This chapter outlines the research design and methodology. In this chapter we highlights the theory that supports the research and presents the research hypotheses derived from the research questions. This study will use quantitative approach as a means to measure the effect of independent variables towards dependent variables. It also provides a detailed description of the research questionnaire, research procedure as well as the approach for validation of the measurement instrument.
Research Framework
The research framework GCSR (see figure 3.1.) was developed by Shih and Nguyen (2013) relying on the review of the literature discussed in chapter 2 and the research purposes. This research seeks to analyze the effect of UNGC on CSR and business performance. United Nations principles in UNGC have done the most to promote and help companies to encompass CSR and sustainability topic in companies, therefore the researcher assume that there is relationship between UNGC and CSR. Also theoretical and empirical studies suggested that CSR affect business performance (Barnett & Salomon, 2006;
McWilliams & Siegel, 2001; Waddock & Graves, 1997; Wood, 2010). In addition, there is a common consensus among scholars and researchers that CSR have effect on financial performance but as discussed for this work, other aspect besides financial will serves as the research framework for this study.
CSR Vision, values and strategy
Management systems, organization and processes Products and services Performance from financial perspective Performance from customer perspective Performance from internal business process perspective
Performance from learning and growth perspective
Figure 3.1. GCSR model, developed by Cheng-Ping Shih and Linh Nguyen H1
H2
28
Research Hypotheses
On the bases of the research questions, the literature review, the research framework and Baron and Kenny (1986) guidelines for mediation testing, the research null-hypotheses are formulated as follow:
Hypothesis1: Global Compact have no effect on CSR.
Hypothesis2: CSR have no effect on Business Performance.
Research Procedure
A review of the literature on CSR revealed several key topics commonly linked to UNGC and business performance. Further literature review has been carried out on the basis of the interests generated from these variables to determine the nature of their connections.
The literature also revealed that many studies related to CSR and performance has been conducted on differential sector and countries, this has prompted this research to focus on multinational companies in Taiwan as an area for investigation. Even though that the topic was previously well researched; not all the constructs have been identified and described and theory have been selected from the literature and practice in order to build a good theoretical model. Moreover, the relationships among the constructs were also addressed so as to form a clear idea about the research objectives and frame the questions for the study and provide a comprehensive research framework. For order of research procedure see Figure 3.2.
29 Identify Research Questions
and Hypothesis Review of Literature
Developing Theoretical Framework Research Motivation
Developing of Instrument Identify Problems
Developing Research Method
Translation and Expert Review of instrument
Proposal Meeting
Conduct Pilot Study
Review of Instrument
Data Collection
Data Analysis Data Coding
Report Completion Conclusion and Suggestions
Final Defense
Thesis Submission Revision
Figure 3.2 Research procedure.
30
Research Method
This research used a quantitative case study and this method was selected for several reasons. Firstly, with quantitative method the researcher can classify, count and construct a statistical model to explain what is observed (Neill, 2007). This research will use modified statistical model to explain the way all the variables interact. Secondly, to explain process through variables interact with each other; precise measurements are necessary for validity of research. With quantitative approach researcher can objectively seek precise measurement and analysis by using questionnaires and surveys. The more objective research will be, the more accurate results research will get (Neill, 2007).
Survey was employed for the following reasons. First, surveys can be readily analyzed and interpreted, because they are structured around key items and topics. Second, they ensure confidentiality for the respondents who may be more inclined to provide honest feedback if their identity remains undisclosed (Debowski, 2007, p. 278). Since the researcher was interested in determining the factors affecting employees’ tacit and explicit CSR behavior, self-reported measures re used. Behaviors and attitudes are determined by the individuals’ perceptions of reality rather than by objective reality (Rentsch, 1990), self-reporting quantitative surveys allowed respondents to report their own perceptions of reality.
Self-reported measures are considered as the most suitable methodology for the study of individuals’ perception (Howard, 1994; Spector, 2011). This quantitative research will focus on Taiwanese multinational companies, whilst the results cannot be generalized; they can lead to “thick description” or “deep data” that can be basis for new theory development.
Participant
The research is aimed at profit multinational companies with established business practices in Taiwan and also in the world. As multinational company this research describe a corporation that has its facilities and other assets in at least one country other than its home country. The target population for this study are employees of multinational companies in Taiwan. Taiwan is a suitable choice for this research because of recent developments, but still in some ways CSR still lacks. Louis Chen, deputy chairman of the Taiwan Corporate Governance Association, says that according to the Foundation of Taiwan Industry Service, only 8 percent of Taiwan's publicly listed companies issue CSR reports, far behind the rate in Japan (35%) and China (33%). This population was chosen because there is great opportunity to focus on differentiation and processes occur within them. Therefore it is a good testing
31
opportunity for this proposed framework. Companies of larger scale like multinational ones have greater capability of the resources required regarding society environment, and environmental communication capacities (Waddock & Graves, 1997). Small companies do less CSR related activities then larger companies while the community also has higher expectations of social responsibilities to larger companies. Larger companies are more mature and attract the attention of the public more easily, so they respond more to the needs of public interest of stakeholders.
Population
The population for this study consist full-time employees, mostly managers working for multinational companies in Taiwan. Employees were randomly selected and contacted through email and personal contact. All participants are employees of multinational companies. Employees with self-evaluation can provide a more realistic account of the organization they belong to. The selection of the research subjects seeks to include different types of industries. The sample size chosen for the pilot study was 40 randomly chosen employees. Twenty five participants per group should be probably be considered the lower threshold of sample size for aims related to instrumentation, although 35-40 per group would be preferable if estimating test-retest reliability (Hertzog, von Oertzen, Ghisletta, &
Lindenberger, 2008, p. 190). For the main study the researcher sent 200 questionnaires for 200 randomly selected employees. Paper questionnaire was sent to 60 participants and the rest 140 participants filled the interned based questionnaire.
Instrument Development
The instrument is largely adaptation and modification of different instruments. For this research, instrument consists of 3 dimensions with total amount of 69 questions. Statistical analysis was initially conducted after the pilot test to prove the validity of the instrument. The number of questions was reduced after the pilot test.
The questionnaire is divided into four distinct part; which are: UNGC, CSR, business performance; and demographics. For parts UNGC, CSR and business performance respondents were asked to rate each item that describes identified dimensions using a 5-point Likert scale with scale anchors ranging from “strongly disagree” (1) to “strongly agree” (5).
For the last part, respondents were asked to choose one of different available options. A 5-point Likert scale was chosen because without neutral mid5-point (with 6-5-point scale or
8-32
point) respondent should choose more positive or negative response and (Gwinner & Bennett, 2008) point out that in many cases respondents will accent the negative in experience.
Because for purpose of research, researcher is not interested in neutral opinion, even though is legitimate, this research didn’t include it to the scale. With the 5-point scale this research can still have a midpoint where with calculated mean weighted average above 3.4 is above neutral and 2.8 is below(Gwinner & Bennett, 2008).
First part UNGC, consist of 4 factors, in total 20 questions were adopted from The Global Compact Self Assessment Tool, the shared property of the Danish Institute for Human Rights, the Confederation of Danish Industries, the Ministry of Economic and Business Affairs and the Danish Investment Fund for Developing Countries. The content of the human rights and labour sections of the Global Compact Self Assessment Tool originate from the Human Rights Compliance Assessment (HRCA) and its sole property of the Danish Institute for Human Rights. The questionnaire were modified and adopted for this research, because there is no exact valid questionnaire that uses the same model like this study, therefore items are adopted from The Global Compact Self Assessment Tool for the purpose of this study.
The validity and reliability of the items were tested in the pilot test.
Second part CSR, consist of 6 items, in total 30 questions. The questionnaire was adopted and modified assessment tool for evaluation of corporate responsibility from Fondation Guilé. As mentioned in review part with complexity of CSR construct researchers cannot have consistent results of CSR (Entine, 2003). Adequate measure of corporate social performance is a challenging task and there are many different measures and tools for CSR.
This research is using Fondation Guilé approach where they are using analysis of annual and sustainability reports and other information concerning the integration of corporate responsibility principles and practices published by the company. This measure has also two systematic assessment aspects:
A) Normative approach
Rating of depth of the CSR criteria consideration in corporate reporting
B) Systemic approach
Rating of the consideration of corporate responsibility criteria in Vision, value and strategy
Management systems, organisation und processes Products and services
33 Resource and environment management Stakeholder management
Communication and reporting (Fondation Guilé; T. Streiff, 2013)
Lastly the content of the Foundation’s programme is guided by the principles of the UN Global Compact. The assessment focus on the four topical areas of the UNGC - human rights, labour standards, the environment and anti-corruption - and are based on publicly available information only. Fondation Guilé analyze and benchmark the annual Communication on Progress (COP) and the sustainability reporting of all portfolio companies. CSR part of questionnaire is modified and adopted for this research, because there is no exact valid questionnaire that uses the same model like this study. The validity and reliability of the items were tested in the pilot test.
Third part, business performance, the 4 items are adopted from Balanced Scorecard (BSC) and Chen & Fong (2012). The business performance construct was adopted from scales developed by a construction-specific study (Chen & Mohamed, 2008). These scales adopted the BSC (Kaplan & Norton, 1996) as a measurement framework, since the BSC has been used by firms for strategic management, and in particular for performance evaluation (Robinson, Anumba, Carrillo, & Al-Ghassani, 2005). Questionnaire scales measure performance from the four perspectives of the BSC, namely financial performance (PF), performance from the customer perspective (PC), performance from the internal processes perspective (PP), and performance from the learning and growth perspective (PL). According to Chen & Fong (2012) the fully subjective self-report measures allow researchers to address latent performance constructs directly and relative measures have been used by many empirical studies ( Yan Ling, Hao Zhao, & Baron, 2007). The validity and reliability of the 4 items was also sufficiently established in the authors’ original work and been undergone various validation from prior studies.
Validity and Reliability of Instrument Face Validity
The reliability of questionnaire created from these scales needs to be reassessed because items are from mutually independent studies and samples. The reliability of these scales were reexamined in a pilot study. Reliability of instrument is the external and internal
34
consistency of measurement. On the other hand, validity of instrument is the degree to which scales measure what researchers claim they measure (Williams & Monge, 2001).
For the convenience of respondents, the questionnaire was available in English and Chinese Language. Originally the instrument was constructed in English. It was translated from English to Chinese language by the Chinese speaking researchers. Two bilingual experts, English – Chinese Language translators, then revised the translation. The questionnaire was also undergoing a back translation to ensure the meaning of each items will be preserved.
Two Chinese academic professionals in Human Resource department again reviewed the questionnaire.
Construct Validity
Before construct validity can be established, researcher needs to run two analyses and obtain required values. First is the Kaiser-Meyer-Olkin (KMO) Measure of Sampling Adequacy value. This value shows us appropriateness of using factor analysis on data. The higher is KMO to 1.0 the better the factors extracted from the data will account for the variance in the data (Friel. 2010). Second is the Bartlett’s test for Sphericity, this value shows whether or not the correlations between variables happened by chance. The smaller the p-value is, the less likely will correlations between variables happened by chance. This research is using convergent and divergent validity. First is the EFA (factor loadings), for the convergent validity. This tests if the factors of each scale are consistent with those of previous studies. The total score of the dimension is used, if the item-to-total correlations are less than .40 the item is then dropped from further analysis (Kerlinger, 1986). The divergent validity is tested by CFA. SEM software was used for performing CFA. In this study CFA was obtained from SmartPLS software.
35 Table 3.1
Reliability Types
Reliability Type
Theoretical Meaning How This Study Achieves Reliability Reliability
Analysis The degree to which a scale consistently reflects the construct it is measuring (Field, 2007)
Cronbach's alpha will be computed for each of the set variables measuring the same dimension to provide evidence for reliability. Values above 0.70 are generally considered acceptable number (Nunnaly, 1978).
Individual item reliability (i.e. factor loading) is also computed. Factor analysis will be done to obtain factor loadings of the questions. If the factor loadings are below 0.5 then they will be dropped from further analysis. Factor loadings show discriminant validity because factor loadings less than 0.5 indicate that these questionnaire items do not belong to a particular factor. (Hair, Anderson & Tatham, 2006)
Average variance extracted (AVE) will be also computed. A cut-off value above or equal to .50 or higher is considered acceptable (Al-Busaidi, Olfman, Ryan, & Leroy, 2010).
36 Table 3.2
Validity of Instrument Validity
Type
Theoretical Meaning How This Study Achieves Validity Face
Validity
A tool where validity is achieved by “looks valid” to people who use it (Schiavetti, Metz, &
Orlikoff, 2011).
The questionnaire will be examined and translated by group of native speakers.
Two bilingual experts and two Chinese speaking academic professionals will examine face validity.
Content Validity
It is a type of measurement validity that requires that a measure represent all aspects of the conceptual definition of a construct (Neuman, 2011).
Questionnaire was examined by two peer reviews and expert reviews. Every questionnaire items refer back to the definition.
Criterion Validity
It is a type of measurement validity that relies on some independent, outside verification (Neuman, 2011).
The questionnaire will initially undergo pilot testing. The questionnaire items are only adopted from previously used measures.
Construct Validity
This refers to whether a scale or test measures the construct adequately (Shuttleworth, 2009)
For this validity this research use
convergent and divergent validity methods.
Convergent validity is established by using Explanatory Factor Analysis (EFA). Before the EFA test, KMO and Bartlett’s test for Sphericity are conducted. If the items are confirmed then EFA can be conducted.
Divergent validity is established by using Confirmatory Factor Analysis (CFA).
37
Descriptive Statistics
Descriptive statistics in this research was used to describe the main features of a collection of data quantitatively without employing a probabilistic formulation (Mann, 1995).
This research used frequencies and percentages to compare the sample and population according to the categorical variables age, educational background, length of experience and type of industry employees are working for. Descriptive will also helps to identify the main aspects of each variable that seems to be most influential within the population under study.
The main used descriptive statistics will be mean and standard deviation. These will be used in pilot study and main study.
Statistical Analysis
For this research SPSS 21 was used for descriptive statistics. Validity and reliability test was carried out to ensure the accuracy of the measurements. Reliability tests were conducted to evaluate the consistency of the participants’ answers to the items in the questionnaire. Reliability tests included Cronbach’s alpha item if deleted, Cronbach’s alpha coefficient, and Corrected item-total correlation. The data were then further analyzed by evaluating the coefficient of determination (R2), path coefficients, and t-value (t). These analyses were conducted by using SmartPLS 2.0 software.
Correlation Analysis
A correlation analysis was used to investigate the direction and strength of linear relationship between variables. With compute method-using SPSS test a statistical value of VIF (Variance Inflation Factor) was obtained to detect whether multicollinearity is a problem in the study. In the finding, the researcher can conclude the relationship between dependent variable is positively or negatively related to independent variable. The researcher have to point out that a high correlation coefficient may be not necessarily implied multicollinearity existed between the variables. If the data will be confirmed by CFA, the constructs investigated in the study will be also confirmed as distinct constructs. Therefore, if CFA is confirmed and the correlation coefficient value is still high, it means that the variables are highly related, and the value of regression can be accepted.
38
Coefficient of Determination (R2)
Coefficient of determination or regression will be used to explain the endogenous latent variables to the total variance. The higher the percentage, the data are closer to explain and to predict in reality. The structural model will be evaluated by examining the value of R2 of the latent variables, the path coefficients and the goodness of fit.
Structural Equation Modeling (SEM)
Structural Equation Modeling (SEM) s statistical method for determining the relationship between set of interconnected variables. SEM is also called multivariate or multi-equation method because is not like the traditional multivariate linear models, the response variable is one regression equation in an SEM can appear as a predictor in another equation (Fox & Weisberg, 2002). This research will use SEM method because is usually used to determine if a given model can be applied to explain a given set of data or if a model is applicable to a given situation. That is why this technique is suitable for testing proposed research framework and test the validity. This research will run SEM for two reasons. First purpose is to perform CFA. Second is to perform causal modeling, or path analysis, that hypothesizes causal relationships among variables and tests the causal models with a linear equation system. Path coefficients are used to judge the relationship between variables and to determine the direction of the relationships and its significance. SEM will be used to find evidence to support or reject the hypothesis proposed in the beginning.
Partial Least Square (PLS)
PLS uses a combination of principal component analysis, path analysis and a set of regressions to simultaneously evaluate the theory and the data (Staples & Webster, 2008).
PLS uses a combination of principal component analysis, path analysis and a set of regressions to simultaneously evaluate the theory and the data (Staples & Webster, 2008).