Intro

As member of the sustainable finance team, I’ve had the opportunity to contribute to this comprehensive ESG paper that delves into the complexities of corporate reporting. This paper aims to shed light on the challenges organizations face in accurately and transparently disclosing their environmental, social, and governance (ESG) performance. By examining the current landscape and identifying key obstacles, we hope to contribute to the development of more effective and standardized reporting practices. In particular, we focussed on the anomaly detection issue, in order to find outliers scores.

For the sake of brevity, I only display a few sections here. Please reach out to me if you are interested in reading the whole work.

Table of Content

  1. ESG data
    • 1.1 Growing importance
    • 1.2 Trust Issue
    • 1.3 Refinitiv’s ESG Data
  2. A tool that harnesses Al to detect and explain anomalies
    • 2.1 How does anomaly detection work?
    • 2.2 Anomaly detection example
    • 2.3 Low anomaly magnitude
    • 2.4 High anomaly mgnitude
  3. Applications of anomaly detection
    • 3.1 Portfolio and individual holding review
    • 3.2 Stock weight adjustment
  4. Conclusion and future work
  5. Biographies

1.1 Growing Importance

As issues such as climate change, social inequality, diversity and inclusion are placed in the spotlight, investors, corporations and banks alike are realising the importance of sustainable investment. The demand for global change is fuelled by environmental, social and governance (ESG) data, an increasingly important element in the reporting and analysis process. As millennials lead a drive to invest in ESG-focused companies, organisations seek to improve their ESG scores and performance.

Companies with strong ESG performance have demonstrated higher returns on their investments, lower risks and better resiliency during a crisis. From 2018 to 2020, investments in ESG strategies grew by 42%, and a third of assets under management are invested in ESG strategies. LSEG (London Stock Exchange Group) recognises the critical importance of transparent, accurate and comparable ESG data and analytics for the financial industry, a key focus in 2022 and beyond, with the proven potential for excess returns.

Figure 1 below shows the structure of the ESG dataset. A company’s ESG score is calculated as a weighted sum of three pillar scores: environmental, social and governance. The pillars are further broken down into multiple category scores shown on the right-hand side. Categories are further split into 186 metrics in total which are based on more than 500 ESG measures. Industry group-specific weights define how much impact a category has on the ESG score. For example, in emission-intensive industry groups such as oil and gas, the emissions category has a greater weight than the financial industry group.

2.1 How does anomaly detection work?

The goal of anomaly detection is to highlight unusual changes in ESG scores of a company compared to a group of peer companies in the same industry. Anomalies are characterised by their magnitude and polarity. To dive deeper and verify why anomalies occur, we quantify how much each ESG category contributes to an anomaly.

3.0 Applications of anomaly detection

This scatter-plot in Figure 9 shows the distribution of all stocks within a portfolio, with each dot representing a stock. The position is defined by the anomaly magnitude and polarity, while the size of the dot is relative to its portfolio weight. (The bigger the dot size, the greater the weight). Each dot is also colour-coded to reflect the relevant industry sector.

In this example, most of the stocks are concentrated towards the left along the median line, indicating that the probability of their being anomalous is low.

The stocks which do require attention are distributed on the right-hand side of the chart. The stocks on the top right of the median line represent an increase in the scores, and the stocks on the bottom right of the median line represent a decrease in the scores.