Im digitalen Zeitalter sind Social-Media-Plattformen wie LinkedIn integraler Bestandteil des Rekrutierungsprozesses geworden und bieten eine Fülle von Informationen über Bewerber. Vollständigkeit und Struktur dieser Informationen variieren jedoch stark, was zu potenziellen Verzerrungen und Missverständnissen führt. Dieser Artikel untersucht, wie unvollständige LinkedIn-Profile die Einstellungsraten beeinflussen und enthüllt, dass Anwerber häufig Kandidaten mit weniger detaillierten Profilen aufgrund von empfundener Mehrdeutigkeit und Misstrauen bestrafen. Die Studie beleuchtet die psychologischen Mechanismen hinter diesen Urteilen, einschließlich der Wahrnehmung von Vertrauenswürdigkeit, Professionalität, Wärme und Kompetenz. Es untersucht auch die mäßigende Rolle anderer Anwendungsmaterialien wie Lebensläufe bei der Abmilderung der negativen Auswirkungen unvollständiger LinkedIn-Informationen. Durch die Untersuchung dieser Dynamik vermittelt der Artikel ein differenziertes Verständnis, wie Social-Media-Bewertungen Einstellungsentscheidungen beeinflussen und bietet praktische Einsichten sowohl für Personalvermittler als auch für Arbeitssuchende. Die Ergebnisse unterstreichen die Bedeutung umfassender und professioneller Online-Profile sowie die Notwendigkeit strukturierter und fairer Bewertungspraktiken bei der Rekrutierung.
KI-Generiert
Diese Zusammenfassung des Fachinhalts wurde mit Hilfe von KI generiert.
Abstract
Assessing social media platforms like LinkedIn has become popular in personnel selection but remains controversial due to varying relevance and availability of job-related content. To better understand the implications of this low information uniformity, we introduce the incomplete LinkedIn information paradigm: How do recruiters react to incomplete LinkedIn profiles? In Study 1 (N = 460), we found a significant decrease in hireability ratings when job seekers provided just basic LinkedIn information and no details, even outweighing the effect of their qualifications. Perceptions of professionalism and trustworthiness, reflecting overall warmth and competence, served as mediators. An interview study (Study 2, N = 32) confirmed that incomplete LinkedIn profiles diminish hireability ratings and increase suspicion. In studies 1 and 2, raters were presented with no application materials beyond the LinkedIn profile, framing LinkedIn assessments as the active sourcing of passive candidates. Thus, to investigate the effect of incomplete LinkedIn information in a more traditional context, we examined this incomplete LinkedIn information paradigm again in Study 3 (N = 363), where we framed LinkedIn screenings as background checks beyond the applicant’s complete résumé. Here, incomplete LinkedIn information did not negatively affect hireability, suggesting that providing a complete résumé can offset the negative impressions from incomplete LinkedIn profiles. In summary, we demonstrate that incomplete information on professional platforms has new relevance in digital assessments, but its impact varies by context: Incomplete LinkedIn profiles can harm a passive candidate’s prospects, whereas they do not seem to impact background checks in the same way.
All data, analysis syntax, a reference to R packages, and materials are available in the electronic supplementary, https://osf.io/utgp6/. This study was preregistered; see https://osf.io/x3p2a, https://aspredicted.org/fgjr-pz5z.pdf, and https://osf.io/vkdra. Portions of this manuscript were presented at the 21 st Congress of the European Association of Work and Organizational Psychology, Katowice, Poland; the 13 th Congress of the German Psychological Society, Section for Work, Organizational and Business Psychology, Kassel, Germany; and the 39 th Society for Industrial and Organizational Psychology Annual Conference, Chicago, IL, United States. The authors thank Filip Lievens, Chiara-Maria Frieler, Celeste Brennecka, and Marie Therese Bartossek for their support and helpful comments.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Recent surveys estimate that two out of three recruiters screen job seekers’ social media to form impressions about their knowledge, skills, abilities, and other characteristics (KSAOs; Hartwell & Campion, 2020; Roth et al., 2019; Smith, 2017). However, relatively few studies have addressed social media assessments in selection, and the validity of this practice is reported as contradictory or low, with decision processes poorly understood (Mönke et al., 2024b; Roth et al., 2016). Accordingly, this practice remains controversial, with the lack of clear guidelines adding to the confusion surrounding its role in decision-making. For instance, Wilcox et al. (2022) cautioned that “organizations may miss out on talent, [and] become less diverse and innovative” (p. 329) if they use social media assessments. A potential reason for the low validity of social media assessments may be that social media serves as an information source with low structure and consistency (Roth et al., 2016). That is, social media profiles vary greatly in both the job-relatedness and completeness of the information they provide, much more so than traditional application materials like résumés. Social media assessments often yield incomplete or disjointed information, as many users omit entire sections of their LinkedIn or Facebook profiles due to privacy concerns, context, or preferences (Neubaum et al., 2023; Roth et al., 2016; Zhang et al., 2020). We examine how recruiters respond to such incomplete profiles.
Our primary objective is to examine how varying degrees of LinkedIn information influence hireability ratings. Many studies on incomplete application materials were conducted decades ago using analog formats (e.g., Jagacinski, 1991; Stone & Stone, 1987). However, in contrast to traditional selection procedures, the interpretations associated with incomplete information on digital platforms like LinkedIn may differ fundamentally. We argue that incomplete information is now much more common, and thus its implications have gained new relevance in digital contexts. We focus on LinkedIn because it is the most relevant platform for social media assessments (e.g., Smith, 2017).
Anzeige
Drawing from prior research in the marketing literature (e.g., Johnson & Levin, 1985), decision-making (e.g., Ebenbach & Moore, 2000; Jaccard & Wood, 1988), and (analog) personnel selection (e.g., Jagacinski, 1991), we argue that job seekers who provide incomplete LinkedIn information will receive less favorable ratings than other applicants. Our rationale is based on (1) the salience of more complete LinkedIn profiles, which accentuates the absence of information in less complete profiles; (2) the increased ambiguity arising when incomplete profiles undermine the utility of LinkedIn assessments as a background check tool; and (3) the predominantly positive framing of LinkedIn content that may elicit suspicion among recruiters regarding omitted profile details. This approach—which we will refer to as the incomplete LinkedIn information paradigm—offers a framework that extends the traditional incomplete information paradigm (e.g., Highhouse & Hause, 1995; Johnson & Levin, 1985) and provides a more nuanced understanding of social media assessments (Mönke et al., 2024b; Roth et al., 2016; Wilcox et al., 2022). That is, in Studies 1 and 2, we examine how incomplete information influences recruiter impressions, drawing on frameworks from social cognition (warmth and competence; Fiske et al., 2007) and cybervetting research (privacy intention; Hartwell & Campion, 2020; trust; Roth et al., 2016). Finally, we compare LinkedIn use cases for (a) screening passive job seekers in active sourcing (Study 1) and (b) a more traditional application review process, where LinkedIn serves as part of a background check of applicants alongside résumés (Study 3). Thereby, Study 3 examines whether the impact of incomplete LinkedIn profiles depends on the availability of other application materials.
Study Background
Cybervetting: Social Media Assessments in Personnel Selection
Social media assessments (also referred to as cybervetting) describe the screening of social media to evaluate a job seeker’s hireability, check consistency with other application materials, and screen for red flags (e.g., drug abuse, badmouthing a former employer; Berkelaar, 2014; Roth et al., 2016; Wilcox et al., 2022). Such social media assessments are most often conducted on professional platforms like LinkedIn, but hedonic platforms like Facebook are also screened (e.g., Smith, 2017). HR professionals consider social media information as a digital residue of candidates’ KSAOs: Assessing social media might offer a first impression of the candidate’s hireability (McFarland & Ployhart, 2015; Mönke et al., 2024b). Thus, social media assessments grew to be more popular than cognitive ability tests and personality questionnaires (Farndale et al., 2023; Roth et al., 2019). They are established as background checks, and they are also used to screen passive job seekers (active sourcing; Nikolaou, 2014).
Alongside this popularity, social media assessments are also controversial. Frantz et al. (2016) labeled cybervetting the “Wild West” (p. 332) of personnel selection. This is because validity evidence is scarce: Convergence with self-reported personality traits is substantial, but criterion-related validity is low (Mönke et al., 2024b). Another major concern is that social media assessments target the applicant’s personal life. For instance, political affiliation, religion, and hobbies are prominent on social media (Zhang et al., 2020). This increases the risk of adverse impact, e.g., similarity attraction (Mönke et al., 2024a; Roth et al., 2020).
Relatedly, the quantity and job-relatedness of social media information differ widely (Roth et al., 2016). For many professionals, social media assessments only yield disjointed or incomplete information. With the term incomplete social media information, we refer to the absence or omission of expected content on a social media platform’s profile that is typically used to form impressions in social media assessments. On LinkedIn, such low information quantity can range from minor gaps—such as lack of detail in skill listings or a lack of posts and interactions that could otherwise contribute to a comprehensive understanding of the job seeker’s KSAOs, qualifications, or intentions—to major omissions like entire missing sections (e.g., work history, education, or profile picture; see also Fernandez et al., 2021; Hartwell et al., 2024; Hartwell & Campion, 2020; Roulin & Bangerter, 2013). For example, Aguado et al. (2019) reported that about one-third of surveyed LinkedIn users did not include a profile picture. Relatedly, Zhang et al. (2020) reported that only 56.8% of surveyed Facebook profiles included information on the job seeker’s education, and only 41% included information on prior work experience. Further, while Facebook and LinkedIn are the two most used sites for social media assessments, about one-third of US workers (18–64 year olds) do not use Facebook, and about two-thirds do not use LinkedIn at all (Auxier & Anderson, 2021; Smith, 2017). So, it is very likely that HR professionals encounter incomplete social media information in cybervetting.
Anzeige
What Happens When Job Seekers’ Information is Incomplete?
Incomplete information plays an important role in various fields, such as marketing (e.g., Johnson & Levin, 1985), decision-making (e.g., Ebenbach & Moore, 2000; Jaccard & Wood, 1988; Vuolevi & Van Lange, 2012), and (traditional) personnel selection (Garcia-Retamero & Rieskamp, 2009; Highhouse & Hause, 1995; Jagacinski, 1991). When facing incomplete information, raters could react with (1) devaluation of this person/object, (2) average imputation of the missing information, (3) inferential imputation (i.e., inferring the value of the missing information from the values provided on other dimensions), or (4) disregard, i.e., ignoring that information is missing (Jaccard & Wood, 1988). Some authors have argued for the devaluation-hypothesis (e.g., Ebenbach & Moore, 2000), others for a hybrid model of imputation and devaluation (Jaccard & Wood, 1988); but there is a strong consensus that decision-makers prefer options with complete information over incomplete information (see Highhouse & Hause, 1995). Johnson (1987) referred to this as an incomplete information bias.
Decision-makers dislike incomplete information because it introduces ambiguity into their decisions, and such ambiguity is disliked (Frisch & Baron, 1988; Highhouse & Hause, 1995). Ambiguity refers to the “subjective experience of missing information relevant to a prediction” (Frisch & Baron, 1988; p. 152). Thus, ambiguity in personnel selection is heightened when application materials are incomplete. To minimize ambiguity in their decisions, recruiters assign lower ratings to candidates with incomplete information, effectively screening them out (Highhouse & Hause, 1995; Jagacinski, 1991). For example, applicants who omitted a certain test score (Jagacinski, 1991) or did not answer a question regarding former convictions (Stone & Stone, 1987) were penalized. As Jagacinski (1991) pointed out, recruiters “did not want to take a chance on the person having poor skills on the missing dimension” (p. 27).
The New Relevance of Incomplete Information in Social Media and LinkedIn Assessments
Is the impact of incomplete information still relevant in the era of social media assessments and LinkedIn recruiting? That is the key question for this study. Several authors argued that the effect of incomplete information depends on the rating context (e.g., Garcia-Retamero & Rieskamp, 2009; Highhouse & Hause, 1995). We argue that LinkedIn assessments are a context in which incomplete information may have a particularly negative impact.
First, screening and actively sourcing job seekers via social media platforms like LinkedIn has become a new standard in HR departments (Hartwell & Campion, 2020; Nikolaou, 2014; Wilcox et al., 2022). Thus, recruiters might screen dozens of social media profiles per job opening and have a lot of opportunities to compare the digital footprint of different applicants. Some job seekers will have more detailed social media profiles (even publicly sharing personal details; Zhang et al., 2020), and others will be more reluctant (e.g., due to privacy concerns; Trepte & Masur, 2023). On LinkedIn, other candidates with more information are readily accessible with minimal effort. Being aware that other job seekers provide more complete information increases the salience of incomplete information, and this is an antecedent for experiencing ambiguity (and a negatively biased impression; Highhouse & Hause, 1995).
Beyond that, incomplete information might stem from privacy settings, oversights, or intentional omissions, leaving recruiters unsure whether the candidate is hiding something or lacks the relevant KSAOs. Since most organizations lack policies on how to evaluate (incomplete) social media information (McDonald et al., 2022; Wilcox et al., 2022), recruiters are left to interpret the data. We predict that, in such cases, raters will respond negatively: Ambiguity is heightened, and the LinkedIn assessment fails to fulfill its purpose as a background check—a key objective of cybervetting aimed at reducing uncertainty (Pike et al., 2018; Roulin & Levashina, 2019).
Second, most social media postings are framed positively. For instance, users post about their achievements and promotions on LinkedIn but less about their failures (e.g., Johnson & Leo, 2020; Roulin & Levashina, 2016). Leaving out LinkedIn information usually equals leaving out positive information about KSAOs that could enhance hireability ratings. This is relevant because the positive framing of complete information dimensions may increase raters’ expectations that they “lose” when choosing an applicant with incomplete information. LinkedIn may establish a normative expectation for a certain level of positive content, prompting negative reactions when job seekers fall short. Incomplete LinkedIn sections could signal that the applicant may lack competencies or fail to fully present their qualifications online. In line with this, with analog application materials, HR decision-makers were more likely to reject job candidates with incomplete information when other candidates demonstrated average or high levels of the missing attribute (Highhouse & Hause, 1995).
Finally, recruiters are likely to attribute incomplete social media information to the applicant’s traits (see Johnson & Levin, 1985). They may perceive missing LinkedIn details as a deliberate choice, reflecting the applicant’s attempt to conceal certain aspects or signaling low conscientiousness, rather than a random omission. This is important because prior studies often framed incomplete information as being the interviewer’s error or that the information was assessed later—attributing incomplete information to external factors and not the job seeker (Highhouse & Hause, 1995; Jagacinski, 1995). Even in these cases, candidates with incomplete information were penalized. When incomplete information is viewed as intentional and potentially indicative of the applicant’s KSAOs, its negative impact increases (Garcia-Retamero & Rieskamp, 2009). This adds to our reasoning that raters will penalize those with incomplete LinkedIn information.
In sum, we propose that recruiters will react negatively to job seekers with incomplete LinkedIn profiles. This is our basic rationale as we lay out the incomplete LinkedIn information paradigm below. We suspect that such thinking could even overshadow a good fit between the candidate and the job: Why choose a candidate with incomplete social media information when other candidates provide more complete and positive information?
Preliminary evidence for our notion was provided by Berkelaar’s (2014) interview study. They found that “employers reported being more likely to hire those who gave a little something more because it showed they have nothing to hide” (p. 494) in social media assessments. Also, LinkedIn profile length was positively related to hireability perceptions in two field studies (Roulin & Levashina, 2019; Roulin & Stronach, 2022). We aim to substantiate these findings in a robust experimental design that separates the effect of incomplete information from other applicant and profile characteristics (e.g., the applicant’s conscientiousness, number of connections; see Roulin & Levashina, 2019). Our design provides more nuanced insights into the decision-making processes involving incomplete online information. Hence, we hypothesize:
(H1)The more detailed content presented on the applicant’s LinkedIn profile, the better ratings of the applicant’s hireability.
The Special Role of Missing Pictures
Building on this, we propose that the (non-)presence of an applicant’s profile picture will amplify the negative associations with incomplete information, as visual cues such as a profile picture may have a stronger influence on raters’ judgments than verbal or textual cues. This rationale is rooted in Sundar (2008), who predicted that raters rely on web images more than textual information because images better resemble the real world, and higher realism increases credibility. So, not including a profile picture on a LinkedIn profile might be especially salient to cybervetting raters, potentially leading to an even stronger devaluation of applicants who lack this information (Domahidi et al., 2022; Roulin & Levashina, 2019). Thus:
(H2a)If the applicant’s LinkedIn profile includes a picture, hireability ratings are higher.
(H2b)The presence of the applicant’s profile picture has a stronger effect on hireability ratings than the LinkedIn information quantity (incomplete information).
Disentangling Decision Processes in the Incomplete LinkedIn Information Paradigm
While the consequences of incomplete information were extensively studied (e.g., Highhouse & Hause, 1995; Jagacinski, 1991), less is known about why incomplete information leads to an applicant’s devaluation (see Roth et al., 2016). Hence, we propose several key perceptions that might explain a recruiter’s decision when they face incomplete LinkedIn profiles. For an overview of this incomplete LinkedIn information paradigm, see Fig. 1.
Fig. 1
Theoretical model: the incomplete LinkedIn information paradigm. Note. Higher LinkedIn information quantity = lower incompleteness of LinkedIn information, see Table 1. Figure 1 also illustrates our stepwise approach to modeling recruiters’ decision-making process: That is, specific impressions of trust and professionalism shape overall perceptions of broad warmth and competence (for a similar rationale see Roth et al., 2017, 2020). An additional analysis (see ES09 and footnote 4) revealed that when warmth, competence, professionalism, and trustworthiness were simultaneously treated as mediators, the effects of professionalism and trust were masked by the broader dimensions of warmth and competence
First, we focus on trustworthiness. Social media assessments are often conducted to evaluate trustworthiness (e.g., as a background check; Berkelaar, 2014), i.e., a perception of the applicant’s benevolence, integrity, and ability (Klotz et al., 2013). When information on a social media profile is incomplete, recruiters may wonder whether the omission is an oversight or if the job seeker is intentionally concealing something. The perceived trustworthiness of the applicant might mediate this process by shaping the recruiter’s interpretation of incomplete LinkedIn information and reacting to the increased ambiguity about the job seeker. Specifically, incomplete information may undermine the perceived trustworthiness of the applicant, which in turn diminishes an applicant’s perceived hireability (see Jagacinski, 1991; Roth et al., 2016; Vuolevi & Van Lange, 2012). This notion of trustworthiness as a key mediator in LinkedIn assessments is supported by recent findings: A manager’s suspicion in social media assessments—a state characterized by uncertainty, heightened cognitive activity, and perceptions of malintent—was shown to reduce managers’ expectations of applicant performance (Pu et al., 2023; Roth et al., 2024). Suspicion is a precursor to distrust (Bobko et al., 2014); thus, we argue that trustworthiness is not merely a contextual factor but a crucial mediating mechanism that explains why incomplete LinkedIn information impacts recruiters’ decisions.
(H3)Perceived trustworthiness mediates the relation between the applicant’s LinkedIn information quantity (incomplete information; profile picture) and hireability ratings.
Next, Hartwell and Campion (2020) reported that impressions of online professionalism are among the most sought-after attributes in cybervetting. Professionalism refers to maturity and character and might be seen as an indicator of the applicant’s competence (Hartwell & Campion, 2020; Roberts, 2005). For instance, the appropriateness of pictures, postings, and interaction with other LinkedIn users may shape impressions of professional behavior. Relatedly, recruiters may view incomplete information on professional platforms such as LinkedIn as inappropriate. Since social media assessments are becoming a standard procedure in selection, recruiters likely expect job seekers to provide a professional online self-presentation, and this could include a certain level of LinkedIn details to complement their professional identity. If the LinkedIn profile is incomplete, recruiters may perceive it as unprofessional, leading to disappointment and reduced hireability perceptions. Hence, such perceptions might be another key mediator in LinkedIn assessments, especially when information is incomplete. We hypothesize:
(H4)Perceived professionalism mediates the relation between the applicant’s LinkedIn information quantity (incomplete information; profile picture) and hireability ratings.
Cybervetting is often used to inform initial impressions about a job seeker (e.g., Berkelaar & Buzzanell, 2015). Thus, next, we draw from research in the domain of first impressions. There is a large consensus that impressions about people or groups can be attributed to two underlying dimensions: warmth (also referred to as communion) and competence (also referred to as agency; Abele & Wojciszke, 2007; Fiske et al., 2007). That is, warmth refers to liking a person and believing in their good intentions, and competence refers to respecting a person due to their capability to pursue their intentions (Fiske et al., 2007).
Previous research has demonstrated that these dimensions of social cognition play a key role in hiring decisions. For example, women may face discrimination in vying for managerial positions, as they are often perceived as warm but not competent (Cuddy et al., 2011). Broad perceptions of warmth and competence may have a particular impact when applicants belong to groups associated with stereotypes, such as sexuality or political affiliation (Roulin et al., 2023). This personal information, often readily available on social media profiles, may encourage recruiters to categorize applicants (Zhang et al., 2020). Perceptions of warmth and competence may be crucial for understanding the incomplete LinkedIn information paradigm: If incomplete social media information is seen as suspicious or indicative of low trustworthiness, raters may question the applicant’s intentions (i.e., warmth/communion) and qualifications (i.e., competence/agency). Hence, we predict:
(H5)Perceived warmth mediates the relation between the applicant’s LinkedIn information quantity (incomplete information; profile picture) and hireability ratings.
(H6)Perceived competence mediates the effect between the applicant’s LinkedIn information quantity (incomplete information; profile picture) and hireability ratings.
Finally, incomplete social media information can be attributed to either (a) missing target information or (b) the candidate restricting public access to it. As Hartwell and Campion (2020) found, recruiters considered privacy settings in cybervetting. That is, if social media information was incomplete due to strict privacy settings, recruiters reported more favorable perceptions of the applicant. Attributing incomplete LinkedIn sections to the candidate’s intention may attenuate—or even suppress—the impact of incomplete information: Applicants with stringent privacy settings and lower LinkedIn information may be perceived as more technically savvy, thus avoiding devaluation (Roth et al., 2016). Thus, we hypothesize:
(H7)The perceived intention that an applicant is protecting their privacy moderates the effect of incomplete LinkedIn information on hireability criteria: The higher the perceived privacy intention, the smaller the effect of LinkedIn information quantity on the hireability ratings.
Figure 1 outlines our hypotheses and summarizes the theoretical model of our study, referred to as the incomplete LinkedIn information paradigm. We posit interconnections among perceived trustworthiness, professionalism, warmth, and competence, with trustworthiness representing an aspect of warmth and professionalism reflecting a facet of competence. These relationships are further elaborated in the results section (Study 1, see also footnote 4).
Overview of Objectives and Studies
In Study 1, we examine how raters react to incomplete information in LinkedIn assessments. This study tested if the incomplete information paradigm’s rationale transfers to contexts of digital selection (active sourcing, passive job seekers), with LinkedIn being the most prominent example of cybervetting (Smith, 2017). This is relevant because incomplete information is common in digital assessments, and such screenings differ fundamentally from more traditional selection procedures. At the same time, by testing key perceptions between incomplete information (i.e., low information quantity) and hireability perceptions, we deepen our understanding of the incomplete information paradigm. For an overview, see Fig. 1. Then, in Study 2, we delve into the decision processes of HR professionals confronted with incomplete LinkedIn profiles through interviews. Finally, in Study 3, we revisited the incomplete LinkedIn information paradigm in a more traditional context, where LinkedIn is screened alongside the applicant’s résumé as part of a background check.1 This tested whether raters would still perceive incomplete LinkedIn profiles as ambiguous when a complete résumé is also available.
The Ethics Committee of the Faculty of Psychology and Sports Science at the University of Münster approved our procedure (Decision 2022-41-FM). We conducted our analyses in R (Version 4.4.2) and RStudio (Version 2024.12.1). All data, analysis syntax, a reference to R packages, and materials are available in the electronic supplementary, https://osf.io/utgp6/. We preregistered our procedure2; see https://osf.io/x3p2a, https://aspredicted.org/fgjr-pz5z.pdf, and https://osf.io/vkdra.
Study 1
Method and Materials
Sample
As a power analysis, we ran 1000 Monte Carlo simulations (Wang & Rhemtulla, 2021) with N = 360 for a basic model (i.e., no interaction effects, residual covariances), assuming Cronbach’s α = 0.80 and a small to medium correlation between factors (r = 0.20). The power to detect the regression paths was adequate to test our hypotheses in SEM and MANOVA (α = 0.05), ranging between 0.95 and 0.99 (MANOVA required fewer participants, via G*Power; Faul et al., 2009).
After excluding 132 careless responders (self-report, instructed responses; Meade & Craig, 2012), our final sample consisted of 460 German working professionals (SoSci panel). The participants were on average 45.5 years old (SD = 10.9, range 20–76), with 59.7% women and 38.5% men. 74.3% held a university degree. Further, 53.9% had prior hiring experience and 39.8% were LinkedIn users; 67.6% used social media at least once a week.
Procedure
After informed consent, and similar to earlier research in this field (Mönke et al., 2024a; Pu et al., 2023; Roth et al., 2020, 2024), we instructed participants to act as a recruiter for a job (project manager) with typical job requirements (e.g., a degree in a related field, prior experience; see ES01). We asked participants to review the job seeker’s LinkedIn profile and rate hireability by evaluating expectations of task performance, organizational citizenship behavior (OCB), and counterproductive work behavior (CWB), representing core dimensions of employee performance (Borman & Motowidlo, 1997; Spector et al., 2010). We offered a raffle draw of vouchers (3 × 50€).
Experimental Materials: LinkedIn Profiles and Pilot Study
We built mock LinkedIn profiles that we manipulated in information quantity (i.e., incomplete information), the presence of a profile picture, and applicant qualification, resulting in a 3 × 2 × 2 between-subjects design. First, we manipulated these profiles with three levels of incomplete LinkedIn information by changing the quantity of available LinkedIn information. These levels of information quantity indicated three levels of LinkedIn incompleteness and mirrored common LinkedIn privacy settings. That is, we built a full LinkedIn profile, filled in all LinkedIn sections, and then restricted the profile’s detail and length step-by-step (see Roulin & Levashina, 2019; Shields & Levashina, 2016). In Level 1, representing low information quantity and high incompleteness, we presented a LinkedIn profile with maximum privacy settings, i.e., only the candidate’s name and current occupation were available. To simulate that this LinkedIn profile was widely incomplete, we deliberately masked important job-related information, e.g., the applicant’s education and prior occupation. Next, in Level 2, we presented a basic LinkedIn profile: All relevant information about education and the applicant’s job experience was available, but they lacked details like job descriptions, majors, and specific interests. So, in Level 2, we increased the information quantity, but the profile was still incomplete. Finally, in Level 3, we presented a full LinkedIn profile with postings, details about prior jobs, study majors, grades, specific interests, and skills. So, in Level 3, the LinkedIn information was fully complete, i.e., all LinkedIn sections were complete, and the information quantity was at its maximum. Our manipulation is summarized in Table 1. Second, the LinkedIn profiles did or did not include a profile picture. Third, we manipulated the job seeker’s qualification separately to investigate the role of applicant qualification in the incomplete LinkedIn information paradigm. We assumed that a highly qualified candidate for this job would (1) have prior experience related to the position, (2) possess a master’s degree, and (3) have excellent grades and a short duration of studies. Conversely, a less qualified candidate would lack these attributes. Group assignment was random, with 33 to 41 participants per group.3
We conducted a pilot study with 19 LinkedIn users (63.2% women, 36.8% men; M = 27.2 years old, SD = 7.1, range 22–53), for details see ES03. They perceived the LinkedIn profiles as authentic (5-point scale; low qualified: M = 4.3, SD = 0.8; high qualified: M = 4.2, SD = 0.8). As expected, the more qualified profile received higher hireability ratings (Wilcoxon V = 5, p = 0.002). Finally, we pretested several pictures: To manipulate the mere presence of a LinkedIn picture, we selected an image rated as average in both attractiveness and hireability. The chosen picture received medium ratings (5-point scale; attractiveness: M = 2.9, SD = 0.9; hireability: M = 3.2, SD = 0.5). All stimuli are available in the ES04.
Table 1
Overview of the manipulation of linkedin information quantity/incomplete information
Note. x = when the LinkedIn profile contained the feature in this condition; blank space = when the respective profile feature was incomplete/missing. Level 1 = low LinkedIn information quantity, highly incomplete LinkedIn information. Level 2 = medium LinkedIn information quantity, medium LinkedIn incompleteness. Level 3 = high LinkedIn information quantity, low LinkedIn incompleteness. Interests (general) = just adding the current employer and university as interests; Interest (specific) = two magazines relevant to the field
Measures
All measures used 7-point Likert scales (1 = strongly disagree, 7 = strongly agree). For reliability, we used the weighted McDonald’s ω (e.g., Bacon et al., 1995).
Perceived Privacy Intention. We measured perceptions of the candidate’s intent to protect their privacy using four items adapted from Xu et al. (2011), e.g., “The candidate is concerned that their LinkedIn information might be misused.” Reliability was excellent, ω = 0.95.
Trustworthiness. To measure perceptions of the candidate’s trustworthiness, we adapted a 5-item-subscale from the Interpersonal Trust Scale (Buck & Bierhoff, 2012), e.g., “I could expect the candidate to tell me the truth.” Reliability was good, with ω = 0.89.
Perceived Online Professionalism. We assessed perceptions of professional online behavior using a three-item scale adapted from Hartwell (2015). A sample item is “The candidate presents themselves advantageously on the internet.” Due to a low λ = 0.31, we excluded one item after an initial confirmatory factor analysis (CFA), resulting in a 2-item scale with good reliability, ω = 0.88.
Warmth and Competence. We used the 3-item scales from Asbrock (2010) to assess warmth, e.g., “This candidate is good-natured” and competence, e.g., “This candidate is competent.” Reliability was good for both warmth (ω = 0.83) and competence subscales (ω = 0.82).
Expected Task Performance. As the first dimension of hireability ratings, and rooted in one of the main dimensions of performance ratings (Borman & Motowidlo, 1997), we assessed ratings of expected task performance. We applied the 6-item scale from Williams and Anderson (1991), as translated by Staufenbiel and Hartz (2000). A sample item is “This job candidate will adequately complete assigned duties.” Reliability was good, ω = 0.93.
Expected Organizational Citizenship Behavior (OCB). Drawing from the other main dimension of employee performance (Borman & Motowidlo, 1997), we measured ratings of expected OCB with the 7-item OCB-I scale of Williams and Anderson (1991). A sample item is “This job candidate will help others who have heavy workloads.” Reliability was good, ω = 0.92.
Expected Counterproductive Work Behavior (CWB). As an indicator of future negative employee performance, we used the 10-item Counterproductive Work Behavior Checklist (Spector et al., 2010). An example item is, “This candidate will stay home from work and said that they were sick when they were not.” We used a 5-point scale (1 = never, 5 = every day). Reliability was good, ω = 0.92.
Overall Hireability. Adapted from Pu et al. (2023), we assessed overall ratings of the candidate’s hireability with two items: “I would invite this candidate for an interview” and “Overall, I consider this candidate to be suitable for the position.” Reliability was good, ω = 0.91.
Perceived Overall Similarity. Finally, we included perceptions of similarity as a control variable in the SEM because such perceptions can influence cybervetting ratings (Mönke et al., 2024a; Roth et al., 2020). We measured perceptions of similarity with the candidate using five items (Roth et al., 2020; Tepper et al., 2011), e.g., “The job candidate and I are similar in terms of our outlook, perspective, and values.” Reliability was good, ω = 0.94.
Results
Measurement Assessment and Preliminary Analyses
Descriptive statistics and correlations are provided in Table 2. First, we conducted a CFA to test the factor structure. We adhered to established evaluation criteria for good model fit: Comparative fit index (CFI) > 0.90 (McDonald & Ho, 2002), root-mean-square error of approximation (RMSEA) < 0.10 (Browne & Cudeck, 1992), and standardized root-mean-square residual (SRMR) ≤ 0.08 (Hu & Bentler, 1998). Fit of the CFA model was good, χ2Y-B (989) = 2074.3, p < 0.001; CFI = 0.92, RMSEA = 0.052, SRMR = 0.056, with λs ranging between 0.56 and 0.92. Details, including all item loadings, are provided in the ES09. As our focal variables did not follow multivariate normality (Mardia skewness = 966.1, p < 0.001; kurtosis = 21.1, p < 0.001), we used a robust estimator (MLR; Yuan & Bentler, 1997).
Table 2
Descriptive statistics and correlations of study variables
Variable
Study 1
M (SD)
Study 3
M (SD)
1a
2a
3a
4
5
6
7
8
9
10
11
12
13
1. Qualificationa
—
—
—
—
—
.02
.11**
.03
–.08
.16***
.02
.21***
.01
.01
.21***
2. Picturea
—
—
—
—
—
.14***
.22***
–.03
.39***
.09*
.07
.08
.12**
–.01
.10*
3. Info. Quantitya
—
—
—
—
—
.11**
.50***
–.29***
.06
.41***
–.02
.29***
.26***
–.08*
.34***
4. Trustworthiness
3.98 (1.09)
—
—
—
—
—
.28**
.00
.46***
.35***
.39***
.33***
.47***
–.20***
.28***
5. Professionalism
4.20 (1.55)
4.41 (1.63)
.09*
.33***
.46***
—
—
–.25***
.36***
.70***
.23***
.56***
.37***
–.13**
.65***
6. Privacy Intent
2.78 (0.92)
2.84 (0.96)
.02
−.20***
−.30***
—
–.28***
—
–.05
–.15**
.03
–.09*
–.15**
.08
–.06
7. Warmth
3.34 (0.67)
3.78 (0.57)
−.03
.08
.00
—
.20***
–.15**
—
.37***
.42***
.32***
.52***
–.25***
.31***
8. Competence
3.21 (0.79)
3.80 (0.68)
.31***
.01
.03
—
.33***
–.08
.38***
—
.34***
.68***
.44***
–.18***
.70***
9. Perc. Similarity
3.57 (1.00)
4.13 (1.00)
.09*
.01
−.07
—
.18***
–.04
.44***
.39***
—
.24***
.35***
–.19***
.30***
10. Exp. Task Perf
4.69 (1.00)
5.27 (0.98)
.19***
.01
.02
—
.26***
–.08
.37***
.68***
.30***
—
.53***
–.25***
.72***
11. Exp. OCB
4.51 (0.78)
4.98 (0.81)
.05
.02
.01
—
.18***
–.11*
.56***
.44***
.39***
.51***
—
–.34***
.41***
12. Exp. CWB
1.58 (0.52)
1.58 (0.56)
.01
.04
−.02
—
.00
.07
–.37***
–.24***
–.10
–.37***
–.35***
—
–.16***
13. Hireability
4.10 (1.67)
—
—
—
—
—
—
—
—
—
—
—
—
—
—
N = 460 for study 1 (only LinkedIn) above the diagonal. N = 363 for study 3 (LinkedIn and résumé) below the diagonal. Info. Quantity = LinkedIn Information Quantity: Higher information quantity = less incompleteness of LinkedIn information. Perc. = perceived. Exp. = expected. OCB = organizational citizenship behavior. CWB = counterproductive work behavior
aThese variables are experimental factors. Hence, no means, standard deviations, or intercorrelations are reported. For these variables, all reported correlation coefficients are Kendall’s τ (i.e., the correlation between ordinal and continuous data)
*p <.05. **p <.01. ***p <.001
Because Mardia’s test indicated no multivariate normality for the hireability criteria (skewness = 316.7, p < 0.001; kurtosis = 15.8, p < 0.001) and Box’s test indicated no equality of covariance matrices for information quantity (χ2(20) = 90.5, p < 0.001; but for all other predictors ps > 0.05), we used the robust Pillai trace test statistic for the MANOVA (Field et al., 2012). Regarding the assumptions of ANOVA, QQ plots indicated normality for task performance, OCB, and overall hireability; CWB was log-transformed to achieve normality. Levene’s test indicated homoscedasticity for all hireability criteria (task performance: F(100) = 1.0, p = 0.55), OCB: F(100) = 0.8, p = 0.87, CWB: F(100) = 0.8, p = 0.87, overall hireability: F(100) = 0.8, p = 0.93). To control for α-inflation, we used robust Tukey–Kramer tests for pairwise comparisons.
Hypotheses Testing
In line with Hypothesis 1, a MANOVA indicated a significant influence of LinkedIn information quantity (i.e., LinkedIn incompleteness) on the hireability ratings, F(8, 880) = 15.9, p < 0.001, η2partial = 0.13. That is, ANOVAs (Type 2) indicated that incomplete LinkedIn information had a large effect on ratings of overall hireability and task performance, medium-sized effects on expected OCB, and small effects on CWB, see Table 3. Specifically, supporting H1 and as illustrated in Fig. 2, more LinkedIn detail increased hireability ratings, as Tukey’s ps < 0.05 between all levels of LinkedIn information quantity for ratings of overall hireability and task performance. For ratings of OCB and CWB, the difference between information quantity Level 1 (only name and current occupation available, high incompleteness) and Level 2 (basic information about education and prior employers, medium incompleteness), as well as between Levels 1 and 3 (fully complete profile with all details), was significant, Tukey’s ps < 0.05; but the rating difference was not significant between Levels 2 and 3 (OCB: Tukey p = 0.75, CWB: Tukey p = 0.99). Notably, the effect of information quantity was larger than the influence of candidate qualification in the MANOVA, F(4, 439) = 10.7, p < 0.001, η2partial = 0.09, and it was also larger for the ratings of overall hireability, task performance, OCB, and CWB in the ANOVAs (see Table 3). In addition, we found no significant interaction between incomplete information and qualification.
Table 3
Study 1 (LinkedIn Profile): results from ANOVAs
Predictor
Overall hireability
F
df
p
Partial η2
Qualification
30.89
1
<.001
.07
Picture presence
9.97
1
.002
.02
LinkedIn information quantity
61.52
2
<.001
.22
Interaction terms
Qualification x Info. quantity
3.18
2
.04
.01
Qualification x Picture
0.01
1
.92
.00
Picture x Info. quantity
4.99
2
.01
.02
Qualification x Picture x
Info. quantity
0.12
2
.89
.00
Control variables
Prior hiring experience
6.02
1
.02
.01
Gender
1.33
4
.26
.01
LinkedIn use
0.32
1
.57
.00
Task performance
F
df
p
Partial η2
Qualification
25.50
1
<.001
.05
Picture presence
6.11
1
.01
.01
LinkedIn information quantity
74.45
2
<.001
.14
Interaction terms
Qualification x Info. quantity
0.87
2
.63
.00
Qualification x Picture
0.20
1
.61
.00
Picture x Info. quantity
1.12
2
.55
.00
Qualification x picture x
Info. quantity
0.28
2
.81
.00
Control variables
Prior hiring experience
10.29
1
.001
.02
Gender
0.86
4
.49
.01
LinkedIn use
5.97
1
.01
.01
Organizational citizenship behavior
F
df
p
Partial η2
Qualification
0.001
1
.98
.00
Picture presence
11.00
1
.001
.02
LinkedIn information quantity
27.63
2
<.001
.11
Interaction terms
Qualification x Info. quantity
1.38
2
.25
.01
Qualification x Picture
0.77
1
.38
.00
Picture x Info. quantity
0.15
2
.86
.00
Qualification x Picture x
Info. Quantity
0.54
2
.59
.00
Control variables
Prior hiring experience
6.28
1
.01
.01
Gender
0.50
4
.73
.00
LinkedIn use
2.23
1
.14
.01
Counterproductive work behavior
F
df
p
Partial η2
Qualification
0.24
1
.62
.00
Picture presence
0.03
1
.85
.00
LinkedIn information quantity
4.34
2
.01
.02
Interaction terms
Qualification × Info. quantity
1.97
2
.14
.01
Qualification × Picture
0.18
1
.67
.00
Picture × Info. quantity
0.24
2
.78
.00
Qualification × Picture × Info. quantity
1.66
2
.19
.01
Control variables
Prior hiring experience
0.01
1
.94
.00
Gender
3.34
4
.01
.03
LinkedIn use
3.48
1
.06
.01
Note. N = 460. Info. quantity = LinkedIn information quantity. Higher LinkedIn information quantity = lower incompleteness of LinkedIn information, see Table 1
Fig. 2
Results from the ANOVAs, study 1 (only LinkedIn profile available). Note. N = 460. Displayed are estimated marginal means. Error bars represent the 95% CI of the estimated marginal means. Higher LinkedIn information quantity = lower incompleteness of LinkedIn information, see Table 1. Level 1 = maximum privacy settings, Level 2 = basic LinkedIn profile, and Level 3 = full LinkedIn profile
Hypothesis 2a focused on the impact of (not) presenting a LinkedIn picture profile on hireability ratings: Indeed, a MANOVA indicated a significant effect F(4, 439) = 4.4, p = 0.002, η2partial = 0.04. As displayed in Fig. 2, the presence of a LinkedIn picture substantially increased ratings of overall hireability (Tukey p = 0.004), task performance (Tukey p = 0.03), and OCB (Tukey p = 0.002), but it did not impact expectations of CWB (Tukey p = 1.00). Thus, we found partial support for Hypothesis 2a. In contrast to Hypothesis 2b, effect sizes were substantially smaller than the η2 for LinkedIn information quantity for overall hireability. The effect of presenting a picture did not depend on the candidate’s qualification, as we found no interaction effect.
To test Hypotheses 3 to 7, we applied structural equation modeling (SEM; Rosseel, 2012). We used the residual-centering approach to model latent interaction variables (Little et al., 2006) and calculated robust 95% confidence intervals (CI) to test indirect effects and path differences (Monte Carlo method; see MacKinnon et al., 2004; Preacher & Selig, 2012). As outlined in the hypotheses, we modeled warmth, competence, professionalism, and trust as mediators between LinkedIn information quantity (i.e., varying levels of incomplete LinkedIn information) and hireability ratings. Given that warmth and competence are broad judgment dimensions, they may encompass trustworthiness and professionalism as lower-level facets (see Cuddy et al., 2008; Fiske, 2018). Specifically, the medium to high intercorrelations (see Table 1) suggest that trust could be a facet of warmth, and professionalism might be a facet of competence perceptions. Therefore, we tested whether more LinkedIn details (i.e., less incomplete information) predicted perceptions of online professionalism and trustworthiness, which in turn influenced overall impressions of competence and warmth. These impressions were expected to predict hireability ratings (i.e., serial mediation to task performance, OCB, and CWB), see Fig. 3.4
Fig. 3
Results from the SEM: structural model, Study 1. Note. N = 460. χ.2 (1546) = 3079.1, p <.001; CFI =.91, RMSEA =.050 (90% CI [.048,.053]), SRMR =.07. For better readability, we report only incremental paths: That is, we included info quantity, qualification, and picture as a predictor for every DV. Dotted lines indicate p >.05. Full results are available in Table 4. Path coefficients are standardized, β. Info. quantity = LinkedIn information quantity. Higher LinkedIn information quantity = lower incompleteness of LinkedIn information, see Table 1. * p <.05
Overall model fit was good, χ2 (1546) = 3079.1, p < 0.001; CFI = 0.91, RMSEA = 0.050 (90% CI [0.048, 0.053]), SRMR = 0.07. Results of the structural model are presented in Table 4 and Fig. 3. In partial support of Hypotheses 3 and 5, perceptions of trustworthiness and warmth mediated the effect between LinkedIn information quantity and hireability ratings: Having more LinkedIn details predicted perceived trustworthiness (β = 0.11, p = 0.03), as did profile picture presence (β = 0.11, p = 0.02), but qualification did not (β = 0.00, p = 0.95). Trustworthiness predicted warmth (β = 0.36, p < 0.001). Warmth eventually predicted expectations of task performance (β = 0.26, p = 0.01), OCB (β = 0.71, p < 0.001), and CWB (β = –0.43, p < 0.001). The respective indirect effect was significant for OCB (b = 0.03, 95% CI [0.003, 0.06]) and CWB (b = –0.01, 95% CI [–0.02, –0.001]) but not for task performance (the 95% CI included zero; b = 0.01, 95% CI [0, 0.02]). The indirect effect from picture presence via trust and warmth was significant for task performance (b = 0.02, 95% CI [0.001, 0.04]), OCB (b = 0.05, 95% CI [0.01, 0.10]), and CWB (b = –0.02, 95% CI [–0.04, –0.002]). We found no indirect effect of qualification on task performance through these mediators (b = 0.00, 95% CI [–0.02, 0.02]), OCB (b = –0.001, 95% CI [–0.04, 0.04]), and CWB (b = 0.00, 95% CI [–0.01, 0.02]).
Table 4
Results from the structural model, study 1
Variable
b (SE)
[95%-CI]
z
β
Hireability
Task performance
1.33
(0.12)
[1.08, 1.57]
10.68*
0.70
OCB
0.02
(0.09)
[–0.15, 0.20]
0.26
0.01
CWB
0.13
(0.13)
[–0.12, 0.38]
1.04
0.04
Info. Quantity
0.37
(0.08)
[0.22, 0.51]
4.81*
0.18
Picture
0.19
(0.11)
[–0.03, 0.41]
1.66
0.06
Qualification
0.28
(0.12)
[0.04, 0.52]
2.30*
0.09
Task Performance
Warmth
0.33
(0.13)
[0.08, 0.59]
2.59*
0.26
Competence
0.99
(0.11)
[0.77, 1.21]
8.85*
0.86
Similarity
–0.14
(0.06)
[–0.27, –0.02]
–2.25*
–0.17
Info. Quantity
–0.13
(0.06)
[–0.25, –0.002]
–1.99*
–0.12
Picture
–0.21
(0.10)
[–0.40, –0.02]
–2.13*
–0.12
Qualification
0.13
(0.08)
[–0.02, 0.28]
1.70
0.08
OCB
Warmth
0.92
(0.18)
[0.57, 1.27]
5.12*
0.71
Competence
0.25
(0.12)
[0.02, 0.48]
2.16*
0.22
Similarity
–0.05
(0.07)
[–0.18, 0.09]
–0.64
–0.05
Info. Quantity
0.10
(0.06)
[–0.02, 0.23]
1.64
0.10
Picture
–0.38
(0.13)
[–0.63, –0.13]
–2.98*
–0.22
Qualification
0.00
(0.09)
[–0.17, 0.18]
0.05
0.00
CWB
Warmth
–0.32
(0.09)
[–0.49, –0.15]
–3.72*
–0.43
Competence
–0.02
(0.07)
[–0.17, 0.12]
–0.32
–0.04
Similarity
0.02
(0.03)
[–0.05, 0.08]
0.49
0.03
Info. Quantity
–0.04
(0.04)
[–0.11, 0.04]
–1.00
–0.06
Picture
0.19
(0.07)
[0.06, 0.33]
2.76*
0.20
Qualification
–0.02
(0.06)
[–0.14, 0.10]
–0.36
–0.02
Warmth
Similarity
0.22
(0.04)
[0.14, 0.30]
5.55*
0.33
Trust
0.23
(0.03)
[0.16, 0.29]
7.20*
0.36
Qualification
–0.14
(0.05)
[–0.24, –0.03]
–2.61*
–0.11
Picture
0.57
(0.07)
[0.42, 0.71]
7.70*
0.43
Info. Quantity
0.06
(0.03)
[–0.01, 0.13]
1.82
0.08
Competence
Similarity
0.16
(0.03)
[0.09, 0.22]
4.61*
0.21
Professionalism
0.38
(0.04)
[0.30, 0.46]
9.66*
0.74
Qualification
0.17
(0.05)
[0.06, 0.28]
3.14*
0.11
Picture
–0.16
(0.06)
[–0.27, –0.04]
–2.64*
–0.10
Info. Quantity
0.05
(0.06)
[–0.06, 0.16]
0.95
0.06
Professionalism
Similarity
0.33
(0.06)
[0.21, 0.45]
5.41*
0.23
Privacy Intention
–0.05
(0.07)
[–0.19, 0.09]
–0.73
–0.03
Qualification
0.29
(0.10)
[0.09, 0.49]
2.83*
0.10
Picture
0.83
(0.11)
[0.63, 1.04]
7.90*
0.29
Info. Quantity
1.14
(0.07)
[1.01, 1.28]
16.66*
0.64
Privacy Intention × Info. Quantity
–0.07
(0.08)
[–0.23, 0.09]
–0.87
–0.03
Privacy Intention × Picture
0.25
(0.14)
[–0.02, 0.51]
1.83
0.07
Trust
Similarity
0.44
(0.07)
[0.31, 0.58]
6.47*
0.42
Privacy Intention
0.03
(0.07)
[–0.10, 0.16]
0.43
0.02
Qualification
–0.01
(0.09)
[–0.19, 0.18]
–0.06
0.00
Picture
0.23
(0.10)
[0.03, 0.43]
2.30*
0.11
Info. Quantity
0.14
(0.06)
[0.01, 0.26]
2.18*
0.11
Privacy Intention × Info. Quantity
0.06
(0.07)
[–0.08, 0.20]
0.81
0.04
Privacy Intention × Picture
0.00
(0.13)
[–0.26, 0.25]
–0.02
0.00
N = 460. β = standardized path coefficient. Predictors are indented, dependent variables are not indented. The boldface indicates significant predictors. Info. Quantity = LinkedIn information quantity. Higher LinkedIn information quantity = lower incompleteness of LinkedIn information, see Table 1. We note that some negative direct effects reported above may reflect a suppression effect, where the mediator absorbs shared variance and inverts the residual relationships. *p <.05 (two-tailed)
In partial support of Hypotheses 4 and 6, perceptions of online professionalism and competence mediated the effect between LinkedIn information quantity and hireability ratings: More LinkedIn information was associated with higher ratings of professionalism (β = 0.64, p < 0.001), as were profile picture presence (β = 0.29, p < 0.001) and qualification (β = 0.10, p = 0.01). As expected, professionalism predicted perceptions of competence (β = 0.74, p < 0.001). Finally, competence predicted ratings of task performance (β = 0.86, p < 0.001) and OCB (β = 0.22, p = 0.03), but not CWB (β = –0.04, p = 0.75). The hypothesized mediation effect of professionalism and competence between LinkedIn information quantity and hireability ratings was significant for task performance (b = 0.43, 95% CI [0.33, 0.54]) and OCB (b = 0.11, 95% CI [0.01, 0.20]) but not CWB (b = –0.01, 95% CI [–0.07, 0.05]). We found similar results for the decision process regarding profile picture presence: The mediation effect was significant for task performance (b = 0.31, 95% CI [0.22, 0.42]) and OCB (b = 0.08, 95% CI [0.01, 0.15]) but not CWB (b = –0.01, 95% CI [–0.05, 0.04]). We observed similar effects for candidate qualification: Professionalism and competence were mediators for task performance (b = 0.11, 95% CI [0.03, 0.19]) and OCB (b = 0.03, 95% CI [0.001, 0.06]), but not for CWB (b = –0.003, 95% CI [–0.02, 0.01]). All mediation analyses are provided in detail in the ES09.
In contrast to our assumptions, perceptions of privacy intention (Hypothesis 7) did not moderate the relationship between LinkedIn information quantity and trust (β = 0.04, p = 0.81), nor that between profile picture presence and trust (β = 0.00, p = 0.99). Perceived privacy intention was also not a moderator between LinkedIn information quantity (β = –0.03, p = 0.38), profile picture presence (β = 0.07, p = 0.07), and professionalism.
As an exploratory analysis, we revisited the visual primacy for Hypothesis 2b: Interestingly, contrary to our expectations, the presence of a LinkedIn picture had a smaller impact on professionalism ratings than the quantity of LinkedIn information, Δb = –0.31, 95% CI [–0.55, –0.07]. For trustworthiness, we found no difference (Δb = 0.10, 95% CI [–0.11, 0.31]).
Effect of Control Variables
As control variables in the (M)ANOVAs, we assessed gender, prior hiring experience, and LinkedIn usage. With a small effect size, participants who indicated prior hiring experience provided lower ratings of overall hireability, F(1) = 6.02, p = 0.02, η2partial = 0.01; task performance, F(1) = 10.29, p = 0.001, η2partial = 0.02; and OCB, F(1) = 6.28, p = 0.01, η2partial = 0.01. LinkedIn users rated task performance higher, F(1) = 5.97, p = 0.01, η2partial = 0.01; but all other effects were non-significant, p > 0.05. Finally, we included perceptions of similarity as a control variable in the SEM. Similarity predicted only ratings of OCB and CWB, see Table 4.
Discussion
The first goal of this study was to transfer the (traditional) incomplete information paradigm (Jagacinski, 1991; R. D. Johnson & Levin, 1985) to the context of social media assessments on LinkedIn. As expected, and in line with prior research (e.g., Hartwell & Campion, 2020; Roulin & Levashina, 2019), presenting more LinkedIn details increased ratings of a job seeker’s overall hireability as well as task performance and OCB, and reduced expectations of CWB. Conversely, incomplete LinkedIn information, i.e., fewer LinkedIn details and lower LinkedIn information quantity, reduced hireability perceptions. Thus, we find support for the core concept of our incomplete LinkedIn information paradigm: Raters favored job seekers with more complete LinkedIn profiles and penalized those with incomplete profiles. Importantly, the effects of incomplete LinkedIn information were substantially larger than the effects of qualification. Beyond research in traditional incomplete information paradigms, we showed that even candidates with high qualifications faced a significant devaluation in hireability. Even in times of digital assessments and online active sourcing, raters had negative views of candidates with incomplete social media. From a holistic perspective, it seems that incomplete LinkedIn information was used as a signal in the screening of job candidates: When the LinkedIn profile was incomplete, they seemed to expect a lower fit (see also Highhouse & Hause, 1995; Roth et al., 2016). Thus, our results support the notion that recruiters value more comprehensive LinkedIn profiles (Roulin & Levashina, 2019).
As a second objective, we explored the key perceptions linked to fewer LinkedIn details to unpack the underlying mechanisms of the incomplete LinkedIn information paradigm. As expected, incomplete LinkedIn information attenuated perceptions of online professionalism and trustworthiness, and more LinkedIn details increased such perceptions. This underlines the notion that suspicion (Pu et al., 2023; Roth et al., 2024) and ambiguity avoidance (Roth et al., 2016) play an important role in cybervetting. Ambiguity about the candidate’s trustworthiness and professionalism led to a devaluation of the broader dimensions of warmth and competence. This ultimately decreased hireability ratings and, through expectations regarding the candidate’s task performance, reduced the likelihood that raters would pursue the candidate or invite them to a job interview (overall hireability). We note that such impressions could reflect an incomplete information bias (Johnson, 1987) because there is no clear connection between LinkedIn profile length and job seekers’ performance or KSAOs. By uncovering these perceptions associated with incomplete information, our results connect incomplete LinkedIn information to the literature on social cognition and first impressions (Cuddy et al., 2011; Fiske et al., 2007; Highhouse & Hause, 1995; Jaccard & Wood, 1988). In a broader context, we believe that this also enhances our understanding of the underlying processes in traditional incomplete information paradigms (Jaccard & Wood, 1988; Jagacinski, 1991; Johnson & Levin, 1985), as similar processes may be relevant.
Contrary to our expectations, the effect of incomplete information was not mitigated by the perception of the candidate’s privacy intention. We expected that less LinkedIn information would be strongly linked to the perception that the applicant intends to protect their privacy on social media, but the association was only small. Our manipulation may have been too weak to evoke such perceptions. For example, Instagram’s “This account is private” disclaimer could make the use of privacy settings more salient.
Study 2
After establishing the incomplete LinkedIn information paradigm, we then conducted an interview study to further verify the negative effect of incomplete information and the positive effect of providing more detailed LinkedIn profiles. Also, we aimed to make the implicit process of devaluing applicants with incomplete LinkedIn information more tangible (e.g., McFarland & Ployhart, 2015). Thus, we collected verbal protocols from HR professionals evaluating more or less incomplete LinkedIn profiles and asked them to explain if and how the LinkedIn information quantity influenced their considerations. This approach enabled a more open identification of decision processes and the implications of incomplete LinkedIn profiles.
Method
Sample
Our convenience sample consisted of 32 German HR professionals. They were on average 51.9 years old (SD = 9.7, range 28–67). Twenty-five percent identified as women and 75% as men. In the sample, 84.4% held a university degree, and they averaged 16.7 years of hiring experience (SD = 9.1, range 3–40). Most participants (87.5%) had a managerial role. Whereas 56.3% used LinkedIn, all participants indicated at least some familiarity with the platform.
Procedure and Design
After informed consent and reviewing the job posting, participants were asked to inspect one of the LinkedIn profiles from Study 1. As in Study 1, the profiles were manipulated in a 3 (levels of information quantity/incomplete information) × 2 (qualification) between-subject design. Mirroring the hireability dimensions applied in Study 1, we asked participants to rate how they expected the candidate to meet the performance requirements of the job role (task performance; Williams & Anderson, 1991), how much they would expect the candidate to go out of way to help new employees (OCB, Williams & Anderson, 1991), how often they expected that the candidate stays home from work and says that he was sick when he was not (CWB, Spector et al., 2010), and then rated overall hireability. After each rating, we probed participants to elaborate on their considerations, verbalize how they came to their rating, and what criteria they applied (see interview guideline in ES10). This approach followed the principles of cognitive interviewing (e.g., Miller et al., 2014). We analyzed data with MAXQDA, the coding scheme was induced from Study 1.
Results and Discussion
First, the information quantity provided on LinkedIn was indeed relevant to the decision processes of HR professionals. Of the 470 interview passages that referred to the LinkedIn content, 31.3% mentioned that LinkedIn information was incomplete (e.g., “[I rate 4 out of 7], not more, because I do not see more detailed work experience”). Incomplete LinkedIn information was highlighted more often when the profiles had maximum privacy settings (Level 1 of information quantity, 63 codes) or only basic information (Level 2, 31 codes) than when the full profile was presented (Level 3, 11 codes; χ2(2) = 68.5, p < 0.001), see Table 5. For instance, recruiters referred to incomplete information regarding work experience (Level 1: 21 coded passages, Level 2: 8, Level 3: 0; χ2(2) = 23.8, p < 0.001) and education (Level 1: 24 coded passages, Level 2: 8, Level 3: 1; χ2(2) = 25.9 p < 0.001) more often.
Table 5
Results from study 2, distribution of coded interview passages
Code
Level of LinkedIn Information Quantity
(Sub)
Total
Level 1
Maximum privacy
Level 2
Basic profile
Level 3
Full profile
Reference to present LinkedIn content
323
61
124
138
Picture
20
11
7
2
Work experience
136
31
50
55
negative
81
23
32
26
neutral
18
4
6
8
positive
37
4
12
21
Occupation
13
5
5
3
Volunteering
54
–
32
22
Education
53
–
20
33
Interests
15
6
3
6
Skills
15
–
–
15
Postings
0
–
–
0
Contacts
5
2
3
0
Other (e.g., name, follower)
12
6
4
2
Reference to incomplete LinkedIn content
147
94
37
16
Work experience
29
21
8
0
Volunteering
2
2
0
0
Education
33
24
8
1
Skills
21
5
7
9
Postings
9
5
4
0
Other (e.g., age, hobbies, general comment)
53
37
10
6
Frustration
18
9
6
3
Note. N = 32. Dashes indicate that this profile feature was not presented in this group and hence was not mentioned. Level 1 = low LinkedIn information quantity, highly incomplete LinkedIn information. Level 2 = medium LinkedIn information quantity, medium LinkedIn incompleteness. Level 3 = high LinkedIn information quantity, low LinkedIn incompleteness. See also Table 1
Detailed results of the hireability ratings are available in ES 09, Table ES7. In general, and replicating the core idea of the incomplete LinkedIn information paradigm (Study 1), more detailed LinkedIn profiles received higher ratings of overall hireability (Level 3 vs. Level 1: Wilcoxon W = 7.5, p = 0.001), expected task performance (Wilcoxon W = 15.5, p = 0.004), and OCB (Wilcoxon W = 22.5, p = 0.02) but not lower ratings of CWB (Wilcoxon W = 62, p = 0.31). For example, a professional asked: “He did an internship, but was it in HR? Was it in project development? In which area did he do it? I’m just not getting enough information [out of the profile], I can’t grasp the young man.” Another professional explained: “I would say 1 [out of 7, regarding OCB]. I can’t see from this profile that he takes much initiative. […] You can’t see [OCB] because there haven’t been any posts or activities recently. There’s not much initiative, the same goes for interests. Someone proactive would have [posted] more varied interests.”
Again, incomplete information had the power to mask the candidate’s true qualifications. For example, ratings of overall hireability did not differ between less and more qualified candidates in LinkedIn information quantity L evel 1 (Wilcoxon W = 11, p = 0.91) and Level 2 (Wilcoxon W = 8.5, p = 0.25), but they did when the full profile was presented (Wilcoxon W = 1, p = 0.02). Also, supporting the proposition that incomplete information causes negative feelings about a candidate (Jaccard & Wood, 1988; Jagacinski, 1991), participants expressed more frustration when evaluating a profile with more incomplete information (LinkedIn information quantity Level 1: 9 passages, Level 2: 6, Level 3: 3). For instance, a participant said that “without further information, I can’t make anything of this profile, absolutely nothing. I find [him] uninteresting due to his profile.”
Across conditions, the most considered parts of the LinkedIn profiles were prior work experience (136 codes, 42.1%), volunteering activities (54 codes, 16.7%), and education (53 codes, 16.4%). Notably, work experience was referred to less when the LinkedIn profile was more incomplete (LinkedIn information quantity Level 1: 31 coded passages, Level 2: 50, Level 3: 55; χ2(2) = 6.5, p = 0.04). This indicates that when less LinkedIn information was available to raters, they relied more on non–job-related information, such as the profile picture (Level 1: 11 coded passages, Level 2: 7, Level 3: 2; χ2(2) = 6.3, p = 0.04). For instance, a participant in Level 1 said that “just from his photo, I would say [he] gives an alert, young, dynamic, and tidy impression, which plays a big role. It seems like he is a team player.”
In contrast to some prior research (Zhang et al., 2020), we found that recruiters focused more on job-related information in social media assessments (work experience, education) on LinkedIn than on personal information (pictures, postings). For example, raters referenced the profile picture in only 20 interview passages (6.2%). However, as a limitation, we recorded only direct mentions and implicit processes could be important, too.
Study 3
Studies 1 and 2 highlighted that incomplete LinkedIn profiles can attenuate hireability perceptions, and increasing the LinkedIn information quantity was associated with better hireability ratings. Yet, in both studies, we approached LinkedIn assessments from the perspective of active sourcing, i.e., raters screened passive job seekers solely on their LinkedIn profile without considering other application materials. We modeled this procedure on cybervetting practices (Suen, 2018) and the latest studies in this field (e.g., Mönke et al., 2024a; Roth et al., 2020; Roulin & Levashina, 2019; Zhang et al., 2020). However, several interview studies reported that social media assessments are also often used as extensions of background checks, and hence, other application materials like the applicant’s résumé are available to cybervetting raters as well (Berkelaar, 2014, 2017; Hoek et al., 2016). For example, Berkelaar and Buzzanell (2015) reported that cybervetting is often done “soon after receiving résumés” (p. 93). If a résumé is available as well, recruiters have access to a more structured information source than social media: How much value will they place on LinkedIn information in this context? To test the robustness of our incomplete LinkedIn information paradigm, we think that it is critical to examine such a context as well. We suppose that if a rater encounters very little LinkedIn information, an applicant’s complete résumé might buffer the incomplete LinkedIn information. Providing the résumé could shift the rater’s attribution of incomplete LinkedIn information, for example, toward the use of privacy settings rather than hiding a potential person-job misfit. Thus, we hypothesize:
(H8)Résumé availability moderates the effect of incomplete information: If a complete résumé is presented, the effect of incomplete LinkedIn information on hireability ratings is attenuated.
Method
Sample
As an a priori power analysis, we ran 1000 Monte Carlo simulations with N = 350, assuming Cronbach’s α = 0.80 and small correlations between factors (r = 0.15). Power to detect these effects with α = 0.05 ranged between 0.80 and 0.95, indicating an adequate sample size. After excluding 26 careless responders via instructed responses and a self-report item, our final sample consisted of 363 German working professionals. The participants (Prolific panel) were on average 33.2 years old (SD = 10, range 18–66). 40.2% identified as women and 57.3% as men. Further, 68.3% held a university degree and 61.2% had hiring experience; 62.8% used LinkedIn. 93.1% used social media at least once a week.
Procedure
The procedure was identical to Study 1, except that it was applied to a more traditional selection context, where we presented both the applicant’s résumé and LinkedIn profile. Screening the LinkedIn profile now served as a background check, as the résumé provided all relevant job-related details, such as prior work experience, education, or volunteering. We consider the résumé complete because it contained all the information presented in the complete LinkedIn profile (Study 1: Level 3, see Table 1) and typically desired by HR managers (like education, prior experience; see Knouse, 1994). The résumé was manipulated for qualification (low or high), whereas the LinkedIn profile was manipulated in terms of qualification, information quantity (varying degrees of incomplete information), and profile picture presence (2 × 4 × 2 between-subjects design). To examine how recruiters would rate applicants with no LinkedIn presence, we added another minimum level of LinkedIn information quantity: a screenshot indicating that the applicant had no LinkedIn profile. Group assignment was random, resulting in 22 to 30 participants per group. The participants received £1.05.
Measures
Measures were identical to Study 1.5 Again, due to a low loading, the scale of perceived online professionalism was reduced to two items. For all scales, reliability was good (warmth: ω = 0.73, competence: ω = 0.92) to excellent (perceived online professionalism: ω = 0.91, privacy intention: ω = 0.94, similarity: ω = 0.92, task performance: ω = 0.93, OCB: ω = 0.91, CWB: ω = 0.91).
Results
Descriptive statistics and correlations are provided in Table 2. Fit of the CFA model was good, χ2Y-B (712) = 1452.0, p < 0.001; CFI = 0.91, RMSEA = 0.058, SRMR = 0.058, see Table ES8. We used the robust Pillai trace test statistic for the MANOVA because Mardia’s test indicated no multivariate normality for the three hireability criteria (DVs, skewness = 170.4, p < 0.001; kurtosis = 10.2, p < 0.001), and Box’s test indicated no equality of covariance matrices for LinkedIn information quantity, χ2(18) = 29.5, p = 0.04; all other predictors ps > 0.05. For the ANOVA, we log-transformed the CWB variable to achieve normality, whereas QQ plots indicated normality for task performance and OCB. Levene’s test indicated homoscedasticity for task performance, F(110) = 1.0, p = 0.54; OCB, F(110) = 0.9, p = 0.84; and CWB, F(110) = 0.9, p = 0.77.
Hypotheses Testing
A MANOVA indicated a significant effect of applicant qualification, F(3, 349) = 9.3, p < 0.001, η2partial = 0.07, but no significant effect of LinkedIn information quantity, F(9, 1053) = 0.3, p = 0.98, η2partial = 0.002, and profile picture presence, F(3, 349) = 0.6, p = 0.59, η2partial = 0.005, on hireability ratings. As illustrated in Fig. 4, information quantity and profile picture presence did not predict ratings of task performance, OCB, or CWB. Notably, not even the complete absence of a LinkedIn profile influenced hireability ratings negatively. Our results support Hypothesis 8: The negative effect of incomplete LinkedIn information was attenuated (and even absent) when the applicant’s complete résumé was available. For detailed results, see Table 6.
Fig. 4
Results from the ANOVA, study 3 (LinkedIn and résumé available). Note. N = 363. Displayed are estimated marginal means. Error bars represent the 95% CI of the estimated marginal means. Higher LinkedIn information quantity = lower incompleteness of LinkedIn information, see Table 1. Level 1 = maximum privacy settings, Level 2 = basic LinkedIn profile, Level 3 = full LinkedIn profile
Table 6
Study 3 (LinkedIn Profile and Résumé): results from ANOVAs
Predictor
Task performance
F
df
p
Partial η2
Qualification
21.97
1
<.001
.06
Picture presence
0.00
1
.95
.00
LinkedIn information quantity
0.33
3
.80
.00
Interaction terms
Qualification × Info. quantity
0.46
3
.71
.00
Qualification × Picture
1.38
1
.24
.00
Picture × Info. quantity
2.06
2
.13
.01
Qualification × Picture × Info. quantity
0.42
2
.66
.00
Control variables
Prior hiring experience
1.29
1
.26
.00
Gender
1.91
4
.11
.02
LinkedIn use
1.16
1
.28
.00
Organizational citizenship behavior
F
df
p
Partial η2
Qualification
1.60
1
.21
.01
Picture presence
0.48
1
.49
.00
LinkedIn information quantity
0.52
3
.67
.01
Interaction terms
Qualification × Info. quantity
0.87
3
.46
.01
Qualification × Picture
1.27
1
.26
.00
Picture × Info. quantity
1.06
2
.35
.01
Qualification × Picture × Info. quantity
0.88
2
.42
.01
Control variables
Prior hiring experience
0.14
1
.71
.00
Gender
1.14
4
.34
.01
LinkedIn use
8.97
1
.003
.03
Counterproductive work behavior
F
df
p
Partial η2
Qualification
0.09
1
.76
.00
Picture presence
0.94
1
.33
.00
LinkedIn information quantity
0.26
3
.86
.00
Interaction terms
Qualification × Info. quantity
0.82
2
.48
.01
Qualification × Picture
0.01
1
.94
.00
Picture × Info. quantity
4.17
2
.02
.02
Qualification × Picture × Info. Quantity
0.16
2
.85
.00
Control variables
Prior hiring experience
0.00
1
.96
.00
Gender
0.96
4
.43
.01
LinkedIn use
0.62
1
.43
.00
Note. N = 363. Info. Quantity = LinkedIn information quantity. Higher LinkedIn information quantity = lower incompleteness of LinkedIn information, see Table 1
N = 363
As an additional analysis, available in the ES09, Tables ES9 and ES10, we revisited decision processes in this rating context as well. We fitted an SEM: Overall, we found support for the key decision processes proposed in Study 1 (professionalism, warmth, and competence), but they were only relevant as mediators between applicant qualification and hireability ratings and not between LinkedIn information quantity and hireability ratings. Again, perceptions of applicant privacy did not moderate the assumed relations (all ps > 0.05).
Effect of Control Variables
We assessed gender, prior hiring experience, and LinkedIn usage as control variables in the ANOVAs. LinkedIn users rated OCB higher, F(1) = 9.0, p = 0.003, but all other ps > 0.05. Also, in the SEM, we included perceptions of similarity as a control variable. Similarity only had a main effect on ratings of CWB, β = 0.24, p = 0.001; all other ps > 0.05.
Discussion
The goal of this study was to test whether incomplete LinkedIn information still reduces hireability ratings when LinkedIn is considered alongside an applicant’s complete résumé. Unlike Studies 1 and 2, which presented raters with the scenario of LinkedIn assessments in active sourcing (i.e., no additional application materials available), this study framed the LinkedIn screening as a background check (i.e., in addition to the applicant’s résumé). As expected, the negative effect of incomplete LinkedIn information was absent: Providing raters with the applicant’s résumé attenuated the negative perceptions reported in Studies 1 and 2.
This has important implications. First, when incomplete information could be inferred from the résumé, raters did not penalize the applicant (in contrast to the rationale of Jaccard & Wood, 1988; Jagacinski, 1991). Instead, they appeared to rely on the résumé for their decisions, without ambiguity about applicants with incomplete information. Therefore, incomplete LinkedIn information only led to negative perceptions when the applicant’s résumé was absent, such as in active sourcing. Generally, incomplete information in one part of a candidate’s application (social media: LinkedIn) had no negative impact when the information was available from another source (résumé).
General Discussion
Hiring practices have changed dramatically in the last decades: Assessing online information and social media has emerged as a new context for hireability ratings (see recent calls for action by Behrend et al., 2024; Roulin et al., 2024; Mönke & Schäpers, 2022; Roth et al., 2016; Wilcox et al., 2022). However, social media assessments remain controversial because, unlike traditional sources, the availability and relevance of job-related information on platforms like LinkedIn vary widely. Thereby, incomplete information became much more common. In this study, we sought to disentangle how raters react to incomplete information in cybervetting—and whether, why, and when they perceive applicants with incomplete LinkedIn information more negatively. To examine this, we manipulated the information quantity on LinkedIn profiles, indicating varying levels of incomplete LinkedIn information.
Implications for Theory
In line with the incomplete information paradigm (Jaccard & Wood, 1988; Jagacinski, 1991; Johnson & Levin, 1985), we found that applicants who had incomplete social media information (i.e., less information on their LinkedIn profiles) or did not post a profile picture received lower hireability ratings. Lower hireability was indicated by lower expectations of applicant task performance, OCB, and increased CWB (key tenets of future employee performance; Borman & Motowidlo, 1997; Spector et al., 2010). This supports the idea that this theory applies to digital assessments: Effect sizes were substantial, and beyond research in the analog paradigms, we observed that incomplete information could obscure an applicant’s true qualification for the job. Our findings extend prior research on social media assessments (e.g., Roulin & Levashina, 2019) by offering a more nuanced understanding of incomplete LinkedIn information and by uncovering the processes underlying the devaluation of candidates with incomplete profiles. That is, impressions of the candidate’s professionalism and trustworthiness were diminished, which led to lower perceptions of warmth and competence (Study 1). Also, suspicion and ambiguity played an important role (Study 2). This connects the incomplete information paradigm (e.g., Johnson & Levin, 1985) with the literature on social cognition and stereotype use (Cuddy et al., 2008; Fiske et al., 2007): When LinkedIn information was incomplete, raters relied on attributions (trustworthiness, professionalism) tied to the fundamental dimensions of social cognition used to evaluate strangers (warmth, competence).
Raters interpreted complete LinkedIn information as an indicator of hireability, possibly associating LinkedIn information quantity with KSAO-related outcomes (Garcia-Retamero & Rieskamp, 2009; see also recruiters’ comments, Study 2). The raters’ rationale aligns with prior research showing that conscientiousness correlates with LinkedIn profile length (Roulin & Levashina, 2019), while extraversion and openness predict greater social media activity (e.g., Correa et al., 2010; Davis et al., 2020). Thus, incomplete social media information may reflect aspects of a candidate’s personality, a cue the raters appeared to rely on.
As an important context effect, we found that incomplete LinkedIn information did not negatively impact hireability ratings when LinkedIn was assessed alongside a complete résumé, reflecting a more traditional selection process where cybervetting serves as a background check (e.g., Berkelaar & Buzzanell, 2015; Hoek et al., 2016). Thus, negative perceptions about applicants arise not merely from incomplete information bias (Jaccard & Wood, 1988; Johnson, 1987) but from the perceived loss of KSAO-related content relative to other candidates (Garcia-Retamero & Rieskamp, 2009; Highhouse & Hause, 1995). Also, this finding highlights a broader distinction in decision processes between active sourcing (screening passive job seekers) and background checks (integrating LinkedIn profiles with résumés). We emphasize the need to explore how traditional information sources integrate with social media assessments (Roth et al., 2016). Our findings suggest that, when available, raters prioritized résumés over LinkedIn profiles. However, the buffering effect observed in Study 3—where a complete résumé mitigated the negative impact of incomplete LinkedIn information—may not generalize to other information sources. For example, could a complete LinkedIn profile offset an incomplete résumé? Could a complete résumé compensate for a missing test score? Recruiters might follow an implicit hierarchy of information sources, assigning different weights to each when making decisions. It remains an open question what relative importance social media assessments (should) have in selection decisions, particularly when other application materials are available.
Implications for HR Practice
For organizations and HR professionals, we recommend exercising caution when selecting candidates based on impressions from social media platforms like LinkedIn. Although raters in our study indicated a focus on job-related information (Study 2), their decisions were influenced by gut feelings and stereotypes, with even the length of a LinkedIn profile impacting evaluations substantially. Suspicion and distrust emerged as critical factors (see also Pu et al., 2023; Roth et al., 2024), leading recruiters to penalize less detailed LinkedIn profiles.
However, there is no robust evidence to suggest that incomplete LinkedIn profiles indicate a candidate’s fit. So, recruiters should not overestimate the validity of social media assessments: The association between LinkedIn profile length, content, and KSAOs such as conscientiousness is only small or close to zero (Cubrich et al., 2021; Mönke et al., 2024b; Roulin & Levashina, 2019). Thus, the effect of LinkedIn profile length on hireability ratings may reflect an incomplete information bias (Johnson, 1987). Demographic disparities in LinkedIn usage add to this concern—Hispanics, for example, use LinkedIn less frequently than White or Black U.S. adults (Auxier & Anderson, 2021). Consequently, biases related to incomplete LinkedIn profiles could disproportionately harm underrepresented groups. Hence, we underline that overreliance on LinkedIn assessments or profile length may conflict with established guidelines (e.g., Society for Industrial & Organizational Psychology, 2018).
At the same time, our Study 3 points toward a possible strategy to mitigate biases arising from social media assessments: Providing an applicant’s complete résumé appeared to shift raters’ focus toward inferring job-related KSAOs from the résumé rather than relying on LinkedIn. So, in active sourcing, this implies that recruiters could build a pool of potential candidates and request their résumés irrespective of the LinkedIn profile content. Incorporating social media assessments as part of a structured background check alongside reviewing other application materials could serve the purpose of increasing the structure of social media assessments (e.g., Hartwell et al., 2022). That said, we encourage organizations to critically evaluate their current cybervetting practices. Applicants frequently view cybervetting negatively (e.g., Cook et al., 2020; Manroop et al., 2022; Schäpers et al., 2025), and there are ongoing concerns about its legality (Drouin et al., 2015; Roth et al., 2020). Balancing the potential benefits of social media assessments with their psychometric, ethical, and legal implications remains crucial for fair and effective recruitment.
For (passive) job seekers, our findings highlight the importance of maintaining an up-to-date LinkedIn profile. Recruiters tended to penalize incomplete LinkedIn profiles, interpreting them as ambiguous and lowering hireability ratings when no full résumé was available to compensate for missing details. Conversely, more comprehensive LinkedIn profiles were associated with higher evaluations (see also Roulin & Levashina, 2019). In practice, this may result in recruiters prioritizing candidates with more complete information (Highhouse & Hause, 1995; Jagacinski, 1991). Therefore, when being screened by potential employers outside formal application processes—such as during active sourcing without a submitted résumé—incomplete LinkedIn information may negatively impact career opportunities. Providing thorough, job-relevant details and a professional picture might help mitigate this risk and prevent recruiters from making negative assumptions about incomplete profiles.
Limitations and Future Research
A limitation of our study is its vignette-based design, where raters evaluated hypothetical scenarios. While this approach enabled a controlled experimental setup to isolate the effects of incomplete LinkedIn and minimized confounding factors related to profile completeness (e.g., the applicant’s conscientiousness; Roulin & Levashina, 2019), it may not fully reflect decision-making and LinkedIn content in real-world HR practice. For example, although recruiters typically compare candidates, our between‐subjects design had raters evaluate each LinkedIn profile individually, yielding absolute rather than comparative ratings. Hence, the applicability of our findings to actual recruitment contexts warrants further investigation. That said, our results align with studies that used other design approaches like interviews (Berkelaar, 2014) and field studies (e.g., Roulin & Levashina, 2019).
Next, we acknowledge that recruiters’ demographics differ between contexts and companies, e.g., regarding gender, LinkedIn use, or prior hiring experience. Thus, we included these characteristics as control variables (see Tables 3 and 6). In line with earlier studies (Mönke et al., 2024a; Roth et al., 2020; Sallach et al., 2024), we found no substantial effects. Beyond that, we did not investigate different genders, ages, and races of applicants; all raters assessed a mid-20s White man. However, applicant pools in practice are much more diverse, and future research should bear this in mind. For example, although adding a LinkedIn profile picture raised hireability ratings in Study 1, it also discloses an applicant’s race and may enable discrimination against minority candidates (Cuddy et al., 2011). Henderson and Welsh (2024) showed that applicant race can influence the impact of incomplete LinkedIn information.
Additionally, the perception of incompleteness may vary based on both the applicant’s and the rater’s cultural or contextual background. For example, as noted by a reviewer, we included information on the job seeker’s high school education in the LinkedIn profiles, but this may be uncommon in some cultural contexts. Future research should investigate how our findings generalize across different settings and industries (Mönke & Schäpers, 2022).
Furthermore, we point out that we used LinkedIn assessments as a proxy for the HR practice of social media assessments, as LinkedIn is the platform most frequently utilized for cybervetting (Smith, 2017). Thereby, we also focused on incomplete job-related information because LinkedIn is a professional platform. However, we recognize that recruiters also often use other platforms—e.g., Facebook—especially to get insights into an applicant’s personality (Hartwell & Campion, 2020; Nikolaou, 2014; Smith, 2017). As the information provided on hedonic platforms like Facebook differs substantially from professional platforms like LinkedIn, future research should address whether the incomplete LinkedIn information paradigm transfers to such platforms as well. For instance, as Facebook contains much more personal information (Zhang et al., 2020), recruiters might have more sympathy with applicants who restrict public access. So, when cybervetting is conducted on hedonic platforms, incomplete information might have a reduced impact during active sourcing, and perceived privacy intentions could play the moderating role described above.
Future research should also explore the extent to which recruiters believe they can make valid KSAO evaluations (e.g., warmth, competence) based on social media profiles. Consistent with prior studies (Berkelaar, 2014; Hartwell & Campion, 2020; Pu et al., 2023), we found that social media postings influence hireability impressions; however, it remains to investigate which specific cues activate such judgments and how incomplete social media information relates to KSAOs. This could help clarify how recruiters weigh different contents of a social media profile and identify biases that may influence hiring decisions.
Conclusion
This study examined the implications of incomplete LinkedIn profiles on hireability perceptions. Our findings revealed that job seekers with incomplete LinkedIn profiles—those providing fewer details—received lower hireability ratings due to reduced perceptions of trustworthiness, professionalism, warmth, and competence. This effect was significant in the context of active sourcing, where candidates were evaluated solely on their LinkedIn profiles. However, we observed no negative effects when the applicant’s complete résumé was presented alongside the LinkedIn profile, with the LinkedIn assessment framed as part of a traditional background check. These results offer a nuanced understanding of how incomplete information influences hireability perceptions and decision-making processes in social media assessments.
Declarations
Ethical Approval
All studies were conducted in compliance with APA standards and the Declaration of Helsinki. Informed consent was obtained from all individual participants included in the study. The Ethics Committee of the Faculty of Psychology and Sports Science at the University of Münster approved our procedure (Decision 2022-41-FM).
Conflict of interest
The authors declare that they have no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We note that the sequence and wording (but not the content) of some hypotheses were adjusted during the revision process to enhance the manuscript’s readability.
Due to a high drop-out rate (~ 50%), we excluded participants from a group tasked with assessing a job seeker without a LinkedIn profile. This condition was represented by a screenshot indicating that no matching profile was found.
If warmth, competence, professionalism, and trustworthiness are treated as mediators within the same step, the effect of the facet variables (professionalism, trust) was masked by the broader dimensions (warmth, competence; see ES09). In this model, only warmth and competence were significant mediators. However, when warmth and competence were removed from the model, trust and professionalism emerged as mediators. In conclusion, we suggest that it is crucial to model the decision process in two steps: First, specific impressions of professionalism and trust, which then lead to overall impressions of warmth and competence. For a similar rationale, see Roth et al., (2017, 2020).
For parsimony, we did not assess the applicant’s trustworthiness and overall hireability: As discussed in Study 1, overall hireability was highly correlated with task performance, and trustworthiness was linked to warmth.
