We designed a machine learning algorithm that identifies patterns between ESG profiles and financial performances for companies in a large investment universe. The algorithm consists of regularly updated sets of rules that map regions into the high-dimensional space of ESG features to excess return predictions. The final aggregated predictions are transformed into scores which allow us to design simple strategies that screen the investment universe for stocks with positive scores. By linking the ESG features with financial performances in a non-linear way, our strategy based upon our machine learning algorithm turns out to be an efficient stock picking tool, which outperforms classic strategies that screen stocks according to their ESG ratings, as the popular best-in-class approach. Our paper brings new ideas in the growing field of financial literature that investigates the links between ESG behavior and the economy. We show indeed that there is clearly some form of alpha in the ESG profile of a company, but that this alpha can be accessed only with powerful, non-linear techniques such as machine learning.

No abstract

No abstract

No abstract

We assess the value added of a multifactor portfolio from a performance-agnostic point of view. First we introduce a broad general definition of factor, that encompasses usual factors like Size or Value, and then we prove that static long–short multifactor strategies (as the equal weighting of factors) are indeed factors according to our definition. This result is new in the literature and states that, by investing in a long–short static multifactor strategy, one is indeed investing into a new (synthetic) factor. Finally we test the strength of such a synthetic factor compared to each single factor by looking at its predictive power. We empirically test the equal-weighting of Value, Size, Momentum and Low Volatility in the US and Europe. Our conclusion is very clear in both regions: the equal-weighting of these four standard factors is a synthetic factor that has no predictive power on stocks’ return, while each of the factors shows clear ability to distinguish among stocks. In other words, the measure that underlies this equal-weighting of factors has zero predictive power on cross-sectional differences in stocks’ returns.