Milevsky, Moshe A., Andrew Aziz, Al Goss, Jane Thomson, and David Wheeler. "Cleaning a Passive Index: How to Use Portfolio Optimization to Satisfy CSR Constraints." Journal of Portfolio Management, Spring 2006.

This study demonstrates how modern optimization techniques can be used to maintain the risk/reward ratio of a socially screened portfolio of Canadian stocks.

The social screening techniques used are based on Thomson and Wheeler (2004), and are strongly focused on assessment of employee relations, which the authors explain, Thomson and Wheeler judge "likely to be especially salient to firm performance and risk minimization."

Using this screening technique, the authors progressively eliminate companies from the S&P/TSX 60 index. Interestingly, the screening method used did not have obvious sectoral impact, a very different result from typical screening practice in the U.S., where entire industries (such as gold mining and integrated oil) may be eliminated.

Using a proprietary optimization algorithm, the authors then demonstrate that tracking error can be kept low as companies are removed from the portfolio, show that tracking error rises as the number of stocks excluded increases. Mean squared error is 0.078 index points (index level of 462.28) when two stocks are excluded, 0.677 when 10 are excluded, and 2.392 when 20 are excluded.

The authors conclude that: "Our results are encouraging for those who want to implement CSR constraints and still maintain a comparable level of diversification. We find that removing companies with the lowest CSR scores does not materially impact any of the first four moments of the investment return distribution, providing the funds released by divesting from the low scoring companies are properly redistributed amongst the remaining stocks."