Chapter Overview
This chapter contains the research hypothesis, conceptual framework, research procedure, measurement instrument, and data analysis methods. The chapter explains the research framework by showing the hypothesized interrelation among variables.
Also, the chapter explains the sampling procedure and the necessity for the researcher to follow this procedure to obtain the required sample. Details of the measurement instrument are given with the reliability and validity explained. Finally, data collection, and data analysis methods are introduced.
Research Hypotheses
Previous studies have indicated that human capital is positively associated with structural capital and relational capital (Bontis, 1998; Bontis et al., 2000; Chen, 2001;
Cabrita and Bontis, 2008); also, structural and relational capital respectively mediate the impact of human capital on business performance. Therefore, the following hypotheses are developed.
H1. Human capital is positively associated with structural capital.
H2. Human capital is positively associated with relational capital.
H3. Structural capital is positively associated with relational capital.
H4. Structural capital is positively associated with business performance.
H5. Relational capital is positively associated with business performance.
20
Conceptual Framework
This research framework was developed in accordance with the literature review. From the review, it was noticed that intellectual capital is related to business performance. The Intellectual Capital Variables defined in the study are in relation to Cabrita and Bontis’ (2008) classification of intellectual capital: Human Capital, Structural Capital, and Relational Capital. Their interrelation and their impact on Business Performance will be tested.
Figure 3.1. Conceptual framework of this study Source: Revised from Cabrita and Bontis’ (2008) study
Human Capital (HC)
Structural Capital (SC) Relational Capital (RC)
Business Performance (P) H1 H2
H3
H5 H4
21
Research Procedure
A pilot test was administered in December 2008 and the data were collected by paper questionnaire. Before the distribution of questionnaire, it was reviewed by four experts in this field. For the pilot test sample, four executives of Taiwanese design companies and six students from the extended education division of Department of Fine Arts, National Taiwan Normal University were chosen using convenience sampling method. All participants are managers or directors who come from ten different design companies in Taiwan and their permissions to participate in the pilot study were obtained. The questionnaire items come from the empirical study of Cabrita and Bontis (2008), which are 71 items in total. All items are translated into Chinese by a bilingual translator and are revised by experts to suit the study. Also, the items are placed categorically as Cabrita and Bontis’ (2008) classification of intellectual capital.
For the main study, the researcher contacted Taiwan Design Center (TDC) requesting permission to mail surveys electronically using their design industry catalog. The researcher explained by telephone and mails the research background, research purpose, along with a note of confidentiality detailing that the data collected will be used solely for the researcher’s thesis and all names of companies will be excluded. Meanwhile, the researcher made phone calls by using the public catalog provided by the website of TDC (http://www.boco.com.tw). For every phone call, the researcher explained the purpose of the study and the contributions it may have to Taiwanese design industry. The participants are assured their anonymity and that the results will be sent to them if requested. Moreover, the researcher also reminded that the survey should be answered by managers or directors of the company as recommended by Bontis (1998) and Bukh et al. (1999). Electric surveys are mailed to these respondents so as to reduce the trouble of replying to paper questionnaires and
22
increase respondents’ willingness of reply.
After all the phone calls are made, the researcher waited and collected all the data. The data was coded and the information was keyed into the Statistical Package for Social Sciences (SPSS) PC 12.0 statistical software program. Figure 3.2 shows the research process of this study.
Figure 3.2. Research process Design of Final Report Conclusions and Suggestions
Analysis of the Data Implementation of the Survey
Design of Research Methods
Development of the Framework of the Study Establishment of Research Questions and Hypothesis
Discussion of the Literature Review Identification of the Research Subject
23
Measurement Instrument
To collect the data the survey is conducted by questionnaires. The researcher used the questionnaire developed by Cabrita and Bontis (2008) to assess the intellectual capital and business performance of Taiwanese design industry. All the items in the survey are translated from English to Chinese by one English major and then revised by four experts for face validity. It was six pages in length, containing a total of 71 items and a cover letter explaining the academic purpose of the study, the concept of intellectual capital, and the assurance of their confidentiality. Some of the items were reworded from the original in order the suit the characteristics of the sample. (Please see Appendix B for the survey of the study)
The survey consists of five parts. Respondents were asked to refer to their experience of working in their company and to fill out the questionnaire that has a range of items with regard to Human Capital, Structural Capital, Relational Capital, as well as Business Performance.
The first three parts assess the intellectual capital of the sample company using a 7-point Likert Scale. (1 = strongly disagree, 7 = strongly agree). They are respectively Human Capital (20 items), coded as H1 to H20; Structural Capital (16 items), coded as S1 to S16; Relational Capital (25 items), coded as R1 to R25.
The fourth part measures business performance (10 items), coded as P1 to P10.
The respondents are asked to state how their companies’ performance compared to their key competitors in the sector. The answer ranges from 1 to 10. (1 = my company performs the worst in the sector; 10 = my company performs the worst in the sector).
The last part requires respondents to provide the information of them (seniority and job title) and their company (location, company age, industry, type of property right, employee number and sales of year 2007).
24
Data Analysis Methods
There are two major data analysis methods in this study: Partial Least Squares (PLS) and multiple regression analysis. PLS is used to analyze simultaneously the interrelationships among all the constructs, however, one disadvantage of it is that PLS is not able to evaluate each individual item’s impact on business performance.
Therefore, this study also used multiple regression analysis to assess each item’s influence on the dependent variables. These two methods are complimentary to each other.
Partial least squares
Partial least squares (PLS) is a kind of structural equation modeling (SEM) technique. It is based on regression and originates from path analysis. As stated by Cabrita and Bontis (2008), it is a powerful tool in social and behavioral sciences where theories are formulated in terms of hypothetical construct, which are theoretical and cannot be observed or measured directly. Besides, PLS estimation does not require assumptions of normality or independence of observations. Moreover, it works well with small samples and is better suited for exploratory work. These are also the reasons that make PLS a more suitable analyzing method for this study.
Therefore, in this study, PLS is used to analyze intellectual capital data and business performance data. Through the use of PLS, the researcher can conduct confirmatory factor analysis and path analysis. Bontis (1998) reports the benefits of using PLS for such studies:
The objective in PLS is to maximize the explanation variance. Thus, R² (r-squared) and the significance of relationships among constructs are measures indicative of how well a model is performing. The conceptual core of PLS is an iterative combination of principal components analysis relating measures to constructs, and path analysis permitting the construction of a system of constructs. The hypothesizing of relationships between
25
measures and constructs, and constructs and other constructs is guided by theory. The estimation of the parameters representing the measurement and path relationships is accomplished using ordinary least squares (OLS) techniques. The first step in PLS is for the researcher to explicitly specify both the structural model and the construct-to-measures relationships in the measurement model. The exogenous constructs are consistent with the idea of independent variables (antecedents). Similarly, the endogenous constructs are consistent with the idea of dependent variables (consequents). The constructs can be specified as “formative” indicators or “reflective” indicators. Formative indicators imply a construct that is expressed as a function of the items (the items form or cause the construct).
Reflective indicators imply a construct where the observable items are expressed as a function of the construct (the items reflect or are manifestations of the construct). One looks to theory to decide on which type of epistemic or construct-to-measure relationship to specify. In this case, all constructs were “reflective” indicators. Once specified, the measurement and structural parameters are estimated using an iterative process of OLS, simple and multiple regressions. The process continues until the differences in the component scores converge within certain criteria (p. 69).
Due to the exploratory feature and small samples of this study, the researcher decided to adopt Visual PLS 1.04b1 as one of the major tools to investigate causal relationship between intellectual capital and business performance. Figure 3.3 demonstrates the methodological approach to test research hypotheses using PLS.
To examine the measurement model, items with loadings greater than 0.5 should be retained to reach adequate individual item reliability; internal consistency and Cronbach’s α ought to exceed 0.7 to achieve adequate convergent validity; the square root of average variance extracted of a construct should exceed its correlation coefficients with other constructs to get adequate discriminant validity and no items should load in wrong constructs when examining the cross-loading matrix. To analyze
26
the structural model, the model explanatory power is showed by the R² value and the significance of path coefficients is examined by t-test with degree of freedom N-1.
Finally, the “rule of thumb” for sample size requirements suggests that it will be equal to the larger of the following (Cabrita & Bontis, 2008):
1. 10 times the scale with the largest number of formative indicators (scales with reflective indicators can be ignored) or
2. 10 times the largest number of antecedent constructs leading to an endogenous construct.
In our study we applied the second requirement as all indicators are reflective.
The final full test would have 2 constructs. Therefore, a minimum of 20 (2 x 10) was required. Our sample size (87 samples) met the criterion.
Figure 3.3 Methodological approach to test research hypotheses using PLS Source: Cabrita and Bontis’ (2008) study
Individual item reliabilities
z Loadings (λ’s) – retained items with factor loadings >0.5
Internal consistency
z Convergent validity (Cronbach’s α and internal consistency) z Discriminant validity (average variance extracted,
examination of loadings and cross-loadings)
Explanatory power z R-squares (R²) for each dependent variable
Testing of hypotheses
z Estimation of path coefficients (Standardized β’s)
Significance of hypotheses
z Significance testing path estimates (jack-knifing procedure to examine the stability of estimates)
Measurement
27
Multiple regression
The multiple regression is conducted by SPSS. Before analysis, the data were coded using number sequences. The 61 intellectual capital questions were coded using a 7-point Likert scale, and the 10 performance questions were coded from 1 to 10 as previously mentioned, including seniority, job title, company location, company age, industry, type of property right, number of employees, and sales of year 2007.
Table 3.1 Coding System Used in SPSS Data Analysis (N=87)
Variables Seniority
(Unit: years)
1 = Less than 3
2 = Above 3, less than 5 3 = Above 5, less than 10 4 = Above 10, less than 15 5 = Above 15, less than 20 6 = Above 20
Job Title 1 = Chairman
2 = President 3 = Vice President 4 = Associate President 5 = Manager
6 = Deptu Manager 7 = Director
8 = Senior Designer 9 = Others
Company Location 1= Northern Taiwan
2 = Central Taiwan 3 = Southern Taiwan 4 = Eastern Taiwan
28
Industry 1 = Service design
2 = Activity design 3 = Product design 4 = Space design 5 = Others
Type of Property Right 1 = Corporation limited 2 = Limited
3 = Partnership
4 = Sole proprietorship
5 = Foreign enterprise Taiwan branch 6 = Others
Number of Employees (Unit: people) 1 = Less than 3
2 = Above 3, less than 5
(Unit: NTD) 2 = Above 500 thousand, less than 1 million 3 = Above 1 million, less than 2 million 4 = Above 2 million, less than 3 million 5 = Above 3 million, less than 5 million 6 = Above 5 million, less than 10 million 7 = Above 10 million, less than 20 million
29
The researcher used from the SPSS software, descriptive and inferential statistics to analyze and interpret the data collected from the sample population. These statistical procedures allowed the researcher to understand the impact of intellectual capital on business performance. The descriptive statistics helped the researcher to arrange the data into a more interpretable form by calculating numerical indices such as maximum value, minimum value, means, and standard deviation. All this data can be summarized easily or can be examined on their interrelation.
The use of inferential statistics helped the researcher to examine relationships, differences and trends, a process also known as hypothesis testing or significance testing. The researcher analyzed the data to investigate the relationship between intellectual capital and business performance. The inferential statistics provided the researcher with the means to test whether two variables are associated and to assess the strength between the independent and dependent variable; it also assisted the researcher to predict the value of dependent to independent variable, and compare relative changes in trends.
30