Chapter 3: Research Method
Procedure and participants
This study was constructed in a field setting of social recruiting that includes actual recruiters’ and job seekers’ profiles for various job vacancies on LinkedIn to frame the posited relationships between the independent and dependent constructs – as specified in our conceptual research model (see Figure 1). This research began with a pilot test that consisted of five in-depth interviews with human resource (HR) professionals who have hiring experience with LinkedIn and two focus groups (one with the five HR professionals and one with five actual job seekers using LinkedIn).
The participants in the pilot study are members of the authors’ local and personal networks. These participants were invited by telephone and participated in the study in a conference room in a local HR association in Shanghai, China. The pilot participants (the HR professionals had a mean age of 39 and three were female, whereas the job seekers had a mean age of 41 and two were male) were located in China and worked in different industrial sectors (e.g., consumer goods, consulting, financial services, hi-tech manufacturing, and computer software).
The pilot test suggested that all the paths in our structural model were
significant and recommended that LinkedIn’s self-presentation categories of ‘updated activity on the personal page’, ‘connections’, ‘number of connections’, ‘joined groups’, ‘joined organisations’, and ‘following’ be excluded for this study because this information is seldom noticed by recruiters or used for self-presentation by job seekers and because it is irrelevant to perceived P–J/P–O/P–P fit and hiring
recommendations when recruiters review the profiles of potential candidates for
‧
The survey profiles of job seekers reflect the majority of LinkedIn members:
currently employed individuals who are interested in obtaining information concerning new career possibilities and who are prepared to act upon these
opportunities in this online community (Dekay, 2009). In conducting the research, we joined a professional LinkedIn group that job-seekers and recruiters commonly use to search for jobs and candidates in the HR field. From this group, we connected with five recruiters in charge of hiring full-function HR Managers in the consumer goods, consulting, financial services, high-tech manufacturing, and computer software
industries in China. The mean age of the recruiters was 38, and three participants were female (60%). Within this group, 90% of members are HR managers or superiors and senior recruiters who focus on sourcing candidates.
We selected the recruiters who posted similar job vacancy information (China HR Manager) on the “jobs” discussion board in the group, and then we requested to connect with them (if they were not previously the authors’ connections on LinkedIn) and then invited them to participate in this study. There were five participants who accepted our invitation.
After the recruiters agreed to participate in the study, we instructed them in the survey procedure through a web meeting and then sent a questionnaire to each of them. Each recruiter was asked to randomly and carefully review 20 LinkedIn profiles within both their and the researchers’ connections who meet the basic requirements in terms of relevant experience, educational background, and work location, which is the information that determines recruiters’ initial judgements regarding hiring
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
recommendations (Cole et al., 2007). If the selected profile was from a job seeker who had already been interviewed, then the recruiters were asked to find another profile. The purpose of this step was to eliminate or minimise the possibility that the recruiters’ interview-based impressions of a job seeker would contaminate their evaluation of the job seeker’s self-presentation (Tsai et al., 2011). By the end of this process, the recruiters had reviewed a total of 100 LinkedIn profiles for the five job vacancies.
If the selected candidates were not previously connected with the recruiters, we introduced (a LinkedIn’s networking function) the recruiters to connect with the job seekers because some candidates’ profiles were visible only to their direct
connections on LinkedIn. Thus, the recruiters could evaluate job seekers’ full profiles and self-presentations on LinkedIn after they were “within” connected.
Of the 100 LinkedIn profile owners (see Table 1), 58% were female, and 66%
had a master’s degree or above. A total of 68% of the profile owners had worked for their current employer for more than three years. Furthermore, 12% of the companies were in the consumer goods industry, 10% were in the consulting industry, 15% were in the financial services industry, 21% were in high-tech manufacturing, 22% were in computer software, and 20% were in other industries. In addition, 79% of the
companies had more than 1000 employees.
‧
Table 1. Demographic of the job seekers
Items
Description
Count PercentageHigh-tech manufacturing 21 21%
Computer software 22 22%
To avoid the potential problem of social desirability (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003), the respondents were simply told that the purpose of this study was to identify factors that influenced recruiters’ perceptions when reviewing job seekers’ LinkedIn profiles. When reviewing a job seeker’s LinkedIn profile, the recruiters themselves decided how much time to spend screening the profile. Once they felt that they had sufficient information to form an opinion regarding a job seeker,
‧
they were asked to complete the surveys evaluating the job seeker’s self-presentation information quality (argument quality and source credibility), fit perceptions, and hiring recommendations. With the information openly provided by each job seeker once connected on LinkedIn, recruiters can review the self-presentation categories for each job seeker, which includes (1) portrait, (2) profile summary, (3) experience, (4) volunteer experience and causes, (5) projects, (6) languages, (7) certifications, (8) publications, (9) education, (10) discussion posts and comments, (11)
recommendations, (12) endorsed skills and expertise, (13) interests, and (14) honours and awards. Each self-presentation category is described based on job seeker’s perspective according to LinkedIn:
(1) Portrait: It is a job seeker’s picture or a look shot that allows people to recognize him/her. This means that this little, square image is a job seeker’s first chance to make a good impression on their target audiences.
(2) Profile summary: It is information about a job seeker’s mission,
accomplishments, and goals. This 2,000-character space is where the LinkedIn algorithm searches for key words. Job seekers fill it with information they think their target audience is looking for, and to jazz it up with awards or anything that will make job seekers stand out. Recruiters will want to put in keywords that match those of the candidates they are looking for, and job seekers will want to use terms that their potential employers might be looking for.
(3) Experience: The information is to demonstrate a job seeker’s professional positions and experience, including jobs, volunteer posts, military, board of
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
directors, non-profit, or professional activities. Under experience, job seekers can go beyond their resume by sharing media such as relevant videos, images, presentations or articles quoting them. Also, job seekers can spruce up their profile with visuals, take advantage.
(4) Volunteer experience and causes: Job seekers can share their volunteer experience, the organizations they support, and the causes they care about with their entire network. A LinkedIn survey reports that volunteer experience can give job seekers an edge with recruiters. Recruiters consider volunteer work equally as valuable as paid work experience.
(5) Projects: The section is an ideal place to show off some of job seekers’
actual work, similar to a portfolio of experiences. One of the great ways in which job seekers can communicate to potential employers what the profile owners can do for them is by showing them what the job seekers have done for other similar employers or clients.
(6) Languages: If being able to speak or write in other languages is important in job seekers’ prospective jobs, job seekers can list the language skills along with their proficiency level by adding this section to their LinkedIn profile.
(7) Certifications: Having a certification on job seekers’ profile can be a beacon for opportunity and a powerful way to showcase expertise. LinkedIn allows job seekers to add professional certifications to their profile in one-click. Any certification issuer can easily add to their website.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(8) Publications: The Publications section can describe and provide links to articles, books and e-books with uploaded files. Job seekers can also include links to blog posts, particularly if users have contributed articles on sites other than their own.
(9) Education: Job seekers can add more education to their profile and list details, including their credential's name, major of study and attendance dates.
They can also add extracurricular activities or clubs in which they participated, as well as list any awards and accomplishments they achieved throughout their degree programs. LinkedIn supports multiple entries and allows job seekers to add as many credentials and schools as needed.
(10) Discussion posts and comments: LinkedIn Groups are where individuals with common interests, professions, and university affiliations connect. The group members can create great exposure to their target audiences through posting discussion and comments, which can help job seekers to build visibility and personal brand within the group.
(11) Recommendations: A recommendation is a comment written by a LinkedIn member to recognize or commend a job seeker. Viewers of a
LinkedIn profile often view the recommendations the profiles owner received on her or his profile to see what others have to say about her or his work.
Recruiters searching for new candidates may prefer to work with people who come recommended by someone they know and trust.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(12) Endorsed skills and expertise: Skill and expertise endorsements are a great way to recognize a job seeker’s first-degree connections' skills with one click. They also let the job seekers’ connections validate the strengths found on their own profile. These endorsements are a simple and effective way of building job seekers’ professional brand and engaging their network.
(13) Interests: Job seekers can use this section to incorporate keywords and outline their professional passions. Also, job seekers can add interests that display personality and that take advantage of its search engine optimization potential for recruiters.
(14) Honours and awards: Industry honours and achievements can show off a job seeker’s hard-earned awards, and this section may help a job seeker’s profile stand out from the electronic slush pile. While much of a job seeker’s profile is a subjective characterisation of his/her abilities, honours and Awards provide objective validation for his/her accomplishments.
‧
The measures were adapted primarily from previously validated questionnaires when possible. Minor modifications were made to fit the context of the present study.
All the items used a six-point Likert scale with anchors ranging from strongly
disagree (1) to strongly agree (6). The preliminary instrument was pilot tested using a convenience sample of 30 HR professionals in a LinkedIn group who have experience recruiting in online communities. The results of the pilot test were evaluated using Cronbach’s reliability and factor analysis. Cronbach’s alpha indicator was used to assess the initial reliability of the scales. The standard lower bound for Cronbach’s alpha is 0.6 (Hair, Anderson, Tatham, & Black, 2010). Any items that did not significantly contribute to reliability were eliminated. A factor analysis was then performed to examine whether the items produced the expected number of factors and whether the individual items loaded on the appropriate factor as expected. The
criterion for factor loading suggested by Hair et al. (2010) is greater than 0.5. The measurement was then refined by removing the items that did not load significantly onto the expected constructs. As a result, the Cronbach’s alpha (α) values ranged from 0.77 to 0.97, which indicated a satisfactory level of reliability.
Argument quality
This study measured argument quality for each self-presentation category using three items adopted from Bhattacherjee and Sanford (2006): ‘The information presented by the job seeker on LinkedIn was informative’, ‘The information presented by the job seeker on LinkedIn was valuable’, and ‘The information presented by the job seeker on LinkedIn was persuasive’. The α score for these items ranged from
.87-‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
.99.
Source credibility
The study measured source credibility for each self-presentation category using three items adopted from Bhattacherjee and Sanford (2006): ‘The job seeker presenting the information on LinkedIn was trustworthy’, ‘The job seeker presenting the information on LinkedIn was credible’, and ‘The job seeker presenting the information on LinkedIn appeared to be experienced and professional’. The α score for these items ranged from .94-.99.
Perceived P–J fit
The study measured perceived P–J fit using Kristof-Brown’s (2000) three-item scale: ‘The job seeker fits the demands of the job’, ‘Other employees will think this job seeker is qualified to do this job’, and ‘I am confident that this applicant is qualified for this job’. The α score for these items was .90.
Perceived P–O fit
The study measured perceived P–O fit from Cable and DeRue’s (2002) three-item scale: ‘The things that the job seeker values in life are very similar to the things that the hiring organisation values’, ‘The job seeker’s values match the hiring
organisation's values and culture’, and ‘The hiring organisation's values and culture provide a good fit with the things that the job seeker values in life’. The α score for these items was .94.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Perceived P–P fit
The study measured perceived P–P fit on the basis of Howard and Ferris’
(1996) three-item scale for measuring ‘affect toward applicant’: ‘The job seeker has qualities that I like’, ‘I would like to do something with the job seeker’, and ‘I would like to spend free time with the job seeker’. The α score for these items was .98.
Hiring recommendation
The study adopted three items from Tsai, Chen, and Chiu (2005) to assess recruiters’ intentions in terms of hiring recommendations: ‘I consider the job seeker to be suitable for hiring into the hiring organisation’, ‘The job seeker would have a good future in the hiring organisation’, and ‘The job seeker would perform well for the hiring organisation’. The α score for these items was .91.
‧
To test the hypotheses, the partial least squares (PLS) method was used. PLS is suited for explaining complex relationships, as it avoids two serious problems:
inadmissible solutions and factor indeterminacy (Fornell & Bookstein, 1982).
Moreover, PLS offers the benefit of lower sample size requirements (Chin, Marcolin,
& Newsted, 2003). In the context of this study, PLS was employed to examine the proposed paths from argument quality to P-J fit and P–O fit and the path from source credibility to P–P fit for each of the 14 self-presentation categories. To evaluate convergent validity, three criteria were used. First, the standardised factor loadings were greater than 0.7. Second, the composite reliability (CR) was greater than the cutoff value of 0.7. Third, the average variance extracted (AVE) was greater than the 0.5 threshold (Fornell & Larcker 1981). Each research construct of the 14 structural models conforms to the above three criteria, indicating adequate convergent validity for this exploratory study. To assess discriminant validity, the root square of AVE and all reflective interconstruct correlations were compared (Sánchez-Franco & Roldan, 2005). Because the square root of the AVE was greater than all the interconstruct correlations, this result provides evidence of sufficient discriminant validity.
Because this study collected data from a single respondent regarding each job seeker, common method variance (CMV) might possibly have inflated the
relationships among the variables. To examine this possibility, we first followed Podsakoff et al.’s (2003) approach to examine the CMV using Harman’s single factor test for the 14 models. To complement Harmon’s test, this study conducted an
additional analysis as outlined by Klein, Rai, and Straub (2007) and Liang, Saraf, Hu,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
and Xue (2007). The results demonstrate that the average substantively explained variance of the indicators is between 0.70 and 0.78, whereas the average method-based variance is between 0.016 and 0.029. The ratio of substantive variance to method variance is between 26:1 and 39:1. Second, we adopted Malhotra, Kim, and Patil’s approach (2006) and modelled all items as indicators of a factor representing the common method effect. The results indicated a poor fit with the 14 models. For example, the goodness-of-fit index (GFI) was 0.538 (<0.90), and the root mean square error of approximation (RMSEA) was 0.133 (>0.08). Given the results of both tests, we believe that CMV is not a significant problem in our research.
‧
Two tables were created to help describe the data in this sample. The
correlations among all measured variables appear in Table 2. A summated score was saved for the construct of cognitive response to preserve the multiple aspects of the concept when estimating the 14 complete models. We tested the hypotheses with PLS, and Table 3 presents all of the hypothesised paths.
Table 2. Correlations between variables
Variables Argument