This research project deals with the challenge of developing new Key Performance Indicators (KPI) to assess the companies¿ Corporate Social Responsibility (CSR) explained as a function of Socially Responsible Investments (SRI), also known as Environmental, Social and Corporate Governance (ESG) investments. The recent attention to SRI has become crucial for regulators and governments whose directives focus on sustainable investments to generate sustainable growth. In this context, Europe has set important international agreements and investments in ¿green¿ projects. Therefore, the identification and evaluation of the degree of responsibility represents a research question of interest also for investors.
Rating agencies introduced sustainability rating based on firms' ESG business model that account for specific features such as emissions, environmental product innovations, human rights, and the companies' structure. Measuring the risk profiles of SRI becomes crucial for guidance and benchmarking. However, there is not a unified framework for ESG ratings and their reliability is strongly criticized calling for the identification of suitable KPI able to express the real impact of SRI on the firms¿ performances.
The aim of this project is twofold. First, we propose to build new KPI to assess sustainable investments using structural information on the company¿s business model as well as using balance sheet data. The KPI will be built with tree-based machine learning algorithms that detect hidden relationships and describe potential complex patterns while preserving the interpretability property. The second aim is to analyze the relationship between KPI and the main financial risks, such as market and credit risks. Market risk will be measured using companies¿ performances in terms of stocks¿ return conditional volatility whereas the impact of the KPI on credit risk is related to the firms¿ probability of default computed as an extension of the Altman¿s Z score.
This research is the first aiming to build new KPI for sustainable investments. We construct the new KPI based on balance sheet data of European companies and empirically investigate their impact on financial risk mitigation, contributing to the literature in different ways. On the one side, the construction of new KPI is targeted to express the companies¿ effort for choosing sustainable investments and to overcome the major critique of the not availability of a unique set of indicators for SRI. We study the relationship between structural data emerging from the company¿s balance sheet and the newly built KPI, using a novel approach based on machine learning, which is considered a very prominent methodology for detecting hidden relationships and describing complex patterns. However, machine learning algorithms are frequently considered as black-boxes because the process between input and output is opaque. Such a problem is sometimes a barrier to the adoption of machine learning models in regression and classification problems. A new field of literature focusing on the interpretability of artificial intelligence is becoming increasingly important with the aim to interpret and explain individual model predictions to decision-makers, end-users, and regulators. To understand how a model operates, we need to unfold the various steps in order to know how it works and which decisions it takes. We will deal with the model interpretability considering a measure of the feature importance and the partial dependency plot.
On the other side, we further analyze possible effects of the new KPI on the risk mitigation of two major financial risks, market and credit risks, exploring portfolio optimization techniques adapted to time series methodologies. The market risk for investing ¿green¿ is assessed through the analysis of such new built KPI on companies¿ performances in terms of stocks¿ return conditional volatility whereas the evaluation of the impact of the KPI on credit risk is related to a firm¿s probability of default which can be computed as an extension of the Altman¿s Z score.