Feature Article: Equal Risk Weighted Portfolios in 2007 and 2008

As regular readers know, our primary benchmark for evaluating the performance of our model portfolios is based on equal asset class weighting. This methodology assumes no ability to forecast either asset class returns or risk (i.e., volatility, correlation, and other risk metrics). The key to implementing this approach is to define asset classes broadly enough to create a portfolio of exposures to underlying return generating processes while minimizing the overlap between them. However, we have also noted that future asset class returns and risk are not equally hard to forecast. Rather, history suggests that the relative ranking of asset class riskiness is more stable over time than the ranking of asset class returns. This is not to say that volatilities and correlations do not change over time - as was vividly demonstrated in 2008. However, even when these variables change, the overall riskiness ranking tends to be more stable over time than the ranking of asset class returns. This observation provides the logic for another approach to creating a benchmark portfolio - allocating asset class weights so that each makes the same contribution to overall portfolio risk. Mechanically, this requires making assumptions about the standard deviation of returns (i.e., volatility) for each asset class, and the correlations between them. If any of these forecasts prove to be wrong, the actual contribution of different asset classes to overall portfolio risk will not be equal.

We thought it would be interesting to see how equally risk weighted portfolios that were formed at the end of 2006 (based on returns data from the previous five years) performed in 2007 and 2008. We formed equal risk weighted portfolios using twelve different asset classes for portfolios covering AUD, CAD, CHF, EUR, GBP, JPY and USD. All portfolios included domestic and foreign bonds, domestic and foreign commercial property, commodities, timber, emerging markets equity, uncorrelated alpha strategies (defined as a 50/50 mix of an equity market neutral and a global macro index) and volatility (defined as the VIX index). Where real return bonds were available, we also included them; where they were not (in CHF and JPY, the latter because of the short data series), we included gold instead. Finally, in the USD portfolio, we combined domestic and foreign equity (which were used in all other portfolios) into developed market equity, and added gold (in the form of the GLD ETF). After reading that, long-time subscribers may start scratching their heads, as we have traditionally taken the position that while an allocation to gold coins makes sense as part of one's cash or liquid reserve, its role as a stand alone asset class (separate from commodities, where most indexes already include an allocation to gold) is more open to question, not the least because of the difficulty in establishing a fundamental value for it (since it provides no regular stream of income). To be honest, the inclusion of gold in this analysis was a bit of an experiment. Clearly, gold ETFs have performed well, and provided diversification benefits, during a period of heightened uncertainty, and may also do well if inflation eventually spikes as a result of the past two year's explosion in global liquidity. This is not necessarily true of a similar asset class, short-term U.S. Treasuries, which did very well in the former case, but should not repeat that performance if inflation spikes. Equally important is our evolving view of the fundamental valuation challenge. In this regard, deeper commodity futures markets have provided more information that could be used in a fundamental valuation analysis, as has the IMF's continuing research into gold's evolving role and the factors driving its price. So consider the appearance of gold in this analysis as yet another step in our exploration of gold's potential role as a permanent asset class investment option.

In each currency region, our analysis made two points clear. First, as shown in the following table, the equal risk weighted portfolios had lower expected real return volatility than the equal asset weighted portfolios:

Second, in the equal asset weight portfolios, just three asset classes usually accounted for over half the portfolio's volatility. The following table shows the top three contributors (and their percentage contribution) to the volatility of the equal asset weight portfolios:

It is also interesting to note that in all cases, the inclusion of an allocation to volatility as an asset class in our 12 equal asset class weight portfolios had a negative contribution to overall portfolio risk (i.e., it reduced it). This provides yet another reason for including it as an option in our updated model portfolios. The next table shows the composition of each of our equal risk weighted portfolios:

Last but not least, the next table shows how our equal asset and equal risk weighted portfolios performed between 31 December 2006 and 31 December 2008, assuming they were rebalanced monthly (and ignoring the associated transaction costs). Both portfolios have a starting value of 100:

As you can see, even under conditions which were quite favorable to the equal risk weighting approach (due to the high weighting it gave to fixed income asset classes in 2007 and 2008), the extent of its outperformance versus equal asset weighting (using 12 asset classes) was not very large. Put differently, it might well have been the case that the performance differences we observe could have been eliminated by the choice of a different approach to forming our equal risk weighted portfolios (e.g., collecting historical data from a different time period, or using a different methodology, like shrinkage estimators or an exponential moving average, to weight it). In other words, the differences we observe between the two portfolios do not strike us as very significant, at least over the 2007 - 2008 period. Of course, that begs the question of how an equally risk weighted portfolio might have performed over another time frame. Yet that raises a series of additional issues, such as the methodology for forming the portfolio, and how often it is updated -- interesting questions for another article to address.

That said, this analysis produced some useful insights, not only about the risk weights of different asset classes in the equal asset weighting portfolio, but also about the roles of volatility (which appeared in all equal risk weighted portfolios), gold (which received similar weightings were it was available, even in the presence of real return bonds), and uncorrelated alpha (which received a much heavier weighting in the USD portfolio than in the others, which no doubt reflects the impact of changing exchange rates on its non-USD returns and risk).

| October 2009 Issue: Key Points | Global Asset Class Returns | Uncorrelated Alpha Strategies Detail | Table: Market Implied Regime Expectations and Three Year Return Forecast | Table: One Year Asset Class Valuation Conclusions and Recent Momentum | Market Phase Change Risk Analysis | This Month's Letters to the Editor: Oil and Gas Partnerships - Do they fit in your portfolio?; What's the best Asset Allocation Today; Is it Possible to Over-diversify a Portfolio?; PIMCO Heavy Weighting for Emerging Markets, II's thoughts?; Timber Followup | October 2009 Economic Update | Global Asset Class Valuation Updates Detail | Feature Article: Equal Risk Weighted Portfolios in 2007 and 2008 | Product and Strategy Notes: Challenges Facing Investors in Venture Capital Funds; Improving Warning of Future Financial Crises; A New View on the Fundamental Drivers of Equity Market Returns; and The Long-Term Impact of the 2007-2008 Crisis on Financial Advisers' Compensation |