media and other correlated political factors. Nevertheless, our analysis underscores the importance of understanding the substantial variation in the volume and tone of coverage that Americans actually experience in their local environments. Our findings to date suggest that media messaging offers a source of variation in the local environment that appears to be different from other sources of influence on health insurance enrollment outcomes, such as availability of or costs of plans or the local political climate. Evaluations of ACA implementation by analysts or policy makers that do not account for these media factors may lead to misleading conclusions: a finding of low uptake of insurance in a particular geographic region does not necessarily mean that citizens in that area are not interested or that the insurance plan options are unaffordable; it could be that dominant media messages in that area (whether in news or advertisements) contributed to lower public enthusiasm toward insurance options. Given substantial evidence that media messages about health reform affect public opinion (Cappella and Jamieson 1997; Soroka, Maioni, and Martin 2013), we believe that variation in messages about ACA implementation will be important to understanding differences not only in exchange uptake but also in public perceptions of US health care options, cost, and quality.

This work was supported by Wesleyan University. We are grateful for the dedication and tireless research assistance of Darcie Binder, Oluwaremilekun Ojurongbe, Oladoyin Oladapo, Michael Yee, Marshal Lawler, Michael Linden, Eliza Loomis, Zachary Wulderk, and Olivia Horton, along with the rest of the undergraduate research team at Wesleyan University. We also thank Ross Petchler, Leonid Liu, Mansoor Alam, Manolis Kaparakis, and Wesleyan's Quantitative Analysis Center and are indebted to CommIT Technology Solutions for invaluable support and system infrastructure to complete the analysis.

Notes

References