LITERATURE REVIEW
3.4 Other Variables
3.4.1 Degree of Internationalization
Ietto-Gillies and London (2009) identified three major dimensions in the internationalization concept: intensity, geographical extensity, and concentration dimension. These three dimensions have been incorporated into the study. The intensity dimension focuses on the dichotomy measure of foreign versus domestic activities. From the geographical aspect of internationalization, extensity dimension measures the number of countries in which activities take place, while concentration dimension focuses on the degree to which activities are concentrated within the foreign countries. The market and locational advantages deviate across geographic regions due to differences in socio-economic environment (Qian, 2000), and have varying degrees of emphasis on
environmental concern (Ö zen and Küskü, 2009). Therefore, when considering geographical dimension of internationalization, not only number of countries should be taken into consideration, the geographical concentration or regional effect should also be incorporated.
Foreign sales to total sales revenue: For the internationalization intensity, the most common measure used by researchers has been the percentage of foreign sales to total sales revenue (FSTS), and is adopted here as investment intensity. FSTS is computed according to the total revenues and international revenues shown in ENR report (ENR, 2012).(Engineering News Record (ENR), 2012)(Engineering News Record (ENR), 2012) To compute the geographical extensity, and concentration, regional classification adopted in 2012 Environmental Performances Index data files are invoked (EPI, 2012).
There are a total of six regions, according to the countries listed in EPI 2012. These regions are further sub-divided into four developed regions and six developing regions.
The former includes Asia and Pacific; Europe; Middle East and North Africa; and Americas. The latter includes Asia and Pacific; Eastern Europe and Central Asia; Europe;
Americas; Middle East and North Africa; and Sub-Saharan Africa. The countries of six developing regions are identical to World Bank’s country classification, thereby ensure the convergent validity. The countries in which each firm worked in 2011 can be found in the ENR report (2012).
Network Spread Index (NSI): Developed by Ietto-Gillies (1998), NSI has been used to measure the percentage of foreign countries a firm is affiliated with in relation to the total number of foreign countries in which, potentially, the firm could occupy. As indicated in Pheng and Hongbin (2004) study, NSI has been adopted in this study for the country-level analysis of a firm’s international business distribution and used as proxy to geographical extensity.
Regional diversification index (RDI): Geographic regions are substantially different in socio-economic environment (Qian, 2000). The imperative for regional study underscores insufficiency of purely country-level analyses in the evaluation of a firm’s operations across multiple locations that are distinct but not entirely independent of each other (Ghemawat, 2003). As in Qian et al. (2008) study, entropy measure is adopted to measure the geographical concentration. The entropy measure of regional diversification index is defined as:
𝑅𝐷𝐼 = [∑ 𝑃𝑖ln (1 𝑃𝑖)
𝑚
𝑖=1
] /ln(𝑚)
where Pi is the probabilities of number of countries where a firm had its subsidiaries to regional market i, and ln (1/Pi) is the weight that is given to each global market region, m is the number of total regions considered in the computation.
Three NSI and three RDI related variables have been derived for comparison. First, NSI and RDI are derived from a global standpoint comprising all 10 regions (NSIoverall
and RDIoverall); second, NSI and RDI are used to measure the degree of internationalization according to the respective number of countries in the four developed regions (NSIdeveloped and RDIdeveloped) and six developing regions (NSIdeveloping and RDIdeveloping).
3.4.2 Financial Performances
The dependent variables comprise of different aspects of financial performance and all the data are extracted from the Datastream database. Generally, accounting profitability is assumed as measures of past or short-term financial performance and market performance as measures of future or long-term performance (Hoskisson et al., 1994).
The study uses accounting-based measures of return on assets (ROA) and return on sales
(ROS) to identify the short-term financial outcomes of environmental practices. Return on equity has been ruled out due to its sensitivity to capital structure differences (Hitt et al., 1997). ROA and ROS were calculated as the mean value over the 3-year period from 2011 to 2013. Firms will maintain higher ROA if their environmental management depend less on end-of-pipe control and rely on firm’s capability to facilitate the environmentally sustainable economic activity. On the other hand, efficient utilization of raw material would reduce unnecessary waste and minimize input which would result in higher ROS. In addition, revenue growth and market-based measure, Tobin’s Q ratio are selected to examine long-term financial performance from the environmental practices.
Tobin’s Q is computed by dividing the market value of assets to the replacement value of assets, and averaging the values over the 3-year period 2011 to 2013. Tobin’s q is able to reflects what cash flows the market thinks a firm will provide per dollar invested in assets and higher Tobin’s Q ratio is mirroring market expectation of future cash flows to be greater or less risky (King and Lenox, 2001). Revenue growth is computed as revenue change from 2011 to 2013. It is expected environmental management would assist greater business expansion which manifests in revenue growth. The samples of construction firms span across three developed continents, namely Asia and Pacific, Europe, and America. The financial performance differences attributable to the continent of origin and its concomitant effects, such as differences in accounting practices employed, might increase the potential for confounding effects in the statistical analysis (Michael Geringer et al., 1989). In order to control for these confounding effects, this study employed the method suggested by Michael Geringer et al. (1989), to standardize all the financial performance variables in accordance with the continent of origin. A list of financial variables is summarized in Table 3.5.
Table 3.5 Financial performance variables
Measure Calculation Notes
ROA ROA = Net Income / total assets Mean value over the or if they are expected to be less risky. response in host countries (Kolk and Fortanier, 2013; Sharma et al., 2007). The influences of a firm’s home country conditions are captures in two-ways. First, the environmental governance of a home country was measured according to the Environmental Performance Index (EPI), published jointly by Yale University and Columbia University in year 2012 (EPI, 2012). Next, the gross domestic product per capita (GDPCAP) of a construction firm’s home country is included in the study. For firm-level considerations,
many studies in environmental management consider firm size effect. The natural logarithm of number of employees is used to measure firm size (Size) in the study.
Another firm-level variable considered is revenue growth of a firm which depicts the difference of revenue over 2009 to 2011 (REVG0911). Number of employees and revenue over 2009-2011 are extracted from Datastream database.
3.5 Analysis
3.5.1 Analysis of Variance and Other Extension Methods
In Test 1, one-way ANOVA tests are adopted to test whether the means of each internationalization variables were statistically significant different across the environmental strategy clusters. The dependent variables used in this test are FSTS, NSI and RDI. In addition, post hoc Tukey's honest significant difference (HSD) tests are performed to further investigate the statistical differences between the pairwise clusters.
Next, multivariate analysis of variance (MANOVA) was conducted based on overall internationalization variables (excluded FSTS) and environmental strategy clusters.
Control variables such as EPI, GDPCAP, Size, and Revenue Growth over 2009-2011, are entered into the analysis as covariates and one-way analysis of covariance (ANCOVA) is performed to verify whether each of the dependent variable were still associated with differences among the strategy cluster after the home condition effects and firm control variables have been accounted for. A similar multivariate analysis of covariance test (MANCOVA) is performed by taking all the dependent variables of internationalization (except FSTS) together with the covariates, and the result was compared with MANOVA result.
The paired sample t-tests have been conducted to explore possible influences of business distribution portfolio within a specific strategy. With respect to each
environmental strategy cluster, the pairwise RDIdeveloped- RDIdeveloping and NSIdeveloped -NSIdeveloping are used for comparison.
The Levene’s tests would be carried out to test either the assumption of homogeneity of variance of ANOVA is being violated (p>0.05). As a remedy, the robust Welch’s F-ratio would be reported if this assumption is violated. For ANCOVA, assumption of homogeneity of regression slopes can be tested by customizing the ANCOVA model in SPSS to look at the independent variable×covariate interaction and the p-value should be greater than 0.05.
3.5.2 Stepwise Regression
Test 2 is conducted to explore potential environmental management practices that associates with the short and long term financial performances. Stepwise multiple linear regression is adopted in this analysis. In stepwise regression, a predictor is added to the equation each time, a removal test is made of the least useful predictor. As such the regression equation is constantly being reassessed to see whether any redundant predictors can be removed (Field, 2009). The screening process of stepwise regression is based on the F-statistic, the entry threshold is set at 0.05 and the removal threshold is 0.10.
The stepwise selection assists in screening for those environmental practice variables that appear to significantly impact financial performance. The variables of environmental management practices are shown in Topic 3.3.2. The dependent variables used in the regression consist of ROA, ROS, Tobin’s Q ratio and revenue growth from 2011 to 2013. Control variables such as firm size, and Revenue Growth over year 2009 to 2011 are included in the first stage before proceeding with the stepwise regression. In this analysis, ROA and ROS are both indicators of short-term financial performances,
while Tobin’s Q ratio and Revenue Growth over 2011 to 2013 are indicators of long-term financial performances.
Several regression diagnostics are further conducted to ensure that basic assumptions for ordinary least squares (OLS) regression are satisfied, such as:
(i) White test was performed to examine whether the sample met the homoskedasticity assumption of the OLS regression.
(ii) Ramsey reset test was performed to test specification errors such as omitted variables and non-linearity of functional form.
(iii) Variance inflation factor (VIF) test against each regressor was performed to test the no-multicollinearity assumption for OLS, especially for those with significant correlation.
3.5.3 Moderated Regression Analysis
Moderated regression analysis, a form of ordinary least squares regression method is adopted to examine the moderating effect of RDI on the relationship between environmental management practices and financial performances. Among the three variables of DOI, RDI is selected over the other two variables, as it is theoretical proven that regional differences increase the cost of coordinating geographically dispersed operations, can reduce or negate potential benefits associated with increased internationalization scope (Lu and Beamish, 2004; Qian et al., 2008). The three environmental management practices used in Test 3 are environmental surveillance, process-related pollution abatement, and environmental innovation. Five interaction terms would be constructed, three terms are formed by multiplying the linear terms of environmental management practices variables with RDI, and another two squared-term variables of process-related pollution abatement and environmental innovation are also
used to form the interaction terms. The dependent variables used in the regression consist of ROA, ROS, Tobin’s Q ratio and revenue growth from 2011 to 2013. Control variables consist of firm size, and Revenue Growth over year 2009 to 2011.
For each financial dependent variable, the regression starts with a control model which includes the control variables, RDI, linear and square-term of environmental management variables. Next, each interaction term of environmental surveillance, process-related pollution abatement, and environmental innovation would be included in the regression equation separately in order to minimize multicollinearity among the independent variables.
Significant regression coefficients for the interaction terms and significant increases in the explanatory power of the model through inclusion of the interaction terms support the hypotheses regarding moderating effects. T-tests were used to assess the significance of regression coefficients and F-tests to assess the significance of the increase in the explanatory power of the models.