A recent article in The Wall Street Journal explores a hypothetical market-timing performance of various investment newsletters.
To eliminate the effect of individual stock picks, equity positions in each newsletter’s portfolio were virtually substituted with the Vanguard Total Stock Market ETF (ticker VTI) in the the same dollar amount. In periods when the newsletter eliminated equity positions in the portfolio, VTI would be replaced with a typical money market fund. Then the returns and volatility of the virtual portfolio would be compared to those of a fixed portfolio of 100% VTI. Using this methodology, one of the top newsletters would generate almost twice the return with one-third less volatility of VTI in the five-year period through May 2013.
Unfortunately, such market timing does not generate RealAlpha™. The article underscores that
“It also is worth noting that the best-performing timers primarily focus on market trends of several months’ duration or longer, not trying to profit from shorter-term market gyrations.”
As discussed in the FAQ, this is a classic situation where one fixed benchmark, or even a combination of two fixed benchmarks, is not appropriate. In this case, a fixed 100% VTI benchmark is not applicable to prolonged (multiple-month) periods in which the hypothetical portfolio was invested in a money market fund with practically zero volatility.
A benchmark composed of VTI and a reference money market in fixed proportions in the entire evaluation period is better in that it attempts to match the overall risk of the hypothetical portfolio. For example, if over a five-year period the portfolio held VTI roughly two-thirds of the time and a money market one-third of the time, then a composite benchmark with such weights might be used. (Incidentally, this is similar to devising a proper benchmark for target date funds that consist of multiple types of assets, whose weights change over time.) However, this fixed dual-asset benchmark is also imperfect — at any given time, one part of that benchmark is not applicable to what the “binary” portfolio holds.
In the Alpholio™ approach, the benchmark would dynamically adapt to the holdings of the analyzed portfolio in each multiple-month sub-period. This is necessary for a true risk adjustment of the portfolio. Unfortunately, this would also result in little to no credit given for market timing.
Finally, the article only covers a five-year period starting right after the trough of the recent major downturn, i.e. a rebound portion of the market cycle. A longer period encompassing the entire cycle should have been used to assess market timing strategies.