Gender differences and similarities in news media effects on political candidate evaluations: a meta-analysis

How do different types of media coverage shape—and potentially bias—voter evaluations of women and men politicians? Theoretically reviewing 50 experimental studies and statistically synthesizing 671 evaluation outcomes from more than 23,000 participants, this meta-analytic review shows that gender bias in media-induced voter evaluations is conditional rather than universal. Our ﬁndings suggest that voters respond similarly to most media messages about women and men candidates. When gender-differentiated media effects are found, for instance, based on trait, appearance, or family coverage of politicians, this is mostly harmful for women candidates as it reafﬁrms gender stereotypical beliefs and lowers their viability ratings and vote preferences. Shedding light on the conditional nature of media-driven voter bias, this study adds to a better understanding of how the mediation of gender stereotypes sustains the underrepresentation of women in politics.

In nearly all countries around the world, women hold fewer elected political offices than men (Dahlerup, 2013).This structural and almost universal descriptive underrepresentation of women in politics entails an underrepresentation of the preferences and interests of women in political decisionmaking (Bratton & Ray, 2002;Celis, 2007) and reinforces the association between masculinity and political leadership roles (Koenig et al., 2011;Thomsen & King, 2020).An oftenheard claim in the scholarly and public debate is that political media coverage presents a major obstacle to women's electoral success, as it constitutes the main information source for voters' knowledge about politicians and their qualifications, personality, positions, deeds, and misdeeds (Balmas & Sheafer, 2013;Shehata & Stro ¨mba ¨ck, 2014).But how do the media shape-and potentially bias-voter evaluations of women and men politicians?What types of media coverage are detrimental to women's electoral chances, and which are not?
These fundamental questions at the intersection of gender, the media, and voter attitudes can be approached in two ways.First, the media representation approach construes bias as structural differences in political media reporting, with women politicians receiving more of certain (largely detrimental) types of media coverage but less of other (largely beneficial) types than men politicians.Another way in which the media contribute to the representational deficit of women in politics is by inducing gendered voter responses (e.g., Bauer, 2020a;Cassese & Holman, 2018;Hayes et al., 2014).This second media effects approach thus construes gender biases as effects of political coverage on voters' information processing, connecting cues in media coverage to literature on gender stereotyping and electoral behavior.
In this same journal, Van der Pas and Aaldering (2020) presented a comprehensive meta-analytic review of the media representation approach.However, a systematic overview of gendered media effects of coverage about women and men politicians is still missing.The media effects approach currently lacks a common framework for the underlying mechanisms of media-induced candidate evaluation-that is, how different types of media coverage conceptually and causally relate to evaluations of women and men politicians.Moreover, existing studies focus each on specific types of media coverage or voter evaluation outcomes, sometimes producing elusive if not conflicting empirical evidence (Bauer, 2015, p. 691;Hayes & Lawless, 2015, pp. 97-98).
This study takes up the cumulative endeavor and provides the missing second half of the overview of the field by conducting a meta-analytic review of the effects of media messages on gender-differentiated candidate evaluations.Theoretically reviewing 50 experimental studies and statistically synthesizing gender differences and similarities in over 671 media-induced evaluation outcomes of politicians from more than 23,000 participants, the main goal of this review is to assess which type of media coverage results in different (or similar) evaluations of men and women candidates for which type of evaluation outcomes.
In doing so, this study offers a holistic view of the multidisciplinary field and bridges several theoretical and empirical gaps in existing scholarly and public debates on gender stereotyping in politics.We first review the literature and reorganize different lines of theorizing about the intersection of gender, media, and politics into a single conceptual framework.Integrating concepts from role congruity theory, gendered mediation, and the literature on mediated leadership effects, we highlight the moderating role of media coverage in the candidate evaluation process.In this view, gender bias does not emerge in isolation but as a moderation effect: that is, media messages may activate gender role (in)congruent expectations which are then applied to evaluations of women politicians.By conceptually clarifying if and when media coverage should lead to which gender-differentiated evaluations, we gain a better understanding of how gender stereotypes sustain the underrepresentation of women in politics.
Exploiting the benefits of Bayesian meta-analytic techniques, we then empirically show that voters respond to media messages about men and women politicians in mostly similar-rather than different-ways.Quantifying, for the first time, the evidence of absence and presence of gender differences, we add to the growing body of research challenging the conventional wisdom of ubiquitous gender differences (Brooks, 2013;Dolan, 2014;Hayes & Lawless, 2015) and call for a narrative of conditional gender effects in political communication.However, results also reveal some critical exceptions.Notably, media coverage on candidate traits, appearance, and family, and coverage that is negative in tone disproportionally hurt women politicians, as it reaffirms existing gender stereotypical beliefs and lowers their viability assessments and vote preferences.On a positive note for women, coverage focusing on scandals and candidate attacks is less detrimental to their electoral support than for men.

Reviewing and reorganizing the gendered mediation of politics
The gendered mediation thesis (Gidengil & Everitt, 1999, 2003;Wagner & Everitt, 2019) serves as a-direct or indirect-conceptual point of departure for theoretical thinking about the intersection of gender, media, and politics.It posits that political media coverage follows an inherently "masculine narrative" that portrays politics in a "stereotypically masculine fashion" (Gidengil & Everitt, 1999, p. 50;Rakow & Kranich, 1991, p. 8).For example, with conflict as an important news value (Han & Federico, 2018), journalists routinely portray politics using references to warfare, violence, fighting, and sports.
The media representation approach outlined above, links gendered mediation to structural differences in media coverage along the lines of politicians' gender (e.g., Bystrom et al., 2001;Kahn, 1994;Meeks, 2012;Wagner et al., 2022).A recent meta-analysis of this field shows that on average women politicians receive not only less media attention than their male colleagues (albeit only in proportional and not in majoritarian electoral systems), but also that their media coverage is more strongly focused on their appearance, family, and gender, and that the media tend to associate both men and women politicians with personality traits and political issues consistent to their respective gender stereotype (Van der Pas & Aaldering, 2020).The mediated leadership effects literature suggests that media representation of politicians directly affects voters' evaluations and their electoral support (Aaldering, 2018;Mughan, 2000).Thus, genderdifferentiated political media coverage adversely affects women politicians because they receive less coverage overall and on less electorally relevant aspects.
However, the focus of this review is the second and arguably more insidious way that the gendered mediation of politics can sustain gender inequality, which is through gendered media effects on voters.Even if journalists cover men and women politicians similarly, the effects of that coverage can still vary by politician gender.Because political leadership itself is stereotypically linked to masculinity (Koenig et al., 2011;Schneider & Bos, 2014), then the journalistic application of these stereotypically masculine associations to women politicians indirectly biases voter evaluation by creating the impression that they are unfit for politics.
The literature offers various terms to capture this idea of a disconnect between (masculine) leadership stereotypes and the social roles of that members of gender groups are expected to fulfill, such as the double-bind or think-manager-thinkmale effect (see Brooks, 2013 for an overview).Role congruity theory (Eagly & Karau, 2002) predicts that the incongruity between leadership stereotypes and the feminine social role(s) puts women in political functions in a lose-lose situation: either they fail the demands of political leadership, or they breach gender norms related to femininity-with potential backlash due to stereotype violation in both cases.By rendering stereotype-incongruent attributes of politicians more (or less) salient, media coverage can thus reinforce (mitigate) the impact of stereotypes on subsequent evaluations (Bauer, 2015(Bauer, , 2017)).
Figure 1 provides an illustration of the concepts reviewed thus far.It causally connects the three intersected sides of the gendered mediation of politics: (a) women and men politicians, (b) the "mediated reality" in form of media coverage of women and men politicians, and (c) the evaluations of political candidates by voters.The dashed lines show the first way in which political media coverage may affect the electorate: Structural differences arise in media coverage from politicians' gender (arrow a) and these differences in representation then influence voters directly, as established by the literature on mediated leadership effects (arrow b).The combination of the two solid lines depicts the relations studied in this article and represents the second way of media-based gender bias.The horizontal arrow (c) is the "unmediated" main effect of politicians' gender on evaluations investigated in candidate choice experiments; the vertical arrow (d) captures the moderation effect of media coverage on voter evaluations of women and men politicians.It is the combination of these last two arrows (cd)-which we call media-moderated candidate evaluations-that we study in this meta-analysis.As this concerns a symmetric moderation, another way of thinking about this is as the effect of types of media coverage, moderated by gender of the politician, on evaluations by the public.

Media moderators of candidate evaluations
In the framework outlined above, a media moderator constitutes a specific aspect of media coverage that is part of the gendering of media coverage; they are categories that we know are relevant media representation outcomes of gendered mediation (for an overview see Van der Pas & Aaldering, 2020).While the theoretical underpinning may vary for each media moderator, the role congruity framework outlined above provides a common mechanism for their moderating effects: By emphasizing (de-emphasizing) gender-stereotypical attributes of the candidates in a media message, a moderator will strengthen (weaken) the extent to which a person receiving the message will perceive incongruent role expectations and thus result in more different (similar) candidate evaluations for men and women.The gender and media effects literature has discussed several such aspects (e.g., Bauer, 2020b;Cassese & Holman, 2018;Hayes et al., 2014), of which six have been most frequently linked to candidate evaluations: (a) tonality of the coverage, (b) trait, (c) issue, and (d) appearance coverage, and reporting on politicians' (e) personal life, or (f) behavior.
First, valence frames in media messages anchor the general tonality (favorable vs. unfavorable) of the subsequent evaluation process by providing a particular valence framing (De Vreese & Boomgaarden, 2003;Lodge & Taber, 2013).Negative media content has been shown to not only elicit stronger attitudinal reactions than positive or neutral news (Soroka & McAdams, 2015) but may also disproportionately hurt women politicians by reinforcing incongruent affective expectations (Bakker et al., 2021;Rohrbach, 2022).Positive tonality, however, may override this incongruence and preempt potential backlash against women (Ditonto, 2017).
Second, coverage of politicians' personality traits can be an effective means for activating gender stereotypes, which are then applied to portrayed politicians during the evaluation (Bauer, 2015).In line with role congruity theory, the association of women politicians with communal traits (caring, likeable, cooperative, etc.) violates leadership stereotypes while with linking them to agentic traits (strong, competitive, ambitious, etc.) conflicts with gender norms (Schneider & Bos, 2019).
Third and similarly, reporting on political issues can be seen as moderating candidate evaluations by modulating the salience of gender norms.For example, emphasizing economic crises (Lei & Bodenhausen, 2018) or threat of terrorism (Holman et al., 2016(Holman et al., , 2017))-two issues typically cast as masculine-can lead to disproportionately more negative evaluations for women candidates (Huddy & Terkildsen, 1993).Unlike communal traits, coverage of feminine issues (welfare, healthcare, education, etc.) has been suggested to be less effective at activating feminine stereotypes and may thus entail fewer gendered evaluation outcomes (Bauer, 2020b).
Fourth, media attention to candidates' appearance renders heuristic gender cues more salient which "exaggerate perceptions of gender-stereotyped attributes" (Sigelman et al., 1986, p. 231).Appearance-focused coverage may also hurt women because it diverts attention from their qualification and issue positions, which they struggle to establish in the first place (Hayes et al., 2014;Ross, 2017).
Fifth, media attention to candidates' personal life "politicizes gender identities and roles" (Trimble et al., 2021, p. 166) by inciting voters to factor in personal background, problems, civil status, parenthood, or sexuality in their appraisal of politicians (Balmas & Sheafer, 2013).As a way of imprinting the traditional gender dualism between a stereotypically feminine private realm and a masculine political realm, coverage of the personal can thus compound the expectation of double standards for women politicians (Stalsburg, 2010).
Sixth, gendered evaluations can result from coverage of a range of candidate behaviors.Potentially gendered behaviors include media accounts of candidates' involvement in political scandals (Cucchi & Cavazza, 2021;Courtemanche & Connor Green, 2020) or campaign attacks (Cassese & Holman, 2018;Gordon et al., 2003), display of emotions (Brooks, 2013), and even specific gestures (Everitt et al., 2019).The different behaviors all share stereotypical assumptions about the (in-)appropriateness of their display for political leaders or women and men.For instance, voters might punish women portrayed as emotive either for breaking with the prescriptive expectation of the levelheaded political leader or for breaking with the gender stereotype of emotional and caring women (Boussalis et al., 2021;Brescoll, 2016).Likewise, because women are cast as more honest and moral, they may face more severe consequences when they are portrayed as engaging in immoral or aggressive behaviors (Eagly & Karau, 2002).

Gender and evaluation outcomes
There is no single relevant outcome measure on which voters evaluate women and men politicians; instead, existing research has measured gendered candidate evaluations across a plethora of evaluation outcomes.In her review of gender stereotype measures in political contexts, Bauer (2013)   Evaluations of personality traits reflect voters' beliefs about the degree to which politicians (ought to) possess a set of character qualities.Inferences about candidates' traits are an integral part of political and social information processing (Funk, 1999).Though varying in their labels and operationalizations, the methodological and theoretical choice of measured candidate traits is usually guided by genderstereotypical expectations distinguishing between the two broad categories of communality and agency (Schneider & Bos, 2014).Men and political leaders are generally perceived as high in agency and low in communality while the opposite pattern applies to women.
Evaluations of issue competencies measure voters' perceived skill of politicians to handle specific domains of political life.Perceived issue competencies align with gender stereotypes in that women politicians are cast as more competent on the so-called feminine issues (e.g., health, education, welfare, women's issues) and men politicians are seen as more skilled on masculine issues (e.g., economy, military, defense, finance) (Huddy & Terkildsen, 1993).Crucially, the masculine norms embedded in political structures imbue masculine issues with higher priority and relevance than feminine issues (Schneider & Bos, 2019).
Measures of vote preferences range from hypothetical candidate support or rankings (Bauer, 2020b), vote likelihood (Lizotte & Meggers-Wright, 2019) to favorability or feeling thermometer scores (Holman et al., 2017).A final group of outcome includes evaluations of candidates' viability.Assessments of viability typically relate to the quality of candidates' political profile, such as their ability to win elections, overall qualification for a position, or their perceived effectiveness in elected office (e.g., Bauer, 2020a).The differentiation between these groups of evaluation outcomes is important as it allows for a more nuanced view on the impact of media moderators on gendered candidate evaluations.

Study selection
The following eligibility criteria were used to select studies: 1) The research must focus on candidates in a political context.This includes evaluations of real or fictitious elected officials or candidates running for an institutional office but excludes candidates for positions related to professional or leisure activities.
2) The research design must be an experiment containing at least (a) a (binary) candidate gender comparison (woman vs. man) and (b) an aspect of media coverage (i.e., a media moderator; see subsection above).We focus on experiments because our goal is to synthesize causal relationships while controlling for as much contextual influences as possible.
3) The manipulated stimulus material must consist of an aspect of media coverage that reflects a journalistic decision, which shows a moderating influence of the mediation process.This notably excludes forms of strategic communication and self-presentation (political advertisements, flyers, press releases, and candidates' own social media posts, etc.) as well as candidates' biological features (candidates' facial structure, skin complexion, etc.), as these aspects are beyond journalists' influence.
4) The data analysis must contain a candidate evaluation as an outcome measure, that is a participant rating of characteristics relating to the politicians' personality, competence, qualifications, etc. (see subsection above).
The systematic search was conducted based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Moher et al., 2009) The second round repeated the screening for all references cited in relevant records from the first round, together with serendipitously identified records via suggestions from other scholars (after public outreach on Twitter and mailing lists) or publication alerts (a ¼ 0.92).Finally, the same authors read the full text of all potentially relevant records using the complete set of eligibility criteria (a ¼ 0.93 for the decision to include).This procedure yielded a total of 39 publications containing 50 studies that were included in the meta-analytic review (see Supplementary Materials Section A2 for an overview of included studies).

Coding
Each included study was independently coded by at least two authors on three hierarchical levels-namely, (a) the level of the individual publication, containing (b) one or several experimental studies that (c) report each multiple effect sizes (see Supplementary Materials Section D for the full codebook).On the publication level, we coded the gender and number of author(s), year, title, and publication type and outlet.On the study level, coding included information relating to the experimental design (factors and levels), stimulus material, and the sample.
On the effect size level, to account for the moderating effect of media coverage and to capture gender differences across a range of evaluation outcomes, the single-gender difference in one outcome measure per experimental media condition represents the smallest unit of coding.For each effect size, we first coded the experimental condition of the moderating aspect of media coverage (e.g., mention or not of candidate attractiveness in a 2 Â 2 design with candidate gender being the other factor) in the most fine-grained manner possible, including three-way interactions.Second, we assessed specifics of the outcome measure (description, scaling, item composition, and measure direction).As outlined in the theory section, we recoded the individual experimental conditions into six media moderator groups and the measures into four outcome groups (see Supplementary Materials Sections A3.1-A3.2 for the detailed category scheme).Finally, we extracted means, standard deviations, and group sizes for women and men candidates to calculate standardized mean differences, as

Data analysis: Bayesian meta-analysis
We estimated the meta-analytic summary effects using Bayesian random-effects models (Ro ¨ver, 2020;Ro ¨ver et al., 2021).The Bayesian approach provides several advantages for meta-analysis over traditional null hypothesis statistical testing (NHST) approaches.First, the Bayesian models circumvent some of the common statistical issues involved in estimating summary effects and heterogeneity of effect sizes (Williams et al., 2018), even when the studies included in the synthesis are few and their observed effect sizes are small (Friede et al., 2017).Second, Bayesian inference is more intuitive as conclusions from posterior distributions of estimated parameters allow for direct statements about the likelihood of the observed data (Gelman et al., 2013).Third, unlike traditional NHST, Bayesian analysis can both quantify the evidence provided by the data against and for the null hypothesis.The relative evidence is indicated with the Bayes factor (BF), which describes two models' predictive performance in relation to each other; BF 01 is calculated as the ratio of the likelihood of H 0 (absence of gender difference) over the likelihood of H 1 (presence of gender difference)-given the data (Keysers et al., 2020;Wagenmakers et al., 2018).This is a crucial aspect for this meta-analysis as it enables the synthesis of gender similarities in the field, which have so far been neglected with studies overwhelmingly focusing on gender differences.
Our statistical approach consists of three steps.We first pool all 671 coded effect sizes in a single model to estimate the overall gender difference as Hedges' g across all experimental conditions and across all evaluation outcomes. 1Second-and this is the focus of this article-we run a series of subgroup analyses for each group of media moderator and evaluation outcome to disentangle the conditional effects of gender.For ease of interpretation, we report the results of our subgroup analyses in a series of figures (see Figures 2-5).The figures show estimated summary effects broken down for individual media moderators (rows) and different evaluation outcomes (columns).The reported effects reflect the combined influence of gender and the media moderator, where each summary effect is the standardized gender difference in voters' evaluation of politicians for the given moderator and outcome group.For each effect size in the text, we also report its associated Bayes factor (BF 01 ), which represents the relative strength of evidence from the data about the absence of a gender difference over its presence. 2Third, we conduct a series of sensitivity analyses to investigate the robustness of our results to a range of alternative model specifications and to the inclusion of covariates in our models.Due to limited space, please refer to the Supplementary Materials file for additional explanations of the used models and prior distributions (section A3.4), a detailed account of all sensitivity analyses (C1), and additional analyses relating to the risk of biases in metaanalysis (C2).

Tonality of coverage
We first turn to the combined effects of gender and the tonality of media coverage on voters' evaluation of politicians, which are depicted in Figure 2. First, coverage with mixed tonality-that is, combining both positive and negative elements-does not disproportionately affect voters' evaluations of women and men candidates' perceived agency (black box in figure, g ¼ À0.27, CrI ¼ À0.97 to 0.44, BF 01 ¼ 1.60) or communality (white box in figure, g ¼ À0.17, CrI ¼ À0.81 to Note.In this and following figures, the plots present summary effects as standardized mean differences (Hedges' g) along with 95% credible intervals (CrI) and the number of synthesized effects (k).
0.46, BF 01 ¼ 2.28).Moreover, we find substantial evidence that positively toned coverage does not influences candidate trait evaluations in a gendered way.For example, voters are 10.13 (¼ BF 01 ) more likely to not discriminate between women and men candidates in terms of perceived agency (g ¼ À0.02, CrI ¼ À0.19 to 0.17) when they are both positively portrayed in the media.Additionally, positive media coverage equally affects men and women politician's perceived masculine and feminine issue competencies and vote preferences.
However, we do find some evidence for genderdifferentiated effects of negative coverage.On the one hand, women candidates who receive negative media coverage are perceived as more communal than men candidates receiving the same coverage (g ¼ 0.34, CrI ¼ À0.01 to 0.71, BF 01 ¼ 0.47).The Bayes factor of 0.47 indicates that this gender difference is about twice as likely to be present than absent.A gender effect of negative media coverage on agentic trait evaluations of candidates is not found (g ¼ 0.11; CrI ¼ À0.27 to 0.48; BF 01 ¼ 3.58).On the other hand, there is substantial evidence (BF 01 ¼ 0.12) that voters are considerably less likely to vote for women rather than men candidates as a result of negative media coverage (g ¼ À1.01, CrI ¼ À1.77 to À0.43).The presence of this gender effect on vote preferences is about 8 times more likely than its absence. 3

Trait and issue coverage
Figure 3 shows the summary effects for the moderating influence of different types of trait and issue coverage.First, the media coverage on candidates' communal or agentic traits (upper half of the figure) shows a gender-differentiated effect on voter evaluations, but only in some cases.The evidence shows that the effects of trait coverage in media messages on candidates' perceived communality are stronger for women candidates than for men candidates: the presence of this gender effect is about twice as likely than its absence for communal trait coverage (g ¼ 0.18, CrI ¼ 0.04 to 0.33, BF 01 ¼ 0.51), and about seven times as likely for agentic trait coverage (g ¼ 0.24, CrI ¼ À0.09 to 0.40, BF 01 ¼ 0.14).This means that trait coverage, irrespective of the kind of trait coverage, more strongly affects the communality assessments of women.The impact of communal trait coverage on politicians' perceived agency, on the other hand, is weaker for women than for men candidates (g ¼ À0.16; CrI ¼ À0.25 to À0.07, BF 01 ¼ 0.10, presence of this gender effect is 10 times as likely than its absence), while there is strong evidence for the absence of a gender effect for agentic trait coverage (g ¼ À0.03, CrI ¼ À0.13 to 0.07, BF 01 ¼ 13.18).
Furthermore, no type of trait coverage seems to induce a gender-differentiated effect in evaluations of candidates' perceived issue competencies or lead to gendered vote preferences.However, we find very strong evidence (BF 01 ¼ 0.006) that communal trait coverage disproportionately diminishes assessments of women candidates' viability (g ¼ À0.24, CrI ¼ À0.33 to À0.14, the presence of this gender effect is about 167 times more likely than its absence).In summary, trait coverage induces gender-differentiated effects for evaluations of politicians' traits and viability to the detriment of women candidates, but similar effects for men and women candidates are found for assessments of their issue competencies and of vote preferences.
Additionally, the results imply that there is no genderdifferentiated effect of the impact of masculine issue coverage on candidates' viability assessment (g ¼ À0.03; CrI ¼ À0.74 to 0.07; BF 01 ¼ 2.96), while strong evidence is shown for the absence of a gender effect in the influence of masculine issue coverage on vote preference (g ¼ 0.02; CrI ¼ À0.11 to 0.17; BF 01 ¼ 10.87).Although slightly more ambiguous than for masculine issue coverage, the findings also indicate the absence of gender-differentiated effects on voter evaluations based on feminine issue coverage.Thus, coverage mostly affects evaluations of men and women candidates in similar ways.

Coverage of candidate appearance and personal life
Figure 4 displays the summary effects for the moderating influence of media coverage based on candidates' physical appearance and personal life.First, the upper half of the figure shows that media discussing a candidate's appearance does not result in voter's assessment of issues competencies, traits, or viability differently for men or for women politicians.However, we do find moderating effects on vote preferences: Media coverage in which a candidate's appearance is mentioned without a valenced overtone (g ¼ À0.58, CrI ¼ À0.90 to À0.26; BF 01 ¼ 0.40), or is described positively (g ¼ À0.37, CrI ¼ À0.63 to À0.10; BF 01 ¼ 0.08), negatively affects vote preferences for women candidates compared to their men colleagues.The presence of the gender effect to the detriment of women's electoral changes is 2.5 times more likely than its absence for neutral appearance coverage, and 12.5 times more likely for positive appearance coverage.By contrast, negative appearance coverage (g ¼ 0.27, CrI ¼ À0.18 to 0.72, BF 01 ¼ 1.60) and media messages that describe candidates as attractive (g ¼ À0.26, CrI ¼ À0.97 to 0.45, BF 01 ¼ 1.55) do not seem to result in gender-differentiated vote preferences.Thus, although in most cases appearance coverage is equally beneficial or harmful for men and women candidates, discussing politicians' looks can hurt women disproportionality under certain circumstances.Second, the lower half of Figure 4 shows the moderating influence of personal media coverage on gendered evaluations of politicians.Family coverage seems to impact women candidates' evaluations differently than that of their men colleagues: Media messages about candidates' children result in stronger feminine issue competency evaluations for women candidates than for men candidates (g ¼ 0.25, CrI ¼ 0.02 to 0.47, BF 01 ¼ 0.46).A gender difference is not found for masculine issue competencies (g ¼ 0.11; CrI ¼ À0.10 to 0.31, BF 01 ¼ 3.38).Moreover, this type of family coverage disproportionally lowers women candidates' viability evaluations (g ¼ À0.44, CrI ¼ À0.96 to 0.07, BF 01 ¼ 0.48).The presence of both gender effects is about twice as likely than their absence.Furthermore, the results suggest that media attention to candidates' personal problems or experience influences evaluations of women and men politicians similarly.For both media moderators, no gender differences emerge in all but one instance: women are perceived as more communal than men when media coverage zooms in on politicians' prior experience (g ¼ 0.17, CrI ¼ 0.05 to 0.28, BF 01 ¼ 0.32, the presence of this gender effect is about 3 times as likely as its absence).
The presence of this effect is 100 times more likely than its absence.
Other noteworthy findings are that scandalization coverage and media accounts of politicians' attacking behavior are less detrimental to the electoral support of women candidates than men candidates (scandalization: g ¼ 0.23, CrI ¼ 0.13 to 0.34, BF 01 ¼ 0.02, the gender effect being about 50 times more likely than its absence; candidate attack: g ¼ 0.67, CrI ¼ 0.07 to 1.28, BF 01 ¼ 0.47, the effect being about twice as likely than its absence).

Discussion
We draw four key insights from these findings.First, the evidence from the subgroup analyses tends to favor the absence rather than the presence of gender differences in mediainduced voter evaluations.For instance, we find no indication in favor (and even strong evidence against) genderdifferentiated effects on candidate evaluations of mixed or positive tonality coverage, media messages focusing on politicians' gestures and personal problems, and all types of issue coverage.Moreover, if a media moderator is shown to result in gender-differentiated effects, this is usually for specific rather than a wide range of evaluation outcomes.This conclusion that media coverage of politicians tends to affect the evaluation of men and women candidates similarly complements the evidence from observational research (Bridgewater & Nagel, 2020;Dolan, 2014;Van der Pas, Aaldering & Steenvoorden, 2022) and candidate-choice experiments (Schwarz & Coppock, 2022), suggesting that, all things equal, voters display little overt bias against women politicians.
Second, certain types of media coverage strengthen stereotypical beliefs and result in a stronger effect on feminine issue 108 Meta-analysis of gendered candidate evaluations competency or communal trait evaluations, or a weaker effect on masculine issues and agentic trait evaluations for women than for men candidates.These findings are most apparent following media coverage of candidates' traits, but are also shown for negative tonality media coverage, and media messages focusing on a candidate's past experiences, family, and emotionality.In line with role congruity theory, these genderdifferentiated media effects reaffirm gender stereotypical role expectations (e.g., Eagly et al., 2020) and reinforce the link between women and communality.Perceptions of increased communality are likely to be an electoral liability rather than an asset, as they are at odds with the leadership stereotype and the (agentic) qualities usually demanded of politicians (e.g., Eagly and Karau, 2002;Koenig et al., 2011).Thus, emphasizing women politicians' communality may prompt voters to appraise them primarily in their role as women instead of political leaders (Bauer, 2020a).By the same logic, the rather striking absence of genderdifferentiated effects of issue coverage corroborates the observation of past studies that issues are less likely to activate gender stereotypes than traits (Bauer, 2020b).On the one hand, issue coverage may divert voter attention to the political rather than the personal dimension of politicians and therefore be an effective means of reducing perceived incongruence between gender and leadership roles.This may imply that women risk little by potentially running on the "wrong" issues.On the other hand, however, issue coverage may still consolidate the link between masculinity and political leadership (e.g., Koenig et al., 2011;Schneider & Bos, 2014) and thus indirectly feed the notion of politics as a masculine domain (e.g., Van der Pas & Aaldering, 2020;Gidengil & Everitt, 1999, 2003;Wagner & Everitt, 2019).
Third, specific types of media coverage directly hurt women candidates' electoral changes compared to those of their men colleagues.The results show that negative tonality in media messages about politicians, and discussions of their appearance and their families trigger gender-differentiated responses that hurt women in terms of lower viability and vote preferences.Further supporting role congruity theory (e.g., Eagly & Karau, 2002), coverage of communal traits is detrimental for women candidates' viability assessment compared to their men colleagues, as it renders the stereotypical incongruity of women and leadership roles more salient.These mediainduced impressions of less viable women politicians present a particularly pernicious barrier to women's political leadership, as voters may pragmatically withhold support if they perceive women's prospects of winning an election to be too uncertain (Bateson, 2020;Corbett et al., 2022).These forms of media coverage can therefore directly contribute to the ongoing underrepresentation of women in political functions.
Fourth, the results also offer more positive news for women candidates, as the results indicate that voters tend to punish women less than men if their involvement in scandals or campaign attacks are discussed in the media.One explanation of this beneficial moderation effect for women politicians could be the role of voter motivations.Media content that is too blatantly gender-stereotypical or even sexist may motivate voters to overcompensate the reliance on stereotypes by consciously reversing them in women's favor (Rohrbach, 2022).Another possibility is that media negativity directed toward women politicians elicits forms of benevolent sexism (Glick & Fiske, 1996), as voters feel compelled to protect women from their exposed vulnerability (Barnes et al., 2020).

Conclusion
Does coverage of political news affect evaluations of women and men politicians differently?If so, which types of media coverage result in gender bias?This systematic review of 50 studies containing 671 evaluations from over 23,000 voters shows that similar responses to mediated candidate messages are the norm and gender differences are the exception.
One implication of the overall tendency of voters to equally respond to media messages discussing men and women candidates may be a shift in the field's narrative traditionally focused on gender differences toward a narrative of gender similarities (for a similar narrative shift in psychology see Hyde, 2014).Assuming gender similarities as a future default mode of research on gender and political communication could prove a viable strategy to attenuate the negative impact of claims of gender differences.The current scholarly focus on gender differences may implicitly reinforce or even validate gender stereotypes-and thus uphold the notion of politics as a gendered sphere-by, for example, recommending women to run "as a woman" or "as a man" (Herrnson et al., 2003).Moreover, a gender similarities approach does not collapse the varying experiences of women and men politicians but rather highlights their conditionality by drawing attention to the "importance of context in creating and erasing gender differences" (Hyde, 2014, p. 392; see also Rohrbach et al., 2020).Crucially, such gender narratives matter beyond academia (Rakow, 1986), as the public framing of women's relationship with the media and voters has implications for women's political ambition, fundraising and mobilization capabilities, and electoral support (Brooks & Hayes, 2019;Haraldsson & Wa ¨ngnerud, 2019;Fox & Lawless, 2014).
The most important exceptions to the pattern of gender similarities in voters' candidate evaluations are predominantly detrimental to women politicians, most notably following coverage of politicians' traits, appearance, and family.Thereby, media coverage of politicians is disadvantageous for the electoral chances of women candidates and directly (through viability assessments and vote preferences) and indirectly (through reaffirming gender stereotypes and the masculinity of the political domain) sustains the underrepresentation of women in politics.Moreover, this gender bias in the effects of media coverage combines with biases in media representation, where research shows that women candidates are substantially more often discussed in relation to their appearance and their families and in accordance with their gender stereotypical traits (Van der Pas & Aaldering, 2020).Therefore, women on the campaign trail may still face a double loss: they receive more non-political and detrimental media coverage, which hurts their candidacy in the first place, but this coverage also disproportionately diminishes their evaluations by voters thus further lowering their chances at the ballot box.
This meta-analytic review of the field comes with several caveats.Although our meta-analysis showed that certain types of media coverage can moderate voter evaluations, it cannot establish a casual mechanism.Although initial work has proposed stereotype activation as an underlying driver of gendered candidate evaluations (Bauer, 2015), there is still little theoretical and empirical knowledge of how media coverage influences the stereotyping process.Although the synthesis of the field's experimental results echoes some key findings from observational research, it remains unclear to what extent our findings hold in real-life electoral campaigns.There is still need for future experimental research that, for example, connects media effects to observational evidence from content analyses or voters' media use.Furthermore, our analysis does not account for intersectional aspects and individual differences in voter characteristics.On top of potentially activating gender stereotypes, media coverage can also trigger partisan or racial stereotypes (e.g., Philpot & Walton, 2007;Sanbonmatsu & Dolan, 2009).However, the circumstances in which different types of media coverage result not only in gender differentiated but also in intersectional biases in voter evaluations remains an underexplored topic.Likewise, future studies should examine the conditioning impact of voters' own gender, ethnicity, or partisan identity on the extent of gender bias by modulating motivations to resist stereotype reliance.Lastly, the majority of included studies examine the US context, suggesting a strong American-centric focus of the field.Future research can strengthen crosscultural generalizability by studying cases besides the US or by adopting a comparative approach.
In conclusion, this study reinforces the notion that gender bias in voter evaluations is conditional rather than universal.In trying to understand this conditionality, this study has highlighted the role of media coverage in moderating voter bias in evaluations of women and men politicians.With a conceptual and statistical review of the field, this study offers insight into the complex ways in which media coverage can both result in little gendered consequences or significantly exacerbate or mitigate the impact of gender.By shedding light on the conditional nature of media-driven voter bias, we add to a better understanding of when and how gender stereotypes sustain the underrepresentation of women in politics.
distinguishes three groups of evaluation outcomes based on what aspect of the politician is being evaluated.

Figure 2 .
Figure 2. Moderating influence of tonality in media coverage on gendered evaluations of political candidates.

Figure 3 .
Figure 3. Moderating influence of trait and issue coverage on gendered evaluations of political candidates.

Figure 4 .
Figure 4. Moderating influence of personalization and appearance coverage on gendered evaluations of political candidates.

Figure 5 .
Figure 5. Moderating influence of coverage of candidate behavior on gendered evaluations of political candidates.
, as depicted in FigureA1in Supplementary Materials.The search strategy consisted of an initial database search string containing combined terms for the concepts politician, evaluation, gender, and media coverage (see Supplementary Materials Section A1 for full strings).All results from Web of Science, Scopus, ProQuest IBSS, ProQuest Political World abstracts, EBSCO communication and sociological abstracts, and Ovid Psychinfo were downloaded.To reduce publication bias, three complementary databases (Worldcat, opengrey.eu,and DOAJ) were searched specifically for unpublished, open access, grey literature, or book publications.In a first round of abstract screening, three authors used Criteria 1, 2a, and 4 to assess the potential relevance of the 1,107 retrieved records (intercoder reliability Krippendorff's a ¼ 0.85).
(Borenstein & Hedges, 2019)d candidate evaluations indicated by Hedges' g(Borenstein & Hedges, 2019); positive values indicate a difference in evaluation outcome favoring women politicians and negative values favoring men politicians (see Supplementary Materials Section A3.3).The coding was conducted in several rounds allowing for discussion between all authors to resolve any disagreement.