Here's my comments on Mebane Faber's "A Quantitative Approach to Tactical Asset Allocation" paper. Sometimes it helps me digest what I'm reading when I'm an "active reader". I have modeled this commentary after something I saw in a translation of "The Art of War", where the various students commented on each of Sun Tzu's statements. If anyone has study tips like this that work for them, please post them in this thread or PM me.
Jon: This is right after an international fixed gold price in dollars was abandoned, so the entire time period under study is consistent within that context.
Jon: Holding cash through the entire period would keep you at the starting line, $100. Poor performance compared with every other asset class shown, which have all grown from $100 into the thousands, a more than ten-fold increase. In nominal terms, the cash position exhibited the least amount of volatility, but was the clear loser.
Jon: In the late 1920's and early 1930's during the observed drawdown, the dollar had a fixed gold price, so it should not be compared to modern stock market drawdowns in dollar terms because it is a drawdown in gold terms. Today the S&P 500 is presently in a drawdown over 80% in gold terms. This comment is not intended to refute the findings of this paper, but to illustrate that the 1930's drawdown was from a different market "epoch".
Jon: I agree for the need of this type of market timing model. Rational investors and traders are funding positions in these "risky asset classes" with borrowed dollars because of the "cash vs everything else" phenomenon illustrated in Exhibit 1. When the funding component of these trades (dollars) become scarce and liquidation events occur, it helps to have models that describe the process and how to navigate a portfolio through the liquidation events.
Jon: I will assume this constraint is only because there is sufficiently accurate price data available throughout the test period, and easy to feed into models without tinkering with the raw data. The market timing model I use also has this constraint.
Jon: I find the same results myself, but it varies greatly by when the study starts or stops.
Jon: The end date of these types of studies really does matter. Everybody probably remembers hearing the numbers bandied about by financial planners in 1999 that the stock market returns 10% or 11% per year. Today those numbers look ridiculous, and any financial planner doing linear projections of 10% per year in stocks to plan somebody's retirement would be tossed out on the street.
Jon: The lower risk profile of a good market timing system allows you to use position size and leverage to "engineer" a better risk/return situation. If market-like risk is acceptable to you, then a leveraged market timing scheme can be used to increase the long-term return dramatically. This idea is the holy grail of market timers, and explains the great emphasis placed with the historic maximum drawdown. The mention of 70% of time invested in the market is also important, because the other 30% of the time is spent collecting interest while sitting in a "risk free" vehicle. This could be a compelling sales point to investors, that their money isn't even at risk 30% of the time! This goes a long way to explaining the lower volatility profile of the market timing system. By watching prices and measuring the risk in them, we can reduce the volatility of our portfolio. This is the function of a good market timing model.
Jon: This sounds like a structural edge that could be exploited by shorting the leveraged etfs and going long the unleveraged etfs. We'll save that study for another day.
Jon: Studying history is a really big deal for investors and traders. There is no limit to the kinds of life-saving information like this to be found in the history books and data archives. I think it was Jim Rogers who taught a university course on finance and asked his students to pick an interesting market in history and write about its rise and subsequent decline.
Jon: This is a wonderful statistic, I'm impressed by the idea of measuring this.
Jon: This is one of the best uses for market timing models I have identified also. The REIT's just trend so well. It might be tempting to think that you don't need a market timing on something like this, but if you are using a lot of leverage or are a levered real estate investor, this kind of stuff can save your fortune.
Jon: I find these kinds of "sanity checks" are important in determining the robustness of a system. Of course I want to find and use the best parameters, but not because I expect similar results in the future. I consider the best parameters to be the ones in the middle of the robust range, and I expect performance similar to what I've seen near the boundaries of that range, rather than the middle.
Jon: I find this to be true in practice as well. A great example from this year was the March crash in the Nikkei caused by the massive earthquake, which nobody probably had advance knowledge of. The market didn't crash immediately from its highs. The real market panic didn't come until March 15, several trading days later. This out-of-sample exhibit corroborates the story that market timing models can be effective, although one would need to be making portfolio adjustments far quicker than monthly to escape this particular Nikkei crash.
