Scientific Beta

Full paper available at SSRN: Bruno, Giovanni and Goltz, Felix and Naly, Antoine. 2025. "Does ESG Information Deliver Investment Value? A High-Dimensional Portfolio Perspective" This research overview provides a new perspective on investment benefits of using ESG information (Environmental, Social, and Governance metrics).

We tested whether ESG data improves investment performance by analyzing 200+ ESG metrics. Our results show that ESG information allows to increase performance in traditional backtests but adds no value to portfolio construction when assessed in out-of-sample tests. In addition, financially optimal use of ESG data requires taking tilts towards ESG laggards on several dimensions. For investors, this means they need to set realistic expectations about what ESG integration can achieve. And when faced with claims about performance benefits, they need to require out-of-sample testing, not just traditional backtests.

Institutional investors and asset managers use an increasing number of ESG metrics when constructing their portfolios, yet a fundamental question remains unanswered: does all this sustainability data help improve portfolio performance? ESG information is inherently high-dimensional, with hundreds of metrics spanning climate impact, labour practices, board diversity, and more. We test whether this vast information set helps investors build better portfolios beyond what traditional financial data already provides.

Our approach is deliberately focussed on financial objectives. We adopt the perspective of "ESG aware" investors who care only about risk-adjusted returns, not about the ESG score of their portfolio. As one pension fund manager put it: "Having more information when making investment decisions can be informative and better than having less information." But is it? Our study finds that adding 200+ ESG metrics to standard financial data does not improve portfolio performance out of sample.

While ESG information is notoriously high-dimensional - encompassing countless issues measured in countless ways - empirical studies typically treat ESG as a monolith. They rely on either a single aggregate ESG score or focus on individual issues like carbon emissions or employee satisfaction. This not only ignores the rich, multifaceted information available to investors, but also opens the door to cherry picking metrics that showed high performance in the past. In contrast, we harness the full set of high-dimensional ESG data and test its value using out-of-sample analysis.

Why does out-of-sample testing matter? Traditional studies examine whether ESG variables earned a premium by looking backward over the entire history of returns - essentially asking, with perfect hindsight, whether an ESG tilt would have improved performance. But this tells us little about whether ESG data helps investors make better decisions going forward. We take a different approach: we evaluate ESG metrics using only the information that would have been available to investors at the time they constructed their portfolios.