Invoking Identity? Partisan Polarization in Discussions of Race, Racism, and Gender in 2022 Midterm Advertising in the United States

Abstract

Context: Media messaging matters for public opinion and policy, and analyzing patterns of campaign strategy can provide important windows into policy priorities.

Methods: The authors used content analysis supplemented with keyword-based text analysis to assess the volume, proportion, and distribution of media attention to race-related issues in comparison to gender-related issues during the general election period of the 2022 midterm campaigns for federal office in the United States.

Findings: Race-related mentions in campaign advertising were overwhelmingly focused on crime and law and order, with very little attention to racism, racial injustice, and the structural barriers that lead to widespread inequities. In stark contrast to mentions of gender, racial appeals were less identity focused and were competitively contested between the parties in their messaging, but they were much more likely to be led by Republicans.

Conclusions: The results suggest that discussions of race and gender were highly polarized, with consequences for public understanding of and belief in disparities and policies important to population health.

A large body of work has documented growing polarization in American society (Hetherington 2009), including how policy issues get covered in media (especially in partisan media; Levendusky 2013) and in corresponding attitudes among the public about social and political issues (Levendusky and Malhotra 2016). Health care policy has not been spared; in fact, polarization is a defining feature of public opinion about health care reform, with a stark gulf between Democrats and Republicans in favorability toward the Affordable Care Act (ACA) that has persisted more than a decade (Brodie et al. 2020). Competing messaging about the ACA by political parties and their allies has also featured centrally in political advertising (Gollust, Fowler, and Niederdeppe 2020), so much so that the 2018 midterm election was dubbed “the health care election” (WMP 2018). Republicans who had previously relied on nearly a decade of campaign promises to “repeal and replace Obamacare” were on defense for the first time after pushing a highly unpopular and controversial ACA repeal plan in 2017 (Oberlander 2020). At the same time, Democrats, after nearly a decade of not talking about the ACA, were finally emboldened by growing public support for the law to attack Republicans for repeated attempts to repeal the law and their failure to protect those with preexisting conditions (Gollust, Fowler, and Niederdeppe 2020). Public opinion data from 2019 suggested bipartisan agreement that high health insurance costs were among the most urgent issues for elected officials to address (Blendon, Benson, and McMurtry 2019), contributing to expectations that health and health care policy would continue to be a large focus in the 2020 election cycle.

Ironically, the onset of a global (health) pandemic in early 2020 did not result in a long-term amplification of health and health policy discussions in campaign rhetoric. Shifts in electoral strategy were evident as early as the summer of 2020 as President Trump's campaign sought to turn attention away from his administration's handling of the pandemic response. While the Biden campaign outlined how he might have handled the pandemic, the economic fallout, and the needed investment in jobs, the Trump campaign sought to change the subject, focusing almost exclusively on crime, protests/riots, and to a lesser extent, immigration. This focus was notable given the public attention and outrage over George Floyd's murder and the widespread protests that followed. During a period when the Biden campaign devoted more than a quarter of its television ads to discussion of racial injustice and discrimination, the Trump campaign was talking about race but in a way to potentially stoke racial animus and backlash (WMP 2020). Although Trump's approval numbers did not improve and the campaign changed course to focus on more traditional Republican economic issues during the fall 2020 campaign, references to crime and safety remained prominent and continued into the 2022 midterm cycle, with other Republicans picking up the torch (Franz, Ridout, and Fowler 2023).

Health, health policy, and health care remained on the campaign agenda in 2022, at least for Democrats. The Supreme Court's decision in Dobbs v. Jackson Women's Health Organization, which overturned Roe v. Wade, led to abortion and women's rights becoming a central focus of Democratic messaging, while Republican discussion of health care dropped off the agenda (Franz, Ridout, and Fowler 2023). Despite the fact that “the simultaneous calamities of the COVID-19 pandemic and the murder of George Floyd underscored the urgency and interconnectedness of racism and health” (Michener and LeBrón 2022: 111), issues of race and racism have not received as much attention in political ads as gender, particularly women's issues. Mass public attention and sentiment moved strongly in favor of racial justice movements in the immediate aftermath of Floyd's killing and early racial justice protests (Horowitz, Hurst, and Braga 2020; Long and McCarthy 2020), likely in part the result of increased media attention and public awareness (Dunivin et al. 2022); yet public support for the racial justice movement has declined since then (Horowitz, Hurst, and Braga 2020). Thus, it is important to draw attention to and understand how themes related to race, racism, and racial equity have played out in election advertising, as campaign messages provide important cues to the public about timely social issues (Fowler, Franz, and Ridout 2021).

Our inquiry proceeds as follows. We begin by reviewing the literature on the importance of campaign appeals for shaping subsequent political processes and how the public perceives and understands inequality more broadly, and why racialized and gendered appeals matter for attitudes and policies relevant to health. We proceed to outline our approach to assessing the content of campaign ads. We then analyze the use of racial appeals in campaign ads, and we provide additional context by undertaking a similar analysis with regard to gender.

Drawing on election advertising aired on television in federal races during the general election period of the 2022 midterms, we demonstrate that there were clearly polarized patterns in the discussion of identity, but we note a key distinction in how this emerged with regard to race relative to gender. Specifically, we show that while Democrats dominated the conversations about women and health, which were also much higher in volume, discussions of topics related to race were much more competitive between the parties, and to the extent that one party was leading the conversation, it was Republicans.

We conclude by discussing our findings and the limitations of our analysis, and we examine the potential implications of these patterns. Given the lack of attention to connections between health and racism and because conversations that did address or implicitly raise race in advertising mostly invoked contentious and stereotypical issues (i.e., crime, public safety, immigration), our results suggest that not only does campaign advertising not equip the public to understand racism as a health policy issue, it may actually provide the public with fodder for opposition to such arguments.

Why Advertising Patterns Matter for Election Strategy, Public Opinion, and Policy

Messages in paid advertising have consequences for election strategy, public opinion, and policy (Fowler, Franz, and Ridout 2021). Ad content is indicative of what issues and tactics the candidates, parties, and outside groups think will most help their side with voters, and ads are also important for shaping the postelection narrative parties and activists will craft about reasons for success or failure that affect future campaign strategy. Advertising is also a key tool for communication between aspiring political office holders and their constituents, one that conveys a policy agenda that can influence both news media coverage (Roberts and McCombs 1994; Ridout and Mellen 2007) and the issues the public views as most important for policy makers to address (Bowers 1973; Druckman 2004; Hayes 2008; Ridout and Mellen 2007; Sulkin and Swigger 2008).

Campaign ads are also important to understand as a signal and predictor of policy behavior for politicians once in office (Sulkin and Swigger 2008). Advertising appeals are important not just for their short-term goal of helping a party win office but also because they signal which options are on the table for solving identified problems. In addition to providing proposed solutions, these messages signal who is responsible for societal problems and who might benefit from policy changes, and they can therefore shape how Americans view politicians and the policies associated with them as well as who is helped or hurt by these policies.

A long empirical literature on campaign advertising on television demonstrates—in contrast with many public perceptions of the medium—that these messages contain substantive policy information, and that policy information is more likely to be present in contrast and attack advertising (Geer 2008; Goldstein and Freedman 2002; Mattes and Redlawsk 2014; Fowler, Franz, and Ridout 2021). Earlier work has also demonstrated that messages about health care policy have been prominent in advertising in recent years (Fowler et al. 2019; Gollust, Fowler, and Niederdeppe 2020) and in fact were the dominant topic in 2018, primarily the result of Democrats’ focus on the issue, although Republicans also talked a lot about it (Fowler, Franz, and Ridout 2020; Gollust, Fowler, and Niederdeppe 2020). Earlier work has also demonstrated that population health–relevant topics (such as the broader social determinants of health) that expand beyond traditional public health topics (such as health care or infectious disease) are also prominent in campaign advertising; however, political messages rarely explicitly connect topics such as jobs and the economy, education or power, or crime and public safety to health specifically (Fowler et al. 2019). Yet these ads and the identity appeals in them have implications for health policy, as we describe below.

How Racial and Gendered Discussions in Ads Matter for Health

Attitudes around race and racism have long been central in US politics, acting as a lightning rod around which many have organized their political beliefs (Hutchings and Valentino 2004; Kinder and Sears 1981; Mendelberg 1997). Gender politics, particularly abortion (Adams 1997), also functions as a critical source of division. Even though health is not always a focus of (or even explicitly mentioned in) campaign ads that feature broad social issues (see Fowler et al. 2019), discussion of social-determinants issues such as employment, education, and criminal justice have the power to influence the public's attitude toward policies that can ameliorate—or expand—health inequities (Niederdeppe et al. 2023). The racial and gendered appeals that are interwoven into discussions of these broad social issues can further affect support for or opposition to policies that have clear and large implications for population health.

Importantly, the advertising appeals do not even need to be explicitly about race or gender appeals, or health, to have an influence on support for health-relevant policies. For example, a paradigm has emerged from much of the racial and ethnic politics literature that race is often subtly or carefully brought into the broader conversation as a way to activate aversive attitudes among white Americans (Stephens-Dougan 2021), which can bolster support for or opposition to policies with clear implications for health (e.g., criminal justice). Despite this consensus, more recent campaigns have given much reason to question whether subtlety is required for racialized messages to activate racial resentment and make it politically consequential. As a candidate and as president, Donald Trump engaged in explicit racial appeals that were not roundly rejected by the GOP, as existing theories might have predicted (Mendelberg 2001). Instead, he was able to successfully galvanize resentment toward several different racial and racialized groups on his way to the presidency (Lajevardi and Oskooii 2018; Newman et al. 2021). The success of this strategy provides ample reason to believe that we might find a deviation from the pattern of implicit racial appeals, resulting in more explicit appeals in the 2022 midterms.

In the wake of the protests around the murder of George Floyd, Democrats in the United States expressed significantly higher levels of support for reducing racism via policy (Reny and Newman 2021). However, for some among the public, particularly Republicans, these shifts were temporary, and they faded by the election in fall of 2020; for Democrats these shifts were substantively large and lasted through the fall (Reny and Newman 2021). The unprecedented mass mobilization around the issue of racial bias in policing might have led us to expect Democrats to focus heavily on this issue in campaign ads into 2022, since the issue is a priority for many within their broader coalition. By the same token we might have expected Republicans to focus on race in a very different light, activating the resentment that Trump tapped into in his campaigns and term as president.

To further contextualize shifts in identity-related discussions, we also look at gender, another identity that has been key to campaign messaging strategies. As was the case with racial identity, current events in the lead-up to the election offer some clear expectations about how this topic might show up in ads. The Supreme Court's decision in Dobbs to overturn Roe v. Wade in the summer of 2022 vaulted issues of women's rights and women's health care to the top of many Americans’ minds (Schaeffer and Van Green 2022). Abortion access has long been a topic that divided the parties, and the seismic shifts in policy that accompanied the Dobbs decision made the issue very relevant to many candidates running for office. The salience of this highly polarized and gendered issue likely exacerbated already growing gender gaps in party identification, with women becoming much more likely to identify as Democrats and men conversely being more likely to identify as Republicans (Gillion, Ladd, and Meredith 2020), which can further affect policy proposals and attitudes.

As outlined in the previous section, policy and identity discussion (or even implicit identity references) in campaigns can substantially impact the lived experiences of everyday citizens. To provide one concrete example, many of the implicit racialized campaign appeals studied in the 1990s and 2000s focused on heavily racialized policy areas such as criminal justice and welfare policy. The direction of policy in those areas during that time was very much shaped by public opinion and mood (Enns 2016; Gilens 2009). And there is very clear evidence that appeals in campaigns activated racial animus, which increased support for tough-on-crime policies and reductions in the broader social safety net (Gilens 1996; Peffley and Hurwitz 2002). It is difficult to overstate the degree to which these shifts affected lives in marginalized communities (Alexander 2020). Especially in the domain of health, there is considerable evidence that policy in both of these areas—the criminal justice system and the social safety net—contributed to growing racial gaps in health outcomes through this time period (Cottrell et al. 2019). In the same way, many gender-related policy areas have similar impacts on the lived experience of everyday Americans and on the health of the population as a whole. Abortion policy has particularly clear implications for maternal mortality, childhood poverty rates, and many other key areas (Kozhimannil, Hassan, and Hardeman 2022).

Campaign Communication and Processing of Information about Inequality

Campaign communication also contributes to how the audience processes information about policy, identity (both race and gender), and inequality. Although campaign ads on television typically omit partisan labels because they seek to persuade voters who are not already committed to a candidate (Fowler, Franz, and Ridout 2021; Motta and Fowler 2016), the partisanship of many high-level politicians is more well-known, and the inclusion of cues tying politicians to party leaders in Congress or the White House helps to reinforce partisan perceptions. When partisan cues are clearly available, they can enhance a polarized mode of processing information about social issues (Leeper and Slothuus 2014; Slothuus and Vreese 2010).

Polarized information processing matters because it can contribute to how the public thinks about new policy proposals, including those related to policies that might enhance or reduce equity, including racial equity. For example, the public has politically polarized perspectives around the existence of racial disparities in COVID-19 mortality, and cues in the media (including ads) likely enhance and contribute to these differences (Gollust et al. 2022). The combination of racialization and polarization of health policy issues is not new; many researchers have demonstrated that the ACA became a racialized symbol because of strategic political efforts to remind citizens of President Obama's central role in getting the law passed (Henderson and Hillygus 2011; Knowles, Lowery, and Schaumberg 2010; Tesler 2012). Similarly, Hillary Clinton's close association with national health care reform in the 1992 presidential campaign reinforced the gendered implication in the issue discourse (Winter 2006). These cues, among other factors, contributed to the deep and persistent partisan gap surrounding the ACA that exists today (Brodie et al. 2020).

In short, there is ample evidence to suggest that media messaging contributes to polarization in attitudes and information processing and that these effects are reinforcing and relevant not just for campaign strategy but also for public opinion and policy, with consequences for health and racial equity. As recent critiques of the literature in political communication and health policy make clear, more explicit attention to and centering of race is needed (Coles and Lane 2023). In this article, we explicitly focus on the issue policy and identity conversations relevant to race and gender that occurred in 2022 midterm advertising on television surrounding federal races, using gender as a comparison and contrast to discussions of race.

Data and Methods

To analyze patterns in party discussions on race-related issues (in comparison to gender), we relied on election advertising data from the Wesleyan Media Project (WMP). Although WMP tracks both television and digital advertising that airs on online platforms, earlier work has shown that television advertising is the more substantive of the two media (Fowler et al. 2021), so we focused our analysis on television specifically. Through a contract with Vivvix (formerly Kantar) CMAG, WMP received tracking information on all election advertising airing on broadcast television and national cable in all 210 media markets in the country. Vivvix CMAG provided a video of each unique advertisement and a frequency database containing information about where, when, on what station, and on which program each ad aired, along with an approximate cost for each airing based on audience factors such as time of day, market, and program.

WMP trained undergraduate research assistants to classify the content of each unique ad according to an extensive battery of questions, including which substantive policy issues were discussed. The issue battery was intended to capture broad policy areas (not necessarily specific mentions of words or phrases) and to allow comparisons within and across campaign cycles over time. After initial training, student research assistants received multiple rounds of practice before being cleared to code, and all of WMP's variables were double coded on a sample of roughly 20% to assess intercoder reliability.

To analyze issue content that relates specifically to race and to gender, we first examined the topics within WMP's standard issue battery to identify those that would be most closely associated with race and gender specifically. This means that either the category itself is closely connected to racial or gender identity, or that the policy area is (based on earlier literature) typically discussed in a way that conveys clear connections to race or gender identities. This identification is imperfect, as policies can be discussed in a variety of ways depending on context, and we chose to exclude some categories that could be related if we were less certain about how they are currently discussed. For example, discussion of income inequality, poverty, and minimum wage are likely to disproportionately affect nonwhite populations, but because economic hardship was more widespread in the aftermath of economic recovery from pandemic shutdowns, we chose to exclude these issues because they would require further investigation to determine whether they were racialized. These discussions are low volume, so even if they had been included, they would not meaningfully change our results.

Table 1 contains all the policy areas from the WMP issue battery that literature and current context suggest would be most likely to be racialized or gendered in the 2022 midterms. We included two mentions of specific phrases (critical race theory and Planned Parenthood) that are captured in the mention battery. Several of the issue categories that we believed to be relevant did not meet the traditional 0.7 alpha threshold for intercoder reliability. Most of these did not meet this threshold because they are low volume, which makes reliability much harder to achieve because they appear infrequently in our intercoder reliability sample. We include these categories anyway because of their substantive importance, but we urge caution in interpretation of their results.

We rely on the WMP's issue categorization as outlined in table 1 to understand the broad contours of how the parties discuss issues relevant to race and gender.¹ However, to capture additional nuance, such as how ads might refer to race or gender in other issues not solely or traditionally about race or gender, we also perform textual keyword searching for racial and gendered identity- and policy-based words. To do this, we first process all of the unique television ad videos using automatic speech recognition (ASR) of Google's speech-to-text API (using their video model) to get the audio information into a text format, which we use as our basis for text-based searching.

We then use two methods to arrive at a set of race and gender textual keywords to identify identity-relevant content in ads. In the first method, we draw on the literature for guidance on what language indicates that content is related to race or gender. The gender literature has clearly defined dictionaries based on how women politicians talk about identity compared to men (see Pearson and Dancey 2011) that have been widely used in studies on women and politics (e.g., see Dietrich, Hayes, and O'Brien 2019). We are unaware of a comparable dictionary for race-related terms,² perhaps in part because scholars acknowledge contested conceptualizations and measurement of race (Coles and Lane 2023; Dunivin et al. 2022).

Because we lacked an existing dictionary for race, we started with the “women” dictionary for gender (Pearson and Dancey 2011) containing the list of terms for women as targets (e.g., “women,” “woman,” “woman's,” “girl,” etc.) and then supplemented this list by adding terms such as “mothers” and “daughters” that were likely to be present in campaign advertising. To capture the broader array of policy discussion, we added words most likely to appear in gender-owned issues (Herrnson, Lay, and Stokes 2003; Shogan 2001; Windsor et al. 2022), which included abortion, marriage, lesbian and gay issues, social welfare, violence against women and children, elderly/senior issues, education/schools, health care related to women, and gun control/violence. We then created our own list of comparable terms for racial identity targets (e.g., “Black,” “Asian,” “Latin,” etc.) and then used the literature to guide us on which policy issues have historically been racialized to add additional terms for those issues, which included crime, immigration, drugs, and welfare.

We also supplemented both the gender- and race-related textual keywords through an empirically based approach to find important keywords that co-occur with specific issues. To do this, we combined the WMP issue battery with the transcriptions (obtained through ASR) of the ads and then used two methods to obtain the words most associated with a given issue: (1) the informed Dirichlet method described in Monroe, Colaresi, and Quinn (2017), and (2) a chi-square-based test (see Dunning 1993; Yang and Pedersen 1997). This yields, for each issue, a list of words that distinguish the issue from other issues. For example, for abortion, the top words would be abortion, rape, incest, health, ban, decisions, and exceptions; for crime, the top words were crime, criminals, police, violent, safe, dangerous, defund, bail, streets, enforcement, radical, cash, and prison. Since there was overlap between these keywords and the literature-based ones, we primarily relied on the literature-based ones and supplemented them with empirically based ones as needed. See appendix A for specific words we used for each issue.

For this analysis we include advertising from all federal sponsors (candidates, parties, coordinated ads sponsored jointly by candidates and parties, and outside groups). Although outside groups airing ads in support of Republicans or Democrats are not formally part of the Republican or Democratic Parties, many of the most prominent outside sponsors are run by former staff of party politicians; and because outside group advertising is concentrated in some of the most competitive races, we argue that the content of group messaging is just as important to understand because it, too, is intended to bolster one partisan side (often in a way that attacks). We do, however, exclude advertising that supports third-party candidates in all of our analysis (including total counts of airings), focusing instead on the pro-Republican and pro-Democratic content only.

We limit results to only include advertising airings from the general election period: from September 5, 2022 (the day after Labor Day, which is the traditional start of the fall campaign), to November 8, 2022 (election day). We do this to avoid making claims about how the parties are discussing race-related issues that might conflate primary-based discussion within the party with how the parties compete against each other in the heat of the general election cycle.

We calculate the volume (total count) of federal ad airings related to race or gender (excluding advertising in support of third-party candidates) and the proportion that the race- or gender-related airings represent of the total volume of two-party federal advertising (by dividing the volume of race- or gender-related airings by the total number of airings supporting Democrats and Republicans overall). We also calculate the relative attention that the advertising supporting each partisan side devoted to race- or gender-related issues by dividing the volume of airings sponsored by or on behalf of each party by the total volume of airings supporting each side. We repeat this process first for WMP issue categories and then for our keyword approach.

We focus on the distribution of attention by each partisan side as a whole, so the bulk of our analysis assesses aggregate patterns between the parties, although we do also look at some patterns of geographic variation by state. Even though television advertising airs at the media market level, our primary interest in this analysis is in the strategic calculations of the advertising sponsors themselves. Since advertising is targeted at bolstering or discrediting candidates for specific state-based offices, we simplify our analysis of geographic patterns to the state-based targets of these messages, rather than examining the media market–level information that would contain messaging from multiple states in markets that cross state lines. For our geographic analysis, when we combine across issue categories and/or keywords, we do so at the ad level—scoring the ad with a 1 when it contains at least one issue category or keyword hit, and 0 otherwise—so that we do not double-count airings.

Results

Descriptive Results of Race- and Gender-Related Advertising Patterns

As a first step in understanding discussions relevant to race in the 2022 midterms, we examine aggregate patterns by party in addressing race-related topics from WMP's traditional issue battery. Figure 1 plots the overall volume of attention to these issues (upper-left panel), the proportion of overall airings by each party (upper-right panel), and the proportion of pro-Democratic airings (lower-left panel) and pro-Republican airings (lower-right panel). Especially because Democrats have held television ad advantages in the past few cycles (Franz, Ridout, and Fowler 2023; Ridout, Fowler, and Franz 2021), the lower panel provides insight into how each of the parties (and their surrogates) allocate their attention relative to the total amount of advertising being aired for each side respectively.

As shown in the top-left panel of figure 1, discussion of issues related to criminal justice were the most prominent race-related topics in 2022, with crime topping the list at nearly 270,000 airings (16% of total ads supporting either of the two parties), followed by police (including police brutality/racial violence) with a little more than 230,000 airings (14%) overall. Incarceration and sentencing ranked fourth overall at nearly 80,000 airings (5%), followed by protests and riots (a little more than 54,000 airings; 3%). Mentions of capital punishment (more commonly referred to as the death penalty in advertising) occurred in fewer than 5,000 ad airings (0.3%). Illegal drugs received roughly double the attention that opioids did (more than 50,000 airings, compared to almost 25,000). Of the non–criminal justice and drug topics, immigration ranked highest at third overall, with a little more than 19,000 airings (8%), whereas mentions of DACA or Dreamers occurred very rarely (roughly 4,400 airings or 0.3%). Although critical race theory made headlines quite prominently in the past and featured heavily in the Virginia gubernatorial contest in 2021, references to the topic in 2022 were virtually nonexistent (fewer than 700 in total, 0.04%) but were pro-Republican when they occurred. Mentions of civil rights, racial discrimination, and affirmative action were also rare, constituting a mere 1% of overall airings. While earlier work found that civil rights/racial discrimination (to include affirmative action) accounted for only 1% of pro-Democratic airings and fewer than 0.1% of pro-Republican airings in the 2011–2012 and 2015–2016 cycles combined (Fowler et al. 2019), it is notable that just two years following widespread national attention devoted to racial injustice and during an unequal economic recovery that exacerbated racial inequities, this issue category would remain so minimal (although see Chudy and Jefferson 2021).

Turning to the upper-right panel showing the distribution of airings by party, race-related topics in the 2022 midterms were often led by pro-Republican voices. On just three law-and-order topics (police, civil rights, and capital punishment) did pro-Democratic airings outweigh pro-Republican ones, and two out of three of those issue categories were very infrequent. Even though pro-Democratic airings were more likely to mention police, the combination of issues that make up discussions of criminal justice led to pro-Republican messaging in the aggregate being more prevalent on air. The only other category where pro-Democratic messages were more prominent than pro-Republican ones occurred in references to opioids, which suggests a willingness among Democratic-leaning sponsors to discuss drugs when they have been constructed as an issue among white populations (Mendoza, Rivera, and Hansen 2019). Pro-Republican airings, on the other hand, were much more likely to mention illegal drugs than they were opioids.

This partisan tilt is even more striking when one considers that pro-Democratic messages were more common on air overall, meaning there were more chances for pro-Democratic airings to swamp pro-Republican ones. For this reason, we turn to the lower panels of figure 1, where we calculate how much attention each side spent on race-related issues as a proportion of all messaging supporting Democratic candidates (on the left) and Republican candidates (on the right). Crime was a much bigger focus of the pro-Republican campaign agenda than it was for Democratic-leaning messages, with 23% of pro-Republican airings mentioning issues of crime and law and order compared to 11% of pro-Democratic messaging. Relative to overall airings for each side, references to police made up a similar proportion of each pro-partisan agenda. Incarceration and sentencing appeared in nearly 1 out of every 10 pro-Republican airings during the general election, more than 4 times the proportion of pro-Democratic airings.

Figure 2 displays the overall volume of references to keywords that we divide roughly into nonidentity words (which we label policy-related in the first row) and identity words (in the second row). The left panels display the total count of the number of airings containing each keyword, and the right shows how the total airing distribution breaks out between sponsors supporting the two parties.

Starting with the upper left, consistent with our WMP issue category analysis, crime is the most commonly used word. Adjectives likely to be related to discussions of crime such as dangerous, violent, prison, murder, bail, release, and convict all were mentioned in tens of thousands of airings, whereas words like justice, immigration, and sentence(-s,-ing) were less common. References to discrimination and discriminating (through the stem discriminat*), equal, and voting right(s) were virtually nonexistent.

Turning to the party breakdown of these keywords in the upper right, racial discourse as it relates to policy-relevant keywords was dominated by pro-Republican messages. While the adjectives dangerous and violent were used slightly more frequently by Democratic-leaning sponsors, pro-Republican sponsors were far more likely to use many others. The word defund likely indicates that both parties were referencing “defund the police,” suggesting the issue was contested between the parties. Words we might expect to be used to reference race in pro-Democratic advertising on issues of racial justice—including justice, discrimination/discriminating, voting right, and equal—were indeed most likely to appear in messages from Democratic sponsors, but they were low in volume relative to the preponderance of attention to discussions of crime.

The bottom row of figure 2 explores race-related identity words. Police and criminal were the most commonly referenced identity terms, followed by illegal (which could arguably be a policy reference or an identity reference), whereas specific mentions of groups defined by race and ethnicity were much less common. Immigration-related terms (illegal, immigrant, and illegal alien) were more likely to occur in pro-Republican messaging, and there were no references to the term undocumented. Of the four explicit racial identity terms most frequently used (race, Black, white, and racis*, which would pick up both racist and racism), two were nearly equivalent between the parties in volume of airings (race and Black), while pro-Democratic messages were slightly more likely to use the word white. Although racis* (capturing both racism and racist) occurs less frequently than the other three, it was used solely in Republican affiliated advertising.³ In general, the race-related identity words used during the general election tell a story of disinclination by both parties to explicitly name racial identities (although note that our chart does not include terms that we examined but that did not occur in advertising, such as Indigenous).

We also examined how prominent each of the race-related keywords was to the overall volume of messaging for each party (fig. 3). Looking first at the policy-related discussion (top row), we see that Republicans and their allies devoted much more of their overall ad agenda to crime along with most of the scary adjectives that were likely used in law and order–related messaging. Justice was mentioned in fewer than 2% of pro-Democratic advertising, roughly equal to references to the word murder. The analysis of identity-related keywords (bottom row) demonstrates that both parties devoted only limited attention to explicit racial identities, but when they did, pro-Republican airings discussed criminals and racism more than did pro-Democratic airings.

To summarize: the dominant issue and keywords related to race-relevant messaging in the midterms were those most closely connected to law and order. Discussions of racial inequality, racism, injustice, and explicit references to racial identities were extremely rare. Although both parties include content about race in their advertising, the Republican Party appears to be driving the conversation.

We now turn our attention to an assessment of how the 2022 midterm content on issues of gender and gender identity compared to 2022 midterm content on race. As shown in figure 4, abortion was far and away the top issue topic related to gender (top left), and the change in the x axis to account for the nearly 300,000 airings devoted to the topic conveys that this volume is well beyond the attention given to crime. Women's health ranked a distant second, on par in overall volume of mentions with mentions of incarceration and sentencing from the race-related issue topics, which ranked fourth overall behind crime, police, and immigration. The upper-right quadrant shows that in stark contrast to the results for race, issues related to women specifically (although not to LGBTQIA+ communities) were dominated by pro-Democratic messaging. In contrast, LGBTQIA+ discussion was much lower in overall volume, and pro-Democratic voices were largely absent, ceding the conversation nearly entirely to Republicans.

The bottom left panel of figure 4 shows that nearly a third of all pro-Democratic federal airings discussed abortion and abortion rights, with a smaller subset focusing on women's health issues beyond abortion. The bottom right panel shows that Republican-leaning advertising was more likely to reference LGBTQIA+ topics than to discuss abortion or women's health, and their space devoted to this issue was more than twice that of pro-Democratic messaging.

Figure 5 displays results of our keyword searches on policy- and identity-related searches for gender. In the top panel, we again see that abortion was the most frequently mentioned term, followed by the word rape. Other health care–related terms (Medicare, health care, prescription drugs) and Social Security were the next most frequent words, and all six of these terms appeared predominantly in pro-Democratic messaging. References to pro-life, reproductive, pregnancy, marriage, and contraception were less prominent but similarly dominant in pro-Democratic messaging. In fact, the only pro-Republican advantage in any of these gender-related policy words that had any substantial volume behind them were references to gun issues (NRA and the Second Amendment).

Turning to the identity-related keywords, in stark contrast to racial identity references, there were abundant explicit mentions of women in 2022 federal advertising. Women is the most common keyword, followed closely by woman, suggesting both a group-based and a more singular or personal focus of discussion. These references each occurred in more than 10% of pro-Democratic advertising (figure not shown). Seniors and victims along with kids, children, and girl round out the secondary tier of gender identity–related keywords (seniors and victims were included because of their association with issues owned by women, and the frequency with which women politicians typically use these words). Although seniors and victims were words used more frequently in airings by Democratic-leaning sponsors, references to children, including girls (plural) and boys, were more common in pro-Republican ads. At least some of the “boys” references in pro-Republican advertising appeared to be mentions of Proud Boys, which would be both a racial appeal and a gendered one. Notably, Republicans were the sole voices using explicit references to transgender, gender, and identity (also “pronoun,” although that term was extremely uncommon).

Geographic Patterns in Race- and Gender-Related Appeals

Advertising is unequally distributed across the country, with higher levels of airings targeted at competitive races in particular. Advertising sponsors typically tailor their messaging to their audiences (in this case voting constituents), which is why it is important to understand how messages vary geographically. Figure 6 displays the overall volume of attention (in total airings) to race-related issues by state in the top-left panel. We perform a similar analysis using our keyword approach (available in appendix B), but because these results are much more sensitive to inclusion/exclusion criteria, we stick to the broader issue categorization for presentation. The upper right shows what percentage race-related airings were of the total amount of pro-Democratic and pro-Republican advertising in each state. In the bottom row, we display how much attention race-related issue mentions garner as a proportion of advertising in support of each party respectively, with pro-Democratic airings on the left and pro-Republican airings on the right.

As shown in the upper-left panel of the figure, the volume of airings containing race-related issue content exceeded 25,000 airings per state in seven states (Arizona, Georgia, Florida, New York, Ohio, Pennsylvania, and Wisconsin). Turning to the prominence of race-related issues among all federal election ads targeted at the state (upper right), there is substantial variation in the extent to which race-related issues received prominent billing in election ads. Some states—for example, Mississippi and West Virginia, which did not have very many ad airings overall (a total of 281 airings targeted Mississippi contests and 316 targeted West Virginia)—were much more likely to mention race-related issues in their advertising as a proportion of the small total. On the other hand, some states that saw high volumes of ads overall—for example, New York, Florida, and Arizona, which each had more than 50,000 ad airings in the general election cycle—also saw a nontrivial percentage of ads related to race.

The partisan breakdowns on the bottom row of figure 6 show (similar to our aggregate analysis above) that Republican-leaning airings are much more likely to be talking about race-related issues across the United States. Comparing the map on the left to the one on the right, in almost every state, race-related issues are more prominent in pro-Republican advertising as a proportion of their overall agenda than in pro-Democratic advertising. The exception to that pattern is Florida, where Democratic Senate hopeful Val Demings aired an unusual number of ads mentioning police. This is an important exception to note, because if we were to exclude Florida airings from the overall count of race-related airings, we would find that pro-Democratic attention to race is even lower overall than our aggregate statistics suggest. In other words, Democratic-leaning messages appeared to have ceded most of the race-related conversations in election advertising to Republicans and their allies.

Figure 7 displays the comparable results for gender, which rely heavily on the issue of abortion because it was most prominent among the gendered issue topics, as shown in figure 4. The upper-left volume panel shows that gender-related issues were prominent, but not quite as high in volume as the race-related topics (almost certainly driven by the fact that Republicans were all but silent on the issue of abortion during the general election). Discussion was widespread, with states across the country getting thousands of ads. As a percentage of all airings (upper-right panel), gender-related issues constituted more than 40% of airings in a few states: Massachusetts, Delaware, Florida, Alabama, and Tennessee, the latter two of which had very few total airings, as shown in the volume panel.

The bottom row of figure 7 shows the stark contrast between the parties and an opposite pattern from the one displayed for race. More specifically, pro-Republican discussion of gender-related issues was low as a function of their overall agenda, with the two exceptions being Tennessee and Alabama, where the total amount of advertising was very low. Democratic-leaning ads, however, were much more likely to make gender-related topics a prominent part of their campaign agenda, with multiple states across the country featuring the issues in a substantial percentage of their ads.

Discussion and Limitations

Race-related mentions in 2022 midterm advertising were overwhelmingly crime- and law and order–focused, with very little attention to racism, racial injustice, and the structural barriers that lead to widespread inequities. In stark contrast to gender, racial references in advertisements came from both parties and were less likely to call out specific identity groups while also using more coded language such as dangerous, criminals, and violent. In this way, the 2022 election discussion seems to be consistent with earlier research that demonstrates the prominent use of implicit, rather than explicit, racial appeals (Mendelberg 1997; Stephens-Dougan 2021). However, the fact that these references are implicit does not mean they have no effect on the politics of racial and health equity. Decades of evidence suggest that implicit cues can activate racial animus and increase support for punitive policies over social safety net provisions (Gilens 1996; Peffley and Hurwitz 2002). Our geographic analysis clarifies that most of the aggregate volume on the pro-Democratic side came from an outlier in Florida, and if we were to exclude those messages, it becomes even clearer that racial discussions in advertising were predominantly led by Republicans and their allies.

The patterns we take note of here—with Democrats making a concerted effort to emphasize gender and highlighting abortion policy and its connection to health in their messaging, while failing to prioritize a similar strategy with regard to racial inequality—suggest a coordinated strategic effort. Indeed, since campaign advertising is deliberately planned and executed with careful attention to audiences and message effects, this divergence was not incidental. Especially given recent public scholarly debate over the effectiveness of appeals rooted in racial identity (English and Kalla 2021), which received considerable attention and made the pages of The New York Times in early 2021 (Edsall 2021), concern about the effects of messages centering race or racial justice may have influenced consultant decisions for 2022 about not only which issues to emphasize but also how to discuss them. A recent scoping review of racial equity–related messaging research highlights that most scholarly evidence focuses on short message treatments and short-term effects, and that much more work is needed to make firm conclusions about the effectiveness of racialized appeals in galvanizing support for policies to reduce disparities (Niederdeppe et al. 2023).

Regardless of what is leading to the distinctions in the discussions of race and gender, these differences have big implications for how the public understands these issues. There is considerable evidence that attitudes with regard to race and health equity are already quite polarized, and this messaging environment is unlikely to reduce these effects (Gollust et al. 2022). We also note that the distinct approach the two parties have taken to discussing race is likely to exacerbate this polarization around racial equity among the mass public. By focusing on race through a narrow lens of its connection to crime, Republicans are likely to reinforce this connection in the minds of their partisans (Liu et al. 2023). In sum, messaging around racial politics and identity in 2022 campaign communication is unlikely to contribute to improved public understanding of the links between racism and health. In fact, it is more likely (albeit untested here) that the aggregate impact of this messaging could enhance negative responses among the public to policies that seek to advance health equity, by making negative and stereotypical imagery of racial groups more salient in the information environment.

Before turning to conclusions, it is important to acknowledge some limitations of our analysis. First, we only examine television advertising. Although earlier literature suggests that this is where the substantive policy discussion occurs, there may be different ways in which issues along with group identities are discussed in other campaign rhetoric online or in speeches that are worthy of future exploration. Although our issue battery analysis relies on a more comprehensive approach to measuring policy discussion, our keyword analysis relies on imperfect speech-to-text captioning. Proksch, Wratil, and Wäckerle (2019) validate the use of automatic transcriptions in political text analysis, yet our analysis may still have false positives and false negatives of keywords related to racial and gender appeals. For example, a keyword search for alien resulted in a false positive where ASR transcribed Pennsylvanian as “passive alien.” As another example, ASR does not transcribe acronyms such as ICE and DACA correctly. Moreover, our analysis relies on explicit mentions of identities to identify racial and gender appeals. Future work may explore the sentiment surrounding specific group entities to better understand the use of identity-based appeals. In addition, although our results provide some insight into race-related patterns, we urge more attention to and validation of building more robust keyword dictionaries of terms that are likely to appear in issues related to people of color and other racialized identities. Finally, while we make claims about the potential impact of exposure to the types of appeals and cues we observe in advertising, we are not testing those effects on the public empirically here; such analyses would require linking exposure to advertising with public opinion.

Conclusion

The priorities reflected in the issue content of campaign ads are meaningful and shape public discourse as well as the priorities of legislators once in office. As we examine the discourse in ads around issues of race, a clear story emerges. There is a clear focus on policy issues, and particularly that of criminal justice, and much less focus on racism more broadly and racial injustice, even in light of the recent mass mobilization around this issue. Furthermore, for these policy-related discussions of criminal justice, Republicans led the charge, with greater overall volume of discussion on these issues and a larger proportion relative to the rest of their messaging.

These distinctions are instructive and will likely have many downstream consequences for policy in coming years. Particularly, the potential set of policies that the GOP has prioritized in the arena of criminal justice have clear implications for health, especially for people of color, who have suffered and continue to disproportionately suffer from criminalization and punitive policing strategies. Research in public health continues to show that contact with the criminal justice system has population health consequences (see, e.g., Cottrell et al. 2019; Bowleg et al. 2020). It is equally important to note that the dearth of discussion of remedies for persistent health inequities fueled by racial discrimination and structural barriers makes it less likely that policy to address these issues will be at the top of the political and policy agenda. Connecting these intersecting inequalities with health and acknowledging their common roots is crucial to rectifying their harmful effects (Michener and LeBrón 2022).

We also find clear evidence that the polarized patterns were different for gender. The discourse around gender in the midterms was dominated by Democrats. We note one exception to this pattern, wherein discussion of LGBTQIA+ issues was higher among Republican-leaning messages; but for all other issues within this set, Democrats and their allies far and away outpaced pro-Republican messaging about gender. Abortion was the centerpiece of this discourse, with pro-Democratic advertising making clear connections to the population health impacts of the potential denial of important medical care.

We highlight this distinction between the patterns by race and gender not to detract from the focus on gender and abortion specifically, a topic that was extremely salient, and a top issue for voters. But the way this issue was prioritized and contextualized in messaging was a strategic choice. And it demonstrates that highlighting a polarizing and contentious issue is certainly possible. This raises an important question: why were racism and its deleterious impacts not shown similar care and consideration by Democrats and Democratic proponents? Instead, on this issue, Republicans led the charge, focusing on the issue through a criminal justice lens and ultimately playing a larger role in shaping the broader messaging environment surrounding race.

This pattern has a precedent. In the immediate aftermath of the passage of the ACA, assessment of advertising indicated that Democrats did not mount an affirmative defense of the policy, in light of constant Republican attacks (Gollust et al. 2022). In this messaging environment, without consistent competitive framing from Democrats to match the GOP's output on the issue, the policy came under considerable threat. Republicans in Congress came within a handful of votes of overturning the policy. Only once this threat was clear did Democrats rally around the policy and affirmatively support it at a level comparable to the Republican countermessaging. It is difficult to say what the effects of this strategy were, but it is clear that the Republican messaging did galvanize opposition, despite it not being enough to overturn the policy.

Polarization has become an ever-present reality in American politics, and our analysis of campaign messaging in the 2022 midterms shows that it was no exception to this trend. Even though we note clear distinctions in the ways Democrats and Republicans made racial appeals in their messaging—and a lack of connection of racial inequality to health, despite the myriad ways these issues are intertwined—we are only able to speculate about the consequences of this. More work is necessary to understand how exactly this will influence the attitudes of the public and subsequent policy making. But the results with regard to gender make clear that it is possible to prioritize the connections of policy to population health, even in these polarized times. Given the clear and dire consequences of racial disparities in health outcomes for the lived experiences of people of color, it is imperative to better understand how to reinforce an understanding of this reality in the mass public.

Acknowledgments

This article is a product of the Collaborative on Media and Messaging for Health and Social Policy. We would like to acknowledge the contributions of Dominique Monserrat and the Wesleyan Media Project undergraduate research assistants for their efforts in coding election advertising; in particular, we would like to thank Cecilia Smith for going above and beyond in ensuring we could include Spanish-language advertising. This work was funded by the Robert Wood Johnson Foundation (grant no. 79754). The views expressed here do not necessarily reflect the views of the Robert Wood Johnson Foundation.

Notes

References