<p><span class="p-body"><a href="https://www.sps.nyu.edu/about/academic-divisions-and-departments/preston-robert-tisch-institute-for-global-sport.html">Tisch Institute</a> Clinical Assistant Professor <a href="https://www.sps.nyu.edu/faculty-directory/21710-nico-rosamilia.html">Nico Rosamilia</a>&nbsp;recently coauthored a peer-reviewed academic article entitled </span><i><span class="p-body">AI-driven ESG Investing: Robust Portfolios with Dynamic Ensemble Learning </span></i><span class="p-body">that&nbsp;introduces the Dynamic Ensemble Learning Strategy (DELS), a machine learning framework designed to integrate Environmental, Social, and Governance (ESG) considerations into portfolio construction while optimizing financial performance.</span></p> <p><span class="p-body">The study, published in <a href="https://www.springernature.com/gp">Springer Nature</a>, addresses two persistent challenges in sustainable investing: the heterogeneity of ESG information and the difficulty of capturing nonlinear relationships between sustainability indicators and financial returns. Unlike conventional screening approaches, DELS incorporates ESG characteristics directly into the asset selection process, seeking to reconcile sustainability objectives with competitive risk-adjusted performance.</span></p> <p><span class="p-body">The analysis examines 578 companies from the STOXX Europe 600 universe, using monthly stock returns from 2021 to 2024 and 24 ESG and financial indicators obtained from Refinitiv. Three portfolio construction objectives are investigated: the Sharpe Ratio, emphasizing total risk-adjusted returns; the Treynor Ratio, focusing on systematic risk; and the Calmar Ratio, prioritizing drawdown management. The resulting portfolios are evaluated against an iShares STOXX Europe 600 ETF, an equally weighted benchmark, and a Fama–French-inspired portfolio.</span></p> <p><span class="p-body">The study develops a volatility-based recalibration mechanism that adjusts the relative importance of the three performance measures according to anticipated market volatility, captured through the VSTOXX index. Applied over 2022–2024, this annually rebalanced strategy generated an annualized return of 9.59%, compared with 0.76% for the benchmark ETF and 4.20% for a static balanced DELS portfolio.</span></p> <p><span class="p-body">The authors' findings suggest that machine learning can improve sustainable portfolio construction by combining multidimensional ESG information with financial performance criteria. DELS contributes an interpretable, consensus-driven selection mechanism capable of adapting to different investment objectives and market conditions, offering an alternative to conventional ESG screening and static portfolio optimization.</span></p>