I've had some questions about this December 14th press release from QED International, which states that social investors' aversion to vice industries has cost them returns. I'll start with my critical remarks, but I also have some positive comments (down near the bottom).
Any time I hear about a study of historical returns, I have some basic questions:
Question 1: Can I see the study?
Answer: In this case the press release appears to be the study - there's no information on how to get a more detailed look at the work.
LK Comment: There is a often a big gap between what the data shows and what the authors say it shows - if there's an underlying study it's good to have it. In this case we'll work from the press release, which is pretty detailed.
Question 2: Has the study been reviewed by anyone else or is it likely to be published somewhere?
Answer: There's no mention of a more thorough writeup, nor of any attempt to have the work reviewed or published somewhere. That doesn't mean it won't be, they just aren't telling you in the press release.
LK Comment: It's a lot easier to write a press release than to publish an article in a refereed journal.
Question 3: Does the study look at real portfolios, or backtests?
Answer: Backtests (with the exception of Domini Index/ S&P 500 comparison).
LK Comment: Hindsight is 20-20 - at any moment in time you can pick a group of stocks excluded by social investors that has performed well, and write an article about how social screens are costing investors money. (For some reason everyone wants to talk about tobacco lately but not about the auto companies, which social investors have also avoided and which have underperformed badly in recent years.)
Anyway, the real test of an investment strategy is whether it works forwards, not backwards. The only analysis in the press release I could call prospective is the table at the bottom of page three. It shows their vice composite with a marginally higher estimated growth rate than the Value Line universe (9.7% vs. 9.3%), and slightly more attractive valuation ratios. Okay, that's a good point. Vice stocks are expected to grow a bit faster and look to be a little cheaper.
Question 4: Do the numbers look ok?
LK Comment: The table at the top of page 2 presenting Domini Social Index vs. the S&P 500 appears to be in error. It shows the Domini underperforming the S&P 500 for the 10 years ended September 30th, when KLD's 9/30 press release shows outperformance during that period. They show the Domini Social Index for the 10 years at an annualized 8.74%, while KLD reports 9.97%. I don't know, but I'll bet they used the Domini Social Equity Fund's performance instead of the underlying Domini Social Index, comparing a mutual fund with expenses to an index which has none.
The QED press release claims the Domini Social Index is behind the S&P "for annualized periods of one, three, five, and ten years." But comparing index-to-index using KLD's reported returns, it looks like Domini is ahead on its ten-year record, about tied on its five-year record, and behind over just the past one- and three-year periods.
Question 5: Are returns risk-adjusted or presented in the context of a risk model?
LK Comment: Sometimes a study shows a significant performance difference between portfolios, but does not explain where the difference might come from. In fact, there's a well-developed literature on determinants of differences in portfolio return. The usual suspects are:
- Risk (beta)
- Size (market capitalization)
- Valuation (price/book ratio)
- Momentum (relative price strength)
Usually, these factors explain most of the differences in historical investment performance among portfolios. If you use the variables listed above you're using a Carhart model, if you drop momentum you're using a Fama & French model. The authors of this press release appear to have used neither.
And that's a problem. Their argument is: the Domini Social Index left returns on the table - given the data problems it is clearly not true for all the time periods they list, but it certainly has been true over the past one and three years. The authors would like to show that this shortfall was because of the failure to own vice stocks.
But maybe it was just a failure to own value stocks or small stocks, both of which have had great performance lately. The authors include some commentary on this, but without a risk model we can't know the answer with any precision. And it's really important to know that answer: if the Domini Social Index is underperforming because it doesn't own value stocks, that's a solvable problem (investors can supplement it with a value fund, or pursue other diversification strategies). But if there's something really special about vice stocks, that makes it impossible to create diversified portfolios without them - well, that would be big news, and a big problem for social investors.
So those would be my questions, and after looking at the press release they haven't persuaded me. To show that social investors are making a BIG MISTAKE by not owning vice stocks, they need to show that that excluding them creates unavoidable diversification costs. And I'm not seeing it here. From that perspective I'm a lot more concerned about Energy than the sectors they presented in this study.
Let me finish with a positive comment. In addition to their retrospective analysis, the authors do some APT analysis of the excluded sectors (there is a good explanation of APT here). This is new and useful work and deserves attention - I have not seen APT analysis of specific excluded sectors before. Dan Dibartolomeo and I did an overall APT analysis of the Domini Social Index ten years ago (we used a different model than the one used here) and found that its macroeconomic bets differed significantly from the S&P 500 (see Kurtz, Lloyd and Dan DiBartolomeo, "Socially Screened Portfolios: An Attribution Analysis of Relative Performance." Journal of Investing, Fall 1996). Dan updated this work in 1999 and that paper ("Managing Risk Exposures of Socially Screened Accounts") is available on the Northfield website. Our analysis suggested the Domini index was particularly oil price-sensitive, and could underperform during a period of rising oil prices. And we have certainly seen that.
Hopefully QED will follow this up with a more detailed white paper, or better, a journal article. I think the APT aspect of this work could be a journal article in itself, particularly if the authors computed the APT coefficients for the Vice Composite they present on page three.