How the coronavirus killed portfolio diversification
The extraordinary market movements over the past few weeks have thrown long-established cross-asset class correlations into disarray. The worldwide scramble for USD cash meant that traditional safe havens, such as government bonds, the Japanese yen, the Swiss Franc and even gold, dropped in value, as stock markets tumbled. In other words, conventional diversification strategies have not been effective.
The figures below highlight the interaction between the major drivers of risk in Qontigo’s global model portfolio, used in our weekly Multi-Asset Class Risk Monitor report. The numbers represent correlations between daily risk-factor returns over three months. A red field with a negative value indicates offsetting movements. For example, the -0.68 for equity versus interest rates as of February 14 means that prices of stocks and government bonds mostly moved in opposite directions in the 12 weeks before that date. Blue fields, on the other hand, imply co-movement. FX returns are measured against the US dollar, which is also the portfolio’s base currency. The blue 0.36 between FX and equity on March 20, therefore, indicates that foreign currencies depreciated as share prices declined.
Short-term correlations between major risk drivers as of Feb. 14 (left) and Mar. 20 (right)
Source: Axioma Risk
The fact that the matrix on the right is all blue confirms the perception shared by many market participants that “everything is going one way right now.” Yet, not all correlations are close to one. Share and bond prices—which, as we already noted, traditionally move in opposite directions—have now simply uncoupled, as Treasury yields first dropped, but then bounced back, while stock markets continued to fall. As investors around the globe rushed into the US dollar, the resulting depreciation of foreign currencies further amplified local stock-market losses, pushing risk even higher.
Even more noteworthy was the flipping of the interaction between risk-free interest rates and credit spreads from negative to positive. As we noted in our recent blog post on corporate bonds, the volatility and returns of those securities seemed now to be primarily driven by changes in risk premia. Formerly, this had predominantly been the case for sub-investment grade issues, where the risk of repayment is a bigger consideration for investors than changes in the overall level of interest rates.
The co-movement of most security types in the portfolio also meant that there were much fewer diversification opportunities for multi-asset class investors. This is highlighted in the two graphs below, which show the weighted contributions by risk type for February 14 and March 20. If all the major factors were perfectly positively correlated (i.e., stock, bond, commodity prices, exchange rates, etc., all moving in the same direction), total portfolio risk before the stock-market downturn would have been slightly above 10%. However, because of the offsetting movements of share prices on one side and bond prices and exchange rates on the other, the actual volatility was less than half that.
Weighted volatility contributions by risk type as of Feb. 14 (left) and Mar. 20 (right)
Source: Axioma Risk
Looking at the same graph in the current environment, we see that the diversification effect has all but disappeared. Revisiting the correlation matrices above, the near-zero correlation between equities and government bonds is now almost the only source of risk-reduction left. This is also reflected in the graphs below, which show percentage risk contributions by asset class, compared with market-value weights. While traditional safe havens, such as higher-quality bonds, the Japanese yen and gold, actively lowered overall portfolio volatility in mid-February, all risk contributions are positive in the current environment. However, US Treasuries still exhibit the lowest percentage of overall volatility compared with the size of the holding.
Asset-class weight versus risk contributions as of Feb. 14 (left) and Mar. 20 (right)
Source: Axioma Risk