Can Emotions Capture the Elusive Gain-Loss Framing Effect? A Meta-Analysis

Abstract

Although recent streams of research have suggested that emotions play a key role in generating framing effects, little is known about the affective dimension of gain and loss framing and its potential impact on persuasion. The current study adopted a meta-analytical approach, synthesizing over 30 years of literature (k = 25, N = 5,772), to investigate this issue. The results indicate that message frame type directs the emotional response elicited in the audience, with gain frames inducing positive emotions (d = .31, p = .02) and loss frames inducing negative emotions (d = .22, p = .001). In turn, the experience of positive emotions enhances the influence of gain frames (b = .18, p = .045), whereas negative emotions augment the effects of loss frames (b = −.70, p = .01). These findings confirm that emotional responses may offer a pathway through which gain- and loss-framed messages exert persuasive influence. The study integrates the results with the emotions-as-frames perspective and proposes several promising avenues for future research.

Keywords

gain-loss frames emotions persuasion meta-analysis message design

Framing is among the most heavily referenced paradigms used to guide communication research, and gain-loss framing is among the most frequently employed framing strategies across a wide variety of communication contexts. From health communication campaigns that integrate framed messages to increase compliance with organ donation (e.g., Reinhart, Marshall, Feeley, & Tutzauer, 2007) and anti-smoking interventions (e.g., Zhao & Nan, 2010), to the framing of political statements regarding equal protection and free speech (e.g., Kimble & Wiener, 2016), to marketing campaigns designed to encourage travel (e.g., Seo & Dillard, 2016), to science communication messages designed to promote earthquake preparedness (e.g., Marti, Stauffacher, Matthes, & Wiemer, 2018) or willingness to sacrifice for climate change (e.g., Bilandzic, Kalch, & Soentgen, 2017), gain-loss framing has been embraced by researchers as a potentially powerful persuasive message construction.

Despite its wide acceptance, previous meta-analyses have shown gain and loss frames to have minimal differential direct persuasive effects (O’Keefe & Jensen, 2007, 2009). This finding has thus led scholars to consider the possibility that more substantial effects can be observed when taking into account moderating forces (e.g., topic and contextual factors, individual ideologies and prior attitudes; Quick & Bates, 2010). More generally, as research on gain-loss frames develops and matures, there is a need to look beyond the more simplistic direct effects to understand the process through which frames may influence both cognitive and affective responses, which, in turn, may influence attitudinal and behavioral outcomes. As argued by Hayes (2018), “answering such questions of how and when results in a deeper understanding of the phenomenon or process under investigation, and gives insights into how that understanding can be applied” (p. vii). As such, research on gain-loss framing would benefit greatly from the adoption of a more nuanced, process-oriented approach, examining more closely the moderators that may augment or attenuate effects, as well as the mediators that can intervene between frame exposure and persuasive effect.

One potentially important mediating variable is that of emotional response. Historically, framing research has focused primarily on cognitive explanations for ensuing effects (Scheufele & Iyengar, 2014). Yet, in recent years, the potential for emotions to explain framing effects has received growing theoretical (e.g., Nabi, 2003; Updegraff & Rothman, 2013) and empirical attention (Lerner & Keltner, 2001; Marti et al., 2018), which is especially timely given the extensive literature of the persuasive influence of emotions (e.g., Nabi, 2002). As noted by Kahneman (2011), “emotion now looms much larger in our understanding of intuitive judgments and choices than it did in the past” (p. 28).

Although a growing number of studies have investigated emotion’s role in framing effects, the literature lacks a clear synthesis of the existing evidence speaking to this issue. A meta-analysis can thus help determine whether emotions are indeed central to the process of influence stimulated by gain and loss frames. Further, such analysis can offer much needed enrichment of framing theorizing more generally, frequent calls for which have gone largely unanswered (e.g., D’Angelo & Kuypers, 2010; Reese, 2007; Scheufele, 1999), and provide insights into more effective persuasive message design (Pearl, 2014).

We begin first by reviewing extant theorizing linking emotions to framing effects and the arguments supporting its potential role as a mediating variable. Particular attention is given to the emotions-as-frames model (EFM; Nabi, 2003, 2007), in which emotions are conceptualized as lenses through which incoming stimuli are interpreted to ultimately generate emotion-consistent decisions and action. We then provide a meta-analytic synthesis of studies that focused on emotions as potential mediators of gain and loss framing effects on persuasive outcomes. In particular, this study examines the ability of gain and loss frames to induce positive and negative emotions, respectively, which, in turn, can augment the persuasive influence of the framed message. Theoretically relevant moderators are also considered. In essence, this study integrates the emotions-as-frames perspective with gain-loss framing to examine the extent to which gain and loss frames evoke emotions and their subsequent influence on persuasion outcome.

Gain-Loss Framing

The framing literature is quite vast. However, one particularly productive area of framing research, stemming from prospect theory (Kahneman & Tversky, 1979), has been the effect of gain-loss frames on decision-making. In Nobel Memorial Prize winning work, Kahneman and Tversky developed prospect theory to explain how people make real (vs. optimal) decisions in the face of probabilistic information about risk under conditions of uncertainty. Emphasis is placed on the value of the gains or losses to be experienced, with a key contribution being that even equivalent information can lead to quite different decisions, depending on whether that information is presented, or framed, in terms of benefits or costs. Although rooted in cognitive psychology, prospect theory has wide application to a host of real-world decision-making contexts that involve risk, including health, economic, and political decision-making. Indeed, its relatively early introduction in the broad scheme of the framing literature writ large and its far-reaching application make it a particularly fitting context for this investigation.

To elaborate, gain-framed messages make salient the positive outcomes likely to result from taking a certain action, whereas loss-framed messages highlight negative outcomes (Rothman & Salovey, 1997). For example, a gain-framed message about smoking behavior might state that “by smoking less, you can extend your lifespan,” whereas a loss-framed message would claim that “by smoking, you will shorten your lifespan.” In addition, gain-loss frames may show different kernel states or the basic state of the consequence described by the message (O’Keefe & Jensen, 2006). When gain-loss frames combine with kernel states, such as the attainment or avoidance of certain outcomes, there are four possible frames: (1) reaching a positive outcome (gain-framed), (2) avoiding a negative outcome (gain-framed), (3) suffering a negative outcome (loss-framed), and (4) missing out on a positive outcome (loss-framed).

By highlighting the potential benefits to be gained from an action, gain-framed messages have been argued to be well-suited to promoting risk-averse behaviors, or ones in which positive outcomes are more ensured. Given prevention health behaviors are considered risk-averse (i.e., they focus on actions that encourage one to maintain good health), their promotion via gain frames has been argued to be especially effective relative to loss frames, and some supportive evidence for this prediction has been found in the context of exercise and flossing (Rothman & Salovey, 1997). Conversely, by highlighting the potential risks to be incurred from an action, loss-framed messages have been argued to be well-suited to motivating risk-seeking behavior, in which greater risk is taken with the hope that a good outcome might result. Detection behaviors, in which one’s behavior may bring about awareness of disease and the associated unpleasant sequela, are expected to put one in the mind set of risk (Rothman & Salovey, 1997). As such, loss-framed messages are predicted to be more successful at encouraging illness detection behaviors, such as colonoscopies and mammograms (Meyerowitz & Chaiken, 1987; Rothman & Salovey, 1997). Of note, O’Keefe and Jensen’s (2006) meta-analysis found gain-framed appeals to be more persuasive for encouraging disease prevention behaviors, but gain- and loss-framed appeals did not differ significantly in persuasiveness for detection-related behaviors. This discrepancy may, in part, be attributed to issue involvement, as systematic processing of a framed message is a necessary precondition to observe the predicted advantage of gain framing for prevention behaviors and loss framing for detection behaviors (Rothman & Salovey, 1997). For instance, if people do not perceive themselves to be susceptible to HIV/AIDS, there is little reason for them to associate personal risk with HIV testing. As such, focusing on the potential negative outcomes associated with putting off HIV testing (i.e., loss frame) is unlikely to put one in a mind-set of risk.

When considering different framing techniques, it is essential to highlight the critical difference between equivalence and emphasis framing. Equivalent framing presents identical information, but through different lenses. For example, a 95% survival rate is equivalent to a 5% mortality rate, though one is clearly framed in terms of positive versus negative outcome. Emphasis framing, however, highlights certain elements or perspectives of a topic over others (Scheufele & Iyengar, 2014), for example, emphasizing the personal versus societal benefits of vaccinations. Given equivalent frames contain conceptually identical information, any differences in outcomes can be attributed to the phrasing or presentation of that information. This is not the case, however, with emphasis framing, which may precipitate different outcomes across conditions as a function of different message content. Although gain-loss framing from early prospect theory studies was based on equivalent frames, gain and loss frames have been operationalized in both ways in the extant literature.

Despite individual studies suggesting that gain or loss frames are more or less effective relative to each other, meta-analyses have shown that mean effects of gain and loss frames in persuasive messages across diverse contexts are quite small (O’Keefe & Jensen, 2007, 2009). This finding suggests that neither frame type is inherently more effective than the other and that moderating forces are likely at play (Quick & Bates, 2010). At a more conceptual level, though gain and loss frames may not differ in their direct influence on persuasive outcome, there are likely factors associated with these message devices that result in different processes of influence. To uncover these processes would be an important advance in understanding how frames persuade more generally. We focus on the emotions associated with those frames as one possible mechanism of influence.

Emotion and Framing Effects

The role of emotions in explaining framing effects has received growing theoretical attention (e.g., Updegraff & Rothman, 2013). Most directly relevant to this line of inquiry is Nabi’s (2003, 2007) EFM in which emotions are conceptualized as frames, or perspectives, through which incoming stimuli are interpreted. Specifically, when a message contains information that is relevant to an emotion’s core relational theme (e.g., imminent threat for fear; possible positive outcomes for hope), an emotion (i.e., fear, hope) is aroused. Once experiencing that emotion, the EFM predicts that individuals will both have more emotion-consistent information accessible from memory and seek out information related to the emotion’s motivational goals, which combined will generate emotion-consistent decisions and action.

Although other appraisal-based perspectives address the link between emotion and decision-making, they tend to be less directly linked to message framing contexts. For example, the appraisal tendency framework (ATF) suggests that prior emotional states predispose individuals to cognitively appraise subsequent events using emotion-consistent appraisal patterns (Lerner & Keltner, 2001). Although the ATF relates to decision-making outcomes, it focuses on the influence of incidental affect, or the emotions the receiver brings to the message context that are not typically relevant to the message topic. Further, that preexisting affect and the decision task may be non-message based. As such, the ATF is far better suited to contexts of judgment and evaluation rather than the message-based influence that is at the heart of framing research (Kim & Cameron, 2011).

The EFM, however, focuses specifically on message-induced emotion as a mediator between message framing and persuasive outcomes. Although the model suggests that emotions are frames themselves, it is clear that the emotions are stemming from particular message themes. That is, any message that captures the theme of a particular emotion could be conceptualized as falling within the broader category of an emotion frame. Given the wide array of frames identified in both the news and health contexts, in which both issue-specific and generic frames abound (e.g., de Vreese, 2005), the notion that they ultimately have influence because of the emotions they evoke by virtue of the appraisal patterns they generate is a more molar and parsimonious orientation to the study of framing and influence with theoretical, analytic, and practical value. Theoretically, the EFM highlights the potentially important role emotion may play in documented framing effects. Analytically, it directs researchers to include measures of emotion and place them as mediators in analyses, and practically, it suggests that message framing efforts should consider the core relational themes that messages likely reflect as they have clear implications for emotional arousal and subsequent effects.

Applying this perspective to gain-loss framing specifically, the EFM suggests that the presentation of information as such would generate discrete emotional experiences, which would, in turn, influence persuasive outcome. That is, loss-framed messages point out the harms one might incur as a result of action or inaction. Such framing likely captures the essence of fear (i.e., imminent harm), though it may also relate to other negative emotions, like sadness (i.e., irrevocable loss), guilt (i.e., violating an internalized moral code), or anger (i.e., demeaning offense). Conversely, gain-framed messages emphasize potential positive future outcomes, which associate with more positive emotions, most likely hope (i.e., fearing the worst while yearning for better) or happiness (i.e., making progress toward a goal; see Lazarus, 1991, for a description of a range of discrete emotional states.) The intensity of those emotional experiences is then predicted to influence relevant dependent variables. For example, greater fear evoked by a loss frame should translate into stronger framing effects—a supposition supported by meta-analyses of fear appeals (e.g., Tannenbaum et al., 2015). Finally, though the EFM emphasizes discrete emotions, the combination of emotions into assessments of positive and negative affect arguably follows the same principle. That is, message frames evoke affect, and it is the affective response that directs outcomes.

Recent research has supported the predictive power of the emotions-as-frames perspective, demonstrating that positive frames produce positive emotions, and negative frames produce negative emotions, and further that emotions mediate the relationship between frames and attitudinal or behavioral effects (e.g., Bilandzic et al., 2017). For example, research has found loss frames to generate greater fear, leading to greater willingness to sacrifice for climate change (Bilandzic et al., 2017) and greater intentions to have a colonoscopy (Lee-Won, Na, & Coduto, 2017). In a more complex mediation model, loss frames led to greater fear, which increased perceived message effectiveness, and ultimately generated greater behavioral intention to follow the message’s recommendation (Seo, Dillard, & Shen, 2013). Similarly, guilt has also been found to be associated with exposure to a loss-framed message in the context of smoking (Wong, Harvell, & Harrison, 2013) and climate change mitigation (Bilandzic et al., 2017). In both cases, the message-evoked guilt generated intentions to perform behaviors recommended in the message.

Positive emotions are less frequently tested as mediators between frames and persuasive outcomes, though supportive evidence exists. Bilandzic et al. (2017) and Nabi et al. (2018) found that exposure to gain frames in the context of climate change led to more hope, and, in turn, to greater message-consistent intentions and behavior. Of interest, Bilandzic et al. (2017) did not find hope to mediate the relationship between gain frames and negative kernel states, and Nabi et al. (2018) found no difference in effects between different kernel states. Yet, in contrast to hope, Major (2011) did not find happiness to mediate the link between gain-loss frames and attribution of responsibility for lung cancer and obesity.

In sum, there is only tepid support for gain and loss framing’s direct influence on persuasion outcomes. However, there is growing evidence seemingly in support of both the influence of gain-loss framing on emotions and the influence of emotions on attitudinal and behavioral effects, which implicates the mediating role of emotion. However, there have been some equivocal results regarding emotion’s role in producing outcomes in response to gain-loss framing. This meta-analysis thus aims to shed light on the emotion-related processes and effects of gain and loss frames across a host of contexts.

Based on this foundation, we assert the following hypotheses. First, given previous meta-analytic results (O’Keefe & Jensen, 2007, 2009) we expect the following:

Hypothesis 1 (H1): There is a minimal direct effect of exposure to gain-loss framing on persuasion outcome.

Loss-framed messages point out the harms one might incur as a result of action or inaction. Such framing is more likely to capture the core themes of negative emotions (e.g., imminent harm for fear, irrevocable loss for sadness). Conversely, gain-framed messages emphasize potential positive future outcomes, which are associated with the core themes underlying most positive emotions (e.g., hope, happiness). Several empirical tests of framing effects support this idea (e.g., Cho & Boster, 2008). Thus, we predict the following:

Hypothesis 2 (H2): Gain frames are associated with positive emotions, whereas loss frames are associated with negative emotions.

Assuming gain and loss frames evoke emotions as expected, the question of emotional intensity becomes salient. There is extensive evidence suggesting that the ability of gain-loss frames to induce emotions will be contingent on an individual’s level of involvement with a message. Indeed, the inconsistent effects of message framing are often attributed to differential levels of issue involvement (Rothman, Salovey, Antone, Keough, & Martin, 1993). In the persuasion literature, studies have repeatedly demonstrated that persons who are involved with an issue are more likely to carefully process the message, compared to their less involved counterparts (see Johnson & Eagly, 1990, for a review). To illustrate, there is no reason to expect that a loss-framed message regarding the risks of smoking would engender substantial levels of fear, sadness, or guilt among nonsmokers, as it would among smokers. By the same token, drivers are more likely to feel joy or hope as a result of exposure to a gain-framed message designed to promote safe driving as opposed to those who do not operate a vehicle. Following this logic, we expect the following:

Hypothesis 3 (H3): The effect of exposure to gain-loss framing on emotions is moderated by issue involvement, such that highly involved participants experience more intense emotions.

As to the effect of emotional intensity from gain and loss frames on persuasive outcome, if positive emotions are, in fact, consistent with and the logical outgrowth of exposure to gain-frame messaging, then we expect that gain-framed messages will evidence persuasive effect to the extent that they evoke stronger positive emotions, but less so if they evoke negative emotions, which are likely unintentional and counterproductive in this particular context. In contrast, assuming negative emotional intensity is consistent with loss-framed message exposure, we expect loss frames will evidence persuasive effect to the extent that they evoke stronger negative emotions, but not positive emotions. Thus, we predict that:

Hypothesis 4 (H4): The effect of exposure to gain frames on persuasion is contingent on the intensity of positively valenced emotions.

Hypothesis 5 (H5): The effect of exposure to loss frames on persuasion is contingent on the intensity of negatively valenced emotions.

After establishing the average persuasive effect of gain-loss frames and the specific emotions engendered, we then offer some exploratory analyses regarding the interplay between emotion-inducing frames and some message-related constructs that have been identified in previous literature as relevant to framing operationalization and effects as previously discussed. These include the following: frame equivalency (logical equivalency vs. emphasis framing), kernel state (desirable vs. undesirable), behavior type (detection vs. prevention), and message topic. Hence, a final research question is proposed:

Research Question 1: Is the effect of gain-loss framing on persuasion moderated by frame equivalency, kernel state, behavior type, or message topic?

Method

Sampling

Literature search

The literature search included the following steps. First, electronic databases (All Academic, Communication and Mass Media Complete, Educational Resources Information Center, Google Scholar, JSTOR, Medline, ProQuest, PsycINFO, and PubMed) were searched using relevant key terms (and their derivations), such as “emotions,” “gain-loss,” and “frame.” Those key terms were coupled with discrete emotions, such as “anger,” “fear,” “hope,” “guilt,” and “sadness.” This initial search yielded potentially relevant journal articles, books, dissertations, and conference papers that were screened for retrieval on the basis of their titles or abstracts. Second, reference lists were examined, and additional relevant studies were retrieved. Third, recent programs of leading conferences in communication, political science, public health, and social psychology (American Public Health Association [APHA], American Pharmacy Student Alliance [APSA], International Communication Association [ICA], National Communication Association [NCA], Society for Personality and Social Psychology [SPSP]) were systematically searched using the same key terms. We next shared a brief explanation of this project along with its inclusion criteria on two large list servers to solicit further relevant studies (Communication, Research, and Theory Network [CRTNET] and the SPSP). Finally, 15 leading scholars in the field of emotion and persuasion/judgment were contacted and asked to review our corpus and identify potential omissions (see Figure 1 for further details regarding the literature search strategy).

Figure 1.

Search strategy flowchart.

Inclusion criteria

Studies were selected based on four inclusion criteria: (1) include an experimental comparison between a condition exposed to a gain frame and an equivalent condition exposed to a loss frame¹; (2) include measurements of emotions that constitute a response to the framing manipulation, as opposed to the result of an ambient emotional induction (e.g., biographical recall); (3) provide a quantitative estimate for the effect of framing on message-relevant attitudes, intent, or behavior; and (4) report on relevant statistics (e.g., means, SDs, t values, exact p value, Cohen’s d, odds ratio, frequencies, zero-order correlations). When appropriate statistical data were missing (k = 9), relevant information was obtained from the corresponding authors. In total, the sample included 25 individual studies (24% unpublished),² with a total sample of 5,772 (M = 230.88, Mdn = 217, SD = 123.01).

Variable Coding

Outcomes

The effect size estimate employed was Cohen’s d, the standardized mean difference between the gain and the loss frame for a relevant research outcome (i.e., positive/negative emotional intensity and persuasive outcomes), allowing easy interpretation of both directionality and strength (O’Keefe, 2017). In addition, the transformation of effect sizes to Cohen’s d adjusts for different response scales, precision of measurement, and the size of the sample (Faraone, 2008). Given the current study focuses both on the ability of gain-loss frames to induce positive and negative emotions, as well as to affect persuasion-related outcomes, all effect sizes of exposure to gain-loss framing on positive emotions (k = 8), negative emotions (k = 22), attitudes (k = 13), intent (k = 20), and behavior (k = 2) were calculated per sample. Thus, this study is better understood as a combination of three separate meta-analyses (meta-analysis of framing on positive emotions, meta-analysis of framing on negative emotions, and meta-analysis of framing on persuasion). Exposure to loss frame was coded as a reference category. As such, positive and significant Cohen’s d coefficients indicate that the gain frame was superior to the loss frame in influencing positive emotions, negative emotions, attitudes, intent, or behavior, whereas negative and significant Cohen’s d coefficients suggest that exposure to the loss frame exerted more influence on these variables. Notably, the paucity of neutral control conditions (k = 3) in the framing literature has prevented us from assessing the efficacy of gain-loss frames in comparison with a no-frame control.

Moderators

Several theoretically driven variables were tested as moderators, including emotional valence, emotional intensity, and issue involvement. Broadly speaking, these moderators were grounded in the EFM (Nabi, 2003, 2007), as well as other relevant frameworks used to explain how frames affect persuasion (Cacciatore, Scheufele, & Iyengar, 2016; O’Keefe & Jensen, 2007, 2009). To protect against violations of independence of effect sizes, moderators were coded at the level of the study, treating the study as the unit of analysis (see Borenstein, Hedges, Higgins, & Rothstein, 2009; Hunter & Schmidt, 2004). When several emotions were measured within the same study, first, we coded the discrete emotions into two exclusive categories—positive and negative emotions. Then, we averaged the intensity of all positive emotions per study and all negative emotions per study. Later, information pertaining to positive and negative emotion was analyzed separately to account for potential dependency.

Following the taxonomy provided in Nabi (1999, 2002), discrete emotions were categorized based on their valence, with eight samples reporting on positively valenced emotions and 22 samples reporting on negatively valenced emotions. The specific discrete emotions measured included fear (k = 20), anger (k = 7), happiness (k = 4), guilt (k = 2), hope (k = 2), sadness (k = 1), and contempt (k = 1). Given the relative scarcity of nonfear discrete emotions, the analyses focused on emotional valence as opposed to discrete emotions. In addition, emotional intensity (based on manipulation checks for experienced emotions) was recorded. Specifically, emotional intensity was calculated as the standard difference in means (Cohen’s d) of experienced emotion between the gain-frame and loss-frame conditions ( $\bar{d} = . 35$ , SD = 0.27). Finally, when studies explicitly indicated that they took some measures to ensure that participants were involved with a message (e.g., focusing on female samples to study the framing of Pap test–related information in Balbo, 2010), samples were coded as highly involved (k = 13).

The coding also focused on a series of theoretically relevant variables that were identified as relevant for gain-loss frames, though not specific to the interplay between frames and emotions. These variables include the following: framing type (equivalent framing: k = 17, emphasis framing: k = 8), kernel state (desirable: k = 12, undesirable: k = 10, both: k = 3), and behavior type (prevention: k = 16, detection: k = 3, cessation: k = 1, various: k = 5). Additional exploratory potential moderators included message topic (health: k = 17, social/political: k = 1, science: k = 3, marketing: k = 3, various: k = 1), sample type (students: k = 12, nonstudents: k = 9, both: k = 1, N/A: k = 3), and gender (the relative percent of females in the sample: M = 68.55, SD = 19.96; see Table 1 for a summary of studies included in the meta-analysis and the online appendix for the study references).

Table 1.

List of Studies Included in the Meta-Analysis.

Study	Year	N	Outcome	Emotion	Intensity	Topic	Involvement
Balbo^a,b,c,d,e,f	2010	210	Intent	Negative	0.29	Pap test	High
Bassett-Gunter et al.^a,c,d,e,g	2014	75	Attitude	Negative	0.25	Physical activity	High
Bilandzic et al.^a,c,h,i	2017	247	Attitude/Intent	Negative/Positive	0.28/0.41	Climate change	Low
Block^c,d,e,f,g,h	1993	118	Attitude/Intent	Negative	0.06	Skin cancer	High
Cho and Boster^a,e,h	2008	246	Attitude/Intent	Negative/Positive	0.90/0.96	Substance abuse	High
Detweiler et al.^a,c,d,e,h	1999	217	Intent	Negative	0.03	Sun safety	High
Elbert and Ots^a,c,d,e	2018	258	Intent	-	0.03	Produce intake	Low
Jensen and Ratcliff^a,c,d,e,f	Under review	313	Attitude	Negative	0.001	Physical activity	Low
Lee and Cameron^a,d,e,g,h	2017	82	Attitude/Intent	Negative	0.36	Diet	Low
Lee-Won et al.^b,c,d,e	2017	140	Intent	Negative	0.16	Colonoscopy	High
Maheswaran and Meyers-Levy^a,c,e,g,h	1990	96	Attitude/Intent	Negative	0.20	Heart attack	Both
Mays and Zhao^a,c,h,e	2016	475	Intent	Negative	0.18	Indoor tanning	High
Meyerowitz and Chaiken^b,c,d,e,g	1987	44	Attitude/Intent/Behavior	Negative	0.42	Breast self-exam	High
Millar and Millar^a,c,d,j	2000	278	Intent	Positive	0.65	Driving	High
Nabi et al.^a,c,g,h,i	2018	337	Attitude/Behavior	Negative/Positive	0.60/0.70	Climate change	Low
Seo and Dillard^d,f,k	2016 #1	455	Intent	Negative/Positive	0.04/0.11	Free travel	Low
Seo and Dillard^d,f,k	2016 #2	455	Intent	Negative/Positive	0.13/0.14	Free tablet	Low
Seo and Dillard^d,f,k	Under review	424	Intent	Negative/Positive	0.25/0.38	Free travel	Low
Seo et al.^a,c,d,e,g	2013	252	Attitude/Intent	Negative	0.22	Flossing/Sun safety	Low
Spence and Pidgeon^a,c,g,h,i	2010	161	Attitude	Negative	0.56	Climate change	High
Stark et al.^c,g,h	2017	211	Attitude	Positive	0.79	Various	Low
Wong and McMurray^c,d,e	2002	70	Intent	Negative	0.38	Smoking cessation	High
Wong et al.^b,e,g,h	2013	152	Intent	Negative	0.74	Tobacco use	Low
Yu et al.^a,e,g,h	2010	213	Intent	Negative	0.49	Fetal alcohol spectrum	High
Zimmerman et al.^a,e,g,h	2014	243	Attitude/Intent	Negative	0.43	Marijuana use	High

Note. ^aPrevention; ^bdetection; ^cequivalence framing; ^ddesirable state; ^ehealth; ^funpublished; ^gstudent sample; ^hundesirable state; ⁱscience; ^jsocial/political; ^kmarketing. Elbert & Ots (2018) was not coded for emotional valence because the study only evaluated whether the framed message was perceived as emotional, as opposed to focusing on discrete or valenced emotions.

Intercoder reliability

One author trained an independent coder on a subsample of related studies that were not included in the final sample. Reliability was then calculated on 35% of the database using Krippendorff’s alpha. The overall reliability was satisfactory, with agreement ranging from 0.84 to 1.00.³ Disagreements were resolved through discussion.

Analysis

All analyses were conducted using the statistical package Comprehensive Meta-Analysis (CMA; Version 3; Borenstein, Hedges, Higgins, & Rothstein, 2005). The results were based on uncorrected estimates of random-effects models (Hedges & Vevea, 1998), computed by assigning more weight to the studies that carry more information based on the inverse of a study’s variance (Borenstein et al., 2005). Although some approaches to meta-analysis advocate for correction of estimates that potentially attenuate measurement artifacts (e.g., Hunter & Schmidt, 2004), numerous researchers have voiced their opposition to this logic, arguing that the

proper goal of a meta-analysis . . . is to teach us better what is, not what might some day be in the best of all possible worlds when all our independent and dependent variables are perfectly measured, perfectly valid, perfectly continuous, and perfectly unrestricted in range. (Rosenthal, 1991, p. 25)

After assessing the average effect of framing on relevant outcomes (H1-H2), heterogeneity was assessed using Q statistics. Then, moderator analyses were conducted to explore potential causes of heterogeneity (Ashford, Edmunds, & French, 2010). Specifically, for categorical variables (e.g., issue involvement), moderation analyses focused on Q statistics by comparing the mean variability in effect size estimates across different values of the moderator. For continuous variables (e.g., emotional intensity), a meta-regression was employed to test different moderators as potential predictors of effect sizes. For all moderator analyses, persuasion-related outcomes were combined and averaged per study (see O’Keefe, 2013, for data in support of utilizing combined estimates of attitudes, intent, and behavior, rather than individual outcomes). Due to concerns over statistical power, all moderation analyses were conducted only if there were at least five cases for each value (for a discussion of statistical power in meta-analysis, see Jackson & Turner, 2017).

Results

Framing, Emotion, and Persuasive Outcome

As predicted (H1), the average effects of exposure to gain-loss frames were weak and nonsignificant for all persuasive outcomes, including attitudes (d = −.02, confidence interval [CI] = [–.19, .16], p = .85, k = 13), behavioral intent (d = −.08, CI = [–.20, .04], p = .19, k = 20), and behavior (d = −.18, CI = [–.90, .55], p = .63, k = 2). Further, the framing effects on attitudes (Q(12) = 62.22, p = .001, I² = 80.71), behavioral intent (Q(19) = 74.54, p = .001, I² = 74.51), and behavior (Q(1) = 5.22, p = .02, I² = 80.84) exhibited significant heterogeneity, suggesting that the framing effects could potentially be contingent on emotions induced by the frame (see the online appendix for forest plots of framing effects on attitudes and behavioral intent).

In line with H2, compared with gain frames, exposure to loss frames had a significant effect on negative emotions (d = −.22, CI = [–.33, –.11], p = .001, k = 22), whereas exposure to gain frames was more likely to engender positive emotions (d = .31, CI = [.05, .55], p = .02, k = 8). Indeed, the effects retrieved from primary studies were overwhelmingly consistent with only two significant deviations from this pattern, that is, greater fear among gain-frame subjects in Meyerowitz and Chaiken (1987) and greater happiness/contentment among loss-frame subjects in Seo and Dillard (2016). Likewise, as predicted for intensity of negative emotions (H3), stronger effects, Q(1) = 2.49, p = .021, were recorded for highly involved participants (d = −.30, CI = [–.47, –.13], p = .001, k = 10) as opposed to lowly involved participants (d = −.16, CI = [–.30, –.02], p = .023, k = 12). Given the sample included only one study that assessed the influence of frames on intensity of positive emotions for highly involved individuals, H3 could not be tested for positive emotions in a statistically meaningful way.

To test the role of positive emotions in enhancing the effects of gain frames on persuasion (H4), a meta-regression was performed. The model included eight studies that measured positive emotions induced by exposure to gain- and loss-framed messages, attempting to predict framing effect sizes by the intensity of positive emotions. According to the model, the occurrence of positive emotions was a significant predictor of framing effects, enhancing the influence of gain frames on persuasion (b = .18, SE = .10, CI = [.01, 35], p = .045). The model accounted for 10% of the variance in framing effects (Q(1) = 2.56, p = .045). A similar meta-regression model tested H5, which suggested that negative emotions will augment the influence of loss frames on persuasion. The model included 22 studies that measured negative emotions induced by the gain-loss framed messages, treating the intensity of negative emotions as a predictor of gain-loss effect sizes. As hypothesized, negative emotions’ intensity enhanced the influence of loss frames on persuasion (b = −.70, SE = .28, CI = [–1.24, –.15], p = .01). The model explained 24% of the variance in gain-loss effect sizes (Q(1) = 6.34, p = .01). Figure 2 presents the results as a conceptual model of emotional gain-loss framing effects on persuasion.

Figure 2.

Conceptual model of the influence of emotional gain-loss frames on persuasion.

Moderator Analyses

Focusing on sample-related moderators, the analysis did not record a significant difference between gain- and loss-framed messages (Q(1) = 0.06, p = .81) by student samples (d = −.06, CI = [–.31, .19], p = .62, k = 12) and nonstudent samples (d = −.03, CI = [–.17, .11], p = .68, k = 9). Likewise, the gender distribution of the sample did not moderate gain-loss frames’ effect sizes (b = .001, SE = .03, CI = [–.008, .011]; Q(1) = 0.09, p = .77, k = 19).

With respect to message characteristics, framing type (emphasis vs. equivalency) significantly moderated the influence of gain-loss frames (Q(1) = 4.16, p = .04), with loss-framed messages being more effective when coupled with emphasis framing (d = −.54, CI = [–1.08, .01], p = .05, k = 8), whereas equivalency framing did not produce a difference in gain-loss framing effects (d = .05, CI = [–.08, .18], p = .40, k = 17). Next, the kernel state of the message was not a significant moderator (Q(1) = 0.31, p = .58), when gain- and loss-framed messages addressed both desirable (d = −.07, CI = [–.19, .06], p = .31, k = 12) and undesirable behaviors (d = .02, CI = [–.24, .27], p = .90, k = 10). Regarding message topic, there was a significant difference in effect sizes of gain-loss frames by message topic (Q(2) = 6.81, p = .03), with gain frames being more influential when communicating science-related information (d = .16, CI = [.01, .33], p = .048, k = 3). In contrast, health (d = −.14, CI = [–.30, .02], p = .09, k = 17) and marketing messages (d = −.01, CI = [–.14, .12], p = .85, k = 3) did not significantly moderate the effects of gain-loss frames. Finally, there was significant moderation by behavior type (Q(1) = 7.48, p = .006), as gain-framed messages were more effective for prevention behavior (d = .06, CI = [.01, .14], p = .05, k = 16), whereas loss-framed messages exerted more influence on detection behavior (d = −.80, CI = [–1.41, –.19], p = .01, k = 3).

Assessment of Publication Bias

Publication bias, or the file-drawer problem, can threaten the validity of meta-analytic results and lead to overestimation of effect sizes (van Assen, van Aert, & Wicherts, 2015). Given that published studies are easier to obtain and because published studies are more likely to report on significant results, it is important to investigate this phenomenon (Rosenthal, 1979). In recent years, however, various methods of estimating publication bias have drawn substantial criticism (Du, Liu, & Wang, 2017), reaching the conclusion that it is best to triangulate several different approaches when addressing the file-drawer problem. Accordingly, to check for possible bias, four steps were taken. First, publication bias was tested with Egger, Davey Smith, Schneider, and Minder’s (1997) regression test. The results indicated that the intercept from the analysis did not differ from zero (β = −2.05, p = .17), offering no support for a publication bias. Second, a potential bias was assessed with a funnel plot, where the effect sizes of all studies were plotted (on the x-axis) relative to a measure of standard error (on the y-axis). It is expected that if publication bias is present, then the funnel plot will show asymmetry, with small sample studies (those with larger standard errors) that found small effects missing from the funnel plot. Visual inspection of the funnel plot does not suggest a publication bias as studies that recorded small effects are highly represented (Figure 3). Third, Begg and Mazumdar’s (1994) rank correlation test, which computes an adjusted rank correlation between effect size and standard error for all included studies, did not record a publication bias (Kendall’s τ = −.15, p = .29). Finally, a Q test that compared effect sizes of published (d = −.03, CI = [–.16, .10], p = .68) and unpublished studies (d = −.07, CI = [–.22, .09], p = .41) from our data set did not find significant difference (Q(1) = 0.13, p = .72).

Figure 3.

Funnel plot for the detection of a potential publication bias.

Discussion

Gain-loss message framing has received considerable attention in a range of communication contexts, with much effort devoted to understanding their relative effects and the conditions under which they may be differentially influential. When O’Keefe and Jensen’s (2007, 2009) meta-analyses documented minimal evidence to support the differential effects of frame type on persuasive outcome, researchers began to look more closely at mediating and moderating variables that could potentially reveal gain- and loss-frame influence, including emotional intensity. The purpose of this meta-analysis was to assess the state of the literature to determine whether there is any merit to the notion that emotions might mediate the effect of gain-loss framing on subsequent persuasive influence and to consider other theoretically relevant moderators of framing influence.

Consistent with past meta-analyses of gain-loss framing generally, the two message frames did not differentially influence persuasive outcome. Although exposure to gain- and loss-framed messages did not result in a global effect on persuasion, consistent with predictions, gain frames were associated with the arousal of more positive emotions and loss frames were associated with the arousal of more negative emotions. As expected, emotional intensity was proportional to the level of involvement with the issue, with stronger negative emotions emerging for highly involved participants. Finally, gain-framed messages were more effective when they were able to produce more intense positive emotions, whereas loss-frame message effects were facilitated by the experience of stronger negative emotions. In other words, variations in emotional intensity were associated with corresponding variation in framing effects, such that gain-framed messages were more effective when individuals experienced more positive emotions and loss-framed messages gained their potency from intense negative emotions. Thus, emotional intensity appears to be a pathway through which both gain and loss frames exert persuasive influence.

The support for emotional intensity as a potential mediator of framing effects is significant both theoretically and practically. Theoretically, these findings are consistent with the EFM (Nabi, 2007), in that the nature of the message content and presentation (i.e., focus on benefits vs. consequences) generates unique message appraisals and thus emotional responses, the intensity of which directs persuasive outcome. However, the EFM was not designed with gain-loss framing specifically in mind. As such, it fails to consider unique elements of this context. Most especially, how does the presumed association of risk aversion with gain frames and risk-seeking with loss frames, which originated from the articulation of prospect theory (Kahneman & Tversky, 1979), relate to the emotions engendered by such frames? And are there implications for how the produced emotion influences message engagement in other ways, for example, processing depth, perceived message effectiveness, and the like? A closer consideration of the psychological state induced by gain-loss framing would be helpful in building theory around its process of influence via emotional intensity.

From a practical standpoint, these findings suggest that incorporating appeals to positive emotions within gain frames and appeals to negative emotions within loss frames may enhance effects. This could be accomplished in multiple ways, for example, through more vivid language or images, narrative, inclusion of efficacy information (for positive emotions), or risk information (for negative emotions). By identifying message elements that could intensify the desired emotional response matched to the frame condition, the effects of both gain- and loss-framed messages may be enhanced. Likewise, more careful attention should be given to pilot studies and formative research to ensure that the framed message is able to (1) induce a congruent emotion (positive emotions with gain frames and negative emotions with loss frames) and (2) generate sufficient emotional intensity to affect a desired outcome.

The moderator analyses also revealed interesting findings. The notion of prevention behaviors being linked to gain frames is consistent with past research and reviews (see Rothman & Salovey, 1997). However, the difference between equivalency versus emphasis framing was especially notable. It is logical that equivalent information would have similar effects, regardless of framing type. This is fully consistent with past meta-analytic results (O’Keefe & Jensen, 2009). The greater influence of emphasis frames when presented through a loss frame, however, was interesting. Perhaps, when researchers allow information to vary, the skew is toward including more powerful negative information in the loss-framed message, leading to more intense negative emotions, which we found to increase loss-frame effectiveness. This is, of course, just a supposition that should be assessed empirically.

More generally, these findings have broader theoretical and practical implications. Theoretically, the idea that differently framed equivalent information evokes different emotions, but those frames have comparable downstream effects, highlights the important role emotion plays in the process of message effects. In other words, even when messages are logically identical and the only varied information is the presentation of risk (Druckman, 2004), people’s emotional reactions significantly differ based on whether the information is presented in terms of gains or losses. Thus, one could argue that more drastic message variations are needed to affect message-consistent attitudes but even minimal variations in the framing of risk are enough to induce message-consistent emotions. Further, in line with recent evidence that emotions (e.g., hope) can influence behavior without changing attitudes, it might be the case that the capacity of equivalent gain- and loss-framed messages to induce relevant emotions can directly translate into behavioral change, without developing a favorable attitude toward the behavior (Nabi & Myrick, 2019). Although this remains speculative, future studies should look closely at the mechanism of equivalency framing, its ability to induce emotions, and its subsequent influence.

Similarly interesting, the data revealed science information was better off framed in terms of gain but there was no difference between frame types for health information. This could be an anomaly given only three science information studies were identified for inclusion in the analyses. But there might also be systematic differences between message context types that could be relevant. For example, some contexts, like climate change, might be heavily biased toward risk aversion such that gain frames resonate more squarely. A close consideration of how message context may suggest alignment with one frame type over another would be useful.

There are clearly additional questions that remain unanswered and unaddressed in these analyses. Most pressing is the question, do discrete emotions influence the success of gain-loss frames? Unfortunately, the state of the literature precludes the opportunity for addressing this question well as fear was the only emotion to be assessed in more than 10 studies. As such, a meta-regression model with each discrete emotion entered separately revealed no particular emotion as predictive of loss or gain-frame effectiveness. Future research that takes a closer look at the effects of particular positive and negative emotions within the gain- and loss-frame contexts is essential for us to gain a more complete view of the role of emotion generally, and particular emotions specifically.

Further, given that understanding the process by which gain and loss frames may influence persuasive outcomes can offer tremendous insight into how to effectively design such messages, future research would do well to consider other potential mediators or moderators of effects. For example, unlike the carefully constructed scenarios developed in tests of prospect theory, gain- and loss-framed messages in health or science, for example, have more complex structures. They may include narrative structure and appealing characters. They may be attributed to more or less credible sources. They may even contain multiple frames. For example, a problem (e.g., climate change) may be framed with individual or corporate attribution of responsibility, but the solution may be gain or loss framed. Indeed, across all these message elements, different emotions are likely to be evoked, as suggested by the concept of emotional flow (Nabi, 2015). As such, the emotions aroused in response to the gain- or loss-framed information should be further contextualized within the broader emotional flow experience of the full message.

At this point, it might be worthwhile to raise the possibility that gain-loss frames are simply a special case of an emotional appeal. Indeed, the focus on gains/losses appears to serve an emotion-inducing function similar to that achieved through exposure to emotional facial expressions and recall of emotional memories. Keeping in mind that emotional persuasive appeals are often conceptualized as messages attempting to arouse emotions to promote precautionary motivation and self-protective action (Rogers, 1983), the gain-loss framing effects seem to fit this role. In fact, Riet, Ruiter, Werrij, Candel, and De Vries (2010) have argued that loss frames might be conceptualized as fear appeals. The findings of the current meta-analysis support this contention, especially when comparing the moderate-large effects of gain-loss frames on message-consistent emotions and the negligible effects on other research outcomes.

Ultimately, this research not only supports the critical role emotion plays in the influence of gain and loss frames, but it also supports the EFM’s heuristic and analytical value for future message framing research more generally. First, it highlights the value and importance of including a range of emotion measures in framing research. This is critical as gain-loss frames could evoke a range of different emotions to different effects. For example, a loss frame that focuses on threat would be expected to elicit fear, whereas a loss frame focused on negative outcomes caused unjustly would likely elicit anger. Relatedly, a gain frame emphasizing personal achievement would encourage pride, whereas a gain frame highlighting the potential for improved circumstance in the future would elicit hope. Such research would not only allow for additional testing and development of the EFM but more importantly would help illuminate the process through which a host of different frame operationalizations in contexts ranging from politics to health and science communication and beyond have influence. If emotions prove to be critical mediators, the EFM offers an economical and parsimonious model for explaining such findings and can aid in future message framing selection.

To conclude, though individual studies have found differential persuasive advantage between gain- and loss-framed messages, the evidence is clear that as a general rule, they do not differ in their direct persuasive influence. This meta-analysis, however, demonstrates clearly that gain- and loss-framed messages do systematically differ in the emotions they engender and those emotions link directly to persuasive outcomes. As such, the most meaningful impact gain-loss framing may have is in the emotional responses they produce.

Supplemental Material

Appendix_-_Forest_plot_figures – Supplemental material for Can Emotions Capture the Elusive Gain-Loss Framing Effect? A Meta-Analysis

Supplemental material, Appendix_-_Forest_plot_figures for Can Emotions Capture the Elusive Gain-Loss Framing Effect? A Meta-Analysis by Robin L. Nabi, Nathan Walter, Neekaan Oshidary, Camille G. Endacott, Jessica Love-Nichols, Z. J. Lew and Alex Aune in Communication Research

Supplemental Material

Supplement_1_-_Included_Study_References – Supplemental material for Can Emotions Capture the Elusive Gain-Loss Framing Effect? A Meta-Analysis

Supplemental material, Supplement_1_-_Included_Study_References for Can Emotions Capture the Elusive Gain-Loss Framing Effect? A Meta-Analysis by Robin L. Nabi, Nathan Walter, Neekaan Oshidary, Camille G. Endacott, Jessica Love-Nichols, Z. J. Lew and Alex Aune in Communication Research

Footnotes

Acknowledgements

The authors are particularly indebted to Daniel O’Keefe for his invaluable advice and kind support. Likewise, the authors gratefully acknowledge the Editor and three anonymous reviewers for their helpful insights and constructive comments.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Robin L. Nabi

Nathan Walter

Supplemental Material

Supplemental material for this article is available online.

Notes

Author Biographies

Robin L. Nabi (PhD, University of Pennsylvania) is a professor of communication at the University of California, Santa Barbara (UCSB). Her research interests focus on discrete emotions’ influence on media message processing and effects. She has published over 75 articles and book chapters and co-edited the SAGE Handbook of Media Processes and Effects. She has served as a managing editor of the journal Media Psychology, as associate editor of the Journal of Communication, and on the editorial board of numerous top communication journals. She is also a fellow of the International Communication Association.

Nathan Walter (PhD, University of Southern California) is an assistant professor in the School of Communication at Northwestern University. His research concerns the effects of mass media, evaluation of strategic health messages, naive theory, and correction of misinformation.

Neekaan Oshidary is a graduate student in the Department of Communication at UCSB. His interests lie in wellbeing and media use, as well as media multitasking and study habits. He has a master’s degree in symbolic systems (cognitive science) and a BA in psychology from Stanford University.

Camille G. Endacott is a PhD candidate in the Department of Communication at UCSB with an emphasis on organizational communication. Her research interests center on organizational and occupational identity issues in emerging work arrangements. She earned her BA in communication studies from Azusa Pacific University in 2015 and her MA in communication from UCSB in 2017.

Jessica Love-Nichols is a PhD candidate in linguistics at UCSB. Her research investigates climate change communication and activism within rural, western communities in the United States. She received her bachelor’s degree in linguistics from Reed College and holds an MS in teaching from Pace University and an MA in linguistics from UCSB.

Z. J. Lew is a PhD student in the Department of Communication at UCSB. His research interests focus on social influence, interpersonal communication online, communication technology, and computer-mediated communication.

Alex Aune is a PhD student in the Department of Communication at UCSB. He primarily studies evolutionary psychology and how that perspective can inform our understanding of communication phenomena. He received his BA in communication from the University of Arizona and MA in communicology from the University of Hawaii.

References

Ashford

Edmunds

French

D. P.

(2010). What is the best way to change self-efficacy to promote lifestyle and recreational physical activity? A systematic review with meta-analysis. British Journal of Health Psychology, 15, 265-288. doi:10.1348/135910709X461752

Balbo

(2010). The message framing of health communications: How to elicit higher intention to get an annual pap test? Retrieved from https://halshs.archives-ouvertes.fr/halshs-00534779/

Begg

C. B.

Mazumdar

(1994). Operating characteristics of a rank correlation test for publication bias. Biometrics, 50, 1088-1101. doi:10.2307/2533446

Bilandzic

Kalch

Soentgen

(2017). Effects of goal framing and emotions on perceived threat and willingness to sacrifice for climate change. Science Communication, 39, 466-491. doi:10.1177/1075547017718553

Borenstein

Hedges

Higgins

Rothstein

(2005). Comprehensive meta-analysis version 2.0. Englewood, NJ: Biostat.

Borenstein

Hedges

L. V.

Higgins

J. P. T.

Rothstein

H. R.

(2009). Introduction to meta-analysis. Chichester, UK: John Wiley.

Cacciatore

M. A.

Scheufele

D. A.

Iyengar

(2016). The end of framing as we know it . . . and the future of media effects. Mass Communication and Society, 19, 7-23. doi:10.1080/15205436.2015.1068811

Cho

Boster

F. J.

(2008). Effects of gain versus loss frame antidrug ads on adolescents. Journal of Communication, 58, 428-446. doi:10.1111/j.1460-2466.2008.00393.x

D’Angelo

Kuypers

J. A.

(Eds.) (2010). Doing news framing analysis: Empirical and theoretical perspectives. New York, NY: Routledge.

10.

de Vreese

C. H

. (2005). News framing: Theory and typology. Information design journal & document design, 13, 51-62.

11.

Druckman

J. N.

(2004). Political preference formation: Competition, deliberation, and the (ir)relevance of framing effects. American Political Science Review, 98, 671-686. doi:10.1017/S0003055404041413

12.

Liu

Wang

(2017). A Bayesian “fill-in” method for correcting for publication bias in meta-analysis. Psychological Methods, 22, 799-817. doi:10.1037/met0000164

13.

Egger

Davey Smith

Schneider

Minder

(1997). Bias in meta-analysis detected by a simple, graphical test. British Medical Journal, 315, 629-634. doi:10.1136/bmj.315.7109.629

14.

Elbert

S. P.

Ots

(2018). Reading or listening to a gain- or loss-framed health message: Effects of message framing and communication mode in the context of fruit and vegetable intake. Journal of Health Communication, 23, 573-580. doi:10.1080/10810730.2018.149305

15.

Faraone

S. V.

(2008). Interpreting estimates of treatment effects: Implications for managed care. Pharmacy & Therapeutics, 33, 700-703.

16.

Hayes

A. F.

(2018). Introduction to mediation, moderation, and conditional process analysis: A regression-based perspective (2nd ed.). New York, NY: The Guilford Press.

17.

Hedges

L. V.

Vevea

J. L.

(1998). Fixed- and random-effects models in meta-analysis. Psychological Methods, 3, 486-504. doi:10.1037/1082-989X.3.4.486

18.

Hunter

J. E.

Schmidt

F. L.

(2004). Methods of meta-analysis: Correcting error and bias in research findings (2nd ed.). Newbury Park, CA: SAGE.

19.

Jackson

Turner

(2017). Power analysis for random-effects meta-analysis. Research Synthesis Methods, 8, 290-302. doi:10.1002/jrsm.1240

20.

Johnson

B. T.

Eagly

A. H.

(1990). Involvement and persuasion: Types, traditions, and the evidence. Psychological Bulletin, 107, 375-384. doi:10.1037/0033-2909.107.3.375

21.

Kahneman

(2011). Thinking, fast and slow. New York, NY: Macmillan.

22.

Kahneman

Tversky

(1979). Prospect theory: An analysis of decision under risk. Econometrica, 47, 263-292. doi:10.2307/1914185

23.

Kim

H. J.

Cameron

G. T.

(2011). Emotions matter in crisis: The role of anger and sadness in the publics’ response to crisis news framing and corporate crisis response. Communication Research, 38, 826-855. doi:10.1177/0093650210385813

24.

Kimble

K. M. K.

Wiener

R. L.

(2016). Equal protection versus free speech rights: When gains loom larger than losses. Applied Cognitive Psychology, 30, 722-735. doi:10.1002/acp.3247

25.

Lazarus

R. S.

(1991). Emotion and adaptation. New York, NY: Oxford University Press.

26.

Lee-Won

R. J.

Coduto

K. D.

(2017). The effects of social media virality metrics, message framing, and perceived susceptibility on cancer screening intention: The mediating role of fear. Telematics and Informatics, 34, 1387-1397. doi:10.1016/j.tele.2017.06.002

27.

Lerner

J. S.

Keltner

(2001). Fear, anger, and risk. Journal of Personality and Social Psychology, 81, 146-159. doi:10.1037/0022-3514.81.1.146

28.

Major

L. H.

(2011). The mediating role of emotions in the relationship between frames and attribution of responsibility for health problems. Journalism & Mass Communication Quarterly, 88, 502-522. doi:10.1177/107769901108800303

29.

Marti

Stauffacher

Matthes

Wiemer

(2018). Communicating earthquake preparedness: The influence of induced mood, perceived risk, and gain or loss frames on homeowners’ attitudes toward general precautionary measures for earthquakes. Risk Analysis, 38, 710-723. doi:10.1111/risa.12875

30.

Meyerowitz

B. E.

Chaiken

(1987). The effect of message-framing on breast self-examination attitudes, intentions, and behavior. Journal of Personality and Social Psychology, 52, 500-510. doi:10.1037/0022-3514.52.3.500

31.

Nabi

R. L.

(1999). A cognitive-functional model for the effects of discrete negative emotions on information processing, attitude change, and recall. Communication Theory, 9, 292-320. doi:10.1111/j.1468-2885.1999.tb00172.x

32.

Nabi

R. L.

(2002). Discrete emotions and persuasion. In Dillard

Pfau

(Eds.), Handbook of persuasion (pp. 289-308). Thousand Oaks, CA: SAGE.

33.

Nabi

R. L.

(2003). Exploring the framing effects of emotion: Do discrete emotions differentially influence information accessibility, information seeking, and policy preference? Communication Research, 30, 224-247. doi:10.1177/0093650202250881

34.

Nabi

R. L.

(2007). Emotion and persuasion: A social cognitive perspective. In Roskos-Ewoldsen

D. R.

Monahan

(Eds.), Social cognition and communication: Theories and methods (pp. 377-398). Mahwah, NJ: Lawrence Erlbaum.

35.

Nabi

R. L.

(2015). Emotional flow in persuasive health messages. Health Communication, 30, 114-124. doi:10.1080/10410236.2014.974129

36.

Nabi

R. L.

Gustafson

Jensen

(2018). Framing climate change: Exploring the role of emotion in generating advocacy behavior. Science Communication, 40, 442-468. doi:10.1177/1075547018776019

37.

Nabi

R. L.

Myrick

J. G.

(2019). Uplifting fear appeals: Considering the role of hope in fear-based persuasive messages. Health Communication, 34, 463-474. doi:10.1080/10410236.2017.1422847

38.

O’Keefe

D. J.

(2013). The relative persuasiveness of different message types does not vary as a function of the persuasive outcome assessed: Evidence from 29 meta-analyses of 2,062 effect sizes for 13 message variations. Annals of the International Communication Association, 37, 221-249. doi:10.1080/23808985.2013.11679151

39.

O’Keefe

D. J.

(2017). Misunderstandings of effect sizes in message effects research. Communication Methods and Measures, 11, 210-219. doi:10.1080/19312458.2017.1343812

40.

O’Keefe

D. J.

Jensen

J. D.

(2006). The advantages of compliance or the disadvantages of noncompliance? A meta-analytic review of the relative persuasive effectiveness of gain-framed and loss-framed messages. Annals of the International Communication Association, 30, 1-43. doi:10.1080/23808985.2006.11679054

41.

O’Keefe

D. J.

Jensen

J. D.

(2007). The relative persuasiveness of gain-framed loss-framed messages for encouraging disease prevention behaviors: A meta-analytic review. Journal of Health Communication, 12, 623-644. doi:10.1080/10810730701615198

42.

O’Keefe

D. J.

Jensen

J. D.

(2009). The relative persuasiveness of gain-framed and loss-framed messages for encouraging disease detection behaviors: A meta-analytic review. Journal of Communication, 59, 296-316. doi:10.1111/j.1460-2466.2009.01417.x

43.

Pearl

(2014). Interpretation and identification of causal mediation. Psychological Methods, 19, 459-481. doi:10.1037/a0036434

44.

Quick

B. L.

Bates

B. R.

(2010). The use of gain- or loss-frame messages and efficacy appeals to dissuade excessive alcohol consumption among college students: A test of psychological reactance theory. Journal of Health Communication, 15, 603-628. doi:10.1080/10810730.2010.499593

45.

Reese

S. D.

(2007). The framing project: A bridging model for media research revisited. Journal of Communication, 57, 148-154. doi:10.1111/j.1460-2466.2006.00334.x

46.

Reinhart

A. M.

Marshall

H. M.

Feeley

T. H.

Tutzauer

(2007). The persuasive effects of message framing in organ donation: The mediating role of psychological reactance. Communication Monographs, 74, 229-255. doi:10.1080/03637750701397098

47.

Riet

V. T.

Ruiter

R. A. C.

Werrij

M. Q.

Candel

M. J. J. M.

De Vries

(2010). Distinct pathways to persuasion: The role of affect in message-framing effects. European Journal of Social Psychology, 40, 1261-1276. doi:10.1002/ejsp.722

48.

Rogers

R. W.

(1983). Cognitive and physiological processes in fear appeals and attitude change: A revised theory of protection motivation. In Cacioppo

Petty

R. E.

(Eds.), Social psychophysiology (pp. 153-176). New York, NY: Guilford Press.

49.

Rosenthal

(1979). The file drawer problem and tolerance for null results. Psychological Bulletin, 86, 638-641. doi:10.1037/0033-2909.86.3.638

50.

Rosenthal

(1991). Meta-analytic procedures for social research (2nd ed.). Beverly Hills, CA: SAGE.

51.

Rothman

A. J.

Salovey

(1997). Shaping perceptions to motivate healthy behavior: The role of message framing. Psychological Bulletin, 121, 3-19. doi:10.1037/0033-2909.121.1.3

52.

Rothman

A. J.

Salovey

Antone

Keough

Martin

C. D.

(1993). The influence of message framing on intentions to perform health behaviors. Journal of Experimental Social Psychology, 29, 408-433. doi:10.1006/jesp.1993.1019

53.

Scheufele

D. A.

(1999). Framing as a theory of media effects. Journal of Communication, 49, 103-122. doi:10.1111/j.1460-2466.1999.tb02784.x

54.

Scheufele

D. A.

Iyengar

(2014). The state of framing research: A call for new directions. In Kenski

Jamieson

K. H.

(Eds.), The Oxford handbook of political communication theories (pp. 1-28). New York, NY: Oxford University Press.

55.

Seo

Dillard

J. P.

(2016). The persuasive effects of two stylistic elements: Framing and imagery. Communication Research. Article first published online. doi:10.1177/0093650215626979

56.

Seo

Dillard

J. P.

Shen

(2013). The effects of message framing and visual image on persuasion. Communication Quarterly, 61, 564-583. doi:10.1080/01463373.2013.822403

57.

Tannenbaum

M. B.

Hepler

Zimmerman

R. S.

Saul

Jacobs

Wilson

Albarracín

(2015). Appealing to fear: A meta-analysis of fear appeal effectiveness and theories. Psychological Bulletin, 141, 1178-1204. doi:10.1037/a0039729

58.

Updegraff

J. A.

Rothman

A. J.

(2013). Health message framing: Moderators, mediators, and mysteries. Social and Personality Psychology Compass, 7, 668-679. doi:10.1111/spc3.12056

59.

van Assen

M. A.

van Aert

Wicherts

J. M

. (2015). Meta-analysis using effect size distributions of only statistically significant studies. Psychological Methods, 20, 293-309. doi:10.1037/met0000025

60.

Wong

N. C.

Harvell

L. A.

Harrison

K. J.

(2013). The unintended target: Assessing nonsmokers’ reactions to gain-and loss-framed antismoking public service announcements. Journal of Health Communication, 18, 1402-1421. doi:10.1080/10810730.2013.798376

61.

Zhao

Nan

(2010). Influence of self-affirmation on responses to gain- versus loss-framed antismoking messages. Human Communication Research, 36, 493-511. doi:10.1111/j.1468-2958.2010.01385.x

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.04 MB

0.02 MB