Cybervetting prospective employees of SMEs

. Cybervetting has become an integral part of HR practices, even though it is generally perceived negatively, as an invasion into employees’ privacy. In response, scholars have begun investigations of the ways social networks impact HRM in organizations. In this paper, we present and discuss the results of our exploration into whether SMEs will be screening job applicants. We have used binary logistic regression to build the model based on the predictors: company's size, HR formality score, company's age and turnover. The units of the analysis in this study are SMEs operating in Czech and Slovak Republics. Random sample technique was applied to select 665 respondents. The resulting model contained only Formality Score (FS) as a predictor with the association to SCR. We can claim that inclination to social networks’ screening in SMEs decreases as FS rises. The study also reveals that social networks screening is widely spread among SMEs in both countries under study. The study extends the research on networks’ screening and expands the knowledge about how SMEs are replacing formalised HR practices. Finally, it offers recommendations for future work that can enhance knowledge in social networks’ impact on SMEs.


INTRODUCTION
Social networking sites (SNSs) have become popular, and today are playing a vital role in our lives (Tulu, 2017).According to Hootsuite (2019), in January 2019, 54% of the total population in the Czech Republic were active social media users, in the Slovak Republic, their share was 50%.Massive adoption of SNSs has fundamentally reshaped the way people make decisions and interact with each other (Qualman, 2010).HR professionals are no exception in this regard.Social networks have long become an essential HR supporting tool (Poba-Nzaou et al., 2016), they help to attract job applicants and gather additional information about job candidates before the interview (Evuleocha & Ugbah, 2018;Krasnova et al., 2010).Information obtained from SNSs may change job candidates' evaluation in a range of rather unexpected ways (Berkelaar, 2017).Social networks may cause bias and subjective judgments and thus may prevent individuals from obtaining a job without knowing why (Black & Johnson, 2012).Since employers do not make cybervetting activities or their outcomes visible, applicants may feel they lack opportunities to explain or correct misunderstandings (Berkelaar, 2014).
Cybervetting is not an unusual activity anymore as it is becoming the norm in HR practices.There is evidence that recruiting via social media is growing with 84% of organizations using it currently and 9% planning to use it.To compare, back in 2011, only 56% used social media for recruitment purposes.Overall, 43% of organizations said they use social media or online search engines to screen job candidates, and this shows an increase from 2013.Over one-third (36%) of organizations have disqualified job candidates in the past year because of concerning information (e.g., illegal activity, the discrepancy with application) due to information they found on a social media profile or through an online search (SHRM, 2016).
However, there is relatively little research on these issues in human resource management literature (Gentina & Chen, 2019; Wang et al., 2018;Won & Seo, 2017).In order to address the mentioned research gaps in the existing literature, this study examines how to predict the probability of whether SMEs will be doing social media screening of job applicants.
The paper is organised as follows.We first discuss literature on social networks and cybervetting in the HRM domain.This is followed by the methodology section, which explains the research context, methods of data collection and data analysis.After that, we discuss the results and their implications for HRM literature and practice.Finally, we conclude with some insights for future research.

Definition of social networks
Researches conducted in the HRM domain mainly utilise SNSs' definitions of Boyed & Ellison (2007) (Black & Johnson, 2012;Brown & Vaughn, 2011;Kluemper et al., 2016;Nayak et al., 2018;Rienties et al., 2010;Suen, 2018) or Kaplan &Haenlein (2010) (Aguado et al., 2016;Arjomandy, 2016;Kluemper et al., 2016;Martin et al., 2015;Poba-Nzaou et al., 2016;Vetráková et al., 2018).Boyed & Ellison (2007) define social network sites as "web-based services that allow individuals to (1) construct a public or semi-public profile within a bounded system, (2) articulate a list of other users with whom they share a connection, and (3) view and traverse their list of connections and those made by others within the system.The nature and nomenclature of these connections may vary from site to site".Kaplan &Haenlein (2010) in their work firstly create a classification scheme by creating three categories for "social presence/media richness" (low, medium, high) and two categories for "self-presentation/self-disclosure" (low, high).Secondly, they define six categories of social media one of which are SNSs (with medium social presence/media richness and high self-presentation/self-disclosure), and others are blogs, virtual social worlds, collaborative projects, content communities, and virtual game worlds.Thirdly, they define SNSs as "applications that enable users to (1) connect by creating personal information profiles, (2) inviting friends and colleagues to have access to those profiles, and (3) sending e-mails and instant messages between each other.These personal profiles can include any type of information, including photos, video, audio files, and blogs".However, we should note that Boyed & Ellison (2007) definition of SNSs was already updated in order to accurately characterizes the today SNSs landscape.According to Boyed & Ellison (2013), SNSs are a networked communication platforms in which participants (1) have uniquely identifiable profiles that consist of user-supplied content, content provided by other users, and/or system-provided data; (2) can publicly articulate connections that can be viewed and traversed by others; and (3) can consume, produce, and/or interact with streams of user-generated content provided by their connections on the site.Also, Kane et al. (2014) entered discussion and state that "SNSs possesses four essential features, such that users (1) have a unique user profile that is constructed by the user, by members of their network, and by the platform; (2) access digital content through, and protect it from, various search mechanisms provided by the platform; (3) can articulate a list of other users with whom they share a relational connection; and (4) view and traverse their connections and those made by others on the platform".Table 1 presents SNSs definitions in terms of their properties important in the context of our research.

Source: as cited and authors processing
Other authors are also involved in discussing the essence of SNSs by highlighting some of their characteristics.For instance, SNSs allow users to create content (Kaplan & Haenlein, 2010); share and communicate content among and across hundreds or thousands of other individuals (Ham et al., 2014); build social relations with others who have similar interests, activities, backgrounds (Berthon et al., 2012;Cheung et al., 2015;Obar & Wildman, 2015).And all of that quickly and with relatively little effort (Mangold & Faulds, 2009), because of their collaborative, flexible, agile, spontaneous, unstructured, and informal nature (McAfee, 2006).
Another SNSs key characteristic is that they can take the form of either personal (or nonprofessional) networks (e.g., Facebook, Instagram) or professional networks that connect applicants with potential employers (e.g., LinkedIn) (Hanna et al., 2011;Hartwell & Campion, 2019;Nikolaou, 2014;Smith & Kidder, 2010;Suen, 2018;Zide et al., 2014).Within personal SNSs, users share substantial amounts of private information, regularly communicate and collaborate (Lewis et al., 2008;Subrahmanyam et al., 2008) and/or use game or applications (Rienties et al., 2010).In contrast, professional SNSs, such as LinkedIn, are primarily used for business purposes (Rienties et al., 2010) and almost exclusively for building professional relations (Zide et al., 2014).The use of professionally and non-professionally oriented SNSs has become widespread (Nikolaou, 2014) due to their increased accessibility through mobile technology (Kaplan and Haenlein, 2010).According to Liu et al. (2016) from SNSs can also be excluded networks for micro-blogging activity (Twitter, Tumblr, Sina-Weibo) because they focus on information dissemination (typically in the form of shorter blogging) (Liu et al., 2016).

Cybervetting in the HRM domain
Cybervetting occurs when employers use nongovernmental, non-institutional online tools or sites (e.g.search engines and SNSs) to extract informal often personal information about prospective or current employees (Berkelaar & Buzzanell, 2014;McDonald et al., 2016).There is a relatively small body of literature that is concerned with cybervetting.Existed works can be divided into two areas, specifically from an employee and an employer perspective (Bartosik-Purgat & Jankowska, 2017).
From an employee perspective, scholars have investigated applicants' reactions to cybervetting.Madera (2012) studied the perceived fairness and job pursuit intentions of college students, who attended a career fair for hospitality positions.The results showed that organizations that used social networking websites as a selection tool were perceived as less fair than those that did not.Similarly, job pursuit intentions were lower for organizations that used social networking websites as a selection tool than organizations that did not.Berkelaar (2014), using semi-structured protocols to interview applicants about their job-seeking approaches, found that applicants felt the lack of opportunities to explain or correct misunderstandings.Although many applicants noted that employers could check an applicant's digital footprint for legitimate needs, most of them typically considered online information as personal; cybervetting deemed as an invasion to personal space and wanted from organizations to be transparent about the process and outcomes.Also, they asked for opportunities for correction.Curran et al. (2014) asked 12 college students to describe what they used SNSs for and what types of information they thought were available to others who looked at their sites.They found that college students are aware that potential employers will use SNSs to gather information about them, but at the same time, they are not overly concern about it.Aguado et al. (2016) study looked at differences in applicants' reaction to SNSs screening depending on whether it is a personal or professional network.They had also evaluated the role of different sociodemographic variables on the applicants' reactions.The results showed that both age and gender play a role in participants' attitudes towards personal networks but not towards professional ones.Contrary to what was hypothesized, older participants showed more positive attitudes towards the use of personal networks than younger ones.Hurrell et al. (2017) explored how Generation Y employees experience and perceive employer use and monitoring of SNSs and whether employment-related concerns influenced their online behaviour.The study pointed out that cybervetting is in generally perceived negatively, such as the invasion into privacy.Nevertheless, many students displayed alertness through actively managing online profiles which, in turn, marginally yet significantly increased their justice perceptions.Suen (2018) examined similar research questions on data from Taiwan Facebook users using Partial Least Square Structural Equation Modelling (PLS-SEM).The results indicated that a candidate who can better control his/her SNSs information is less likely to perceive that his/her privacy has been invaded during SNSs screening by potential employers, thus mitigating his/her perception of procedural unfairness.Moreover, when SNSs screening is more transparent, the candidate is less likely to perceive the selection procedure as unfair, which will reduce his/her intention to withdraw from employment selection.Jeske& Shultz (2019) grappled with the issues of the relationship between perceived respect for privacy and potential job pursuit intentions when it comes to working with vulnerable groups.They found that perceived respect for privacy differed across the jobs.Specifically, they found that applicants were willing to accept the privacy violation if it was connected with recruitment for jobs involving work with children.
From employer perspectives, scholars have investigated the scope of cybervetting use and the prediction ability of published information.De La Llama et al. ( 2012) through 19 in-depth semi-structured interviews with employers belonging to Generation-X found, that thanks to Facebook, they can remove applicants who do not meet the company profile.The employers initially noted that the weight of Facebook is on the latter end of the hiring-decision spectrum, after the results of the interview with candidates and their resumes evaluation.However, the way how candidates portrayed themselves through their wall posts, as well as what they wrote on the walls were also crucial in hiring decision processes.Roulin & Bangerter (2013) turned their attention to the fact whether recruiters and potential applicants (students and graduates) both perceive professional SNSs as a potential antecedent of Person-Job fit information and personal SNSs as a potential antecedent of Person-Organization fit information.Their results suggest that applicants should worry less about smaller faux pas, such as pictures from parties with friends published on personal SNSs and focus more on developing the professional information section in their professional profiles.Nikolaou (2014) explored how HR professionals use SNSs.Specifically, he studied how they use SNSs in the recruitment process, and also their effectivity.Conducted research seems to indicate that HR professionals are more engaged in LinkedIn than Facebook and considered the information from professional SNSs as more effective than from private ones, in the recruitment process.Berkelaar & Buzzanell (2015) qualitative analysis of 45 in-depth employer interviews emphasized how employers evaluate extracted information about applicants during personnel selection.Their work showed that employers disqualified applicants who posted photos of inappropriate behaviour and/or have spelling, grammar issues.The level of used language presumably evidences about applicants' levels of professionalism, attention to detail, and suitability for employment.Results also showed that SNSs are used to assessing the trustworthiness of applicants' character or reputation.Employers also questioned the motivation of candidates who often played games.Machado (2016) acknowledged the fact that the practice of cybervetting is often seen as professional duty and urges greater awareness of the impact that online activity may have on applicants' image.Berkelaar (2017) offers a preliminary framework for understanding how and for what purposes employers using cybervetting in their personnel selection process.Hedenus& Backman (2017) paid special attention to recruiters' expectations and evaluations of confessions regarding information that can be found online about applicants.The presented qualitative analyses of interviews and observations of 36 Swedish hiring managers suggest that jobseekers are either expected to disclose information that can be found online in a proactive manner or to confess when confronted with the findings reactively.Hartwell & Campion (2019) results indicate that SNSs content is considered useful and is regularly utilized during hiring.Some SNSs content is viewed positively (e.g., information supporting qualifications), while other information leads to negative perceptions (e.g., discriminatory comments).Finally, results support differentiation between personal and professional SNSs, as the kind of information sought and the effectiveness of assessing various work-related constructs differs between these SNS categories.
The prediction ability of SNSs information was assessed mainly by the ability to distinguish between those individuals who are high on Big Five personality traits from those who are low on that characteristic; and also, whether this information can be manipulated.Kluemper et al. (2009) support the notion that based solely on viewing social networking profiles, judges are consistent in their ratings across subjects and typically able to accurately distinguish high from low performers.In the next study, they suggest that evaluators trained to assess participant profiles can provide reasonably reliable estimates of Big Five personality traits from SNSs (Kluemper et al., 2012).Back et al. (2010) suggest that people are not using their SNSs profiles to promote an idealized virtual identity.Instead, SNSs might be an effective medium for expressing and communicating real personality.Schroeder & Cavanaugh (2018) indicate that individuals can indeed manipulate Facebook profiles to present themselves in a specific way.Faking ability had an inverse relationship with age and daily Facebook usage, and a positive association with cognitive ability, Facebook impression motivation, self-monitoring ability and sensitivity, and neuroticism.Interestingly, cognitive ability was only a marginally significant predictor of faking ability in the "fake bad" condition, and unexpectedly, this was a negative relationship.

AIM, METHODOLOGY AND DATA
The study aims to predict whether a company will cybervetting job applicants.The units of the analysis in this study are SMEs operating in Czech and Slovak republics.Random sample technique was applied in two public databases ("Albertina" and "Cribis") to select the respondents.An online questionnaire survey was used to obtain data during October 2019.Company owners or persons in charge of recruitment were asked to complete the questionnaire.
There final sample consists of 665 usable questionnaires, where 62.71% per cent (N=417) were completed by Czech SMEs and 37.29 per cent by Slovak.According to their size, 83.46 per cent of them (N=555) belonged to the micro-enterprise category, 15.46 per cent (N=104) were small size, and 0.90 per cent (6) were medium size.For more extensive sample data, see Table 2.

Variables Measuring the company's size (CS) -independent variable
We construct SMEs size within the European Commission recommendation: "The category of micro, small and medium-sized enterprises (SMEs) is made up of enterprises which employ fewer than 250 persons and which have an annual turnover not exceeding EUR 50 million, and/or an annual balance sheet total not exceeding EUR 43 million.Within the SME category, a small enterprise is defined as an enterprise which employs fewer than 50 persons and whose annual turnover and/or annual balance sheet total does not exceed EUR 10 million.Within the SME category, a microenterprise is defined as an enterprise which employs fewer than 10 persons and whose annual turnover and/or annual balance sheet total does not exceed EUR 2 million" (EC, 2003).

Measuring HR formality score (FS) -independent variable
In SMEs, the HR processes may be managed by one person who determines the relevant factors and requirements for the job.(Abraham et al., 2015;Dvorský et al., 2018).However, HR formality is defined as the extent to which HR practices are documented, systemized, and institutionalized (Nguyen & Bryant, 2004).
There are relatively few studies in the area of measuring HR formality in SMEs.Most of published works adopt the 2004 Workplace Employment Relations Survey (WERS) approach (Lai et al., 2016;Saridakis et al., 2013;Storey et al., 2010), which collected data on a wide range of HR practices including recruitment, appraisal, and pay as well as methods of handling communication with employees.To measure degree of formality, WERS collected data about indicators as follows: ( 1 12) Any non-payment benefits (Forth et al., 2006).Nguyen & Bryant (2004) took similar approach when they compiled a list of the following criteria of HR formality: (1) Presence of HR specialist(s); (2) Written criteria for hiring; (3) Written criteria for firing; (4) Professional source of recruitment; (5) Existence of HR plan for the company; (6) Spending on training for employees, (7) Written job descriptions, (8) Written criteria for performance appraisal.Psychogios et al. (2016) measured the degree of HRM formality with questions whether the organization has HRM department or who takes responsibility for HRM, respectively, who is in charge of HRM.
Based on the previous studies, we created a list of 10 items that were relevant to the Czech and Slovak SMEs: (1) HR specialist(s) and presence of written: ( 2 Responses to these questions were measured on a binary scale: "yes" (1) or "no" (0).An overall HR formality score was calculated as an equally weighted average of the items.

Measuring turnover (TO) -independent variable
To measure the company turnover, we asked the respondent to provide the exact amount of turnover reported in the tax return for the year 2018 as sales of goods, products and services.ECB euro reference exchange rate for Czech koruna (CZK) from 31st December 2018 was used to calculate turnover of Czech companies (1EUR=25.724CZK).

Measuring the company's age (CA) -independent variable
The age of the companies was measured in three intervals coded as follows: (1) less than 5 years; (2) from 5 to 10 years; (3) more than 10 years.

SNSs screening (SCR) -dependent variable
The depended variable was coded as a dummy binary variable whose value is 1 if the respondent realised SNSs screening on the last employee selection and 0 if not.

Analysis
We used binary logistic regression to model the relationship between a binary target variable (SCR) and a set of independent variables (CS; FS; TO; CA).Binary logistic regression is a generalised linear model that estimates the probability of a binary outcome based on one or more independent predictors and can be used as a classifier by setting a decision rule.This method offers the advantages of fast training time and interpretability and is suitable for problems where a linear relationship between the predictors and log odds can be assumed (Bigsby et al., 2019).The target level for logistic regression was set to 1 (the respondent realises SNSs screening).SAS JMP version 15 was used to make calculations.

Descriptive statistics
Table 2 presents the sample characteristics and responses to each measure.The results yielded some interesting findings.Firstly, even 65.11 per cent of SMEs (N=433) realised SNSs screening on the last employee.Secondly, the average formality score of SMEs in Czech and Slovak republic is low (Mean=0.19Std Dev=0.13).Further analysis also showed that significant differences in formality score exist according to company size (p-value <.0001) and age (p-value <.0001).Difference between Czech and Slovak companies was not significant (Dif.SK-CZ=0.00693,p-value 0.589).However, these results are out of the scope of this paper.Source: Authors' calculations

Binominal logistic regression
The goal of the building model is to estimate the likelihood that dummy binary variable is either "realise SNSs screening" (1) or "do not" (0) and thus which predictors (CS; FS; TO; CA) lead to SNSs screening of employees (SCR).
To build the model, we started by examining the impact of each predictors separately.Results showed statistically significant associations between the CS and SCR (Whole model test p value <.0001, misclassification rate=0.2316);FS and SCR (p value <.0001, misclassification rate=0.0135);TO and SCR (p value <.0001, misclassification rate=0.2376);CA and SCR (p value <.0001, misclassification rate=0.0361).
In the second step, we tried to create a model that contain all the predictors as effects.The constructed model's lack of fit report showed that the model fits the data well (0.9694) but there were some predictors without a statistically significant association with the response (CS p value=0.8833; CA p value=0.6175; TO p value=0.9698; CA p value=0.3394), so we decided to reduce the model by gradual insignificant predictor deletion.As Table 3 presents, the resulting model contained only FS as predictor with the association to SCR.The misclassification rate of the final model also proved that only 1.35 per cent responses are not the observed category.We also use prediction profiler to see how the crossed effects show as curvature in the prediction traces.It reveals, that inclination to SNSs screening decreases as FS rises.First, organizations that used SNSs during the candidates' selection are perceived as less fair than organizations that did not (Madera, 2012).Furthermore, when SNSs screening is transparent, the candidates are less likely to perceive the selection procedure as unfair, which will reduce their intention to withdraw from employment selection (Suen, 2018).Perceived respect for privacy is positively correlated, and information privacy concern is negatively correlated, with job pursuit intention (Jeske & Shultz, 2019).In this context, it should be stressed that the company's internal potential is most strongly influenced by the potential and commitment of its employees that can be directly influenced by the individual activities of human resources management (Stachová et al., 2019).This is becoming even more critical in small and medium enterprises (SMEs) as they play a vital role in job creation (Koisova et al., 2017;Appiah et al., 2018) and are recognized as one of the main contributors to economic, development and employment growth (Cepel et al., 2018;Kljucnikov et al., 2016;Mura, 2019).Extensive and unregulated use of cybervetting by SMEs can reduce their credibility in potential employees' eyes, which could have, given the important position of SMEs, enormous consequences even though the Czech and Slovak economy still focuses on cheap labour which is not so sensitive about privacy.(Habánik et al., 2019).
Second, our results confirmed a low level of HR formality in SMEs even lower than was unveiled by works of Psychogios et al. (2016) or Saridakis et al. (2013).Therefore, the recruitment and selection stages and processes are not systematic and highly sophisticated instead they are at best, rudimentary and modest and managed by one person who determines the relevant factors and requirements for the job (Abraham et al., 2015).This may have at least two consequences, one for the company itself and second for society in general.This may have at least two consequences, one for the company itself and second for society in general.Such a person is unable to evaluate the findings sufficiently and is very likely to be mistaken in the selection of the employee which can also lead to hardly detectable discrimination regarding race, religion, national origin, age, gender and disability, since, as results showed, SMEs supply the existence of formalised HR practices with cybervetting.
Third, potentially more invasive monitoring of SNSs for current employees (rather than applicants) may occur over issues such as suspected illegal activity, defamation of the organization, or inappropriate conduct (Hurrell et al., 2017).Thus, we would also like to draw attention to ethics and legality of screening via SNSs screening.Specifically, we would like to emphasize employers to avoid bypassing an applicant's privacy settings to obtain additional information about the applicant (Chauhan et al., 2013).

CONCLUSION
The relationship between employers and potential employees is further changed by new practices (Ladkin & Buhalis, 2016).Even if SNSs screening does not replace other e-recruitment tools, many experts consider it more and more important.Nevertheless, the essential nature of SNSs remains for the moment subject to controversy (Girard et al., 2013) mainly because of significant legal, ethical and professional implications.
The study aimed to estimate which changes in predictors (CS; FS; TO; CA) lead to SNSs screening of employees (SCR) in SMEs.We started an examination with the impact of each predictor separately and continued with creating a model including all the predictors.The results showed that although each of the predictors was associated with changes in the response separately, together in one model, they were without a statistically significant association with response.Finally resulting model contained only FS as predictor with the association to SCR, and therefore we can claim that inclination to SNSs screening in SMEs decreases as FS rises.The study also revealed that 65.11 per cent of SMEs realise SNSs screening and as the by-product that the average formality score of SMEs in Czech and Slovak republic is low (Mean=0.19 Std Dev=0.13).
This study makes several contributions to theory and practice in human resource management.First, it extends the research on SNSs screening.Second, it expands the knowledge about how the SMEs replace the existence of formalised HR practices.
There are also some limitations which should be discussed.First, data collected from mainly from micro size SMEs might raise issues of generalizability, but this fully reflects the state of the SMEs in Czech and Slovak Republic with 82 per cent share of micro-size SMEs (Eurostat, 2017).However, what can affect the generalizability to a certain extent is that the sample did not contain the sole traders.Second, the self-reported questionnaire could cause response bias from a misunderstanding of what a proper measurement is to social-desirability bias, where the respondent wants to 'look good' in the survey, even if the survey is anonymous (Rosenman et al., 2011).This fact can be accentuated by the fact that SNSs screening is frequently implemented without legal support, even on the edge of the law.
In future research, it would be useful to address the limitations of this research with enlarging sample to include more small and medium-size companies; the implementation of some experiments in research design such us verification of fictitious job applications evaluation according to different SNSs profiles; and extension of predictors by generational dimension, because of different generations managers can have different levels of trust to applicants.Also, other variables, such as gender, social embeddedness and educational background, should be addressed, because they influence the trust too (Lazányi & Bilan, 2017).A separate field of research, in the context of this issue, should be to verify how the information obtained from the SNS screening is evaluated, which is assessed as positive and which is not, according to job category.

Table 2
The sample characteristics

Table 3
The final binominal logistic regressions