Kempf, Alexander and Peer Osthoff. "The Effect of Socially Responsible Investing on Financial Performance." Working Paper, University of Cologne, Germany. June, 2006.

Compares the returns of two trading strategies for the 1992-2004 time period, one focused on companies with high social responsibility scores, the other focused on companies with low ones.

Using KLD data, the authors construct an overall social performance score based on sub-criteria in the Community, Diversity, Employee Relations, Environment, Human Rights, and Product categories, as well as sub-scores for the individual categories. Examining these independent variables the authors find that they are independent, with no evidence of multicolinearity (a weak correlation is found between alcohol and tobacco, r=0.37).

Portfolios are formed using two methods, a best-in-class approach, and a negative screening policy. For the best-in-class approach, strong and weak social performers are selected by SIC code, then value-weighting portfolios by industry. "Thus, the best-in-class approach mitigates the sensitivity of financial performance to certain industries and leads to industry-balanced investment portfolios." The negative screening policy eliminates "all the stocks involved in controversial business areas." This portfolio is then compared with a portfolio of the excluded companies.

The dependent variable is stock returns, using the CRSP database. These are evaluated using a Carhart Model in which stock returns are explained by beta, size, valuation, and momentum.

The study finds that the lowly-rated portfolios generally underperform the highly-rated ones after adjusting both for the influence of the Carhart factors. This appears to be due primarily to underperformance by the lowly-rated companies, as the highly-rated portfolios performed roughly in line with the broader market. Also finds that "negative screening did not lead to any significant performance effect," and that "the performance of the socially responsible portfolios are never significantly negative."

This is an exceptional study given its long time period (including performance for the first half of the 2000s), careful industry adjustments, and use of risk models in evaluating returns.