Abstract
A crisis can severely tarnish a company’s image and attempts to restore the damaged image may negatively affect postcrisis recruitment efforts. Because company image plays an important role in job seeker attraction, and company crises are on the rise, we examined the effects of postcrisis image restoration on job seeker attraction. We applied image restoration theory to a repeated-measures 2 (negative or neutral company information at Time 1) × 3 (two types of image restoration, and neutral for the same company at Time 2) factorial experimental design. We used the psychology of apologies to understand the underlying mechanisms by which a company’s image changes in the context of image restoration. Results showed that Time 2 ratings of image were more positive after restoration attempts; however, there was no significant difference between the two types of restoration or with the neutral condition. We also found that attraction to the organization assessed at Time 2 fully mediated the relationship between perceptions of image and intentions to pursue a job opportunity. The implications of our findings are that the valence of image restoration influences job seekers’ pursuant behavior more than what is said within image restoration.
Company crises are on the rise. In 2016 and 2017 alone, companies such as Pepsi, United Airlines, Equifax, Samsung, Kobe Steel, Mitsubishi Materials, Toray Manufacturing, Wells Fargo, Apple, Uber, and 21st Century Fox experienced significant crises affecting their business, consumer confidence, and in some cases, put consumers at both financial and physical risk. Crises can quickly damage company image and reputation, which can often take years to recover. Crises create a challenging context for companies because they are forced to respond immediately, regardless of the uncertainty. As these companies attempt to recover their image and restore consumer faith, they may also continue to hire employees, which means they engage in recruitment and selection practices during image restoration.
Recruitment is influenced by individuals’ perceptions of organization image (a collection of knowledge, beliefs, and feelings individuals have about a company; Tom, 1971), which has been shown to be particularly influential on applicants’ attraction to an organization (Aiman-Smith, Bauer, & Cable, 2001; Cable & Turban, 2003; Lievens & Highhouse, 2003; Tsai & Yang, 2010; Überschaer, Baum, Bietz, & Kabst, 2016). Recruitment, a human resource function in organizations, focuses on attracting high-quality applicants expected to progress through a series of review stages (Boudreau & Rynes, 1985; Collins & Han, 2004; Collins & Stevens, 2002), with the ultimate goal of their employment and personal perceptions of fit (e.g., Swider, Zimmerman, & Barrick, 2015; Uggerslev, Fassina, & Kraichy, 2012). Attracting applicants is a crucial step in the recruitment process; the more qualified applicants in the pool, the greater the utility of the selection process and opportunity for competitive advantage (Boudreau & Rynes, 1985; Collins & Han, 2004). Furthermore, attraction leads to applicants’ intentions to apply to the organization (Allen, Scotter, & Otondo, 2004; Chapman, Uggerslev, Carroll, Piasentin, & Jones, 2005), which in turn lead to actual application efforts, known as pursuit behavior (Highhouse, Lievens, & Sinar, 2003; Jaidi, Van, Edwin, & Arends, 2011). Consequently, attraction to the organization is a critical outcome of recruitment efforts (Chapman et al., 2005), as it is the starting point in the hiring cycle (i.e., intentions to apply, application, selection, hire).
However, potential applicants’ attraction to a company depends to a degree on company image, which can be tarnished by crisis events, such as product recalls, employee strikes, bankruptcy, insider trading, environmental disasters, extensive or continuous layoffs, or ethics scandals (Carney & Jorden, 1993). The consequential damage to an organization’s image reduces its attractiveness to prospective applicants (Kanar, Collins, & Bell, 2008). Furthermore, although researchers suggest that companies apologize immediately after a crisis event, the content of the apology message may negate the good intentions of the apology, and worse yet, cause further image damage. For example, in 2017 when United Airlines aggressively pulled a paying passenger off an overbooked plane, the CEO apologized but used language that suggested the passenger deserved the treatment for being difficult and noncooperative. This poorly worded apology resulted in even more negative publicity for United, exacerbating the initial damage to its image. Research shows that how job seekers view the organization affects their pursuit intentions (Cable & Turban, 2003), and negative events that damage the company’s image hurt its attractiveness to job seekers. What remains unclear, however, is whether restoration efforts can reduce or even counteract those negative effects for job seekers, and if so, what must be in the content of the message to ensure the negative effects are ameliorated and not exacerbated? Research shows people have an internal mechanism that automatically directs their attention to negative information (Pratto & John, 1991) and they in turn evaluate it more strongly than even extreme positive information (Ito, Larsen, Smith, & Cacioppo, 1998). Thus, the question remains as to whether image restoration is effective at all for job seekers.
Given the increasing prevalence of corporate crises and image restoration, the purpose of this study was to advance the recruitment literature by proposing that models of recruitment and specifically employer image must account for the context of the situation. We examine how job seekers respond to image restoration efforts after a crisis. The theoretical framework of our study relies on integrating Highhouse, Brooks, and Greguras’s (2009) conceptual model of image types, Benoit’s image restoration theory (1995, 1997) from the crisis management literature, and research on apologies (e.g., Lee & Atkinson, 2019). Our main contribution to the recruitment literature is in identifying the criticality of the context under which recruiting takes place and how that affects employer image, which plays a role at the start of the recruitment cycle.
Organizational Image
Recruitment is a broad area of research, and as such, a comprehensive review of the literature is beyond the scope of this article (for excellent reviews, see Rynes & Cable, 2003; Dineen & Soltis, 2010). Instead, because image is one of the most crucial components of generating applicants, our review of the literature focuses on organizational image and its relation to attraction. Without employer image, prospective employees may not be attracted to the organization, and consequently, may not even apply for a job (Chapman et al., 2005).
Employer image influences applicants’ initial attraction to the organization (Aiman-Smith et al., 2001; Allen, Mahto, & Otondo, 2007; Chapman et al., 2005; Tsai & Yang, 2010) and their subsequent intentions to apply for employment with that organization (Belt & Paolillo, 1982; Chapman et al., 2005; Gatewood, Gowan, & Lautenschlager, 1993; Turban & Greening, 1997). Thus, organizational image perceptions are an important antecedent to the three stages of the recruitment process (i.e., generating applicants, maintaining applicant status, influencing job choices; Barber, 1998).
Though related and often conflated, employer image and reputation are distinct constructs (e.g., Williamson, King, Lepak, & Sarma, 2010). Organizational reputation results from multiple images of the organization shared by all constituents of an organization (Barnett, Jermier, & Lafferty, 2006; Fombrun & Van Riel, 1997; Highhouse, Brooks, & Greguras, 2009), and refers to the collective judgment of an organization based on assessment of its financial, social, and environmental impact (Barnett et al., 2006). In contrast, employer image is a specific perception of an organization held by an individual. Furthermore, an organization has a single reputation, but individuals can hold different types of image perceptions resulting from their relationship or interest in the organization (Highhouse et al., 2009; Treadwell & Harrison, 1994). For instance, job seekers, current employees, and investors likely have different organizational image perceptions.
Job seekers can have distinct images for the same company: an overall image and an employer image (Gatewood et al., 1993). Gatewood et al. found the two images were related (r = .44), but the image of a company as an employer better predicted recruitment outcomes, such as likelihood to respond to a job advertisement, than did overall image. Thus, because our focus is on job seekers and the recruitment process, we focus on employer image in the current study.
Organizational Crisis
Although recruitment research shows that image and attraction perceptions for job seekers exposure is influenced by media or word of mouth messaging (Kanar, Collins, & Bell, 2010; Van Hoye, 2008; Van Hoye & Lievens, 2005, 2007), much less is known about the specific content of those messages especially within a restoration context. The focus of research to date has been on examining the source (e.g., word of mouth, news, job advertisement, web; e.g., Frasca & Edwards, 2017) and the valence of information (i.e., positive or negative), while ignoring the message content. For example, job seekers’ image perceptions change in response to varying sources of information (Barber, 1998; Kanar et al., 2008; Rynes & Cable, 2003), such as public relations, word of mouth, and media exposure. When word of mouth and media exposure are negative, job seekers’ attraction to the organization is low (Kanar et al., 2010; Van Hoye & Lievens, 2005, 2007).
In these studies, the context has been a typical job-seeking/recruiting situation in which organizations strategize and carefully develop their marketing literature to enhance their employer image. Crisis and image restoration comprise a very different context because of the immediacy of information based on uncertainty inherent in crises. Moreover, when the crisis is the result of humans (as opposed to a natural disaster), attempts to recover from the crisis are fraught with attributions of blame, questions about whether the situation could have been avoided and about what will be done to repair damage. The context is further complicated as there may be at least two possible conditions, if not more, that job seekers confront: (1) exposure to negative information about the company and then exposure to an image restoration attempt or (2) job seekers have only been exposed to an image restoration attempt without prior exposure to the event triggering the image restoration effort. In both cases, information associated with the crisis and recovery establishes the context.
Image Restoration Theory: Five Strategies
Based on actual company responses, Benoit (1995, 1997) developed a five-part typology for how companies respond to crises. The five general approaches companies’ employ to restore their image in the wake of a crisis include (1) denial, (2) evade responsibility, (3) reduce offensiveness, (4) mortification, and (5) corrective action. Researchers, including Benoit, refer to the typology as a theory and apply it as such (Arendt, LaFleche, & Limperopulos, 2017; Avery, Lariscy, Kim, & Hocke, 2010; Coombs & Schmidt, 2000; Dardis & Haigh, 2009; Haigh & Brubaker, 2010; Harlow, Brantley, & Harlow, 2011). Consistent with previous research, we refer to his typology as image restoration theory. Each approach in the image restoration theory subsumes one or more specific strategies, reviewed next.
The first strategy is denial, which includes two specific tactics: simple denial in which you claim the negative event did not happen and shift the blame for the negative event to someone other than the company or entity. Research shows the most commonly used but least effective strategy is denial (Arendt et al., 2017).
The second strategy, evading responsibility, can take four different forms. The first is to cast the event as a reasonable response to another’s provocation. A company may also use defeasibility, which is a claim of insufficient information or lack of control over the situation. Companies also report that the negative event occurred by accident. Last, a company can evade responsibility by asserting it had good intentions.
Reducing the offensiveness of a situation, the third tactic, includes six specific tactics. Namely, a company can (1) bolster positive feelings by emphasizing positive qualities of the company or past positive acts; (2) minimize negative feelings about the crisis by downplaying the severity of the situation; (3) differentiate themselves from other similar but more offensive actions, casting the current crisis as not as serious as other previous situations; (4) engage in transcendence, or place the act in a more favorable context thereby emphasizing how the good of the company’s actions outweighs the bad; (5) attack accusers such as competitors, government agencies, or the media; and (6) reduce offensiveness by offering compensation to victims of the crisis.
Mortification, the fourth tactic, is simply confessing to wrongdoing and then begging forgiveness, but not offering corrective action or compensation to victims. The last approach in image restoration theory is corrective action, in which a company can promise to correct the problem, such as taking steps to return to its state before the crisis, or the company can promise to prevent reoccurrence, such as taking steps to prevent a future crisis. Recent research shows corrective action is the most successful strategy (Arendt et al., 2017).
Findings on Image Restoration Theory
Coombs (1998) expanded Benoit’s framework by plotting crisis response strategies onto a continuum from accommodative to defensive, based on the amount of responsibility an organization takes for the crisis. Coombs and Schmidt (2000) later added an additional strategy called separation. Accommodative strategies acknowledge fault and accept the most responsibility, defensive strategies deflect responsibility, and separation shifts blame to a separate group. Mortification and corrective action are at the accommodative end of the continuum. Denial, reduce offensiveness, and separation lie at the defensive end of the continuum. Though Coombs (2007) proposed that accommodative, responsibility-taking strategies should result in higher image restoration than defensive actions that acknowledge the least amount of responsibility for the crisis, empirical findings have demonstrated mixed results, with most recent experiments (e.g., Dardis & Haigh, 2009; Haigh & Brubaker, 2010) suggesting that defensive strategies may be best for restoring image. To date, although image restoration theory offers an initial framework for actions taken by organizations to restore corporate social responsibility and market image after a crisis, the theory is still without a clear underlying explanation or mechanism for how employer image restoration works.
Toward Understanding Employer Image Restoration in Recruitment
Constituents of an organization draw on similar general information, yet they form specific images based on their goals (Elsbach, 2006; Highhouse et al., 2009). Relevant to our study, job seekers form their image perceptions based on information specific to their goals of finding employment. However, organizations in the midst of crisis and image restoration focus on communications in reaction to the crisis, and not on ensuring a positive image for attracting potential job applicants. Yet, these organizations still recruit and hire during crisis events, as well as after, while image restoration is well underway.
Research on interpersonal apologies and forgiveness after transgressions provides a psychological perspective on the process of image restoration that may provide guidance for how employer image is viewed and/or potentially restored using some of Benoit’s (1995, 1997) strategies. Specifically, interpersonal transgressions, such as revealing a secret or telling a lie, may be considered similar to crises where employer image is damaged. That is, in both transgressions and crises, the transgressor (an individual or a company) must restore his or her image in the eyes of an audience (interpersonal relation or organizational constituent, such as job seekers). People apologize to maintain their favorable relationship and to restore the positive evaluations (e.g., liking) that others have of them (Schlenker & Darby, 1981). As such, interpersonal apologies act analogously to image restoration by repairing image damage and potentially increasing likability (or attraction). Holtgraves (1989) found that a full-blown apology yielded the most satisfaction from those receiving the apology. In addition, Lee and Atkinson (2019) found that employees directly involved in the crisis recovery efforts were positively affected by apologies, likely because they, in effect, were asking for forgiveness on behalf of the organization.
Apologies are generally effective because they increase empathy for the transgressor (McCullough et al., 1998; McCullough, Worthington, & Rachal, 1997), demonstrating concern for the victim and restoring relational balance (Fehr & Gelfand, 2010). Including compensation and showing empathy as part of an apology renders it more effective and appears to map onto reduced offensiveness restoration strategies (Fehr & Gelfand, 2010).
A company’s acceptance of responsibility may also improve image perceptions because of similar psychological mechanisms that drive interpersonal forgiveness. Apologies promote forgiveness because they reduce negative affect directed toward the transgressor (McCullough & Witvliet, 2002; Ohbuchi, Kameda, & Agarie, 1989). As with an interpersonal transgression, accepting responsibility may reduce negative affect directed at the company (Coombs, 2007). In turn, reduced negative affect, which is related to interpersonal forgiveness, may increase empathy for the company (Fehr & Gelfand, 2010; Fehr, Gelfand, & Nag, 2010; Konstam, Holmes, & Levine, 2003; McCullough et al., 1997; McCullough et al., 1998).
Negative Information
Research has shown that among job seekers negative information results in more negative image perceptions and greater subsequent reduction in company attractiveness than positive information results in improvements in image perceptions and increases in attraction (Kanar et al., 2008; Kanar et al., 2010; Van Hoye & Lievens, 2005, 2007). These findings are consistent with social psychological theory and research suggesting that negative information tends to have a stronger impact on attitudes than positive or neutral information (Baumeister, Bratslavsky, Finkenauer, & Vohs, 2001; Rozin & Royzman, 2001).
Furthermore, signaling theory (Rynes, 1991; Spence, 1973) suggests that job seekers rely on general impressions, such as image, to judge company attractiveness. Despite the presence of positive or negative information, job seekers also view information that has no valence (neutral) when learning about a potential employer, namely because they tend to possess very little information about an employer. As such, reactions to neutral information can serve as a baseline comparison for reactions to negative information. Based on the potential threat associated with and additional scrutiny given to negative information (Rozin & Royzman, 2001), we expect our research to be consistent with previous findings.
Restoration: Ameliorating Negative Information
Image restoration information can positively influence image perceptions after a crisis and ameliorate the effect of negative information on image perceptions (Benoit, 1995, 1997; Coombs & Schmidt, 2000; Dardis & Haigh, 2009; Haigh & Brubaker, 2010). In particular, restorations that include transgressions and apologies function to increase empathy and likeability, which improve image perceptions. A combination of research on interpersonal transgressions and crisis management suggests that reduce offensiveness and/or corrective action should function as effective strategies for employer image repair. Therefore, based on this integrated research, employer image should be higher after image restoration actions than before.
Because past theoretical or empirical work is unclear as to which of the two image restoration strategies should be most effective, we propose the following research question:
Restoration: Where There Is No Awareness of Crisis
To date, image restoration theory only includes cases where stakeholders are aware of the negative crisis prior to restoration communication. We propose that job seekers who initially know nothing of the crisis, but are exposed to details of a crisis as part of image restoration will likely have more negative image perceptions than if they had seen no image restoration. Namely, a company engaging in image restoration has to acknowledge prior negative actions to cast the organization in a more positive light, provide compensation to victims, or promise that the action will not happen again. Because people find negative phenomena more salient than positive (Baumeister et al., 2001; Rozin & Royzman, 2001), job seekers with no prior exposure to the crisis hearing the image restoration communication are likely to attend to the negative information (Pratto & John, 1991), giving it substantial weight (Ito et al., 1998) in their image formation. Viewing negative information at this stage in recruitment cues negative attributes about the company, thus reducing attraction to the company (signaling theory: Rynes, 1991; Spence, 1973). In particular, the negative content of the image restoration should have greater potency and complexity than the restoration content, thus making it salient to participants (Rozin & Royzman, 2001), and eliciting more scrutiny of the company.
Image Restoration, Attraction
Image is a key antecedent of organizational attraction (Allen et al., 2004; Chapman et al., 2005; Lyons & Marler, 2011), but it is also a distal predictor of behavioral outcomes of recruitment. Research has shown that intentions to pursue a job mediates the relationship between attraction and job choice (Allen et al., 2004; Chapman et al., 2005). Thus, to adequately predict behavioral recruitment outcomes, researchers need to assess intentions toward specific recruitment behaviors (such as pursuit behaviors), in addition to attitudes, consistent with the theory of planned behavior (Ajzen, 1991; Ajzen & Fishbein, 1977). The theory states that attitudes are only one predictor of behavior and that intentions mediate the relationship between attitudes and behavior. Subjective norms and perceived behavioral control are also antecedents of behavior in Ajzen and Fishbein’s theory, though all three predictors are not required in every research model. Moreover, subjective norms, what others’ think of the applicant applying for the job, and behavioral control, the applicant’s self-efficacy beliefs that he or she has the ability to apply for the job, are particularly relevant in the later stages of recruitment when job choice is the key outcome (Barber, 1998), and applicants are beyond the initial attraction stage. Therefore, we focus on attitudes and intentions only, excluding subjective norms and behavioral control, to maintain a parsimonious model and consistency with the goal of focusing on generating applicants.
Image restoration theory suggests that the application of restoration strategies should improve image after a crisis and several studies support these suppositions. Furthermore, recruitment literature provides ample evidence that positive image perceptions relate to attraction, which in turn should predict indirectly (via the theory of planned behavior) pursuit behaviors. We hypothesize that after image restoration, image will positively predict attraction, which will in turn predict intentions to pursue a job.
Materials and Method
Participants
Power analysis with G*Power (Version 3.1.9.4; Faul, Erdfelder, Lang, & Buchner, 2007) indicated a sample of 183 participants would be adequate to detect a medium multivariate effect size (.25) with a power of .80. We recruited 200 undergraduates through a psychology department subject pool, in which students receive course credit for participation in research. Of the 200 we recruited, 114 completed both Time 1 and Time 2 surveys for a response rate of 57%. Participants averaged 19.42 years old (SD = 1.72 years) and were mostly female (68%). The majority of the participants (60%) were unemployed at the time of the study. Of those employed, 30% worked part-time, 9% worked in a temporary setting, and 1% reported full-time status. Those who were employed had an average job tenure of 13 months (SD = 15.37). Although most participants were not working, 44% reported actively seeking a job at the time of the study.
College students may have more malleable employer image perceptions than those who have been in the workforce for some time (Barber, 1998). This malleability may be due to how college students base image perceptions largely on familiarity (Gatewood et al., 1993). For this reason, companies devote considerable marketing energy to cultivating positive image perceptions among college students (Yeager, 1991). Thus, college students are a good audience for this study because they are the target of company recruitment, and importantly, they tend not to follow the news as closely as working adults (Jarvis, Stroud, & Gilliland, 2009; Mindich, 2005), which suggests they likely do not have as much background on the crisis we chose for this study.
Design and Procedure
We conducted a preliminary study with 227 students (different from the study sample) to examine college student exposure to news about technology companies, as a means of testing students’ exposure to Hewlett-Packard (HP), the target company for our experiment. Too much exposure might confound our manipulations by introducing a halo or predetermined bias against the organization, but no exposure might suggest an inability to understand the manipulation or appreciate the severity of an organizational crisis. Preliminary study participants had read on average one article about HP in the past 3 months. Though participants read more news about HP than comparison companies: Cisco Systems, Oracle, and Research in Motion, they read less news about HP than about Apple Inc. Participants read similar amounts of news related to Samsung and Dell as they did about HP. Thus, the study indicated students were only moderately exposed to information about HP in the months prior to the study.
The main study design was a repeated-measures 2 × 3 factorial experiment, conducted over two sessions separated by 2 weeks, and offered online (via web-based survey and materials). Time was the within-subjects factor. We measured the dependent variables, image and attraction, at two times. The within-subjects component allowed for an examination of image perceptions over time and participants served as their own baseline control group. Time 1 information (negative or neutral), image restoration (reduce offensiveness, corrective action), and neutral information (comparison for image restoration) were the experimentally manipulated between-subjects factors.
Our cover story was that the university was assisting a number of technology companies in understanding what college students look for in internships, thereby placing the study into a recruitment context. Participants were encouraged to believe their responses were going to a company that at that time was recruiting interns. Considering the majority of participants were freshmen and sophomores, we used internships in this study rather than full-time jobs that would be more appropriate for students closer to graduation. Internships are an important recruitment and selection tool for companies, with 60% of internships turning into job offers for full-time employment (Zhao & Liden, 2011).
Participants viewed stimuli and completed measures on a web-based survey system that randomly presented the stimuli at Time 1 and Time 2 to ensure an equal number of participants in each condition. We coded participants for condition, based on the combination of Time 1 and Time 2 stimuli they viewed. We collected data via the web, consistent with previous investigations of image and attraction in recruitment (e.g., Cable & Yu, 2006; Kanar et al., 2008; Walker, Feild, Giles, Bernerth, & Short, 2011). At Time 1, participants created a unique ID number to match their Time 2 responses with their Time 1 responses. We randomly assigned participants to view either negative information about HP or a neutral overview of HP from Reuters. Participants then answered image, attraction, and familiarity items about HP, as well as source credibility, demographic, and vocational preference questions. We continued the cover story by telling participants they would need to complete a second part of the study because the company wanted to do a follow-up with them in 2 weeks. We chose 2 weeks as the wait period to allow more time to mitigate the recall of negative information than a 1-week span as previous studies showed participants still had some recall of negative recruitment information after one week (Kanar et al., 2010). We reminded participants that they were not finished with the study and would not receive full credit for the study until they completed the second part of the study. Two weeks later, we sent participants an e-mail invitation to complete the second part of the study.
At Time 2, we randomly assigned participants to view one of three stimuli: a reduce offensiveness image-restoration article, a corrective action image-restoration article, or the neutral article about the computer release. We exposed participants to one of six conditions derived from the combination of Time 1 and Time 2 random assignment. Participants completed image and attraction items, as well as source credibility items, and then viewed the HP jobs website described below. They selected an internship of interest to them and then answered questions about their interest in the internship. They completed items on their employment, recruitment, and organization crisis experience. Presenting the crisis items at the end of the study prevented priming of negative events that could influence image perceptions.
Materials
We selected HP as the target company for image restoration because of its repeated scandals involving the board of directors, including the chairperson of the board (Kaplan, 2006), and because HP was long considered a well-regarded company on Fortune’s list of most admired companies in 2006, before the original scandal (Fortune, 2006). Fortune’s list of most admired companies is a common metric that researchers have used to assess organization image and reputation (Brooks, Highhouse, Russell, & Mohr, 2003; Flanagan & O’Shaughnessy, 2005; Fombrun & Shanley, 1990; Gatewood et al., 1993). HP’s prior reputation is important to consider because a halo effect of previous positive reputation can shield a company from the effects of crisis (Coombs & Holladay, 2006). HP recovered from the scandal in 2006 and ranked highly on Fortune’s list of most admired companies in 2010 (Fortune, 2010). However, after its successful recovery in 2010, HP again experienced over a year of sustained negative publicity and instability, including the dismissal of two CEOs, removal of board members, mismanaged high-profile product releases, and contradictory announcements regarding the direction of the company’s business. Because of its history of profitability, care for stakeholders (employees, shareholders, customers), reputation of high integrity, and yet experienced several major scandals (Marconi, 1997; Schwartz, 2000), HP is a good target company for this study.
HP employed image restoration in response to its many crises, but the messages were targeted at investors and customers, not job seekers. Generally, image restoration research has focused on response to a single instance of crisis (e.g., Benoit & Czerwinski, 1997; Coombs & Schmidt, 2000; Dardis & Haigh, 2009; Haigh & Brubaker, 2010; Harlow et al., 2011; Jaques, 2008; Stromback & Nord, 2006); therefore, this study contributes to the literature by examining sustained negative publicity due to several crises.
Negative information comprised the actual content of one news article and two opinion pieces by technology journalists summarizing a previous year of negative publicity targeted at HP. We presented the articles as screenshots taken from their original websites. Job seekers rarely view a single piece of information about a company. However, most recruitment research uses a single source (e.g., one news article or one employee testimonial) of information in manipulations. Hence, by providing multiple articles, our study closely resembles the actual experiences job seekers have as they learn about a company. Our study thus incorporated a rich manipulation of negative information, addressing a weakness of previous studies (Lievens & Chapman, 2009).
At Time 1, neutral information was an overview of HP from Reuters’ stocks section. Reuters is a news organization with a history of providing news articles and financial data to other news providers. The article offered an overview of HP’s history and business structure, but did not discuss recent news or provide information about HP as an employer. As such, the article does not include evaluative information or recruitment-like statements. At Time 2, neutral information was a news article on a technology website announcing a new HP tablet computer. The article did not discuss recent news or HP as an employer, only the features of the tablet computer. We presented both Time 1 and Time 2 neutral stimuli as screenshots from their respective websites to provide a realistic source of information.
There were two image restoration articles about HP: reduce offensiveness (e.g., “It’s really important to me to make the right decision, not the fast decision”) and a corrective action (e.g., “To make HP a great company once again, we need more than competitive costs and operational efficiency. We’re in the process of assessing and refining our growth strategy, and the same concepts that were behind our operational changes will be at work here: simplicity, focus, alignment, and execution”). Though created for the study, the articles included genuine publicly made quotes from HP leaders so that the image restoration content came directly from the company (Dardis & Haigh, 2009; Haigh & Brubaker, 2010). We presented the articles as screenshots from the same website as the Time 2 neutral information. The articles retained the formatting (including header, sidebar links) of the neutral article, but with text changed to reflect image restoration.
We showed a screenshot of the HP student job home page, which includes general information about the company and their internship program, along with six screenshots of internship pages. The internship opportunities were standardized for content and format, but retained the format of HP’s actual job site. Internships were in the following areas to represent a broad range of interests and majors included within the psychology subject pool: engineering, computer science, business administration, marketing, graphic design, technical writing, and human resources. Participants provided a rating of the internship they chose and a narrative on why they did or did not prefer the available internships.
Measures
All items included a 5-point Likert-type response scale (1 = strongly disagree and 5 = strongly agree). Alpha reliability estimates are for the data in this study.
General image was assessed with a three-item measure by Schwoerer and Rosen (1989). We obtained acceptable reliability of scores at Time 1 (α = .82) and Time 2 (α = .79). Trait-based image was assessed using Cable and Yu’s (2006) 8-trait, 16-item measure. The scale assesses powerful, achievement-oriented, stimulating, self-directed, universal, benevolent, traditional, and conforming traits using two adjectives each. For example, achievement-oriented is composed of “successful: achieving goals” and “capable: competent, effective, efficient.” Participants rated their agreement as to how well each adjective described HP. Reliability of scores at Time 1 (α = .89) and Time 2 (α = .88) were acceptable. Participants completed four dimensions of instrumental and symbolic image items (Lievens & Highhouse, 2003). We assessed four dimensions of instrumental characteristics: teamwork opportunity, advancement opportunity, pay and benefits, and task diversity and five symbolic traits: sincerity, innovativeness, competence, prestige, and robustness. Both reported acceptable reliability of scores at Time 1 and Time 2 respectively: instrumental items, α = .88, .90, and symbolic, α = .89, .84.
Participant attraction to organization (Time 1 α = .92; Time 2 α = .87) was assessed with a five-item measure (Highhouse et al., 2003). An example item is “For me, HP would be a good place to work.”
A popular recruitment outcome construct is intentions to apply to the company (Chapman et al., 2005; Rynes & Cable, 2003). Intentions to apply was assessed with four items (Roberson, Collins, & Oreg, 2005) at Time 2. An example item is as follows: “If I were searching for a job, I would apply to HP.” Reliability of scores was acceptable (α = .90).
Familiarity is related to possessing both positive and negative information and judgments about a company (Brooks et al., 2003), which may affect image perceptions and attraction. Research suggests that job seekers with less familiarity with a company change their image perceptions more over time than do job seekers with more familiarity with the company (Kanar et al., 2008). Familiarity with the target organization (α = .84) was measured at Time 1 using Cable and Turban’s (2003) three-item scale and used as a control variable.
Credibility relates to image perceptions, such that negative information perceived as credible is associated with more negative image perceptions among job seekers (Cable & Yu, 2006). Therefore, we assessed credibility as a control variable using a four-item measure of credibility for online sources (Johnson & Kaye, 2002). Reliability of scores were acceptable at Time 1 (α = .70) and Time 2 (α = .69).
We assessed vocational preferences of participants to determine whether potential lack of interest in internships was based on the specific internships reflected on the jobs website, rather than on image. Participants reported in which occupational category they were most interested in pursuing employment. Nine categories based on the O*Net system Job Families were listed with example jobs for each category (National Center for O*NET Development, 2001). O*Net Families that do not typically require a college degree (e.g., military specific occupations, maintenance) were not included because our sample comprised degree pursuing students. We collapsed categories with similar traits to yield a smaller number of categories amenable to analysis. For instance, business and financial operations occupations, management occupations, and sales and related occupations were collapsed into one group. Participants had the option to write-in another occupation, if available groups did not fit their preferences.
To understand the extent to which participants perceived HP as responsible for the negative events presented in the study, we assessed attribution of responsibility using a six-item scale (Struthers, Eaton, Santelli, Uchiyama, & Shirvani, 2008) adapted for this study. The original scale items referred to a specific individual as the referent; therefore, in this study, we modified items to refer to HP. Additionally, we standardized the response scale to match the other measures in this study. The scale assesses three components of responsibility based on Weiner’s (1995) theory of attribution: internal locus of control, controllability, and inference of responsibility. Internal consistency reliability was acceptable (Time 1: α = .74; Time 2: α = .67) after the removal of the two items assessing internal locus of control.
Demographics included age, sex, college major, and grade point average. Work experience items included number of jobs held, current employment (part- or full-time), and number of job searches undertaken in the past 5 years. At the end of the study, participants also indicated whether a past or current employer had experienced any of the common organizational crises: product recalls, employee strikes, acquisitions, bankruptcy, insider trading, environmental problems, layoffs, and ethics scandal (Carney & Jorden, 1993; Kline, Simunich, & Weber, 2009).
Results
Table 1 includes descriptive statistics and correlations between variables, and Table 2 includes cell means for each condition. Univariate effects for each image dimension mirrored the multivariate effects (see Table 3 for all multivariate and univariate effects).
Correlations Between Study Variables.
Note. T1 = Time 1, T2 = Time 2. Coefficient alpha reliability coefficients along the diagonal bolded and italicized.
p < .05. **p < .01.
Image Means by Study Conditions.
Note. T1 = Time 1; T2 = Time 2; SD = standard deviation.
Repeated Measures Multivariate Analysis of Variance and Univariate Effects for Each Image Dimension.
Note. T1 = Time 1; T2 = Time 2; Time = within-subjects effect; df = degrees of freedom.
p < .01. ***p < .001.
There was a significant multivariate main effect for Time 1 condition (neutral vs. negative), Λ = .60, F(4, 102) = 16.84, p < .001, η2 = .40, consistent with Hypothesis 1. Participants who read negative information reported more negative image perceptions than those who read neutral information at Time 1. There was not a significant multivariate main effect for Time 2 condition (neutral vs. reduce offensiveness vs. corrective action), Λ = .95, F(8, 204) = 0.61, p = .77, η2 = .02, suggesting that image perceptions did not differ as a function of information read at Time 2. The within-subjects main effect was significant, Λ = .41, F(4, 102) = 17.45, p < .001, η2 = .41, indicating that image ratings increased from Time 1 to Time 2. However, the change from Time 1 to Time 2 did not differ between the image restoration strategies and the neutral condition. These results indicate partial support for Hypothesis 2.
There was a significant interaction indicating that image perceptions increased significantly at Time 2 (participants read either positive or neutral information at Time 2) for participants who read negative information at Time 1, but did not increase significantly for those who read neutral information at Time 1, Λ = .23, F(4, 102) = 7.48, p < .001, η2 = .23. Examining the interaction plots (see Figure 1) illustrates that image ratings increased from Time 1 to Time 2 for those who viewed negative information at Time 1 (right graph), but remained equivalent between Time 1 and Time 2 for those who viewed neutral information (left graph).
Interaction effects for each image dimension.
There was not a Time 2 condition interaction, Λ = .95, F(8, 204) = 0.67, p = .72, η2 = .03, nor a three-way interaction with Time 1 condition, Λ = .97, F(8, 204) = 0.42, p = .91, η2 = .02. These findings indicate there was no difference in change between corrective action and reduce offensiveness strategies, which answers the research question. Image ratings were higher at Time 2, but so were image ratings for those who viewed neutral information. Participants who viewed neutral information at Time 1 and image restoration at Time 2 showed no change. Thus, Hypothesis 3 was not supported.
To examine Hypotheses 4 to 7, we conducted mediation analyses using PROCESS (Hayes, 2013). PROCESS is an SPSS macro that tests mediation, as well as conducts a Sobel’s test and computes the bootstrapped indirect effect for the mediation. PROCESS is as effective as structural equation modeling for assessing mediation (Hayes, Montoya, & Rockwood, 2017). Results show that attraction at Time 2 mediated the relationship between each image dimension and intentions to pursue (see Table 4). Each image dimension significantly predicted intentions to pursue, but coefficients became nonsignificant when attraction was included in the model. Furthermore, the indirect effect confidence intervals did not include zero and each of the Sobel’s test z values were significant. Thus, Hypotheses 4 through 7, outlining the mediation relationship, were all supported.
Intentions to Apply Regressed on Image With Attraction as a Mediator.
Note. SE = standard error; CI = confidence interval.
p < .05. **p < .01. ***p < .001.
Neutral Article, Participant Familiarity, and Assumed Responsibility
The neutral article was perceived as significantly more credible (M = 3.56, SD = 0.47) than the negative article (M = 2.90, SD = 0.50), F(1, 110) = 41.54, p < .001. However, there was not a significant difference in the perceived credibility of the three articles at Time 2, F(2, 110) = 1.15, p = .32, suggesting that participants saw the neutral source and image restoration articles as having equivalent credibility. Source credibility at Time 1 was a significant covariate in a multivariate analysis of covariance (MANCOVA), Λ = .89, F(4, 96) = 3.02, p = .02, η2 = .11, but did not significantly interact with the other variables. This suggests that although the Time 1 neutral article was perceived as more credible than the negative information, this difference did not influence the process of image restoration. We conducted similar analyses to examine the influence of participant familiarity with HP on the process of image restoration. Familiarity was not a significant covariate in a MANCOVA, Λ = .91, F(4, 95) = 2.25, p = .07, η2 = .09, suggesting that initial familiarity with HP also did not influence participants’ change in image perceptions.
To assess the influence of assumed responsibility (a driver of forgiveness in interpersonal apologies) on image restoration, we conducted repeated-measures MANCOVA. Results of preliminary one-way analyses of variance show that perceived responsibility for HP’s action did not differ between the conditions, both at Time 1, F(1, 110) = 1.53, p = .22, and Time 2, F(2, 110) = 0.24, p = .79. Responsibility at Time 2, however, was a significant covariate, Λ = .85, F(4, 96) = 4.26, p = .003, η2 = .15. For ease of interpretation and visual illustration, we separated responsibility at its mean to yield high and low responsibility groups. Table 5 shows the image rating means separated by high and low responsibility and Time 1 condition. Table 6 shows multivariate and univariate effects. Figure 2 illustrates the interaction between Time 1 condition and responsibility between Time 1 and Time 2. When participants viewed HP as more responsible for its actions (right graph) image ratings increased from Time 1 to Time 2 rather than remaining flat, as they did when participants viewed HP as less responsible for its actions (left graphs). This trend appeared only for those who viewed negative information at Time 1 (dotted line), whereas those who viewed neutral information at Time 1 had equivalent image ratings between Times 1 and 2. These interactions suggest that perceiving high responsibility enhances the process of image restoration.
Image Means by Responsibility and Time 1 Condition.
Note. T1 = Time 1; T2 = Time 2; SD = standard deviation.
Repeated Measures Analysis of Covariance With Responsibility as a Covariate.
Note. T1 = Time 1; T2 = Time 2; Time = within-subjects effect.
p < .01. ***p < .001.
Interaction of responsibility with T1 condition.
Discussion
In general, participants who viewed negative news articles and opinion pieces about HP had lower image ratings than those who viewed a neutral overview of the company. Considering the results for image restoration did not differ from the neutral source as expected, it may be that the sources of image restoration were not strong enough manipulations to differentiate them from the neutral source of information. The two image restorations strategies we used were previously shown to be effective in restoring image (Dardis & Haigh, 2009; Haigh & Brubaker, 2010). Less effective image restoration strategies, such denial or evade responsibility, may have not shown the positive shift in image perceptions that the corrective action and reduce offensiveness strategies did.
Post restoration image perceptions were more positive for those who initially viewed negative information, but there was no difference between the corrective action and reduce offensiveness image restoration strategies; nor were either significantly different from the neutral condition. Additionally, image perceptions for those who viewed a restoration attempt without knowledge of the crisis were not lower than initial image perceptions, as expected.
We did not compare image restoration and neutral information to additional negative information about the company; therefore, this is an area for further study. Researchers can clarify if the mechanism for image restoration is a lack of negative information or if it is the nature of the positive information provided. Additionally, less effective image restoration strategies (e.g., denial or evade responsibility) should be used as comparisons to more effective image restoration strategies (i.e., corrective action and reduce offensiveness).
Hypotheses derived from the theory of planned behavior (Ajzen, 1991; Ajzen & Fishbein, 1977), suggesting a chain of recruitment outcomes, were supported. Results demonstrated that image perceptions were related to intentions to apply for employment, but only through their relationship with organizational attraction. Though previous research has implied such a causal chain, this is one of the first studies to test the full chain in an image restoration context.
We combined crisis communications theory and the psychology of apologies to propose that HP showing responsibility for its actions would influence the process of image restoration. Our findings supported our suppositions—those who perceived HP took responsibility reported more positive image ratings after restoration than before and in contrast to those who perceived HP had low responsibility. The implications of our findings are that accepting responsibility enhances image restoration, much as it fosters forgiveness in interpersonal apologies.
Although Benoit (1995, 1997) proposed distinctions between the various strategies for image restoration, there may be overlap between the strategies with regard to acceptance of responsibility. Specifically, the corrective action strategy is very accommodative (Coombs, 1998) and incorporates aspects of the mortification strategy, not examined in this study. Holtgraves’s (1989) full-blown apology is similar to Benoit’s corrective action strategy, but it also includes aspects of mortification: “It was a terrible thing to do and I’m very sorry” (p. 11). Mortification alone was not tested as an image restoration strategy, thus it is impossible to ascertain whether the observed improvement in image perceptions was due to the apology for doing wrong (i.e., mortification), or if it is the company’s promise to correct the problem and prevent its reoccurrence (the core of corrective action). Future research should test more pure image restoration strategies (i.e., mortification without corrective action and corrective action without mortification) to remove this confound. Such a delineation of strategies will clarify the theoretical distinctions between image restoration strategies and provide evidence for which is more effective in restoring image.
Although college students are frequently used in recruitment research (Barber, 1998), the use of students limits the groups to which our results generalize. Using internships as the target job rather than a full-time position made the study more relevant to the younger undergraduate population who consisted mainly of first- and second-year students. However, freshman and sophomore college students might react differently to recruitment media than graduating seniors or those already employed. Researchers should measure image perceptions of actual job seekers in future studies.
The repeated measures experimental methodology was a strength of the study because participants acted as their own control group and the fully crossed factorial design provided the examination of all combinations of stimuli at Time 1 and Time 2. In practice, job seekers rarely view a single source of information about a potential employer at any one time; thus presenting multiple sources of information over time approximated actual recruitment contexts.
In addition, the type of stimuli used in the study is another strength of the study. Anecdotally, questions asked and comments made by participants during and after the study period suggested they believed and were invested in the cover story. This suggests the stimuli likely achieved experimental realism (i.e., participants take the study seriously; Aronson, Wilson, & Brewer, 1998). Additionally, the study may have been high in mundane realism (i.e., study mimics real world; Aronson et al., 1998) as screenshots of news websites, the sources of negative, neutral, and image restoration information all accurately depicted the types of web-based sources that job seekers use and encounter in industry.
Last, there were limitations in the application of the psychology of interpersonal apologies to the recruitment context. Research into interpersonal apologies examines forgiveness in an existing relationship (e.g., friends, family, coworkers, or romantic partners). Recruitment represents the beginning of a relationship between the potential employee and the company. The image restoration process may be more complex for stakeholders (e.g., applicants, current employees, investors, customers) who have established relationships with the company experiencing a crisis. Researchers should examine image restoration for applicants who are in the recruitment and/or selection process already, either at the maintaining applicant status or influencing job choice phases of recruitment (Barber, 1998). Rather than focusing on outcomes like attraction and intentions to apply that are more relevant in the generating applicants phase, such future work should include outcomes, such as retention in the recruitment/selection system and eventual job choice, as these may be relevant outcomes to applicants further along in the recruitment process.
We advance the literature on organizational image in the recruitment context by applying image restoration theory and research to interpersonal apologies. The integration of theory and research from different disciplines helps clarify the process of image restoration in recruitment. Through the application of image restoration theory, our study is one of the first to examine the best strategy for image restoration after a crisis. Moreover, we examined the underlying psychological mechanism as to why image perceptions change by focusing on the influence of perceived responsibility for a company’s actions on the process of image restoration. Thus, the theoretical contributions of this study lie in the integration and expansion of several theories across disciplines, providing a rich explanation and test of how recruitment occurs post crisis.
Our findings have implications for practice in recruitment by providing empirical support for the chain of recruitment outcomes after image restoration. The implications of our findings are that companies need to respond with image restoration when facing a crisis. The study shows how image restoration influences not only image perceptions after a crisis but also job seekers’ attitudes and behavioral intentions.
Conclusion
Organizational image serves as an early view of how a company is as an employer. As shown in this study, exposure to negative information results in negative image perceptions, and thus lowers attraction and intentions to pursue employment. Companies should expose job seekers to positive information about the company, especially in the wake of a crisis. Moreover, what is said within the communications does not matter as much as the valence of what is said. Exposing potential applicants to image restoration can ameliorate the deleterious effects of a crisis on job seeker image perceptions, thus increasing the likelihood that job seekers will apply to the company.
Footnotes
Authors’ Note
Zachary Steiner is now with Wipro Digital. This article is based on his dissertation, which was conducted under the supervision of Z. Byrne.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
