Derwall, Jeroen, Nadja Guenster, Rob Bauer, and Kees Koedijk. "The Eco-Efficiency Premium Puzzle." Financial Analysts Journal, March/April 2005.

This study examines the performance impact of using environmental ratings (from the research firm Innovest) as part of an active management strategy. The authors construct two matched value-weighted portfolios of U.S. equities, one consisting of firms with high environmental ratings, the other consisting firms with low ones. Portfolios are constructed using a "best-in-class" approach - unlike most currently marketed social investment portfolios, the basis for inclusion/exclusion is relative standing within the industry, not a fixed statistical test or exclusion criteria.

Returns were computed for the 1995-2003 time period (the Innovest data starts in July 1997 - the authors acknowledge that look-ahead bias is possible, but note that shortening the time period to 1997-2003 does not affect the results of the paper). Annualized mean return for the high-ranked portfolio was 12.2% vs. 8.9% for the low-ranked portfolio. CAPM alpha was computed as 1.29 (p<0.01) for the high-ranked vs. -1.76 (p<0.01) for the low-ranked. After adjusting for industry, the difference in these alphas grew from 3.05 to 3.82. The authors also use a Carhart-type model and find that the high-ranked portfolio's alpha of 3.98 (p<0.10) was dramatically higher than the low-ranked portfolio's -1.08. The industry-adjusted difference is again larger, rising from 5.06 to 6.04. The authors employ a variety of tests of robustness to demonstrate that the effect persists despite changes in regression variables or imposition of transaction costs.

Although this study is a backtest, the magnitude of the reported alphas are so large, even after adjusting for risk using a variety of models, that the findings deserve further attention.

This study can be downloaded here. See also the closely-related Guenster et al (2005).