Portfolio Allocation Model

I've built a portfolio allocation model to give a little more structure to my previously ad hoc investment decisions. It is a set of rules and equations for determining which investment categories — commodities, equities, and bonds — are seeing investor interest as a whole, determining what percentage of the portfolio each category will be, and within each category, selecting and weighting the subcomponents which are seeing increasing investor interest.

One of the worrying things in the three years before the financial collapse was that all asset classes had become correlated. In effect, there was a giant hose of money pointed at anything with a price, and price changes were more due to credit conditions than they were to fundamental value. To compensate for this, the model evaluates the each subcomponent relative to the most inclusive index for its category. For example, the FTSE All World Index (FAW) attempts to aggregate the entire universe of investable equities, something like 8000 worldwide. If I'd like to know how the US stock markets are doing, then looking at the relative performance of the S & P 500 (SPX) to FAW tells me whether or not my investing dollars should be in the US market, or somewhere else. It filters out (to some extent) the effect of the hose, and panics, calamities, policy discontinuities, etc. and it becomes quite clear that the US was a crappy place to have your money from 2003-2007, despite a very nice bull market in the SPX.

Having an allocation model compensates for having a lizard brain stuck inside an otherwise perfectly reasonable mammalian brain. Buy low, sell high to our lizard brain looks like "chase the leopard, run from the banana". When the prices of your investments are rising, it is a very comfortable feeling, and when they are falling, there is uneasiness. But if you'd like to buy low, then you have to buy when you feel uneasy, and likewise to sell high, you have to sell when you feel comfortable. This seems unnatural. Having a portfolio model gives you some leverage against the lizard. Issues that have done well will be overallocated and some amount sold for profit, and those that have not done well will need to be topped up, buying at attractive prices. Rather than having fixed weights, I've added a dynamic weighting system, so that an accelerating increase in investor interest triggers overweighting, and a decelerating increase triggers underweighting.

The issues I have chosen for the subcomponents are all unlevered long ETFs, which are themselves tradable versions large aggregates. My investable set currently consists of 9 commodity ETFs, 27 industry sector ETFs, 42 regional ETFs, and 10 bond ETFs.

Every weekend, I'll plug the latest numbers into the model and pull out the new target portfolio, and make adjustments on Monday.

Here's the allocations based on the state of the market as of last Friday's close:

Commodities
  18%   DBE (Base Metals)
   6%   DBO (Oil)
  10%   SLV (Silver)

Equities
   2%   BRF (Brazil Small Cap)
   2%   ECH (Chile)
   2%   INP (India)
   2%   IDX (Indonesia)
   2%   EWM (Malaysia)
   2%   EPU (Peru)
   2%   THD (Thailand)
   2%   TUR (Turkey)
   2.5% MOO (Agribusiness)
   2.5% KOL (Coal)
   2.5% BJK (Gaming and Casinos)
   2.5% HAP (Hard Asset Producers)
   2.5% XLB (Basic Materials)
   2.5% XES (Oil & Gas Equipment & Services)
   2.5% KWT (Solar Energy)

Bonds
   7.5% BWX (International Sovereigns)
   4.5% COY (Corporate High Yield)
   7.5% EMB (Emerging Market)
   5.5% JNK (Junk Bonds)
   4.5% LQD (High Quality Corporates)
   1.5% SHY (1-3 Year US Treasuries)
   2%   TLT (20+ Year US Treasuries)