Helpful at the extremes, less helpful elsewhere.
Last year I examined the predictive power of a number of data points, but I didn't test turnover. Because many investors use it in their selection process, I thought I'd put turnover through the paces to see where it shakes out versus the other data points.
I'll walk you through what turnover means, how I tested it, what the tests show, and finally, what turnover tells you about a fund's strategy.
An Old Formula
The SEC mandates how fund companies calculate turnover, and that method could stand to be updated. The algorithm is a sum of the dollar value of the lesser of purchases and sales during the fiscal year divided by average net assets over that period. As you'd expect, a 100% turnover ratio means the manager has turned over an amount of shares equal to the total value of the portfolio. It doesn't actually mean he's held nothing for more than one year--he may have held half the portfolio unchanged throughout the year and turned over the other half twice.
The SEC uses purchases or sales rather than both because that would be double-counting. However, by choosing the lesser of the two rather than summing them and dividing by 2, the SEC is allowing some quirky results to sneak in. If a fund has had large inflows or outflows in a year, the managers may have made far more purchases than sales or vice versa. In that case, the turnover rate would be significantly understated.
Bond funds present an additional challenge. A maturing bond constitutes a sale, so a short-term bond fund where most of the holdings are maturing in a year will run near 100% turnover simply by buying an equal amount of bonds to replace the ones that mature.
How We Ran Our Test
Using rolling five-year periods beginning in 2000, we ran a test that was similar to past tests of predictive value. We calculated returns and success rate to find out what the chances are that a fund would survive and outperform its peers over the ensuing period. So, a 40% success rate over a five-year period tells you that 40% of the funds in a certain group survived and outperformed. If another group had a 20% rate, you know the former was better than the latter.
The idea is to use real data that existed in the past and see how one might have done using that past data rather than using today's data and projecting backward. Also, asking whether a fund survived is a way of accounting for the fact that many weak-performing funds are killed off each year in a way that can skew studies of fund data.
For turnover, I added another wrinkle. I looked at pre-expense returns. I wanted to separate turnover from expenses. The lowest-cost funds are mostly index funds, and those funds generally have low turnover. Were we to run the data including expenses, we might find noisy results in which low turnover simply captured low costs in the form of index funds.