Channel of communication
may disclose CSR information through several channels, including annual
reports, separate CSR reports and the company websites. Beyond that, companies
can channel information through media channels, for example using press
releases. In order to limit the scope of the research and pertain a more even
frame of reference across the sampled companies, this study will focus solely
on annual reports and supplementary CSR/sustainability reports for the
accounting year 2017. Scholars consider these channels the most important for
CSR disclosure (Gamerschlag
et al., 2011).
CSRD variable constructed in this study will consider whether the predetermined
keywords are present in the published reports, but will not consider the
frequency. In practical terms, each company will get a score based on the
proportion of keywords disclosed in the reports. Thus, for a company that
discloses information about all 32 keywords, the dependent variable will be 100
Independent and Control Variables
Data for independent and control variables will be collected from an
array of readily available resources, including annual reports, ORBIS and
The independent and control variables included in this research is based
on a review of existing literature on the topic. The hypotheses tested relates
to ownership concentration, proportion of independent directors, proportion of
female directors and degree of institutional ownership. All hypotheses. Previous
studies (eg., Ali et al., 2017) suggest that
company size, industry profile and financial performance affect CSRD. The study
will, therefore, incorporate these variables in the analysis.
variables are constructed in line with previous research. A description of how
each variable is constructed is outlined in Table 2.
5.3 Data Analysis
To test the outlined hypotheses, the study will apply regression
analysis. The study will apply multiple regression, using ordinary least
squares (OLS), to regress the independent variables against the dependent
variable (see Table 2 for explanation of variables). This analysis will be
conducted using statistical software such as STATA or JMP.
to the multiple regression, each of the variables will be examined using graphs
and descriptive statistics, in order to identify any possible outliers,
skewness or multicollinearity (Agresti, Franklin and Klingenberg, 2017).