TAA Model for February, 2011
This is a new monthly feature at the MarketSci Blog.
Our Tactical Asset Allocation (TAA) model selects up to four assets from a diversified basket of asset classes on the final trading day of each month. Below is the new allocation for today’s close. Click to read more about the TAA model.
I eat my own cooking, so I’ve devoted a healthy share of my own net worth to the TAA model (read why). On the last day of each month I share my new allocation (see above) and real-time performance (see below) excluding trading frictions and taxes.
The model underperformed the benchmark in January for the first time since inception, returning -1.5% (versus -0.4%) as of the close on 01/28.
Note that five asset classes have been chosen for February. The model is designed to hold up to four in any given month, but two competing asset classes are on the cusp, and with the cutoff time for placing trades quickly approaching, I’m hedging my bets and taking a partial position in each.
I opted to trade small cap (IWM) instead of large cap (SPY) U.S. stocks in January due to a seasonal tendency for small caps to outperform (read more). That play did not work out as large caps beat out small caps by a healthy margin. This month the model is reverting back to large caps.
Happy Trading,
ms
. . . . .
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Tactical Asset Allocation | 14 Comments
michael, IEF? wow, now that’s surprising. did I miss something or is this model momentum oriented? IEF is at the bottom of my list. what am I missing?
Hello Steve – good question – UST is the only asset class the model trades that doesn’t respond well to trend-following/momentum (either in terms of it’s own performance or performance relative to other asset classes). I started using a unique non-momentum criteria for UST last month. Would have been nice to have it in place in December, but I digress…michael
thanks. figured something like that.
Mike, do you trade TAA model with MOC orders (15min before close @ NYSE)?
RE to Blink: in the “official” tracking account I trade (and will be posting account statements from shortly) I exclusively use MOC orders. michael
What happens if you delay TAA model and buy/sell next day MOO or MOC? At that time you know your new trade candidates for sure and since you are catching monthly market strength, it shouldn’t be much difference…?
I’m curious about what the effect of combining this with your seasonality information would be. There are several times when you change your position in the stock market between small cap and large cap. If you’re going to be changing the whole position would it make sense to sell a day before the end of the month and purchase on the last day to avoid the typical downtrend?
Hello Fred – for the TAA model, there is only one month where I replace large caps with small caps (if large caps would have been chosen by the model). That’s January b/c of the month long seasonality play.
Would trying to time large vs small caps (or any other asset class) intramonth perhaps improve returns? Sure, but it’s really beyond the scope of what I want the TAA model to be. I intentionally only want this model to trade robust observations that have existed for the last century. In other words, I don’t want this model chasing more fleeting observations that are constantly in flux. That’s why I refer to the TAA model as a “generational strategy” (i.e. not something we have to constantly revisit). More:
http://marketsci.wordpress.com/2010/10/04/tactical-asset-allocation-it-really-is-that-good/
Just my $0.02…michael
nice post. I will be watching your TAA model with interest.
Michael, does your TAA model backtesting show January to be one of the weaker months for momentum in terms of performance ? My rudimentary backtesting of simple ETF momentum models showed negative returns more frequently for January than any other month. Without a shred of evidence, I put this down to end of year readjusting.
Hello Ian – good question – I took a look at my own data, but I don’t see any significant strong/weak month seasonality in my tests. michael
Michael, are you sure that gold should be included among your investment options? I guess it’s easy to pick on it based on January’s results…if your GLD investment had been in GSG instead, you would have gotten >1% positive return last month, albeit with higher volatility. I’m not suggesting that you would have simply made that switch, but rather that you might have better results in general if you take that wild bucking bronco called GLD out of your model entirely!
GLD sure does wreak havoc on my (simple) TAA method, so I simply removed it from consideration. I do however choose to use DBC for commodities exposure, which does contain some gold and some base metals.
Hello Ed – couple of thoughts here: (1) I agree that gold can often be a volatility monster which is why I reduce exposure relative to other less volatile asset classes, (2) my criteria for inclusion in the TAA model is that the asset class is a “store of wealth” – gold clearly fits the criteria, commodities are a stretch, so it doesn’t make sense to forsake gold in favor of commodities, (3) because of the low weight gold is given in GSG/DBC, I view them as two entirely different widgets, (4) in recent history, gold exhibits consistently lower correlation to most asset classes than commodities (which is almost as important to a TAA model as pure returns are), and (5) you’re right, it’s easy to pick on gold based on January, but this would be a very different discussion in almost any month over the last few years (i.e. one month does not my mind change).
Just my $0.02…michael
Michael, I have been trying to develop my own TAA model and after working on it for a while the devil is really in the details. It is amazing how difficult it is to get squeaky clean data. It seems like even the best sources have some bad data points.
So far one of the hardest data to replicate is 10 year government bonds. It is easy to get your hands on 10 year note data from the Feds website back to 1956, but interest rates and the performance of 10 bonds are not the same. For example, I can’t take recent 10 year interest rate data and replicate the performance of the IEF or 30 year interest rate data and replicate the performance of futures data on the the 30 year. I am hoping you can show me how or possible tell me where I can find data on 10 years government bonds going back 30 or or more years.
In your early post on your research, you state that you used 10 year government notes. Do you have a clever way to replicate the performance of 10 year bonds based on 10 year interest rate data. If so, can you throw a want-to-be geek a bone and explain how I can closely replicate the theoretical performance of the IEF using 10 year note data. What data source have you found to be the overall best source or are do you use a combination of sources like you posted in an earlier post on various data sources.
As always, thanks for helping us that are less geeky than you.
John