TAA Model for August, 2011
This is a 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 outperformed the benchmark by a healthy margin in July, returning (as of yesterday’s close) +2.4% versus +0.3%.
The source of that outperformance was mostly the model’s position in gold, with the ETF GLD returning 7.8% for the month. Random fact: the model has held gold in all 10 months since inception, and over that time, GLD has returned 28% annualized. Not sure how long this trend holds, but I’m glad we’ve been on board.
For the upcoming month of August, the model is selling out of real estate in favor of Japan equities. Because of the lower expected volatility of Japan equities (relative to real estate), the size of the protective Treasuries position is being reduced.
There hasn’t been a significant difference in performance between the model and the benchmark in the 10 months since inception. As I’ve discussed before, I think the real strength of TAA becomes apparent when equities and related asset classes go through a protracted downturn. Until then, I’m happy to keep pace with the benchmark, given how diversified our holdings are.
The benchmark has been uber strong over the last 10 months relative to historical norms because of (a) strong performance from equities (which has boosted absolute returns), and (b) a strong inverse relationship between equities and bonds (which has boosted relative returns). Again, happy to keep pace at the moment with the intention of locking in these gains when equities and related asset classes turn sour.
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, what is the ytd return on this model?
thx
steve
Excluding trading frictions (which should be negligible for a reasonable sized account) 5.1% YTD. michael
Dear Michael, I am puzzled with your concept of relative volatility. I understand why you measure the volatility of each etf versus a target volatility. However I don’t understand why you need a future level (implied vol ?) rather than a realized level.
In your ranking of asset class returns, you use realized returns and not expected returns.
Could you please clarify this concept or recommend some research paper about it ?
Thank you
Henry
Hello Henry – I think we’re saying the same thing. When I mentioned the “expected volatility” of Japan equities, I’m basing expected vol on historical realized vol (relative to other asset classes), not the implied volatility of some derivative (which is what I assume you’re asking about). Sorry if that was confusing. michael
Hi Michael,
You seem to have a very low turnover for such a model.
Do you use a very slow MA for defining uptrend?
Also I’ve notices that US Treasuries don’t exhibit as much momentum as the other assets. Timing them actually reduces return and increases DD (at least with simple MAs) over the past 40 years.
Do you still use your uptrend criteria on Treasuries?
Best,
Rom
Hello Rom – I guess you could say that the model implicitly uses long’ish MAs for defining the uptrend because I’m trading based on monthly data. I think the model can reach it’s long-term goals with the low turnover we’ve exhibited so far (about 3 trades / month on average) and I think that has value in terms of ease-of-use.
I stopped using the uptrend criteria for Treasuries after the first few months of trading. I agree with you, doesn’t make sense with Treasuries. I just hadn’t analyzed the issue to that degree when I wrote the posts linked to from this one. For that one particular asset I’m using something totally unrelated to trend-following.
michael
I agree with you, low turnover is a good thing.
I’ve re-read your posts and I’ve seen you were using std as risk measure.
It turns out using std reduces turnover much more than using other $-loss based metrics that I use. Although the performance is slightly worse the lower turnover makes up for it.
Treasuries have basically been straight up for the last 4 decades. However with recent devaluations, the $ loosing its reserve status as well as uncertainties around USA solvability I’m afraid the Treasuries could become a constant negative bias on the portfolio for the decades to come.
I need to find an uptrend criteria for bonds that can mitigate that risk (atleast until eurobonds come out :) ).
Would you be willing to sharing what you use or point to some interesting ideas?
Best,
Rom
I actually had a similar question to Rom’s; however, instead can I ask what look-back period you use in your analysis when determining whether to put money in Treasuries or not?
Also, kind of going along Rom’s line of thinking with UST and adapting the model to long term changes in the market: given the point of tactical asset allocation and ideally investing in hopefully non-correlated assets, any commentary on the recent correlation between gold and treasuries and speculation that some investors may be viewing them interchangeably (I think I saw an R-value of 0.89 between the 20 year and gold for the past month or so).
And as always, thanks for posting. Especially this month.
RE to Jon: unfortunately, don’t want to share details on the UST criteria (apologies).
My thoughts on the gold vs UST relationship (which are not by an means revolutionary) is that yes, in this specific market, the two are both holding a strong negative correlation to equities (and a positive correlation to each other), but none of that has always been true and at some point in the future won’t hold true again. Another tangentially-related example is commodities which for most of the market’s history has had little correlation to equities, but more recently have almost been a perfect match.
As for the “why”. No idea. I have my own personal thoughts, but don’t feel qualified to express them in the public sphere.
As a side note: it should go without saying that all of this underscores why it’s important to build a TAA model that’s not based on what we know to be true today, but rather, one that adapts to the market as it evolves.
michael
Hello Rom – unfortunately, I’m not going to share the details of how I determine when to trade Treasuries, but I will say that it’s in no way related to the trend. Side note: I think the benefit of Treasuries extends to beyond just generating returns. Treasuries have maintained over the long-term consistently low correlation to equities, something that can’t be said for any other major asset class beyond gold and, up until the last decade or so, commodities. michael
Hey Michael, I was re-reading prior posts on TAA andI noticed that you filter with trend then you rank the remaining assets. my first inclination was to rank and then filter. i’d be curious to know if you’ve tested your system in this manner.
either way, how do you make sure that you have a diversified portfolio unless you rank within asset class as well?
also, i was wondering what software you use in your testing.
Hello Steve – I treat the trend as a “binary” criteria. Either the asset is uptrending and might be chosen that month or not and is completely out of the running. So either way, the result would be the same. Difficult to explain in text, so I hope that made sense.
I limit the number of similar assets that can trade each month, regardless of what the trend/ranking is calling for. So I can’t ensure that, for example, equities and offsetting assets (ex. Treasuries) are chosen in a given month. But I can ensure that equities are not given too large a weighting in any given month.
And I use a mixture of Excel, Wealth-Lab, and custom programming depending on the task at hand.
michael
Alright, I understand.
I agree with you on intermarket effects.
However if you exclusively base your uptrend criterie on these effects (I’m not saying you do) you might have a problem down the road with UST. Basically you’d be betting on the correlation effects persisting which might not be the case if the USA loose the superpower status (loosely speaking).
Best,
Rom