Creating More Stable and Diversified Socially Responsible Investment Portfolios

This study is the first to apply a robust estimation technique when constructing Socially Responsible Investing (SRI) portfolios and to highlight that the selection of the optimisation process in this industry matters. We go beyond the mean-variance Markowitz framework in order to bypass issues surrounding the significant estimation risk that causes unstable, poorly diversified and suboptimal portfolios. Using data from MSCI KLD on the social responsibility of US firms, we construct SRI portfolios which exhibit higher risk-adjusted performance, lower total risk, greater stability and diversification compared to conventional and SRI equity indexes, as well as more naïve forms of optimization. Our main conclusions are robust to a series of tests, including the use of different estimation windows, stricter screening criteria, and alternative ways of evaluating portfolio performance.