In this chapter, the research structure, research design, definition and measures of the studied variables are described. The research design section includes sampling and data collection method, sample description, and the data analysis procedures to test the hypotheses.
The definition and measures section includes definition and instrument of study variables, the reliability and validation of each scale is reported as well.
Research Framework
Based on the prior literature review, the framework of this study adopted two theories, role theory and task complexity theory to examine the possible antecedents in the research.
As the researcher stated in the previous chapter, the complexity of role expectation and the complexity of task characteristics were the independent variables to test the potential linkage with the e-HR practice adoption. Departmental outcomes were treated as dependent variables.
Moreover, the possible moderating effect of the IT capability was examined as well. E-HR practice adoption is the mediator in this research framework. The research structure of this study is shown in Figure 3.1.
44
Figure 3. 1. Research framework
As discussed in Chapter 2, this study intends to test the hypotheses listed above.
Research Design
Sampling and Data Collection
Quantitative method was implemented in this study by collecting data from multiple channels: campus recruitment activities, online HR forums, in-service master program class, and personal networks. A total of 278 responses were collected. When multiple responses were collected from one company, a representative response was selected based on the following criteria: 1) human resource working experience, 2) seniority in the company, 3) functional content, 4) computer literacy. The reverse coded items were used to screen out invalid responses. As a result, data from 182 sample companies were retained in the data analysis.
45
The HR professionals in the participant samples were dominantly comprised by females, with 75.1%, while the male only accounted for the other 24.9%. Out of the 182 samples, 59.3% of the participants lied in the age segment of 26-35, the others comprised of the rest (40.7%). As for the seniority, 47.3% of the participants lied in the 1 to 5 years experience category, followed by 20.9% of the participants with 6 – 10 years seniority in the same company; the others accounted for the rest (31.8%). As for the human resource working experience, 47.3% of the participants had accumulated 1 – 5 years in HR relevant work, followed by 24.2% of the participants with 6 -10 years in HR relevant work; the others accounted for the rest (28.5%). Generally speaking, the participants are confident about their computer skills; the majority of the participants rated themselves good on the computer skills;
the other 30% rated themselves as medium. No participant rated themselves as poor or very poor in computer skills. As for the functional expertise, the majority of the HR professionals now perform more than functional content, with 14.1% of the HR professionals performing two functions and 38.5% of the HR professionals performing at least three functions. Profile information regarding the research samples are presented in the following tables. Table 3.1 shows the personal information concerning the respondents.
Table 3. 1
Descriptive Information Concerning the HR Respondents
Sample characteristics Frequency Percentage
Gender 1. Male 45 24.9%
46
Sample characteristics Frequency Percentage
6. 66 and above 0 0%
Functional department Recruiting 34 18.70%
Training 12 6.60%
C&B 11 6.00%
Employee relationship 6 3.30%
Performance management 5 2.70%
Organizational design 1 0.50%
Account service 8 4.40%
Other functions 9 4.90%
Two functions 26 14.10%
Multiple functions 70 38.50%
Table 3. 1. (continued)
47
The participants represented different industries. Almost half of the participants (46.7%) were from manufacturing industries. Participants from service industries accounted for 29.7%
of the population. Detailed information was demonstrated in Table 3.2.
Table 3. 2.
Descriptive Information Concerning the Company Participants
Organizational information Frequency Percentage
Manufacturing 85 46.7%
Service 54 29.7%
Others 30 16.5%
manufacturing and service 3 1.6%
missing value 10 5.5%
Measurement
Reliability and Validity Testing For Formative Measurement The usage of e-HR practices
Haines and Lafleur (2008) proposed a comprehensive list of e-HR practices. However, the list had not been validated in Taiwan. This list was first reviewed by experts and followed by a focus group interview with Taiwanese human resource practitioners to validate as the measurement of the actual e-HR usage of the questionnaire. The question items adopted a 5-point Likert scale, representing different levels of organizational automation. The scale ranged from 1 to 5; 1 represents that the organization did not have the practice, 2 the organization adopts the practice manually, 3 the organization adopts office-software, such as word, excel and access, to assist the companies, 4 the organization uses packaged software for specific function, and 5 the organization uses integrated package of software for the e-HR practices.
Some procedures were performed for the usage of e-HR practices prior to questionnaire
48
distribution: three expert reviews on the questionnaire to ensure the content validity and a focus group interview to validate the usage of E-HR practices in Taiwan. The original scales contain 78 items and 9 functions. In addition, the scale was translated into Chinese Mandarin to accommodate the sample in Taiwan. The scale went through focus group interviews by experts to ensure the content validity after translation. After the focus group interview, the scale of health and safety were removed so the industrial differences in the construct would not affect the validity. The original 78 items were incorporated and new items were added into the scale after the focus group interview. A total of 51 items was generated after the focus group interview. The revision of the questionnaire is organized in Appendix A for reference.
After revision, the revised questionnaire went through a one-by-one item review by 4 human resource professionals. Subtle wording was adjusted according to their suggestion.
Exploratory factor analysis was performed to examine the factor structure of the usage of e-HR practices. The questionnaire went through factor analysis four times to ensure the factor structure is valid with no cross loaded items. Items with factor loading below .5 were eliminated to ensure the validity. The final result was organized in Table 3.3.
49
Table 3. 3.
The Exploratory Factor Analysis of the Usage of E-HR Practices in SPSS
New dimension Question items 1 2 3 4 5 6 7
50
New dimension Question items 1 2 3 4 5 6 7
EHR35 .249 .190 .808 .155 .153 .219 .012
EHR34 .284 .219 .783 .124 .101 .117 -.063
EHR33 .142 .192 .637 .068 .323 -.062 .323
EHR37 .205 .342 .577 .004 .151 .065 .232
Compensation and rewards
EHR14 .103 .169 .070 .811 .184 .092 .153
EHR19 .176 .177 .115 .803 .104 .027 .183
EHR9 .103 .120 .118 .784 .113 .169 .010
Employee benefit
EHR2 .120 .129 .278 .244 .698 .127 .031
EHR4 .290 .117 .149 .203 .692 .213 .065
EHR5 .415 .280 .224 -.060 .577 .165 .245
EHR10 .332 .154 .074 .132 .550 .102 .070
Human resource planning
EHR24 .105 .189 .095 .236 .218 .752 .058
EHR25 .342 .117 .123 .050 .243 .657 .161
Internal compensation administrative work
EHR8 .181 .127 .204 .216 .015 .209 .732
EHR7 .174 .291 .018 .353 .327 .025 .575
Table 3. 3. (continued)
51
The Cronbach’s alpha value for each dimension in e-HR practices was calculated to ensure the internal consistency reliability. The information for the Cronbach’s alpha was organized in Table 3.4.
Table 3. 4.
The Cronbach’s Alpha Value for Usage of E-HR Practices in SPSS Dimension of usage of e-HR practices Number
of items formative measurement items for the construct of e-HR practices. The formative constructs do not have composite reliability (CR) or average variance extracted (AVE). Since the scale is formed by different concepts, the internal consistency of the concepts is not believed to be correlated. Therefore, the internal consistency problem does not exist. For the formative measurement, the factor loading at the .5 level indicates that an indicator is relevant for the construction of the formative concept (Chin, 1998). The dimension of internal compensation administrative work was eliminated since the dimension did not generate sufficient level of t-value to prove validity on the scale. The factor loadings and t-values for e-HR practices are presented in Table 3. 5.
52
Table 3. 5.
The Factor Loadings and T-values for E-HR Practices in SmartPLS
Construct Loadinga T-valuea
Performance management 0.83 (.83) 11.81 (12.03)
Training and development 0.66 (.66) 7.61 (7.71)
Recruiting and selection 0.50 (.50) 4.04 (4.00)
Compensation and rewards 0.51 (.51) 3.87 (3.89)
Employee benefits 0.62 (.62) 6.33 (6.36)
Human resource planning 0.85 (.85) 11.66 (12.35)
Internal compensation administrative work 0.01 0.02b
Note: a Numbers in the parentheses are the values generated after internal compensation administrative work is deleted.
b This dimension is deleted since the t-value does not prove to be significant.
Collinearity diagnostics were calculated to examine if an indicator’s variance is explained by the other indicators of the construct. All dimensions of e-HR practices were under 10 (Cassel et al., 1999; Fornell & Bookstein, 1982) and 5 (Hair et al., 2011) and met the criteria. The information for the variance inflation factor for the usage of e-HR practices is organized in Table 3. 6.
Table 3. 6.
The Variance Inflation Factor (VIF) for the Usage of E-HR Practices
Construct VIF
E-HR a (performance management) 2.045
E-HR b (Training and development) 2.464
E-HR c (Recruiting and selection) 1.989
E-HR d (Compensation and rewards) 2.150
E-HR e (Employee benefits) 1.895
E-HR f (Human resource planning) 1.597
The correlation of the usage of e-HR practices and the other constructs are all below .7 (MacKenzie et al., 2005). As a result, the constructs were regarded sufficiently different from
53
the others. Therefore, the discriminant validity is proven. The information for the discriminant validity between the usage of e-HR practices and the other constructs is organized in Table 3. 7.
Table 3. 7.
The Discriminant Validity between the Usage of E-HR Practices and the Other Constructs
Construct 1 2 3 4 5 6
1. Competence 1.00
2. E-HR 0.45 1.00
3. Information
technology 0.46 0.37 1.00
4. Role complexity 0.72 0.41 0.36 1.00
5. Strategic focus 0.60 0.35 0.50 0.49 1.00
6. Task complexity -0.01 0.00 -0.04 0.03 -0.04 1.00
IT capability
This study adopted the task-technology fit measure developed by Goodhue and Thompson (1995) as the measure of IT capability. Only Ease of use/ Training, system reliability and relationship with users were retained in the research to serve as the indicator to assess the IT capability or the support from IT capability; the other factors were not appropriate for the context of the research. The three dimensions are best to measure the information technology, since the three constructs measure different aspects of IT capability.
“System reliability” measures if the information system is stable and provides adequate dependability of access for the participants; three items were used to assess the system reliability. Sample items include “I can count on the system to be ‘up’ and available when I need it” and “The computer systems I use are subject to unexpected or inconvenient down times which makes it harder to do my work”. “Ease of use” measures if the users perceive that it is easy to do what the employees want to do with the system hardware or software for
54
accessing and analyzing data. Four question items were used to assess the dimension; sample items include “It is easy to learn how to use the computer systems I need” and “The computer systems I use are convenient and easy to use”. “Relationship with users” emphasizes on the relationship of the information technology department with the other functions. Ten question items were used to assess the dimension; sample items include “My work group feels that IS personnel can communicate with us in familiar business terms that are consistent” and “IS takes my business group's business problems seriously.” The IT capability is measured by a 7-point Likert scale; the higher the score is, the higher the IT capability of the organization is.
After the internal consistency testing, the dimension of “ease of use” did not meet the .7 threshold for Cronbach’s alpha (Nunnally & Bernstein, 1994). Therefore, two items were eliminated, since the two items conveyed a different concept, that is, training instead of ease of use. After the elimination, the Cronbach’s alpha value soared and exceeded .7. The Cronbach’s alpha for the IT capability is listed in Table 3. 8.
Table 3. 8.
The Cronbach’s Alpha for the IT Capability Dimension of IT capacity Number
Note: Numbers in the parentheses are the values generated after item deletion.
For the formative measurement, the factor loading at the .5 level indicates that an indicator is relevant for the construction of the formative concept. The t-values of all the dimensions are over 2.54. The factor loading and relevant information is listed in Table 3.9.
55
Table 3. 9.
The Factor Loadings and T-values of the IT Capability in SmartPLS
Construct Loading T-value (as in outer loading)
Ease of use 0.70 12.35
System reliability 0.63 4.79
Relationship with users 0.97 4.11
Collinearity diagnostics were calculated to examine if an indicator’s variance can be explained by the other indicators of the construct. All dimensions of IT capability were under 10 (Cassel et al., 1999; Fornell & Bookstein, 1982) and 5 (Hair et al., 2011) and met the criteria. The VIFs are organized in Table 3. 10.
Table 3. 10
The Variance Inflation Factor for IT Capability
Construct VIF
Ease of use
1.690 System reliability
1.688 Relationship with users
1.963
Discriminant validity is tested by examining correlation between the formative construct and the other constructs. If the correlation is less than .7, the construct shall be regarded as sufficiently different from the others. The correlation of IT capability and the other construct are all below .7. The analysis is presented in Table 3. 7.
56
Reliability and Validity Testing For Reflective Measurement Perceived role complexity
The complexity of perceived role expectation was measured by adopting integrated HR role model proposed by Ulrich in Human Resource Champions: the Next Agenda for Adding Value and Delivering Results (1997). The four roles represent what HR professionals need to perform: strategic partner, change agent, employee champion and administrative expert. The sample question to assess strategic partner is “HR participate in the process of defining business strategies”; the sample question to assess change agent is “HR participate in shaping culture change for renewal”; the sample question to assess employee champion is “HR participate in improving employee commitment”; the sample question to assess administrative expert is “HR participate in delivering HR processes”.
The perceived role complexity was measured by the number of roles that the HR department was expected to perform. Each role was assessed by ten items. A total of 40 items were used to assess the HR role expectation. The overall Cronbach’s alpha value was .982;
the Cronbach’s alpha values for each HR roles were respectively: strategic partner .948, administrative expert .921, employee champion .953 and change agent .955. The Cronbach’s alpha for the HR role expectation is organized in Table 3. 11.
Table 3. 11.
The Cronbach’s Alpha for HR Role Expectation Dimension of perceived HR role
57
The score in each category would reflect the perceived level of expectation in each of the roles and the complexity of role expectation perceived by human resource professionals.
The mean score in each HR role categories was considered; those scores which exceeded 4 were considered to bear role expectation. Therefore, the perceived role complexity was assessed by counting the number of roles which had a mean score above 4. The scores for this variable would range from 0 to 4, with 0 representing no positive expectation on any of the four roles, thus zero role complexity, and 4 representing positive expectation on all four roles, thus very high on role complexity.
The values of the role complexity for the internal consistency reliability (Composite reliability, CR), indicator reliability (indicator loading) and convergent reliability (Average variance extracted, AVE) are organized in Table 3. 12. The composite reliability exceeded .7;
indicator loading exceeded .7 (Chin, 1998) and AVE exceeded .5 (Fornell & Larcker, 1981) respectively. The overview of factor loading, composite reliability for role complexity is listed in Table 3. 12.
Table 3. 12.
The Overview of Factor Loading, Composite Reliability for Role Complexity
Construct Item Loading Composite
Discriminant validity test were performed to ensure the construct validity with its items:
cross loading comparison and Fornell-Larcker criterion (Fornell & Larcker, 1981) were performed. The smallest factor loading scores of the role complexity outweighed the largest score in other constructs; the discriminant validity was ensured. The cross loadings among all the construct comparison is organized in Table 3. 13.
58
Table 3. 13.
The Cross Loadings among All Constructs Construct
59
To ensure discriminant validity, the square root AVE of each latent variable shall be greater than the other latent variable’s highest correlation. The square root AVE of role complexity outweighed the correlation in other constructs. Therefore, the discriminant validity is also ensured in the method. The Fornell-Larcker criterion (Fornell & Larcker, 1981) among all the construct comparison is organized in Table 3. 14.
Table 3. 14.
The Overview of Composite Reliability, AVE and Discriminant Validity Testing among All Variables
Note: 1. The Composite Reliability (CR) are all above 0.7
2. The Average variance extracted (AVE) are all above 0.5 3. The square root of average AVE is in parentheses.
Perceived task complexity
This study adopted Withey, Daft and Cooper’s (1983) scale, which was based on Perrow’s (1967) dimensions of work unit technology. The dimensions were task analyzability and repetitiveness (Number of exception). Task analyzability refers to the way that the individuals respond and deal with the problems in the task completion process. Number of exception refers to the task variety, which is the frequency of unexpected events that could occur during the task completion process. Each measurement contained 5 items to assess the two dimensions. Sample question for the analyzability include “To what extent is there an
60
understandable sequence of steps that can be followed in doing your work?” and “To what extent is there a clearly known way to do the major types of work you normally encounter?”.
For repetitiveness, sample questions include “People in this unit do about the same job in the same way most of the time” and “How many of these tasks are the same from day-to-day?”.
The overall Cronbach’s alpha for the task complexities were .856, with .869 for repetitiveness and .903 for analyzability respectively. The values of Cronbach’s alpha for the task complexity is organized in Table 3. 15.
Table 3. 15.
The Cronbach’s Alpha for Task Complexity Dimension of perceived task scored high (above 4) on both task repetitiveness and analyzability dimensions, that case was coded 1 for low task complexity. On the other hand, if the case scored low (below 4) on both task repetitiveness and low analyzability, that case was coded 3 for high task complexity. All the other cases were coded as two for medium task complexity.
The values of the task complexity for the internal consistency reliability (Composite reliability, CR), indicator reliability (indicator loading) and convergent reliability (Average variance extracted, AVE) are organized in Table 3. 16. The composite reliability exceeded .7;
indicator loading exceeded .7 (Chin, 1998) and AVE exceeded .5 (Fornell & Larcker, 1981) respectively.
61
Table 3. 16.
The Overview of Factor Loading, Composite Reliability for Task Complexity
Construct Item Loading Composite
Discriminant validity test were performed to ensure the construct validity with its items:
cross loading comparison and Fornell-Larcker criterion (Fornell & Larcker, 1981) were performed. The smallest factor loading scores of task complexity outweighed the largest score in other constructs; the discriminant validity was ensured. The cross loadings among all the construct comparison is organized in Table 3. 13.
The square root AVE of each latent variable shall be greater than the other latent variable’s highest correlation. The square root AVE of task complexity outweighed the correlation in other constructs. Therefore, the discriminant validity is also ensured in the method. The Fornell-Larcker criterion (Fornell & Larcker, 1981) among all the construct comparison is organized in Table 3. 14.
Strategic focus
So far, assessment of strategic focus does not exist for the human resource department.
The existing measurement is mainly concerned with organizational strategic focus, and consequently not suitable for the study. Therefore, this study applied the strategic partner measurement in Ulrich’s (1997) questionnaire to assess the participants’ perception toward the human resource strategic focus.
Strategic focus was measured by a 7-point Likert scale; 10 items were used to evaluate
62
the unidimensional variable. Sample questions include “We spend most of our time (or resources) on accomplishing business goals” and “We spend most of our time (or resources) to make sure HR strategies are aligned with business strategy”. The higher the score is, the more strategic focused the HR department is. The value of Cronbach’s alpha for strategic focus is organized in Table 3. 17.
Table 3. 17.
The Cronbach’s Alpha for the Strategic Focus Dimension of strategic focus Number
of items
Cronbach’s Alpha
Mean SD
Strategic focus 10 .936 4.75 1.00
After running the SmratPLS, the factor loading of strategic focus item 4 did not pass .7 (Chin, 1998). Hence, item 4 was eliminated. After elimination, the factor loading complied with the standard of .7. The information is organized in Table 3. 18.
Table 3. 18.
The Overview of Factor Loading, Composite Reliability for Strategic Focus in SmartPLS
Construct Item Loading Composite
Note: The numbers in the parentheses are the result of the second analysis without the strategic focus
Note: The numbers in the parentheses are the result of the second analysis without the strategic focus