The influence of firm size on the corporate social responsibility - corporate financial performance relationship

This thesis explores the impact of firm size on the relationship between corporate social responsibility (CSR) and corporate financial performance (CFP). It examines the relationships between CSR and CFP (and the reverse relationship) and tests whether there is a change in that relationship depending on the size of the firm. It evaluates whether stakeholder theory, which predicts that good CSR ratings result in better CFP, holds in the US and the UK over an 11-year period. The study also considers the slack resources theory, which posits that in order to achieve a good CSR rating, firms need to have good CFP in order to afford to carry out the CSR activities on which the ratings are based. The resource-based view (RBV) is also evaluated in this study to consider whether firm size has an impact on CSR and CFP. Two datasets are used, the largest 1,180 firms in the US and 325 firms in the UK listed in their respective stock exchanges over a period of eleven years (2007 to 2017). Whilst the US CSR-CFP relationship has been studied extensively previously, the UK has not. Equally, with the exception of Orlitzky’s (2001) US-based study, the specific role of firm size has not been studied, and his study only addressed firm size as a mediator in the relationship, not a moderator. This study addresses both aspects of firm size on the CSR-CFP relationship. A traditional ordinary least squares (OLS) regression methodology was used to test the relationships initially. The primary reason to use OLS was so that the results could be used to compare with previous studies. Both the US and UK findings were statistically significant and negative, refuting much prior literature. However, given the issue of simultaneity, these results are undermined and so a more novel approach was subsequently adopted. Structural Equation Modelling (SEM) was used to derive three new composite latent constructs to depict CSR, CFP and firm size for each country, from an iterative evaluation of thirty different variables. This resulted in different variables and different weightings of variables for the constructs for each country. The CSR-CFP and firm size relationship were re-tested using these constructs. Both countries demonstrated positive relationships between CSR and CFP (and the reverse). Firm size did alter these relationships, however, the magnitude of the effect of firm size was small. This research has contributed to the CSR-CFP field in various ways. The main contribution is methodological, as this thesis has introduced an approach which has not been widely used in the CSR-CFP field by developing multivariate latent constructs to encapsulate the multi-faceted nature of the CSR, CFP and firm size using SEM.
A second contribution to knowledge is in the specific role of firm size in the CSR-CFP relationship which has hitherto not been specifically addressed. It has concluded that firm size can affect the relationship in some cases, although not to a degree which might have previously been assumed – the overall magnitude of the effect of firm size is small. A final contribution is theoretical: by using constructing country-specific multivariate constructs to reflect individual country jurisdictions, it has been proven that stakeholder and slack resources theories hold in different countries. This suggests a wider applicability of these theories to other countries other than the US where the majority of studies have been previously focused. This thesis, as all research, has some limitations. It examines only two countries, hence the generalisability of the findings outside those two countries may be limited. It also examines all industries together, and hence the findings of individual industries could differ from this overall picture. Equally, this research is taken at a point in time, using one data source: different results could be obtained using data from different periods or indeed from different data sources. These industry, time and data source effects could affect the variables used to construct the latent constructs, which could also alter the findings.