Promoting Conservation Behaviors by Leveraging Optimistic and Pessimistic Messages and Emotions

Abstract

Meeting the challenge of the global biodiversity crisis requires evidence-driven communication strategies to engage public and political audiences in conservation. This study used a real-world conservation campaign to test how messages framed as pessimistic and optimistic, emphasizing species losses or gains, affected emotional response and behavioral intent, including willingness to donate and adopt specific conservation behaviors. We administered a national, web-based survey to a random sample of U.S. adults (n = 1,998). Respondents exposed to pessimistically-framed messages were significantly more willing to donate than those exposed to optimistically-framed messages, and emotions mediated effects. Intention to engage in behaviors was greatest when a respondent’s emotions aligned with the valence of the frame they received, such as when they experienced negative emotions in response to pessimistic frames or positive emotions in response to optimistic frames. These findings contribute to the growing body of evidence guiding the strategic use of communications to promote conservation.

Keywords:

• Communication
• conservation
• emotions
• framing
• messaging
• pessimistic

Introduction

Addressing the planet’s climate and biodiversity crises requires broad multistakeholder engagement across public and political arenas (Mace et al. 2018), but this is often stymied by ineffective communication tactics that ignore evidence-driven strategies (Kidd et al. 2019; Martell & Rodewald 2020a; Pidgeon and Fischhoff 2011). One successful strategy to improve the salience of messages is to subtly present certain aspects or information within a deliberate “frame” (Entman 1993). Effective framing is a core element of communication because it can set a train of thought into motion that highlights key points of the issue (Nisbet and Scheufele 2009). Though often confused with other mechanisms of audience influence (e.g., priming, persuading, and agenda setting), framing is invoked across many different contexts, disciplines and, when effectively deployed, can achieve similar outcomes (Cacciatore, Scheufele, and Iyengar 2016). Such outcomes are meaningful, as the amplification and spread of well-framed messages can influence attitudes, preferences, and behavior in ways that shape politics, and even influence subsequent legislation and policies (Lakoff 2004; Sheshadri, Hang, and Singh 2018). It is little surprise, then, that framing has long been used by the broader conservation and environmental communities to deliver a wide array of messages and plans. More recently, emphasis has been placed on testing and applying empirical insights from the broader framing literature to specific issues related to biodiversity conservation (e.g., Dean and Wilson 2023; Echeverri et al. 201; de Lange et al. 2022; Gregg et al. 2022; Kidd et al. 2019; Kusmanoff et al. 2020; Niemiec et al. 2020; Weinstein et al. 2015).

Optimistic vs. Pessimistic Framing

One approach to framing that has led to lively discourse in the conservation community is the effectiveness of optimistic versus pessimistic message framing (or negative vs. positive, see Jacobson et al. 2019), which draws upon core tenants of prospect theory (Kahneman and Tversky 1979). Prospect theory, also called “loss-aversion theory”, shows that logically equivalent frames (known as equivalency frames) about gains and losses are assessed and valued differently depending upon the context (see Asian disease study in which equivalent descriptions are presented in terms of lives lost or lives saved, Kahneman and Tversky 1982). People tend to be risk averse when gains are associated with a high degree of uncertainty, and most will trade off gains for greater certainty. In contrast, when there is uncertainty associated with losses, individuals may tolerate greater risk to avoid losses, perhaps because losses provoke stronger emotions. As distinctions between traditional types of frames have blurred (Cacciatore, Scheufele, and Iyengar 2016), loss and gain frames have been commonly deployed as negative and positive, or pessimistic and optimistic frames, respectively (Ledford and Nan 2020). Although positive and negative frames may draw upon prospect theory and concepts related to equivalency framing in framing theory (Druckman 2001), they primarily operate as “emphasis frames” (or “non-equivalency” frames) by drawing attention to specific aspects of an issue and- unlike a true equivalency frame- are not precisely logically equivalent.

Framing of Climate and Environmental Issues

Emphasis frames purposely use descriptor words or phrases that will influence preferences, opinions, and even decision making. For example, a well-established body of empirical work about communicating issues related to climate change indicates that both word choice and positive vs. negative framing can influence public perception and behavioral intentions (e.g., Hart et al. 2015, Gifford and Comeau 2011; Morris et al. 2020; Morton et al. 2011, Myers et al. 2012; Scannell and Gifford 2013; Spence and Pidgeon 2010; Xue et al. 2016). Moreover, an individual’s perceptions about climate change can be profoundly shaped by seemingly small differences in words, such as climate-science denier rather than climate skeptic, or biofuels rather than ethanol (O’Neill and Boykoff 2010; Cacciatore, Scheufele, and Shaw 2012). Past studies have also showed that Americans were more likely to believe in the evidence for, and serious consequences of, “climate change” rather than “global warming”, with Republicans being most susceptible to such wording differences (Schuldt, Konrath, and Schwarz 2011; Villar and Krosnick 2011). Notably, the strong and consistent reactions to these climate frames informed the Republican communication strategy during the George W. Bush Administration (Luntz 2002), and the Democrats subsequent attempts to reframe the debate (Lakoff 2004). More recently, as the political landscape has evolved, research has shown that Independents are more susceptible to framing effects, while those with strong partisan beliefs are most resistant (Benjamin et al. 2017).

Much of the research on strategic communication within the context of environmental behavior has focused on areas like recycling (e.g., White and Hyde 2012), energy use (e.g., Steinhorst et al. 2015), and consumer behavior (e.g., Sachdeva, Jordan, and Mazar 2015), as well as in framing perceptions of environmental risks related to emerging biotechnologies (Freedman 2013; Lomberg 2013; Petit, Needham, and Howe 2021; Zahry and Besley 2019), watershed protection (e.g., Marquina et al. 2022), and hydraulic fracking for gas extraction (Fry, Briggle, and Kincaid 2015; Clarke et al. 2015). Within this space, emerging work is starting to examine the role that optimistic and pessimistic message framing may play in influencing environmental perceptions or behavior (Jacobson et al. 2019; Morris et al. 2020; MacKinnon, Davis, and Arnocky 2022; Petit, Needham, and Howe 2021; Salazar et al. 2022). Yet despite the rich body of literature on framing of environmental behavior and risk, few studies have examined the applicability of findings to contexts like biodiversity conservation that seldom provide direct benefits to individuals and for which links between individual behavior and impact may be poorly defined (Reddy et al. 2017; Selinske et al. 2018; Kidd et al. 2019).

Framing in Donation Appeals

Research examining donation behavior has explored the efficacy of negative vs. positive charity appeals (see Erlandsson, Nilsson, and Västfjäll 2018 for a review). Whereas most negatively-framed charity appeals focus on unpleasant or undesirable events in hopes of persuading the target audience that the requested action will prevent such events from happening, positive appeals do the opposite (Putrevu 2014; Erlandsson, Nilsson, and Västfjäll 2018). Although positive charity appeals can be effective, negative appeals are more consistently successful (see Erlandsson, Nilsson, and Västfjäll 2018), perhaps because strong negative emotions, like sadness, may elicit more empathy and willingness to help or donate (Bagozzi and Moore 1994; Bagozzi, Gopinath, and Nyer 1999; Baberini et al. 2015). The comparatively greater success of negative appeals is also consistent with prospect theory which predicts that people are averse to, and will act to avoid, negative consequences (Kahneman and Tversky 1979). In other words, successful charity appeals, need to “harm” the reader by inducing negative emotions, such as sadness, guilt, anger, and general aversion, in order to motivate them to donate (Erlandsson, Nilsson, and Västfjäll 2018). In contrast, very little is known about the efficacy of messages that combine positive and negative frames in the same message (Ledford and Nan 2020).

Framing to Elicit Emotional Responses

Within this context, the strategic use of emphasis frames to elicit and leverage emotions in biodiversity conservation messages is surfacing as a promising area of inquiry, and one that may be an important tactic to engage people in wildlife conservation (Manfredo, Teel, and Henry 2009; Jain et al. 2019. Early examples demonstrate that eliciting negative (e.g., guilt and shame), positive (e.g., pride, compassion), or neutral emotions can change attitudes and encourage pro-conservation behaviors and environmental behaviors, like recycling (Bamberg and Möser 2007; Harth, Leach, and Kessler 2013; Harth, Leach, and Kessler 2013; Onwezen, Antonides, and Bartels 2013; Phillips et al. 2019). These seemingly mixed findings may be explained by the theory that emotional responses are guided by two complex, interacting, and unconscious decision-making pathways that create heuristics (mental shortcuts) to guide behavior: core affect (i.e., simple, raw emotional reactions or moods) and emotional disposition (i.e., more predictable emotions in response to a stimulus, as with fear of heights) (Finucane et al. 2000; Larson, Cooper, and Hauber 2016). Moreover, the degree to which emotions strengthen behavioral intent also may be a function of the type of behavior, specific emotion, and intensity of emotional arousal (e.g., Bamberg and Möser 2007; Harth, Leach, and Kessler 2013; Onwezen, Antonides, and Bartels 2013). Framing messages in negative or positive terms may also influence people’s emotional reactions to them. Recent research looking at physiological changes in audiences exposed to positively and negatively framed messages about wildlife and insects has shown that both frames are effective at arousing emotions, which can lead to further engagement with the content of the message (Jain et al. 2019).

The Present Study

Here we use a real-world, multi-organizational conservation campaign to investigate how optimistic and pessimistic messages that utilize emphasis frames, and emotional responses to these frames, could influence behavioral intentions. We specifically focus on intentions to donate to wildlife conservation causes and to engage in pro-conservation and political actions. The context of our research was a study published in Science that documented the loss of three billion birds, or over one-quarter of breeding populations, from North America since 1970 (Rosenberg et al. 2019). Following the article, several environmental and conservation-focused organizations launched a web-based communication campaign (3 Billion Birds; https://www.3billionbirds.org/) that promoted “Seven Simple Actions” that individuals could adopt to conserve birds (Figure 1). While the campaign was organized around the sobering loss of three billion birds, it also included hopeful messages that certain birds, like raptors and waterfowl, had increased. We were particularly interested in how these different, and broadly framed, pessimistic and optimistic messages might interact with each other and affect people’s intentions to engage in different conservation behaviors. By examining this question within the context of the 3 Billion Birds campaign, we were able to test a “live” message to provide real-time insights for guiding effective communication strategies.

Figure 1. Seven simple actions to conserve birds (https://www.birds.cornell.edu/home/seven-simple-actions-to-help-birds/.

We administered a national, web-based, experimental survey to test how behavioral intentions of individuals across four treatment groups receiving either (1) a pessimistic message that emphasized the severity of bird population losses, (2) an optimistic message about birds with increasing populations, (3) a combined message that include the previous pessimistic and optimistic frames, or (4) no message if in the control group.

We hypothesized that individuals who are exposed to pessimistic messages would be more likely to indicate behavioral intent to (a) donate to conservation causes and to (b) engage in pro-conservation and (c) political actions than those who are exposed to optimistic messages (H1), drawing upon evidence that negative appeals successfully motivate diverse behaviors (e.g., Baberini et al. 2015; Echeverri, Chan, and Zhao 2017; Erlandsson, Nilsson, and Västfjäll 2018; Witte and Allen 2000). We explored the mediating role of positive and negative emotions on the relationship between framing and a respondent’s intent to (a) donate to conservation causes and to (b) act in a pro-conservation manner, and (c) engage politically (RQ1). We also postulated that the behavioral intent of respondents exposed to the combined pessimistic and optimistic message would not be affected (H2). We speculated that the combined message would disrupt or confuse the respondent’s ability to emotionally regulate (Schneider and Schwarz 2017), which in turn would lead to a lack of willingness to engage in the three behaviors we studied.

Methods

Sample and Pretest

We collected data via an online, experimental survey administered in May 2020 to a national sample of US-based respondents (n = 1998) throughout all 50 states. The survey was managed and administered by the Survey Research Institute (SRI) at Cornell University who obtained the sample through Univox, an online paid survey community, for the e-mail panel survey. The sample was representative by age and gender across the four treatment groups using a between-subjects experimental design. Panel respondents were recruited via an email invitation that employed screening criteria to ensure respondents were over 18 years of age. All survey materials and administration procedures were reviewed and approved by the Cornell University Institutional Review Board (protocol #1912009295).

We received 2250 survey responses and used data from 1998 respondents. We excluded 161 surveys with incorrect attention check answers and 91 surveys in which answers to the variables we included in our analyses were categorized as “Refused/Not Answered.”

Of the 1998 survey responses used in our analysis, 479 respondents were presented the loss frame, 498 the gain frame, 478 the combined loss and gain frame, and 543 were in the control group that were not presented a frame. Our sample was predominantly white (n = 1,462, 73.2%), with 11.9% of respondents identifying as Black (n = 214) and 7.9% as Latino, Hispanic, or Spanish (n = 158). Respondents averaged 49 years old (SE = 0.798), and the majority were female (n = 1109, 55.5%), and college educated (n = 1,498, 74.9%), with over one-third (n = 705) holding a bachelor’s degree or higher. Additionally, 43.4% of respondents identified as male (n = 867), 0.5% (n = 10) as trans male, 0.3% as trans female (n = 5), and 0.2% as non-binary (n = 4). Politically, the sample leaned Democratic (n = 783, 39.2%), while 30.5% identified as Republican (n = 610) and 25% as Independent (n = 500). The sample was demographically balanced across the four treatment groups, including by age, gender, race, education, marital status, and political party. A summary of demographic variables by treatment is included in Table 1.

Table 1. Demographic detail across message frames (treatments).

Variable	Pessimistic Frame	Optimistic Frame	Combined Frame	Control
Age	48.82 (17.55)	49.59 (17.20)	48.64 (18.33)	49.08 (18.17)
Gender Identity
Male	219 (45.7%)	220 (44.2%)	212 (44.4%)	216 (39.8%)
Female	258 (53.9%)	274 (55%)	257 (53.9%)	320 (58.9%)
Trans Male	2 (0.4%)	3 (0.6%)	1 (0.2%)	4 (0.7%)
Trans Female	0 (0%)	1 (0.2%)	3 (0.6%)	1 (0.2%)
Non-Binary	0 (0%)	0 (0%)	4 (0.8)%	0 (0%)
Race
White	389 (81.2%)	390 (78.3%)	374 (78.2%)	421 (77.5%)
Black	55 (11.5%)	60 (12%)	55 (11.5%)	68 (12.5%)
Latino, Hispanic, or Spanish	37 (7.8%)	47 (9.4%)	43 (8.9%)	65 (12%)
Education
Bachelor Degree or higher	175 (36.6%)	173 (34.8%)	179 (37.5%)	178 (32.7%)
Political Party
Republican	127 (26.5%)	155 (32.4%)	155 (32.4%)	172 (31.7%)
Democrat	196 (40.9%)	195 (39.2%)	186 (38.9%)	206 (37.9%)
Independent	134 (28%)	125 (25.1%)	112 (23.4%)	129 (23.8%)

*Numeric variables are presented as Mean (SD), and categorical variables are presented as Count (%).

Prior to launching the survey, we conducted a pretest (n = 100) in March 2020 to test the timing, the reliability of the attention check question, and the reliability and validity of the measures. The pretest survey performed as expected. The pretest was administered by SRI and sent to respondents recruited through Amazon’s MTurk (www.mturk.com), which provides a more representative convenience sample of the general U.S. population for online sampling than other commonly used sampling methods (Berinsky et al. 2012).

Design and Experimental Stimuli

After consenting to the survey, respondents were asked some background questions and then randomly assigned to one of four treatments: (1) control (no message), (2) pessimistic message about bird population losses (3) optimistic message about bird population increases, and (4) a combined pessimistic and optimistic message about bird population losses and certain bird population increases (see exact language in Table 2). The optimistic message and the combined message both included the reminder that action must come soon at the end, as the three treatment versions were adapted from actual language on the Cornell Lab of Ornithology’s website (https://www.3billionbirds.org/). Each message ranged from 108 to 189 words. Respondents in the three experimental groups were then asked a knowledge-based, attention check question (Was the story you just read about: 1. Insects, 2. Birds, 3. Fish) after the experimental stimuli, as well as some questions about how what they read made them feel. The questions measuring emotion were adapted from the Cornell Climate Action Policy Survey (McComas, Stedman, and Sol Hart 2011). All respondents then completed a series of questions measuring their behavioral intentions (see below) before answering demographic questions. The control group was not presented with the attention check question or the question about how they felt and, instead, they simply completed the questions that measured their behavioral intentions.

Table 2. Language used for each of the four* message conditions (column headings).

Pessimistic message	Optimistic message	Pessimistic + Optimistic message
The first-ever comprehensive assessment of net bird population changes in the U.S. and Canada, published in Science in 2019, reveals across-the-board declines that scientists call “staggering.” All told, the North American bird population is down by 2.9 billion breeding adults, with devastating losses among birds in every biome. Forests alone have lost 1 billion birds. Grassland bird populations collectively have declined by 53%, or another 720 million birds. Habitat loss is likely to be a driving factor in these declines, say the authors, particularly agricultural intensification and development. The losses include favorite species seen at bird feeders, such as Dark-eyed Juncos (or “snowbirds,” down by 168 million) and sweet-singing White-throated Sparrows (down by 93 million). Eastern and Western Meadowlarks are down by a combined 139 million individuals. Even the beloved Red-winged Blackbird—a common sight in virtually every marsh and wet roadside across the continent—has declined by 92 million birds.	The first-ever comprehensive assessment of net bird population changes in the U.S. and Canada, published in Science in 2019, reveals that some groups of birds are currently doing well, and for good reason—governments and societies have invested in saving them. Raptors, waterfowl, and turkeys show what’s possible when commitments are made to bird conservation. These species numbers have increased to over 15 million raptors, 35 million waterfowl, and 250 million turkeys and grouse since 1970. As Rosenberg, the study’s lead author, points out, there are still 70% of North America’s birds left to spur a recovery if conservation measures can be implemented. But action must come soon.	The first-ever comprehensive assessment of net bird population changes in the U.S. and Canada, published in Science in 2019, reveals across-the-board declines that scientists call “staggering.” All told, the North American bird population is down by 2.9 billion breeding adults, with devastating losses among birds in every biome. Forests alone have lost 1 billion birds. Grassland bird populations collectively have declined by 53%, or another 720 million birds. Habitat loss is likely to be a driving factor in these declines, say the authors, particularly agricultural intensification and development. Not all the news out of this analysis is dire. Some groups of birds are doing well, and for good reason—governments and societies have invested in saving them. Raptors, waterfowl, and turkeys show what’s possible when commitments are made to bird conservation. These species numbers have increased to over 15 million raptors, 35 million waterfowl, and 250 million turkeys and grouse since 1970. As Rosenberg, the study’s lead author, points out, even if 30% of North America’s birds are lost, there are still 70% left to spur a recovery if conservation measures can be implemented. But action must come soon.

*The fourth message condition was the control group which did not receive any text or images.

Survey Measures

Respondents completed a series of questions that allowed us to explore the influence of the frames and emotions on behavioral intent, emotional intensity (experimental groups only), willingness to engage in the Seven Simple Actions (pro-conservation behaviors), willingness to engage in political actions to support conservation policy, and willingness to donate to bird, wildlife, and biodiversity causes.

Emotional Intensity

Respondents in the three groups that received the experimental stimuli (but not the control group) first completed seven items (sad, hopeful, anxious, cheerful, angry, afraid, and optimistic) that assessed the extent to which they experienced different emotions in response to the text (experimental stimuli) they had just read (“How does what you have just read make you feel?”). All items were assessed on a 5-point scale ranging from 1 (not at all) to 5 (very). Positive emotions (hopeful, cheerful, optimistic) were averaged to create an index termed Positive Affect (Cronbach’s α = 0.853), whereas negative emotions (anxious, afraid, angry, sad) comprised an index of Negative Affect (Cronbach’s α = 0.887). A bivariate correlation confirmed that the two measures are negatively correlated and statistically distinct (r(1443) = −0.357, p < 0.001).

Willingness to Engage in the Seven Simple Actions

Respondents in all four groups next completed seven items that assessed their willingness to engage in one of the Seven Simple Actions to helps birds, which we considered to be “pro-conservation behaviors”. The question was worded as following: In light of what you just read, how likely would be to engage in these behaviors? (items included: Make windows more visible to birds to prevent collisions (e.g. by putting designs on them), Keep my cat indoors, Plant native plants and reduce lawn to make more bird habitat, Avoid or reduce pesticide use, Buy certified shade grown coffee, a traditional growing method that uses little to no chemicals, creates habitat for birds, and improves soil, Avoid or reduce plastic use, and Birdwatch or count birds and share the info online (e.g. to citizen science projects like eBird or Project FeederWatch){5= Extremely likely to 1= Not at all likely}. These items were averaged to create a single-measure index of likelihood to engage in pro-conservation behaviors (Cronbach’s α = 0.854). All pairs of variables were significantly correlated using a Pearson correlation coefficient (p < 0.01).

Willingness to Engage Politically

All respondents then completed seven items that assessed willingness to act in support of wildlife conservation policy. The question was worded as following: In light of what you just read, how likely are you to engage in any of the following actions to support wildlife conservation policy? (items included: Contact a government official at any level of government, Attend a demonstration, rally, or march, Attend a government hearing (at any level of government), Write a letter to a public official or newspaper, Volunteer time to support an advocacy organization (e.g. an NGO, nonprofit, or community org), Volunteer time to support a political candidate with a strong environmental or wildlife conservation platform, and Volunteer time to support a wildlife conservation cause) {5= Extremely likely to 1 = Not at all likely}. These items were averaged to create a single-measure index of likelihood to engage in political behaviors to support conservation policy (Cronbach’s α = 0.944).

Willingness to Donate

We next created an index of the willingness of respondents to donate (WTD) to bird, wildlife, and biodiversity causes from the questions: Would you donate to: an organization that works to protect bird populations and habitat; an organization that works to protect wildlife and habitat; an organization that works to protect biodiversity) {yes, no}. These items were averaged to create a single-measure index of WTD (Cronbach’s α = 0.843).

Demographics

Respondents were asked to identify their age, gender identity, race/ethnicity, level of education, income, marital status, and political affiliation. Age was requested as an open-ended response (In answer to the question: What is your age?), while the remainder of the questions were fixed response. For gender identity, respondents chose one of six categories (male, female, trans male/trans man, trans female/trans woman, non-binary, not listed- open ended). Marital status included five categories (married, widowed, divorced, separated, never married). Race/ethnicity was modeled on the US Census question at the time (2020), while the political affiliation question was based on the question asked by American National Election Study (Generally speaking, do you consider yourself a Republican, Democrat, an Independent or what?, 2020).

Data Analysis

To test how framing and affective response influenced willingness to engage in certain conservation-oriented behaviors, we used two complementary approaches. First, to test our hypotheses (H1 & H2), we used linear regressions to test for main effects of message framing treatments on dependent variables reflecting behavioral outcomes related to pro-conservation, political, and donation intentions. Regression models also included several demographic attributes, as covariates, that are known to influence the behaviors of interest to us such as education and donation behavior (e.g., Lundberg et al. 2019), gender (e.g., Lee and Chang 2007; Petrovici, Golden, and Orazbek 2019), marital status (e.g., Lee and Chang 2007) and others. We used a backwards stepwise model selection to determine which variables to include in the final model. Second, because responses to message frames may depend upon emotional state (RQ1), we tested for mediation effects (i.e., emotion*frame interactions) using Model 4 of PROCESS (boot = 10,000 and 95% bias-corrected confidence intervals, Hayes and Preacher 2013). All models included demographic variables as controls. Last, all tests were conducted using SPSS version 28 (Chicago, Illinois).

Results

Message Effects

The first set of hypotheses predicted that individuals exposed to pessimistic messages are more likely to indicate behavioral intent to donate to conservation causes (H1a), and to engage in pro-conservation (H1b) and political actions (H1c) than those who are exposed to optimistic messages. The regression analyses revealed that pessimistic message framing influenced the intentions of respondents to donate to conservation causes (b = .09, SE = 0.03, p < 0.001) but not to engage politically (b = .09, SE = 0.07, p = 0.20) or to adopt other conservation behaviors (b = .04, SE = 0.06, p = 0.46). No other message frames differed significantly from the control.

To further investigate the effect of the pessimistic message frame on donation behavior, we ran pairwise comparisons of the treatments. We found that participants had an average difference between the pessimistic and optimistic frames of 0.075, indicating that the pessimistic frame was more likely to influence donation intentions than the optimistic frame (SE = 0.026, p < 0.004). Thus, H1 was only partially supported Table 4.

Table 3. Regression coefficients predicting donation, political engagement, and proconservation behavioral intentions.

Independent variable β (standardized coefficient)
	Donation	Political engagement	ProConservation
Pessimistic Frame	0.09***	0.08	0.04
Optimistic Frame	0.01	−0.07	−0.07
Combined Frame	0.02	−0.01	−0.01
Control^a	–	–	–
Age	−0.00*	−0.02***	−0.01***
Gender Identity	0.02	− 0.11*	0.18***
Race/Ethnicity: White	−0.01	−0.10	0.00
Race/Ethnicity: Black	0.03	−0.05	0.05
Race/Ethnicity: Hispanic, Latino	−0.01	0.05**	−0.11
Spanish
Education Completed	0.02**	−0.00	0.03
Household Income	0.03**	−0.01	0.01
Marital Status	0.01	−0.05**	−0.04**
Political Party	0.01	−0.08*	−0.01

^aAll coefficients are compared against the control group

*p < 0.05, **p < 0.01, ***p < 0.001

Table 4. Pairwise comparison of treatment for donation behavior.

	Pessimistic	Optimistic	Combined	Control
Pessimistic	–	0.075**	0.071**	0.087***
Optimistic	−0.075*	–	−0.005	0.012
Combined	−0.071**	0.005	–	0.016
Control	−0.087***	−0.012	−0.016	–

The values in the table represent the row variable minus the column variable (e.g., –0.075 is Optimistic minus Pessimistic).

*p < 0.05, **p < 0.01, ***p < 0.001

The second hypothesis speculated that the combined pessimistic and optimistic message would not affect behavioral intent of respondents (H2). The regression results did not show a significant relationship between the combined optimistic and pessimistic message frame and respondents’ behavioral intentions regarding donation behavior (b = 0.02, SE = 0.03, p = 0.53), political engagement (b = −0.02, SE = 0.07, p = 0.80) and pro-conservation behaviors (b = −0.01, SE = 0.06, p = 0.85) (Table 3). Thus, H2 was supported.

Mediation Effects

RQ1 explored whether emotions might mediate participants’ responses to message frames and subsequent behavioral intentions. Mediation analyses showed that positive and negative emotions do indeed mediate the relationship between message frame and behavioral intentions to engage in donation, political engagement, and pro-conservation behavior. When compared to the optimistic message frame, respondents reported greater negative emotions after exposure to the pessimistic message frame (b = 1.09, SE = 0.06, t(1441) = 18.30, p < .001) and the combined pessimistic and optimistic message frame (b = −0.33, SE = 0.06, t(1441) = −5.47, p < .001). When compared to the pessimistic message frame, respondents reported more positive emotions after exposure to both the optimistic message frame (b = 1.40, SE = 0.06, t(1441) = 24.65, p < .001) and the combined pessimistic and optimistic message frame (b = 0.81, SE = 0.06, t(1441) = 13.99, p < .001).

Donation Intention

We found evidence that negative emotions mediated the extent to which framing affected intention to donate (Figure 2). Our results showed that negative emotion was positively related to intention to donate (b = 1.09, SE = 0.01, t(1440) = 6.70, p < .001) when controlling for message frame. Likewise, intentions to donate were greater when those presented a pessimistic frame also experienced negative emotions (b = 1.09, SE = 0.06, t(1441) = 18.30, p < .001). Specifically, the indirect effect of the change of frame from pessimistic to optimistic (CI = [–0.1125, −0.0599], and from pessimistic to the combined pessimistic and optimistic frame (CI = [–0.0384, −0.0155]), were both significant, indicating that participants expressed fewer negative emotions after viewing the optimistic and combined frames, which reduced their intentions to donate. After accounting for this indirect effect across the frames, the direct relationship between frame and intention to donate was no longer significant (p = 0.76).

Figure 2. Mediating role of negative emotion on the pessimistic message frame and donation behavior (compared to the optimistic frame).

*p < 0.05, **p < 0.01, ***p < 0.001.

In contrast, we found no evidence that positive emotions mediated the intention to donate after exposure to messages. When controlling for message frame, positive emotion was not a significant predictor of donation intentions (b = −0.00, SE = 0.01, t(1440) = −0.13, p = .89). Although the direct effect on donation intention was significant when comparing the pessimistic frame to the optimistic frame (p = 0.02) and to the combined frame (p = 0.01), the indirect effects of the change of frame from pessimistic to optimistic (CI = [–0.3455, 0.0313]), and from pessimistic to the combined pessimistic and optimistic frame (CI = [–0.0199, 0.0181]), were not.

Political Engagement Intention

Mediation analysis showed that, after controlling for message frame, individuals experiencing negative emotions expressed greater intentions to engage in political behaviors (b = 0.39, SE = 0.03, t(1440) = 13.05, p < .001). Specifically, negative emotions increased the effectiveness of the pessimistic frame at raising intentions to engage politically (b = 1.09, SE = 0.06, t(1441) = 18.30, p < .001). Indirect effects further revealed that participants expressed fewer negative emotions after viewing the optimistic (CI = [–0.5114, −0.3427]) and combined (CI = [–0.1798, −0.0801]) frames, which decreased their intentions to engage in political behaviors and demonstrates mediation (see Figure 3).

Figure 3. Mediating role of negative emotion on the pessimistic message frame and political engagement behavior (compared to the optimistic frame).

*p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.

Conversely, we also found that positive emotions mediated the relationship between message frame and intentions to engage politically (b = 0.19, SE = 0.03, t(1440) = 5.85, p < .001; Figure 4). Indirect effects showed that the change of frame from pessimistic to optimistic (CI = [0.1706, 0.3699]) and from pessimistic to the combined pessimistic and optimistic frame (CI = [0.0961, 0.2177]), were also significant, indicating that participants who experienced high positive emotions, after exposure to the optimistic message frame, were more likely to express intentions to engage politically.

Figure 4. Mediating role of positive emotion on the optimistic message frame and political engagement behavior (compared to the pessimistic frame).

*p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.

Pro-Conservation Intention

Mediation analysis further showed that negative emotion was a significant predictor of proconservation behavior (i.e., Seven Simple Actions) when controlling for message frame (b = 0.32, SE = 0.02, t(1440) = 13.92, p < .001). Indirect effects demonstrated that the change of frame from pessimistic to optimistic (CI = [–0.4166, −0.2829]), and from pessimistic to the combined pessimistic and optimistic frame (CI = [–0.1465, −0.0663]), were both significant, indicating mediation. Direct effects indicated partial mediation when comparing the pessimistic frame to the optimistic frame (p < 0.001), and full mediation when comparing the pessimistic frame to the combined pessimistic and optimistic message frame (p = 0.34). This shows that participants expressed fewer negative emotions after viewing the optimistic and combined frames, which decreased their intentions to engage in proconservation behaviors (see Figure 5).

Figure 5. Mediating role of negative emotion on the pessimistic message frame and pro-conservation behavior (compared to the optimistic frame).

*p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.

After controlling for message frame, positive emotion was also a predictor of proconservation intentions (b = 0.14, SE = 0.03, t(1440) = 5.56, p < 0.001). The direct effect on intentions was significant when comparing the pessimistic frame to the optimistic frame (p < 0.001) and to the combined frame (p = 0.004), as were the indirect effects of the change of frame from pessimistic to optimistic (CI = [0.1179, 0.2759]), and from the pessimistic to the combined pessimistic and optimistic frame (CI = [0.0677, 0.1632]). This showed that participants who experienced high positive emotions, after exposure to the optimistic message frame, were more likely to express intentions to engage in proconservation behaviors (Figure 6).

Figure 6. Mediating role of positive emotion on the optimistic message frame and pro-conservation behavior (compared to the pessimistic frame).

*p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.

Discussion

Deployment of strategic, evidence-driven messages about the biodiversity crisis is more urgent than ever to inspire people to take actions that support conservation (Kidd et al. 2019; Kusmanoff et al. 2020; Gregg et al. 2021), but there continues to be disagreement about the efficacy of optimistic versus pessimistic message framing (McAfee et al. 2019). Though optimistic messages have become popular among many in the conservation field (McAfee et al. 2019; Papworth, Thomas, and Turvey 2019), our study found that pessimistic messages, framed to emphasize species loss, best provoked intentions to donate. We found no evidence that message framing alone increased willingness to engage in pro-conservation and political actions.

Strikingly, the influence of message frames on behavioral intentions depended upon the self-reported emotional states of respondents. Though intentions generally rose with the intensity of emotions, willingness to adopt conservation behaviors was greatest when respondents’ emotional state was congruent with the positive or negative valence of the message. In other words, intentions to engage in any three of the behaviors we studied was most likely when individuals either experienced (1) strong positive emotions when exposed to messages emphasizing species gains or (2) strong negative emotions when exposed to messages emphasizing population losses. This finding is reminiscent of the concept of “complementary” or “intersecting” frames (Schuldt, McComas, and Burge 2021), when multiple frames emphases build upon each other to support the same issue position (e.g., two pro-conservation frames that emphasize biodiversity loss and the effects of climate change). We therefore propose that such congruent emotions (i.e., negative + pessimistic or positive + optimistic) may have an additive effect that enhances the pessimistic or optimistic framing effect, much like complementary frames emphases support the same message position (Wise and Brewer 2010; Schuldt, McComas, and Burge 2021; Detenber et al. 2018; Ho 2021). Our results are consistent with the idea that emotional responses are guided by heuristics that influence decision making and suggest that emotional intensity drives whether these mental shortcuts are utilized (Finucane et al. 2000; Larson, Cooper, and Hauber 2016).

Our findings align with previous work demonstrating that negative emotions play a key role in influencing donation behavior (Anik et al. 2009; Penner et al. 2005). Past research has shown that negative emotions, like sadness, correspond with greater intentions to donate, suggesting that people are averse to negative consequences and will make decisions to avoid them (e.g., Kahneman and Tversky 1982; Cialdini 1991; Baberini et al. 2015). This also supports the idea that people are susceptible to negativity biases, in which negative, rather than positive, information is most likely to move people to action (Liu, Xin, and Lin 2014). Similarly, a meta-analysis by Witte and Allen (2000) showed that appeals framed with fear messages were most successful at changing behavior, especially when recipients believed they could take specific actions to mitigate the threat. Although some research suggests that people in a positive mood may be more willing to donate (Murat kyzy 1998), that happier people may act more altruistically (see Anik et al. 2009 for a review), and that positive charity appeals have been shown to be effective at eliciting donations (Burt and Strongman 2005; Putrevu 2014; and see Erlandsson, Nilsson, and Västfjäll 2018 for a review), our results did not support this. Therefore, we suggest that conservation fundraising appeals may be most successful when organizations strategically deploy messages that highlight negative outcomes (pessimistic messages) and elicit negative emotions.

The role played by emotions in our study is also consistent with previous work showing that emotions can drive behavior across a wide range of contexts – from assisting victims of child abuse to participating in citizen science projects (e.g., Bagozzi and Moore 1994; Bagozzi, Gopinath, and Nyer 1999; Carrus, Passafaro, and Bonnes 2008; Dean et al. 2018; Phillips et al. 2019) to engaging in political activism (Iyer, Schmader, and Lickel 2007; Ford et al. 2019; Skitka and Wisneski 2011; Tausch et al. 2011). Indeed, respondents were most likely to indicate intentions to donate, engage politically, or adopt at least one of the Seven Simple Actions when their emotional state was congruent with the positive or negative valence of the message they received.

Although combining loss and gain frames within the same message may seem like an effective hedging strategy, we found that combined frames were ineffective, even when emotions were elevated. We suggest that combined messages stymie action by disrupting the ability of respondents to emotionally regulate, and potentially creating attitudinal ambivalence, in which the simultaneous experience of both negative and positive emotions in response to a stimulus triggers aversion (Schneider and Schwarz 2017). Another possibility is that behavioral intentions are less likely to change in response to “competitive” frames, where multiple, opposing frames are presented in one message and lead to different interpretations of the issue or decision-making (Chong and Druckman 2007). Early research into competitive frames shows that opposing frames in one message can serve to neutralize and cancel out both framing effects (Borah 2011; Ho 2021). Our findings are consistent with this potentially counterproductive effect and, hence, we suggest that messages avoid simultaneously emphasizing both species losses and gains until the underlying mechanisms are better understood.

While our research points to strategies that can be used by conservation organizations, several study limitations should be considered. First, we relied upon self-reported behavioral intentions, rather than actual behavior. Although self-reports are reliable, cost-effective, and convenient in many cases, there can be discrepancies between self-reported and observed behaviors, especially as it relates to donation behavior (Kormos and Gifford 2014; Veríssimo et al. 2018). Second, surveys were administered during the COVID-19 pandemic, which raises the possibility that individuals responded differently than they might ordinarily. Third, our study design does not allow us to discern between reactions to visual versus written (text) messages, as both were used in our treatments. Fourth, because respondents in the control group did not receive the emotion measure (“How does what you have just read make you feel), we cannot directly attribute emotional state to exposure to a specific message frame. Fifth, the optimistic and combined message include language about government and social investment in bird conservation and a phrase that “action must come soon” that is not included in the pessimistic message treatment. This language may have primed or triggered an element of response-efficacy in respondents before they answered the subsequent questions measuring their behavioral intentions, however we saw no evidence of this in the results. Sixth, our study is only one, and additional studies are needed to ensure that the results are replicable. Seventh, like most framing research (Badullovich, Grant, and Colvin 2020), our U.S.-based sample, which was predominantly white and educated, was not representative of the American population overall. Culturally aware, multidirectional messaging is essential for engaging underrepresented and marginalized populations in order to include, and activate, all impacted communities in the critical fight to preserve biodiversity and save our planet. Research testing messages with, and in, diverse populations should be a priority for future framing scholars.

Our study builds upon a growing body of research examining how strategic, evidence-driven messaging can benefit biodiversity conservation and encourage pro-environmental behaviors (Kidd et al. 2019a, 2019b, 2019c; Jacobson et al. 2019; Kusmanoff et al. 2020; Martell & Rodewald 2020a; Martell & Rodewald 2020b; Gregg et al. 2021). Our results provide additional evidence that when messages are the primary communication tool, pessimistic frames may be most successful at encouraging donation intentions. Moreover, the efficacy of frames may be leveraged by eliciting emotions that align message valence. Collectively, these results may guide the strategic deployment of pessimistic and optimistic messages to effectively arouse emotions that translate to conservation actions.

Acknowledgments

We appreciate helpful insights provided by K. McComas, B. Lewenstein, S. Allred, D. Balog-Way, and M. Thomas, and the Cornell Lab of Ornithology for their funding support.