TAA Backtest and Expectations
Unfamiliar with Tactical Asset Allocation (TAA)? Click for a primer.
I want to put this discussion about TAA into perspective. Just how much juice can we really expect from this TAA thingamajig?
Note that when I talk about TAA, I’m using Mebane Faber’s model as a jumping off point, so I’m trading a diversified basket of asset classes infrequently (once per month or less) with a focus on trend-following and momentum (read more).
In this post I’ll show a backtest of my take on a TAA model, and in a follow up post I’ll take a more technical look at what I considered when building the model.
The graph above shows backtested results of the TAA model (red) versus the S&P 500 (grey) since 1971. The model does not employ leverage. See end of post for assumptions about return on cash and trade frictions.
The real benefit of this flavor of TAA is NOT generating returns, it’s managing losses. To illustrate, below I’ve included a chart showing drawdowns for the model (red) vs the S&P 500 (grey) since 1971. I like this very much…
And lastly, numbers for the number-lovers…
1. IGNORE RETURNS. The fact that the model outperformed the market (or any other asset class) is meaningless. Returns are an illusion; they’re just a function of risk. Much more important are returns relative to volatility/drawdown and the smoothness of the equity curve over different market regimes. On both, the model shines.
2. These results are made even better by the fact that this model is very much not curve-fitted. My rules are similar to Faber’s: select asset classes that are in an uptrend and showing positive momentum. These are tried-and-true rules that have worked since the dawn of ticker tape.
3. There are limitations to this flavor of TAA because of how infrequently it trades and how broad the asset classes are that it’s holding. I don’t claim that my take on TAA is the best one, but I do think we’re pretty close to the ceiling of what one could reasonably expect from this type of approach. Without more active trading there’s only so much blood to squeeze from this turnip.
4. There are two other variations on the model I’m also tinkering with: (a) trading with leverage, and (b) holding positions for at least a year to take advantage of reduced long-term CG rates in taxable accounts.
In a follow up post I’ll take a more technical look at what I considered when building the model. As always, more to follow.
[Edit: click for a summary of all posts in this series on TAA]
Test assumptions: (a) when an appropriate ETF existed, I ran this test using actual ETF data. For periods prior, I used index or futures data but did not adjust for an ETF expense ratio, which would add some drag to the results presented here, (b) results do not account for transaction costs or slippage, but given how infrequently the model trades, an investor with a reasonably sized portfolio should be able to closely reproduce these results,(c) taxes have been ignored, and (d) I’ve assumed a return on cash of HALF the nearest 13-week Treasury.
. . . . .
Filed under: Tactical Asset Allocation, Trading Strategies | 72 Comments