THE INFLUENCE OF CRISIS SITUATION AND CONSTRUAL LEVEL ON CONSUMER EVALUATIONS OF ORGANIZATIONAL REPUTATION
Thesis in completion of the degree Master of Science in Business Administration
Michael Naudé
Student number 11363088
Final version 26 January 2018
MSc in Business Administration – Marketing Track Amsterdam Business School
University of Amsterdam
STATEMENT OF ORIGINALITY
This document is written by student Michael Naudé who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
1
TABLE OF CONTENTS
ABSTRACT ... 2
1. INTRODUCTION ... 3
2. LITERATURE REVIEW... 5
2.1 Background: Marketing and Public Relations ... 5
2.2 Corporate Reputation ... 6
2.3 Situational Crisis Communication Theory ... 8
2.4 Crisis Type ... 9
2.5 Construal Level Theory ... 12
2.6 Conceptual Framework and Hypotheses ... 14
3. METHODS ... 16
3.1 Pretesting ... 16
3.2 Main Experiment: Variables ... 18
3.3 Design ... 18 3.4 Procedure ... 19 4. RESULTS ... 21 4.1 Preliminary Steps ... 21 4.2 Testing Hypotheses ... 24 5. DISCUSSION... 28
5.1 Limitations and Future Research ... 30
5.2 Theoretical and Managerial Implications ... 31
5.3 Conclusion ... 33 REFERENCES... 34 APPENDIX A ... 37 APPENDIX B ... 39 APPENDIX C ... 42 APPENDIX D ... 44 Acknowledgements
2
ABSTRACT
For all organizations, for-profit or non-profit, a good reputation among stakeholders is essential to the accomplishment of organizational objectives, yet there are many risks that threaten reputation. Prior research has examined how the type of reputational crisis and the level of organizational responsibility can influence stakeholder reactions, but there is a large gap because psychological construal level theory has not been applied in this context. The purpose of this thesis was to investigate whether and how construal level can influence consumers’ attitudes about a company involved in a brand scandal. Hypotheses were tested by collecting data through an online experiment with a 2(crisis type: corporate ability vs. corporate social responsibility) by 2(crisis responsibility: internal vs. external) by 2(construal level: low/concrete vs. high/abstract) between-subjects factorial design. A convenience sample of N = 57 was analyzed primarily using analysis of variance. The findings show a weak (but not statistically significant) moderation effect of construal level, such that a more abstract construal level leads to more favorable attitudes only if the crisis type is corporate ability and the crisis responsibility is external. The implication is that organizations might benefit by framing their communications abstractly in these conditions and concretely otherwise, and by predicting stakeholder reactions, but limitations to this research are also noted. This paper is the first work to propose a theoretical model and empirically examine construal level theory in this context, and is thus a worthwhile contribution.
3
1. INTRODUCTION
Reputational damage can accrue from corporate scandals or brand scandals linked to a certain firm, especially those that receive much media coverage. Recent examples include the Volkswagen emissions scandal in 2015, when Volkswagen programmed their vehicles to defeat laboratory tests of nitrogen oxide emissions; and the 2016 case of Samsung’s Galaxy Note 7 smartphone catching fire and exploding. However, many of us can recall the 2001 Enron accounting scandal and the 1989 Exxon Valdez oil spill, showing the long-lasting impact of such events. Good reputations take time to build, and bad reputations are hard to recover from.
Having a strong positive reputation corresponds to higher profitability, competitive advantage, customer satisfaction, resource acquisition and lower operating costs (Fombrun, 2012; Walker, 2010). Conversely, these outcomes are threatened by a negative reputation. Research has found that brand scandals can also reduce a brand’s perceived authenticity by signaling a lack of moral integrity, and that authenticity is important for emotional brand attachment and positive word-of-mouth (Morhart, Malär, Guèvremont, Girardin, & Grohmann, 2015).
Ideally, firms would strive to prevent such scandals from happening in the first place, but sometimes the risk is hard to eliminate, and firms should be prepared to respond to a crisis as it unfolds. Part of the corporate-level marketing umbrella is the field of corporate communications, which includes the specialized subfield of crisis communications. One framework in this area is Coombs’s (2007) Situational Crisis Communication Theory (SCCT). According to this approach, an organization should first act to ensure that its stakeholders are out of harm’s way, and then respond to negative publicity with a communications strategy that matches the level of responsibility that stakeholders attribute to the organization. For example, if perceived responsibility is low, the organization may deny that there is a crisis or that it is responsible; but if perceived responsibility is high, the organization does best by apologizing and offering compensation to victims.
Under this framework, it is clearly relevant to ask which factors influence stakeholders’ perceptions of the organization’s culpability. Previous research has found that the type of crisis – whether it relates to the firm’s ability to deliver its product (corporate ability), or its commitment to ethical values (corporate social responsibility) – can influence the reactions of stakeholders. For example, a product-harm (or performance-related) crisis may trigger more anger among consumers than a moral-harm (or values-related) crisis, presumably because product failure threatens consumers’ basic functional needs (Dutta & Pullig, 2011; Hegner, Beldad &
4
Kraesgenberg, 2016). As an individual-level variable, thinking style (analytic vs. holistic) has been found to influence how people interpret negative publicity (Monga & John, 2008). Naturally, there is no one grand unified theory of consumer psychology, so progress in the field has to be made by developing explanations of why people think, feel or behave a certain way in a certain context.
Construal-level theory (CLT) explains how the psychological distance of an object (in the mind of a subject) influences the level of abstraction at which the object is perceived and evaluated (Trope & Liberman, 2010). CLT has been studied in a variety of contexts, including interpersonal conflict and advertising, and has gathered empirical support. However, there has been a lack of research studying the role of CLT in the context of corporate scandals and reputation management. This presents an interesting gap in the literature, because there is reason to believe that construal level would influence how stakeholders perceive a crisis situation, and it therefore has implications for how organizations should respond to repair their reputations. Hence, this thesis explores the following research question: How does construal level influence consumers’ post-crisis evaluation of organizations? By answering this question, managers can build crisis communication strategies that more accurately reflect the mental states of their audiences. In particular, one implication of construal level is how it may affect SCCT response recommendations for different crisis types.
The rest of this thesis is structured as follows: first, a review of the extant literature on corporate reputation and construal level theory leads to new hypotheses. Second, the methodology section describes how the hypotheses were investigated through a preliminary study. Third, the hypotheses are tested against the data. Finally, the thesis concludes with a discussion of the implications and limitations of this research together with suggestions for future research.
5
2. LITERATURE REVIEW
2.1 Background: Marketing and Public Relations
Marketing is about attracting and retaining customers by creating superior customer value, in order to achieve strategic goals like increasing firm revenue or market share. Brands play a crucial role in this process by allowing consumers to identify, learn about, or even resonate with the offering of a particular firm (Keller, 2001). While branding has traditionally focused on products and services, in recent years there has been increasing interest in corporate branding – that is, treating the organization itself (whether a for-profit enterprise, nonprofit or public agency) as a brand. This means that standard branding concepts become relevant at the corporate level.
Building a successful brand entails creating strong, favorable and unique brand associations in the minds of consumers. These associations comprise the brand’s image. A good brand image, together with a high level of brand awareness, contribute to what Keller refers to as “customer-based brand equity”, which is the extent of a customer’s response to the firm’s marketing that can be attributed to the customer’s brand knowledge (Keller, 1993). Applying this to a corporate brand, we can talk about “corporate image”, keeping in mind that corporate brands must appeal not only to customers but also various other stakeholders including employees, investors, suppliers, regulators and the general public. The aggregation of images held by stakeholders over time is known as “corporate reputation”. Like its brands, a firm’s reputation is an important intangible asset.
The field of marketing is separate from, but overlaps with, the field of public relations (PR). While marketing deals with tools like advertising, sales promotions, pricing, distribution and new product development, PR handles relations with investors, communities, governments, the media and charity (De Pelsmacker, Geuens, & Van den Bergh, 2010). However, the boundary between the two can be blurred in cases like corporate advertising, sponsorships and product publicity. Moreover, the trend toward integrated marketing communications (IMC) means that the integration of various departments and functions within an organization is increasingly important despite the tendency for turf wars (De Pelsmacker et al., 2010).
Indeed, the task of crisis communications has traditionally been within the purview of PR. However, the relevance of marketing to building a reservoir of goodwill should not be understated. Caywood (2013) notes that certain products and services carry reputational risk due to the potential harm they might cause, for example pharmaceuticals, food, smoke alarms, cleaning chemicals, propane tanks, vehicles, seat belts, cosmetics, plastic surgery, sports gear, and X-rated movies. Additionally, organizations might target vulnerable groups, for example
6
children, the elderly, racial minorities, lower IQ individuals, the mentally ill or recent immigrants. This is another source of reputational risk. In both cases, IMC plays a role in the reputation building process (Caywood, 2013). According to Varey (2013), marketing communications experts view corporate reputation and image as “a crucial aspect of business strategy”. The reason why corporate reputation is a marketing issue is that “[w]hat a company stands for and goodwill towards the company are closely linked to assessment of quality, and thus the salability of product offerings” (p.105). Insofar as reputation is part of the corporate brand, more effective marketing communications can help solve the management problem of protecting and strengthening corporate reputation, especially by enhancing customer familiarity with the corporate brand (Varey, 2013). The next section reiterates the importance of reputation as an organizational asset.
2.2 Corporate Reputation
Reputation is one of the most important assets that organizations can have, because reputation influences profitability, competitive advantage, operating costs, customer satisfaction, media coverage and resource acquisition (Abratt & Kleyn, 2012; Fombrun, 2012; Walker, 2010). Good reputations take time to build and bad reputations are hard to recover from (Abratt & Kleyn, 2012), making it crucial to manage this intangible resource wisely. Reputation is therefore relevant to the topics of corporate-level marketing and corporate branding, which have become increasingly important in recent decades as companies have moved from traditional product branding (focused on attracting customers) towards concern for a broader set of stakeholders, including not only customers but also employees, managers, investors, suppliers, competitors, regulators, the media, influence groups and communities (Abratt & Kleyn, 2012; Walker, 2010).
Corporate reputation is defined as “the collective judgment by observers of a firm based on assessments of the financial, social and environmental impacts attributed to the corporation over time” (Clardy, 2012) or as “a collective assessment of a company’s attractiveness to a specific group of stakeholders relative to a reference group of companies with which the company competes for resources” (Fombrun, 2012), although there is no consensus on a definition. Reputation is not exclusively a concern of corporations, but almost any type of organization. Reputations are multidimensional, as they are evaluated on a number of issues including the organization’s performance, socially responsible behavior, marketing mix activities, innovation, workplace conditions and good governance (Abratt & Kleyn, 2012; Shamma, 2012). This relates back to Keller’s (1993) associative network memory model of the brand concept.
7
Organizational reputations are formed over time from stakeholders’ aggregated perceptions and impressions (or images) arising from their interactions with the organization (Abratt & Kleyn, 2012; Fombrun, 2012). Organizational images are shaped by the media, the organization’s strategic posture (e.g. product quality and price) and the organization’s projected image, which is a projection of its identity (Fombrun, 2012). Organizational identity refers to the distinguishing characteristics of an organization, including its philosophy, culture, vision, strategy and structure (Abratt, 1989; Dowling, 1993). Stakeholders experience the organization’s identity through its communications, visual identity and symbolism, corporate brand promise and personality, product offerings, and employees (Abratt, 1989; Dowling, 1993; de Chernatony, 1999; Stuart, 1999). The following figure illustrates how reputations are formed.
Due to their strategic importance as valuable intangible assets, organizational reputations need to be actively managed (Coombs, 2007; Shamma, 2012). This is especially true in the case of reputational damage. Opportunistic misconduct that harms customers, suppliers, employees or investors can adversely impact the bottom-line of an organization (Karpoff, 2012). Unless a company’s product is relatively unknown, the benefits of increased awareness due to negative publicity is outweighed by the negative consequences, like lower product evaluation, sales and net present value (Berger, Sorensen & Rasmussen, 2010). So what should companies do in such situations? Organizational identity Philosophy Culture Vision Strategy Structure Projected image Communications Visual identity Symbolism Corporate brand Employees
Perceived image Reputation
Media coverage Strategic posture Product quality Product price
8 2.3 Situational Crisis Communication Theory
W. Timothy Coombs devised a framework for organizational crisis communications to protect reputational assets based on empirical research, known as Situational Crisis Communication Theory (SCCT) (see Coombs, 2007). By harming stakeholders physically, emotionally or financially, a crisis can undermine trust that the organization will continue to meet stakeholders’ expectations, and it therefore damages reputation. SCCT anticipates the reputational threat posed by a crisis in terms of stakeholder reactions, and recommends crisis response strategies according to their effects on stakeholders (Coombs, 2007).
According to SCCT, there are two response phases following a crisis: as part of an ethical obligation, the first phase involves providing instructing and adjusting information to stakeholders; i.e. telling them how they can protect themselves from harm and helping them cope with psychological threats by providing information that reduces uncertainty, and expressing concern for them (Coombs, 2007; Liu, Austin & Jin, 2011). In the second phase, organizations choose among primary and secondary response strategies. Primary response strategies include denial, diminish, and rebuild. Denial strategies seek to remove any connection between the organization and the crisis, for example by denying that there is a crisis, attacking the accuser or shifting blame onto a scapegoat. Diminish strategies seek to reduce the perceived severity of the crisis or the organization’s responsibility for it, for example by explaining why it happened and offering excuses and justifications. Rebuild strategies aim to create a new positive reputation by apologizing for the crisis and offering compensation to victims. In addition to these primary strategies, the main secondary strategy is bolster, which entails reinforcing previous good reputation by reminding stakeholders about the organization’s past good deeds, praising stakeholders or reminding them that the organization is also a victim (Coombs, 2007; Grappi & Romani, 2015; Hegner, Beldad & Kraesgenberg, 2016; Liu et al., 2011).
Which strategy should an organization use and when? SCCT is inspired by Attribution Theory, which argues that stakeholders’ attributions of motives and responsibility for a crisis can trigger emotions such as anger or sympathy, which can then influence their behavioral intentions (Coombs, 2007). Indeed, McDonald, Sparks and Glendon (2010) note that when the cause of a crisis is internal to an organization and under the organization’s control, stakeholders attribute more responsibility to the organization and therefore experience more anger and negative behavioral intentions. According to SCCT, organizations should first assess the level of reputation threat, which is a function of initial crisis responsibility, crisis history and prior relational reputation; and then match the response to the threat level (Coombs, 2007; Kim &
9
Cameron, 2011). The aim of crisis communication is to reduce negative stakeholder emotions and behavioral intentions, and accordingly it follows that the more severe the reputational threat, the more accommodative the strategy is required to be (Coombs, 2007; McDonald et al., 2010). In essence, a response strategy is a way for an organization to show accountability for its actions, so the more responsibility stakeholders attribute to the organization, the more responsibility the organization should accept (Claeys & Cauberghe, 2014; Coombs, 2007).
2.4 Crisis Type
A key factor that influences stakeholder attributions of responsibility is the type of crisis situation, i.e. how the crisis is framed (Coombs, 2007). Crisis types fall into three clusters: victim, accident, and preventable. Victim crises include natural disasters, false rumors, workplace violence, product tampering and so on, where the organization is also a victim and is not attributed responsibility. Accident crises include challenges by stakeholders that the organization is operating inappropriately, technical-error industrial accidents, and technical-error product harm, where the organization’s actions did not intentionally cause the crisis. Finally, preventable crises are those where the organization knowingly broke laws or put stakeholders at risk, such as human-error accidents, human-error product harm, organizational misdeeds (with or without injuries) and management misconduct (Claeys & Cauberghe, 2014; Coombs, 2007; Grappi & Romani, 2015). Therefore, SCCT recommends that organizations match their primary response strategy to the crisis type. More specifically, organizations should use denial strategies for victim crises, diminish strategies for accident crises, and rebuild strategies for preventable crises (Claeys & Cauberghe, 2014; Grappi & Romani, 2015). Although rebuild strategies (in combination with adjusting information) are generally the best for mitigating negative affect, they are also more expensive for the organization, which is why they are not always the optimal answer (Coombs, 2007; McDonald et al., 2010).
An alternative crisis typology is offered by Snyder, Hall, Robertson, Jasinski and Miller (2006), which is not strictly related to SCCT. They classify crises into a 2-by-2 framework, based on the organizational distance to the crisis’s original “center of gravity”, and the frequency (or “normality”) of the crisis. In other words, does the crisis originate from within or outside the organization, and how predictable is it? Thus, any scenario can be classified as shown in the table below, which gives some examples.
10
Internal - normal Internal - abnormal
Product recall Personnel strike
Internal supply chain breakdown
Executive kidnapping Information theft Product boycott
External - normal External - abnormal
Economic depression
Product category obsolescence Industry-wide cyberattack
Political expropriation Industry deregulation Malicious rumors/slander Natural disasters
Major supplier bankruptcy
Table 1: Typology of crises, based on Snyder et al. (2006)
The purpose of presenting this framework is to show that there are different possible ways to categorize a crisis situation besides the three SCCT clusters. Nonetheless, there is some overlap: the victim – accident – preventable types reflect the extent to which the organization’s actions were intentional, and the matrix by Snyder et al. (2006) reflects the predictability of the crisis (with internal/normal being the most predictable, and external/abnormal being the least predictable). Unintentional events are likely to be less predictable. Therefore, the practical usefulness of each of these two approaches is likely to be similar.
Claeys and Cauberghe (2014) challenge SCCT’s conclusion that the crisis response should always match the crisis type. In their experiment they found that a matching response strategy only leads to better post-crisis attitudes toward the organization if the subjects have high crisis involvement (it affects them personally) or when the organization’s communication had a rational framing (i.e. straightforward and objective). Conversely, when crisis involvement is low or when the crisis communication is framed emotionally, it does not matter whether or not the response strategy matches the crisis type (Claeys & Cauberghe, 2014; Grappi & Romani, 2015). Note, however, that this does not say anything about the effectiveness of the strategies themselves, just the condition of matching or non-matching.
Studies generally support the conclusions of SCCT, with a few boundary conditions. Crisis involvement and emotional vs. rational framing are two moderators. Another variable is whether the crisis relates to product performance (so-called Corporate Ability, or CA crises) or moral values (so-called Corporate Social Responsibility, or CSR crises). CA crises call into question the organization’s capability to produce quality products, for example the discovery of carcinogens in a shampoo. CSR crises call into question the organization’s commitment to its social obligations, for example the revelation of inhuman working conditions of employees (Jeon & Baeck, 2016). According to Hegner et al. (2016), CA crises may cause more anger in stakeholders, especially consumers. Dutta and Pullig (2011) find that rebuild strategies work best for CA crises, while
11
rebuild and diminish work equally well for CSR crises. This supports the general SCCT proposition that more severe consumer reactions require more accommodative approaches.
To summarize this section: SCCT is essentially a prescriptive theory, arguing that organizations should match their response to a reputation-damaging event, to the threat level of the event. The threat level is influenced by the type of crisis, of which there are roughly three clusters: victim, accident, and preventable (Claeys & Cauberghe, 2014, 2015; Grappi & Romani, 2015). Note that “crisis type” in SCCT refers to the cause of the crisis, i.e. whether it was external or internal to the organization, and whether it was within or beyond the organization’s control (McDonald et al., 2010). However, there are other ways of categorizing crisis events. Many authors distinguish between product-harm (or corporate ability) crises and moral-harm (or corporate social responsibility) crises (Dutta & Pullig, 2011; Hegner et al., 2016; Jeon & Baeck, 2016; Sohn & Lariscy, 2015). Thus for the sake of clarity, the SCCT typology will be called “crisis responsibility” and the CA/CSR distinction will be called “crisis type”. These concepts are independent of each other, because a crisis being accidental or preventable is separate from it being related to the organization’s performance or its moral integrity.
Greater organizational responsibility for the crisis is known to cause negative affect (such as anger) among stakeholders (Grappi & Romani, 2015; Hegner et al., 2016; McDonald et al., 2010), thereby damaging reputation. This is the reason why SCCT recommends more accommodative responses for preventable crises. For reasons of parsimony, this paper will conceptualize crisis responsibility in terms of “internal” and “external” sources, the former being regarded as more preventable. Additionally, CA crises may also cause more anger compared to CSR crises, because product failure directly harms customers and threatens their basic need for personal safety, and because functional benefits are more important than symbolic benefits for brand choice (Dutta & Pullig, 2011; Hegner et al., 2016). Thus, a more thorough accommodative strategy is also recommended in the case of CA crises as the organization needs to do more to earn back consumers’ trust, while a diminish strategy is acceptable in the case of a CSR crisis (Dutta & Pullig, 2011; Hegner et al., 2016). Essentially, the mechanism of action is the same for both variables. Therefore, the following hypotheses are proposed:
H1: Corporate ability (CA) crises will cause more negative consumer responses toward the
organization compared to corporate social responsibility (CSR) crises.
H2: Crises that originate from internal sources will cause more negative consumer
12
As we have seen, organizational responses to reputation crises can affect the emotion, cognition and behavior of stakeholders. Bridging the organizational level and the individual stakeholder level, however, requires a theory of how stakeholders perceive their environments. 2.5 Construal Level Theory
Construal Level Theory (CLT) is a psychological theory about how our mental construals are influenced by psychological distance (Trope, Liberman & Wakslak, 2007). Construals are mental representations of targets – events, activities, objects or people – that are removed from the self in time, space, hypotheticality and social difference (Rizvi & Bobocel, 2016; Trope & Liberman, 2010). These dimensions (temporal, spatial, hypothetical and social) constitute psychological distance, and psychological distance influences the level of abstraction at which the target is construed. Construal level, in turn, has consequences for prediction, evaluation, preference and action (Trope & Liberman, 2010; Trope et al., 2007).
Psychologically distant targets are represented using more high-level, abstract, central, stable, schematic, global and decontextualized construals, whereas near targets are represented using more low-level, concrete, subordinate, detailed, local and context-dependent construals (Rizvi & Bobocel, 2016; Trope & Liberman, 2010; Trope et al., 2007). This has a number of implications. For example, when making decisions, people focus more on the primary, goal-relevant and essential features of the choice when using high-level construals, while they focus more on the secondary, peripheral, incidental and nonalignable features when using low-level construals. Pro arguments, desirability concerns, “why” questions and idealistic values are more salient at high construal levels, while con arguments, feasibility concerns, “how” questions and pragmatic considerations are reflected more by low construal levels (Trope & Liberman, 2010; Trope et al., 2007).
A key feature of CLT is the idea that the various dimensions of psychological distance are interrelated. For example, an event that is perceived to be higher in temporal distance (farther into the future) is likely to also be perceived as higher in spatial distance (in a more remote location). Another premise is that psychological distance and construal level have a bidirectional relationship. In other words, increasing the distance of a target should increase the abstraction of its construal, and vice versa, the more abstract a person’s mindset the more easily distant targets are brought to mind (Trope & Liberman, 2010).
CLT has been applied to a number of contexts. For example, in an interpersonal context, a higher construal level can make people more likely to forgive interpersonal transgressions by reducing the perceived severity of the transgression (Rizvi & Bobocel, 2016). This is probably
13
because abstract processing makes concrete negative details less accessible while making the virtue of forgiveness (an idealistic value) more salient (Rizvi & Bobocel, 2016). Fujita, Trope, Liberman and Levin-Sagi (2006) found that higher construal level can facilitate self-control with regard to resisting temptations that undermine high-level goals. However, McCrea, Liberman, Trope, and Sherman (2008) found that lower construal level can reduce procrastination by making the concrete requirements of a task more salient.
CLT has also been investigated in advertising, where Septianto and Pratiwi (2016) found that emotional appeals in ads (e.g. emphasizing the image of a product rather than its quality) work better when consumers have a low construal level. More intense emotional responses are associated with psychologically proximate events, whereas high-level construal causes consumers to focus on the central positive arguments. However, it may be that only simple emotional states are low-level, while more complex ones like shame are high-level (Septianto & Pratiwi, 2016).
As of yet, CLT has not been applied in the context of reputational crises. However, the influence of people’s processing styles has. Individuals may be categorized as holistic thinkers or as analytic thinkers. Holistic thinkers tend to consider the context as a whole, while analytic thinkers tend to focus on objects detached from their contexts (Monga & John, 2008). Because holistic thinkers are more likely to consider context-based explanations for a brand’s negative publicity (e.g. industry or supplier problems), they are less likely to attribute responsibility solely to the organization and therefore less likely to shift their brand beliefs as a result of the negative publicity (Monga & John, 2008). Holistic thinkers become more like analytic thinkers under conditions of high cognitive load (when it becomes difficult to account for contextual factors), and analytic thinkers become more like holistic thinkers when contextual cues are made more salient (Monga & John, 2008).
Thinking style is linked to construal level. Wang, Wang, Keller and Li (2016) find that holistic thinkers prefer when organizations frame their apologies and remedial solutions in “why” terms (i.e. emphasis on the purpose) and analytic thinkers prefer a framing in “how” terms (i.e. emphasis on the action to be taken). We have already seen that “why” questions are high-level and that “how” questions are low-level. Wang et al. (2016) add that holistic thinkers may pay more attention to and respond faster to global, abstract high-level construals (and symmetrically, that high-level construal facilitates holistic thinking). Conversely, analytic thinkers use more local, concrete low-level construals. Why vs. how framing influences psychological distance and therefore construal level and thinking style (Wang et al., 2016). However, the authors call for
14
more research to directly test whether construal level itself (not just thinking style) can serve as a moderating variable (Wang et al., 2016). This thesis responds to that call.
To summarize: a higher level of construal should decrease the perceived severity of the crisis (Rizvi & Bobocel, 2016), reduce the intensity of negative emotions experienced by consumers (Septianto & Pratiwi, 2016), and increase holistic processing which reduces perceived responsibility (Monga & John, 2008). The effect is expected to be strongest for internal crises and weakest for external-responsibility crises (because internal crises are seen as more severe to begin with). High construal level (but not low level) might also improve consumer responses to CA crises, but worsen consumer responses to CSR crises because moral violations tend to be judged more severely at a higher level of construal (Rizvi & Bobocel, 2016). Therefore, the following hypotheses are proposed:
H3a: Consumers with high construal level will have more favorable responses to the
organization compared to consumers with low construal level in the case of CA crises (but not CSR crises).
H3b: Consumers with high construal level will have more favorable responses to the
organization compared to consumers with low construal level in the case of internal responsibility crises (but not external responsibility crises).
2.6 Conceptual Framework and Hypotheses
In general, there is a lack of research on the role of construal level in the contexts of corporate scandals and reputation repair. This is an interesting gap, because construal level theory has been applied in many different contexts (e.g. advertising), and there are reasons to believe that it would also play a role in this context because it is a psychological factor, which – like emotion and thinking style – influences the way we process communications.
More specifically, previous literature did not look at construal level per se, but the closely-related concept of thinking style (analytic vs. holistic). For example, thinking style moderates the effect of negative publicity (Monga & John, 2008) and the effects of framing a company’s remedial solution in “how” or “why” terms (Wang et al., 2016) on consumers’ brand evaluations. However, an important difference between thinking style and construal level is that thinking style is more like a personality trait, whereas construal level is more context-variable and therefore easier to influence. It is interesting to note that construal level has not been linked to Situational Crisis Communication Theory, one of the key theories in reputation repair. Therefore, linking
15
CLT and SCCT may be a promising approach to begin to address this gap in the literature. A conceptual model doing this is shown below.
An open question is whether the strategy recommendations might change depending on consumers’ level of construal. As shown in the figure above, construal level is expected to moderate the relationship between crisis type/responsibility and consumers’ post-crisis evaluation of the organization. An overview of the hypotheses is as follows:
H1: Corporate ability (CA) crises will cause more negative consumer responses toward the
organization compared to corporate social responsibility (CSR) crises.
H2: Crises that originate from internal sources will cause more negative consumer
responses toward the organization compared to crises with external origins.
H3a: Consumers with high construal level will have more favorable responses to the
organization compared to consumers with low construal level in the case of CA crises (but not CSR crises).
H3b: Consumers with high construal level will have more favorable responses to the
organization compared to consumers with low construal level in the case of internal responsibility crises (but not external responsibility crises).
The contribution of this theoretically novel research is that, by linking CLT to SCCT, managers in marketing and public relations will be better-equipped to select the appropriate response strategy following a scandal or negative publicity, or perhaps to influence the construal level used by audiences in order to mitigate reputational damage.
crisis type (CA/CSR) consumer mindset (low CL/high CL) consumer attitudinal responses crisis responsibility (internal/external) H1 H2 H3a H3b crisis situation
16
3. METHODS
3.1 Pretesting
The purpose of the pretest was to determine which crisis scenarios to use for the main experiment, since one has to be a CA crisis and one a CSR crisis, but it is possible that different scandals may differ in terms of perceived severity, meaning that the measured effect of crisis type on consumer response could actually be the effect of perceived crisis severity. Therefore, to control for perceived severity, the two scenarios selected for the main experiment should score as close as possible in their perceived severity when measured. Recall, however, that it has been postulated earlier that product-harm crises are generally perceived as more severe by consumers than moral-harm crises, so if the effect of crisis type on consumer attitudinal responses were mediated by perceived crisis severity, it would not change the implications of this research.
A secondary aim of the pretest was to determine which crisis scenarios are perceived as CA crises and which are perceived as CSR crises. This is needed due to the lack of an existing typology that exhaustively categorizes crisis scenarios into these two types. Moreover, this also ensures the construct validity of the “crisis type” variable.
The pretest was conducted using an online survey. A sample of N = 12 respondents were reached through convenience sampling. The survey used a within-subjects (repeated measures) design, and presented each respondent with a plain text description of two crisis scenarios, adapted from Study 2 of Jeon and Baeck (2016). For the CA crisis, a fictional coffee brand (“Dominicana Coffee”) received customer complaints that the taste of the coffee was unusually bad. For the CSR crisis, a different brand (“Sonna Caffé”) was being investigated for its executives allegedly having a slush fund. Respondents were asked, for each scenario, to rate its severity on a scale from 0 to 10. They were then asked to rate their agreement with six statements on a sliding scale. These statements were adapted from Berens, van Riel and van Bruggen (2005) and Brown and Dacin (1997), and were designed to measure whether the crisis was associated with corporate ability (CA) or corporate social responsibility (CSR). The statements are shown in the table below.
17
Corporate ability
The company offers sophisticated products.
The company offers high-quality, innovative products. The company’s people have talent and expertise.
Corporate social responsibility
The company contributes to societal welfare. The company is environmentally responsible. The company supports good causes.
Table 2: Scale items for CA and CSR associations
The results of the pretest show that the CA crisis had a mean perceived severity of 5.25 (SD = 1.865) while the CSR crisis had a mean perceived severity of 6.83 (SD = 1.403). According to a paired-samples t –test, this difference is significant at the p < .001 level (t = -6.092). The CA crisis was associated more with corporate ability (mean 51.81 out of 100, Cronbach’s Alpha = .774) than with corporate social responsibility (mean 42.19, α = .719). However, the CSR crisis was not associated more with corporate social responsibility (mean 49.92) than with corporate ability (mean 49.69).
These findings contradict those of Jeon and Baeck (2016), who write that their manipulation of CA/CSR associations was successful, and also that “there was no significant difference in seriousness between the two scenarios”. Given that this pretest used the same brand crisis scenarios included in the appendix of Jeon and Baeck’s paper with minimal alterations to the text, the difference in findings can be attributed to differences in the sample characteristics (e.g. sample size, South Korean students vs. Europeans) and methodology (Jeon and Baeck presented subjects with the crisis scenarios after giving them information on the brands twice per week for two weeks).
Nevertheless, the CA scenario was more associated with corporate ability than the CSR scenario was, and the CSR scenario was more associated with corporate social responsibility than the CA scenario was. This suggests that they are at least somewhat representative of the construct. Therefore, it may still be appropriate to use these crisis scenarios for the main experiment. Furthermore, there was a strong correlation (r = .886, p < .01) between the perceived severity of the two scenarios, indicating that subjects (who saw both scenarios) had a tendency to rate their severity similarly. This was a drawback of the within-subjects design. One possible reason why the CSR scenario was rated more severely on average could have been the text, which described the slush fund in terms of an “outflow of millions”. This number may intuitively seem like a lot, and therefore appear severe, especially when the CA scenario did not
18
include a similar number. It was therefore decided that for the main experiment, the slush fund would instead be described as “a suspected outflow of cash”.
3.2 Main Experiment: Variables
For this research, the three independent variables are crisis type, crisis responsibility, and construal level (i.e. consumer mindset). To instantiate crisis type and responsibility, the same CA and CSR scenarios used in the pretest will be used, with some modifications (including the addition of an internal/external crisis origin). To make the crisis “internal” or “external” to the organization, the source of the scandal was either within the organization (e.g. with its managers) or with a partner in the supply chain (e.g. the plantation supplying the coffee beans). The text stimuli are included in Appendix A. Consumer mindset was operationalized as thinking in concrete terms (low CL) or abstract terms (high CL), and there are validated scales that exist to measure this – for example, asking people to choose if making a list is an example of getting organized (abstract) or writing things down (concrete). The experiment included a manipulation check to see whether respondents were thinking abstractly or concretely, based on 8 items from the “global concerns” index by Fujita, Trope, Liberman and Levin-Sagi (2006).
The dependent variable (outcome) is consumer attitudinal responses to the organization following a scandal, which was measured using Likert scales (7-points) for favorability, trust and purchase intentions. These are commonly used measures. It is not expected that age, gender, income etc. will influence the results; nonetheless age, gender, highest educational level attained, and cultural region were included as control variables. Furthermore, given that familiarity with the organization (as in customer loyalty) might affect the outcome, it was important that the experiment used fictional brands.
3.3 Design
The experimental design is a 2 (CL: high/low) x 2 (crisis type: CA/CSR) x 2 (responsibility: internal/external) between-subjects design carried out online, with all independent variables manipulated for each subject. Therefore there are eight experimental conditions in total. Each subject was randomly assigned to one condition.
Internal responsibility External responsibility
CA crisis CSR crisis CA crisis CSR crisis
Concrete mindset 1 2 3 4
Abstract mindset 5 6 7 8
19 3.4 Procedure
Participants for the online experiment were recruited via email, personal contacts, public Facebook groups, and a website called SurveyCircle. The experiment was implemented using Qualtrics. The experiment began with a briefing, asking participants for their informed consent. They were told the general topic of the experiment (“consumer perceptions of coffee brands, health activities, and art”) but not the details, so as not to skew the results. Participants were not filtered by demographic characteristics.
After consenting to participate, participants were randomly allocated to either the concrete mindset or abstract mindset priming task. Construal level was manipulated using a combination of the procedure of McCrea, Liberman, Trope, and Sherman (2008) and the procedure of Freitas, Gollwitzer and Trope (2004). Participants were first asked to study a painting by Georges Seurat, either by having a general overview (abstract condition) or by examining part of the painting in detail (concrete condition). This was the same stimulus used by McCrea et al. (2008). Afterwards, participants were asked to think about why they would perform the activity of maintaining and improving their health (abstract) or how they would perform that activity (concrete). This was the stimulus used by Freitas et al. (2004). Both procedures were used in order to increase the chances of the manipulation being successful. See Appendix B for the questions.
Next, each participant was asked to read the description of a brand scandal (see Appendix A). They were randomly assigned to one of four versions of the text. The question presented it as an extract from a news service, so as to increase credibility. Afterwards, the participants were asked to evaluate the company. They were asked, “Regarding the story you just read, please answer the following questions, giving your honest personal opinion about Dominicana Caffé”. Three items were included, each scored on a 7-point Likert scale (ranging from “disagree completely” to “agree completely”). These items were (a) I trust the company; (b) I see the company favorably; and (c) I would buy from this company.
After evaluating the brand, participants were asked to complete the Fujita et al. (2006) task, which requires them to match an activity to the description they think fits best. For example, “sweeping the floor” can be described as “moving a broom” (low level) or “being clean” (high level). “Meeting new people” can be described as “small talk and shaking hands” (low level) or as “enhancing one’s social network” (high level). The order of questions was randomized for counterbalancing purposes. Finally, subjects were asked about the significance of the painting they saw earlier (to double-check which mindset manipulation they received), about whose
20
responsibility the crisis was (to check if they perceived it as external/internal), and whether the situation was more an example of Dominicana Caffé’s product performance or social responsibility (to check if they perceived it as CA or CSR). They were further asked a number of demographic questions. The participants were then debriefed, and told the true purpose of the study. They were informed that the brand scandals were fictional.
21
4. RESULTS
A total of N = 80 participants accessed the online experiment, out of which 22 did not complete the experiment and were subsequently deleted from the dataset. This left a sample of N = 58 to be analyzed. There were no further missing values in the dataset. Of these participants, 32.8% were male and 65.5% were female. A cumulative 91.4% of participants fell between the ages of 18 and 34. Participants reported that the cultural region they most identify with is Western Europe and North America in 87.9% of the cases, with the second-most common being Eastern Europe and former USSR (5.2%). The highest levels of education attained were a bachelor’s degree for 48.3%, a master’s degree for 20.7% and a high school diploma for 20.7%. Based on these statistics, we can conclude that the sample mostly comprised well-educated, young individuals from Western cultures.
It was important to check the sample distribution across the eight experimental groups, because they should ideally not differ in any characteristics except those manipulated by the experiment. The following table shows how many of the 58 total participants were allocated into each experimental group.
Internal responsibility External responsibility
CA crisis CSR crisis CA crisis CSR crisis
Concrete mindset 5 (8.6%) 10 (17.2%) 9 (15.5%) 7 (12.1%)
Abstract mindset 8 (13.8%) 5 (8.6%) 7 (12.1%) 7 (12.1%)
Table 4: Number (and percentage) of participants per experimental group
A more detailed cross-tabulation and breakdown by demographic variables is shown in Appendix C. There are some notable differences: participants in the abstract mindset condition were more likely to report having a Master’s degree (29.6% of them) versus those in the concrete mindset condition (12.9%). The abstract mindset groups have a higher proportion of males (44.4%) versus the concrete mindset groups (22.6%). Participants who saw an internal responsibility crisis were more likely to fall into the youngest age bracket (71.4% aged 18-24) and have only a high school diploma (25%) than those who saw an external crisis (43.3% aged 18-24 and 16.7% high school diploma). Participants who saw a CSR crisis were more likely to be male (38% of them were) versus those who saw a CA crisis (27.6%).
4.1 Preliminary Steps
To prepare the data for analysis, the dependent variable was checked for scale reliability. For the three items, Cronbach’s Alpha was .893, which indicates that this was a good scale (using the α > .7 rule of thumb). All three items had corrected item-total correlations above 0.3, indicating that
22
they were good items. To check for normality, a histogram and Q-Q plot were graphed in SPSS. They show a slight positive skewness and kurtosis, but are almost normally distributed.
Fig. 3: Histogram of attitude toward the organization
23
Furthermore, a Shapiro-Wilk test and Kolmogorov-Smirnov test both found significant p-values, indicating that the distribution was non-normal. Therefore, to normalize the distribution, an X* = log10(X) transformation was applied to reduce skewness. An outlier with a zscore of
-3.08 was deleted from the dataset (reducing N to 57 cases). After checking for normality again, it was found that the new distribution had a Shapiro-Wilk statistic of .971 (p = .180) and a Kolmogorov-Smirnov statistic of .095 (p = .200) which were both non-significant, indicating that the distribution is normal.
Regarding the manipulation check, participants’ true experimental conditions (0 = concrete; 1 = abstract mindset) were significantly correlated (r = -.509, p < .001) with their responses to the control question “What was the significance of the Seurat painting you saw earlier?” Participants could answer 1 (“It demonstrates Neo-Impressionism and uses color in order to evoke harmony and emotion” [abstract]) or 2 (“It demonstrates how contrasting individual points of color are used as part of the pointillism technique” [concrete]). This suggests that the mindset manipulation was successful.
However, none of the items from Fujita et al.’s (2006) 8-item list was significantly correlated with the experimental manipulation. Checking the scale reliability showed a Cronbach’s α = .479, meaning that it was not reliable. The alpha could not be raised to .7 by deleting any items. Nevertheless, some items were correlated with each other. Participants could choose to describe each activity with either a low-level or a high-level description. Answers were correlated for “sweeping the floor” and “skydiving” (r = .485, p < .001); for “sweeping the floor” and “recycling” (r = .317, p < .05); for “skydiving” and “attending a family reunion” (r =.333, p < .05); and for “skydiving” and “recycling” (r = .339, p < .01). In other words, participants who chose the concrete description for one item were likely to also choose the concrete description for the other item. None of the significant correlations were the opposite of what was expected. Furthermore, there was a statistically significant correlation (r = -.294, p < .05) between descriptions of “staying home to study” (coded as 1 = “reviewing one’s notes” [concrete], 2 = “exerting self-discipline” [abstract]) and participants’ answers to the control question about the Seurat painting. This may be interpreted as further support for the success of the mindset manipulation.
The crisis responsibility and crisis type manipulations also had control questions. Crisis responsibility (coded as 0 = “internal”, 1 = “external”) was significantly correlated with participants’ answers to the question of who was to blame for the scandal: Dominicana Caffé or its supplier (r = .407, p < .01). Crisis type (coded as 0 = “CA”, 1 = “CSR”) was significantly
24
correlated with participants’ answers to the question of whether the situation reflects the company’s product performance or its social responsibility (r = .369, p < .01). Therefore these manipulations may be regarded as successful.
After the data clean-up, reliability analysis, normality check, outliers check, and manipulation check were satisfactory, the data was ready for analysis.
4.2 Testing Hypotheses
After the preliminary steps, a correlation matrix was computed. This is shown in table 5 below. There were five statistically significant correlations. Age is positively correlated with attitude toward the organization (r = .352, p < .01). Age is also positively correlated with educational level (r = .410, p < .01) and cultural region (r = .403, p < .01). Gender (coded as 1 = “male”, 2 = “female”) was correlated with cultural region (r = .267, p < .05) and educational level (r = -.261, p < .05). Cultural region was not an ordinal variable, but was coded as 1 = “Western Europe and North America”, so the correlation statistic here may suggest that older respondents and female respondents were more likely to identify with a cultural region other than Western Europe and North America. No other correlations were statistically significant.
Variables M SD 1 2 3 4 5 6 7 8 1. Crisis responsibility 0.53 0.504 - 2. Crisis type 0.51 0.504 -0.089 - 3. Construal level 0.47 0.504 -0.015 -0.122 - 4. Attitude 3.2164 1.133 0.089 -0.04 -0.026 - 5. Age 2.67 1.17 0.212 -0.192 -0.061 0.352** - 6. Gender 1.68 0.506 -0.037 -0.129 -0.244 -0.188 0.06 - 7. Education 4.44 1.5 0.162 -0.064 0.216 0.206 0.41** -0.261* - 8. Cultural region 1.46 1.351 0.14 -0.111 -0.14 0.121 0.403** 0.267* 0.129 - **. Correlation is significant at the 0.01 level (2-tailed)
*. Correlation is significant at the 0.05 level (2-tailed)
Table 5: Means, standard deviations, correlations
Since the dependent variable is continuous and the three independent variables are categorical, it was decided that a three-way factorial ANOVA would be the best approach to testing the hypotheses. The assumptions of normality, no outliers, and a between-subjects design are all met. Furthermore, a Levene’s Test did not return any significant (p < .05) statistic, meaning that the “equality of variances” assumption is satisfied. The results are presented in the table below.
25
Sum of Squares DF Mean Square F Partial η2 Sig.
Crisis type 0.396 1 0.396 0.425 0.009 0.517
Crisis responsibility 0.147 1 0.147 0.158 0.003 0.693
Construal level 0.007 1 0.007 0.007 0.000 0.932
Crisis type * Crisis
responsibility 0.146 1 0.146 0.156 0.003 0.694
Crisis type * Construal level 0.262 1 0.262 0.281 0.006 0.598
Crisis responsibility *
Construal level 0.610 1 0.610 0.655 0.013 0.422
Crisis type * Crisis responsibility * Construal
level 0.440 1 0.440 0.472 0.010 0.495
Error 45.663 49 0.932
Total 47.524 57
Table 6: Factorial ANOVA with attitudinal responses as dependent variable
There were no statistically significant interactions at the p < .05 level. There were no moderate or high effect sizes using the η2 > .06 rule of thumb. A table of group means can be
found in Appendix D. The interaction effects are graphically plotted below; the plots show non-parallel slopes.
26
Fig. 6: Construal level * Crisis type interaction
From the plots we can conclude that, in general, participants in the abstract mindset condition had more negative attitudes toward the organization (compared to participants in the concrete mindset condition) in the case of the internal responsibility crisis, but not in the case of the external responsibility crisis. Participants in the abstract mindset condition had more positive attitudes toward the organization (compared to participants in the concrete mindset condition) in the case of the corporate ability crisis, but not in the case of the corporate social responsibility crisis.
To follow up the factorial ANOVA, a one-way ANOVA was conducted for each of the independent variables to isolate their main effects. The mean attitudinal response was more positive for the external responsibility crisis than for the internal responsibility crisis. It was also more positive for the CA crisis than for the CSR crisis, and more positive for the abstract construal level than for the concrete construal level.
27
Sum of Squares DF Mean Square F Sig.
Crisis responsibility 0.051 1 0.051 0.059 0.808 Error 47.306 55 0.86 Total 47.357 56 Crisis type 0.239 1 0.239 0.279 0.6 Error 47.119 55 0.857 Total 47.357 56 Construal level 0.002 1 0.002 0.002 0.962 Error 47.355 55 0.861 Total 47.357 56
Table 7: One-way ANOVA results
Again, there were no statistically significant effects. Finally, a three-way factorial ANOVA was conducted on the control variables of age, gender, and educational level, as they showed statistically significant correlations earlier. The results show no statistically significant effects of these variables on the dependent variable. However, there were moderately-sized main effects of gender on attitudinal responses (F(1, 38) = 3.624, p = .065, η2 = .087) and educational level on
attitudinal responses (F(3, 38) = 1.266, p = .300, η2 = .091). In general, males had more
favorable attitudinal responses than females, and those with higher levels of education (e.g. Master’s degree) had more favorable attitudinal responses than those with lower levels (e.g. high school diploma).
Hypothesis Supported?
H1: CA crises will cause more negative consumer responses toward the organization than
CSR crises. No
H2: Crises with internal origins will cause more negative consumer responses toward the
organization than crises with external origins.
Partially but not Sig.
H3a: Consumers with high CL will have more favorable responses than those with low CL,
in the case of CA (but not CSR) crises.
Partially but not Sig.
H3b: Consumers with high CL will have more favorable responses than those with low CL,
in the case of internal (but not external) responsibility crises. No
Table 8: Overview of hypotheses
Table 8 above shows an overview of the status of each hypothesis following the analysis. The discussion section will elaborate on these results, as well as potential implications and limitations of this research.
28
5. DISCUSSION
Corporate reputation is a crucial asset that must be safeguarded against reputational risks such as brand scandals and crises. Previous research in the field has made forays into the gamut of variables that affect individual stakeholders’ perceptions of an organization following a crisis, for example analytic vs. holistic thinking style, yet much remains unexplored. In particular, the flexible theory of psychological construal level has not yet been applied to the context of organizational crises, and hence this thesis has attempted to answer the question: How does construal level influence consumers’ post-crisis evaluation of organizations? Using a 2 (construal level: low vs. high) x2 (crisis type: corporate ability vs. corporate social responsibility) x2 (crisis responsibility: internal vs. external) between-subjects factorial design, an online Qualtrics experiment with N = 57 participants was conducted to test four hypotheses.
The first hypothesis was that organizational crises that concern the organization’s corporate ability (e.g. product harm) would engender more negative responses on the part of consumers than crises concerning its corporate social responsibility (e.g. moral harm). In the experiment, attitudinal responses toward a fictional coffee brand were solicited using a 7-point scale with three items, measuring trust, favorability, and purchase intentions. The results show a Pearson coefficient of r = -.04 between crisis type and consumer attitude, and an F-statistic of F(1, 49) = .425, η2 = .009 derived from a three-way factorial ANOVA. Neither of these was statistically
significant at the p < .05 level, and therefore the null hypothesis of no effect of crisis type cannot be rejected. How can we interpret this result?
It could be the case that merely reading about a scandal will elicit weaker reactions than being actually affected by a scandal in real life, especially for product-harm crises. Additionally, perhaps the relationship between crisis type and consumer evaluations of an organization is mediated by other variables; extant literature (see Dutta & Pullig, 2011; Hegner et al., 2016) proposes that consumers care more about functional benefits than symbolic benefits when evaluating products. However, this might be too broad a generalization – some cases of product failure may well be seen as less serious than some cases of moral transgressions. In that case, perceived crisis severity could be the mediating variable, and the main experiment of this thesis did not test for that variable. However, the pretest showed that the CSR scenario was perceived as more severe on average. One possible reason why the differences in attitude in the main experiment was not significant is that the attitudinal responses were measured after being exposed to the priming stimuli, such that the stimuli could have closed the gap in perceived severity.
29
The second hypothesis was that organizational crises where the organization itself was responsible – because the event originated from inside the organization and was potentially preventable – would engender more negative responses on the part of consumers than crises where an external agent was responsible. The correlation between crisis responsibility and consumer attitude was r = .089 and the main effect showed an F(1, 49) = .158, η2 = .003, neither of which
was statistically significant. Looking back at the manipulation check, the correlation between the experimental manipulation and participants’ perceptions of who was responsible was statistically significant (at p < .01), but the coefficient size (r = .407) is not very large. If the manipulation were 100% successful, we would expect a correlation closer to 1, but the fact that r was smaller than 0.5 indicates that some participants who saw an external-responsibility vignette still placed most of the blame on the company itself rather than the supplier. However, even participants’ perceived crisis responsibility shows only an association of r = .185 (p = .168) with attitude. This finding is surprising given that a core claim of Situational Crisis Communication Theory is that preventable crises cause more negative feelings among stakeholders than do accidents or events where the organization itself is a victim. Once again, there is a possibility that the experimental interventions closed the gap in perceived severity.
The third and fourth hypotheses were about the moderation effect of construal level. Respectively, it was expected that under conditions of high construal level (i.e. an abstract mindset), the CA crisis will be evaluated more favorably than the CSR crisis, while the reverse would be true under conditions of low construal level (i.e. a concrete mindset). In addition, it was expected that under conditions of high CL, the internal-responsibility crisis will be evaluated more favorably than the external-responsibility crisis, while the reverse would be true under conditions of low CL. The results show a crisis type – construal level interaction of F(1, 49) = .281, η2 = .006 and a crisis responsibility – construal level interaction of F(1, 49) = .655, η2 = .013, neither of which was statistically significant. Despite the small effect, the interaction was still present and could be seen on the plots in figures 5 and 6.
Figure 5 shows that under an abstract mindset, the internal-responsibility crisis resulted in stronger negative attitudes than the external-responsibility crisis while the reverse was the case under a concrete mindset. This relationship was the opposite of the expected relationship. Figure 6 shows that under an abstract mindset, the CA crisis resulted in more positive attitudes than the CSR crisis; however, the reverse was not true under a concrete mindset – the CA crisis was also evaluated more favorably, but the difference was much smaller. Again, this finding may partly be a consequence of the experimental manipulation not being 100% successful. It could also be due to the CSR crisis being perceived as more severe in general.
30 5.1 Limitations and Future Research
A number of shortcomings may help to explain the lack of significant effects in this experiment. The first is to point out the relatively small sample size (N = 12 for the pretest and N = 57 for the main experiment). Ideally, the experiment should have had at least 20 to 40 participants per “treatment” group, for a total of 160 to 320 participants. Not only do small groups reduce the generalizability of findings, but also allow for a greater “noise” to “signal” ratio, making an effect harder to detect.
The second is an issue of methodological approach: this experiment was conducted online rather than in the lab or field, and online experiments have a couple of disadvantages. For example, there is a lack of control over the participants’ environment – they could have taken the survey on their laptop at home while listening to music, or on their mobile phone while travelling on public transport. This means that extraneous influences could not be limited to the same extent possible in a lab setting. At the same time, an online experiment is still an “artificial” setting, meaning that it does not gain the ecological validity of a field experiment. A further disadvantage of this approach is that participants may not always fill in the questionnaire honestly, carefully or seriously – they might be lazy or even deliberately try to thwart the researcher’s efforts.
The third is that this experiment used a hypothetical crisis scenario involving a fictional coffee brand. While there was a good reason to use a fictional vignette (i.e. to control for the potential influence of brand familiarity), the choice may nevertheless carry other implications, for example reducing the potential effect size. This is because a real brand, which participants have actually used, may be more relevant to the participants and therefore lead to a higher degree of crisis involvement and accompanying intensity of attitudinal responses.
The fourth is that ideally, all sample differences between experimental groups should be randomized, but in this case, the groups were not perfectly identical in their demographic characteristics. For example, the average level of educational attainment was slightly higher for participants in the high CL condition and the external-responsibility condition; and there were slightly higher proportions of male respondents in the high CL condition and the CSR crisis condition. Therefore it was not possible to rule out the possibility that men might be more sensitive to the effects of high construal level or judge moral-harm crises more harshly. Again, this might have increased the noise-to-signal ratio, making it harder to isolate the effects of the independent variables.
