RESEARCH DESIGN AND METHODOLOGY
4.5. Multiple Regression Analysis.
4.5.2. Model Specification of the relationship between CSRD, CFP and IO
After the conceptual framework is constructed (see Figure 3.1 in chapter three) and the hypotheses are presented, the next step is to construct the multiple regression models for hypotheses testing procedures. Five multiple regression equation models are performed to examine the relationship between CSRD, CFP and IO. Utilized panel data analysis combines cross-sectional and time series data (for detailed discussion see section 4.7.2). Each model specification is presented in the following sections.
Model 1: Relationship between CSRD and CFP
In this model a multiple regression model is constructed to examine the relationship between CSRD and CFP. Three alternative dependent variables are used as measures of CFP, one independent variable and six control variables are also used to estimate the following multiple regression equation model:
CFPjt = β0 + β1CSRDjt + β2BETAjt + β2LEVjt + β3 LSIZEjt + β4 LSALESjt + β5ATRjt
+ β6EPSjt + εjt (4.1)
Where:
CFPjt : three alternatives of CFP variables presented by ROA, Rijt and Qjt
CSRDjt:: CSRD scores value of company j at period t. BETAjt : the systematic risk of company j at period t. LEVjt :total debt to total assets of company j at period t.
LSIZEjt: measured by natural logarithm total assets of company j at period t.
LSALESjt: measured by natural logarithm total sales of company j at period t.
ATRjt : ratio of total sales to total assets of company j at period t.
EPSjt :ratio of net earnings to number of shares outstanding of company j at period t.
152
Model 2: Relationship between Dimensions of CSRD and CFP.
This model is constructed to examine the relationship between dimensions of CSRD and CFP. Three alternative variables are used as measures of CFP as dependent variables, four dimensions of CSRD as independent variables and six control variables are also used to estimate the following multiple regression equation model:
CFPjt = β0 + β1 MPLDjt + β2COMDjt+ β3PRODjt + β4ENVDjt + β5BETAjt +
β6LEVjt + β7 LSIZEjt + β8 LSALESjt + β9ATRjt + β10EPSjt + εjt (4.2)
Where:
CFPjt : three alternatives of CFP variables presented by ROA, Rijt and Qjt
MPLDjt : score value of employee relations disclosure of firm j at period t.
COMDjt : score value of community involvement disclosure of firm j at period t.
PRODjt : score value of product disclosure of firm j at period t.
ENVDjt : score value of environment disclosure of firm j at period t.
BETAjt : the systematic risk of firm j at period t. LEVjt : total debt to total assets of firm j at period t.
LSIZEjt: measured by natural logarithm total assets of firm j at period t.
LSALESjt: measured by natural logarithm total sales of firm j at period t.
ATRjt : ratio of total sales to total assets of firm j at period t.
EPSjt : ratio of net earnings to number of shares outstanding of firm j at period t.
εjt : error term.
Model 3: Relationship between CSRD and IO
This model is constructed to examine the relationship between CSRD and IO. One dependent variable as a measure of institutional ownership, one independent variable represented by CSRD variable and seven control variables are also used to estimate the following multiple regression equation model:
153 PERCIOjt = β0 + β1CSRDjt + β2BETAjt + β3LEVjt + β4 LSIZEjt + β5 LSALESjt +
β6ATRjt + β7EPSjt + εjt (4.3)
Where:
PERCIOjt: Percentage of shares held by institutional investors in firm j at period t.
CSRDjt:: CSRD scores value of firm j at period t. BETAjt : the systematic risk of firm j at period t.
LEVjt : total debt to total assets of firm j at period t.
LSIZEjt: measured by natural logarithm total assets of firm j at period t.
LSALESjt: measured by natural logarithm total sales of firm j at period t.
ATRjt : ratio of total sales to total assets of firm j at period t.
EPSjt : ratio of net earnings to number of shares outstanding of firm j at period t.
εjt : error term.
Model 4: Relationship between Dimension of CSRD and IO.
This model is constructed to examine the relationship between dimensions of CSRD and IO. There is one dependent variable represented by percentage of shares held by institutional investors, four dimensions of CSRD as independent variables and six control variables to estimate the following multiple regression equation model: PERCIOjt = β0 + β1MPLDjt + β2COMDjt+ β3PRODjt + β4ENVDjt + β5BETAjt +
β6LEVjt + β7LSIZEjt + β8LSALESjt + β9ATRjt + β10EPSjt + εjt (4.4)
Where:
PERCIOjt: Percentage of shares held by institutional investors in firm j at period t.
MPLDjt: score value of employee relations disclosure of firm j at period t.
COMDjt: score value of communityinvolvement disclosure of firm j at period t.
PRODjt : score value of product disclosure of firm j at period t.
154 BETAjt : the systematic risk of firm j at period t.
LEVjt : total debt to total assets of firm j at period t.
LSIZEjt: measured by natural logarithm total assets of firm j at period t.
LSALESjt: measured by natural logarithm total sales of firm j at period t.
ATRjt : ratio of total sales to total assets of firm j at period t.
EPSjt : ratio of net earnings to number of shares outstanding of firm j at period t.
εjt : error term.
Model 5: Relationship between CSRD and IO on CFP
This model is constructed to examine the relationship of both CSRD and IO on CFP. Three measures of CFP are used, namely return on assets (ROA), stock return (Ri) and Tobin‘s q ratio (Q), two independent variables and six control variables are also used to estimate the following multiple regression equation model:
CFPjt = β0 + β1CSRDjt + β2PERCIOjt + β3BETAjt + β4LEVjt + β5 LSIZEjt +
β6LSALESjt + β7ATRjt + β8EPSjt + εjt (4.5)
Where:
CFPjt : three alternatives of CFP variables presented by ROA, Rijt and Qjt
CSRDjt:: CSRD score value of firm j at period t.
PERCIOjt : percentage of shares held by institutional investors of firm j at period t.
BETAjt : the systematic risk of firm j at period t. LEVjt : total debt to total assets of firm j at period t.
LSIZEjt: measured by natural logarithm total assets of firm j at period t.
LSALESjt: measured by natural logarithm total sales of firm j at period t.
ATRjt : ratio of total sales to total assets of firm j at period t.
EPSjt : ratio of net earnings to number of shares outstanding of firm j at period t.
155 The equation of these regressions will be used on the panel data comprising cross sectional and time series data observations. The panel data usually gives the researcher a large number of data points increasing the degree of freedom and reducing collinearity among the independent variables while also improving statistical estimates efficiency (Hsiao, 2003). The panel data is also utilized to analyze the dynamic change and to improve in identifying the measured effect that cannot be obeyed in pure time series or cross-section data. The other benefit of panel data over cross-sectional data or time series data is that it enables the study of more complicated models, for instance, phenomena such as the scale of economics and technological change (Gujarati, 2003).