This material comments on “‘Will My Risk Parity Strategy Outperform?’: A Comment” (March/April 2013).
We are grateful to the Financial Analysts Journal for facilitating an open exchange of ideas about the evaluation of investment strategies. We are also grateful to Clifford Asness, Andrea Frazzini, and Lasse H. Pedersen for their detailed comment on our article, in which we raised concerns about the backtest methodology in the January/February 2012 FAJ article by Asness, Frazzini, and Pedersen (AFP). In their letter, AFP make three basic points.
First, they argue that our concern about period dependence applies to “any strategy.” Of course that is true, and analysis of how particular strategies perform in different economic climates is essential to sound financial decision making. A broad cross section of readers—including financial practitioners, researchers, and students—have expressed appreciation for the period analysis shown in our Figures 2, 4, and 6. That analysis supports the notion (but does not prove) that risk parity tends to outperform in turbulent periods but not otherwise. We are aware of an appetite in the finance community for more of this type of analysis.
Second, they argue that their implementation of risk parity is superior to ours. We disagree, but we’ll put that aside and focus on something we believe investors ought to care about. In their letter, AFP claim that “the outperformance of [their risk parity implementation] survived [those financing] costs even for the most conservative estimates.” However, they fail to mention that by their own measurements, this “outperformance” lacks statistical significance.1 Moreover, their preferred risk parity strategy assumes that the investor adopted a fixed volatility target in 1929 and held it constant for 82 years—all the way through the Great Depression, several wars, the stagflation of the 1970s, the stock market boom of the 1980s and 1990s, the collapse of the dot-com bubble, and the recent financial crisis.
Third, they write that
using current data in a large broad-based universe of a variety of asset classes, we can estimate that current transaction costs and funding costs reduce expected returns for a risk parity portfolio leveraged to 15% volatility by about 0.50% a year, a modest number compared with the historical outperformance of risk parity.
In conclusion, we’d like to express our appreciation for the elegant theory of leverage aversion developed by Frazzini and Pedersen. We find it perfectly plausible that heterogeneous dispositions toward leverage lead to a priced risk factor. However, empirical studies of strategies that attempt to exploit such a factor would benefit from a more balanced perspective.
1 According to Table B1 in their January/February 2012 FAJ article, the t-statistics for their levered risk parity strategy minus the value-weighted strategy are 1.88 for the long sample and 1.16 for the broad sample when the financing is with Fed funds. When the financing is at LIBOR, the t-statistics are 1.16 for the long sample and 0.97 for the broad sample. None of the four t-statistics are statistically significant at the 5% level under Gaussian assumptions.
2 Wall Street Journal Market Data Center, Money Rates (1 February 2013): http://online.wsj.com/mdc/public/page/2_3020-moneyrate.html?mod=topnav_2_3000.