{"paper_id":"2310.02163","title":"Navigating Uncertainty in ESG Investing","authors":["Jiayue Zhang","Ken Seng Tan","Tony S. Wirjanto","Lysa Porth"],"paper_date":"2023-10-03","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2310.02163v3","page_url":"https://thequant.space/flowcharts/navigating-uncertainty-in-esg-investing/","scores":{"empirical_rigor":6,"math_complexity":7,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes four ESG ensemble strategies, a Double-Mean-Variance (DMV) model, and ESG‑adjusted CAPM, integrating ESG scores into a reinforcement‑learning reward function for portfolio optimization.","quote":"In response to this ambiguity, we propose four ESG ensemble strategies that cater to different risk and ambiguity preference profiles, rather than providing a single, optimal ESG scoring system.","summary":"The widespread confusion among investors regarding Environmental, Social, and Governance (ESG) rankings assigned by rating agencies has underscored a critical issue in sustainable investing. To address this uncertainty, our research has devised methods that not only recognize this ambiguity but…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"A linear mean‑variance preference structure is assumed in the Double‑Mean‑Variance model for analytical tractability.","type":"model","stated":"explicit","quote":"We assume a linear mean-variance (MV) preference structure in the Double Mean-Variance (DMV) model to ensure analytical tractability and clear separation of pecuniary and non-pecuniary returns."},{"assumption":"ESG ratings from the four major providers are assumed to be available consistently for the Dow 30 from 2007 onward.","type":"data","stated":"explicit","quote":"We download daily data for DOW 30 from 06/30/2007 to 06/30/2022. This time window is chosen to ensure consistent ESG rating availability across all four major providers as ESG coverage becomes more comprehensive from 2007 onward."},{"assumption":"The reinforcement‑learning environment follows a Markov Decision Process framework.","type":"model","stated":"explicit","quote":"A foundational concept in RL is known as Markov Decision Process (MDP), which provides a framework for modeling decision-making in a stochastic environment."},{"assumption":"Actions are limited by a maximum allowable trading proportion and are subject to transaction‑cost, leverage, and liquidity constraints.","type":"execution","stated":"explicit","quote":"Actions are constrained by practical factors such as transaction costs, leverage limits, and liquidity considerations, ensuring that the trading strategy remains realistic and implementable."},{"assumption":"The reward function adds a weighted ESG score to the portfolio return.","type":"model","stated":"explicit","quote":"Reward Function: We update the reward function by adding a weighted ESG score of the portfolio in the RL framework."},{"assumption":"ESG rating scales from different agencies are harmonized to a common scale.","type":"data","stated":"explicit","quote":"To harmonize different ESG rating scales across agencies, MSCI’s scale is t…"}]},"datasets":{"status":"available","items":[{"name":"ESG scores from Sustainalytics, RobecoSAM, MSCI, Refinitiv Asset4","provider":"Sustainalytics, S&P Global, Refinitiv, MSCI","kind":"alternative","access":"commercial","period":"mostly 2020, fallback to 2019","universe":"Dow Jones 30 stocks","quote":"The ESG data used in our study are acquired from four major rating agencies: SA (Sustainalytics), RobecoSAM (S&P Global), Asset4 (Refinitiv), and MSCI.","known_id":"refinitiv"},{"name":"Daily price and fundamentals data for Dow 30","provider":"CRSP and Compustat via WRDS","kind":"market-prices","access":"commercial","period":"06/30/2007 to 06/30/2022","universe":"Dow Jones 30 constituents","quote":"Financial data and relative features used in the model are also sourced from CRSP and Compustat via WRDS, and matched with ESG ratings.","known_id":"crsp"},{"name":"ESG ratings via Bloomberg and Thomson Reuters Eikon","provider":"Bloomberg, Thomson Reuters","kind":"alternative","access":"commercial","period":null,"universe":null,"quote":"The data from these rating agencies are downloaded from the Thomson Reuters Eikon database (for Asset4) and Bloomberg (for RobecoSAM, Sustainalytics, and MSCI scores) database respectively.","known_id":"bloomberg"}],"known":[{"id":"refinitiv","name":"Refinitiv / LSEG (Datastream, Eikon, Tick History)","provider":"LSEG","kind":"market-prices","access":"commercial","quote":"The data from these rating agencies are downloaded from the Thomson Reuters Eikon database (for Asset4) and Bloomberg (for RobecoSAM, Sustainalytics, and MSCI scores) database respectively.","mentions":3,"used_in":2},{"id":"bloomberg","name":"Bloomberg","provider":"Bloomberg L.P.","kind":"market-prices","access":"commercial","quote":"The data from these rating agencies are downloaded from the Thomson Reuters Eikon database (for Asset4) and Bloomberg (for RobecoSAM, Sustainalytics, and MSCI scores) database respectively.","mentions":2,"used_in":2},{"id":"wrds","name":"WRDS","provider":"Wharton Research Data Services","kind":"other","access":"commercial","quote":"These ratings cover large-cap U.S. equities and are obtained via subscription to Wharton Research Data Ser-vices (WRDS).","mentions":2,"used_in":1},{"id":"compustat","name":"Compustat","provider":"S&P Global","kind":"fundamentals","access":"commercial","quote":"Financial data and relative features used in the model are also sourced from CRSP and Compustat via WRDS, and matched with ESG ratings based on firm identifiers and reporting periods.","mentions":1,"used_in":1},{"id":"crsp","name":"CRSP","provider":"Center for Research in Security Prices","kind":"market-prices","access":"commercial","quote":"Financial data and relative features used in the model are also sourced from CRSP and Compustat via WRDS, and matched with ESG ratings based on firm identifiers and reporting periods.","mentions":1,"used_in":1}],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"daily"},"transaction_costs":{"status":"available","modeled":"unclear","headline_net_of_costs":null,"detail":null,"quote":null,"source":"conflict","rule":"R4b","candidates":{"front_matter":"no","evidence_card":"partial","evidence_card_quoted":true},"conflict_kind":"graded","card_detail":null,"card_quote":"Actions are constrained by practical factors such as transaction costs, leverage limits, and liquidity considerations, ensuring that the trading strategy remains realistic and implementable.","meaning":"the platform classifier and the paper-text extraction disagree, so no value is asserted"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"In the short‑term, incorporating ESG considerations reduces financial performance, as shown by lower Sharpe Ratios in the one‑year test set.","kind":"performance","quote":"decisions generally leads to a lower short-term financial performance, as evidenced by reduced Sharpe Ratios in the one-year test set."},{"finding":"Long‑term, firms with strong ESG performance may achieve better financial resilience and returns.","kind":"performance","quote":"However, in the long term, companies with a strong ESG performance are better positioned to manage policy or sustainability risks, potentially enhancing their financial resilience and performance over time."},{"finding":"Performance differences between Type‑I and Type‑U investors remain ambiguous, reflecting a trade‑off between ESG‑related return loss and uncertainty compensation.","kind":"other","quote":"We observe that the performance difference between Type-I and Type-U investors remains ambiguous. This reflects the trade-off between the loss of returns due to ESG considerations and the compensation for uncertainty through different ESG ratings."},{"finding":"Ensemble ESG strategies are needed to address rating heterogeneity and investor ambiguity.","kind":"theoretical","quote":"In response to this ambiguity, we propose four ESG ensemble strategies that cater to different risk and ambiguity preference profiles, rather than providing a single, optimal ESG scoring system."}],"reported_performance":null,"out_of_sample":{"method":"walk-forward","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":24,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T18:21:22Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}