In general, I think we have to be very careful about claiming performance benefits from social screening. Many studies show that social screening, as it is usually practiced, has little or no impact on risk-adjusted returns over the long term. This annoys both social investing's proponents (it's good to be good, but outperformance would clinch the deal) - and critics (who have long predicted disaster).
But there are a few variables where I think a close look is at least in order. The environment and corporate governance have attracted some attention, and good research is being done in these areas. But I also believe security analysts could improve their work by doing research in the area of employee relations.
Back in 1998 Chris Luck and I took look at the performance of the companies mentioned in the book The 100 Best Companies to Work for in America. Looking at both the original 1984 edition and the updated 1994 list, we concluded that these companies were performing, as a group, better than their risk profiles would lead you to expect. We updated the analysis in 2002, and I gave a talk on our findings at the Northfield quantitative conference that year (that presentation is here).
So I was delighted to learn last week that Alex Edmans, a researcher at MIT's Sloan School of Management, had done a careful analysis of the performance of the "100 Best" list and come to the same conclusion (Alex says the latest version of the paper will always be here). The study has many points to recommend it - Alex's introductory discussion is excellent, the returns analysis covers a long time period (1998-2005), and a multi-factor risk model (Carhart) is used to filter out style, size, and momentum effects.