Article

Deleted Text: <xref ref-type=

gift giving. For instance, in-kind donations (e.g., nicotine patches, as opposed to cash) are associated with greater empathy (Jacobsson, Johannesson, and Borquist 2007) and believed to confer social approval, prestige, and respect to the giver (Ellingsen and Johannesson 2011), all concepts with significant emotional content. It is not surprising, then, that in-kind contributions are more likely from donors with higher social capital (Bin and Edwards 2009). Relatedly, Reed, Aquino, and Levy (2007) establish a posi- tive link between the importance to consumers of their moral identity—again, an intrinsically emotional construct (Reed et al. 2007, 191)—and their preference for donating time over money due to the former’s greater perceived mo- rality and self-expressiveness.

Finally, several threads of research on the gifting of money versus in-kind gifts coalesce around the sense, at least among gifters, of the greater thoughtfulness (Gino and Flynn 2011) or, more generally, psychic costs (Robben and Verhallen 1994), of in-kind gifts. In particular, some gift-giving research (Pieters and Robben 1999; Webley, Lea, and Portalska 1983) points to the perceived unacceptability of monetary gifts within self-relevant, inherently emotional relationships (e.g., mother-child) because these convey a lack of effort on the part of the gifter. This is consistent with research in a broader set of domains (Morales 2005) pointing to the greater emo- tionality of more effortful actions. Interestingly, research (Ellen et al. 2000) contrasting a company’s collection of prod- ucts versus cash from its consumers for the purpose of disas- ter relief finds that these consumers ascribe greater effort to the former than to the latter. If a company’s own in-kind con- tributions are seen, similarly, as more effortful than its cash contributions, then this difference is also likely to contribute to the greater emotionality of in-kind contributions over cash.

Next, we focus on a key dimension of the issue eliciting corporate contributions—its perceived controllability— that is likely to interact with the varying emotionality of the monetary versus in-kind contributions to affect con- sumer reactions to the company.

Perceived Issue Controllability

The CSR issues that garner support from companies vary in the extent to which their causes are perceived by consumers to be controllable by some human agent or agents, as opposed to nature or happenstance (Lerner 1980; Schul et al. 2007; Weiner 1980). For instance, issues such as climate change and the AIDS crisis are deemed by most to be largely controllable, whereas breast cancer, Alzheimer’s, and other genetic diseases are perceived as more uncontrollable (Huffington Post 2013; Kolata 2010; Weiner, Perry, and Magnusson 1988). In the context of di- sasters, consumers perceived the 2010 Haiti earthquake as largely uncontrollable (Geeta 2016), whereas the 2015 landslide in Shenzhen, China, was perceived to be more controllable (Dou and Wong 2015).

We assert that the perceived controllability of a CSR is- sue determines the intensity of consumers’ emotional reac- tion to it, influencing, in turn, their reactions to the type of contribution (i.e., monetary vs. in-kind) made by a com- pany. Why might this be so? A broad body of research (Alicke 2000; Bohner et al. 1988; Lerner 1980; Weiner 1980) suggests that consumers typically react in an emo- tionally intense manner to the negative, often unexpected, event at the heart of most (if not all) CSR issues, spurring them to repair this aversive, threatening emotional state by making sense of the event, often quickly and automatically (Weiner, Osborne, and Rudolph 2011; Wilson and Gilbert 2003). A crucial step in this sense-making process is cen- tered on consumers’ appraisals of the extent to which any such event is controllable by human agents (Alicke 2000).

Controllability appraisals allow consumers to assess whether or not the event was intentional, predictable, and, if necessary, preventable, reducing the uncertainty con- nected to the event itself as well as their own ability to deal with it should such an event befall them (Alicke 2000; Bohner et al. 1988; Lerner 1980; Weiner 1980). In fact, as- criptions of controllability to a human agent(s) contribute to what Wilson and Gilbert (2003) refer to as an “ordinization” of the event. Such ordinization renders the event more explainable and understandable, turning the ex- traordinary into ordinary and consequently reducing the event’s emotional power and the intensity of consumers’ emotional reactions to it. This negative relationship be- tween perceived controllability and the intensity of felt emotion is underscored by more recent neuropsychological research (Carlsson et al. 2006; Maier 2105), which docu- ments the link between greater controllability assessments and a lower activation of affective neural processing of aversive, painful stimuli.

It is worth noting that the effect of controllability percep- tions on the emotional intensity of one’s reactions is perva- sive, affecting the reactions of both victims and onlookers, regardless of the precise locus of control (i.e., the onlooker, the victims, or a third party). This notion is supported by re- search both neuropsychological (Buchanan et al. 2012; Carlsson et al. 2006; Maier 2105) and motivational (Bar- Anan, Wilson, and Gilbert 2009; Burger and Arkin 1980) showing that the intensity of people’s affective responses to a misfortune, whether their own or another’s, diminishes with heightened perceptions of the misfortune’s controllability, re- gardless of how controllable it actually is.

Next we theorize about how a company’s CSR contribu- tion type might interact with consumers’ perceived controlla- bility of the CSR issue to affect their company evaluations.

Fluency, Fit, and Company Evaluation

Central to our conceptualization of consumers’ contin- gent responses to CSR contribution type is the assertion that the relatively lower intensity of felt emotion induced

740 JOURNAL OF CONSUMER RESEARCH

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017

Deleted Text: see

Deleted Text: p.

Deleted Text: e.g.,

Deleted Text: ; Weiner, Osborne, and Rudolph 2011

Deleted Text: consequently its

Deleted Text: 2105

Deleted Text: are

Deleted Text: ;

by CSR issues perceived to be largely controllable will cause consumers to process a company’s monetary contri- bution, characterized by lower emotionality, more fluently than its higher emotionality, in-kind counterpart. Conversely, the higher intensity of felt emotion induced by uncontrollable CSR issues will cause the higher- emotionality in-kind contributions to be processed more fluently. Support for this notion comes from research (Halberstadt and Niedenthal 1997; Hsee and Rottenstreich 2004; Lai, Hagoort, and Casasanto 2012) that contrasts, in diverse settings, consumers in an emotional state (i.e., higher intensity of felt emotion) to those in a deliberative state (i.e., lower intensity, if any, of felt emotion) to show that while the former are more likely to attend to subse- quent information that is higher in emotionality, the latter are more likely to attend to information lower in emotion- ality. This process is guided, at least in part, by the prefer- ential activation of emotional-intensity-consistent concepts in memory (Lai et al. 2012; Levine and Ramirez 2013; Niedenthal, Halberstadt, and Innes-Ker 1999) and, in the case of emotional information, not necessarily restricted to specific emotion types (Brosch, Pourtois, and Sander 2010; Lai et al. 2012; Levine and Ramirez 2013; Niedenthal et al. 1999). The increased attention to emotional-intensity- consistent information is likely to result, in turn, in a greater speed and ease of its processing (Dweck, Mangels, and Good 2004; Lai et al. 2012), producing a subjective ex- perience of processing ease, or fluency (Alter and Oppenheimer 2009; Topolinski and Strack 2008, 2009).

Extant research (for a summary, see Alter and Oppenheimer 2009, 228) points to the positive affect pro- duced by the experience of fluency as a key determinant of consumers’ favorable evaluations of fluently processed in- formation. Given that consumer evaluations of the contri- bution type undertaken by a company are likely to spill over to the contributing company itself (Yoon, Gürhan- Canli, and Schwarz 2006), we expect more fluently processed contribution-type information to produce more positive company evaluations. Additionally, a related but distinct body of work (Topolinski and Strack 2009) points to the possibility that feelings of fluency may produce an almost intuitive hunch that the fluency-producing items of information fit together well (i.e., they are coherent and make sense together). Notably, research, much of it in the domain of CSR itself (see Sen et al. 2016 for a recent re- view), suggests that a similar sense of fit or match between a CSR issue and a company’s actions toward it can also produce greater liking for the company. In sum, then, we expect both a direct effect of processing fluency as well as an indirect effect through perceptions of contribution-cause fit to cause consumers to evaluate a company more posi- tively when it makes in-kind contributions, as opposed to monetary contributions, to a CSR issue that is perceived to be relatively uncontrollable. When a CSR issue is per- ceived as relatively controllable, however, these direct and

indirect effects of fluency should cause consumers to eval- uate a company more positively when it makes monetary contributions rather than in-kind contributions.

The Moderating Role of Controllability Accessibility

If consumers’ reactions to CSR contribution type hinge, as implied by our conceptualization, on their perceived controllability of the CSR issue, then the predicted interac- tive effects of contribution type and perceived issue con- trollability on fluency, fit, and evaluation should be more pronounced when the notion of controllability is more ac- cessible to consumers (Bhatia 2013; Sherman, Mackie, and Driscoll 1990; Smith et al. 1993; Taylor and Fiske 1978). A vast body of work points to the accessibility, construed broadly to subsume “notions of salience for externally pro- vided stimuli and retrievability and activation strength for memories and other mental objects” (Bhatia 2013, 524), of an evaluative input or dimension (such as controllability) as a key driver of its influence on relevant judgments (Bhatia 2013; Cervone et al. 2008; Higgins 1996; Sherman et al. 1990). This is because dimensions that are more ac- cessible receive greater attention and have a greater chance of being sampled and weighted in subsequent judgments (Bhatia 2013; Cervone et al. 2008; Higgins 1996; Schwarz and Strack 1981; Sherman et al. 1990).

Accessibility is often a function of the extent to which that dimension has been used in recent judgments (Förster and Liberman 2007; Mussweiler and Damisch 2008; Smith and Brascombe 1987) or presented repeatedly to the con- sumer (Feldman and Lynch 1988; Lu, Xie, and Liu 2015). At the same time, recurrent use of an evaluative dimension can render that dimension chronically accessible. For in- stance, Burger and Hemans (1988) observed that controlla- bility, and attributions thereof, is chronically accessible for those who are high in their general desire for control in that they not only make more controllability-related queries but are also more likely to use controllability- related information in relevant judgments. In other words, consumers with a chronic desire for control tend to con- stantly appraise life events in terms of their controllability, causing this evaluative dimension to, over time, become chronically accessible (Förster and Liberman 2007; Higgins 1996; Smith et al. 1993).

Based on this, we propose that when the notion of con- trollability is either contextually or chronically accessible, consumers will be more likely to focus on the controllabil- ity of the CSR issue. This heightened focus is likely to po- larize, relative to when controllability is less accessible, consumers’ controllability perceptions of high versus low controllability CSR issues, which in turn polarizes the in- tensity of their emotional reaction to CSR issues of varying controllability (Cervone et al. 2008; Higgins 1996; Smith et al. 1993). As a result, consumers with high (vs. low)

HILDEBRAND ET AL. 741

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017

Deleted Text: <xref ref-type=

Deleted Text: emotional

Deleted Text: o

Deleted Text: p.

Deleted Text: s

Deleted Text: p.

Deleted Text: ing, in turn,

Deleted Text: t

Deleted Text: are

Deleted Text: versus

controllability accessibility are more likely to experience fluency, or the lack thereof, while processing CSR contri- bution type information, driving their perceptions of contribution-cause fit and company evaluation.

Overview of Studies

Next, we report five studies that test our framework of consumer reactions to corporate contributions (figure 1) in the CSR context of disaster relief. Study 1 tests our basic outcome prediction. Study 2 provides support for our foun- dational assertions regarding the higher intensity of felt emotions elicited by CSR issues that are perceived to be uncontrollable as opposed to controllable (study 2A) and the greater emotionality of in-kind over monetary contribu- tion type (study 2B). Study 3 explicitly tests for the role of controllability-induced emotional intensity by showing its influence on consumer evaluations of companies making monetary and in-kind contributions. Finally, study 4 pro- vides evidence, through both moderation and mediation, for our fluency-based account while ruling out certain key alternative accounts. Importantly, in all studies we fol- lowed standard procedures (Baskin et al. 2014; Shapiro 1999) to exclude participants who skipped the experimen- tal manipulations, failed attention check questions that asked what condition they were in, were suspicious about study objectives, and/or explicitly reported difficulties in understanding the instructions and experimental materials.

STUDY 1

The main objective of this study was to test our basic prediction that consumers will evaluate a company making monetary contributions, rather than in-kind contributions, to a CSR issue more favorably when that issue is perceived to be largely controllable, but less favorably when that is- sue is perceived to be largely uncontrollable. We also aim to show that this interaction is more likely in contexts wherein the notion of controllability is accessible. Finally, we test for the validity of our conceptualization of per- ceived controllability as one driven, more generally, by some human agent(s) rather than just specific entities, such as the victims (Schul et al. 2007; Weiner et al. 1988).

Design and Procedure

We recruited 216 participants through the Amazon Mechanical Turk platform in exchange for a fee. Two par- ticipants were excluded based on our screening criteria de- scribed earlier, resulting in a total of 214 participants (52.3% female; average age ¼ 37.4). We employed a 2 (contribution type: monetary vs. in-kind) � 2 (issue controllability: high vs. low) � 2 (controllability accessibility: high vs. low) between-subjects design, with company evaluation as the dependent variable. We first primed controllability accessibility using a procedural priming task (adapted from Harris, Griffin, and Murray 2008), wherein repeated judgments involving a dimension (e.g., controllability) increase its likelihood of being used,

FIGURE 1

CONCEPTUAL FRAMEWORK

742 JOURNAL OF CONSUMER RESEARCH

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017

Deleted Text: consumers

Deleted Text: during the

unprompted, in subsequent judgments (Mussweiler and Damisch 2008; Smith and Brascombe 1987). Accordingly, in the high-controllability-accessibility condition, we asked participants to read 15 short descriptions of daily events (e.g., being late for work, developing a stomach ache, re- ceiving a compliment) and indicate, for each, the extent to which they felt they had control over its occurrence (1 ¼ not at all controllable, 7 ¼ completely controllable). In the low-accessibility condition, we asked participants to read the same 15 event descriptions and indicate the frequency with which each happens to them (1 ¼ never, 7 ¼ very fre- quently). We expected respondents’ increased focus on fre- quency, aside from serving as a control condition, to even inhibit the accessibility of controllability-related thoughts (Förster and Liberman 2007; Sherman et al. 1990).

Participants then read two fictitious news reports that systematically manipulated perceived issue controllability and corporate contribution type (see the web appendix for study stimuli). The first report manipulated issue controlla- bility—based on research linking the cause of a disaster, either manmade or natural, to its perceived controllability (Baum and Fleming 1993; Zagefka et al. 2011)—by de- scribing an avalanche in California in 2008 as caused either by snowmobiles (i.e., a manmade disaster: high-issue-con- trollability condition) or a snowstorm (i.e., a natural disas- ter: low-issue-controllability condition). An online pretest (N ¼ 110) verified the effectiveness of the controllability accessibility and issue controllability manipulations. The pretest used a 2 (controllability accessibility: high vs. low) � 2 (issue controllability: high vs. low) between-subjects design. After performing the controllability accessibility manipulation and reading the news report about the disas- ter, participants of the pretest indicated their level of agree- ment (seven-point scale; 1 ¼ strongly disagree, 7 ¼ strongly agree) with the statements “The disaster was pre- ventable,” “Someone had the capability to stop the disaster from occurring,” “The affected community is to be blamed for the disaster,” and “The affected community is responsi- ble for the disaster.” The first two items were averaged to form the issue controllability manipulation check (r ¼ .93, p < .001), and the latter two items were averaged to check the potential role of victim responsibility (r ¼ .83, p < .001). An ANOVA with controllability accessibility and issue controllability as independent variables revealed a significant main effect of issue controllability on the ma- nipulation check (Mhigh ¼ 4.83, Mlow ¼ 2.36, F(1,106) ¼ 67.35, p < .001) and, as theorized, a significant issue controllability � controllability accessibility interaction (F(1,106) ¼ 8.00, p < .01). Planned contrasts revealed that the effect of issue controllability was significantly stronger in the high-controllability-accessibility condition (F(1,107) ¼ 66.51, p < .001, g2p ¼ .38) than in the low- controllability-accessibility condition (F(1,107) ¼ 12.33, p < .01, g2p ¼ .10). We also found, in line with prior re- search (Lerner 1980; Zagefka et al. 2011), a significant

main effect of issue controllability on perceived victim responsibility (Mhigh ¼ 2.21, Mlow ¼ 1.83, F(1,106) ¼ 4.34, p < .05, g2p ¼ .04). Importantly, the role of victim responsibility was, in and of itself, negligible, in terms of both the means as well as the effect size, compared to attri- butions of controllability to, more broadly, some human agent(s) (i.e., manipulation check; gp

2 ¼ .39; Perdue and Summers 1986).

After reading the avalanche news report, participants read a second news report that described a company’s di- saster relief initiative. Specifically, the article reported that in response to the avalanche, a computer technology ser- vices company had decided to either donate $1 million, which would be used for the procurement and allocation of food and medicines to affected communities (i.e., monetary contribution condition), or provide employee volunteers and services to help with the procurement and allocation of food and medicines, an effort that would cost the company $1 million (i.e., in-kind contribution condition). We made donation usage and cost equivalent across the two contri- bution type conditions to control for potentially confound- ing variations in respondents’ perceptions of company involvement. Regardless, we expect, in line with the find- ings of Liu and Aaker (2008), that the theorized differences in emotionality will be produced by the mere mention of each of the two contribution types (a more direct test of this premise follows in study 2B).

Next, participants evaluated the company using a four- item, seven-point scale (1 ¼ not at all favorable; very bad; not at all appealing; not at all helpful; 7 ¼ very favorable; very good; very appealing; very helpful). We averaged these to obtain an overall measure of company evaluation (a ¼ .94). Participants then indicated their gender and age.

Results and Discussion

Company Evaluation. To test our outcome prediction, we conducted an ANOVA with contribution type, issue controllability, and controllability accessibility as indepen- dent variables, and company evaluation as the dependent variable. Both the main effect of controllability accessibility (F(1, 206) ¼ 4.02, p < .05), and the interac- tion between contribution type and issue controllability (F(1, 206) ¼ 3.65, p ¼ .06) were significant, the latter al- beit marginally so. More importantly, consistent with our prediction, we found a significant three-way interaction (F(1, 206) ¼ 4.80, p < .05). Planned contrast analyses re- vealed a significant two-way interaction between contribution type and issue controllability in the high- controllability-accessibility condition (F(1, 211) ¼ 9.25, p < .01). As illustrated in figure 2, participants evaluated the company more positively when it contributed money (vs. services) as relief toward the controllable disaster (Mmonetary ¼ 6.60, Min-kind ¼ 6.06, F(1, 209) ¼ 4.88, p < .05) and when it contributed services (vs. money) as relief

HILDEBRAND ET AL. 743

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017

Deleted Text: high

Deleted Text: controllability

Deleted Text: e

Deleted Text: Aside from serving as a control condition, w

Deleted Text: one

Deleted Text: <italic>&hx2019;</italic>

toward the uncontrollable disaster (Mmonetary ¼ 5.92, Min-kind ¼ 6.45, F(1, 209) ¼ 4.41, p < .05). In the low- accessibility condition, on the other hand, the two-way in- teraction was not significant (F(1, 211) ¼ .05, p > .10): contribution type and issue controllability did not affect company evaluations (Mhigh controllability-monetary ¼ 6.08, Mhigh controllability-in-kind ¼ 5.82, F(1, 209) ¼ 1.34, p > .10; Mlow controllability-monetary ¼ 6.20, Mlow controllability-in-kind ¼ 5.87, F(1, 209) ¼ 1.86, p > .10).

A follow-up lab study with 125 participants, not reported in full for the sake of brevity (data available upon request), obtained an identical pattern of results using a measure of chronic rather than task-triggered controllability accessibility. Specifically, we operationalized controllabil- ity accessibility through the Desirability of Control (DC) scale (Burger and Cooper 1979; see study 2A for measure details), premised on the higher accessibility of controlla- bility among participants with a greater desire for control

(Burger and Hemans 1988; Thompson and Schlehofer 2008). As expected, the regression analysis revealed a sig- nificant three-way interaction (t(117) ¼ 3.04, p < .01). Simple slopes analyses indicated that participants high in DC (þ1 SD) evaluated the company more positively when it contributed money (vs. services) to help the victims of the controllable disaster (t(117) ¼ 2.00, p < .05), and when it contributed services (vs. money) to help the victims of the uncontrollable disaster (t(117) ¼ –2.33, p < .05). There were no significant differences for the participants low in DC (–1 SD) (ps > .10).

Our basic outcome prediction hinges on the two, inde- pendent premises of (a) the greater emotionality of in-kind (vs. monetary) contributions, and (b) consumers’ more in- tense emotional reactions to CSR issues they perceive to be less controllable. We aim to provide support for each of these premises, individually, in the next two studies, 2A and 2B, using the standard response latency (i.e., reaction

FIGURE 2

EFFECT OF ISSUE CONTROLLABILITY, CONTROLLABILITY ACCESSIBILITY, AND CONTRIBUTION TYPE ON COMPANY EVALUATION (STUDY 1)

744 JOURNAL OF CONSUMER RESEARCH

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017

Deleted Text: , relative to

Deleted Text: , as opposed to

time) paradigm (Lai et al. 2012; Topolinski and Strack 2009). Specifically, we expect the relative speed with which participants identify more emotional (vs. less emo- tional) concepts to be greater when they are exposed to less controllable (vs. more controllable) CSR issues (study 2A) and in-kind (vs. monetary) contributions (study 2B). Notably, we decided to use this indirect procedure to re- duce concern with social desirability (Quirin, Kazén, and Kuhl 2009), which could influence more direct assess- ments of consumers’ emotional reactions to disasters and aid provided its victims.

STUDY 2A

This study tests for the greater emotional intensity char- acterizing consumers’ reactions to less controllable CSR is- sues by examining their reaction times to more emotional versus less emotional concepts. We expect, based on re- search linking stimuli evoking a greater intensity of felt emotion to the activation and consequent accessibility of more emotional concepts in memory (Brosch et al. 2010; Lai et al. 2012), that the relative speed with which partici- pants identify more emotional versus less emotional con- cepts will be greater after their exposure to an uncontrollable CSR issue relative to a more controllable one.

Design and Procedure

A total of 142 undergraduate students participated in this study for partial course credit. We examined the perceived CSR issue controllability-based differences in participants’ reaction times to more emotional versus less emotional concepts using a one-factor (issue controllability: high vs. low) between-subjects design, with a measure of chronic controllability accessibility (i.e., desirability of control, or DC) as the moderator. In line with our theorizing, the per- ceived CSR issue controllability-based difference in partic- ipants’ reaction times should be greater for those higher in DC (Burger and Hemans 1988; Thompson and Schlehofer 2008). The scenario used to manipulate perceived issue controllability was identical to that in study 1.

Participants first read the article about an avalanche in California. In line with this study’s objective, participants were not exposed to the article containing the contribution type manipulation. We assessed the accessibility of more and less emotional concepts by measuring participants’ re- action time to four words: “emotional” and “feeling” (i.e., more emotional), and “rationality” and “analytical” (i.e., less emotional). Although response time to a single word can suffice as a measure of concept activation (Wilcox et al. 2009), we decided to use four words to increase the reliability of the measure (see Chatterjee, Irmak, and Rose [2013] for a similar procedure). In line with prior research (Chatterjee et al. 2013), we created an emotion index by

subtracting the reaction times (RT) for the two more emo- tional words from that for the two less emotional words (analyses using just the more emotional words yielded comparable results). Given that a faster reaction time to a word points to greater accessibility of the corresponding concept in participants’ minds (Collins and Loftus 1975; Topolinski and Strack 2008), higher values of the emotion index (i.e., RTlow emotion – RThigh emotion) reflect higher ac- cessibility of more emotional concepts and lower accessi- bility of less emotional concepts (Brosch et al. 2010; Lai et al. 2012).

We used the MediaLab and DirectRT software to pre- sent the instructions and collect response time measures. Specifically, the computer screen presented participants with a series of trials in which a fixation point (þþþþ) would be followed by a string of letters. Participants were told that the string of letters could be a word (e.g., vaca- tion) or a nonword (e.g., favan, niare), and were instructed to identify each word on the screen accordingly by pressing a specific letter on the keyboard (i.e., “W” for a word and “N” for a nonword). They were also instructed to give equal importance to both accuracy and speed in this identi- fication task. After practicing with 10 trials, participants were exposed to 54 strings of letters (27 words, 27 non- words), among which the focal words were randomly pre- sented. Finally, after a filler task, we assessed participants’ controllability accessibility using the DC scale (Burger and Cooper 1979).

Results and Discussion

Thirteen participants were eliminated based on our gen- eral screening criteria. Following convention in response latency studies (Fazio 1990), we further excluded from analyses 15 participants who identified at least one focal word incorrectly and two whose reaction time to the focal words exceeded three standard deviations from the mean, rendering a final sample of 112 participants (52.1% female; average age ¼ 22). We then created a DC index, excluding, in line with prior research (McCutcheon 2000), items with low item-total correlation. The final 10 items were aver- aged to obtain each participant’s DC index score (a ¼ .86; see web appendix for scale items).

Following extant research (Lakens 2011; Lo and Andrews 2015), we analyzed the emotion index using a Generalized Linear Model with gamma distribution and log link. We also analyzed the data using the more com- mon linear regression and found virtually identical results. To test our hypothesis, we included issue controllability, controllability accessibility (i.e., DC index), and their inter- action as predictors of the emotion index; we added a small constant to the index to guarantee positive scores, a re- quirement of gamma distributions (the reported means are untransformed to improve interpretability). Results showed a significant main effect of issue controllability (b ¼ .17,

HILDEBRAND ET AL. 745

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017

Deleted Text: , as opposed to

Deleted Text: , as opposed to

Deleted Text: , thus,

Deleted Text: -

Deleted Text: fifteen

v2 ¼ 9.68, p < .01) and, importantly, a significant interac- tion between issue controllability and controllability accessibility (b ¼ .23, v2 ¼ 6.80, p < .01). As predicted, simple effects analyses revealed that participants high in controllability accessibility (þ1 SD in the DC index) pre- sented higher emotion index scores in the uncontrollable disaster condition than in the controllable one (MLow Controllability ¼ –15, MHigh Controllability ¼ 311, v2 ¼ 14.36, p < .001). Participants low in controllability accessibility (–1 SD in the DC index), on the other hand, were no differ- ent in their emotion index scores across the two issue controllability conditions (MLow Controllability¼ 138, MHigh Controllability ¼ 121, v2 ¼ .04, p > .10).

In summary, this study shows that, as theorized, the per- ceived controllability of a CSR issue affects the intensity of consumers’ emotional reaction: relative to less emo- tional concepts, more emotional concepts were more acces- sible to participants reacting to a CSR issue they perceived to be less controllable. As well, in line with our theorizing, this association was stronger for participants with chroni- cally higher, as opposed to lower, levels of controllability accessibility.

STUDY 2B

This study tests for the greater emotionality of in-kind, relative to monetary, contributions by examining consumer reaction times to more emotional versus less emotional concepts. Similar to the increased accessibility caused by the intensity of an individual’s emotional state observed in study 2A, we expect, based on the vast body of memory re- search linking exposure to a particular construct with the automatic activation of associated constructs (i.e., spread- ing activation; Collins and Loftus 1975), that the relative speed with which participants identify more emotional ver- sus less emotional concepts will be greater after their expo- sure to information about the more emotional in-kind contributions relative to the less emotional monetary ones.

Design and Procedure

A total of 76 undergraduate students participated in this one-factor (contribution type: monetary vs. in-kind) be- tween-subjects study for partial course credit. As in study 1, participants read about an avalanche in California and a company’s contribution to disaster relief. Unlike in study 1, however, the article explicitly stated that the cause of the disaster had not yet been determined (Gruman and Sloan 1983) to ensure that any differences in the emotion indices of participants were due only to differences in con- tribution type (i.e., not muddied by differences in perceived issue controllability). After describing the disaster, the arti- cle reported on a logistics company’s disaster relief contri- bution to be either $1 million that would be used for

logistics support (i.e., monetary) or logistics services worth $1 million (i.e., in-kind).

Next, participants engaged in a lexical decision task that was identical to that in study 2A except for the inclusion, additionally, of the target word “effortful.” This was in- cluded to examine empirically, given the findings of Ellen et al. (2000), the role of effort, if any, in the theorized dif- ference in emotionality across the two contribution types. Upon finishing the lexical decision task, participants evalu- ated (seven-point scales; 1 ¼ not at all, 7 ¼ very much) the company contribution on a series of dimensions: effortful, kind, humane, sincere, and helpful. Corporate responses considered more effortful, kind, and humane are perceived to be more heartfelt and thus produce more emotional reac- tions, particularly when perceived to be sincere (Morales 2005). Accordingly, we chose these items to gauge the dif- ferent potential roles in the varying emotionality of the two contribution types, and thus averaged them to form a contribution beliefs index (a ¼ .85).

Results and Discussion

The exclusion of participants who failed our general screening criteria, as well as those who either incorrectly identified the target words or whose reaction times to these words were three standard deviations above the mean, pro- duced a final sample of 55 participants (49.8% female; av- erage age ¼ 21.4).

Emotion Index. We performed the same analysis as in study 2A, using a Generalized Linear Model with gamma distribution and log link; results using linear regression were virtually identical. The model with contribution type predicting the emotion index revealed, as expected, a sig- nificant effect (b¼–.08, v2 ¼ 7.20, p < .01) such that par- ticipants in the in-kind contribution condition presented higher emotion index scores than those in the monetary contribution condition (Min-kind ¼ 584; Mmonetary ¼ 200). In other words, as theorized, in-kind contributions appear to be associated with emotional concepts to a greater extent than are monetary contributions.

Other Measures. We ran an identical Generalized Linear Model with participants’ reaction times to the word “effortful” as the dependent variable. Interestingly, the ef- fect of contribution type was not significant (b¼–.16, v2 ¼ 1.81, p > .10). This suggests that the ascription of greater effort to a company’s collection of products (vs. cash) from its consumers (Ellen et al. 2000) may not neces- sarily generalize to a company’s own in-kind (vs. mone- tary) contributions, particularly when these are equivalent in overall value, as in our studies. This was reflected, more generally, in the one-way ANOVA with contribution type predicting the contribution beliefs index, which revealed a directional but nonsignificant effect (Min-kind ¼ 5.91, Mmonetary ¼ 5.77, F(1, 53) ¼ .38, p > .10), due likely to the

746 JOURNAL OF CONSUMER RESEARCH

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017

Deleted Text: between-subjects

Deleted Text: one

Deleted Text: , versus

equivalence we imposed across the two contribution types in their overall value.

Taken together, then, studies 2A and 2B provide support for our theorizing regarding the greater emotionality of in- kind as opposed to monetary contributions, on the one hand, and consumers’ greater intensity of felt emotions in response to CSR issues that are perceived to be uncontrol- lable as opposed to controllable. Next, in study 3, we test for our assertion that consumers’ differential reactions, in terms of company evaluations, to the lower emotionality monetary CSR contributions versus the higher emotionality in-kind contributions actually hinge on the greater levels of emotional intensity elicited by CSR issues that are per- ceived to be uncontrollable as opposed to controllable.

STUDY 3

This study had two objectives. The first objective was to test our theoretical assertion that controllability accessibil- ity and perceived CSR issue controllability interact to determine the intensity of participants’ felt emotion, which in turn interacts with CSR contribution types of varying emotionality to affect company evaluations. We did so by manipulating controllability accessibility, issue controlla- bility, and contribution type, and measuring the intensity of participants’ emotional reaction using an indirect task adapted from Quirin et al. (2009). We chose an indirect task to reduce concerns about social desirability (Quirin et al. 2009), as respondents could feel compelled, upon di- rect questioning, to indicate that they felt emotional after reading about a disaster. Our second objective was to en- hance the generalizability of our findings by using alterna- tive manipulations of controllability accessibility and issue controllability.

Design and Procedure

In line with its objectives, this study employed a 2 (controllability accessibility: high vs. low) � 2 (issue controllability: high vs. low) � 2 (contribution type: mone- tary vs. in-kind) between-subjects design. A total of 413 in- dividuals based in the United States were recruited through Prolific Academic to participate in this study for a fee (17 participants who reported having technical issues received the fees but were not included in the sample). We increased our sample size in anticipation of a reduction in the effect size of the indirect effect (Cone and Ferguson 2015). Fifty- six participants (13.5%) were excluded based on our screening criteria, yielding a final sample of 357 partici- pants (40% female; average age ¼ 33).

We first manipulated controllability accessibility by adapting a supraliminal priming procedure developed spe- cifically to manipulate the accessibility of judgmental di- mensions (Sherman et al. 1990). Specifically, we showed participants, in sequence, 10 pairs of words. In the case of

each pair, a “memory” word (e.g., city, control) was pre- sented on the left side of the screen and a “color” word (e.g., green, violet) was presented on the right side. In each word pair, the memory word was written in black on a white background, whereas the color word (e.g., orange) was presented in white on a colored background (e.g., blue) that was different, color-wise, from the color word it contained. For each pair, participants were asked to memo- rize the memory word presented on the left and the back- ground color of the color word presented on the right. In the high controllability accessibility condition, four of the 10 memory words were related to controllability (i.e., controllability, control, controlling, controlled) and the re- maining six were neutral words. In the low-controllability- accessibility condition all 10 memory words, including the four unique to this condition (i.e., computation, conve- nience, certification, city), were neutral. To avoid any po- tential confounds associated with the starting letter (i.e., “c”) of the controllability-related words, all letters in both conditions started with a “c.” For each word pair, partici- pants were given 5 seconds to memorize the memory word and the background color and, after a 10-second pause dur- ing which they saw a blank screen, were instructed to type in the memorized word and color.

After completing the controllability accessibility manip- ulation, participants read the news report used to manipu- late issue controllability. The report described a major flood that hit South Dakota in 2008 and stated that the di- saster was either caused by ill-constructed dams (i.e., high issue controllability) or by an unusually powerful storm (i.e., low issue controllability). This manipulation was underscored in the report through expert opinion that the dam had ruptured due to negligence (i.e., high issue con- trollability) versus a storm of historic proportions (i.e., low issue controllability). We checked the effectiveness of both the controllability accessibility and issue controllability manipulations through an online pretest (N ¼ 97) with the same issue controllability manipulation-check measure (r ¼ .88, p < .001) and victim responsibility measure (r ¼ .79, p < .001) used in study 1. Again, the effect of issue controllability on the manipulation check was significant (Mhigh ¼ 5.77, Mlow ¼ 2.67, F(1,93) ¼ 106.79, p < .001), as was the former’s interactive effect with controllability accessibility (F(1,93) ¼ 5.11, p < .05). Planned contrast analysis revealed, as in study 1, a stronger effect of issue controllability in the high-accessibility condition (F(1,94) ¼ 90.63, p < .001, g2p ¼ .49) than in the low-accessibility one (F(1,93) ¼ 29.98, p < .001, g2p ¼ .24). The effect of issue controllability on victim responsibility was only mar- ginally significant (Mhigh ¼ 1.89, Mlow ¼ 1.61, F(1,93) ¼ 2.86, p ¼ .09), denoting, again, a marginal role (if any) of victim responsibility attributions in participants’ controlla- bility perceptions.

After exposing participants to the news report, we as- sessed the intensity of their felt emotion using an indirect

HILDEBRAND ET AL. 747

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017

Deleted Text: C

Deleted Text: A

Deleted Text: ten

Deleted Text: C

Deleted Text: A

Deleted Text: ten

Deleted Text: .

Deleted Text: five

Deleted Text: <italic>-</italic>

task to avoid potential demand effects (Quirin et al. 2009). Specifically, participants were asked to rate the extent to which they thought three artificial words (i.e., SAFME, TUNBA, and SUKOV) subjectively conveyed the mental states emotional and dull (1 ¼ doesn’t fit at all, 4 ¼ fits very well). We chose these two mental states based on a recent study on the affective meaning of words (Warriner, Kuperman, and Brysbaert 2013), which shows that the word “emotional” [“dull”] rates significantly higher [lower] than the scale mean on the intensity (i.e., arousal) dimension of af- fective meaning, but not on either of the two other dimen- sions, valence and dominance. Because judgments about ambiguous objects such as artificial words are strongly influ- enced by one’s affective state (Quirin et al. 2009), respon- dents experiencing greater emotional intensity should be more likely, compared to those experiencing lower emotional intensity, to ascribe the artificial words to an emotional state as opposed to a dull state. Accordingly, we created an emo- tional intensity index by averaging the scores for the mental state “dull” ascribed to the three artificial words and subtract- ing it from the average score for the mental state “emotional” (Quirin et al. 2009).

After completing the task, participants read one of the two news reports used in study 1 to manipulate contribution type and evaluated the company using study 1’s measure (a ¼ .90).

Results and Discussion

Company Evaluation. We tested our main outcome pre- diction using a 2 (controllability accessibility: high vs. low) � 2 (issue controllability: high vs. low) � 2 (contribution type: monetary vs. in-kind) ANOVA. As in study 1, the three-way interaction was significant (F(1, 349) ¼ 5.87, p < .05). We also found a marginally significant issue controllability � contribution type interaction (F(1, 349) ¼ 3.31, p ¼ .07). Most germane, the issue controllability � contribution type interaction was significant for participants in the high- controllability-accessibility condition (F(1, 354) ¼ 8.76, p < .01), but not for those in the low condition (F < 1). As pre- dicted, participants in the high-controllability-accessibility condition evaluated the company more favorably when it made a monetary (vs. in-kind) contribution to a controllable issue (MMonetary ¼ 6.03; MIn-kind ¼ 5.52, F(1, 354) ¼ 4.40, p < .05), whereas the opposite was true when the issue was per- ceived as uncontrollable (MMonetary ¼ 6.11; MIn-kind ¼ 5.63, F(1, 354) ¼ 4.36, p < .05). There were no differences in company evaluation for those in the low-controllability- accessibility condition (Fs < 1).

Emotional Intensity. We tested for the effects of controllability accessibility and issue controllability on the emotional intensity index using an ANOVA. The main ef- fects of controllability accessibility (F(1, 353) ¼ 2.82, p ¼ .09) and issue controllability (F(1, 353) ¼ 2.74, p ¼ .09)

were both marginally significant. More importantly, the two-way interaction was significant (F(1, 353) ¼ 5.18, p < .05). Contrast analyses supported our theorizing: partici- pants in the high-accessibility condition presented higher emotional intensity scores when the perceived issue con- trollability was low than when it was high (Mlow ¼ .35; Mhigh ¼ .05, F(1, 354) ¼ 7.02, p < .01); those in the low- accessibility condition displayed similar values of the emo- tional intensity index across the two levels of issue controllability (Mlow ¼ .05; Mhigh ¼ .09, F < 1).

To test for the driving role of emotional intensity in the contribution-type-based differences in participants’ com- pany evaluations, we generated bootstrap confidence inter- vals (Hayes 2013) using the R software (R Development Core Team 2008) to capture the indirect effect of emo- tional intensity on company evaluations. In particular, we ran two regressions. First, we regressed the emotional intensity index on controllability accessibility, issue controllability (both included as dummies), and their inter- action. As in the case of the ANOVA, we obtained a signif- icant interaction effect (b¼–.35, t(353) ¼ –2.28, p < .05). Second, we ran a regression that included all three manipu- lated factors (included as dummies), their two- and three- way interactions, along with the main effect of emotional intensity and its interaction with contribution type, as pre- dictors of company evaluation. As expected, we found a significant contribution type � emotional intensity interac- tion (b¼–.32, t(353) ¼ –2.05, p < .05). The main effect of emotional intensity (b ¼ .28, t(353) ¼ 2.45, p < .05) and three-way interaction between the manipulated factors (b¼1.0, t(353) ¼ 2.14, p < .05) were also significant. Finally, we generated 10,000 bootstraps to capture the indi- rect effect (path a¼controllability accessibility � issue controllability ! emotional intensity; path b¼emotional intensity � contribution type ! company evaluation; Hayes 2013). Results revealed a significant indirect effect, as the 95% confidence interval (CI) for the 10,000 boot- straps excluded 0 (b ¼ .11, CI: .01 to .26).

In sum, this study replicated our basic outcome finding (study 1) using a different manipulation of controllability accessibility and issue controllability. More importantly, this study provided evidence for the driving role of con- sumers’ perceived CSR issue controllability-induced inten- sity of felt emotion in the interactive effect of controllability accessibility and issue controllability on consumers’ evaluations of companies making monetary versus in-kind contributions. In the final study, described next, we test for the fluency-based contribution-cause fit component of our theoretical framework (see figure 1).

STUDY 4

This study had two objectives. First, we wanted to test for the theorized role of fluency in the interactive effect of

748 JOURNAL OF CONSUMER RESEARCH

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017

CSR issue controllability and CSR contribution type on company evaluation through both mediation and modera- tion. To do so, we assessed the roles of participants’ self- reported feelings of fluency while reading the news report about the CSR contribution, and their perceptions of contribution-cause fit as sequential mediators of the inter- active effect of contribution type and perceived issue con- trollability on company evaluations. As well, and in line with extant tests for fluency effects (Novemsky et al. 2007; Simmons and Nelson 2006; Topolinski and Strack 2009), we manipulated fluency independently of our focal manip- ulations to examine its moderation of the interactive effects of contribution type and issue controllability on partici- pants’ feelings of fluency, perceptions of contribution- cause fit, and company evaluations.

Second, we wanted to rule out an alternative explana- tion, based on victim deservingness, for our basic interac- tion. Specifically, since the victims of a controllable negative event or issue are more likely to be held responsi- ble for their misfortunes, even when they actually played no part in causing it (results of study 1; Lerner 1980; Zagefka et al. 2011), it is possible that consumers may deem them less deserving of a more emotional type of re- lief, evaluating the company making in-kind contributions less positively. This viewpoint—possibly accompanied by consumers’ more general affinity for in-kind help, and for the providing company as well, when the victims cannot be blamed (i.e., uncontrollable issue)—could produce the pat- tern of company evaluations obtained in studies 1 and 3.

Design and Procedure

This study had a 2 (contribution type: monetary vs. in- kind) � 2 (issue controllability: high vs. low) � 2 (fluency: control vs. low) between-subjects design. A total of 505 participants completed this study online in exchange for a fee. We increased our sample size in anticipation of the re- duced effect sizes in the sequential mediation pathway (Cone and Ferguson 2015). The final sample, after we ex- cluded participants using our general screening criteria, was 431 (41.7% female; average age ¼ 32.7). While this exclusion rate (14.6%) may seem somewhat high, it is on par with previous studies that used similar screening crite- ria and sampling frames (Baskin et al. 2014; Goodman, Cryder, and Cheema 2013).

We manipulated issue controllability as in study 3 by presenting participants with a news report describing a ma- jor flood that hit South Dakota in 2008. Since we were not interested in the moderating effect of controllability acces- sibility in this study, we modified the news report to sys- tematically heighten the accessibility of the controllability dimension for all participants so as to ensure that the basic interaction between issue controllability and contribution type manifests. Specifically, we repeated the information pertaining to controllability (Feldman and Lynch 1988;

Lu et al. 2015) multiple times throughout the news report, including the title as well as its final paragraph (see Zagefka et al. 2011 for a similar operationalization; mate- rials are available in the web appendix).

Next, participants read a second news report containing the contribution type and fluency manipulations. Contribution type was identical to that in study 2B. The news report in the low-fluency condition was presented in embossed italicized gray font, which has been shown to be difficult to read and consequently to reduce processing flu- ency (Alter and Oppenheimer 2009; Novemsky et al. 2007; Oppenheimer and Frank 2008; Simmons and Nelson 2006). In contrast, the article in the control condition was presented in the standard font used in our previous studies.

After reading both news reports, participants evaluated the company using the same scale as in our previous stud- ies (a ¼ .94). Contribution-cause fit was measured with two items: “Does the type of aid provided by [the com- pany] make sense?” and “Does the type of aid provided by [the company] feel right to you?” (seven-point scale; 1 ¼ not at all, 7 ¼ a great deal). The two items, derived from existing measures of fluency-based judgments of fit (Hicks et al. 2010) were averaged to form a fit index (r¼ .80, p < .001). Feelings of fluency were assessed through respon- dents’ ratings of the news item containing the company’s relief effort on two dimensions (seven-point scales): 1 ¼ very difficult to read/unpleasant to read, 7 ¼ very easy to read/pleasant to read (adapted from Song and Schwarz [2009] and Torelli and Ahluwalia [2012]). We averaged these two items to form a feelings of fluency index (r¼ .58, p < .001).

We then checked the effectiveness of our issue controllability manipulation with the two items used in the study 1 pretest (r¼ .92, p < .001). Next, participants indi- cated the extent to which they thought that the affected community was responsible for the disaster (using the same two items as in studies 1 and 3; r ¼ .88, p < .001), and the victims deserved the aid delivered by the company (seven-point scale; 1 ¼ totally disagree, 7 ¼ totally agree; “The affected community deserves the aid delivered by [the company]”). As well, we assessed the extent to which the disaster scenarios triggered mortality salience to exam- ine its potential effect if any; two items (seven-point scales; 1 ¼ not at all, 7 ¼ very much): “I think about how short life really is,” and “I am very much afraid to die” were av- eraged to form a mortality salience index (r¼ .41, p < .001) (Ferraro, Shiv, and Bettman 2005). Finally, partici- pants indicated the extent to which they thought that the news report about the corporate contribution was credible (seven-point scale; 1 ¼ not at all, 7 ¼ a great deal), pro- vided their demographic characteristics (i.e., gender, na- tionality, and age), and completed an attention check that asked them to indicate the cause of the disaster (ill-con- structed dam or historic flood) and the nature of the dona- tion (monetary or services contribution).

HILDEBRAND ET AL. 749

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017

Deleted Text: fluency condition

Deleted Text: as well

Deleted Text: condition

Deleted Text: excluding

Deleted Text: e.g.,

Deleted Text: I

Deleted Text: C

Deleted Text: <xref ref-type=

Results and Discussion

An ANOVA with the three factors and their interactions as predictors of news report credibility revealed no significant main or interactive effects; the corporate contribution news re- port was perceived as equally credible across the two contribu- tion types and fluency levels (Mmonetary-low fluency ¼ 5.42, Mmonetary-control ¼ 5.55, Mservice-low fluency ¼ 5.56, Mservice-control ¼ 5.71, ps > .10).

Manipulation and Other Checks. We ran an ANOVA with the three factors predicting our manipulation check for issue controllability. As expected, we found a strong main effect of issue controllability (Mlow controllability ¼ 2.55, Mhigh controllability ¼ 4.98; F(1, 423) ¼ 225.65, p < .001). We also found an unexpected three-way interactive effect of issue controllability, contribution type, and fluency (F(1, 423) ¼ 4.76, p < .05). However, the negligi- ble effect size of the three-way interaction in comparison to the main effect of issue controllability indicated that the discriminant validity of issue controllability is not a con- cern (gp

2 Issue Controllability ¼ .35; gp2Three-way Interaction ¼

.01; Perdue and Summers 1986). We then checked for the aforementioned potential con-

founds using procedures outlined by Perdue and Summers (1986). An ANOVA with the three factors and their inter- actions predicting mortality salience revealed an unex- pected main effect of contribution type (F(1, 423) ¼ 4.95, p < .05), but, importantly, no significant effect of issue controllability (F(1, 423) ¼ .35, p > .10). A similar ANOVA with victim responsibility as the dependent vari- able revealed, as in studies 1 and 3, only a significant main effect of issue controllability (Mlow controllability ¼ 1.82, Mhigh controllability ¼ 2.54; F(1, 423) ¼ 31.77, p < .001). However, effect size analyses corroborated the findings from studies 1 and 3 regarding the far weaker variation in victim responsibility perceptions (gp

2 ¼ .07) than in that attributed to someone (i.e., manipulation check; gp

2 ¼ .35). Finally, an ANOVA with perceptions of victim deservingness as the dependent variable revealed no signif- icant effects (ps > .10).

Company Evaluation. To test our main outcome pre- diction, we ran an ANOVA with contribution type, issue controllability, fluency, and all their interactions predicting company evaluation. We found a significant two-way in- teraction between issue controllability and contribution type (F(1, 423) ¼ 4.16, p < .05) as well as a marginally significant three-way interaction (F(1, 423) ¼ 3.01, p ¼ .08). No other effects were significant. Consistent with our fluency account, planned simple two-way interaction anal- yses revealed that the interaction between issue controllability and contribution type was significant for participants exposed to the control condition (F(1, 428) ¼ 7.96, p < .01), but not for those exposed to the low-fluency condition (F(1, 428) ¼ .10, p > .10). As illustrated in

figure 3, planned simple main effect analyses revealed that, in the control condition, we replicated our basic study 1 and 3 result: the company was evaluated more favorably when it made an in-kind contribution, rather than an equiv- alent monetary one, to an uncontrollable disaster (Mmonetary ¼ 5.90, Min-kind ¼ 6.28; F(1, 426) ¼ 3.63, p ¼ .05), whereas it was evaluated more favorably when it made a monetary contribution, rather than an equivalent in-kind one, to a controllable disaster (Mmonetary ¼ 6.37, Min-kind ¼ 6.01; F(1, 426) ¼ 4.38, p < .05). Importantly, however, and con- sistent with our fluency account, no contrasts were significant in the low-fluency condition (Mmonetary-low controllability ¼ 5.98, Min-kind-low controllability ¼ 6.16, Mmonetary-high controllability ¼ 5.96, Min-kind-high controllability ¼ 6.19; ps > .10).

Feelings of Fluency. When we analyzed feelings of fluency using the same ANOVA, the three-way interaction was, as expected, significant (F(1, 423) ¼ 6.80, p < .01). Additionally, we found a significant main effect of fluency (Mcontrol¼ 5.83, Mlow fluency ¼ 4.68; F(1, 423) ¼ 76.61, p < .01). Planned contrast analyses supported our fluency account: in the control condition, participants exposed to the uncontrollable disaster felt more fluency when reading about the in-kind contribution, as opposed to the monetary one (Mmonetary ¼ 5.58, Min-kind ¼ 6.05; F(1, 426) ¼ 4.22, p < .05), whereas those exposed to the controllable disaster felt more fluency reading about the monetary contribution, as opposed to the in-kind one (Mmonetary ¼ 5.94, Min-kind ¼ 5.64; F(1, 426) ¼ 3.33, p ¼ .06). In contrast, there were no signifi- cant simple main effects in the low-fluency condition (Mmonetary-low controllability ¼ 4.90, Min-kind-low controllability ¼ 4.45, Mmonetary-high controllability ¼ 4.62, Min-kind-high controllability ¼ 4.75; ps > .10).

Fit. We found a similar pattern of results using the identical ANOVA model with fit as the dependent variable. Only the three-way interaction was significant (F(1, 423) ¼ 16.11, p < .001). Specifically, participants in the control condition perceived monetary, as opposed to in-kind, con- tributions to be more fitting when the disaster was per- ceived to be controllable (Mmonetary ¼ 6.28, Min-kind ¼ 5.77; F(1, 426) ¼ 8.22, p < .01), and in-kind (vs. mone- tary) contributions to be more fitting when the disaster was perceived to be uncontrollable (Mmonetary ¼ 5.82, Min-kind ¼ 6.29; F(1, 426) ¼ 6.00, p < .01). Participants in the low-fluency condition did not perceive any differences in fit across the monetary versus in-kind contribution condi- tions when exposed to the controllable disaster (Mmonetary ¼ 6.03, Min-kind ¼ 6.04; F(1, 426) ¼ .00, p > .10). However, low-fluency participants exposed to the uncon- trollable disaster provided higher fit ratings for monetary contributions than for in-kind ones (Mmonetary ¼ 6.26, Min- kind ¼ 5.69, F(1, 426) ¼ 7.82, p < .01). While unexpected, the potential meaningfulness of this pattern of means is undermined by its failure, unlike in the control condition,

750 JOURNAL OF CONSUMER RESEARCH

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017

Deleted Text: <italic>&hx2019;</italic>

Deleted Text: findings

Deleted Text: was analyzed

Deleted Text: F

Deleted Text: controlling

to manifest for the conceptually related feelings of fluency and company evaluation measures.

Mediation Analyses. To provide stronger support for our fluency-based account, we ran a mediated moderation analy- sis with two mediators modeled in a causal sequence (model 6; Hayes 2013). Specifically, the model included the three factors (coded as dummy variables), which, along with their interactions, predicted, in sequence, feelings of fluency, the perception of fit, and company evaluation (Hayes 2013). As described above, the three-way interaction between fluency, contribution type, and issue controllability was a significant predictor of feelings of fluency (b ¼ .1.13, t(423) ¼ 2.22, p < .05). Further, after we controlled for the main and interac- tive effects of our factors, feelings of fluency had a significant positive effect on fit (b ¼ .23, t(423) ¼ 6.76, p < .001). In the model predicting company evaluation, which included the main and interactive effects of the three manipulated factors as well as the main effects of both mediators, both fit (b ¼ .69, t(423) ¼ 19.18, p < .001) and feelings of fluency

(b ¼ .08, t(423) ¼ 3.08, p < .01) had significant positive ef- fects. Importantly, the conditional indirect effect of the three- way interaction on company evaluation through, sequentially, feelings of fluency and fit was, as expected, positive and sig- nificant (i.e., the 10,000 bootstraps 95% confidence interval (CI) excluded 0; b ¼ .18, CI: .03 to .37), as was the indirect effect of the three-way interaction on company evaluation di- rectly through feelings of fluency (i.e., portion of the indirect effect through fluency that is not further mediated by fit; b ¼ .09, CI: .01 to .23). These results are consistent with our theorizing that feelings of fluency influence company evalua- tions both directly and indirectly, through perceptions of fit. Finally, a mediation analysis pitting our proposed mediators against victim deservingness revealed a nonsignificant indi- rect effect through victim deservingness accompanied by no changes in the significant mediating effects of fluency and fit.

In sum, this study provided strong support for the fluency-based process theorized to underlie the interactive effect of CSR contribution type and perceived CSR issue controllability on consumers’ company evaluations while

FIGURE 3

EFFECT OF ISSUE CONTROLLABILITY, CONTRIBUTION TYPE, AND PROCESSING FLUENCY ON COMPANY EVALUATION (STUDY 4)

HILDEBRAND ET AL. 751

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017

ruling out certain key alternative accounts. One could ar- gue, however, that our operationalization of low fluency through a difficult-to-read font led participants to read the CSR contribution information less carefully, thus prevent- ing our basic outcome prediction to manifest in the low- fluency condition. However, two additional analyses invalidate this alternative attentional account. First, partici- pants in the low-fluency condition did not differ signifi- cantly from those in the control condition in their likelihood of correctly identifying the CSR contribution type: a binary logistic regression of error (i.e., dummy code indicating whether the contribution was identified in- correctly) on the three factors and their interaction indi- cated nonsignificant main and interactive effects (p >.10). Second, and more importantly, we ran a post-test, with 94 online participants, in which we manipulated fluency (con- trol vs. low fluency) and CSR contribution type (monetary vs. in-kind) as in the actual study, and measured (a) their accuracy in identifying the type of contribution (measured as in the actual study); (b) the attention they devoted to reading the information, assessed both objectively (time spent reading the report) and subjectively (“To what extent did you try hard to examine the information?”, “How much effort did you put into examining the information?”, “How much thought did you put into examining the informa- tion?”, “How much attention did you pay to the informa- tion?”, and “How involved were you in reading the task?”; 1 ¼ not at all; 7 ¼ a great deal; Wheeler, Brinol, and Hermann 2007; a ¼ .87); and (c) their fluency in process- ing the information (1 ¼ difficult to read; unpleasant to read; difficult to process; unpleasant to process; difficult to understand; unpleasant to understand; 7 ¼ easy to read; pleasant to read; easy to process; pleasant to process; easy to understand; pleasant to understand; a ¼ .91). Results of the post-test indicated that participants differed, as pre- dicted, in their perceived fluency (F(1, 91) ¼ 23.18, p < .001) but did not differ in their attention to or compre- hension of the information, in terms of both the objective and subjective assessments (Fs < 1, ps >.10).

GENERAL DISCUSSION

Company support of CSR issues comes in many for- ms—yet we know little about the effects of varying contri- bution types on consumers’ evaluations of the contributing company. Some research suggests that relative to monetary support, in-kind support can be viewed by consumers as more effortful and/or emotional (Ellen et al. 2000; Liu and Aaker 2008), pointing to the possibility that they might, in general, react more favorably to a company when it makes in-kind as opposed to monetary contributions. Our research suggests that this evaluative advantage conferred by in- kind contributions actually hinges on characteristics of the contribution-eliciting CSR issue, such as its perceived

controllability. Specifically, a company is evaluated more favorably when it makes in-kind rather than monetary con- tributions to issues that are perceived to be largely uncon- trollable. However, in the case of issues perceived to be largely controllable, monetary contributions elicit more fa- vorable company evaluations than do in-kind contributions of comparable magnitude. We also show that this interac- tion between contribution type and perceived issue control- lability is more likely to manifest when controllability is accessible in the minds of consumers, and is driven by the extent to which the disparate emotionality of each contri- bution type matches intensity of the felt emotion evoked by CSR issues of varying perceived controllability. Together, these insights advance our conceptual under- standing of not only consumer reactions to CSR and, more specifically, disaster relief, but also fluency. We discuss these next.

Theoretical Contributions

CSR. Our findings advance our conceptual sense for consumer reactions to CSR in three ways. First and most fundamentally, this article is the first, to the best of our knowledge, to document the critical role played by con- sumers’ perceptions of a CSR issue in their reactions to the type of contribution made by a company to that issue. This contrasts with the general dominance of in-kind contribu- tions suggested by prior research (Ellen et al. 2000), and underscores the importance of the CSR context in con- sumers’ appraisals of company involvement. As well, the moderation of our basic company evaluation interaction by the accessibility, either contextual or intrinsic, of controlla- bility in the minds of consumers points to the importance of the precise nexus of consumer, CSR, and company- specific factors in a meaningful understanding of consumer reactions to CSR.

Second, and more specifically, we add to the consider- able work on the driving role of the fit between company/ brand and CSR issue in consumer reactions to CSR (Ellen et al. 2006; Gupta and Pirsch 2006; Kuo and Rice 2015; Nan and Heo 2007; Pracejus and Olsen 2004; Sen and Bhattacharya 2001). Widely acknowledged as a multidi- mensional construct (Zdravkovic, Magnusson, and Stanley 2010), fit, as examined thus far, refers broadly to the per- ceived match or closeness, in terms of image, positioning, and/or constituency (Varadarajan and Menon 1988), be- tween a brand/company and the CSR issue(s) it supports. We build on this research by proffering a new dimension of fit, based on the characteristics of the contribution made by a company (as opposed to its more enduring dimen- sions, such as its values) and the perceived controllability of the CSR issue (as opposed to its substantive domain). We also locate this contribution/cause fit in consumers’ ex- perience of processing fluency based on the congruence or match between the emotionality of the contribution type,

752 JOURNAL OF CONSUMER RESEARCH

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017

Deleted Text: a

Deleted Text: , thus,

Deleted Text: <italic>&hx2019;</italic>

Deleted Text: -

Deleted Text: condition

on the one hand, and the intensity of the felt emotion en- gendered by the perceived controllability of the CSR issue, on the other. In this, we join more recent examinations of fit (e.g., perceptual fit based on matching colors; Kuo and Rice 2015) in broadening the bases of the fit between com- pany and CSR issue beyond its traditional conceptualiza- tion as based on the congruence between essentially conceptual dimensions (e.g., product category, target seg- ment, brand image; Ellen et al. 2006; Sen, Bhattacharya, and Korschun 2006).

Third, and relatedly, while the vast majority of research in this domain has focused on the perceptual and cognitive mechanisms underlying consumer reactions to CSR (Sen et al. 2016), our research contributes to the more nascent stream of research focused on the more affective bases for such reactions. Specifically, researchers have recently im- plicated both moral emotions such as pride and guilt (Kim and Johnson 2013; Peloza et al. 2013) as well as feelings of empathy and gratitude (Xie, Bagozzi, and Grønhaug 2015) as drivers of consumers’ reactions to CSR. Our re- search builds on this work to suggest, more generally, that the intensity of consumers’ emotional reactions can also determine their CSR responses.

Disaster Relief. Based on its empirical context, this re- search also extends our understanding of consumer re- sponses to, more specifically, disaster relief. Given that contributions to disaster relief are virtually de rigueur to- day for both global and local companies, a growing body of work examines global patterns in corporate disaster re- lief involvement (Gao 2011; Muller and Whiteman 2009; Patten 2008). However, our only understanding of con- sumer reactions to such corporate action comes, to the best of our knowledge, from Ellen et al. (2000)’s investigation of cause marketing in a retail setting: consumers not only are more likely to support a cause marketing program di- rected toward disaster relief as opposed to an ongoing is- sue, but also react more positively to a program that relies on in-kind contributions from a retailer’s patrons, as op- posed to monetary ones. Our research builds on this main- effect finding by not only examining the extent to which it carries over to consumers’ company evaluations but also, more vitally, fleshing out the disaster- (i.e., perceived con- trollability) and consumer-based (i.e., controllability acces- sibility) contingencies therein to paint a more comprehensive and nuanced picture of how consumers view corporate disaster relief efforts.

Viewed through the broader lens of prosocial behaviors in the context of disaster relief, our findings highlight some interesting differences between consumers’ own charitable donations to disaster victims (Zagefka and James 2015) and their evaluations of companies doing the same. Specifically, prior research suggests that people are gener- ally more likely to help victims of an uncontrollable mis- fortune than those of a controllable one (Weiner 1980;

Weiner et al. 1988). This is consistent with the finding within the context of disaster relief (Zagefka et al. 2011) that disasters ascribed to natural causes elicit greater sup- port than those ascribed to manmade causes. Our findings reveal, in contrast, that when consumers are observers rather than enactors of such helping behavior, there is no main effect of perceived controllability on judgments of the enactor’s (i.e., the company) helping behavior. Instead, consumers’ evaluations of the enactor hinge on the inten- sity of their disaster-induced felt emotions and its impact on their ease of processing subsequent contribution infor- mation. This difference is consistent, more broadly, with research on indirect reciprocity (i.e., when the observer re- wards the enactor for helping a third party; Sigmund 2012), which suggests that the evaluation of a contribution from an enactor to a third party depends less on appraisals of the third party (e.g., victim responsibility) than on those of the enactor’s actions (e.g., type of contribution).

Fluency. Finally, our findings contribute to the vast lit- erature on processing fluency by implicating, for the first time, the intensity of individuals’ felt emotions as yet an- other basis for their fluency experiences. In their integra- tive review of processing fluency, Alter and Oppenheimer (2009) present a wide array of known instantiations of flu- ency, stemming from a diversity of cognitive processes (ta- ble 1; Alter and Oppenheimer 2009). Our findings regarding the driving role of consumers’ perceived controllability-induced emotional intensity in their subse- quent ease of processing contribution-type information of varying emotionality and consequent company evaluations suggest that different emotional states can—like semantic, perceptual, and conceptual priming—comprise, in the words of Alter and Oppenheimer (2009), a fluency manip- ulation. At the same time, our findings complement those of Topolinski and Strack (2008, 2009), who found that a triad of words sharing an underlying semantic structure (e.g., foam, salt, and deep are all semantically linked to the sea) are more rapidly recognized, increasing feelings of fluency and perceptions of fit. Our research suggests that similarly intuitive judgments of fit may emerge for notions (e.g., issue controllability and contribution type) that share a commonality along the affective rather than semantic dimension.

More specifically, in examining the role of emotional state-based fluency in consumers’ CSR judgments, our re- search builds on extant CSR research focused on the driv- ing role of conceptual fluency (e.g., that between the concept of CSR and varying brand concepts; Torelli, Monga, and Kaikati 2012) to broaden the scope of fluency processes in consumers’ responses to CSR. We add to this nascent literature by demonstrating that fluency effects do not occur merely directly by increasing favorability of the fluently processed stimulus (e.g., CSR investment), but also indirectly by increasing consumers’ perception that

HILDEBRAND ET AL. 753

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017

Deleted Text: that

Deleted Text: directly

the stimulus fits or is congruent with previously presented information (e.g., CSR issue). A systematic examination of the underlying roles of these as well as other, thus-far- unexamined bases for fluency in both judgments of com- pany/CSR fit as well as their effects on CSR and company evaluations comprises an interesting future research direc- tion. At the same time, the relative role of such fluency processes in driving consumer preferences for company ac- tions in domains beyond CSR (e.g., brand failures; Smith, Bolton, and Wagner 1999) demands further investigation.

Limitations and Future Research

Naturally, this research has certain limitations, which point to directions for future research. We test our frame- work exclusively in the context of disaster relief. While the type of fit proffered by our research can, importantly, be expected to generalize to any CSR issue wherein controlla- bility looms large in consumers’ minds, empirical corrobo- ration is an essential next step. Future research might also assess whether and to what extent the match between the perceived controllability of a CSR issue (e.g., breast can- cer) and its specific implementation (e.g., Avon’s Breast Cancer Walk vs. Mary Kay’s monetary donations for can- cer research) influences consumer reactions toward the CSR initiative and the company.

In a similar vein, all our studies focused on the role of perceived CSR issue controllability in determining the in- tensity of consumers’ felt emotions. It is entirely conceiv- able, however, that this intensity may also be amplified or reduced by other dimensions of the CSR issue, such as the issue beneficiary’s physical or temporal closeness to the consumer (Loewenstein and Small 2007; Trope and Liberman 2010), the inability of the beneficiary to exert in- tentional control over his or her actions (e.g., children, mentally ill individuals; Malle, Guglielmo, and Monroe 2014), or the beneficiary’s similarity with consumers (Loewenstein and Small 2007). While we took care to con- trol for such potentially confounding differences, it would be interesting for future research to investigate whether the contribution-type-triggered differences in company evalua- tions obtained in our research could be produced through CSR issues that vary in these characteristics.

Additionally, in our research the company contributing to the CSR issue was not in any way responsible for caus- ing or exacerbating the issue. It is possible, particularly in today’s marketplace, that the companies helping to amelio- rate a CSR issue may be the ones responsible for the issue in the first place (e.g., BP compensatory efforts after the Gulf of Mexico oil spill, and more generally, the “green” efforts of companies that have contributed significantly to the degradation of the environment). Clearly, company cul- pability in the CSR issue is likely to alter the dynamics documented by our article. For instance, consumers may expect the offending company to go beyond monetary

contributions and be intimately and comprehensively in- volved in relief efforts, even though—and in fact be- cause—the disaster is deemed as controllable. Interestingly, some research (Muller and Whiteman 2009) suggests that stakeholders may have similar expectations of companies that are merely physically closer to the disas- ter, even in the absence of culpability. It would be impor- tant for future research to investigate consumer responses to remedial rather than prosocial CSR actions taken by a company.

Finally, and more specifically, research (Liu et al. 2012) points to the importance of businesses in helping victims of disasters regain their lives and, through that process, their very identities. The roles of perceived disaster controllabil- ity and company involvement, and indeed other key disas- ter- and company-related factors, in this profoundly important process would comprise a worthwhile research endeavor.

DATA COLLECTION INFORMATION

The first two authors collected the data either at Baruch College/CUNY or online using Amazon Mechanical Turk and Prolific Academic, as well as analyzed these data. They collected data for study 1 in winter 2014 (online), study 2a in fall 2010, study 2b in spring 2014, study 3 in fall 2016 (online), and study 4 in spring 2015 (online).

REFERENCES

Alicke, Mark D. (2000), “Culpable Control and the Psychology of Blame,” Psychological Bulletin, 126 (4), 556–74.

Alter, Adam L. and Daniel M. Oppenheimer (2009), “Uniting the Tribes of Fluency to Form a Metacognitive Nation,” Personality and Social Psychology Review, 13 (3), 219–35.

Bar-Anan, Yoav, Timothy D. Wilson, and Daniel T. Gilbert (2009), “The Feeling of Uncertainty Intensifies Affective Reactions,” Emotion, 9 (1), 123–7.

Baskin, Ernest, Cheryl J. Wakslak, Yaacov Trope, and Nathan Novemsky (2014), “Why Feasibility Matters More to Gift Receivers than to Givers: A Construal-Level Approach to Gift Giving,” Journal of Consumer Research, 41 (1), 169–82.

Baum, Andrew and India Fleming (1993), “Implications of Psychological Research on Stress and Technological Accidents,” American Psychologist, 48 (6), 665–72.

Bhatia, Sudeep (2013), “Associations and the Accumulation of Preference,” Psychological Review, 120 (3), 522–43.

Bin, Okmyung and Bob Edwards (2009), “Social Capital and Business Giving to Charity Following a Natural Disaster: An Empirical Assessment,” Journal of Socio-Economics, 38 (4), 601–7.

Bohner, Gerd, Herbert Bless, Norbert Schwarz, and Fritz Strack (1988), “What Triggers Causal Attributions? The Impact of Valence and Subjective Probability,” European Journal of Social Psychology, 18 (4), 335–45.

Brosch, Tobias, Gilles Pourtois, and David Sander (2010), “The Perception and Categorization of Emotional Stimuli: A Review,” Cognition & Emotion, 24 (3), 377–400.

754 JOURNAL OF CONSUMER RESEARCH

Downloaded from https://academic.oup.com/jcr/article-abstract/44/4/738/3739797 by Adam Ellsworth, Adam Ellsworth on 14 November 2017