The Influence of Identity Salience on Framing Effectiveness: An Experiment

Efforts to influence attitudes on highly polarizing issues, such as climate change, often fail because individuals interpret political messages through the lens of their partisan identities. However, shifting the identity lens through which an individual interprets a message may result in more effective political communication. Through a pre-registered survey experiment (n=978), this study tested how priming a partisan or a non-partisan (parental) identity influenced the effectiveness of a climate change frame on several attitudinal outcomes. Findings suggest that identity salience – specifically partisan identity salience – can influence the effectiveness of a frame. Among Republican parents, receiving a message about the impact of climate change on future generations increased climate change concern and intended pro-climate political behaviors, but this framing effect disappeared when a partisan identity was first primed. Among Democrat parents, framing had no significant effect until a partisan identity was first primed. The findings offer important insight into the role that identity salience plays in framing effectiveness and suggest that political communication on polarized issues is likely to be more effective building bipartisan agreement when non-partisan identities are salient.


Introduction
Today, partisan polarization is the norm rather than the exception on numerous political issues. Data from the Pew Research Center shows that the median Democrat and the median Republican were farther apart on the ideological spectrum in 2017 than at any point in the prior 30 years (Pew Research Center, 2017). This partisan polarization has consequences for effective democratic governance, as polarization on an issue can lead to policy gridlock. At an individual level, polarization leads voters to rely more on their partisan identity when making political decisions, leading to less critical evaluation of policy information and facts (Druckman, 2001;Guilbeault et al., 2018;Kahan et al., 2012; and a reduction in political engagement (Hetherington, 2008).
Recent polling shows that climate change is the most polarizing issue in America (Milman, 2019). In a 2017 Gallup poll, 89% of Democrats worried a "great deal" or a "fair amount" about global warming, compared to just 40% of Republicans (Gallup, 2017), a gap that has increased over time. Studies of climate change public opinion consistently find that political ideology and partisanship dominate other factors as the strongest predictors of attitudes (Brulle et al., 2012;Egan & Mullin, 2017;McCright & Dunlap, 2011a;Nisbet, 2009). Partisan division on climate change has contributed to policy intransigence, with decades passing without any emissions reduction legislation approved by Congress.
To overcome barriers to policymaking presented by issue polarization, policy communicators often seek to shape public opinion through persuasive messaging and information provision through framing. Framing is a communication technique that highlights an aspect of an issue to change the dimension through which people evaluate it (Chong & Druckman, 2007;Entman, 1993;Nelson et al., 1997;Scheufele & Tewksbury, 2007). This technique can lead people to consider new perspectives on issues, influencing attitudes and policy support.
Polarized issues are often the target of framing efforts, as communicators seek to highlight broadly supported aspects of the issues and increase support for new policies or initiatives. For example, policy communicators have attempted to generate public support for climate mitigation policies by emphasizing the economic costs of inaction and the economic benefits of renewable energy (Hoffman, 2011;Hulme, 2009;Stern, 2007), the local impacts of climate change (Scannell & Gifford, 2013;Spence & Pidgeon, 2010), or the public health concerns of increased emissions (Maibach et al., 2010;Myers et al., 2012), among others (for a review, see Nisbet, 2009).
However, conclusions about the effectiveness of framing on highly polarizing issues are mixed. With partisan identities increasingly dominating evaluations of political issues, research suggests that frames often fail to change attitudes among partisans because they disregard information that challenges their party's established position (Druckman, 2001;Druckman et al., 2013;Mutz, 2011;Slothuus & de Vreese, 2010;Taber et al., 2001). Climate change frames, for example, rarely change attitudes among Republicans (whose partisan identity is generally associated with opposition to climate change reduction policies) (Bernauer & McGrath, 2016;Hart & Nisbet, 2012;Zhou, 2016).
Understanding partisan identities as social identities, however, provides a fruitful framework for considering when and how framing efforts can influence beliefs on a highly polarized policy issue. Social identities are psychological constructs that define who we are and how we see ourselves and act as heuristics to drive attitudes and behaviors (Tajfel & Turner, 1986;Turner et al., 1987). Understanding political identities as social identities is not a new concept (Campbell et al., 1960;Converse, 1964), but political identities are increasingly relevant to how Americans think of themselves in society (Baldassarri & Gelman, 2008;Mason, 2018;Westwood et al., 2017).
While political identities are powerful drivers of attitudes on policy issues, they are not the only identities that are relevant to policy attitudes. Social identities come in a number of forms, from racial to religious, and, increasingly, political identities (Green et al., 2002;Huddy, 2001;Huddy et al., 2015). Cross-pressures (conflicting political attitudes driven by opposing identities within an individual) remain across wide swaths of the citizenry (Hillygus & Shields, 2008;Lavine et al., 2012). Specifically, political psychologists find that many non-partisan social identities can inform citizens' opinions on policy issues (Huddy, 2001;Klandermans, 2002;Lewis-Beck et al., 2008). While a citizen's partisan identity may lead them to oppose increased spending on education, for example, their identity as a parent may lead them to support such a policy (Greenlee, 2014).
Research has found that the activation of different identities can influence policy perceptions (Klar, 2013). However, this theory has not been used to test how identity salience interacts with framing. If, as research suggests, framing fails among highly polarizing issues because partisan identities are highly salient (Chong & Druckman, 2007;Druckman, 2001;Egan & Mullin, 2017;Slothuus & de Vreese, 2010), then shifting the lens through which an individual interprets a message could diminish the effect of partisanship. Such a shift could then increase the effectiveness of the frame despite high polarization. I examined this notion by testing how increasing the salience of partisan and non-partisan identities influences how a frame affects attitudes and policy preferences.
Using an online survey experiment, I tested how priming either an alternative, non-partisan social identity (parental identity) or a partisan identity prior to presenting a climate change frame influenced levels of concern about climate change, likelihood to undertake pro-climate behaviors, and support for climate change policies among parents in the United States. I found that priming a parental identity increased climate change concern and likely behavior among Republicans compared to receiving no treatment, although not more than simply providing the frame itself. Meanwhile, priming Republicans' partisan identities eliminated any effect of the frame on attitudes. For Democrats, priming a parental identity had a marginally negative (though not statistically significant) effect on climate change behaviors and policy support (compared to receiving no treatment), while priming a partisan identity increased climate change concern among Democrats.
The difference between partisan and parental identity salience on framing effects was most significant on attitudes towards a carbon tax. Priming a parental identity increased support for a carbon tax among Republicans (compared to priming a partisan identity), while it decreased support for a carbon tax among Democrats (compared to priming a partisan identity). These findings suggest that identity salience does matter in framing effectiveness and identifies a pathway to effectively communicating in highly polarized policy contexts.
In the following sections, I describe the theoretical framework developed to explain the interaction between identity salience and framing effectiveness in greater detail. Then, I describe the experimental design of the survey, followed by a discussion of the results and implications for political psychology and communication.

Background
Social identity theory argues that an important part of one's self-understanding is one's social group memberships and intergroup relations (Tajfel & Turner, 1986). The core of the theory asserts that certain social contexts cause people to think of themselves and others as group members, not just unique individuals (Ellemers & Haslam, 2011). When a social identity is salient, group members aim to exhibit uniform attitudes, behaviors, and preferences, to emphasize similarities with the in-group and differentiate oneself from out-group members (Tajfel & Turner, 1979;Turner et al., 1987).
Social identities are powerful drivers of attitudes and behaviors inside the political world and out (Green et al., 2002;Huddy, 2001;Whitmarsh & O'Neill, 2010).
Social identities can even drive behaviors that would otherwise be inconsistent with an individual's rational interests (Akerlof & Kranton, 2000). Regarding climate change, several studies have shown that attitudes towards climate change are driven more by identities than facts. Notably,  demonstrated that individuals selectively credit or discredit evidence of climate change risk based on whether they believe the experts share their worldviews. Campbell & Kay (2014) noted how antigovernment identities can drive an aversion to many of the solutions for climate change, because they involve government intervention. McCright & Dunlap (2011a) observe that a specific identity groupconservative white malescontributed significantly to the high level of climate change denial in the United States.
The concept of identity salience is key to understanding how identities influence the interpretation of political messages. Identity salience refers to the activation of a certain identity as a response to a social or contextual situation, making membership in a social group function psychologically (Turner et al., 1987). This then increases the influence of a particular identity on an individual's behaviors and perceptions (Stets & Burke, 2000;Turner et al., 1987). Each of us has different identities that become more or less salient in our minds at different points in our lives. For example, in a workplace, one's professional identity may be most salient, but upon returning home to children, a parental identity may be most salient and most likely to drive one's attitudes and behaviors.
Identity salience is an important but underexplored link between the literature on framing and that on identity theory. As Druckman, Peterson and Slothuus (2013) described, polarized issue environments heighten the impact of partisan identities on issue evaluations while decreasing the impact of other considerations such as substantive information. However, it may be possible to lessen the effect of partisan identity salience on framing effectiveness if the message is received while another, nonpartisan identity is salient instead.
One way to manipulate identity salience is through priming. Priming is a technique that increases the salience of a particular identity, leading an individual to evaluate an issue through the lens of that identity. This concept rests on the idea that when individuals are asked about their opinion on a policy issue, they do not necessarily perform a systematic review of everything they know in order to develop a policy attitude (Zaller, 1992). Instead, they form an opinion based on the information (or identity) that is most available and salient at the moment (Turner et al., 1987;Tversky & Kahneman, 1973). If a partisan identity is most salient, policy attitudes will likely be in line with partisan ideologies. However, if another identity is salient, the effect of partisanship may diminish.
Priming can change the identity lens through which an individual interprets a message (Klar, 2013). If a frame conflicts with an individual's salient identity, they are likely to reject the frame and maintain political attitudes that align with the salient identity. For example, on the issue of climate change, Democrats are far more likely to support climate mitigation policies than are Republicans (Egan & Mullin, 2017).
Because the issue is so polarizing, individuals generally approach the issue with their partisan identities salient. If a partisan identity is salient when encountering a framing message about climate change, one might expect Republicans to reject the message while Democrats accept it, leading to further polarization in policy attitudes. However, if the identity lens shifts and a non-partisan identity becomes salient instead, the frame may be more effective at decreasing polarization. In this case, if the frame resonates with the salient, non-partisan identity, then both partisans may be likely to accept the frame, leading to more bipartisan policy agreement.
An important component of this theory is that the frame must resonate with the salient non-partisan identity. A frame about the economic benefits of renewable energy presented while a religious identity is salient is unlikely to be as effective as a frame about the moral or religious imperative to maintain purity and sanctity of the Earth. In short, frames are most likely to be accepted when they match an individual's salient identity. For that reason, in this study I tested how increasing the salience of one specific non-partisan identity, a parental identity, influences the effectiveness of a frame about the impact of climate change on future generations.
The role that parental identities play in driving political attitudes has been welldocumented. Klar (2013) studied the impact of parental identities on political preferences and found that these identities can, in certain situations, overpower partisan identities, particularly when the frame threatens a parental identity. Similarly, Greenlee's (2014) study of the political consequences of motherhood offered significant qualitative and quantitative support for the fact that an individual's identity as a mother has significant impacts on their political attitudes in ways that promote their child's well-being.
The impact of climate change on future generations is also a frame commonly used to elicit concern and policy support. Page (2007) detailed the impacts of climate change on future generations through the idea of "intergenerational justice," while Gardiner (2006) called the issue of climate change a "Perfect Moral Storm" due to its effects on future generations. Barack Obama also used a future generations frame to inspire action on climate change in his final speech as president (Obama, 2017). These frames are a response to a common roadblock found when attempting to generate concern about the issue of climate change: because most of the impacts will be felt in the future, it is difficult to motivate individuals to take costly action now (Markowitz & Shariff, 2012;Weber, 2006). However, evidence finds that priming an individual to consider their legacy increases climate change beliefs (Zaval et al., 2015). As guardians of the future generations who will feel the impacts of climate change the most, parents may be more likely to feel motivated to limit the impacts of climate change now, to secure a healthy and prosperous future for their children and to positively impact their own legacy.

Materials and Methods
This study investigated how identity salience impacts framing effectiveness using climate change and parental identities as a case. Does response to a climate change frame change based on the salient identity? Can priming a parental identity improve the effectiveness of a future generations frame at increasing concern about climate change, likelihood to undertake pro-environmental behaviors, and support for climate change mitigation policies? Are climate change frames less effective in environments when partisan identities are highly salient?
The theoretical framework described in the previous section predicts that priming a non-partisan identity before presenting a framing message may lead to more bipartisan agreement on highly polarizing policy issues. Due to the high levels of partisan polarization on climate change, I expected Republicans and Democrats to respond differently to the experimental treatment, and therefore include distinct hypotheses for each subgroup.
Because of the highly polarizing nature of climate change and what we know about the challenges of climate change framing, I expected that framing alone would be ineffective at changing climate change attitudes from baseline levels. Republicans would be likely to reject the frame as it conflicts with their partisan identities, and Democrats would likely already have high climate change attitudes (a ceiling effect). If identity salience is an important factor in framing effectiveness, however, I expected that increasing the salience of a parental identity (and therefore decreasing the salience of a partisan identity) would increase the effectiveness of a future generations frame, particularly among Republicans. Similarly, I expected that increasing the salience of one's partisan identity would accentuate polarization on climate change, leading to more climate change concern, pro-environmental behaviors and policy support among Democrats and less among Republicans. 2

Republican Subgroup Hypotheses:
1. Framing alone will be ineffective at changing climate concern, proenvironmental behaviors and climate change policy support among Republicans, compared to a control group.
2. Priming a parental identity before presenting a future generations frame will increase climate concern, pro-environmental behaviors and climate change policy support among Republicans, compared to a control group. 2 The experiment and hypotheses were registered prior to receiving the data in the EGAP registration system, ID 20180416AA 3. Priming a partisan identity before presenting a future generations frame will decrease climate concern, pro-environmental behaviors and climate change policy support among Republicans, compared to a control group.

Democrat Subgroup Hypotheses:
4. Framing alone will be ineffective at increasing climate change concern, proenvironmental behaviors and climate change policy support among Democrats, compared to a control group.
5. Priming a parental identity before presenting a future generations frame will increase climate concern, pro-environmental behaviors and climate change policy support among Democrats, compared to a control group.
6. Priming a partisan identity before presenting a future generations frame will increase climate concern, pro-environmental behaviors and climate change policy support among Democrats, compared to a control group.

Sample
I tested these hypotheses using a four-condition survey experiment with U.S. parents fielded online in spring 2018 (n=978). Although it is possible that non-parents could be motivated by a frame about the impact of climate change on future generations, the focus of this study was on the role that identity salience plays in framing effectiveness.
To that end, it was important that participants held the relevant identitythat of parenthood. 3 Similarly, because this study was also designed to manipulate partisan identity salience, individuals that identified as independents or as having no partisan identity were excluded from the analysis. Post-hoc analysis of the subsample of independents found no significant results of any treatment (results available in online appendix). The implications of this finding are considered further in the discussion section.
Participants were recruited through the online survey platform Prolific.ac between April-June 2018. Prolific is an online survey research platform that allows researchers to screen participants based on demographic criteria pre-collected by the platform. Studies have found the participant base of the Prolific platform to be more diverse, more naïve, less dishonest and generally of a higher quality for academic research than other online survey platforms such as Amazon's Mturk (Palan & Schitter, 2018;Peer et al., 2017). Users that met the study qualifications (U.S. residents who were parents) were invited to take this survey for payment. Because political liberals tend to be over-represented in online survey platforms, I oversampled participants who identified as politically conservative to ensure adequate subgroup sizes. The recruited sample was diverse on several key demographics. Fifty-three percent of the sample identified as female, and the average age was 41 years old. Fifty-seven percent had children ages 0-10, and 76% had children under the age of 21. Politically, 48% of the sample identified as Democrats and 32% identified as Republicans. 4 Two multiple-choice attention check questions were embedded in the survey after the frame that asked respondents (1) what organization published the report, and (2) what was the topic of the message. While t-tests show that participants who failed the attention checks were more likely to be male and Republican, including these respondents in the analysis generated additional noise but did not significantly change the results. Considering this and recent findings that eliminating respondents that fail attention checks does not bias scale validity (Kung et al., 2018;Paas & Morren, 2018), responses were excluded if the participant failed the attention checks (a total of 157 participants).

Experimental Design
Survey participants were randomly assigned to one of four conditions at the beginning of the survey. In the first treatment condition (Frame Only), participants read a frame about the impact of climate change on future generations in the form of a short excerpt from a recent UNICEF report on climate change's impacts on children (including an image of the report's cover). UNICEF is largely viewed positively by the public due to its generally non-political work in promoting the welfare of children around the world (Quesnel, 2004). It was therefore chosen as a neutral message source. Participants in the second treatment condition (Parent Prime + Frame) were first primed to think about their identity as a parent before reading the frame about the impact of climate change on future generations, using parental identity priming questions adapted from Klar (2013).
Participants in the third treatment condition (Partisan Prime + Frame) were primed to think about their partisan identity (using questions mirroring the parental identity prime) before reading the same climate change frame. Finally, a control condition provided neither the identity primes nor the frame, moving participants directly to the outcome variables. The frame and identity priming questions can be found in the online appendix.
The participants were randomly assigned to one of the four experimental groups.
The Frame Only group included 213 respondents (22% of the sample), the Parent Prime + Frame group included 234 respondents (24% of the sample), the Partisan Prime + Frame group included 234 respondents (24% of the sample), and the control group included 297 respondents (30% of the sample). Partisanship was relatively evenly distributed across treatment groups, and no statistically significant differences in partisan identity existed between treatment groups. Descriptive statistics and balance checks between treatment groups can be found in the online appendix.

Outcome Measures
I measured treatment effects through three composite outcome variables: level of concern about climate change, likelihood of undertaking personal and political behavior in support of climate change policies, and support for climate change mitigation policies. These three variables were intended to correspond to different levels of effort regarding attitude change. Climate change concern was measured by the subject selecting how strongly they agree (on a 7-point Likert scale, from strongly disagree to strongly agree) with the following statement: "Climate change is a pressing problem." Behaviors, which are "costlier" in terms of effort than attitude change, may be more difficult to shift (although due to the nature of the survey, only intentions to undertake behaviors could be measured). The behavioral questions measured the subject's likelihood (on a 7-point Likert scale, from extremely unlikely to extremely likely) of (1) changing their personal behavior to help combat climate change (recycle, drive less, use less electricity, etc.), (2) signing a petition to urge political action on climate change, and (3) contacting their elected representative to ask them to support legislation to combat climate change. Responses to three questions were highly correlated (Cronbach's a=0.83) and were therefore combined into a single "behavior" variable by taking the average score across the three behaviors.
Finally, policy preferences tend to be most closely aligned with partisan identities, and therefore may be particularly difficult to change on highly polarizing issues like climate change. The policy questions (also on a 7-point Likert scale, from strongly oppose to strongly support) asked respondents how much the subject supported limiting the amount of emissions that power plants can release, increasing government investment in renewable energy research, and a tax on carbon emissions that returns all revenue back to citizens. These questions were also highly correlated (Cronbach's a=0.87) and were combined into a single "policy" variable.

Analytical Strategy
I analyzed the results of the survey experiment using a series of OLS regression models on each outcome variable. Regressions were used, as opposed to simple comparisons of means between treatment groups, to control for several key variables (participant age, child age, political ideology [strongly liberal to strongly conservative] and gender) known to correlate highly with climate change attitudes. The survey firm provided these control variables based on demographic questionnaires completed when the subjects joined the service, mitigating concerns of post-treatment bias (Montgomery et al., 2018).
To measure average treatment effect, I created three dichotomous variables comparing the subjects in each treatment group to the control group. I then used OLS regression models, including the control variables, to estimate the increase or decrease in outcomes compared to the control group. Because the hypotheses predicted different treatment effects based on partisanship, I ran the regressions separately on the subgroups of participants that identified as Republican and Democrat. Finally, to understand the effect of partisan versus non-partisan identity salience on framing effectiveness, I also ran OLS regressions comparing participants in the Parent Prime + Frame and Partisan Prime + Frame groups.

Results
Because a participant's partisan identity was expected to drive divergent reactions to a climate change message, this section presents the experimental results by partisan subgroup. I first describe the results of the regressions testing each treatment group against the control group. Then, I test the effect of the parent prime versus the partisan prime. Detailed regression tables can be found in the online appendix.

Republican Subgroup Treatment Effects
Among Republicans, both the Frame Only and the Parent Prime + Frame treatments significantly increased climate change concern and likelihood to undertake pro-climate behaviors, while the partisan prime eliminated this framing effect. Figure 1 shows the regression coefficients and 95% confidence intervals for each of the treatment groups among Republicans (compared to the control group).

Frame Only vs. Control
In contrast to previous literature finding that framing alone is effective at shifting climate change attitudes among Republicans, receiving the Frame Only treatment increased concern about climate change by 0.71 (10% increase on a 7-point scale, p<.05) and increased likelihood of undertaking pro-climate behaviors by 0.66 (9% increase, p<.05). For reference, a coefficient of 1.0 (representing a 14% increase) would correspond to, for example, moving from "somewhat agree" to "agree" on the 7-point scale. There was no significant effect of the Frame Only on policy preferences.

Parent Prime + Frame vs. Control
In support of the second hypothesis, receiving the Parent Prime + Frame treatment was also associated with greater climate concern (0.70, or 10%, higher than the control group, p<.05) and greater likelihood of undertaking pro-climate behaviors (0.71, or 10%, higher than the control group, p<.05). While there was also a positive treatment effect on policy support among those in the Frame Only and Parent Prime + Frame groups, the effect was not statistically significant.

Partisan Prime + Frame vs. Control
Supporting the third hypothesis, Republicans in the Partisan Prime + Frame group experienced no significant increase in any of the outcome variables compared to those in the control group. Receiving a partisan prime before receiving the frame eliminated any positive framing effect on climate concern and behavior and had a negative (but not statistically significant) effect on policy preferences.

Parent Prime vs. Partisan Prime
To understand how parental versus partisan identity salience changed response to the frame, I also analyzed the effect of receiving the parental prime versus the partisan prime before reading the framing message. For the three composite dependent variables, the effect (of the parental prime versus partisan prime) was positive, but not statistically significant among Republicans.
However, when I broke out the policy composite variable into the three separate policy questions (support for emissions regulation, renewable energy, and a carbon tax), Republicans receiving the Parent Prime + Frame had significantly higher carbon tax policy support than Republicans receiving the Partisan Prime + Frame. Figure 2 shows these treatment effects for each policy variable. Republicans who received the parental prime before the frame reported 13% higher levels of support for a carbon tax compared to Republicans that received the partisan prime before the frame (p<.05), suggesting that identity salience does influence framing effectiveness, especially on certain polarizing policy issues.

Democrat Subgroup Treatment Effects
The results for the Democrats in the sample were fairly consistent with the hypotheses, with the important difference that priming a partisan identity seemed to increase framing effectiveness more than priming a non-partisan (parental) identity. Figure 3 shows the regression coefficients and 95% confidence intervals for each of the treatment groups among Republicans (compared to the control group).

Frame Only vs. Control
Because Democrats generally have high baseline levels of concern about and support for policies addressing climate change, I expected that the framing treatment would have minimal impact on attitudes. Indeed, for the most part, the results matched this expectation: receiving only the message frame did not significantly increase outcome measures among Democrats. Interestingly, however, the Frame Only treatment was associated with a decrease in likelihood of undertaking pro-climate behaviors. For Democrats, receiving the Frame Only was associated with a marginally significant -0.55 (8%) decrease in likelihood to undertake pro-climate behaviors (p=0.051).

Parent Prime + Frame vs. Control
Contrary to the fifth hypothesis, priming a parental identity before presenting the frame was also ineffective at increasing framing effectiveness among Democrats. While there was no real effect on climate concern or policy preferences, results do suggest that receiving the parent prime before the frame may reduce intended pro-environmental behavior. However, this effect was not statistically significant.

Partisan Prime + Frame vs. Control
In support of the sixth hypothesis, Democrats that received the Partisan Prime + Frame treatment did report higher levels of climate change concern than Democrats in the control group. For these Democrats, the treatment was associated with a 0.27 (4%) increase in climate change concern (p<.05). Figure 3 shows the overall treatment effects of each treatment group versus the control group for the three composite variables.

Parent Prime vs. Partisan Prime
Again, to understand the difference between parental versus partisan identity salience, I also ran regressions to measure the treatment effect of receiving a parental prime versus a partisan prime among Democrats. Using the composite dependent variables, there was a marginally significant negative effect for the policy variable. Democrats receiving the parent prime were 0.37 (or about 5%) less supportive of climate change policies than were Democrats receiving the partisan prime (p<0.1). When the policy variable was broken out by question, support for a carbon tax was once again the major contributor to this effect. Democrats in the Parent Prime + Frame group reported levels of carbon tax support that were 0.59 points (or about 8%) lower than the Democrats in the Partisan Prime + Frame group (p<.05). Figure 4 shows the treatment effects and 95% confidence intervals for Democrats receiving a parental versus a partisan prime for the policy variables.

Discussion
In a political environment dominated by partisanship and an increasing number of issues stagnated by polarization, finding ways to build bipartisan agreement is urgently important. While framing has gained a reputation as an ineffective way to change attitudes on highly polarizing issues, activating alternative, non-partisan identities could help increasing framing effectiveness. To better understand how and when framing works to change attitudes on highly polarizing issues, this study examined the role that identity salience plays in climate change framing effectiveness.
Through a survey experiment, I found evidence that identity salience does matter to framing effectiveness. Specifically, presenting a frame about the impact of climate change on future generations increased climate concern and intended proclimate behaviors among Republican parents, but first priming a partisan identity eliminated this framing effect. For Democrats, priming a partisan identity increased climate change concern compared to the control, while priming a parental identity did not lead to any increase in attitudes. Priming partisan identities resulted in significantly more polarized policy support, while priming parental identities resulted in significantly less polarized policy support between Republicans and Democrats.
For both Republicans and Democrats, the biggest difference in outcomes between the parental prime and the partisan prime treatment groups was on support for a carbon tax. Republicans whose parental identity was primed reported 13% higher levels of support for a carbon tax after reading the frame compared to Republicans whose partisan identity was primed. Additionally, Democrats whose parental identities were primed had levels of carbon tax policy support that were 8% lower than Democrats whose partisan identity was primed. These results suggest that polarized responses to climate change frames increase when partisan identities are salient and decrease when non-partisan identities are salient. This is possibly because attitudes towards a carbon tax are already highly polarized in the American public and priming someone's partisan identity will reflect this polarization. Importantly, however, the findings show that priming a non-partisan identity can result in less polarization on this type of policy.
An unexpected finding of this study was that priming respondents' parental identities did not seem to be required for the frame to positively influence climate change attitudes among Republicans. I found that the future generations frame alone effectively increased climate change concern and likelihood to undertake pro-climate behaviors among Republicans, without needing to first prime the parental identity.
While this could suggest that identity salience does not matter to framing effectiveness in the way the theoretical framework predicts, there are several explanations for this result that clarify the role of identity salience in framing effectiveness. The first explanation could be that a parental identity is already highly salient among parents, and therefore additional priming is not necessary for a future generations frame to resonate with this population. To evaluate this, one could replicate this study using another identity that is less likely to already be highly salient among respondentsperhaps a workplace identity or a place-based identity.
Similarly, the frame itself may have activated the parental identity without needing the additional prime. This explanation suggests that perhaps one of the solutions to increasing framing effectiveness is not necessarily priming relevant identities, but simply matching a frame to one of a citizen's already salient identities (using an "identity-relevant frame"). Replicating the study among a sample of nonparents could help enlighten whether the frame worked because it matched subjects' parental identity or because it was simply an inherently effective frame. Indeed, a pretest of this experiment, run in 2016 in a sample including both parents and nonparents, provides some insight into this question. In the subgroup of non-parents in the pretest, no framing effects were found, yet both the frame and the Parent Prime + Frame treatments were effective at changing attitudes among the parents.
The importance of the interaction between identity salience and framing effectiveness is underscored by the results of the Partisan Prime + Frame treatment.
Attempts to increase climate change attitudes among Republicans using a frame were cancelled out when the respondent's partisan identity was salient. Similarly, for climate change concern, the frame became more effective among Democrats when their partisan identity was primed. These finding are important for communicators attempting to use framing techniques to gain bipartisan buy-in on polarizing issues: to be effective, frames should both (a) match with an existing identity and (b) not be presented in a context when an opposing partisan identity is highly salient.
Another notable finding is the negative effect that several of the treatments had among Democrats, and the significantly higher climate change policy support among Democrats whose partisan identity was primed compared to those whose parental identity was primed. These findings suggest that considering climate change through a parental, or non-partisan, identity lens may reduce climate change concern, behavior or policy support for Democrats, especially when compared to viewing the issue through a partisan identity lens. Given that citizens in these groups generally have higher baseline climate change attitudes than Republicans, reframing the issue, particularly in a way that focuses on the future, may make it seem less urgent. Climate change communications have often been criticized as unrelatable when they focus on impacts that will be felt by people in distant countries or far into the future (Spence et al., 2011). Therefore, framing climate change in this way, even among parents, may reduce concern, urgency, and policy support among Democrats. The frame itself, which indirectly focuses on the impacts to children in developing countries (the focus of UNICEF's work), may also have reduced the urgency of climate action.
Another potential explanation for this backfire effect among Democrats, particularly for the behavioral and policy variables, could be related to sentiments around political efficacy in the post-2016 election political environment. In a pretest of this experiment fielded prior to the 2016 election, Democrat parents in the Frame Only treatment group reported significantly higher climate change concern and likelihood to undertake political behaviors in support for climate change, compared to Democrats in the control group. In the full study fielded in 2018, however, Democrats in the Frame Only and Parent Prime + Frame groups were more concerned about climate change, but less likely to undertake political behaviors and less supportive of climate change policies than their counterparts in the control group (although these results are not statistically significant). This backfire effect could be due to a reduction in perceived effectiveness of political action on climate change after the change in presidential administration in 2017 (a "Trump effect"). Subjects who were primed to think about their children may have become more concerned about climate change and its effects on future generations, but more despondent about the futility of any climate change advocacy in a political environment that is outwardly hostile to climate change mitigation policies.
While this study offers an important step in furthering our understanding about when framing works and when it does not, several extensions of this could further expand our understanding of the importance of identity salience. First, particularly considering arguments that identity primes and frames are indistinguishable psychologically (see, e.g. Klar 2013, Chong andDruckman 2007), a useful extension of this study would be to actively test a "frame only" condition against a "prime only" condition. While power limitations prevented this test in the current study, this type of extension would help further clarify the relationship between identity salience and framing.
Second, the impact of identity primes and frames embedded within survey experiments has drawn criticism. Experimental surveys may not mimic a real-world context, and there are many factors within the survey (question order effects, unintentional priming from external experiences, and the attention of subjects) that can limit the effectiveness of experimental treatments. It would be beneficial, therefore, to further test this theory in a real-world context involving natural identity priming. For example, field experiments during situations when an alternative identity is naturally more salient than a partisan identity (for example, parents at a parent-focused event such as a PTA meeting).
Third, a major limitation of single-instance survey experiments involves ambiguity around how long these effects last. Do the changes in policy preferences that we see in the treatment groups have a lasting effect? More longitudinal study of these effects is needed to contribute to the relevance of this work for policymakers and political communicators seeking to generate long-term changes in the public's policy preferences.
The treatment effects of non-partisan identity priming were notably stronger among Republicans than around Democrats, which also raises the question about whether this theory may be more applicable to one party than the other. However, it should be noted that the topic area investigated herethat of climate changeis unlikely to show significant changes among Democrats because these individuals already have consistently high levels of climate change concern and policy support.
Studies have found much greater variability in climate change attitudes among Republicans than Democrats. Because the frame was designed to increase climate change attitudes, it is not surprising that there would be minimal significant difference in attitudes among Democrats. However, it should be emphasized that, in support of the theory presented here, priming a partisan identity among Democrats did increase climate change concern, compared to the control. This suggests that identity salience matters for both Democrats and Republicans. Nevertheless, it would be helpful to test this framework with other issues in the future, particularly those that are less supported by Democrats, to broaden the validity of the findings.
Finally, although not immediately within the scope of this study, it is interesting to consider how partisan and non-partisan identities interact with framing effects among political independents. These individuals were excluded from the main analysis because they lacked a cohesive partisan identity to prime and therefore would have created significant noise in the partisan prime treatment group. However, one might expect that without a salient partisan identity, the priming of a non-partisan identity may have an even more powerful effect on political independents than on Democrats or Republicans.
In a supplementary analysis (available in an online appendix), I found no significant impact of any of the treatments on the subgroup of independents. While this may be due to the heterogenous nature of political independents (the confidence intervals were the widest among this subgroup), the lack of significant results prompts a meaningful question for future research about the role of identity priming among political independents.
The overall findings of this study offer some important insights for political communication. Namely, that communicating messages when partisan identities are highly salient is likely to increase polarized responses, while communicating while nonpartisan identities are salient may be helpful in depolarizing responses. This suggest the importance of microtargeting when using communication frames on highly polarizing issues. As seen in this study, different subgroups are likely to react differently to climate change frames, and in certain circumstances, frames and non-partisan identity priming can backfire. While encouraging Republicans to consider climate change through an alternative identity may effectively increase their levels of concern and pro-climate political behaviors, such efforts may have the opposite effect among Democrats. For Democrats, the greatest increases in climate change attitudes may be achieved by keeping their partisan identities salient. For this reason, it may be particularly important for policy communicators to consider sending different types of messages to different subgroups in the population. In the modern media environment that is characterized by "information bubbles" and consumer choice in media environments, there are numerous opportunities for communicators to introduce messages targeted at certain groups, or presented in situations where a certain identity is salient over others.
The role that partisan and non-partisan identities play in driving political attitudes is controversial, nuanced, and understudied. This study proposes and tests a potential role that identity salience may play in political attitudes and, specifically, the effectiveness of frames at changing political attitudes. The results present a clear picture of two important findings: (1) a highly salient partisan identity can negate any effect of a frame on attitudes when the partisan identity opposes the frame, and (2) matching a frame to a salient, non-partisan identity can decrease the polarized response and lead to more bipartisan policy attitudes. While this study does not find definitive evidence that priming a non-partisan identity in addition to the identity-relevant frame is necessary for this effect, it may still be important when targeting an identity that is less consistently salient than a parental identity. Future research should continue to investigate the role that identity salience plays in framing effectiveness to better inform the strategies of political communicators.