This chapter introduces the design, research’s approach and methodologies used during the investigation process. For that purpose, the research framework, research hypothesis, research procedure, measurement method, data collection, pilot study, validity and reliability and data analysis methods are described.
Research Framework
The model, based on the concepts introduced in the literature review, include different measurement parameters to evaluate the relationship between the independent variables UNGC (X1) and the Local Influences(X2) on the dependent variable CSR (Y1), and the effect of the independent variable CSR (X3) on the dependent variable Business Performance (Y2).
Figure 3.1 Glocal CSR Structural Equation Model
Ho1
Ho2
Ho3
42
Research Hypotheses
The research null hypothesis are based on the research questions described in Chapter 1.
Structural Equation Model hypotheses
Ho1. The variable UNGC (X1) has no significant effects on CSR (Y1).
-Ho1a. The variable UNGC (X1) has no significant effects on CSR (Y1) in Spain.
-Ho1b. The variable UNGC (X1) has no significant effects on CSR (Y1) in Taiwan.
Ho2. The Variable LI(X2) has no significant effects on CSR (Y1).
-Ho2a. The Variable LI(X2) has no significant effects on CSR (Y1) in Spain.
-Ho2b. The Variable LI(X2) has no significant effects on CSR (Y1) in Taiwan.
Ho3. The variable CSR (X3) has no significant effects on BP (Y2).
-Ho3a. The variable CSR (X3) has no significant effects on BP (Y2) in Spain.
-Ho3b. The variable CSR (X3) has no significant effects on BP (Y2) in Taiwan.
T-Test Hypotheses Differences on the Factors between Spain and Taiwan
Ho4: United Nations Global Compact’s Dimension
-Ho4a: Level of the variable GCHR in Spain is equal as in Taiwan.
-Ho4b: Level of the variable GCLR in Spain is equal as in Taiwan.
-Ho4c: Level of the variable GCE in Spain is equal as in Taiwan.
-Ho4d: Level of the variable GCAC in Spain is equal as in Taiwan.
Ho5: Local Influences’ Dimension
-Ho5a: Level of the variable LIE in Spain is equal as in Taiwan.
-Ho5b: Level of the variable LIS in Spain is equal as in Taiwan.
43
-Ho5c: Level of the variable LIMLB in Spain is equal as in Taiwan.
Ho6: Corporate Social Responsibility’s Dimension
-Ho6a: Level of the variable CSRV in Spain is equal as in Taiwan.
-Ho6b: Level of the variable CSRMS in Spain is equal as in Taiwan.
-Ho6c: Level of the variable CSRPS in Spain is equal as in Taiwan.
-Ho6d: Level of the variable CSRREM in Spain is equal as in Taiwan.
-Ho6e: Level of the variable CSRCR in Spain is equal as in Taiwan.
-Ho6f: Level of the variable CSRSM in Spain is equal as in Taiwan.
Ho7: Business Performance’s Dimension
-Ho7a: Level of the variable BPFP in Spain is equal as in Taiwan.
-Ho7b: Level of the variable BPCP in Spain is equal as in Taiwan.
-Ho7c: Level of the variable BPIBP in Spain is equal as in Taiwan.
-Ho7d: Level of the variable BPLGP in Spain is equal as in Taiwan.
T-Test Hypotheses between Different Rank Employees
Ho8: Perceptions between employees are equal as the middle managers:
-Ho8a: Level of the variable GCHR for employees is equal as the middle managers.
-Ho8b: Level of the variable GCLR for employees is equal as the top managers.
-Ho8c: Level of the variable GCE for employees is equal as the top managers.
-Ho8d: Level of the variable GCAC for employees is equal as the middle managers.
-Ho8e: Level of the variable LIE for employees is equal as the middle managers.
-Ho8f: Level of the variable LIS for employees is equal as the middle managers.
44
-Ho8g: Level of the variable LIMLB for employees is equal as the middle managers.
-Ho8h: Level of the variable CSRV for employees is equal as the middle managers.
-Ho8i: Level of the variable CSRMS for employees is equal as the middle managers.
-Ho8j: Level of the variable CSRPS for employees is equal as the middle managers.
-Ho8k: Level of the variable CSRREM for employees is equal as the middle managers.
-Ho8l: Level of the variable CSRCR for employees is equal as the middle managers.
-Ho8m: Level of the variable CSRSM for employees is equal as the middle managers.
-Ho8n: Level of the variable BPFP for employees is equal as the middle managers.
-Ho8o: Level of the variable BPCP for employees is equal as the middle managers.
-Ho8p: Level of the variable BPIBP for employees is equal as the middle managers.
-Ho8q: Level of the variable BPLGP for employees is equal as the middle managers.
Ho9: Perceptions between employees are equal as the top managers:
-Ho9a: Level of the variable GCHR for employees is equal as for top managers.
-Ho9b: Level of the variable GCLR for employees is equal as for top managers.
-Ho9c: Level of the variable GCE for employees is equal as for top managers.
45
-Ho9d: Level of the variable GCAC for employees is equal as for top managers.
-Ho9e: Level of the variable LIE for employees is equal as for top managers.
-Ho9f: Level of the variable LIS for employees is equal as for top managers.
-Ho9g: Level of the variable LIMLB for employees is equal as for top managers.
-Ho9h: Level of the variable CSRV for employees is equal as for top managers.
-Ho9i: Level of the variable CSRMS for employees is equal as for top managers.
-Ho9j: Level of the variable CSRPS for employees is equal as for top managers.
-Ho9k: Level of the variable CSRREM for employees is equal as for top managers.
-Ho9l: Level of the variable CSRCR for employees is equal as for top managers.
-Ho9m: Level of the variable CSRSM for employees is equal as for top managers.
-Ho9n: Level of the variable BPFP for employees is equal as for top managers.
-Ho9o: Level of the variable BPCP for employees is equal as for top managers.
-Ho9p: Level of the variable BPIBP for employees is equal as for top managers.
-Ho9q: Level of the variable BPLGP for employees is equal as for top managers.
Ho10: Perceptions between middle managers are equal as the top managers:
-Ho10a: Level of the variable GCHR for middle managers is equal as for top managers.
-Ho10b: Level of the variable GCLR for middle managers is equal as for top managers.
-Ho10c: Level of the variable GCE for middle managers is equal as for top managers.-
Ho10d: Level of the variable GCAC for middle managers is equal as for top managers.
46
-Ho10e: Level of the variable LIE for middle managers is equal as for top managers.
-Ho10f: Level of the variable LIS for middle managers is equal as for top managers.
-Ho10g: Level of the variable LIMLB for middle managers is equal as for top managers.
-Ho10h: Level of the variable CSRV for middle managers is equal as for top managers.
-Ho10i: Level of the variable CSRMS for middle managers is equal as for top managers.
-Ho10j: Level of the variable CSRPS for middle managers is equal as for top managers.
-Ho10k: Level of the variable CSRREM for middle managers is equal as for top managers.
-Ho10l: Level of the variable CSRCR for middle managers is equal as for top managers.
-Ho10m: Level of the variable CSRSM for middle managers is equal as for top managers.
-Ho10n: Level of the variable BPFP for middle managers is equal as for top managers.
-Ho10o: Level of the variable BPCP for middle managers is equal as for top managers.
-Ho10p: Level of the variable BPIBP for middle managers is equal as for top managers.
-Ho10q: Level of the variable BPLGP for middle managers is equal as for top managers.
47
Research Procedure
The present research has been developed starting from the author’s motivation to investigate within the CSR field (Chapter 1). A detailed and actualized literature review was made (Chapter 2) to develop the GLCSR structural equation model and its questionnaire (Chapter 3). Once developed, samples were collected and the data analyzed (Chapter 4). After analyzed, the results were presented and conclusions explained (Chapter 5).
48
Figure 3.2 Research procedure.
Research motivation
Literature review
Development of the research questions and hypothesis
Adopting the GLCSR as the theoretical framework
Research instrument
Pilot test implementation
Review of instruments
Data gathering
Data analysis
Conclusions and suggestions
Final defense
Thesis submission
49
Research Method
In the present paper, both empirical and non-empirical procedures were conducted. The non-empirical research has been based on literature review, which in addition to the guidance of expertise in the field; they have provided the insight on how to design the research taking into consideration the topics included on it.
The empirical research was conducted according to the literature review and following previous researches and author suggestions. Therefore, the research method selected for the present investigation is based on a quantitative approach.
Using an adapted version of the GCSR questionnaire developed by Nguyen (2014), the general research method consists of a self-report tool, which allows participants to express in a confidential and anonymously way their perceptions about the constructs included in the theoretical framework.
To solve some of the problems that self-report questionnaires may present, we followed the classic recommendations of Podsakoff & Organ (1986), and therefore, statistical and post-hoc analysis were conducted. In addition, to contrast the validity and reliability of the questionnaire, a review of expertise and a pilot test have been conducted before gathering the main sample of the research.
Population
The present research’s population was chosen following the research purposes.
The population is selected from Spain and Taiwan, which in terms of export estimations occupied the 19th and 21st global positions with an exporting volume of
$277,300,000,000 and $262,600,000,000 respectively (CIA, 2015).
Due to the factors included in the theoretical framework (human rights, labour, environmental and anti-corruption values, the pressures from the employees, the media
& lobby groups and other stakeholders, all the factors included in the CSR dimension) Chapter 2 and with the background included in Chapter 1, the targeted population was manufacturing private profit organizations operating in Spain and manufacturing private profit organizations in Taiwan. By selecting organizations from the same sector, we
50
limited potential biases and control possible strange variables detached from the differences in the industry.
Since the governments of Spain and Republic of China Taiwan are in a certain manner, institutionally regulating CSR and implementing CSR policies (UNGC, 2014b).
Thus, the population can be defined as full-time employees, middle managers and top managers/executives of manufacturing private profit organizations from the countries described above.
Sample
The sample of the research is extracted from the population and was collected by convenient sampling method, contacting employees and managers who were willing to cooperate with the research.
Pilot Test Sample
The sample contains middle managers and top managers of manufacturing private profit organizations in Taiwan. 42 questionnaires were personally distributed among the participants. From those 42 questionnaires, 32 (76% from total) were considered valid after their analysis. For sample descriptive analysis, see Chapter IV.
Final Study Sample
With an initial sample of 201 participants including participants from both countries, after their validation, the sample dropped to (N=181). The sample is composed of employees, middle managers and executives of manufacturing companies from Spain and Taiwan.
Spain
In the case of Spain 170, questionnaires were provided to three collaborators (a, b and c). From these 170, only 84 were returned (response rate of 49%), and 17 questionnaires were collected through its online version. However, after the first stage of analysis, only 81 were considered valid, due to statistics problems. Therefore, the Spanish sample contains 81 participants.
51 Taiwan
The Taiwanese sample was collected by distributing 98 questionnaires. From this 98 questionnaires, 81 were fulfilled (83% response rate) while 19 questionnaires were collected through the online version. A total of N=100 samples were used for the main study.
Data Collection
For the pilot test, the data was collected directly from the participants, physical questionnaires were distributed.
The data collection process was done independently from the two targeted populations (Spain & Taiwan) in different stages.
A first stage in which targeted organizations (according to the characteristics defined) were contacted and requested for the permission and assistance to distribute the questionnaires among their employees, middle managers and top managers/executives.
For this stage, a cover letter, an introduction about the National Taiwan Normal University, the questionnaires and references of the researcher were provided to the organizations.
A second stage in which, after receiving the acceptance of the partner organizations, the questionnaires were sent in its paper-based version and in its online version (in the respective language), or handed to the contact person from the organizations.
This method brought unsuccessful results, because of the low response of the organizations to the requests of the researchers and collaborators.
To collect the Taiwanese sample, the researcher contacted employees from multinational manufacturing companies in Taiwan who accepted to distribute the questionnaire.
In the case of Spain, the researcher through 3 different collaborators distributed the questionnaires among their companies respectively. Furthermore, physical questionnaires were given to random employees from outside this companies with very low response rates (70 questionnaires with a response rate of 12.8%).
52
Measurement Method
The measurement instrument Glocal CSR questionnaire utilized to empirically analyze the theoretical framework was adapted and translated from the original questionnaire developed by Nguyen (2014); the property of the questionnaire belongs to the National Taiwan Normal University Graduate Institute of International Human Resource Development. For the purpose of this research, the GLCSR questionnaire items were adapted and translated from the items extracted from the original questionnaire GCSR from Nguyen (2014).
The developed Glocal CSR measurement tool consists of a self-report tool, which allowed participants to express in a confidential and anonymously way their perceptions about the constructs included in the theoretical framework.
With a total of 75 items rated on a 5- point scale ranging from 1 (strongly disagree) to 5 (strongly agree), both the online version and the paper-based questionnaire cover the independent (UNGC-LI) and the dependent variables (CSR-BP) of the theoretical framework. Extra 5 items cover the demographic section.
United Nations Global Compact Factors
The first dimension of the questionnaire focuses on measure the independent variable UNGC including 4 factors (human rights, labour, environmental concern and anti-corruption) contained in 20 Likert scale items. The pilot test confirmed the validity and reliability of the instrument in the Republic of China Twain as had shown in previous research.
Local Influences Factors
The second dimension of the questionnaire was developed for measuring the local influences independent variable; based on 3 factors (employees; stakeholders;
media & lobby groups) 11 Likert scale items were developed following the literature review and adapted to fit the research purposes.
Corporate Social Responsibility Factors
The third dimension of the questionnaire focuses on measure CSR through 6 different factors (vision, values and strategy; management systems, organization and
53
processes; products and services; resource and environmental management;
communication & reporting and stakeholder management) contained in 29 likert scale items. The 29 items, originally adapted from the Fondation Guilé CSR evaluation tool systematic approach included in Nguyen (2014), were adapted and translated to fit the research population and the research purpose.
Business Performance Factors
The fourth dimension of the questionnaire focuses on measure Business Performance of the past financial year through 4 different factors (financial performance, performance from the customer perspective, performance from the internal business processes perspective, and performance from the learning and growth perspective) contained in 15 Likert scale items. The items were adapted and translated from the Nguyen (2014).
Demographic Factors
The last five items of the questionnaire have been developed to categorize the demographic and characteristics of the sample. In addition to the 4 items included in the original questionnaire, an extra item has been added with the purpose to better identify the professional level of each of the participants (employee, middle manager, top management). The item will allow the researchers to compare whether there are significant differences in the self-report results among the different professional levels.
Pilot Study
To ensure the content validity of the items and the reliability of the measurement instrument, a pilot study was conducted on a sample of 32 participants, composed of middle and top managers of multinational private profit organizations operating in Taiwan. The GCSR questionnaire distributed was the Mandarin Chinese version one (See Appendix A).
Validity and Reliability
To ensure the validity and reliability of the research instrument, different statistical analyses are going to be used. KMO and Bartlett’s test for Sphericity, Exploratory Factor Analysis, Confirmatory Factor Analysis and Cronbach’s Alpha are used respectively in both pilot and main studies.
54
Exploratory Factor Analysis (EFA).
To calculate EFA the researcher used computer software SPSS v.15. For KMO, which shows the acceptance of using the factor analysis on the data, values between 0.5 and 1.0 indicate that factor analysis is appropriate (Krizman, 2009). While the Bartlett's Test of Sphericity p<0.05 is considered a good result, which shows whether the correlations between variables occur due to random correlations.
Furthermore, factor loadings are also analyzed to contrast whether the factors within the dimensions are consistent
Confirmatory Factor Analysis (CFA).
For CFA, SmartPLS v.3.0 software was used to analyze and contrast the statistics retrieved from the Glocal CSR model.
Reliability Analysis
Reliability refers to the degree to which a scale consistently reflects the construct it is measuring. To measure the reliability of the research instrument, Cronbach’s alpha analysis, Composite reliability and Average Variance Extracted was calculated after the Exploratory Factor Analysis and Confirmatory Factor Analysis.
Cronbach’s Alpha
To evaluate the reliability of the measurement instrument, Cronbach’s alpha as the most recommended analysis in combination with factor analysis was used.
According to the statistical standards, Cronbach’s alpha between value should be higher than 0.7 (Krizman, 2009).
Composite Reliability
To double check the reliability of the study, the composite reliability analysis was used. For this parameter, values higher than 0.8 are considered valid showing the consistency of the dimensions tested.
55
Descriptive Analysis
Descriptive analyses are conducted to have a clear image of the questionnaire responses and the characteristics of the sample based on the variables age, level of studies, experience in the organizations and their position in the organization.
The descriptive statistics also provide the information to identify general aspects of the dimensions and factors included for analyze.
Structural Equation Modeling (SEM)
Structural Equation Modeling (SEM) s statistical method for determining the relationship between set of interconnected variables.
Partial Least Square
The hypotheses established in the theoretical framework were tested using partial least squares structural equation modeling technique (PLS-SEM). By using SmartPLS software, we estimated standardized regression coefficients (i.e., path coefficients) for the model paths, which can then measure the relationships between the latent variables included in the model, determining the direction and significance of the relationship.
PLS was used to compute path analysis of the Pilot Test sample (See Figure 3.2) and the final study’s sample. Furthermore, we tested the Average Variance Extracted, which according to Hair, Hult, Ringle & Sarstedt (2014) AVE values should be greater than the .50 minimum necessary to justify the use of a construct.
Bootstrapping
Bootstrapping is a nonparametric procedure that can be applied to test whether coefficients such as outer weights, outer loadings and path coefficients are significant by estimating standard errors for the estimates (SmartPLS). In bootstrapping, subsamples are created with observations randomly drawn from the original set of data by assuming the consistency of the sample distribution. Therefore, the estimated coefficient in the PLS-SEM can be tested for their significance (Henseler, Ringle, &
Sinkovics, 2009). Due to the consistency of the surveyed sample bootstrapping will not be counterproductive. Hence, PLS algorithm used the bootstrapping samples to estimate the results, t-values and determine the significance of the PLS-SEM.
56
Thus, in order to evaluate the factor loadings and path coefficient significance of the pilot test and main studies, this technique was applied creating 50 subsamples for obtaining the T-values of the statistics referred above, and therefore obtaining theirs level of significance.
Student T-Test for Independent Samples
For testing the mean differences between the independent samples researched, Pair T-Test analyses were computed.
The t-statistic analysis presents statistical power when the two distributions have similar shapes and the sample data contain no strong outliers.
The t-statistic analyzed using SPSS v.15 software, is calculated using the
The t-statistic analyzed using SPSS v.15 software, is calculated using the