Garco wrote:I've found the Callan periodic fascinating for several years. It's a bit tricky to interpret if you focus only on the rank order of the asset classes because it doesn't take account of the absolute changes in TR magnitude, just the relative ranks. But since those percentages are there in the table, it would be interesting to calculate the mean and standard deviation of the % for each category over the years. This could provide useful information for comparing the risk component for each asset class. (I haven't done it, but if somebody has, please post here.)
Not what you were looking for, but :
Looking at the assets from a yearly equal weighted
angle, then with 9 assets that's no different to having 11.1% weighting in the best asset each year, 11.1% weighting in the worst asset ...etc.
For 1993 - 2012 inclusive
Averaging the top and bottom two (22.2% weights each)Top 2 (22.2% weighting)
Annualised 25.27Bottom 2 (22.2% weighting)
Which is no different to having a single asset, 22.2% weighted, that annualised 25.27% and another asset 22.2% weighted that annualised -6.14%. Whilst it would be nice to just hold the better performing asset in isolation and compound at that rate, in practice we can't know what asset that will be, so we have to equally weight all assets each year in order to capture that characteristic.
As a guide, comparing those figures with a 4x25 Permanent Portfolio comprised of 25% in each of TSM, LTT, STT (2 year) and Gold that over the same period hadAverage 22.2
Annualised 21.9Average -5.575
respectively for the best and worst assets (25% weighting each).
One perspective might be to say - look, the Permanent Portfolio achieved much the same averages as a much more (diverse) stock heavy portfolio (constitutes of the Callan), another might say - look the Permanent Portfolio had much the same multi-year risk as a more diverse stock heavy portfolio.
Similarly extending the concept to Cap weighted versus equal weighted indices, Cap weighted is like substituting (averaging) several categories into a single (more heavily weighted) entity/value. Such that if that single entity performs well then the overall result is better, if it performs poorly the overall result is worse.