Social media and recruitment: examining (counter) productive diversity messages

Purpose – This study aims to explore the consequences of inconsistent diversity-related signals for job seekers. Information sources include strategically crafted corporate signals and independent sources. The authors seek to understand the effect of inconsistent diversity signals on job seekers attitudes and behavior during recruitment. Design/methodology/approach – An experiment was conducted wherein two samples from jobseeking populations were first exposed to a fictitious corporate website and then to LinkedIn profiles of that organization’s employees, with systematically varied diversity signals. Findings – Results demonstrated that conflicting diversity signals had negative effects on perceived organizational attractiveness in the student sample (N = 427) and on organizational agreeableness in the working sample (N = 243). Negative organizational attraction was related to a lower likelihood of participants applying. Practical implications – This work provides a stark but an important message to practitioners: signaling diversity-related values on corporate websites may backfire for organizations that actually lack diversity. Originality/value – Few studies have combined communication theories with recruitment to examine the link between diversity signals and inconsistent information gathered via social media.

Organizational diversity in personnel is important for many reasons including improved innovation and performance (Thomas, 2004), being better able to meet the needs of diverse clientele (Scroggins, Benson, Bendick, Egan, & Lanier, 2010), maintaining compliance with US equal opportunity employment laws (McKay & Avery, 2005) and is related to firm competitive advantage (Richard, 2000). Moreover, high-achieving individuals are more attracted to diverse organizations (Ng & Burke, 2005). Unsurprisingly, organizations use diversity messages (signals) as a key component of their recruitment strategies (Walker, Feild, Bernerth, & Becton, 2012). Most US organizations use their websites to present carefully constructed images of themselves (Young & Foot, 2005), many intentionally signaling employee diversity as a central value (Avery, 2003).
Although this diversity signaling strategy may have been effective during the inception of the internet, the emergence of social networking sites (SNS) fundamentally changed the landscape of information sharing and likely interfered with organizationally controlled signals. Prior to SNS, organizations had a near monopoly over the provision of information about themselves. Today's job seekers leverage both the organizational websites and SNS such as LinkedIn to gather information about organizations (Brouer, Stefanone, Badawy, Egnoto, & Seitz, 2015). According to a study conducted by Glassdoor in 2018, 79% of applicants used SNS for job search. Job seekers rank SNS as more useful than traditional methods, and 73% of millennials found their last position via SNS (Kunsman, 2020). SNS afford job seekers more and varied sources of information giving them access to intelligence not directly controlled by the organization.
As the number of information sources increases, so does the probability that information will be inconsistent (Windscheid et al., 2016), making inconsistencies between espoused and enacted values readily visible. For example, Google's career website clearly emphasizes diversity as a part of their espoused values and includes images prominently portraying diverse employees, yet their actual diversity and diversity initiatives are lacking (Catalan, 2014).
Little is actually known about the unintended costs organizations face when these inconsistencies are exposed. Because so many companies signal diversity as a core value, but actually lack diversity, we chose diversity signals as our manipulation. Further, information on actual diversity can easily be gleaned from SNS from user photos (providing demographics) and organizational affiliations. The purpose of this paper is to explore the consequences of inconsistent diversity signals between intentionally crafted corporate signals and diversity signals available via SNS. Specifically, we seek to understand the effect these inconsistent signals have on job seekers attitudes and behavior during the recruitment process. This work provides a stark but an important message to practitioners: signaling diversity-related values on corporate websites may backfire for organizations that actually lack diversity.

Theoretical background
The warranting principle (Walther & Parks, 2002) suggests that when drawing conclusions about a target (person, organization, etc.), other-provided information is more influential than information generated by the target itself because other-provided information is less susceptible to manipulation by the target. Information about a target, generated independently of that target, is called a warrant. Individuals evaluate warrants and assign values based on how likely the information may have been manipulated by the entity. Indeed, warranting is becoming more important, given the increasing variability of information sources, particularly the rise in user-generated content via SNS. OMJ Because job seekers typically look online for information about prospective organizations (Childress, 2018), organizations invest significant resources in promoting strategic images as vehicles for recruitment (Rynes, Bretz, & Gerhart, 2006;Turban, Campion, & Eyring, 1995). Organizations often make public claims about diversity-related values on their corporate websites (Brouer et al., 2015). Prior to SNS, organizations benefited from information asymmetries (when the organization has more and better information than the job seeker) because job seekers had to rely on corporate-controlled sources (websites and job advertisements), for information about values such as diversity.
However, SNS provide venues for other-generated information about organizations. Because this alternative, and often competing, information is not generated by the entity itself, it is perceived to hold more value and thus is a highly sought after source of data. In fact, SNS are one of the most widely used tools for acquiring additional employment-related information (Stopfer & Gosling, 2013). In other words, SNS provide warrants about organizational diversity, and the low likelihood that these warrants were manipulated by the organizations increases their perceived value to job seekers. Although the research shows third-party websites are perceived as more credible than official organization websites (Van Hoye & Lievens, 2007), the effects of (in)consistent and unintentional signals about target organizations between different sources on job seekers attitudes and behavior remains unknown.

Warranting and job seekers evaluative perceptions
We argue that when organizations make diversity claims, but such claims are not supported by warrants from SNS, this inconsistency should negatively affect job seekers' appraisals of the organization. We focus specifically on warrants from LinkedIn profiles because LinkedIn is one of the primary sources for gathering information about employers (Jobvite, 2013). We explore the impact of these warrants on job seekers' perceptions of the organization's agreeableness, attraction to the organization and applying. In short, we argue that inconsistent information will negatively impact perceptions of agreeableness, attraction and applying.
One approach to investigating an organization's image is through perceptions of corporate character, which help explain the emotional reactions and attachments people have toward organizations (Davies, Chun, Da Silva, & Roper, 2004). We focus specifically on the personality characteristic of agreeableness because it has been found to be the strongest predictor of employee affinity toward an organization and because organizations perceived as agreeable are viewed as trustworthy, likable and friendly. It is measured using warmth (open, straightforward), empathy (concerned, supportive) and integrity (honest, sincere; Davies, 2008;Davies et al., 2004).
Because agreeableness is centered on being open, straightforward, empathetic and honest, inconsistencies between signals from the organization and signals from LinkedIn profiles should lead job seekers to perceive the organization as less agreeable. Based on the warranting principle, other-generated information is considered less susceptible to manipulation and thus more valuable. If information regarding an organization varies across self-generated and LinkedIn sources, job seekers will likely conclude that the messages from the organization are fabricated, leading them to see the organization as having less integrity, warmth and empathy (low agreeableness).
These inconsistencies should impact organizational attraction, which is defined as potential applicants' interest in seeking a job at an organization, and attraction is "the most immediate objective of the early stages of recruitment" (Allen, Mahto, & Otondo, 2007, p. 1700. Early studies have demonstrated that corporate-generated recruitment Social media and recruitment advertisements can positively impact organizational attractiveness (Turban, 2001), but this work was conducted before the significant growth of SNS. Moreover, the research supports that organizational attractiveness suffers when recruitment claims are refuted by other sources such as employee reviews (Windscheid et al., 2016) and word of mouth (Van Hoye & Lievens, 2007).
Inconsistencies between SNS-generated information and corporate websites are likely to negatively impact seekers' evaluations. The instrumental-symbolic framework (Van Hoye & Saks, 2011) coupled with social identity theory (Ashforth & Mael, 1989) suggests that individuals use their associations with organizations to generate information about themselves. Thus, individuals generally want to be associated with organizations they view positively and that have positive traits (Lievens & Highhouse, 2003;Turban et al., 1995), thus seeker will not be attracted to organizations they perceive negatively. Lastly, individuals should be more inclined to take the behavioral initiative to apply for jobs at organizations that they perceive to be more agreeable and attractive. The following hypotheses are therefore proposed: H1. Inconsistent diversity signals from corporate websites and LinkedIn profiles will be related to negative perceptions of agreeableness, organizational attraction and applying behavior.
H2. Agreeableness and organizational attraction will be positively related to applying behavior.

Procedure
After individuals consented to participate, they were randomly assigned to conditions and told they were completing an online survey on behalf of a real organization renamed for anonymity. We used a 2 (organization diversity signal) by 3 (linked profile diversity) experimental design. In all conditions, the participants were directed to a corporate webpage that included either generic information about the organization, REMCO (control condition), or emphasized diversity by including images of employees from a range of diverse ethnicities, genders and ages, and text indicating a commitment to diversity (experimental condition). Mimicking a real function of LinkedIn, all participants were exposed to one of three sets of LinkedIn images presenting lists and photos of what were portrayed as actual employees of REMCO. This was presented as a way to get to know their potential coworkers and managers/leaders. Because we sought to emphasize diversity signals at different hierarchy levels, participants in each condition were shown two LinkedIn results' pages: a list of lowerlevel employees and a list of the top-management team. Avery's (2003) framework was used to manipulate employee LinkedIn profile portraits by race across the conditions, including uniform (all White across both levels), skewed (diverse at lower levels; all White top management) and balanced (diverse at both levels). The order of lower-and upper-level profile exposure was randomly presented to the participants.
After exposure to the LinkedIn stimuli, the participants recorded their perceptions of organizational agreeableness and attractiveness with REMCO in mind. Still, with the idea that the company is a real organization, the participants were given the opportunity to "apply" by clicking a link that would redirect them to a job site. Upon completing the experiment, the participants were fully debriefed about the nature of the experiment. OMJ Participants. We tested the hypotheses by sampling a population of college students (Sample 1) and employed individuals (Sample 2). Although most recruitment research has focused on student populations, students represent only part of the job-seeking population, ignoring the "25-50% of all job seekers who look for other work while still employed" (Schwab, Rynes, & Aldag, 1987, p. 156). This significantly reduces the generalizability of recruitment research findings (Young, Rinehart, & Heneman, 1993). Moreover, in the limited research available, employed job seekers have displayed differences in attraction than college students and further exploration is deemed critical (Breaugh & Starke, 2000). We seek to address this gap by using the same methodology to address these research questions in samples from both the populations.
For Sample 1, junior and senior business students enrolled in university career placement services were invited to participate. Of the 959 invited, 543 individuals responded (56.6% response rate) and 427 completed the survey. The mean age was 22.78 years (SD = 4.71), with 52.8% of the sample identifying as female. The sample was 71.4% White, 9.2% Hispanic/Latino, 8.7% Asian, 5.4% Black, 1.9% Middle Eastern, 2.1% mixed race and 1.4% other.
For Sample 2, the participants were drawn from MTurk. Participation eligibility was restricted to adults currently employed and living in the USA who were either currently looking for another job or planned to in the near future. In total, the 243 participants (40% female) averaged 36.52 years (SD = 8.91) of age. The sample was 74.8% White, 9.1% Black, 8.3% Asian, 4.5% Hispanic, 0.4% Middle Eastern, 2.5% mixed race and 0.4% other. The participants worked an average of 40h per week (SD = 7.84) and average job tenure was 4.08 years (SD = 3.26).
Validation of stimulus materials. In a pilot study, a US sample of 155 individuals (M age = 36.51, SD = 10.88) was drawn from MTurk. Analysis of variance (ANOVA) tests demonstrated that the participants were sensitive to the diversity manipulations (F [2,153] = 52.77, p = 0.00). Bonferroni comparisons revealed significant mean differences between all conditions [balanced condition (M = 4.15, SD = 0.72); skewed condition (M = 3.50, SD = 1.11); uniform condition (M = 2.73, SD = 1.33)] on the following item using a 1-5 Likert-type scale: "I believe Remco is a diverse company."

Measures
The participants were asked to rate organizational agreeableness and attraction after they saw the corporate website and LinkedIn manipulations. They were asked to respond to these questions about the REMCO. All measures were assessed on a five-point Likert scale (1 = strongly disagree, 5 = strongly agree). Agreeableness was measured with the Corporate Character Scale developed by Davies et al. (2004; Sample 1: a = 0.89; Sample 2: a = 0.94), which asked the respondents to personify the organization to uncover their emotional attachments. Thus, they were asked to rate whether they thought that the organization was friendly, straightforward, honest and trustworthy (sample items). Turban and Greening's (1997) five-item scale (Sample 1: a = 0.87; Sample 2: a = 0.90) was used to measure organizational attraction ("I would exert a great deal of effort to work for Remco,"). The behavioral outcome of applying to a job was measured with a binary response to the following question: If you would like to apply for a position at Remco, click 'yes' and you will be redirected to the career website at the end of this survey. If you would prefer not to apply, click 'no.' We controlled for race and gender based on the research, suggesting minorities might be more sensitive to diversity signaling messages (Mor Barak, Cherin, & Berkman, 1998), by entering them into our regression analyses.

Hypotheses testing
Statistical package for the social sciences was used for all analyses. Table 1 reports descriptive statistics. ANOVA results, testing H1a and H1b are reported in Table 2. Results from Sample 1 indicated no mean differences in agreeableness across LinkedIn conditions in both the consistent [F (2,208) = 1.02, ns] and the control and inconsistent conditions [F (2,195) = 2.16, ns]. Thus, H1a was not supported in the student population in that they did not see inconsistent organizations as less agreeable. Sample 2 demonstrated significant mean differences in agreeableness scores between conditions when participants were shown the consistent message [F (2,115) = 6.69, p = 0.00] but not in the control conditions (no diversity signal present) [F (2,119) = 0.53, ns]. Agreeableness was highest in the balanced condition Notes: a Race coded 0 = Non-minorities, 1 = Minorities; b 0 = male, 1 = female; c apply redirect coded 0 = did not click to be redirected to application page, 1 = Clicked to be redirected to application page; alpha reliabilities reported in parenthesis; *p < 0.05; **p < 0.01 Notes: ***indicates means that are significantly different from other means in the column quadrant; * p < 0.05; ** p < 0.01 OMJ (consistent) when the participants viewed the diversity signal. Thus, H1a was supported in the working population sample. That is, the working sample saw organizations' consistent and diverse messages as more agreeable.
In Sample 1, mean differences in organizational attraction were observed between LinkedIn conditions when participants viewed the diversity signal [F (2,216) = 3.35, p = 0.03] but not in the control condition [F (2,206) = 0.81, ns]. For participants in the diversity signal conditions, organizational attractiveness scores were lowest in the uniform/ inconsistent condition (M = 3.14, SD = 0.80), and this mean was significantly lower than organizational attractiveness scores in the skewed/inconsistent (M = 3.34, SD = 0.83) and balanced/consistent (M = 3.47, SD = 0.74) conditions. Participants in the skewed and balanced conditions did not significantly differ in their ratings of organizational attraction. Together, these results supported H1b in Sample 1 that inconsistent organizations were less attractive. However, no significant mean differences in organizational attraction were observed between conditions regardless if participants were shown the diversity signal [F (2,118) = 0.56, ns] or not [F (2,119) = 1.64, ns], not supporting H1b for the working population meaning that inconsistency wasn't related to organizational attraction. Table 3 reports cross-tabulation results testing H1c. Participants in both the samples did not vary their applying behavior based on experimental conditions, finding no support for H1c. Table 4 reports regression results. Although not hypothesized, agreeableness was positively related to organizational attractiveness (Sample 1: B = 0.73, p = 0.00; Sample 2: B = 0.95, p = 0.00), and organizational attractiveness was positively related to the participants applying behavior (Sample 1: B = 1.60, p = 0.00; Sample 2: B = 1.42, p = 0.00), supporting H2b. However, not supporting H2a, agreeableness was not related to applying (Sample 1: B = À0.28, ns; Sample 2: B = À0.07, ns).

Discussion
Although different patterns emerged in each sample, our experimental results suggest that inconsistent diversity signals have negative consequences for organizations. Students perceived organizations with inconsistent diversity signals to be less attractive, and job seekers already in the workforce perceived the organization as less agreeable. However, only negative organizational attraction led to a reduced likelihood of participants applying. Moreover, our results show diversity signals throughout the organizational hierarchy matters. The participants were sensitive to the lack of diversity across upper management in the signals shown. An interesting and unintended pattern that emerged from our results was the different perceptions developed between student and employed job seekers. Student job seekers rated inconsistent diversity signals as less attractive yet not less agreeable. The used sample evaluated inconsistent organizations as less agreeable but not less attractive. One possible explanation is that participants in both the samples were sensitive to different outcomes because of their current stage in the search process. Individuals with experience in professional roles might understand the importance of corporate character and how that might impact the employment relationship and experience. Indeed, individuals with job experience "will be less influenced by recruiters and recruiting practices because they will more likely be aware of, and thus focus on, job and organizational attributes" (Rynes, Heneman, & Schwab, 1980, p. 538).
Our study contributes to the literature in multiple ways. First, a few studies have examined the impact of SNS on recruitment (McKay & Avery, 2005) and even fewer have explored the effects of information about organizations from different sources. Our research begins to answer questions about inconsistent information available from nonorganizationally controlled sources by assessing the impact of (in)consistent signals originating from organizations and publicly available signals about those organizations. Second, we address how employee diversity signals across the organizational hierarchy influence job seekers' perceptions and behavior. This is critically important considering the negative relationship between diversity and hierarchical status in organizations today. We found that although organizations may be diverse at lower levels of the hierarchy, homogenous upper-management teams still drive negative consequences regarding generalized perceptions of organizational attractiveness. We also explored a behavioral outcomeinitiating the application processwhich is not typically used in the literature.

Strengths, limitations and directions for future research
A noteworthy strength of our preliminary research is that we used a multi-sample experimental design that allows us to limit the effects of environmental factors and demonstrate and replicate time-order effects, and our samples enhance the generalizability. Furthermore, we employ a behavioral outcome, demonstrating the importance of job seeker attitudes' on job applying behaviors. Notes: Sample 1, N = 401; Sample 2, N = 243; unstandardized regression coefficients reported; a race coded 0 = non-minorities, 1 = minorities; b 0 = male, 1 = female; c logistic regression summary on binary dependent variable; d 1 = clicked on button to be redirected to apply, 0 = did not click on button to be redirected to apply; e CI = confidence interval; * p < 0.05; ** p < 0.01 OMJ This study should be viewed in light of some limitations. First, we assume that all individuals pay attention to signals from other sources, such as SNS. Moreover, LinkedIn is one of the several SNS for job-relevant information (Brouer et al., 2015). Although MTurk samples have been shown to be high-quality sources of data (Rand, 2012), we sought to overcome some of the limitations of MTruk by using qualification questions for the participants, keeping our task short to minimize inattentiveness and by paying a reasonable rate. Additionally, we used a second sample using a different population. Although we controlled for race and gender, future research also could examine the different reactions of these populations.