I’ve always been a huge proponent of excluding the standard three-, five-, seven- and ten-year performance chart from any of our marketing materials. In my opinion, there are both theoretical and practical drawbacks of pointing investors to this type of analysis. My three main points of contention with this standard table are (1) it assumes investors care about absolute return with no consideration for risk, (2) even if the above assumption is correct, it assumes that trailing three to ten year performance is a good predictor of future performance and (3) the numbers are often skewed by calendar year quirks and are not granular enough to get a good sense as to how the strategy should be expected to perform across different market environments.
As usual, Clifford Asness and Antti Ilmanen of AQR were able to summarize the problem much more eloquently. In their paper “The 5 Percent Solution”, the pair writes “…an additional high-level recommendation would be to alter or at least soften the focus on three-to-five-year evaluation periods for managers and styles. These evaluation periods are death to returns, and nobody ever notices. Financial market data abounds showing short-run (within a year) momentum patterns and multiyear reversal (value) patterns. Yet investors often make asset-class allocation decisions and manager fire-hire decisions using a three- to five-year evaluation period. In short, they act like momentum investors at reversal (value) frequencies. Return-chasing – allocating toward winners or away from losers – at multiyear horizons and procyclic capitulation after disappointing performance are among the most common ways investors can sabotage their own long-run performance.”
“An additional high-level recommendation would be to alter or at least soften the focus on three-to-five-year evaluation periods for managers and styles. These evaluation periods are death to returns, and nobody ever notices.” – Clifford Asness and Antti Ilmanen
A quick experiment can illustrate AQR’s point. Using Morningstar’s mutual fund database, I identified all funds in the US Open End Moderate Allocation category with an inception date prior January 1, 1996 and at least $100 million under management. For each fund, I computed trailing three-year returns. I then formed four portfolios, (“high”, “high mid”, “low mid” and “low”) each representing one quartile of the funds sorted by return. For example, in a given year the “high” portfolio invests in the 25% of the funds with the highest return during the prior three years. Portfolio constituents were changed on an annual basis.
The chart below presents the annualized return and Sharpe ratio of the four portfolios as well as that of a portfolio that invests equally in the entire universe (“equal weighted”). As predicted by Asness and Ilmanen, the funds with the worst performance of the last three years actually returned the most over the course of the following year. In addition, this outperformance was not just a function of the “low” portfolio funds taking more risk as the “low” portfolio also had the highest Sharpe ratio.
Beware placing too much weight on medium to long-term performance when choosing manager’s for your portfolio.