CHAPTER 5: SAMPLE CHARACTERISTICS AND MEASUREMENT SCALES
5.4. The outcomes of formative measurement scales assessment
The structural capital (SCE) variable is built by formative measurement scales, those are innovation capital efficiency (RDCE) and organizational capital efficiency (ORGCE).
This section firstly presents the calculations of the RDCE and ORGCE variables and then, as being presented in the Part 4.2.2, they are assessed on their convergent validity, collinearity issue, statistically significant and relevance of the formative indicators.
5.4.1. Calculation of measurement scale of innovation capital efficiency
As presented in the Part 4.4.1.3, the efficiency of innovation is calculated by the cumulative research and development investment divided by the value added. The cumulative R&D investment is measured by lump sum value of the carrying amount of the prior years’ R&D investments on the basis of 3-year economic life. This study accepts that R&D investment roughly is straight depreciated within a 3-year economic life. The author is unable to apply a long duration of the depreciation because the author may not collect enough data in a young Vietnam stock exchange market where information has been only fully available since 2010. For example, suppose if the first firm in the manufacturing industry spends R&D1,2016 = 3,408 million VND, R&D1, 2015 = 5,658 million VND, R&D1, 2014 = 1,307 million VND, the cumulative R&D investment in year 2016 is: RDC1, 2016 = 3,408 + 0.667 x 5,658 + 0.333 x 1,307 = 7,617.12. Then, the efficiency of innovation of this firm (0.0503) is calculated by the cumulative R&D investment (7,617.12) divided by the value added (151,395).
5.4.2. Calculation of measurement scale of organizational capital efficiency
As demonstrated in the Part 4.4.1.3, the empirical results from the simple correlations and multivariate regressions do not control for the potential endogeneity of SGAi,t and Ei,t
(Shangguan, 2005). In the presence of endogeneity, ordinary-least-squares estimation yields biased and inconsistent coefficient estimate. For this reason, the author conducts the
2-step regression of the equation 4.15 with firm-specific fixed effects and year-specific random effects by industry. In this two-step procedure, the first step is the estimation of a regression that predicts the SGA expenditures driven by some underlying exogenous variables such as total assets, profitability, and the second-stage (OLS) regression uses estimates from the first stage to provide consistent estimates of the parameters between SGA expenditures and annual earnings. Please refer to Appendix 19. Appendix 20 Panel A reports the coefficient estimates across years for different industries. After conducting the 2-step regression, the value of δ1, δ2, δ3 in the equation of SGA expenditures and annual earnings are estimated by industry. For all industries, δ1 and δ2 are significant; however, δ3
is not significant in some industries such as mining and energy, commercials, logistics and transportation equipment, real estate and construction. This demonstrates that SGA expenditure has a two- or three-year useful life. It has the largest impact on earnings in the concurrent year and then depreciates very quickly.
To calculate the amortization rates of SGA expenditures, the author only considers those statistically significant δk’s in Panel A. Each of these coefficient estimates represents the benefits contributed by the associated SGA expenditures to earnings, while the sum of significant δ’s represents the total benefits of SGA expenditures for one year. For example, for the Manufacturing sector, δ1 = 1.694 represents the contribution to earnings by the current SGA expenditures, δ2 = 0.651 represents the contribution to earnings by the previous-year SGA expenditures, δ3 = 0.204 represents the contribution to earnings by the SGA expenditures of the year before the previous year. ∑(δ1, δ2, δ3) = 2.549 represents the total earnings in year t contributed by SGA expenditures over t, t-1, t-2 years, while, for the commercials industry, only the sum of δ1 and δ1 = 1.493 represents the total earnings in year t contributed by SGA expenditures over t, t-1 years because δ3 is not significant enough to confirm the contribution of SGA expenditure arsing in the year t-2. Accordingly, for manufacturing sector, 0.665 = [1.694 / 2.549]; 0.255 and 0.080 are the amortization rate of SGA expenditures in the year t, t-1 and t-2, respectively. Panel B of Appendix 20 illustrates the amortization rates for the selected industries.
After determining the value of the amortization rates, firm-specific level of organizational capital is measured by the equation 4.17. For instances, suppose if the first firm in the manufacturing industry spends SGA1,2016 = 49,854 million VND, SGA1, 2015 = 40,122 million VND, SGA1, 2014 = 19,158 million VND, the cumulative organizational
capital investment in year 2016 is: ORGC1, 2016 = 49,854 x (1 – 0.665) + 40,122 x (1 – 0.255 – 0.655) + 19,158 x (1 – 0.080 – 0.255 – 0.665) = 19,910.85. Then, the efficiency of organizational capital of this firm (0.131516) is calculated by the cumulative organizational capital investment (19,910.85) divided by the value added (151,395).
5.4.3. Assessment of formative measurement scales related to the structural capital efficiency variable
The RDCE and ORGCE indicators are formative measurement scales of the SCE variable. As illustrated in the Part 4.2.2, they are assessed on their convergent validity, collinearity issue, statistically significant and relevance of the formative indicators, as follows:
Convergent validity: This characteristic is measured by correlating between the formatively measured construct with a reflectively measured construct of the same construct. However, established reflective measurement instruments could not be available, and constructing a new scale is difficult and time- consuming (Hair Jr & Hult, 2016). An alternative is to apply a general item that summarizes the essence of the construct the formative indicators purport to measure (Hair, Ringle, & Sarstedt, 2013). For the PLS-SEM on structural capital, an additional question, “Please assess the extent to which your company’s structural capital performs in last three years, compared with your major competitors”, each respondent circles the value measuring on a scale of 1 (extremely poor) to 7 (excellent). This question can be used as an endogenous single-item construct to validate the formative measurement of structural capital.
Figure 5.1. Assessment of convergent validity of formative indicators relative to structural capital
Source: Calculated by the author in SmartPLS 3.1
Figure 5.1 shows the results for the redundancy analysis for the SCE construct.
The original formative construct is labelled with SCE_F, whereas the general assessment of the company’s structural capital efficiency using a single-item construct is labelled with SCE_G. As can be seen, this analysis yields a path coefficient of 0.885, which is above the recommended threshold of 0.70, thus providing support for the formative construct’s convergent validity.
Collinearity issue: Unlike reflective indicators, which are essentially interchangeable, high correlations are not expected between items in formative measurement models. As can be seen in Table 5.5, there is not any collinearity problem between formative indicators due to the fact that each indicator’s VIF value (1.203) is lower than 5.0 (Hair Jr & Hult, 2016).
Significance and relevance of formative indicators: Another important criterion for evaluating the contribution of a formative indicator, and thereby its relevance, is its outer weight. The outer weight is the outcome of the multi- regression between the latent variable and the formative indicators. The results in Table 5.5 show that the outer weights are significant and high enough (i.e.
above 0.50) to be generally retained for further regressions.
Table 5.5. VIF, Significance and relevance of formative indicators
Models with endogenous construct RDCE ORGCE
Asset turnover Outer weight 1.097*** 0.430***
(34.105) (3.617)
Investment efficiency Outer weight 1.090*** 0.561***
(27.045) (5.969)
ROE Outer weight 1.096*** 0.487***
(32.853) (4.312)
Tobin q Outer weight 1.097*** 0.437***
(33.835) (.3794)
IC components Outer weight 1.096*** 0.485***
(32.325) (6.907)
Outer VIF values 1.203 1.203
Note: Significant at: *10, **5 and ***1 percent levels (two-tailed), t value (shown in brackets)
Source: Calculated by the author in SmartPLS 3.1