Università degli studi di Modena e Reggio Emilia

Abstract
The aim of this paper is to compare two philosophically different strategies to obtain (Environmental, Social, Governance) ESG-compliant portfolios and to propose a third one. We confront the risk-adjusted performance of three optimal portfolios: the first results from optimization on an ESG-screened sample, the second is obtained by including a portfolio ESG-score constraint in the optimization on an unscreened sample, the third results from our proposal of taking pros of both by optimizing with an ESG constraint (so as to reach a target) over a slightly screened sample (so as to eliminate companies with lowest sustainability). The optimization approach rests on minimizing portfolio residual risk, it does not require a balanced dataset and it is implemented with Bloomberg ESG scores over a stock sample from the EURO STOXX Index in the period January 2007 – August 2022. Two are the main results. First, the performance of an ESG-compliant portfolio depends not only on the strategy taken (optimizing after screening, constraining optimization, mixed strategy) but also on the feature of the initial investment set in terms of relationship between ESG scores and the risk-reward (monotone, convex, concave). Second, the comparative performance of the three strategies over time does not change much with the financial cycle. The most interesting implication is that, the third strategy we propose, allows sustainable investors to do well by doing good in the presence of both a convex and a monotone investment set, being thus able to attract a larger pool of investors towards ESG-compliant optimal portfolios.