“I Feel Your Pain”: The Effect of Displaying Empathy on Political Candidate Evaluation

Two experiments demonstrate that highly empathetic messages conveyed by a political candidate produce more favorable attitudes and increase the likelihood individuals will vote for the political candidate. Study 1 revealed this Empathetic Communication Effect is stronger among female political candidates than male. Compared to male candidates, female candidates are evaluated more positively when they engage in empathetic language but are more harshly penalized when they fail to display empathy. An analogous pattern emerged for candidate party in Study 2. Namely, the Empathetic Communication Effect is stronger among Democratic political candidates than Republican political candidates. Results also explore the impact of empathetic rhetoric on perceptions of candidates’ socio-emotionality and instrumentality.

understanding likely influences evaluations along the socio-emotional dimension. That is, highly empathetic messages from a candidate should indicate that candidate possesses socio-emotional traits like warmth, compassion, and sensitivity.
But to what extent do interpersonal traits like empathy impact a voter's overall evaluation of a political candidate? Work in social psychology finds warmth and empathy to be crucial in everyday interpersonal relationships (Brambilla et al., 2011;Fiske et al., 2007;Wojciszke, 2005). We like people who are interpersonally warm and emotionally understanding. Research in political science also finds interpersonally warm traits to be important when voters form impressions of political candidates, though the evidence is mixed. Some research suggests socio-emotional traits are not considered or weighted as heavily as instrumental traits like competence and effectiveness (Kinder, 1986;Kinder et al., 1980;McCurley & Mondak, 1995) while other work asserts that socio-emotional traits like empathy and warmth are just as important, if not more important (Clifford, 2018;Laustsen & Bor, 2017;Shogan, 2009).
If candidate evaluation is indeed influenced by the perceived warmth of the candidate, communicating and expressing empathy should be beneficial to those running for office. After all, politicians often must respond to upset voters or distressed constituents, and political experts recognize it is important for candidates to display warmth and compassion to voters on the campaign trail (Fenno, 1978;Gooch, 2018). Thus, being able to accurately surmise the emotional state of others and reflect emotional understanding back to them-in a genuine manner that does not appear cynical or manipulative-should be beneficial in a political context, improving a candidate's image and increasing the likelihood voters will vote for that candidate.

Empathetic Behaviors and Candidate Stereotypes
Candidate behavior is not viewed and evaluated by voters in isolation, however. Past research suggests behaviors performed by political candidates are often judged by how much they confirm or violate cultural stereotypes. For instance, the effects of engaging in negative campaigning are different for male candidates compared to female candidates (Fridkin et al., 2009;Krupnikov & Bauer, 2014). Likewise, voters hold stereotypic expectancies about the two major political parties, including assumptions regarding issue "ownership" (Norpoth & Buchanan, 1992;Petrocik et al., 2003;Pope & Woon, 2009) and assumptions regarding the personality traits of party members (Goren, 2007;Hayes, 2005;Winter, 2010). Such stereotypes involving candidate gender or candidate party may moderate the effect of empathetic communication on candidate evaluation.
Women running for political office often need to address stereotypic expectancies about what traits they do (or do not) possess and what issues they can (or cannot) handle as an elected official (Bauer, 2017;Bligh et al., 2012;Dittmar, 2015;Schneider & Bos, 2014). This may create a dilemma for female political candidates when it comes to displays of emotional empathy. On the one hand, the stereotypic traits associated with women (e.g., warmth, empathy) are not necessarily the key traits associated with political figures, which instead are more masculine or instrumental in nature (e.g., strength, decisiveness) (Best & Williams, 1990;Deaux & Lewis, 1984;Huddy & Terkildsen, 1993a, 1993b. As such, women may feel the need to downplay their stereotypical feminine traits (e.g., empathy) and emphasize more masculine traits to align themselves with the public's expectations of what a political leader is like (Bauer, 2017;Bligh et al., 2012). On the other hand, however, women may be especially vulnerable to attacks from political opponents who challenge them on stereotypically feminine traits (Cassese communication on candidate evaluation may differ when comparing female to male candidates. This will be explored in Study 1. Stereotypes about the candidate's party may also impact the effect of empathetic communication. Voters often hold stereotypes about the major political parties and their members (Goren, 2007;Hayes, 2005;Norpoth & Buchanan, 1992;Petrocik et al., 2003;Winter, 2010). Oftentimes, Democrats are presumed to have greater expertise in policy areas related to social welfare while Republicans are assumed to have more expertise in national security and business. Similarly, Democrats are stereotyped as possessing socio-emotional traits like empathy and compassion whereas Republicans are perceived as possessing more instrumental traits like strength and toughness. These stereotypic assumptions mirror, to a certain degree, gender stereotypes of men and women.
Consequently, showing empathy toward a voter should align with the Democratic Party stereotype but not the typical Republican. Thus, just as a candidate's gender may moderate the effect of candidate empathy on voters' evaluations, a candidate's party membership may also moderate the effect of candidate empathy on voters' evaluations of that candidate. Study 2 investigates this possibility.

Evaluations of Male and Female Political Candidates
The Empathetic Communication Hypothesis predicts a candidate will be judged more positively when the candidate displays empathy toward someone (in this case, a voter) in need or distress. However, this effect will be likely be more strongly pronounced among female candidates than male candidates. We label this prediction the Candidate Gender Role Hypothesis. Empathetic displays by a female (as opposed to male) candidate should elicit greater positive candidate evaluations (see Rudman & Glick, 2001, for evidence on gender role (non)conformity eliciting (un)favorable reactions from others). This is because empathy is more strongly associated with female gender norms than male gender norms (Best & Williams, 1990;Deaux & Lewis, 1984). Conversely, low empathy, which is normatively more masculine in nature (Best & Williams, 1990;Deaux & Lewis, 1984), should elicit a more negative reaction from a female (versus male) candidate because such behavior is incongruent with female gender norms.
The Candidate Gender Role Hypothesis predicts that empathetic displays will be perceived as less attractive coming from a male candidate because masculine gender norms dictate men should be less sensitive and emotionally supportive (Deaux & Lewis, 1984;Jansz, 2000). According to the Candidate-Recipient Gender Role Hypothesis, this may be especially true when a male candidate communicates with another man. That is, men might utilize messages low in emotional sensitivity and empathy, in part, to maintain a masculine gender-identity, especially when interacting with other men Jakupcak et al., 2003;Jansz, 2000). Accordingly, individuals may perceive a highly empathetic communication between two men as a more extreme violation of gender norms and therefore evaluate the empathetic male candidate more harshly.
Finally, it is predicted that empathetic rhetoric will positively influence not only the overall evaluation of and likelihood of voting for a candidate but also trait and issue ratings of that candidate. Candidates who display empathy should be seen as possessing more socio-emotional traits like warmth and sensitivity. Moreover, because certain policy areas like healthcare and welfare are perceived to involve more compassion and warmth (Huddy & Terkildsen, 1993a), such messages should also bolster candidate's perceived competency to handle these socio-emotional issues while serving in office.
The inverse is predicted for the impact of empathy on perceptions of a candidate's instrumentality. That is, empathetic displays should produce more favorable socio-emotional ratings but more unfavorable instrumentality ratings (e.g., strength, ability to handle military issues). Previous research demonstrates this can occur in some cases.
For example, individuals displaying high status and power are perceived as having more instrumental traits but fewer interpersonally warm socio-emotional traits, whereas individuals with low status and power are perceived as lower on instrumentality but higher on socio-emotionality (Gerber, 1996). Empathy should operate similarly.

Method
Participants were presented with verbal statements made by a male or female candidate to a male or female voter at a town-hall meeting. The candidate's statements conveyed either high or low empathy in response to the voter's problem. Empathetic Response (non-empathetic v. empathetic), Candidate Gender, and Voter Gender served as the manipulated independent variables. Effects on attitude toward the candidate, likelihood of voting for the candidate, trait ratings, and issue performance ratings were examined.

Participants
A power analysis via G*Power (Faul et al., 2007;Linear

Procedure and Design
Participants were brought into the lab in groups and randomly assigned to conditions. After consenting, participants read a transcript of a conversation that supposedly occurred between a state senator running for reelection in a neighboring state and a voter at a town-hall gathering where constituents can "meet-and-greet" candidates for office. In the conversation, the voter brings up a personal hardship (i.e., trouble affording college tuition) to which the candidate responds with low or high empathy. The town-hall voter's gender, the candidate's gender, and the content of the candidate's response were manipulated, creating a 2 (empathetic response: low vs. high) X 2 (candidate gender: male vs. female) X 2 (town-hall voter gender: male vs. female) between-subjects design.
After reading the transcript, participants exchanged it for a survey packet where they reported their attitudes toward the candidate, likelihood of voting for the candidate, impressions of the candidate's traits, and perceptions of how well the candidate could handle various issues. Upon completion, participants were debriefed and thanked for their participation.

Materials
Transcript -Participants received a one-page transcript of a conversation that purportedly took place between a state senator and a voter at town-hall meeting. To manipulate the candidate's gender within the transcript, the senator's name was varied across conditions ("Paul" versus "Paula" Johnson). The same was done to manipulate the town-hall voter's gender ("Christopher" versus "Christine" Smith). To avoid effects of political party, the transcript did not mention the candidate's party affiliation and purposefully avoided giving the candidate specific issue stances and policy positions. Thus, participants could not infer whether the candidate was a Democrat or Republican.
In the transcript, the voter at the town-hall expressed sadness and worry because he or she was having difficulty paying for college and might have to drop out due to the financial costs of attendance. The voter's portions of the conversation were identical in all conditions. However, the level of empathy in the candidate's response to this voter's problem was manipulated, modeled after stimulus materials from research on "person-centered" communication (Burleson, 1994;Holmstrom et al., 2005;Samter et al., 1987). Reflecting the key aspects of empathy, "person-centeredness" refers to the degree to which one is comforting, sensitive, and emotionally understanding when interacting with another person. A highly person-centered communication explicitly recognizes and legitimizes the feelings of another distressed person, reflects those feelings back to the person and elaborates on them, and then puts them into a broader context. A low person-centered communication ignores, denies, or challenges the emotions of the person in need, failing to recognize that person's perspective and perhaps even telling the person how he or she should be feeling (Burleson, 1994).
In the low empathy condition the candidate utilized low person-centered messages, remarking on the predictable rise of tuition for everyone and advising the town-hall voter to instead focus on feeling happy about the other great things in his or her life, thus failing to acknowledge the voter's upset feelings and instead telling the voter how to feel. In the high empathy condition the candidate used high person-centered messages: explicitly acknowledging the voter's feelings and reflecting them back, showing the candidate understood the voter's emotional state, and then suggesting a context by which the distressing situation might be viewed. In no condition did the candidate ever solve the voter's problem of financial hardship. Copies of research materials are available from the researchers upon request.
Manipulation check and pilot testing -In preliminary testing, the transcripts were rated by two independent coders trained to identify different empathetic, person-centered messages. Results from their ratings indicate the "high empathy" response by the candidate is indeed high in empathy while the "low empathy" response is low (κ = .98). Further pilot-testing was conducted with a sample of 19 participants who rated the transcripts on several 5-point scales to assess their believability and how empathetic and sensitive the candidate's response is. These individuals rated the high empathy communication as more sensitive (F = 10.52, p < .01), caring (F = 4.02, p = .04), and supportive (F = 5.09, p = .02) compared to the low empathy communication. Crucially, the high and low versions of the transcripts did not differ in terms of believability and overall realism. No differences were found when participants rated how realistic the conversations were (F = .39, p = .68), how authentic (F = .74, p = .49), believable (F = .64, p = .54), and how easy it was to imagine such a conversation actually taking place between a politician and voter (F = .10, p = .90). Thus, the high empathy condition was perceived as relatively authentic, believable, and genuine.
Participants' ratings of the candidate -After reading the transcript, participants completed a survey packet containing the dependent measures. Attitudes toward the candidate were assessed using four items. First, participants indicated how they felt about the candidate on a "feeling thermometer," selecting a number from 0 to 100 with higher numbers representing more positive feelings about the candidate. Participants also rated the candidate Displays of Empathy and Candidate Evaluation 772 on three 7-point semantic differential scales (-3 = extremely unfavorable to 3 = extremely favorable; -3 = extremely unlikeable to 3 = extremely likeable; -3 = negative to 3 = positive). Participants then indicated their likelihood of voting for the candidate, also on a 7-point scale (-3 extremely unlikely to 3 extremely likely).
Next participants reported how well specific traits described the candidate on 5-point scales (1 = Not at all, 5 = Extremely). Modeled after Huddy and Terkildsen (1993a), half of these items measured "socio-emotional" traits (e.g., compassionate, warm, gentle, empathetic) and half assessed "instrumental" traits (e.g., assertive, tough, resolute, aggressive). Participants then indicated how well they thought the candidate could handle certain issues (1 = Not at all, 5 = Extremely). Again modeled after Huddy and Terkildsen (1993a), roughly half of the items assessed performance ratings on "instrumental" issues (e.g., military spending, foreign crisis, budget deficits) and the remaining items assessed performance on "socio-emotional" issues (e.g., healthcare, assisting the poor).

Results
Feeling thermometer and semantic differential ratings were normalized and averaged to form a composite attitude toward the candidate score (α = .93). Likelihood of voting for the candidate was also normalized. The ten instrumental trait ratings were averaged and then normalized to arrive at a single measure of perceived trait instrumentality (α = .82). The nine socio-emotional items were averaged into a socio-emotional trait index which was then normalized (α = .87). Similar procedures were used to form dependent variable scores for perceived performance (issue competency) on instrumental issues (α = .76) and socio-emotional issues (α = .88). For all indices, higher numbers correspond to more favorable ratings.
Using hierarchical regression, each dependent variable was predicted using level of empathy in the candidate's response to the voter (-.5 = low empathy, +.5 = high empathy), candidate gender (-.5 = male, +.5 = female), and gender of the town-hall voter (-.5 = male, +.5 = female). Following Cohen and Cohen (1983), Step 1 in the regression tested all main effects, Step 2 tested all possible two-way interactions, and Step 3 tested the three-way interaction among all the predictors. Two control variables, the participant's own political ideology and party affiliation, were normalized as z-scores and entered into the regression models at Step 1.

Predicting Attitude Toward the Candidate and Voting Likelihood
Regression results are summarized in  Note. Entries are unstandardized regression coefficients with standard errors in parentheses. Reported main effects are from Step 1, two-way interactions from Step 2, and three-way interaction from Step 3 of the regression analyses. † p < .10. *p < .05. **p < .01. ***p < .001.
As predicted by the Candidate Gender Role Hypothesis, a significant interaction between empathetic response and the candidate's gender emerged at Step 2, B = 0.39, SE = .19, t(252) = 2.02, p = .04, η 2 = .02, .01 p . As seen in Figure 1, the effect of empathetic communication on attitude ratings was stronger for the female candidate than the male candidate. In the female candidate condition, the high empathy candidate (M = 0.48) was viewed much   The two-way interaction between empathetic communication and candidate gender on voting likelihood was marginally significant, B = 0.38, SE = .22, t(259) = 1.70, p = .09, η 2 = .01, .01 p . Again, the impact of empathetic communication was more robust for female candidates than male candidates (see Figure 1)

Predicting Performance Ratings: Socio-Emotional and Instrumental Issues
The last two columns of

Supplemental Results
Additional supplementary analyses were run as robustness checks, specifically, to explore whether participant ideology or participant gender potentially moderated the effects of candidate empathy. To examine the first possibility, participant political ideology as well as interactions of participant ideology with the other independent variables were included in an expanded regression model. This expanded model yielded no significant interactions involving participant ideology and candidate empathetic display (p > .21 in all cases), and no interactions involving participant ideology and any other independent variable (p > .14 in all cases).
To examine the second possibility involving participant gender, another regression model was run that included participant gender and interactions of participant gender with all independent variables. This too failed to consistently moderate the effects seen in Study 1. The gender of the participant exerted a modest main effect on most of the dependent variables except candidates' perceived ability to handle instrumental issues, with female participants generally providing lower candidate ratings than male participants (see Table S1 in the Supplementary Materials for all effects). Importantly, the two-way interaction between participant gender and candidate empathy failed to emerge when predicting all outcome variables. Indeed, with the exception of a lone three-way interaction between candidate empathy, candidate gender, and participant gender when predicting instrumental trait ratings (p < .01), all interactions involving participant gender and candidate empathy were nonsignificant (p > .10 for the remaining seventeen interactions tested).

Discussion
The results of Study 1 strongly support the Empathetic Communication Hypothesis. Highly empathetic candidates were preferred to candidates displaying low levels of empathy. This effect emerged when predicting attitudes toward the candidate was well as likelihood of voting for the candidate. These findings align with studies showing individuals who communicate empathetic or emotionally sensitive messages to others are viewed more favorably (Block-Lerner et al., 2007;Burleson & Samter, 1985;Samter et al., 1987;Sened et al., 2017). A main effect of empathetic communication was also found for ratings of the candidate's traits and issue competencies, however these effects depended upon the dimension of evaluation. High empathy led to increased perceptions that the candidate possessed socio-emotional traits like warmth and compassion and could handle socio-emotional issues like welfare. Interestingly, high empathy lowered ratings on instrumental traits like toughness and assertiveness and it decreased perceptions that the candidate could handle instrumental issues like military affairs.
In support of the Candidate Gender Role Hypothesis, a two-way interaction between empathetic response and candidate gender was obtained when predicting attitudes toward the candidate. As predicted, the Empathetic Communication Effect was magnified for female political candidates. Because social norms dictate that women should be emotionally responsive and sensitive, empathetic women were rewarded for conforming to gender norms whereas non-empathetic women were penalized for violating gender norms. This Empathetic Communication Effect was diminished among males, presumably because empathetic men violate gender norms that promote toughness and strength, and non-empathetic males conform to the norm. A similar (marginally significant) twoway interaction emerged when predicting voting likelihood. Against predictions, the gender of the town-hall voter did not seem to matter in these situations. The Candidate-Recipient Gender Role Hypothesis was not supported.

Evaluations of Democratic and Republican Candidates
The results of Study 1 suggest empathetic responses from candidates are especially valued when they conform to gender stereotypes. Hypotheses examined in Study 2 are rooted in research that indicates Republican and Democratic Party stereotypes elicit expectations that, to a large degree, mirror these gender stereotypes (Goren, 2007;Hayes, 2005;Norpoth & Buchanan, 1992;Petrocik et al., 2003;Winter, 2010). Study 2 proposes that analogous findings will emerge when considering the role of partisan stereotypes. That is, empathetic displays from political candidates will be especially valued when they conform to expectations that are activated by a politician's party membership, but they will be less strongly valued when they violate expectations of this nature.
In line with party stereotypes (Goren, 2007;Hayes, 2005;Norpoth & Buchanan, 1992;Petrocik et al., 2003;Winter, 2010), voters may expect Democrats to possess more interpersonally warm socio-emotional traits like empathy, compassion, and sensitivity and they may expect Republicans to possess more instrumental traits like strength and toughness.

Method Participants
A power analysis via G*Power (Faul et al., 2007;Linear  with the remaining reporting some other race).

Procedure and Design
Participants were randomly assigned to read one of the conversation transcripts between the state senator and town-hall voter used in Study 1. The level of empathy in the candidate's response to the voter's problem was again manipulated in the transcripts using materials from the person-centeredness literature. Instead of manipulating candidate gender, here the candidate's party was manipulated to be either Democratic or Republican. The candidate's gender was "female" in all conditions to simplify the experiment and focus more intently on the effects of candidate party. And because Study 1 found no effect of the town-hall voter's gender, it was also held constant at "male" in the transcripts. Thus, the design of Study 2 was a 2 (empathy: low vs. high) X 2 (candidate party: Republican vs. Democrat) between-subjects factorial. After reading the transcript, participants completed a survey packet where they reported their impressions and evaluations of the candidate. Once the packet was completed, participants were debriefed, thanked, and dismissed.

Materials
Transcript -The same conversation transcripts from Study 1 were used in Study 2. The town-hall voter in the transcript recounted a personal problem (concern about affording college tuition) to which the candidate responded with either low or high empathy, the first independent variable. The candidate's name ("Paula Johnson" in all conditions) was preceded throughout the transcript by either "Republican" or "Democrat," the other independent variable, so readers received continual reminders of her party affiliation.
Participants' ratings of the candidate -Measures of attitude toward the candidate, voting likelihood, perceived candidate traits, perceived performance ratings, participant party identification, and participant ideology were identical to those used in Study 1. Two exploratory variables were also measured: participants' political expertise and political interest. Finally, demographic items identical to those contained in Study 1 were administered.

Results
For the dependent measures, composite variables were calculated in the same way as in Study 1. An overall measure of attitude toward the candidate was created by normalizing and then averaging scores on the feeling thermometer and three semantic differential items (α = .91). Voting likelihood was again normalized and served as the sole measure of behavior intention. The average of the ten instrumental trait ratings and the average of the nine socio-emotional trait ratings were both normalized to obtain separate measures of the candidate's perceived instrumentality (α = .80) and socio-emotionality (α = .90), respectively. Analogous procedures were used to create composite variables for perceived ability to handle instrumental issues (α = .80) and socio-emotional issues (α = .90).
Hierarchical regression with dummy coding was used to predict the dependent variables based on the level of empathy in the candidate's response (-.5 = low empathy, +.5 = high empathy) and the candidate's party (-.5 = Republican, +.5 = Democrat) as well as the interaction between the two.
Step 1 of the regressions included all main effects and Step 2 tested the potential two-way interaction between candidate party and empathetic response.
In addition to participants' own political ideology and party affiliation being normalized as z-scores and entered into the regression as control variables at Step 1, an additional control was included at this step: the interaction between participants' party identification and the party affiliation of the candidate in the transcript. This interaction was included to control for the fact that Democratic participants would be more accepting of a Democratic candidate in the stimulus materials and less receptive to a Republican candidate (and vice versa for Republican participants).
The inclusion of this interaction as a control allows the effects of empathetic communication to be seen aboveand-beyond any effects of participant party.

Predicting Attitude Toward the Candidate and Voting Likelihood
Regression results are summarized in Table 2  Note. Entries are unstandardized regression coefficients with standard errors in parentheses. Reported main effects are from Step 1 and twoway interactions from Step 2 of the regression analyses. † p < .10. *p < .05. **p < .01. ***p < .001.
As predicted, the effect of empathetic communication on attitudes toward the candidate was more pronounced for the Democratic candidate than the Republican (see Figure 2). For Democratic candidates, the high empathy  Lastly, a main effect of empathetic communication was the only significant effect found for performance ratings on instrumental issues, B = -0.59, SE = .15, t(162) = -3.92, p < .001, η 2 = .08, .09 p . Again, displays of empathy by the candidate were negatively related to instrumental issue competency. Candidates using highly empathetic messages were assumed to be less adept at handling instrumental issues (M = -0.27), but candidates utilizing messages low in empathy were presumed to be more competent on instrumental issues (M = 0.32).

Supplemental Results
As was done in Study 1, additional analyses were conducted as a robustness check, again analyzing whether the observed effects could be potentially moderated by participant ideology and participant gender respectively. Results involving trait ratings and perceptions of issue performance also largely mirrored Study 1. Participants perceived the candidate as higher on warm socio-emotional traits in the high empathy conditions, and participants saw candidates as lacking those socio-emotional traits when they engaged in low empathy communication. Similarly, participants judged the high empathy candidate to be more adept at handling socio-emotional issues compared to the low empathy candidate. For candidates using messages low in empathy, they were perceived as possessing more instrumental traits and greater competency to handle instrumental issues.

General Discussion
By experimentally manipulating a candidate's rhetoric, the present studies offer a novel approach to examining the impact of empathy on candidate evaluation, allowing us to investigate the connections between how a candidate communicates, what traits the candidate is inferred to possess, and the perceptions of that candidate. Across two studies, we found that empathetic responses from a candidate significantly influence candidate evaluation, suggesting empathetic communication influences impressions within a political context much like it influences interpersonal impressions in non-political contexts (Block-Lerner et al., 2007;Burleson & Samter, 1985;Samter et al., 1987;Sened et al., 2017). Political candidates displaying high (versus low) empathy are evaluated more positively and more likely to attract votes.
Study 1 and 2 showed the effects of empathetic communication are not the same for all candidates, however.
Stereotypes around gender and party play a moderating role. Female candidates accrue larger benefits when using highly empathetic messages but also suffer steeper losses when responding to a voter with low empathy, presumably because the latter violates gender role expectations that stereotype women as warmer and more emotionally sensitive. It is more acceptable for men to convey low empathy messages. In essence, men are in a better position to "get away with" colder, low empathy rhetoric.
Empathetic communication also impacted ratings of the candidate's socio-emotionality and instrumentality, again with implications for gender. Highly empathetic rhetoric positively affects socio-emotionality (e.g., perceived warmth, ability to handle healthcare policy) but is detrimental in terms of instrumentality (e.g., toughness, ability to handle military affairs). This raises unique challenges for female candidates. Female candidates are presumed to have greater expertise regarding socio-emotional issues than male candidates (Huddy & Terkildsen, 1993a, 1993b but may have to "prove" they are competent, strong, or assertive leaders (Bauer, 2017;Rudman & Glick, 2001;Schneider & Bos, 2014). To convey strength and the ability to handle instrumental issues, a female politician might adopt a communication style lower in warmth and empathy. Unfortunately, this strategy may reduce her appeal when voters assess her socio-emotional characteristics (though for contrary evidence, see Leeper, 1991).
The trade-off between instrumentality and socio-emotionality also raises questions for a male political candidate who wants to support a constituent in distress, for this might undermine the perception that he is tough, resolute, and able to handle instrumental issues.
The candidate's party also moderated the effect of empathy on candidate ratings. Democrats receive more positive evaluations compared to Republicans when using the same highly empathetic messages, presumably because such sensitive, emotionally-understanding rhetoric is stereotypically more expected from Democrats (Cassese & Holman, 2018;Winter 2010). But, Democrats paid a bigger price for violating that stereotypic expectancy and failing to be highly empathetic. When using colder, low empathy messages, Democrats were evaluated more harshly than Republicans.

Future Directions and Conclusion
The present research demonstrates that highly empathetic messages from a political candidate positively affect impressions of that candidate. However, future studies should explore additional moderators of these effects, especially in light of the limitations of our sample. While supplementary analyses did not find evidence that participant ideology moderates perceptions of (non)empathetic candidates, there are still questions regarding the degree to which participant gender plays a moderating role. Past research suggests male and female individuals respond similarly to emotionally supportive messages in non-political contexts (Jones & Burleson, 2003;MacGeorge et al., 2004), but further investigation is needed in political contexts. Although the results of Study 1 are largely consistent with these findings, Study 2 suggested male and female voters may sometimes respond differently to a candidate's empathetic displays. Because the samples of both studies were overwhelmingly female, we have limited evidence to draw firm conclusions regarding the moderating role of voter gender. Additional research is needed to more systematically explore potential differences when comparing male and female voters' reactions to candidate empathy.
Further study is also needed to see if these effects extend to other issues. Study 2 showed Democratic candidates are viewed more unfavorably when they fail to speak empathetically about the costs of college tuition, the topic of discussion in the stimulus materials. It may be considered counter-stereotypical for a Democrat to be low in empathy and perhaps especially counter-stereotypical for a Democrat to be non-empathetic about higher education.
Future work can explore whether the effects seen in the present studies are stronger or weaker for other issues (e.g., failed business, unemployment).
Additional aspects of the candidate may also act as moderators. For instance, our results are presumably strongest for relatively unknown candidates. Genuine displays of empathy may convey that a lesser-known candidate is caring and understands the problems of everyday voters, which increases that candidate's favorability ratings and voters' intention of voting for that candidate. But empathetic displays may not meaningfully sway evaluations of politicians who are well-known or for whom empathy (or lack of empathy) is already expected by the public.
Finally, the level of office-holder-local, state, or national-may also be important. Previous research suggests stereotypically "masculine" traits benefit candidates running for national office more than local office (Huddy & Terkildsen, 1993b). Instrumental traits may be viewed as necessary for higher-level offices that require a firm response to national security concerns (president), whereas socio-emotional traits like empathy may be more desirable for local offices, especially those with more face-to-face contact with constituents (mayor). Consequently, the benefits of empathetic messaging may be greater for local politicians. We look forward to future research that investigates conditions that magnify and attenuate the effect of empathetic communication within politics.

Notes
i) Effect sizes are reported throughout. The first η 2 value reflects the eta-squared statistic (semi-partial correlation squared) and the η 2 value with a "p" subscript reflects the partial eta-squared statistic (partial correlation squared).

Funding
This research was made possible, in part, with the support from an Advanced Doctoral Fellowship grant generously provided through Loyola University Chicago.