ESG Scores

ESG evaluation using algorithms and Scores

The evaluation conducted by ESG assessment agencies reflects the implementation of the ESG model, relying on quantified data for assessment purposes. This process requires a substantial budget for data collection and pre-processing. To ensure the ESG evaluation yields useful outcomes, adjustments are made quarterly based on specific events, and continuous data work is performed throughout the year, demanding significant time and effort to disclose the final evaluation results. This study proposes a novel approach to ESG scoring that maximizes the technical advantages of unstructured analysis(Yang and Yang, 2022).

In comparison to existing evaluation methods, the unstructured text quantification method simplifies data collection and provides excellent processing speed, enabling quick analysis of vast datasets. In addition, by utilizing programmed algorithmic calculations, the occurrence of human errors is minimized, and the flexibility of analysis allows for granular ESG scores to be constructed down to the minutest details. , and the flexibility of analysis allows ESG scores to be constructed as penal data down to the smallest detail. Based on these excellent technical characteristics, the characteristics of this ESG evaluation method are as follows.

The distinctive characteristics of this ESG evaluation method include the substantial reduction in data preprocessing time and the minimization of data errors as compared to existing methods. For instance, when considering the K-ESG guidelines, which consist of 61 fundamental diagnostic items, divided into 17 environmental items, 22 social items, 17 governance items, and 5 information disclosure items, the evaluation becomes more extensive in the case of KCGS, encompassing 265 basic evaluation items. However, structured ESG evaluation data often faces challenges and inconsistencies due to variations between industries(Guedhami, Louton, Saraoglu and Zheng, 2022; Liu, Wu, Wu, Fu, and Huang, 2021). Even when data is structured, issues such as format mismatches, errors, or content omissions can hinder analysis(Macpherson et al., 2021; Van and Nijssen, 2021). In the case of KCGS, it encompasses a more extensive set of elements, totaling 265 basic evaluation items. However, ESG evaluation data often doesn't align with the characteristics of structured data, and inconsistencies may arise due to variations between industries. Even when data is structured, issues such as format mismatches, errors, or content omissions can pose challenges for analysis. In contrast, this evaluation method derives ESG scores by collecting unstructured data from ESG reports, simplifying data collection and reducing the likelihood of data errors compared to existing evaluation methods.

In contrast, this evaluation method extracts ESG scores by collecting unstructured data from ESG reports, simplifying data collection and reducing the likelihood of errors. Algorithmic analysis enables rapid processing of large volumes of text data, which is particularly valuable considering the exponential growth in digitized corporate information. Manual analysis of ESG reports, which can span 200 pages, is time-consuming, but this study ensures that the interpretation of unstructured text and calculation of ESG scores for each report can be accomplished in under 10 seconds. Consequently, the analysis of 500 texts can be completed in less than an hour by utilizing programming. By employing algorithms for ESG evaluation, human errors such as duplications(Guedhami et al., 2022; Liu et al., 2021), omissions, and input mistakes are mitigated(Cort and Esty, 2020), even when handling extensive data volumes that surpass human capacity for comparison(Macpherson, Gasperini and Bosco, 2021; Van and Nijssen, 2021). The efficient processing of these repetitive and sequential tasks through programming allows for the rapid derivation of ESG scores.

For instance, the environmental (E) score for Samsung Electronics increased from 0.16 in 2011 to 0.71 in 2021, as demonstrated in <Figure 10>. The governance (G) score for KB Financial Group exhibited an upward trend, rising from 0.14 in 2011 to 0.56 in 2021, as shown in <Figure 11>. By applying the ESG score calculation algorithm developed in this study to other companies, valuable time series data can be extracted. These extracted time series data enable empirical analysis to investigate the performance implications of corporate ESG management or competitiveness, facilitating generalization to other companies. In this way, if the ESG score calculation algorithm developed in this study is applied to other companies, time series data can be extracted. Therefore, the extracted time series data can be used to study the performance implications of corporate ESG management or competitiveness through empirical analysis for generalization to companies.

<Figure 10> Example of ESG Score: Samsung Electronics

<Figure 11> Example of ESG Score: Korean Bank

The current ESG score needs to evolve in a way that reflects the cross-industry influence and weighting characteristics of each ESG activity. Currently, the score is solely based on text similarity analysis between the contents of ESG management reports and ESG best practices. While companies benefit from rich content that aligns with ESG best practice standards, those with limited disclosed keywords, influenced by industry characteristics, may face unfavorable outcomes. Therefore, the ESG score necessitates a weighted calculation method that considers company size and industry-specific characteristics, similar to the approach employed by ESG evaluation agencies. However, excessive artificial intervention in the weighting evaluation method may compromise the reliability of results. Thus, careful consideration should be given to the evaluation factors for each item, incorporating a comprehensive review of methodologies adopted by evaluation agencies.

Reference

Cort, T. and Esty, D. (2020), “ESG Standards: Looming challenges and pathways forward,” Organization & Environment, 33(4), 491-510.

Guedhami, O., Louton, D., Saraoglu, H. and Zheng, Y. (2022), ESG Investing: A Decision-Making Paradox?. The Journal of Impact and ESG Investing. https://doi.org/10.3905/jesg.2022.1.054

Liu, X., Wu, H., Wu, W., Fu, Y. and Huang, G. Q. (2021), Blockchain-enabled ESG reporting framework for sustainable supply chain. https://doi.org/10.1007/978-981-15-8131-1_36

Macpherson, M., Gasperini, A. and Bosco, M. (2021), Artificial Intelligence and FinTech Technologies for ESG Data and Analysis. http://dx.doi.org/10.2139/ssrn.3790774

Van D. E., J. and Nijssen, S. (2021), Extracting ESG data from business documents. http://hdl.handle.net/2078.1/thesis:30732

Yang, B. M. and Yang, O. S. (2022), “Assessing the Effect of Dynamic Capabilities on the ESG Reporting and Corporate Performance Relationship With Topic Modeling: Evidence From Global Companies,” Frontiers in Psychology, 13.