How can a risk parity allocation strategy, equally weighting portfolio components by expected risk contribution, not really spread risk? In their October 2012 paper entitled “The Risk in Risk Parity: A Factor Based Analysis of Asset Based Risk Parity”, Vineer Bhansali, Josh Davis, Graham Rennison, Jason Hsu and Feifei Li examine the essential return and risk drivers embedded in a risk parity diversification strategy applied to multiple asset classes. They apply principal component analysis to extract independent (essential) risk factors from return streams for nine conventional asset classes: U.S. equities; international developed market equities; emerging market equities; real estate investment trusts (REIT); commodities; global bonds; long-maturity U.S. Treasuries; investment-grade corporate bonds; and, high-yield bonds. They then test four commercially available risk parity strategy implementations for dependence on the top two essential risk factors, interpreted as risks to global economic growth (proxied by S&P 500 Index returns) and global inflation (proxied by 10-year Treasury note returns). Using returns for the risk factor proxies and the four commercial risk parity portfolios from portfolio inceptions (ranging from January 1990 to July 2009) through May 2012, they find that:
- The first two factors from principal component analysis account for 68% of the variance in co-movements among the nine asset classes, supporting a view that two risk factors dominate. Other factors account for less than 10% each.
- Pro-cyclical asset classes (equities, commodities, REITs and high-yield bonds) relate strongly to the first essential risk factor (attributed to global growth risk).
- Counter-cyclical asset classes (U.S. Treasuries, investment-grade corporate bonds) relate strongly to the second essential risk factor (attributed to global inflation risk).
- High-yield and investment-grade corporate bonds have sizable positive dependencies on both essential risk factors.
- Across the four commercially available risk parity strategy implementations, dependencies on the essential growth and inflation risk factors vary considerably. Moreover, these dependencies change substantially by subperiod.
- In general, commonly used risk management strategies applied to asset class returns and risks, depending on the set of classes considered, can tilt strongly and dynamically to one of the two essential risk factors (growth or inflation, especially growth). Such a tilt implies unreasonable asset class return assumptions. In other words, diversification across asset classes does not guarantee diversification across essential economic risks.
In summary, evidence indicates that a portfolio diversification strategy based on essential (independent) risk factors provides more intuitive and reasoned allocations than those based on asset class risks.
Cautions regarding findings include:
- Analyses to support essential risk factor allocation may be computationally burdensome (or costly if delegated) for some investors.
- As noted in the paper, sample periods are extremely short for some of the actual risk parity strategy implementations examined. Moreover, the sample period for the strategy with the longest history is mostly simulated and may not realistically reflect net returns.
