Abstract

This paper presents GreenGuard-AU, a pilot RegTech framework for preliminary greenwashing screening that operates under severe label scarcity. The system ingests 38 engineered features organised across four modalities (textual, longitudinal, financial, and sector-weighted) and classifies 633 company-year observations from the Australian securities market (2021 to 2024). Of the 130 positive labels, 19 derive from confirmed ASIC enforcement actions; the remaining 111 are constructed via ESG rating divergence analysis. The ensemble achieves an AUC-ROC of 80.35% (95% bootstrap CI: 75.73% to 84.70%, p < 0.01 by permutation test, n = 100; none of 100 shuffled-label runs matched the real AUC). Nested cross-validation with held-out threshold tuning yields an identical point estimate (AUC 80.35% ± 1.47%), consistent with expected discrimination stability across tuning splits. A transformer baseline reaches 63.7% AUC under the same conditions, consistent with the known failure mode of deep architectures in low-data tabular settings. SHAP attribution identifies longitudinal linguistic consistency, specifically the feature commitment type change (mean vague-verb frequency across reporting years), as the highest-ranked predictor in 4 of 5 cross-validation folds (Kendall’s τ = 0.72 across fold pairs). The framework is fully interpretable, trains in under five seconds on commodity hardware, and is intended as a pilot screening tool rather than an enforcement-ready system. The paper discusses label construction validity, circular leakage from ESG divergence features, and the conditions under which these results would generalise.

References

• Global Sustainable Investment Alliance. Global sustainable investment review 2022. Technical report, GSIA, 2023. URL https://www.gsi-alliance.org/members-resources/gsir2022/. Reports global sustainable investing assets at USD 30.3 trillion.
• Magali A. Delmas and Vanessa C. Burbano. The drivers of greenwashing. California Management Review, 54(1):64–87, 2011. https://doi.org/10.1525/cmr.2011.54.1.64.
• Australian Securities and Investments Commission. Federal Court orders Vanguard Investments Australia Ltd to pay $12.9 million penalty for greenwashing. Media release 24-213mr, ASIC, 2024. URL https://asic.gov.au/about-asic/news-centre/find-a-media-release/2024-releases/24-213mr/.
• Australian Securities and Investments Commission. Federal Court orders Mercer Superannuation to pay $11.3 million penalty for greenwashing. Media release 24-173mr, ASIC, 2024. URL https://asic.gov.au/about-asic/news-centre/find-a-media-release/2024-releases/24-173mr/.
• Thomas P. Lyon and A. Wren Montgomery. The means and end of greenwash. Organization & Environment, 28(2):223–249, 2015. https://doi.org/10.1177/1086026615575332.
• Kent Walker and Fang Wan. The harm of symbolic actions and green-washing: Corporate actions and communications on environmental performance and their financial implications. Journal of Business Ethics, 109(2):227–242, 2012. https://doi.org/10.1007/s10551-011-1122-4.
• Catherine A. Ramus and Ivan Montiel. When are corporate environmental policies a form of greenwashing? Business & Society, 44(4):377–414, 2005. https://doi.org/10.1177/0007650305278120.
• Christopher Marquis, Michael W. Toffel, and Yanhua Zhou. Scrutiny, norms, and selective disclosure: A global study of greenwashing. Organization Science, 27(2):483–504, 2016. https://doi.org/10.1287/orsc.2015.1039.
• Julia A. Bingler, Mathias Kraus, Markus Leippold, and Nicolas Webersinke. Cheap talk and cherry-picking: What ClimateBERT has to say on corporate climate risk disclosures. Finance Research Letters, 47:102776, 2022. https://doi.org/10.1016/j.frl.2022.102776.
• Dominik Stammbach, Nicolas Webersinke, Julia A. Bingler, Mathias Kraus, and Markus Leippold. A dataset for detecting real-world environmental claims. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (ACL), pages 3971–3995, 2022. https://doi.org/10.18653/v1/2022.acl-srw.35.
• Allen H. Huang, Hui Wang, and Yi Yang. FinBERT: A large language model for extracting information from financial text. Contemporary Accounting Research, 40(2):806–841, 2023. https://doi.org/10.1111/1911-3846.12832.
• Xingqiang Du. How the market values greenwashing? Evidence from China. Journal of Business Ethics, 128(3):547–574, 2015. https://doi.org/10.1007/s10551-014-2122-y.
• Olga Hawn and Ioannis Ioannou. Mind the gap: The interplay between external and internal actions in the case of corporate social responsibility. Strategic Management Journal, 37(13):2569–2588, 2016. https://doi.org/10.1002/smj.2464.
• Eun-Hee Kim and Thomas P. Lyon. Greenwash vs. brownwash: Exaggeration and undue modesty in corporate sustainability disclosure. Organization Science, 26(3):705–723, 2015. https://doi.org/10.1287/orsc.2014.0960.
• Florian Berg, Julian F. Kölbel, and Roberto Rigobon. Aggregate confusion: The divergence of ESG ratings. Review of Finance, 26(6):1315–1344, 2022. https://doi.org/10.1093/rof/rfac033.
• Arsha Nagrani, Shan Yang, Anurag Arnab, Aren Jansen, Cordelia Schmid, and Chen Sun. Attention bottlenecks for multimodal fusion. In Advances in Neural Information Processing Systems (NeurIPS), volume 34, pages 14200–14213, 2021. https://doi.org/10.48550/arXiv.2107.00135.
• Bryan Lim, Sercan O. Arık, Nicolas Loeff, and Tomas Pfister. Temporal fusion transformers for interpretable multi-horizon time series forecasting. International Journal of Forecasting, 37(4):1748–1764, 2021. https://doi.org/10.1016/j.ijforecast.2021.03.012.
• Leo Breiman. Random forests. Machine Learning, 45(1):5–32, 2001. https://doi.org/10.1023/A:1010933404324.
• Tianqi Chen and Carlos Guestrin. XGBoost: A scalable tree boosting system. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 785–794, 2016. https://doi.org/10.1145/2939672.2939785.
• Jerome H. Friedman. Greedy function approximation: A gradient boosting machine. Annals of Statistics, 29(5):1189–1232, 2001. https://doi.org/10.1214/aos/1013203451.
• Alik Sokolov, Jonathan Mostovoy, Jack Ding, and Luis Seco. Building machine learning systems for automated ESG scoring. The Journal of Impact and ESG Investing, 1(3):39–50, 2021. https://doi.org/10.3905/jesg.2021.1.010.
• Riccardo Guidotti, Anna Monreale, Salvatore Ruggieri, Franco Turini, Fosca Giannotti, and Dino Pedreschi. A survey of methods for explaining black box models. ACM Computing Surveys, 51(5):1–42, 2018. https://doi.org/10.1145/3236009.
• Cynthia Rudin, Chaofan Chen, Zhi Chen, Haiyang Huang, Lesia Semenova, and Chudi Zhong. Interpretable machine learning: Fundamental principles and 10 grand challenges. Statistics Surveys, 16:1–85, 2022. https://doi.org/10.1214/21-SS133.
• Scott M. Lundberg and Su-In Lee. A unified approach to interpreting model predictions. In Advances in Neural Information Processing Systems (NeurIPS), volume 30, pages 4765–4774, 2017. https://doi.org/10.48550/arXiv.1705.07874.
• Australian Securities and Investments Commission. Regulatory guide 234: Advertising financial products and services (including credit): Good practice guidance. Regulatory guide, ASIC, 2019. URL https://asic.gov.au/regulatory-resources/find-a-document/regulatory-guides/rg-234-advertising-financial-products-and-services-including-credit-good-practice-guidance/.