Financial leverage RSK is employed in

Financial leverage (RSK) is employed in the literature (e.g., Waddock & Grave, 1997) as a proxy for the risk. It is used to reflect management׳s risk tolerance that Z-YVAD-FMK influences its attitude towards social activities and measured by the ratio of total debt to total assets. Dividend per share (DIV) is included to reflect available investment opportunities (Wahba, 2010), and measured by the total dividend paid to ordinary shares divided by number of ordinary shares. Liquidity (LIQ) is added to control for managerial discretion regarding social initiatives and programs (Elsayed & Paton, 2009), and proxied by the ratio of current assets to current liabilities. Capital intensity (CAP) is also included as a control variable for the expected relationship between capital intensity and social investment decision (Rust & Rothwell, 1995). Capital intensity is measured by the ratio between payments in fixed assets and the firm׳s total assets. Controlling for industry effects (SIC) is also important as product differentiation may depend on the industry to which the firm belongs (McWilliams & Siegel, 2001). Consequently, the study supplements the models by experimenting with the inclusion of dummy variables for each two-digit standard industrial classification (SIC) code. Variables descriptive statistics are presented in Table 3.
Econometric analysis To examine whether financial performance mediates the relationship between social responsibility and institutional investors, we used Baron and Kenny׳s (1986) regression approach, while taking into our consideration the recent critique and modifications suggested by Zhao, Lynch, and Chen (2010). According to Baron and Kenny (1986), testing for mediation effect can be done in three steps: first, regressing the mediator on the independent variables. Second, regressing the dependent variable on the independent variables. Third, regressing the dependent variable on both the independent variables and mediator. They pointed out that the independent variable in the first two models is expected to show a statistical significance, while the third model is expected to show a statistical significance of the mediator variable and the insignificance of the independent variable. Recently, Zhao et al. (2010) demonstrated that the significant relationship between independent variable and dependent variable is not necessary and can be misleading. This is because it represents the total effect of the sum of direct and indirect effects, including the mediator, and that mediation must be only established by the existence of an indirect effect. Put simply, to demonstrate mediation “all that matters is that the indirect effect is significant” (Zhao et al., 2010, p. 204). Thus, the following two models of analysis are employed to test for the mediation effect of financial performance, according to the main hypothesis in this study.where () is a constant and () are the parameters for the explanatory variables. The subscript (i) refers to the firm number and the subscript (t) denotes the time period. () is the unobservable individual heterogeneity, and () is the remainder disturbance or the usual disturbance in the regression model that varies with individual units and time. The Hausman specification test (Hausman, 1978) was applied to test for whether corporate social responsibility, institutional ownership and return on assets can be considered as endogenous variables or not. Following the recommendation of Gujarati (2003), the predicted values of corporate social responsibility, institutional ownership and return on assets were estimated and used with original values as well as other control variables as explained above. The χ2-statistics for the predicted values of corporate social responsibility and return on assets, as reported in Table 4, were not significant. Thus, corporate social responsibility, institutional ownership and return on assets can be treated as exogenous variables.