14 juin 2017

Do Investors Overestimate the Benefits of Diversification?

June 14, 2017, Jean-Jacques Ohana, CFA and Dr. Christian Witt (both YCAP Asset Management)

We have witnessed a strong divergence among US sectors over the last 10 years. The S&P 500 Technology sector has almost doubled whereas the S&P 500 Energy has stalled over the same period (see Figure 1). This divergence reveals the impressive winning streak of growth stocks vs. value stocks since 2007 (see Figures 2.a and 2.b). This outperformance of growth stocks is reminiscent of the Technology stock rally at the turn of the 1990s which eventually degenerated into the Tech bubble in the early 2000s.

To assess the contrast between US equity sectors, we can either compute the average of one-year rolling correlations among all US sectors or the implied rolling one-year correlation defined as the average correlation which makes a risk-balanced portfolio attain the same diversification.

To calculate the implied rolling correlation, let us denote xi and ?i as the weights and volatilities of asset i, and ?P the overall portfolio.

Using expressions (2) to (4) to re-arrange equation (1) the implied correlation can be computed as follows:

We choose xito build an unlevered risk-balanced portfolio with

in order to determine the implied correlation at each date.

As shown in Figure 3, the implied correlation is usually close to the average correlation as US sectors are rather homogeneous. Moreover, correlations vary substantially over time. For instance, the present decorrelation trend across US equity sectors could be a side-effect of an emerging bubble characterized by extreme complacency about US technology stocks despite a deteriorating macro environment hitting value stocks (in particular the energy sector). The present correlation falls into the lowest quintile of dynamic correlations since 1993. There are three instances of similar or even lower correlations: 1993-1994, 1996 and 2000-2001. Interestingly, these were all periods of local bubbles when one single investment theme prevailed and when negative news hitting one particular sector spread towards the others.

Most importantly, the US S&P 500 displayed significant monthly underperformance after the correlation reached extremely low levels (see Figure 5). However, these periods occurred mostly in the 1990s when the market structure was different from today. Thus, there are no guarantees that low correlation complacency will produce the same disappointing performance going forward.

However, in a multi-asset universe, implied correlations typically deviate from average correlations. This is possibly due to the availability of safe haven assets (say sovereign bonds, gold) which deeply change the dynamics of the correlation structure. As a matter of fact, the lowest implied correlation was reached in 1993, 2003, 2008 and 2015. Three of these periods (1993, 2008, 2015) were soon followed by major simultaneous corrections across all assets classes. We call such periods « financial deflation » when yields rise, equities drop and commodities fall at the same time. In statistical terms, the future performance of risk-balanced portfolios indeed deteriorates when correlations have been very low. Yet, today’s global implied correlation is neither particularly weak (3rd decile of correlations) nor does it convey any signal of exaggerated complacency about diversification.

Figure 1: Divergence of Performance between US S&P 500 Information Technology Sector and US S&P 500 Energy Sector.

Figure 2: Ratio US Value Stocks / US Growth Stocks (Total Return).

Figure 2.a: since 1990.

Figure 2.b: since 2012.

Figure 3: Average 1-Year Rolling Correlation across US Equity Sectors vs. Implied Correlation of US Equities Sectors.

Figure 4: Average Correlation of the S&P 500 (Blue) on a Monthly Basis and the Lowest Decile of Dynamic Correlations (Orange).

Figure 5: Statistics of Subsequent Monthly Performance Conditional on Preceding Average Correlation across US Equity Sectors.

Figure 6: Average 1-Year Rolling Correlation vs. Implied Correlation across a Basket of Diversified Equities, Commodities, Currencies and Sovereign Bond Futures.

Published on Riskelia’s Blog