One Factor to Rule Them All
Over the past few years there has been a loud and persistent chorus of complaints from market participants about the fact that markets are behaving like simple ‘risk on, risk off’ discounting mechanisms, where almost all of the risk seems to be emanating from a very small number of sources. Gone are the days when investors could seek out independent sources of return in places like emerging market debt, or Japan, or currencies. Now all markets move as one – student body left, or student body right.
Many market participants point fingers at the world’s central banks as the cause of this dynamic, and the ultimate sources of concentrated risks. Charts like Figure 1. below from Ron Greiss at The Chart Store seem to support this view, with the caveat that correlation should not be confused with causation, and this axiom is doubly true in complex systems like markets.
Figure 1. Weekly S&P 500 vs. Total Assets of the Federal Reserve System
Source: The Chart Store
One way to derive the number of systematic factors that are at work in a portfolio of assets is to apply an analytical technique called principal component analysis (PCA) on the correlation matrix. This technique identifies the latent independent sources of risk in the portfolio, which helps to quantify how easy it is to achieve meaningful diversification in a portfolio by spreading one’s bets across a variety of assets. Where several factors contribute meaningfully to portfolio movements it is easier to achieve diversification, whereas if markets are dominated by just 1 or 2 factors, meaningful diversification is elusive.
It’s easy to see from Figure 2 that this has been the case since 2008. We performed a rolling 252 day PCA on a 56 asset universe going back to 2002 (requiring data to 2001 for priming), and captured the proportional contribution to total portfolio correlation from each of the first 10 factors. Note that in the beginning of the period, factor 1 explained about 35% of total portfolio correlation, whereas at the end of the period it explains about 75%. Factors 1 through 3 in aggregate explained 60% of total correlation in 2002, but they explain about 90% today.
Table 1. Asset classes used in our analysis (extended using indexes or mutual fund proxies where ETF history is too short)
Figure 2. Proportion of portfolio correlation explained by principal component factors 1 through 10, 51 asset universe
Data source: Bloomberg
The problem with latent factors derived from PCA is that no causal relationship can be drawn between any factor and an exogenous economic factor, so we cannot draw a conclusive parallel between global Quantitative Easing and the increasing concentration of portfolio risk factors. However, we might suggest that the coincidence is peculiar.
With 1 factor currently accounting for almost 80% of portfolio movement, investors are faced with few options for diversification. Where diversifying assets face prospects for low or negative returns, investors are faced with a difficult choice indeed. Either tie portfolios to one systematic risk factor, and live or die by the success of that factor, or endure lower returns by including the lower return diversifying assets in the portfolio. For example, in 2013 investors were punished for practicing any form of diversification away from developed equity markets, as bonds, credit, commodities, gold, real estate, and emerging equity markets have delivered very low or negative returns this year.
Figure 3. Year-to-date 2013 asset class returns
The median asset return so far in 2013 from Figure 3. is -1.01%, and the average return is 2.71%. This explains the failure of diversified methodologies like risk parity and the Permanent Portfolio this year.
Any way you slice it, diversification is much harder to come by today than it was 10 years ago; someone has taken away most of our free lunch.