Asset Allocation – Risk Parity Revisited: Part 2

In my previous Risk Parity Part 1 post I presented possible ways that funds might be allocated to momentum-ranked assets without any filtering to ensure diversification. One possible weakness of this simple form of momentum ranking is that highly correlated assets may rank similarly within the listings – this may be either a good thing (if all assets continue to outperform) or a bad thing (if all assets underperform).

To overcome this possible weakness I have introduced the concepts and benefits of filtering the asset list by breaking the list into “Clusters” where assets within the clusters are highly correlated but correlation between the clusters is less correlated so as to generate diversification within the chosen portfolio (by limiting the assets from each cluster).

The following figure shows the BuySellHold sheet of the same Feynman spreadsheet used in the last Post with the assets sorted (primarily) by Cluster and (secondarily) by rank (new feature):

For this example I have chosen to restrict the number of assets to four, one from each of the top 4 ranked clusters.

For these 4 assets, possible allocations (without Risk Parity adjustments) are shown in the figure below:

The allocation methods are identical to those described in my earlier Post i.e. Equal Weighted, Rank Weighted and Momentum Weighted (as used in the first figure above).

If we apply the Risk Parity adjustment (involving volatility and correlation factors) we arrive at adjusted weightings as shown below:

In my next Post I will be introducing a new method of momentum ranking, followed by Posts showing possible allocations with/without risk parity adjustments as above.

David