Predictors of U.S. public support for climate aid to developing countries

Effectively responding to global warming requires mitigation and adaptation efforts worldwide. Although developed countries have pledged substantial financial support to help developing countries respond to climate change, these pledges have yet to be fulfilled. A majority of American voters support providing aid to developing nations, however, levels of support differ sharply between Democrats and Republicans. To investigate the predictors of support for climate aid among registered voters in the United States and to assess the effect of political party identification, we conducted hierarchical regression and relative weight analysis on a nationally representative sample of U.S. registered voters (n = 898) surveyed in 2021. Among all voters, the predictors of support for climate aid were: party identification (with more support among Democrats), perceived risk to developing countries, worry about global warming, injunctive beliefs that the U.S. should do more, and global warming belief certainty. Among Democrats, the predictors were: perceived risk to the U.S., worry, injunctive beliefs the U.S. should do more, education, and income. Among Republicans, the predictors were: perceived risk to developing countries, and injunctive beliefs the U.S. should do more. These findings have both theoretical and practical relevance for efforts to build public support for development assistance aimed at reducing climate change.


Introduction
Climate change has global impacts that disproportionately affect developing countries (IPCC 2023).Since the late 1980s-when international climate negotiations began-providing aid to developing countries to help them achieve their climate goals has been a top political issue (Pallemaerts and Armstrong 2009).However, the commitments made thus far by developed nations are not commensurate with the need, and they have largely failed to fulfill their commitments (Timperly 2021).Financial assistance is needed to support both mitigation (such as renewable energy) and adaptation and resilience strategies (such as building coastal defenses) (United Nations Climate Change 2021).During the Twenty-seventh Conference of the Parties (COP27) in Sharm El-Sheikh in 2022, countries agreed to establish a loss and damage fund to provide compensation to developing countries who have contributed less to climate change but are the most vulnerable to climate impacts (United Nations 2022).Although the COP27 agreement represented significant progress, many details about how the fund will be structured, who will pay, and how much have yet to be finalized.Failure to achieve this goal in the past has further exacerbated the impacts of climate change on developing countries (Rowling 2021).
This study investigates the factors that shape and sustain public support for climate aid to developing countries.We focus on the correlates of U.S. public support for such aid, which is important for multiple reasons.First, the U.S. is a global economic and scientific leader (Nye 2008), and as a result it has an outsized capacity to assist developing countries in responding to global warming (Rose and Estes 2021).Moreover, the U.S. is historically the world's largest emitter of carbon emissions, while most developing countries have emitted little by comparison and theyface some of the greatest risks from global warming (Evans 2022).Durable political will to provide foreign assistance depends in part on public will-as elected officials are unlikely to make commitments to policies that do not have support from their constituents (Leiserowitz 2019).Therefore, it is important to understand the dynamics of public opinion in support of and in opposition to providing climate aid to other nations.In the fall of 2021, although a majority of registered voters in the U.S. supported providing financial and technical assistance to developing countries to reduce carbon emissions (i.e., mitigation, 66%) and to help them prepare for the effects of global warming (i.e., adaptation, 61%), Democrats tended to support these policies much more strongly than Republicans (Leiserowitz et al 2021).
A large body of research over the past several decades has identified predictors of public support for global warming policy, political activism, voting issue importance, risk perceptions, and trust in scientific research (e.g., Myers et al 2016, Ballew et al 2019, Goldberg et al 2020, Bumann 2021, Campbell et al 2021).However, to our knowledge, there has been no research on the factors that predict support for providing climate aid to developing countries.In this study, we use data from a nationally representative survey to investigate these factors.By examining theoretically relevant constructs, we aim to (1) identify predictors of U.S. public support for climate aid to developing countries to help them limit their carbon emissions and prepare for the impacts of global warming, (2) offer insights into how to enhance public support for climate aid to developing countries to fight global warming, and (3) suggest guidance on future research into public support for international climate action.This study's hypotheses were largely modeled after previous work of Campbell et al (2021), who sought to identify the factors that predict global warming as an important voting issue in the U.S.

Literature review
Our study draws on multiple lines of prior research to investigate potential predictors of support for climate aid: key beliefs previously associated with climate change attitudes and actions (i.e., belief certainty, human causation, risk perception, and worry), injunctive beliefs, media exposure, religiosity, and party identification (Roser-Renouf et al 2014, van der Linden et al 2015, 2019, Campbell et al 2021).

Key beliefs
Several key beliefs about global warming-that it is real, human-caused, and harmful-consistently predict important attitudinal and behavioral outcomes, like policy support, political advocacy, and consumer advocacy (e.g., Ballew et al 2019, Goldberg et al 2020, Krosnick et al 2006, Leiserowitz 2006, Roser-Renouf et al 2014, Roser-Renouf 2016, Smith and Leiserowitz 2014, van der Linden et al 2015, 2019, van der Linden 2017).Here we investigate whether these key beliefs also predict support for providing aid to developing countries to help them address global warming.

Belief certainty
Belief certainty refers to people's certainty that global warming is, or is not, happening (Roser-Renouf et al 2014, van der Linden et al 2015van der Linden et al , 2019)).Previous studies found that the more certain people are that global warming is happening, the more likely they are to support climate policies or engage in climate action (Ding et al 2011, Roser-Renouf et al 2014, Hall et al 2018, van der Linden et al 2015, 2019, Bromley-Trujillo and Poe 2020, Goldberg et al 2020, Bumann 2021, Campbell et al 2021).Hence, we hypothesize: H1: Belief certainty that global warming is happening will be positively associated with support for climate aid to developing countries.

Human causation
The belief that global warming is human-caused refers to the idea that the increase in global warming is predominantly attributed to human activities, such as burning fossil fuels instead of natural influences (Ding et al 2011, van der Linden et al 2015, 2019).The belief that global warming is caused by human activities is associated with support for climate and clean energy policies (Ding et al 2011, Guy et al 2014, Lee et al 2015, Hall et al 2018, Olson-Hazboun et al 2018, 2018, van der Linden et al 2015, 2019, Goldberg et al 2020, Shao and Hao 2020).Moreover, providing people with information about the human causes of global warming increases their support for climate policy (Bergquist et al 2022).Therefore, we hypothesize: H2: Belief that global warming is human-caused will be positively associated with support for climate aid to developing countries.

Risk perception
Perceiving global warming as a risk is strongly associated with support for climate policies and actions (Ballew et al 2019, Campbell et al 2021, Goldberg et al 2020, Lee et al 2015, Leiserowitz 2006, Roser-Renouf 2014, Spence et al 2012).However, even though global warming is often perceived as a greater risk in developing countries than in the developed world (Kim andWolinsky-Nahmias 2014, Lee et al 2015), to the best of our knowledge no one has yet tested the association between perceived risk to developing countries and support for climate aid to such countries, although we believe it is empirically likely that such a relationship should exist.Therefore, we hypothesize that: H3: Perceived harm to people in developing countries will be positively associated with support for climate aid to developing countries.
There is no clear theoretical or empirical basis, however, to hypothesize a specific relationship between perceived risk to the U.S. and support for climate aid.Therefore, we pose the following research question: RQ1: Is perceived harm to people in the U.S. positively associated with support for climate aid to developing countries?

Worry
Research finds that worry about global warming is strongly associated with support for climate policies (Smith and Leiserowitz 2014, van der Linden 2017, Bouman et al 2020, Goldberg et al 2020, Campbell et al 2021).Moreover, studies of Americans' responses to the COVID-19 pandemic found that worry may lead to support for providing aid to developing countries if there is awareness that such aid would help alleviate impacts in the U.S. (Kobayashi et al 2021).Therefore, because global warming is a global problem that impacts the U.S. and can only be fully mitigated through international action, we expect that worry should be positively associated with support for climate aid.Therefore, we hypothesize: H4: Worry about global warming will be positively associated with support for climate aid to developing countries.

Injunctive beliefs
We also investigate whether injunctive beliefs-the belief that action should be taken to address global warming -predict support for climate aid.Specifically, we examine two forms of injunctive beliefs as predictors of support for climate aid: (1) that the U.S. should take more (or less) action and (2) that developing countries should take more (or less) action.In the U.S., injunctive beliefs are positively related to engagement in climaterelated political and consumer activism (Roser-Renouf et al 2014, Roser-Renouf et al 2016).
Increasing aid to developing countries may be perceived as a way that the U.S. could be doing more to fulfill its global responsibility.However, opponents to climate action often argue that the U.S. should not take action on global warming unless other countries do so as well-in effect redirecting the burden of responsibility to other (often developing) countries (Lamb et al 2020).In Fall 2021, 66% of registered U.S. voters said the United States should be doing more to address global warming, however, a larger proportion (81%) said developing countries (like China, India, and Brazil) should be doing more to address it (Leiserowitz et al 2021).For these reasons, people who have strong views that developing countries should be doing more to address global warming may be less likely to support climate aid meant to help those developing nations.Thus, we hypothesize: H5: Belief that the United States should do more to address global warming will be positively associated with support for climate aid to developing countries.
H6: Belief that developing countries should do more to address global warming will be negatively associated with support for climate aid to developing countries.

Media exposure
It is well established that the media plays an important role in shaping public perceptions on political and policy issues (Scheufele and Tewksbury 2007).Agenda-setting theory proposes that the media coverage of an issue can increase its position on the political agenda of a nation by making it more salient and increasing its relative importance (McCombs and Shaw 1972).
The ongoing evolution of television news beyond objective news reporting into opinionated, partisan formats (both liberal and conservative) is likely to enhance the polarizing impact of the media (Feldman 2011, Stroud 2011).These media effects are specifically important in this study because climate reporting varies substantially in the partisan media (Carmichael et al 2017).
Conservative news outlets, like Fox News, typically reject climate science and climate scientists and dismiss the reality or seriousness of global warming (Feldman 2016).In contrast, liberal news outlets, like MSNBC, accept climate science and the seriousness of global warming and often belittle conservative opinions on the topic (McKnight 2010, Feldman 2016).
Mainstream news networks (i.e., ABC, CBS, NBC) tend to report on global warming in a fact-driven, nonpartisan manner, which may increase climate concerns (Carmichael et al 2017).Research indicates that consumption of mainstream and liberal media is positively related to global warming acceptance and scientific consensus, although exposure to conservative media is associated with the opposite effect (Feldman et al 2012).It is therefore likely that exposure to liberal and mainstream media will increase support for climate actionincluding providing climate aid to developing countries-although exposure to conservative media will erode such support (Hornsey andFielding 2016, Walker et al 2018).Based on this we hypothesize: H7: Exposure to liberal news sources will be positively associated with support for climate aid to developing countries.
H8: Exposure to mainstream news sources will be positively associated with support for climate aid to developing countries.
H9: Exposure to conservative news sources will be negatively associated with support for climate aid to developing countries.

Religiosity
Frequent attendance at religious services is associated with supporting or providing aid to the less fortunate (Smith et al 2008), therefore, how often an individual attends religious services may also be associated with their support for climate aid to developing countries.For instance, Wood (2018) found that Australians who regularly attend religious services are more likely to support Australia giving aid to developing countries.Similarly, Paxton and Knack (2012) found a positive relationship between frequent attendance at religious services in donor countries and support for climate aid.Further, Bayram (2016) found religiosity to be related to support for climate aid to developing countries.Many religious traditions promote generosity, especially to those less fortunate.This relationship may extend to the context of global warming, whereby individuals that attend religious services often may be more likely to support climate aid to developing countries to combat global warming.Thus, we hypothesize: H10: Religiosity will be positively associated with support for climate aid to developing countries.

Party identification
Previous studies have demonstrated that partisanship in the U.S. is associated with differences in policy support (Gollwitzer et al 2020, Grossman et al 2020).Additionally, party identification impacts how Americans perceive policies (Green et al 2004).Political party identification and ideology also predicts public responses to global warming, as it is a highly politicized issue in the U.S. Prior studies also indicate that party identification is related to the assessment of immigration and energy policies, whereby individuals are more likely to favor policies supported by their party (Druckman et al 2013, Bolsen et al 2014).Democrats are more likely to support climate aid than Republicans (Milner and Tingley 2013).Further, there may also be differences in the variables that predict Republicans' and Democrats' support for such aid.Therefore, we hypothesize and ask: H11: Party identification will be positively associated with support for climate aid to developing countries, such that Republicans are less supportive.
RQ2: Do predictors of support for climate aid to developing countries differ between Republicans and Democrats?

Participants
Our data was collected as a part of a nationally representative survey of American adults (ages 18+) in September 2021 (n = 1,006) using the Ipsos Knowledge Panel, a probability sampling-based online panel.The survey took an average of 25 min to complete.For our analyses, we included only registered US voters (n = 898).

Independent variables 3.2.1.1. Belief certainty
To measure belief certainty, we first gave participants a brief definition of global warming and they then answered a series of questions.The first question asked, 'Do you think that global warming is happening?' (1 = No, 2 = Don't know, 3 = Yes) and then, depending on whether they responded 'yes' or 'no' they were asked, 'How sure are you that global warming is [not] happening?' (1 = Not at all sure, 4 = Extremely sure).Responses to these questions were combined and recoded to create a nine-point certainty scale (1 = Extremely sure global warming is not happening, 5 = Don't know, 9 = Extremely sure global warming is happening).We treated 'Don't know' as the midpoint in the nine-point scale (M = 7.06, SD = 2.32).

Human causation
We asked participants one question to assess their beliefs regarding the extent to which human activity is causing global warming: 'Assuming global warming is happening, do you think it is K' with response options of 1 = Caused mostly by human activities, 2 = Caused mostly by natural changes in the environment, 3 = Caused by natural changes and human activities, 4 = None of the above because global warming is not happening, and 5 = Other (please specify).We dichotomized the measure, such that those who chose 'caused mostly by human activities' were coded as 1 and all other responses were coded as 0 (M = .58,SD = .49).

Risk perception
We asked participants two questions about the extent to which they believe global warming will cause harm to 'people in the United States' and 'people in developing countries' on a 4-point scale.The response options were 0 = Don't know, 1 = Not at all, 2 = Only a little, 3 = A moderate amount, and 4 = A great deal.'Don't know' responses were treated as missing (U.S. M = 2.98; SD = 1.05,Developing countries M = 3.20, SD= 1.07).

Worry
To measure participants' level of worry about global warming, we asked, 'How worried are you about global warming?' with responses ranging from 1 = Not at all worried to 4 = Very worried (M = 2.89, SD = 1.04).

Injunctive beliefs
To measure injunctive beliefs about global warming, we asked participants, 'Do you think each of the following should be doing more or less to address global warming?' with separate questions asking about 'the United States' and 'Developing countries (such as China, India, and Brazil)'.Response options ranged from 1 = Much less to 5 = Much more (U.S. M = 2.89, SD = 1.04,Developing countries M = 4.11, SD = 1.12).
3.2.1.6.Exposure to mainstream media sources CBS, ABC, and NBC have been regarded as mainstream media outlets due to their persistent adherence to a less subjective and more impartial approach in their coverage of climate change (Boykoff 2008).Therefore, we measured exposure to mainstream news sources by asking participants, 'How often do you watch, listen to, or read content from the following?The national nightly network news on CBS, ABC, or NBC,' with responses ranging from 1 = Never to 7 = Many times a day.(M = 2.82, SD = 1.81) 3.2.1.7.Exposure to conservative news sources Fox News has a conservative bias (e.g., Feldman et al 2014, Feldman 2016).Therefore, we measured exposure to conservative news sources by asking participants, 'How often do you watch, listen to, or read content from the following?The Fox News Channel,' with responses ranging from 1 = Never to 7 = Many times a day.(M = 2.16, SD = 1.68).
3.2.1.8.Exposure to liberal news sources MSNBC has a liberal bias (e.g., Feldman 2016, Kaye and Johnson 2016, Padgett et al 2019;).Therefore, we measured exposure to liberal news sources by asking participants, , 'How often do you watch, listen to, or read content from the following?MSNBC,' with responses ranging from 1 = Never to 7 = Many times a day.(M = 1.86,SD = 1.43) 3.2.1.9.Religiosity To measure religiosity, participants were asked, 'How often do you attend religious services?',with response options ranging from 1 = Never to 6 = More than once a week.(M= 2.85, SD= 1.69) 3.2.1.10.Party Identification Party identification was measured using responses to two questions.First, we asked, 'Generally speaking, do you think of yourself as aK' with response options of Republican, Democrat, Independent, Other (Please specify), and no party/not interested in politics.If participants selected Independent or Other, they were asked a followup question, 'Do you think of yourself as closer to theK' with responses of 1 = Republican Party, 2 = Democratic Party, and 3 = Neither.These two items were combined to create a 5-point scale (1 = Republican, 2 = lean Republican, 3 = Independent/Other, 4 = lean Democrat, 5 = Democrat).If participants chose 3 = Neither in the follow-up question, responses were coded as 3 = Independent/Other on the combined scale.Likewise, if participants chose 1 = Republican Party or 2 = Democratic Party in the followup question, responses were coded as 2 = lean Republican and 4 = lean Democrat on the combined scale (M= 3.09, SD= 1.70).

Support for climate aid
The measure of support or opposition to climate aid was composed of two items.Participants were asked, 'How much do you support or oppose the following policies?','Provide financial aid and technical support to developing countries that agree to limit their greenhouse gas emissions', and 'Provide financial aid and technical support to developing countries to help them prepare for the impacts of global warming'.Responses to the question were on a four-point scale ranging from 1 = strongly oppose to 4 = strongly support.
The two measures were combined into a single variable for several reasons.Firstly, the measures are logically and conceptually similar as both represent dimensions of support for financial and technical assistance to developing countries.Hence, combining them into one measure captures different aspects of the same underlying construct: support for climate aid.Furthermore, the two measures are highly correlated (Spearman-Brown coefficient = .85,M= 2.72, SD= .98).Lastly, combining them into a single measure provides parsimony, allowing us tostreamline our reporting of results.To demonstrate the consistency in the results, we have included the regression results for each dependent variable individually in the supplementary material (Tables S1-S4).

Controls
The variables for age, race, gender, education, and income were provided by Ipsos Knowledge Panel.Table 1 displays descriptive statistics for age, gender, race, education, and income.Gender was recoded such that 0 = male and 1 = female, 49.9% of participants were male and 50.1% were female (unweighted percentages).Race was recoded such that 0 = White and 1 = Non-White, 75.4% of participants were considered White and 24.6% Non-White (unweighted percentages).

Analytical approach
All study hypotheses, research questions, and analyses were pre-registered at the Open Science Framework5 .The analyses presented in this manuscript deviate slightly from the pre-registered analysis plan.As discussed earlier, we combined the two dependent variables because they were conceptually similar, and highly correlated, which offered parsimony in reporting the results.To demonstrate the consistency in the findings, the regression results for each dependent variable were included in the supplementary material (Tables S1-S4).
To test our hypotheses and answer our research questions, we conducted both hierarchical linear regressions and relative weight analyses (RWA).First, we ran three separate hierarchical linear regressions: one which examined all registered voters in the sample, one which examined only Democrats (including those who lean Democrat) and one which examined only Republicans (including those who lean Republican).
Variables were entered hierarchically in five blocks in the three regression models.We entered the more fundamental variables in the earlier blocks-constructs that we believed would influence a wide range of beliefs, including climate beliefs-and the more distal or exogenous variables in the latter blocks.The first block of independent variables comprised demographic variables (age, gender, race, education, and income).The second block added religiosity and party identification.The third block added exposure to news sources: mainstream, conservative, and liberal.The fourth block added belief certainty, human causation, risk perception of the U.S., risk perception of developing countries, and worry.The fifth and final block added injunctive beliefs that the U.S. should do more and injunctive beliefs that developing countries should do more.
Our RWA analysis modeled the approach of Goldberg et al (2020), who utilized RWA to determine the predictors of U.S. public support for climate policy.In order to investigate the relative importance of the findings from our hierarchical regression model more deeply, we ran three RWA models by using the R code associated with the RWA web-based tool (Tonidandel and LeBreton 2015).Like in our regressions, we ran RWAs for all registered U.S. voters and separately for Democrats and Republicans.
RWAs are useful because they can add clarity to the relative importance of each variable in the models and address the multicollinearity issues related to multiple regression (Lebreton et al 2004, Tonidandel andLeBreton 2011).Further, RWAs provide both raw weights and rescaled weights as measures of effect size.The total model R2 is equivalent to the sum of the raw weights, the raw weights represent the percentage of variance explained in the dependent variable by each of the independent variables.The rescaled weights will always sum to 100%, as they represent the proportion of variance explained by each independent variable compared to that explained by the full model.
Listwise deletion was used to address missing data.The R code associated with the RWA web-based tool provides an option between listwise and pairwise deletion methods, we used listwise deletion to maintain consistency with the multiple regression models (missing n = 189).Further, we examined multicollinearity by assessing tolerance and Variation Inflation Factor (VIF) scores variables in all the models.All VIF scores were below 2.56, and all tolerance scores were above 0.39, indicating that multicollinearity is not a concern (Tabachnick and Fidell 2018).

All registered voters
Table 2 presents the regression findings for all registered voters, the full model explained 59% of the variance in support for climate aid to developing countries.The hypothesized relationships that were supported in the final model were: belief certainty (H1); worry (H3), perceived risk for developing countries (H4), injunctive beliefs of the U.S. (H5), exposure to conservative news sources (H9), and party identification (H11).The hypothesized relationships that were not supported in the final model were: human causation (H2), injunctive beliefs of developing countries (H6), exposure to mainstream news sources (H7), exposure to liberal news sources (H8), and religiosity (H10).The relationship between the perceived risk of global warming in the U.S. and support for climate aid was non-significant (RQ1, table 2).Regarding demographics, education was positively associated with support for climate aid, although income was negatively related (table 2).

Democrats and republicans
For RQ2, we conducted two separate multiple regressions to assess predictors of Democratic (n = 410) and Republican (n = 364) support for climate aid (table 3).The model for Democrats explained 21% of the variance in their support for climate aid.For Democrats, perceived risk in the U.S., worry, injunctive beliefs that the U.S. should do more, and education were positively related to support for climate aid (table 3).In contrast, income was negatively related (table 3).The Republican model explained 44% of the variance in support for climate aid.For Republicans, perceived risk for developing countries and injunctive beliefs that the U.S. should do more were positively related to support for climate aid (table 3).

All registered voters
In order of importance, our results suggest that the five most important predictors of support for climate aid are: worry, injunctive beliefs regarding the U.S., perceived risk for developing countries, belief certainty, and party identification (table 4).Together, these five predictors account for 37% of the variance explained in support for climate aid (raw weight), and 60% of the variance explained by the full model (raw weight).
All voters' regression and RWA results were largely consistent, with one exception (tables 2 and 4).The notable difference was that exposure to conservative news sources variable significantly predicted support for climate aid to developing countries in the final regression model but was not identified as an important predictor in the RWA.This suggests that exposure to conservative news sources provides additional information beyond the other variables in the full regression model, but it does not explain substantial variability in climate aid support in the RWA.

Democrats and republicans
We conducted separate RWAs for Democrats' and Republicans' support for climate aid.For Democrats, in order of magnitude, the five most important predictors of support for climate aid were: worry, belief certainty, injunctive beliefs regarding the U.S., perceived risk for the U.S., and perceived risk perception for developing  5).These five predictors accounted for nearly 18% of the variance explained in support for climate aid (raw weight), and 73.13% of the variance explained by the full model (rescaled weight).For Republicans, in order of magnitude, the five most important predictors of support for climate aid were: injunctive beliefs regarding the U.S., perceived risk for developing countries, worry, belief certainty, and perceived risk for the U.S (table 5).These five predictors accounted for nearly 33% of the variance explained in support for climate aid (raw weight), and 76.8% of the variance explained by the full model (rescaled weight).All variables were entered into the relative weight analysis simultaneously.Raw weights sum to the total R2; rescaled weights sum to 100.01%because it is a rounding issue.US. = United States, DCs = Developing Countries Note= There were differences in the regression and RWA results for Democrats and Republicans (tables 3 and 5).For Democrats, belief certainty and perceived risk perception for developing countries variables were important predictors in climate aid support in the RWA, however they were nonsignificant in the final regression model.This suggests that belief certainty and perceived risk perception for developing countries among Democrats explain substantial variability in climate aid support in the RWA, but they do not offer much additional predictive information beyond the other variables in the regression model (Tonidandel and LeBreton 2015).Conversely, education was significant in the full regression model but explained only a small amount of variance in the RWA.For Republicans, worry, belief certainty, and perceived risk for the U.S. were important predictors of climate aid support in the RWA but were nonsignificant in the full regression model.This suggests that worry, belief certainty, and perceived risk for the U.S. among Republicans explain substantial variability in climate aid support in the RWA, however, they do not offer much additional predictive information beyond the other variables in the regression model (Tonidandel and LeBreton 2015).

Discussion
We sought to identify and examine the predictors of support for climate aid to developing countries to help them limit their greenhouse gas emissions and prepare for the impacts of global warming.Overall, we found that issue-specific beliefs and attitudes about global warming were the strongest predictors of support for climate aid, with media use and demographics explaining relatively less variance.The one exception was party identification, which unsurprisingly was an important predictor of support for foreign assistance.Further, our results showed that there are important differences between Democrats and Republicans in terms of the strength with which certain factors predict their support for climate aid.
In the regression models and the RWA, worry about global warming emerged as a strong and important predictor associated with support for climate aid among registered voters.This finding supports previous research finding that people who are more worried about global warming are more inclined to support climate policies, including signing international treaties, and increasing taxes on gasoline (Smith and Leiserowitz 2014, van der Linden 2017, Bouman et al 2020, Goldberg et al 2020).The findings of the current research reinforce the importance of worry as a means to increase public support of climate policies.Practitioners could, for instance, use striking images (Goldberg et al 2019) of the impact of global warming in developing countries to increase worry and then follow with a message or image showing how helping less developed countries is also an investment in reducing global warming in one's own country.Future research might test whether showing people in developed countries the impacts of global in less developed countries increases support for climate aid.Injunctive beliefs that the U.S. should do more to tackle global warming were also a strong predictor of support for climate aid.This supports previous research that injunctive belief is a strong predictor of climate action in the U.S. (Roser-Renouf et al 2014, Roser-Renouf et al 2016).Strengthening injunctive beliefs that the U.S. should do more may be one of the most effective ways of increasing public support for climate aid.Therefore, practitioners could communicate about the role that the U.S. can and needs to play to address global warming.Future research might test how to most effectively increase people's belief that the U.S. should do more for developing countries.
Previous research found that global warming risk perceptions are strongly associated with support for climate policies (Goldberg et al 2020, Lee et al 2015, Leiserowitz 2006, Roser-Renouf et al 2014) and most people correctly view climate risks as being higher in developing countries than in developed countries (Kim andWolinsky-Nahmias 2014, Lee et al 2015).In this study we found that perceptions of global warming risks to developing countries had a positive association with support for climate aid.This finding implies that educating individuals about the harms of global warming in developing countries may improve public support for climate aid to reduce carbon emissions and prepare for climate impacts.Future research should test specific risk perception strategies that are most effective in increasing public support for climate aid to developing countries.
As hypothesized, we found that global warming belief certainty is related to support for climate aid to developing countries.Past research has identified belief certainty as a predictor of climate policy support generally (Bromley-Trujillo and Poe 2020, Goldberg et al 2020, Bumann 2021).This indicates that people who believe global warming is happening are more likely to support certain kinds of climate policies (Ding et al 2011, Hall et al 2018).Therefore, to increase climate aid acceptance among people, the importance of climate aid should be promoted, along with messages about how global warming is happening and how climate aid addresses global warming.Implementing messages with the goal of increasing belief certainty may be one way for climate communicators to help improve public support for climate aid in the U.S. Previous research found the importance of the media in influencing support for climate policy (Feldman et al 2014, Thaker et al 2017).We found that exposure to conservative news was negatively related to support for climate aid, but that exposure to liberal and mainstream news was not related.The negative relationship found in the present research is consistent with earlier studies on media exposure, which found that conservative media often rejects global warming and disagrees with climate science and scientists (Feldman et al 2012, Feldman 2016).Thus, we suggest that future research examine to what extent conservative news coverage about global warming addresses the topic of climate aid and how.This could provide helpful insights on how best to inoculate people against arguments opposing climate aid.
Party identification was associated with support for climate aid to developing countries, with Republicans showing less support for such policies than Democrats.This is further evidence of the political polarization of global warming among the American public (Lee et al 2015).Prior studies document how party identification relates to climate policies (Druckman et al 2013, Bolsen et al 2014), and this study extends evidence with support for climate aid to developing countries.Overall, our results reinforce the importance of considering party identification in the context of climate policies, this is further exemplified by the differences in predictors of support for climate aid between Democrats and Republicans.
The results of this study align with previous research finding that global warming risk perceptions are a strong predictor of policy support among Republicans and Democrats (Goldberg et al 2020).Interestingly, we found that Republicans' support of climate aid is more related to their perceptions of harm to people in developing countries, while Democrats' support is more related to their perceptions of harm to people in the U.S. Democrats had very high levels of perceived risk to people both at home and abroad, which may explain why only the domestic risk perception was significant.In contrast, Republicans had much lower levels of both forms of perceived risk.Our finding that Republicans are sensitive to the potential harm to people in developing nations is consistent with prior public health research showing strong Republican support for foreign aid during past health crises such as the HIV/AIDS epidemic (RepublicanViews.org.2015).Because of this, the perceived harm of global warming to developing countries may have prompted Republicans to consider climate impacts they had not previously thought, resulting in aid support.Future efforts to build support among Republicans for climate aid should consider communicating the harms to people in the developing world as well as harm to Americans and the U.S.
Among Democrats, worry about global warming was strongly associated with support for climate aid to developing countries.The results of our analysis reveal an interesting discrepancy.In the regression analysis, worry was found to be significant only among Democrats but not among Republicans.However, when we conducted the relative weight analysis, worry was found to be more important among Republicans than Democrats.One possible explanation for this discrepancy is that the relative weight analysis compares the importance of predictors relative to each other, regardless of their significance in the regression analysis.Also, the results emphasize the need for further research on the role worry plays in determining global warming climate aid support among Democrats and Republicans.

Limitations
The current study has limitations.First, regression analyses establish association but not causality.Hence, future experimental studies should investigate whether information designed to strengthen key predictors of support for climate aid actually do so.Second, the findings are based on a nationally representative sample of registered U.S. voters.Thus, the results cannot be generalized to other developed countries.Future research should replicate this study in other developed countries to help climate communicators make informed decisions when communicating to the global public about climate aid.
Additionally, we did not measure people's behavior, but rather relied on self-reported support for climate aid.Therefore, future work should explore the relationship between the significant predictors in this study and behaviors that demonstrate support for climate aid, such as personally donating money to help developing countries tackle global warming or voting for candidates who have a strong history of supporting climate aid.Moreover, previous research suggests the importance of measures of certainty or confidence about people's beliefs about global warming (Thaller and Brudermann 2020).Although we measured certainty of belief that global warming is happening, we did not have a similar measure of certainty in the belief that global warming is human-caused, although future research may benefit from such a measure.Finally, self-reported media exposure measures may overstate actual media use due to inaccurate or biased recall (Prior 2007, 2009, Tewksbury et al 2011).Future research may be able to improve accuracy by integrating self-reports with behavioral measures (Barthel et al 2020).

Conclusion
Global warming is disproportionately harming developing nations that have done relatively little to cause the problem.There is widespread agreement among the nations of the world that high-income nations must help low-and middle-income nations cope with this burden, although to date most wealthy developed countries have not fulfilled the commitments they made to low-and middle-income nations (Timperley 2021).Strengthening public support in developed countries for climate aid is one path to create the political will necessary for developed nations to meet their financial commitments to low-and middle-income nations.Our research points to possible avenues to strengthen public and political will through strategic communication.

(
Ballew et al 2019, Gustafson et al 2019, Goldberg et al 2020).This polarization has led to clear associations between climate opinions, party identification, and political ideology 4 (e.g., McCright et al 2013; Hornsey et al 2016).Republicans usually show lower support for climate policies than Democrats do (Fisher et al 2013, McCright et al 2013).

Table 2 .
Regression results for all registered voters.

Table 4 .
RWA results for all registered voters.

Table 5 .
RWA results for democrats and republicans.