Mel Lindauer wrote:Sure, Wade, but if you choose any other time period, the B&H far outpaced the market timers. So what's the point except that it appears to be cherry-picking a period that makes your point. Change the time period to 1999 or 2006 and you change the entire picture and the results are dramatically different. There's only one period in the entire sequence where your assettion is true, and that's the period you choose to use? Sure looks like data mining to me.
I take the issue of data mining very seriously, and with all due respect, any data mining that I am doing is in favor of buy-and-hold, not in favor of market timing.
I don't think either the 100% stocks buy-and-hold, or the market timing strategy which switches between 100% stocks and 100% Treasury bills, are at all realistic or feasible for a conservative long-term investor. But for now, let's play by the Fisher and Statman rules, as the findings for "market timing" are so robust anyway, that it hardly matters how we do it.
As is considered in Fisher and Statman, this table shows the results of a horserace between 100% stocks, and market timing over the entire period, January 1871 to January 2010, for $1 invested in 1871.
100% stocks provides final wealth of $95,404.
I consider 4 different decision rules for market timing. The one used by Fisher and Statman, "historical median" provides final wealth of $124,147. But the rolling mean and rolling median approaches are more realistic. The worst performing one of these is the "rolling median" one, and with that the market timer has final wealth of $94,866, which is slightly below the buy-and-hold. In the example I gave above, I data-mined in buy-and-hold's favor by basing it on the market-timing strategy that gave the worst results for market timing.
I think a fair conclusion is that market timing will provide you with approximately the same wealth as 100% stocks buy-and-hold.
But before going any further, let's consider risk. 100% stocks is riskier by any imaginable criteria. Its standard deviation of portfolio returns is much higher at 18.02%, its Sharpe ratio defined in two different ways is lower than any market timing. Fisher and Statman argue that the Sharpe ratio is biased, so I also add the "Information Ratio" which compares market timing to a 50/50 fixed portfolio that has the same ex-ante average stock allocation. This ratio compares the active returns from market timing to the active risk of market timing, and the fact that these numbers are positive all help to suggest that market timing is providing superior risk-adjusted returns. Next, the maximum drawdown from market timing is much less. That is how far the portfolio drops from past highs to current lows. The buy-and-hold once experienced a 60.96% drop, whereas the worst drop market timing was 24.16%. Considering returns over maximum drawdown, market timing still does much better. Then, there is another set of risk measures which considers that investors are more sensitive to losses rather than gains. The downside deviation which only penalizes returns less than zero shows that market timing is much less risky, and this in turn favors market timing for the Sortino ratio. Market timing also has a much lower average stock allocation, and market timing changes its allocation on average once every 5 years.
So, market timing provides essentially the same returns, but with much less risk.
This is in spite the fact that market timing has to overcome one major obstacle in its battle with 100% stocks buy-and-hold: the very generous average equity premium enjoyed by U.S. stock investors in the past.
I think it is more realistic to compare market timing to a fixed 50/50 allocation rebalanced each year. With that, the wealth accumulation is only $13,426. That's the equity premium in a nut shell. Fixed 50/50 provides much less wealth than fixed 100/0. At least in historical U.S. data. Mr. Bogle would suggest not to count on this necessarily continuing in the future.
Ex-ante, I would have expected market timing to be much riskier than fixed 50/50. That is because the market timing strategy is rather extreme and unrealistic anyway, as I explained at the start. But comparing risk measures here, market timing is holding its own. It does have higher standard deviations, higher downside deviations and lower Sortino ratios. At the same time, it has higher Sharpe ratios, and I already discussed its positive information ratios which are compared to 50/50, and it has lower maximum drawdowns, and higher returns over maximum drawdowns. Compared to 50/50, I think it is fair to conclude that market timing provides signficantly higher returns at a comparable level of risk.
Now, more specifically, to the point you are making about the choice of end period, and how I am data mining here. I needed to bring up the table above to show how I already am helping out buy-and-hold by comparing it to the worst-case for market timing. This new figure shows the horserace over the entire historical period. Looking at each possible date as the end period:
-market timing provides more wealth in 51.8% of cases
-buy-and-hold provides more wealth in 32.37% of cases
-they are essentially tied in 15.83% of cases, where I define being tied as the difference in wealth accumulations between the two strategies as being less than 1% of the value of the buy-and-hold wealth accumulation. I think there is no point in declaring a winner in cases where they are so close.
So, if you wish to focus on one of the 32.37% of cases in which buy-and-hold is clearly ahead, I respectfully submit that you are the one doing data mining. Even though market-timing is ahead more often than not, I'd be willing to call it a draw on this point, and instead focus on how market timing provides substantially less risk for the same general returns.