Introduction

What is political and what is not? The distinction between what counts as political and what does not have relevant implications. Labeling an issue as political typically implies that it falls within the purview of government and legislative bodies tasked with its regulation. Conversely, matters categorized as non-political may be viewed as lying beyond the realm of formal institutions and therefore subject to other forms of decision-making and power dynamics. The political status of an issue may also influence public discourse and the formation of public opinion. Political issues tend to attract more media attention and scrutiny and therefore can have a greater impact on public opinion and social norms.

However, the concept of politics is multifaceted, encompassing a wide range of ideas, expectations, and dimensions. Despite its centrality in political science, politics remain one of the most contentious concepts. Yet most opinion research assumes that people share a common understanding of the concept and what it encompasses (O’Toole 2003; Walsh 2004). Recent research has challenged the assumption that people share a common understanding of the political realm (Eliashop 1998; Görtz and Dahl 2021; Manning 2010). On the one hand, citizens differ in their perceptions of the political world in terms of the number of issues they consider political (i.e., the breadth of their understanding of politics). While some hold a narrow view of politics, others embrace a broader perspective (Moray and Eveland 2016; Fitzgerald 2013). On the other hand, individuals also diverge in the specific content of their narrow or broad conceptions (i.e., the types of issues deemed political or not political), with these disparities often rooted in social factors (Görtz et al. 2022). Furthermore, these diverse perceptions regarding the concept of politics are pertinent for understanding attitudes and behaviors crucial to people’s citizenship, such as political interest or participation (Görtz and Dahl 2021). While this research has shed light on the diversity of perceptions and understandings of politics, it has largely overlooked the role of the media in sha** these perceptions.

In this article, we make a substantive and methodological contribution to the emerging literature on individual understandings of politics. We do so by theorizing and analyzing the role of the media in sha** citizens’ views about which topics are considered as political or not. More specifically, our research question is how the unexpected salience of abortion regulation in the media influences public perceptions of its political nature. First, we contend that the media can play a significant role in sha** citizens' views regarding the politicization of specific issues. Second, we apply an innovative research design to empirically examine the media’s role in fueling the politicization of these issues. In doing so, we align with recent advancements in the field that seek to approximate experimental benchmark using observational data (see, e.g., Foos and Bischof 2022).

We capitalize on the unforeseen surge of attention that the issue of abortion regulation gained in the media during the fieldwork period of an online survey administered to a representative sample of the Spanish population (N = 1651). To measure people’s perceptions of politics, this survey offered a list of topics and issues, including abortion regulation, prompting respondents to classify them as political or non-political. Our survey was conducted for 1 month. During this period, a proposal to reform Spain’s abortion law unexpectedly captured media headlines, thrusting abortion into the spotlight. Leveraging the Unexpected Event During Survey Design (UESD) methodology (Muñoz et al. 2020), we exploit this media-driven shift during data collection to examine how the prominence of abortion in the media influences citizens’ propensity to consider this issue as political. This approach allows us to empirically assess the media’s impact on citizens’ perceptions of the political sphere in a real-world setting.

Our analyses show that the increasing prominence of the abortion issue in the media has impacted perceptions of its political nature. Those surveyed after the issue gained prominence in the media were about 10 percentage points more inclined to perceive abortion as a political matter. These findings underscore the significant role media coverage can play in sha** citizens’ perceptions of the political world.

What is political?

Politics is a complex and multifaceted concept that elicits a plethora of different ideas and interpretations among people. Its definition can vary widely, encompassing notions such as the exercise of power, the discipline of government, the making of collective decisions, the allocation of scarce resources, the practice of deception and manipulation, the struggle to reach agreements, or a service that politicians provide to citizens, among many others. These varied interpretations underscore the challenge of establishing a unified understanding of this polysemous concept (Bartolini 2018; Palonen 2003). Most research on public opinion, however, assumes that people share a common understanding of politics that revolves around the concepts of government, parties, and conflict over politics (O’Toole 2003; Walsh 2004; Prior 2019). These conceptions are typically associated with governmental functions and procedural aspects of politics.

Recent studies have challenged this assumption by revealing relevant differences in the topics and issues deemed "political" across countries such as Canada, Germany, Norway, Spain, Sweden, and the United States (Görtz and Dahl 2021; Görtz et al. 2022; Görtz 2024; Ferrin et al. 2020; Fitzgerald 2013; Mathé 2017; Podschuweit and Jacobs 2017). These studies suggest that there exists heterogeneity in both the breadth and nature of issues that citizens identify as political. While some people have a broad and inclusive view of the political realm, encompassing a wide array of topics and objects, others adopt a narrower perspective of the political world and consider that only a few topics or issues are political.

Most previous studies conclude that people’s understanding of politics expands with age and higher levels of education (Eliasoph 1998; Fitzgerald 2013; Henn et al 2005; Manning 2010; Mathé 2017). Education and life experiences equip citizens with relevant information, resources, and insights that can expand their views of the political world. Life experiences, such as entering the workforce, forming a family or becoming a parent, expose individuals to new circumstances that they perceive as necessitating government intervention through specific regulations or public services. These private experiences may widen an individual’s conception of what constitutes a political matter, as they recognize the potential need for public regulation addressing these issues. For example, a laborer may regard working conditions as a policy concern and advocate for governmental regulations to protect workers’ safety, whereas a new parent may contemplate the importance of public provision of prenatal care or early childhood education. Without these experiences, individuals might be more likely to believe that these are private rather than political issues.

Hence, existing literature underscores that variations in citizens’ conceptualizations of politics are deeply rooted in social contexts. Socialization processes, life circumstances, and related socio-economic experiences play crucial roles in sha** individuals’ perceptions and demarcating the boundaries between the public and private domains (Görtz et al. 2022). These personal encounters can engender enduring disparities among individuals and social cohorts in their interpretations of what qualifies as political and what falls within the realm of private affairs. Such divergences may arise from disparate cultural backgrounds, educational opportunities, exposure to diverse social environments, or socioeconomic statuses. As a result, individuals and social groups may develop distinct understandings of the significance and scope of political involvement, which can subsequently influence their political attitudes, behaviors, and engagement with public affair.

Aside from one’s own personal circumstances, there are other relevant factors that can influence what one considers political. We assert that the media should play a crucial role in this regard. Previous research in the field of communication underscores the pivotal role of media in sha** individuals’ political attitudes (Iyengar 1990; Iyengar and Kinder 1987). The agenda setting theory highlights the pervasive impact of media on public perception by directing attention to issues considered important. Through its selection of topics, the media not only signals their importance to audiences but also molds the way citizens think about these topics (McCombs 2004). In essence, mass media wields the power to influence which issues are prominent, how they are framed, and how individuals perceive their importance (Zaller 1992).

Despite the fragmentation of media audiences and the decline of a shared agenda due to the rise of targeted broadcasting, recent observational, experimental, and quasi-experimental studies have shown that news media retain their critical role in connecting citizens to the political world. For instance, Beckers et al. (2021) utilized a time-diary approach within the Flemish media context, concluding that different media sources have substantial informative effects. Similarly, King et al. (2017) conducted a field experiment that revealed substantive media effects. By assigning specific small media outlets to publish articles on particular topics on randomly selected dates, they discovered that this intervention increased public expression among Americans. Individuals exposed to these outlets were more inclined to voice their opinions on social media platforms.

This suggests that social media can complement rather than supplant the agenda-setting power of traditional media (Feezell 2018; Gilardi et al 2022). Drawing on a quasi-experimental design, Foos and Bischof (2022) offer additional evidence reinforcing the idea of the ongoing significance of media in sha** public opinion. A boycott of “The Sun” in the Liverpool area, induced externally, had a substantial impact on individuals’ attitudes towards the EU, leading citizens exposed to this tabloid to adopt a more critical stance towards the EU. Overall, these studies provide convincing evidence in favor of the idea that the media retain significant power in sha** public opinion. Given that political issues often encompass diverse interpretations and viewpoints, they are particularly susceptible to media sway.

Bridging the political science and communication studies literatures, we contend that the media play a crucial role in sha** individuals’ understanding of what is and is not political. Our claim is that beyond personal predispositions, contextual factors such as the media also influence the formation of opinions about the blur boundaries between what is deemed political and what is not. This influence manifests for two main reasons. Firstly, the media serves as an environmental stimulus encouraging citizens to pay greater attention to the issue at hand (Cook et al. 1983). This heightened attention increases people’s familiarity with the subject matter, subsequently elevating its importance in the realm of political discourse. Secondly, and perhaps more importantly, the media amplifies authoritative political voices within public debate. These voices, that can provide relevant cues for public opinion formation (Pannico 2017), furnish detailed explanations about the reasons for both supporting and opposing specific policy proposals or reforms, thereby enhancing the likelihood that citizens associate the issue with politics and political competition. In essence, individuals exposed to prominent political figures speaking, debating, and deliberating on a particular issue are more inclined to interpret it as politically relevant, rather than relegating it to the intimate, personal, or private sphere (Cook et al. 1983). Following this reasoning, our first hypothesis states that the emergence and increased visibility of the topic of abortion in the media heightens the likelihood that citizens perceive it as a political issue (H1). Further elucidation on the political context and the subject matter under analysis will be provided in the subsequent section.

Political context: salience of abortion in Spain

When our survey was fielded (December 2013), the Spanish conservative party—Partido Popular (PP)—had been in power for 2 years. During this term, the PP implemented a series of decisions that represented significant steps backward in gender equality policies (Verge 2020). One of the most controversial initiatives of the conservative government during its 2011–2015 tenure was the reform of Spain's abortion law. In 2013, the Ministry of Justice proposed a new abortion law aimed at restricting abortion to two specific circumstances. Firstly, when there is a serious threat to the mother’s health (up to 22 weeks into pregnancy), which must be fully certified by independent doctors not affiliated with abortion providers. Secondly, in case of pregnancy resulting from reported rape (only within the first 12 weeks of pregnancy). Additionally, the draft of the law also included several provisions intended to dissuade women from having an abortion.

The Minister of Justice proposed these changes to appeal to the most conservative constituents, who vehemently opposed the abortion law enacted by the previous socialist government in 2010. The 2010 law, which allowed abortions at the woman’s request up to the 15th week of pregnancy, was based on time limits similar to those in most European countries. The new law proposal, drafted by the conservative party, was also a response to the political campaign organized by pro-life organizations in Spain. For example, the organization DAV (Derecho a Vivir—Right to Life) organized a vigorous campaign that included the dissemination of emotive videos on social media, and the installation of 150 billboards in major Spanish cities, directly addressing the Minister of Justice and urging him to fulfill his electoral promises regarding abortion (see Fig. A1 in the Supplementary Information file). It is important to note, however, that while the draft prepared by the Ministry was presented at a cabinet meeting on December 20, 2013, it never progressed through the legislative process, despite the conservative party holding a sufficient majority in Parliament to pass the law. This suggests internal disagreement within the party regarding the political convenience of such a reform. The content of this draft received extensive coverage and discussion in the media starting on December 11. On that day, leading television channels and national newspapers reported on the Justice Minister's response to a MP from the opposition Socialist Party (PSOE) during a parliamentary debate. In his response, the minister summarized the content of the draft of the new abortion law.Footnote 1

As a result of the parliamentary debate surrounding the content of the new proposed abortion law, the topic of abortion emerged as one of the focal points extensively covered by the Spanish mainstream media during this period. On December 11, 2013, newspapers, radio, and television channels devoted considerable airtime and column space to the parliamentary debate, making the first public presentation and debate of the PP’s proposed reforms for the first time during their conservative tenure from 2011 to 2015. From December 11 until the month’s end, various media platforms extensively covered the contents and implications of the proposed law. For instance, there was a notable surge in print media coverage of abortion the December 11 parliamentary debate. An analysis of the content of the 10 most widely read newspapers in Spain based on the FACTIVA database shows that between December 3 and December 10, only 6 articles mentioned abortion in their headlines.Footnote 2 In contrast, between December 12th and December 19th the headlines of 33 articles mentioned abortion. This indicates a 550 percent increase in news coverage directly related to abortion following the Justice Minister’s discussion of the draft legislation of the new abortion law in Parliament.Footnote 3

The debate over the proposed reform of the abortion law in the Spanish Parliament on December 11, coupled with its extensive media coverage, has resulted in a relevant escalation in the importance of this issue among the Spanish public. summarizes the relative frequency of Google searches for the word “abortion” in Spain between December 2003 (the first year for which these data are available) and December 2021, providing a complementary measure of the importance of the abortion issue at the societal level (see Mellon 2014). Shows a distinct peak in the frequency of searches on the topic of abortion in Spain in December 2013. In fact, the highest number of searches during this period is reached in December 2013, indicating the relevant salience of this topic in Spanish society.

Before delving into our data and identification strategy, it is important to discuss the nature and nuances of the subject under examination: abortion rights. Abortion rights are intrinsically linked to women’s interests, as only biological women can experience pregnancy. Consequently, the regulation of abortion directly intersects with women’s autonomy over their reproductive behavior (Philips 1995).

Scholars in the field of gender and politics have concluded that female legislators are more inclined to champion issues directly impacting women’s bodies, such as abortion, sexual harassment, or violence (O’Brien and Piscopo 2019). Previous studies also show that the concerns and priorities driving women’s engagement often diverge from those of men. While women may express less overall interest in national politics compared to men (Fraile and Gómez, 2017; Fraile and Sánchez-Vitores 2020), they exhibit greater interest and knowledge in matters directly affecting their daily lives, such as local politics, civil rights, and welfare state policies (Campbell and Winters 2008; Coffé 2013; Ferrin et al. 2018, 2020; Sánchez-Vitores 2019).

Building on these findings we expect women to consider abortion a political issue to a greater extent than men, irrespective of its prominence in the media. In other words, women’s views on abortion may be more entrenched, regardless of whether the topic is actively debated in public discourse. Consequently, the amplification abortion-related discussions in the media may have differing impacts on men and women, with potentially stronger effects among men. Due to the nature of the subject matter, women may already have more crystallized attitudes about towards abortion (Campbell and Winters 2008; Ferrin et al. 2020), which could limit the impact of the media among them. In other words, there might be less scope for media-induced shifts in the perception of abortion as a political issue among women compared to men, thus diminishing the media’s sway over the former. Our second hypothesis (H2) posits a moderation effect according to which the potential influence of the media on the probability that citizens consider abortion a political topic will be more pronounced among men than women.

Data and identification strategy

Data and sample

Our empirical analyses draw on an online survey conducted among a sample of 1651 Spanish citizens via the opt-in access panel of the commercial company Netquest. Respondents were selected by quota sampling within the Netquest panel, using quotas for gender, education, age, and region. These quotas ensured that the final sample matched these characteristics of the Spanish population between the ages of 18 and 70.

The survey was fielded between November 22, 2013 and December 23, 2013Footnote 4. While the data are somewhat dated, it provides us with a unique opportunity to address our research question with a quasi-experimental design. Additionally, these data allow us to examine the history of a political process that can be seen as the pivotal starting point for the emergence of feminism in Spain. Precisely, the first large-scale mass demonstration organized in Spain in February 2014 was aimed to oppose the conservative government's proposed restrictions on abortion rights. A decade later, the feminist movement in Spain has rapidly and extensively expanded, moving from theory and organizational structures to the streets. This development positions the feminist movement of Spain among those with the highest mobilization capacity globally (Jimenez et al. 2022).

Given the non-probability nature of our data, we compare the profile of our respondents with that of respondents interviewed in face-to-face probability samples collected in the same month and year (one was collected by the Spanish CIS-ref CIS 3008, and the other was the round 6 of the European Social Survey, see Table A.9 in the Supplementary Information File). The comparison shows that our sample is slightly younger than the general population because respondents over 70 were not targeted in our sample. A younger sample also means higher levels of education and fewer respondents who have retired from the labor force. Our sample is also slightly more interested in politics. Using the same metric, the mean value of declared interest in our online survey is 1.3, while it is 1.2 in the 6th round of the European Social Survey. We consider it unlikely that these minor differences bias our results.

Based on the literature on attitudinal updating (e.g. Zaller 1992) one could think that the relatively younger and more educated sample might both heighten and limit the impact of the increasing salience of abortion in the news. However, the differences between our sample and similar representative samples that we use as a benchmark are relatively small, so any bias caused by the non-representative nature of the sample should be rather limited.

Identification strategy

Analyzing the real-world influence of the media on public opinion faces relevant challenges due to endogeneity issues stemming from media self-selection or economic pressures, leading the media to compete with one another and target specific audiences. While randomized experiments offer a method to isolate causal effects, they often lack realism and external validity. Achieving the experimental benchmark using real-world evidence and observational data is far from straightforward. Our study contributes to this goal by addressing our research question using an unexpected event identification strategy (UESD) (Muñoz et al. 2020). Our design revolves around comparing the responses of those who were surveyed before December 11 and those who answered the survey after that date, when abortion emerged as a prominent issue in the media. Those surveyed after that date represent our treatment group, while those surveyed before December 11 serve as our control group.Footnote 5

Previous studies have examined how shifts in media content over time affect citizens' attitudes. For example, Zaller and Hunt (1995) tracked the association between media coverage of multiple events during a presidential campaign and public opinion, using aggregate opinion data and rolling cross-sectional surveys. Similarly, Barabas and Jerit (2009) compare multiple surveys over time to analyze gains in political knowledge resulting from increased media coverage of specific issues. In contrast, our study takes a different approach. We compare the opinions of citizens interviewed in the same survey before and after the issue of abortion unexpectedly gained prominence in the media during the fieldwork. This approach offers several advantages, including the absence of a time lag between the change in attention and the interviews, the full comparability of the survey instruments used before and after the surge in attention to the issue, and the existence of standardized procedures for examining the main assumptions of this identification strategy (Muñoz et al. 2020).

The UESD enable us to identify the impact of heightened media attention on the topic of abortion on respondents' views regarding whether abortion is a political issue or not. This design hinges on two assumptions: Ignorability and Excludability (Muñoz et al. 2020). Ignorability means that respondents are comparable before and after December 11, while excludability implies that there are no other contextual changes that could alter respondents' opinions about abortion being a political issue.

A key decision related to these assumptions is the selection of the time frame (or 'bandwidth') around December 11 that we incorporate in our estimation. Ideally, our estimate would include only respondents who participated in the survey on dates close to December 11, as this would increase the plausibility of the two aforementioned excludability and ignorability assumptions. However, one must also consider the reduction in the number of observations resulting from narrowing this range of days to prevent a study with insufficient power.

To determine the bandwidth, we performed a power analysis. Specifically, we analyzed the range of control and treatment groups (along with the corresponding sample size) needed to detect a meaningful change in the proportion of citizens viewing abortion as a political issue. We define a meaningful effect as one corresponding to a change of ¼ of a standard deviation in this variable (as measured in the control group), equivalent to approximately a 12% change in the proportion of respondents considering abortion as a political issue. The results, summarized in, indicate that the first range surpassing the threshold of 0.8 is the one that includes eight days before and after December 11. Accordingly, our control group includes those surveyed between December 3 and 10 (637 respondents), while the treatment group consists of those surveyed between December 12 and 19 (174 respondents).

We exclude respondents interviewed on December 11 because we cannot be certain that they were exposed to the news media discussion of abortion before participating in the survey. Our research design assumes that all treatment group participants had been exposed to news media coverage of abortion. Given the salience of the issue during this time period, we find this assumption is reasonable. However, we recognize that significant coverage of a given topic in the mainstream media does not guarantee equal exposure to all citizens. Therefore, we adopt a conservative strategy and treat our estimates as intent to treat effects (ITT) rather than average treatment effects (ATE).

Next, we conduct balance tests between the treatment and control groups. These balance tests include the following covariates: age, gender, labor market status, education level and whether or not one regularly goes to the polls, political knowledge, political interest, and ideology.Footnote 6 Reveals some imbalances between the control and treatment groups. Participants in the treatment group are less educated, more likely to be studying or retired, and more likely to be men.

Respondents in the two groups could also differ in terms of geographic location. To assess potential geographic imbalances, we conduct balance tests concerning the region of residence of the respondents. The results, summarized in Fig. A2 in the Supplementary Information, reveal some regional imbalances in the sample. Specifically, respondents in the treatment group are slightly more likely to reside in the Castilla y León region and slightly less likely to reside in the Madrid region.

These imbalances are not uncommon in UESD studies, often arising due to differences in respondent accessibility and survey organization aspects, such as the use of quota sampling, as in our case. To mitigate these imbalances, we adjust for differences in covariates using entropy adjustment and regional fixed effects (Hainmueller 2012). We rely on entropy balancing because it enables us to weight the units in the control and treatment groups such that the differences in these covariates become ignorable. To prevent any post-treatment bias, the entropy balancing weights include only the predetermined covariates mentioned in the previous paragraph (i.e., political knowledge, ideology, and political interest are excluded).Footnote 7 Table A2 in the supplementary information file shows that there are no discernible differences in the mean, variance, and skewness of these variables between these groups when the entropy balance weights are applied.

Outcome variables and estimation strategy

To analyze which topics and issues citizens consider political (or not), respondents were posed the following question: “Now you will see a list of topics. Please indicate which of these you tend to think as being political versus non-political.” A list of 19 issues was presented to respondents across two consecutive screens, with the order of these items randomized. Respondents were then prompted to classify each issue as either ‘political’ or ‘non-political’ (see Table 1).Footnote 8

Table 1 Share of respondents who state that a topic is ‘political’ in the full sample, in the treatment group and in the control group (ranked by percent who consider a given topic political in the full sample)
Full size table

Our outcome variable measures whether individuals consider abortion to be a political issue (value 1) or not (value 0). Respondents who indicate that they are not sure whether the issue is political or not are excluded from the estimation. We replicated the same analysis with an alternative operationalization of our dependent variable, wherein those who are unsure, along with those who do not perceive abortion as a political issue, are coded with the value 0. Our conclusions do not change when we use this alternative operationalization (see Table A5 in the Supplementary Information). To circumvent the complexities associated with including fixed effects and interactions in logistic models, we use linear probability models (LPM) estimated through ordinary least squares. Nevertheless, to provide robustness check, Table A4, Figs. A4 and A5 in the Supplementary Information replicate all LPM estimates with a binomial logistic specification, yielding similar results and conclusions.

Results

Table 1 summarizes the percentage of respondents identifying each issue as political within our entire sample. Consistent with previous studies, the results indicate considerable variation in the perception of each of these 19 issues as political or non-political. Notably, there is a 55% difference between the topic perceived as political by the highest proportion of respondents (employment, 91.6%) and the topic perceived as least associated with politics (religion, 36.6%). Specifically, regarding abortion, Table 1 illustrates that 59.4% of respondents consider it a political issue. Table 1 also summarizes the share of respondents that consider each topic political or not in the control and the treatment group. This clearly shows an increase in the share of respondents that consider abortion a political issue of around 8 percentage points after the increase in media coverage of abortion (from 57.5 to 65.

2% in the control and treatment groups respectively). However, we would caution against giving this descriptive evidence a causal interpretation. These should be considered as “naïve” descriptive results, since they are agnostic about some of the key assumptions of our identification strategy, which we account for through our modeling strategy.

Has the unexpected increase in media attention toward the issue of abortion contributed to a notable shift in citizens’ perceptions regarding its political nature? We begin our analyses by examining the results of a naïve model that includes only the treatment indicator. This indicator takes the value of 1 for those surveyed after December 11 and 0 for those surveyed before that date (within the ± 8 days bandwidth). This model (M1 in Table 2) shows that, on average, respondents interviewed after the increase in media coverage of abortion were 8% more likely to consider it a political issue, confirming the descriptive evidence summarized in Table 1. However, there is a substantial level of uncertainty surrounding this effect. It is statistically significant at the 10 percent level (with bounds of the 95% confidence interval = [− 0.008, 0.163]). It’s worth noting that this naïve model overlooks the covariate imbalances identified in the preceding section, which pose a threat to the ignorability assumption.

Table 2 Main results
Full size table

In contrast, M2 and M3 in Table 2 include entropy balancing weights and region fixed effects that increase the plausibility of this assumption. These models demonstrate a substantial increase in the proportion of citizens regarding abortion as a political issue due to its heightened prominence in the mass media. According to the fully specified model (M3), following December 11, when abortion significantly gained salience and media attention, citizens were, on average, 10 percentage point more inclined to perceive abortion as a political issue. Additionally, in this instance, there is less uncertainty in the estimate, bolstering our confidence that the surge in media attention toward the abortion issue increased the likelihood of citizens considering it a political topic (p-value = 0.040).

The ITT uncovered by our analyses implies a substantial increase in the proportion of citizens who perceive abortion as a political issue. A change of 0.10 in the proportion of citizens regarding abortion as a political issue corresponds to a change of 0.2 standard deviations in the variable measuring whether abortion is viewed as a political issue (as measured among control group participants). In our survey’s ranking of topics (see Table 1), a 10% increase in the proportion of respondents considering abortion a political issue would elevate it above issues currently closely linked with politics, such as the environment. However, even in times of increased attention, abortion would not emerge among the topics most closely associated with politics. In any case, the intensification in the prominence of abortion in the media resulted in a significant increase in the percentage of citizens perceiving this issue as political. However, considering the particularities of the topic at hand, the media’s influence may not affect all citizens in the same way. Abortion primarily pertains to women's bodies, suggesting that their opinions on this matter may be less susceptible to being swayed by changes in the information environment.

Equation M4 in Table 2 includes an interaction between the treatment indicator and the variable measuring respondents’ gender (= 1 women and = 0 men). While the negative coefficient corresponding to the interaction term aligns with the hypothesized weaker effect for women, the product term is indistinguishable from zero (p-value = 0.425). The uncertainty in this estimate likely stems from the diminished statistical power resulting from splitting our analyses by gender. Indeed, additional power analyses (outlined in Fig. A3 in the Supplementary Information) reveal that the statistical power for estimating these heterogeneous effects falls below the 0.8 threshold for both subsamples (women and men).

When we scrutinize the magnitude and direction of the interaction by calculating the effect of survey participation after December 11th separately for men and women, an interesting pattern emerges. Shows that the increasing salience of the abortion issue in the media has a substantial and statistically significant impact on the likelihood of regarding abortion as a political issue, but this effect is observed only among men. Specifically, the estimated ITT for men is 0.12 (p-value = 0.033), whereas for women this effect is only 0.04 (three times smaller than for men), with a corresponding p-value of 0.669. However, we cannot assert that the ITT is different for men and women. This finding suggests that media effects vary across audiences and depend on the specific topic and its relevance to the audience's interests. To further illustrate these results, in Fig. A9 in the Appendix we summarize the predicted probability of considering abortion a political issue among men and women during the control and treatment periods. The results indicate that among men there was a significant change. While only 54% considered abortion a political issue in the control period, 66% did so in the treatment period. However, among women there was no noticeable change. The predicted probability of considering abortion a political issue among women changed from 59% in the control period to 63% during the treatment period.

Robustness checks

We perform several robustness tests to assess the plausibility of the identifying assumptions of our research design. These follow the recommendations of Muñoz et al. (2020) to add credibility to causal estimates of UESD studies. First, we assess the sensitivity of the results to the bandwidth of fieldwork days included in our analyses. While larger bandwidths could lead to greater imbalances in covariates between the control and treatment groups, minor changes in the bandwidth should not substantively alter our estimates. Therefore, we replicate our analyses using bandwidths of days that include ± 9, ± 10, ± 11, and ± 12 fieldwork days in the control and treatment groups. It is not feasible to extend beyond 12 days because fieldwork concluded on December 23. Furthermore, we do not include bandwidths smaller than ± 8 days because the power analyses (summarized in Fig. 2) indicate that such bandwidths would yield statistical tests with insufficient power.

The findings of these robustness checks, summarized in Fig. A6 in the Supplementary Information, reveal that the results are not very sensitive to the selected bandwidth. Using each of these bandwidths, we consistently conclude that the increasing prominence of abortion in the media has resulted in an approximately 10% raise in the proportion of citizens perceiving abortion as a political issue.

A key threat for the UESD is the presence of pre-existing time trends that may influence the dependent variable. If there were changes in Spaniards’ opinions on abortion before the issue gained prominence in the media, our results might not accurately capture this increase in salience, but rather reflect preexisting trend of change. To address this concern, we adopt a twofold strategy as recommended by Muñoz et al. (2020). This twofold strategy aims to rule out that our results are driven by a jumpy function, which could lead to the misidentification of effects. This robustness check, therefore, relies on the data collected just before the treatment of interest, which provides relevant information about the temporal stability of our variable during the period we analyze.

First, following the recommendations for regression discontinuity design (a similar identification strategy) we assess the effects of placebo treatment among respondents before December 11. In the absence of a pre-existing time trend, this placebo treatment should have no effect. To maximize the statistical power of this test, we estimate it at the empirical median of pre-December 11 respondents (Imbens and Lemieux 2008). Thus, we specify a placebo that treats as a control group those interviewed between November 22 and December 2 and as a treatment group those interviewed between December 3 and December 10. The results, summarized in Table A7 in the Supplementary Information, show that the effect of this placebo treatment is negligible and indistinguishable from zero. Second, in Fig. A8, we conduct a visual inspection of the evolution of our outcome variable before December 11. This, combined with the results of the placebo treatment, leads us to conclude that a pre-existing time trend is not driving our findings.

Another key threat to our research design would be the occurrence of other simultaneous events unrelated to the increase in media attention to the abortion issue. Any such event that could potentially influence opinions about abortion and occurred during the fielding of our survey could confound our estimates. A detailed analysis of the media environment during the studied period suggests that this is unlikely. The draft of the abortion law reform dominated the news during this short period, and we did not identify any other major news stories that might affect our outcome of interest. If our assumptions are correct, we should observe a change in citizens' opinions about whether abortion is a political issue during this period, but not on other issues. If opinions about issues other than abortion also changed during this period, this could indicate a violation of some of our assumptions. We therefore conducted a final robustness check, leveraging the fact that our survey includes 18 additional items that respondents had to classify as “political” or “nonpolitical” (see Table 1). We use these as placebo outcomes. We test whether increased media attention to the abortion issue has an impact on these other unrelated issues. The results, summarized in Fig. A7 in the Supplementary Information, reveal that our treatment (i.e., responding to the survey after December 11) has no effect on these placebo outcomes that is comparable in size to our estimate of the effect on considering abortion a political issue. Furthermore, three of the topics in the list can be considered intrinsically linked to women's interests, and therefore comparable to abortion in this regard: gender inequality in wages, the morning-after pill, and gender-based violence. As Table 1 suggest, on average, gender inequality in wages is already considered a political issue by 73% of total respondents. In contrast, for gender violence and the morning-after pill only a 52% and a 41%, respectively, considered them to be political (even a smaller percentage than in the case of abortion). Figure A7 suggests that there were no significant increases in the percentage of participants regarding gender inequality in wages, the morning-after pill, and gender-based violence as political topics for those answering our survey after December 11.

Only one of these placebo effects returns a statistically significant result (considering the price of drugs as a political issue). These findings bolster our confidence that we are capturing the effects of media coverage of abortion rather than the effects of another concurrent event.

Conclusions

Despite the broad and multidimensional character of the concept of politics, most public opinion studies tend to assume that citizens share a common understanding of its meaning. However, an emerging line of research delves into how politics is interpreted by citizens and which notions, experiences, and issues are considered political or not (Fitzgerald 2013; Gortz et al. 2022; Moray and Eveland 2016). Understanding individuals’ conceptions of politics is crucial for comprehending democratic citizenship, since conceptions of politics are consequential for political engagement (Gortz and Dahl 2021). However, to date this scattered literature has largely overlooked the role of the media in influencing individuals’ cognitive orientations regarding what is deemed political and what is not.

Our study offers a novel exploration of the role of the media in influencing the way people delineate the boundaries of the political sphere. Using a UESD design, we investigate whether the unexpected increase in the salience of abortion regulation in the media translated into a significant rise in citizens’ inclination to regard abortion as a political issue. This approach offers a real-world and, at the same time internally valid, empirical test.

Our findings illustrate the significant role of media in connecting citizens with politics. Exposure to authoritative arguments presented by politicians and commentators in the media, advocating for their viewpoints regarding the proposed reform of abortion regulation, results in a substantive (10% points) increase in the likelihood of citizens regarding abortion as a political topic. These effects underscore sizeable media effects, particularly within a case that has received limited attention: Spain. They also highlight the necessity for systematic exploration of this particular type of media effects. In the current media landscape characterized by increased selectivity and audience fragmentation, understanding traditional and social media’s capacity to shape people’s views and perceptions about the blurred boundaries of politics can yield significant behavioral consequences.

For example, the proposed reform of abortion law in Spain may have contributed to the heightened visibility and scale of feminist mobilization. Several studies have highlighted the large feminist demonstrations and intense lobbying activities triggered by this reform(Astudillo 2014). These events may serve as pivotal moments for opinion formation and activation regarding the public relevance of social and political norms related to women's reproductive decisions (Jimenez et al. 2022). Recent research (Anduiza and Rico 2022; Jimenez et al. 2022) indicates a significant increase in women's mobilization capacity in Spain since December 2013. This mobilization trend has continued to intensify over the past decade aligning with the global wave of feminist activism against sexual harassment, violence, and discrimination against women in general.

In conclusion, the systematic examination of what citizens perceive as political (and what they do not) can offer valuable insights into the factors driving citizens’ engagement and participation in specific political actions. Research indicates that the breadth of one’s conceptualization of politics (i.e., the number of issues considered political) is associated to individuals’ predispositions to engage in political conversations and discussions (Fitzgerald 2013) and even to forms of political participation linked to protest such as signing petitions, boycotting or buying products for political reasons, and specially participating in legal demonstrations (Görtz and Dahl 2021). In essence, cognitive orientations regarding the boundaries of the political world appear crucial in connecting citizens to “real politics” by influencing their political engagement.