November 14, 2004
Keep Your Concerns Off My Returns
Several quarters ago, our global ethics class examined the issue of "social investing." It was presented almost as an unmixed good, self-evidently virtuous, and at little, if any cost. (The Wall Street Week clip we were shown quoted a very specific return on one fund for one year that beat the market, for instance.) This seemed, rather, self-evidently nonsense.
Social investing doesn't necessarily mean putting money into companies that have strong ties to their communities, or who treat their employees better, although it can. Social investing also seeks to reward or punish particular social policies. Funds may screen for tobacco, alcohol, or gambling, involvement in the defense industry, greenness (however one chooses to define that), particular products, or animal testing.
The minute you begin making investment decisions based on something other than return, you're going to do something other than maximize return. If this is so, then an honest appraisal requires knowing the costs of screening stocks according to your particular tastes. And it requires sharing those costs with potential customers. Or with employees whose 401(k) choices you are designing. How important would it be for them to know how much your virtue was costing them?
I decided to take an empircal look for myself. This posting is the first part of an ongoing examination.
Funds are organized by sector. This takes into account both the investment type (large-cap, small-cap, municipal bonds, etc.) and the goals (growth, steady income, capital preservation, etc.) Funds are ranked within their sector.
I used two types of readily-available information: quintile, and Lipper ratings. Quintile is a straight ranking of returns. There must be, by definition, 20% of funds in each quintile. Generally, the Year-to-Date, 1-Year, 3-Year, 5-Year, and 10-Year returns are used. Lipper rates funds on five criteria: Total Return, Consistency, Capital Preservation, Tax Efficiency, and Expenses. Again, these are organized into qunitiles.
I began by looking at perhaps the best-known family of social investment mutual funds, the Calvert Group. So far, I have looked at both quintiles, and at Lipper ratings.
The questions are: 1 ) Are Calvert Group funds more likely to provide above-average returns, or below average returns, or can't we say?, and 2) Are Calver Group funds more likely to have higher Lipper rankings, lower Lipper rankings, or can't we say?
For each test I used the same methodology. I used what's called a one-tailed Test for Proportions. Essentially, I'm asking in each case if the given population of funds, Calvert, has a higher proportion of its funds in the lowest 40%, or highest 40%, than the universe of funds as a whole. I can mix and match fund types this way, since each type of fund has its own set of quintiles.
Caution! Dry Statistical Terminology Ahead! If you can't stand this sort of thing, skip ahead two paragraphs! There will be no further warnings.
I used a 95% level of confidence, meaning that a statistically significant z-score is 1.65.
The z-scores that mattered:
- For 1-Year return, 2.22. (For 3-year return, the score was such that one more fund would have tipped it over; the z-score was 1.43)
- For expenses greater than average, 4.006! (I ran this again for expenses in the 80th percentile, and got a score of 5.37.)
Basically, this means that over the last year, we can state with 95% confidence that Calvert Funds return less than the fund universe, and with 90% confidence that this is true over the last 3 years. At the same time, we can state with practical certainty that they'll take more of your money in expenses.
When I expand the list of funds to include all those listed by SocialInvesting.org, things don't improve much. The 1-year and 3-year returns are lower with 95% confidence, and the 5-year returns are lower with 90% confidence. This implies that the lower returns are more likely systemic, and less likely the result of transient market conditions. This is verified by the only significant Lipper score, a z-score of 2.78 for the notion that Total Returns are in the lower 40%.
So much for the notion of a cost-free conscience.
I realize there's a lot of research out there on this topic already. But I wanted to do a little digging, based on the data that I have at my disposal. Further questions involve looking at volatilities and actual returns, comparing these results by the types of screens the funds employed, and expanding the list of funds involved to be more comprehensive, and reducing redundancy, where similar funds may skew the results. I may also used deciles rather than quintiles, to help start pinning down the actual investment costs of not defending the country, for instance.
Here's a spreadsheet with my original data and calculations. The data is from schwab.com, and was current as of the beginning of November, 2004.