The proposed research deals with the challenge of developing new risk measurement and management tools in the field of Socially Responsible Investments (SRIs), also known as Environmental, Social and Corporate Governance (ESG) investments to assess the level of Corporate Social Responsibility (CSR). In the last decade, investors as well as governments have been increasingly incorporating ESG information into the analysis of investments. For instance, the European Investment Bank contributed for 55 trillion on 'green' projects. The literature is only at the dawn. Rating agencies have recently started to rate companies according to their SRI business model. Such ratings are obtained by analyzing different features such as emissions, environmental product innovations, human rights, and the companies' structure. The understanding of risk profiles of investments becomes crucial for guidance and benchmarking. However, common metrics to assess financial performances still rely on financial accounting tools well-know impact assessment analysis and the need to develop new instruments is called for.
The target of this research is twofold: first we investigate the relationship between structural features of each company (balance sheet data and fundamental information of the company) and the assigned ESG rating using a machine learning approach, promising to assess the accuracy of the current ESG score thanks to the ability of detecting hidden (non-linear) relationship between the scores and balance sheet data.
Second, we study how the ESG rating affects the companies' performance in terms of stocks' return conditional volatility. What drives stock returns' volatility is a largely debated topic. We assume that companies' commitment in the CSR may provide an additional factor to explain conditional volatility dynamics and to capture the financial market reaction to the use of ESG ratings in response to the increasing investors¿ attention to SRI.
This research is the first aiming to empirically investigating the role of sustainability in European financial markets. We, thus, focus on the STOXX 600 index and exploit the ESG score dataset provided by Bloomberg, contributing to the literature in different ways.
On the one side, we analyze possible relationship between structural data and ESG ratings. Balance sheet data will be used in a machine learning approach to assess their role in explaining the ESG ratings. Machine learning will be able to capture non-linear dynamics in the relationship.
On the other side, we investigate the effects of ESG scores on conditional return volatility of stocks' return exploiting GJR-GARCH-type of models. The GJR-GARCH allows to capture very specific features of stock returns, i.e. volatility clustering and leverage effect, providing an accurate measure of conditional volatility. Introducing a dummy variable to measure the level of ESG effort we want to identify how sustainability impacts on volatility and therefore assess its role.
The empirical investigation aims to capture the effects of sustainability in European financial markets. First, we provide a detailed analysis of the ESG scores rated for European companies included in the index. Second, this is the first study, to the best of our knowledge, that investigates the effects of ESG ratings on the volatilities of stock returns. More specifically, according to the ESG score distribution by market sector, we identify ESG thresholds which allow to discriminate against the impact of SRI on companies' return volatility. Then, for the different thresholds specified, we study whether a change in the ESG score of a company leads to swings in its return volatility. The analysis captures the market reaction, for each different sector, to the recent diffusion of SRI. Third, we address the need to have better regulation of the ESG market for an assessment of the companies' CSR efforts. This is achieved by optimally clustering the companies working within the same sector in two groups according their level of sustainability, and test whether the ESG score differently affects the return volatility of the two clusters.