How the All-Weather Fund Got Wet by John Mauldin
There are far fewer market participants today than just ten years ago, managing much larger portfolios across more asset classes, and using much less trading. In future letters I’ll lay out in detail how this structural shift has large and specific consequences for the nature of game-playing in markets, but for the balance of this letter I just want to make a simple, and I hope obvious, point: structural change in any social environment wreaks havoc on historically observed correlations and patterns within that environment.
Unfortunately this sort of structural change is effectively invisible to econometric modeling of portfolios, and as a result understates the risks inherent in portfolios that rely heavily on historical correlation patterns. In a market undergoing structural change, all of the “timeless and universal” relationships that form the backbone of Risk Parity funds like Bridgewater’s All-Weather Fund and similar offerings by Invesco and AQR are much less certain than their econometric justifications would suggest. The underperformance of these strategies in recent weeks and months (“Fashionable ‘Risk Parity’ Funds Hit Hard“, Wall Street Journal, June 27, 2013) takes on new meaning when seen in this light.
There’s nothing wrong with the math of the correlation exercises that underpin Risk Parity funds, any more than there was anything wrong with the math of the correlation exercises that ratings agencies like Moody’s and S&P used to grade Residential Mortgage-Backed Securities (RMBS). But in both cases there is an assumption about market behavior – the relationship of asset performance under varying conditions of growth and inflation for Risk Parity funds; the role of geographical diversity in mitigating the risk profile of mortgage portfolios for RMBS ratings – that is exogenous to the calculation of the projected returns. In both cases, the standard portfolio model of y = ? + ? + ?, where Epsilon is treated as an error term and the preference functions of market participants are assumed, gives a very compelling result: Risk Parity funds demonstrate an excellent risk-adjusted return profile, and trillions of dollars worth of RMBS deserve a AAA rating. But if you are wrong in your exogenous assumptions – if, for example, there is a nation-wide decline in US home prices for the first time since the 1930’s and geographical diversity provides no protection for a mortgage portfolio – then all the Gaussian cupolas and other econometric legerdemain in the world won’t save your AAA-rated security.
Risk Parity funds are a more broadly conceived, less levered version of Long-Term Capital Management. I mean that as a compliment, because it was the narrow conception and over-use of leverage at LTCM that ruined a solid investment premise and made it impossible for that firm to survive even a small disruption in patterns of market participant preferences – in LTCM’s case, the strong historical preference of major sovereign nations not to default on their debt obligations and the strong historical preference of major bond investors not to pay non-economic prices for the safety of US sovereign debt. The investment premise of LTCM was to identify small arbitrage opportunities between securities on the basis of historical correlations and to lever up those opportunities to generate nice returns. If you can take that premise and improve it significantly by expanding the scope and depth of the arbitrage opportunities and by shrinking the leverage turns required for acceptable returns … well, that seems like a really great idea to me. And I have zero doubt that investment giants like Ray Dalio, Bob Prince, and Cliff Asness can design a complex levered bond portfolio that is both safer and more rewarding than a simple unlevered stock and bond portfolio under most conditions. But I get VERY nervous when I am told that the reason these complex levered bond portfolios work so well is that a socially constructed behavior such as the assignment of value to highly symbolic securities is “timeless and universal”, particularly when the composition and preference functions of major market participants are clearly shifting, particularly when monetary policy is both massively sized and highly experimental, particularly when political fragmentation is rampant within and between every nation on earth.