Chapter Overview
This chapter contains two major sections. The first sector gives an overview of the descriptive statistics of the research data, from which the features of Taiwanese design industries are discovered. The second section demonstrates PLS findings, including the analysis of the measurement model and the structural model that are used to test the hypotheses. Finally, a discussion part is given to further investigate the implications of PLS findings.
Descriptive Statistics
All the companies in the sector catalog were contacted. Of the 500 companies approached 78 responded which is a 15.6% response rate. An additional 15 questionnaires were personally delivered and collected, reaching a total of 93 questionnaires. Of the 93 questionnaires the researcher picked out 87 effective questionnaires, which is a rate of 93.5%. The Table 4.1 below provides a view of the sample population and their demographic information.
Table 4.1 Data of Variables by Entries and Values (N=87)
Variables Entries Percentage
31 Table 4.1 (continued)
Variables Entries Percentage
Company Location Northern Taiwan 58 66.7
Central Taiwan 16 18.4
Industry Service design 36 41.4
Activity design 5 5.7 Foreign enterprise Taiwan branch 4 4.6
Others 6 6.9
32 Table 4.1 (continued)
Variables Entries Percentage
Sales of Year 2007 (Unit: NTD)
Less than 500 thousand 2 2.3 Above 500 thousand, less than 1 million 6 6.9 Above 1 million, less than 2 million 11 12.6 Above 2 million, less than 3 million 9 10.3 Above 3 million, less than 5 million 10 11.5 Above 5 million, less than 10 million 11 12.6 Above 10 million, less than 20 million 13 14.9 Above 20 million, less than 30 million 6 6.9 Above 30 million, less than 40 million 1 1.1 Above 40 million, less than 50 million 1 1.1 Above 50 million 17 19.5 Capital (Unit:
33 Human capital
Concerning human capital, the executives showed high agreement to H4, which shows that many managers agree that their employees cooperate in teams. H20 pointed out that the employees gave it their all which makes the company different from the others in the industry. The lowest score of H13R indicated that if certain individuals in the firm unexpectedly left, they would be in big trouble. However this is not too significant to notice (See Table 4.2).
Table 4.2 Human Capital by Likert Scale, Mean, and Standard Deviation (N=87) Min. Max. Mean Std. Deviation
H1 competence ideal level 1 7 4.82 1.317
H2R no succession training program 1 7 4.92 1.894
H3 planners on schedule 1 7 4.56 1.412
H4 employees cooperate in teams 2 7 5.92 1.183
H5R no internal relationships 1 7 5.29 1.670
H6 come up with new ideas 1 7 5.41 1.369
H7 upgrade employees' skills 1 7 5.46 1.429
H8 employees are bright 2 7 5.44 1.158
H9 employees are best in industry 2 7 5.22 1.125
H10 employees are satisfied 1 7 5.18 1.225
H11 employees perform their best 2 7 5.36 1.131 H12 recruitment program comprehensive 2 7 4.98 1.312 H13R big trouble if individuals left 1 7 4.43 1.821 H14R rarely think actions through 1 7 4.54 1.546
H15R do things without energy 1 7 5.37 1.390
H16 individuals learn from others 1 7 5.45 1.265
H17 employees voice opinions 2 7 5.07 1.246
H18 get the most out of employees 2 7 5.30 1.221 H19R bring down to others' level 2 7 5.29 1.405 H20 employees give it their all 2 7 5.56 1.208 Note: The 7-Point Likert scale is used; R represents reverse coded items, but are positively coded before analysis
34 Structural capital
In relation to structural capital, item S8, S13R, and S15 pinpointed that the culture and the atmosphere of most companies are supportive and comfortable and that they support the development of new ideas and products. Also, the organization is not a
“bureaucratic nightmare,” which means the organizational structure is quite flexible.
However, the lowest score of S1 showed the managers’ disagreement and that their companies have the lowest cost per transaction in the industry (See Table 4.3).
Table 4.3 Structural Capital by Likert Scale, Mean, and Standard Deviation (N=87) Min. Max. Mean Std. Deviation
S1 lowest cost per transaction 1 7 3.80 1.598
S2 improving cost per revenue $ 1 7 4.22 1.458 S3 increase revenue per employee 2 7 4.94 1.124 S4 revenue per employee is best 1 7 4.76 1.320
S5 transaction time decreasing 1 7 4.55 1.292
S6 transaction time is best 1 7 4.25 1.323
S7 implement new ideas 2 7 5.06 1.297
S8 supports development of ideas 1 7 5.80 1.199 S9 develops most ideas in industry 1 7 5.26 1.316
S10 firm is efficient 1 7 4.95 1.266
S11 systems allow easy info access 1 7 5.01 1.451 S12 procedures support innovation 1 7 4.90 1.347 S13R firm is bureaucratic nightmare 1 7 5.63 1.356 S14 not too far removed from each other 1 7 5.41 1.394
S15 atmosphere is supportive 1 7 5.51 1.380
S16R do not share knowledge 1 7 5.17 1.740
Note: The 7-Point Likert scale is used. R represents reverse coded items, but are positively coded before analysis
Relational capital
In the dimension of relational capital, five variables showed the managers’
agreement concerning the aspects of customers. Item R13R, R14, R15R, R16, R17 showed that design companies generally care about what customer thinks or wants from them. They capitalize on customers’ wants and needs by: continually striving to make
35
them satisfied, getting as much feedback out of customers as they possibly can, and launching services or products that fits customers’ needs. Also, they feel confident that their customers will continue to do business with them. Nevertheless, R4 pointed out the market share of design companies are not usually high (See Table 4.4).
Table 4.4 Relational Capital by Likert Scale, Mean and Standard Deviation (N=87) Min. Max. Mean Std. Deviation R1 customers generally satisfied 3 7 5.59 1.018 R2 reduce time to resolve problem 1 7 5.00 1.347
R3 market share improving 2 7 4.79 1.374
R4 market share is highest 1 7 3.52 1.477
R5 longevity of relationships 1 7 4.87 1.265
R6 value added service 1 7 5.20 1.310
R7 customers are loyal 2 7 5.30 1.259
R8 customers increasingly select us 1 7 4.90 1.239
R9 firm is market-oriented 1 7 4.72 1.300
R10 meet with customers 2 7 5.56 1.198
R11 customer info disseminated 3 7 5.26 1.289
R12 understand target markets 1 7 5.20 1.284
R13R do not care what customer wants 1 7 6.00 1.248 R14 capitalize on customers’ wants 1 7 5.62 1.287 R15R launch what customers don't want 2 7 5.70 1.202 R16 confident of future with customer 1 7 5.71 1.238
R17 feedback with customer 1 7 5.64 1.161
R18 react to competition 2 7 5.07 1.283
R19 discuss competitors' strength and weakness 1 7 5.00 1.525
R20 contact with sector 1 7 4.44 1.568
R21 consider info from sector 1 7 4.54 1.328
R22 decisions based on info from sector 1 7 4.51 1.311 R23 supports share of info from sector 1 7 4.74 1.316
R24 share competitor info 1 7 5.37 1.192
R25 competitors are sources of innovation 1 7 4.78 1.603 Note: The 7-Point Likert scale is used. R represents reverse coded items, but are positively coded before analysis
36
From all the tables above, the researcher has decided to show the top 5 and the bottom 5 intellectual variables as indicated by the respondents. In Table 4.5 we can see that Taiwanese design companies do care about customers’ opinions and needs, they have confidence in repeat customers, and they launch new products or services that fits customers’ needs. Also, the employees cooperate in teams and the company supports the development of new ideas and products.
Table 4.5 Top Five Intellectual Capital Responses (N=87) Items Score Descriptions
R13R 6.00 We generally do not care about what the customer thinks or wants from us H4 5.92 The firm gets the most of out of its employees when they cooperate
with each other in team tasks
S8 5.80 Our company supports the development of new ideas and products R16 5.71 We feel confident that our customers will continue to do business with
us
R15R 5.70 We often launch something new only to find out that our customers do not want it
In Table 4.6 we can see that Taiwanese design companies generally don’t have a high market share, they don’t focus much on improving cost per transaction and cost per revenue dollar, neither on time to complete a whole transaction. In addition, if certain individuals in the firm unexpectedly left, the company would be in big trouble.
Table 4.6 Bottom Five Intellectual Capital Responses (N=87) Items Score Descriptions
R4 3.52 Our market share is the highest in the industry
S1 3.80 Our organization has the lowest costs per transaction of any in the industry
S2 4.22 We have continually been improving our costs per revenue dollar S6 4.25 The time it takes to complete one whole transaction is the best in the
industry
H13R 4.43 If certain individuals in the firm unexpectedly left, we would be in big trouble
37
Descriptive Statistics Discussion
From the descriptive statistics, we have found out some characteristics of intellectual capital in Taiwanese design industry. The results showed that employees work in teams in design companies (H4) to complete tasks, and they give it their all when they work (H20). Also, if certain individuals unexpectedly left, the firm would be in big trouble (H13R). This might be due to the fact that design companies are usually small-scaled and teamwork plays a crucial role in contributing to company’s performance.
Moreover, the organizational structure of design companies is not bureaucratic (S13) and supports the development of new ideas and products. Also, the culture of the design companies is usually supportive (S15). Additionally, the managers don’t seem to focus on reducing costs (S1). It can be inferred that design companies needs a supportive culture and flexible organizational structure to support creation and innovation. However, to maintain a working environment like this, some efficiency might be sacrificed in replace of more flexibility.
Furthermore, customers’ needs (R14 to R17) are considered crucial in the design industry. Another fact is that design companies don’t seem to have high market share (R4). There are few design companies that possesses high market share in Taiwan’s market.
Validity and Reliability of the Measurement Instrument
Validity of the instrument was determined by content validity. Content validity is basically the extent to which the measurement questions provide adequate coverage of the investigative questions. Before conducting the pilot test, face validity is reached through revising the items and its meaning by four experts, including three professors from National Taiwan Normal University (NTNU), and one design director of China Productivity Center (CPC). These experts provided assessments of each item in the questionnaire by determining if they are appropriate or not. (Please see Appendix C for the list of experts)
To test reliability, the researcher revised the reverse coded questions and conducted a pilot test using 10 samples with Statistical Package for the Social Sciences (SPSS) PC 12.0, which indicated a high internal consistency based on the alpha reliability value:
38
Human Capital 0.816 (20 questions); Structural Capital 0.894 (16 questions); Relational Capital 0.935 (25 questions); Performance 0.856 (10 questions). For the final test: Human Capital 0.928 (20 questions); Structural Capital 0.868 (16 questions); Relational Capital 0.925 (25 questions); Performance 0.958 (10 questions).
Table 4.7 Cronbach’s α Value (Based on Standardized Items) of Survey Instrument
Tests Pilot (n=10) Final (n=87)
Human Capital .816 .928
Structural Capital .894 .868
Relational Capital .935 .925
Performance .856 .958
39
PLS Findings
Cronbach’s alpha and individual item reliabilities
The reliability of the final test was inspected using Cronbach’s alpha. The reliabilities for each of the four constructs were greater than 0.86, which exceeds the criterion of 0.7, considered good for exploratory research (Nunnally, 1978). Then, PLS is used to assess individual item reliabilities so as to confirm factor findings. At early stages of scale development, loadings of 0.5 or greater maybe acceptable if there exists additional indicators for describing the latent construct (Chin, 1998). Therefore, items with loadings of 0.5 or greater are retained. There are other authors (Birkinshaw, Morrison, & Hulland, 1995) who have also followed this criterion in their exploratory studies. Table 4.8 shows the results of PLS loadings on all the items.
Table 4.8 PLS Loadings
Items Loading Items Loading Items Loading Items Loading
H1 0.7116 S1 -0.0429 R1 0.7935 P1 0.7550
H2R 0.2901 S2 0.0522 R2 0.5848 P2 0.8353
H3 0.7169 S3 0.7699 R3 0.7495 P3 0.8061
H4 0.7166 S4 0.6938 R4 0.3555 P4 0.9023
H5R 0.2611 S5 0.6162 R5 0.6736 P5 0.8617
H6 0.7339 S6 0.3335 R6 0.7546 P6 0.8206
H7 0.6619 S7 0.7976 R7 0.8066 P7 0.8563
H8 0.8087 S8 0.8126 R8 0.7590 P8 0.9116
H9 0.6665 S9 0.7866 R9 0.5124 P9 0.8489
H10 0.8365 S10 0.6127 R10 0.6102 P10 0.8996
H11 0.8920 S11 0.6117 R11 0.7459
H12 0.7353 S12 0.6829 R12 0.7647
H13R 0.1934 S13R 0.6052 R13R 0.6469
H14R 0.4964 S14 0.7767 R14 0.7450
H15R 0.5569 S15 0.8279 R15R 0.6802
H16 0.7550 S16R -0.0241 R16 0.8782
H17 0.6453 R17 0.5328
40
Item R17 (We get as much feedback out of our customers as we possibly can under the circumstances) was dropped because it was loaded incorrectly at 0.5449 for the human capital construct when we used PLS techniques (Please see Appendix D for Matrix of Loadings and Cross-Loadings). This left us with 16 indicators for the human capital construct; 12 indicators for structural construct; 16 indicators for relational capital and; 10 items to measure performance. The researcher compared the results with the studies administered in Canada, Malaysia, and Portugal and confirmed that 15 items were reliable in all four researches and 17 were reliable in at least three contexts (See Table 4.9).
Table 4.9 Reliable Items – Comparing Studies in Canada, Malaysia, Portugal and Taiwan Canada Malaysia Portugal Taiwan Canada Malaysia Portugal Taiwan
Human capital Structural capital
H6
41 Table 4.9 (continued)
Canada Malaysia Portugal Taiwan Canada Malaysia Portugal Taiwan
Relational capital Performance
*reliable measures in the Taiwan context and one other country
**reliable measures in the Taiwan context and two other country
*** reliable measures in all four studies
Note: The relational capital was originally coded with “C” in the study of Canada and Malaysia. This study revised them in to “R” in order to avoid misunderstanding.
Source: Revised from Cabrita and Bontis’ (2008) study
According to Cabrita and Bontis (2008), in spite of that the measurement and structural parameters are estimated together, a PLS model is analyzed and interpreted in two stages: the assessment of the reliability and validity of the measurement model, and the assessment of the structural model. The sequence ensures reliable and valid measures of constructs before we try to draw conclusions with regard to the relationships among the constructs.
Testing the Measurement Model
This study uses Cronbach’s alpha in SPSS and PLS approach to assess the measurement model (outer model). All the Cronbach’s alpha values of the four constructs exceeded 0.91 (0.942 for human capital; 0.914 for relational capital; 0.935 for relational capital; 0.958 for business performance).
42
Individual item reliabilities were evaluated by examining the loadings of the measures with their corresponding construct. All loadings were greater than 0.522 except the loading of R9, which is 0.4857; however, it is not too low to be deleted (Table 4.10).
Convergent validity was assessed using the internal consistency measure (Table 4.11), developed by Fornell and Larcker (1981). All values for the four constructs exceeded 0.7, as recommended by Nunnally (1978). Concerning discriminant validity, R17 (We get as much feedback out of our customers as we possibly can under the circumstances) was deleted after examining the cross-loading matrix (Please see Appendix D). Concerning model explanatory power, the R² value of this model (35.5%) is not quite different from those in the Canadian study (56.02% and 56.9%), the Malaysian study (32.1% and 37.3%), and the Portugal study (44.5%).
Table 4.10 Factor loadings Constructs Loadings
H
H1(0.7154), H3(0.7140), H4(0.7156), H6(0.7337), H7(0.6656), H8(0.8157), H9(0.6781), H10(0.8392), H11(0.8938), H12(0.7325), H15R(0.5482), H16(0.7487), H17(0.6493), H18(0.81829), H19R(0.5858), H20(0.8174)
S
S3(0.7704), S4 (0.6929), S5(0.6055), S7(0.7977), S8(0.8164), S9(0.7870), S10(0.6038), S11(0.6163), S12(0.6859), S13R(0.6079), S14(0.7800), S15(0.8314)
R
R1(0.8287), R2(0.6128), R3(0.7605), R5(0.6934), R6(0.7561), R7(0.8230), R8(0.7700), R9(0.4857), R10(0.6177), R11(0.7514), R12(0.7718),
R13R(0.6554), R14(0.7605), R15R(0.6964), R16(0.8873), R17(0.4857), R24(0.5222)
P P1(0.7531), P2(0.8353), P3 (0.8057), P4(0.9025), P5(0.8617), P6(0.8203), P7(0.8562), P8(0.9120), P9(0.8497), P10(0.9000)
Table 4.11 Measurement Model Results
Constructs Number of
43
In order to evaluate the statistical significance of the loadings and the path coefficients (standardized betas), a jackknife analysis was performed. In this case 86 sub-samples were created by removing one case from the total data set. By applying the jackknife formula, PLS estimates the parameters for each sub-sample and compute the
“pseudovalues” (Table 4.12). Three paths (human capital to structural capital, structural capital to relational capital, and relational capital to performance) are proved to be significant at the p-value < 0.1. Results showed that the explanatory power (R²) for the model is 35.5 %. Nevertheless, the path between human capital to relational capital, and structural capital to business performance was not significant and thus didn’t support the hypothesis.
Table 4.12 PLS Path Analysis Results (Standardized Beta Coefficients and Adjusted T-values)
Path Hypotheses β-path Adj. t-value Sig. Support Direction H→S H1 0.870 22.261 *** V +
Figure 4.1 Major Structural Model
* p < 0.1. **p <0.05. *** p <0.01.
Figure 4.1 demonstrates the results for the structural model. The results pinpoint that the three constructs that forms intellectual capital really affect one another. One important
0.616***
44
benefit of the PLS methodology is that it makes it possible to separate direct and total effects of the variables included in the model (Cabrita & Bontis, 2008). As we can see from Figure 4.2, decomposition of effects shows that Human capital has important effects on structural capital (0.870). Additionally, it influences relational capital indirectly through the structural capital (0.870 x 0.616 = 0.536). Furthermore, the total effects of human capital on business performance are HC→SC→RC→P (0.870 x 0.616 x 0.521).
The results pinpointed the features of Taiwanese design industry. Human capital influence structural capital directly, and impact on relational capital indirectly through structural capital. Moreover, the major contribution of Taiwanese design companies comes from relational capital instead of structural capital. That is to say, the talents of a design company is less helpful in enriching the companies’ relationship with customers, suppliers, industry associations or any other stakeholder that influence the organization’s life. Instead, it is favorable for creating the company’s valuable strategic assets, such as information systems, routines, procedures and database. As far as business performance is concerned, maintaining good relationship with the firm’s stake holders is a vital way to pursue for excellent performance.
Table 4.13 Summary of Model Direct and Indirect Effects
Path Equation Influence
H>S Direct effect: H>S 0.870
H>R Indirect effect: H>S>R 0.870*0.616= 0.780 H>P Indirect effect: H>S>R>P 0.870*0.616*0.521=0.406
45
PLS Findings Discussion
All of the hypotheses were supported except hypothesis 2 (Human capital is positively associated with relational capital) and hypothesis 4 (Structural capital is positively associated with business performance). After analyzing the research data, the researcher is interested in finding out how other variables, such as the number of employees in the company, or company age, could influence the intellectual capital structure and the relationship between intellectual capital and business performance. For this reason, the researcher divided the 87 samples into four sample groups to examine if there is a trend of change in the R² value.
To do so, the researcher first divided the sample into two sample groups: Sample B (companies with less than five employees, 34 samples) and Sample C (companies with more than five employees, 53 samples) and ran PLS separately to see the difference. The results indicated that all path coefficients on the structural model of Sample C (Please see Figure 4.2), are significant at p-value <0.05; while in Sample B companies, only three paths are significant at p-value < 0.1.
To examine whether there is a trend in the change, we picked another two sub-samples respectively from Sample B and Sample C. Out of Sample B’s 34 sub-samples, 32 companies of age less than 15 years are picked out and named Sample A; out of Sample C’s 53 samples, 40 companies of age more than 5 years are picked out and named Sample D. Thus Samples A to D are manipulated to represent, respectively, companies with fewer employees and younger companies to those with more employees and older companies.
46
Sample A (less than five employees;
company age less than fifteen years) n = 32
Sample B (less than five employees) n = 34
Sample D (more than five employees;
company age more than five years) n = 40
Sample C (more than five employees) n = 53
Figure 4.2 Structural Models of Different Samples Groups
Note: * p < 0.1. **p <0.05. *** p <0.01. T- stat in brackets. Sample A to D represents companies with different numbers of employees (from few to many) or companies with different age (from young to old)
The results showed some interesting findings. Comparing the R² value of Sample A and Sample D, we can see that the R² value of Sample A (56.0%) is higher than that of
47
Sample D (30.6%). The outcomes indicated that the model of Cabrita and Bontis (2008) has higher explanatory power over business performance when it is used to explain the performance of younger companies with fewer employees. When it comes to older companies with more employees, the model has lower explanatory power. This pointed out the limitation of this model and room for future research improvement. Concerning future research, other variables, such as the scale of the company could be added to the research model in order to investigate what other variables could improve the explanatory power of the model.
48