First, let me acknowledge upfront that deep historical data is messy and is not precise. Depending on the source, you might get different results. Having said that, I still prefer to supplement my analysis with as long a history as I can get in order to learn about potential outcomes unseen in shorter histories. Yes, historical data might be imprecise, but does it lead to accurate / prudent conclusions about future potential risks and rewards? If yes, then it’s informative.
It’s Better to Be Vaguely Right than Precisely Wrong - Rory Sutherland (#15 here)
Second, this blog is just a sketch that can easily be turned into a long and detailed study. Here, the goal is to get a first rough approximation of the risks and returns to Risk Parity as far back as possible (which happens to be January of 1877).
In the spirit of simplicity, I define Risk Parity as a portfolio that has an equal risk allocation to U.S. Equities, U.S. 10-Year Government Bonds, and Commodity Futures - levered to the same volatility as a U.S. 60/40 portfolio.
The earliest one can possibly extend Risk Parity has been constrained by the availability of Commodity Futures Data, which until now has been commercially available only since the 1950’s [note: using commodity spot prices, which are available for longer histories, is not accurate because of the well-known large differences in returns between the futures and the spot markets].
At Two Centuries Investments, we have now created a new, commercially available, data-set of hand collected historical commodity futures prices from the Chicago Board of Trade Annual reports. The data goes back to the beginning of commodity futures trading on January 1877 and ends on December 31, 1962. To build it, we have commissioned two separate data collecting agencies to spend many weeks scraping the raw PDFs and recording monthly commodity prices for all available contract expirations. We then manually checked for differences between the two sources and tried to reconcile the disagreements by looking up the prices ourselves - in this way improving the data collection quality.
The other data required to extend Risk Parity back to 1877 comes from Aswath Damodaran’s and Robert Shiller’s websites - annual returns for U.S. Equities, U.S. 10-Year Government Bonds (approximated from yields) and the Risk-free rate. Prior to 1928, Risk-free rates are from Jeremy Siegel’s old website. Commodity Futures returns after 1962 are from AQR [who have also manually collected commodity futures back to 1877, but our index level results slightly differ, hence I choose to use ours].
Based on the data sources above, from 1877 to 2018, the annual volatility of a 60/40 portfolio re-balanced annually has been 12.23%. A Risk Parity portfolio with the same volatility that is equally allocated to equity, bonds and commodities, results in 29.75% allocation to commodity futures, 122.55% to bonds, and 32.95% to equities. That causes Risk Parity to be 85.24% levered, costing the Risk Free rate (assuming all futures implementation).
So, how does it look?
Risk Parity looks very similar to a 60/40 over the long-run with a slight out-performance since beginning. These are, however, log-scale returns, so the eye can easily miss the ‘wild swings of dispersion’ along the way. Let’s look at the relative performance.
The swings between Risk Parity out-performing and under-performing a 60/40 appear as long lasting cycles. The upswing in favor of Risk Parity that started in the early 1970’s and that has peaked in 2012, has succeeded in propelling Risk Parity into most institutional portfolios, right around the end of the cycle. Most investors did not have the benefit of seeing the prolonged periods prior to the 1970’s that have shown the long-term cycles in the opposite direction.
And how about the relative -60% draw-down from 1940’s to 1960’s? Such 20-year stretches will make any investment committee quit if they have not seen them before.
Overall, Risk Parity definitely appears as a credible alternative to a 60/40. Yet, most of the out-performance came in the last 44 years (albeit with a slightly lower Sharpe ratio), and Risk Parity actually under-performed in the first 50 years. The middle 50 were also in Risk Parity’s favor by 53bps and a better Sharpe ratio [again, the exact magnitudes of returns should be taken with a ‘huge grain of salt’ -but these are the results I get when I run the data, which make me more informed than not seeing them. You might run it differently, or have a different dataset, and so you will need to get comfortable with your results].
What are my take-aways?
1) Risk Parity is not a ‘bad way to invest’ for the Long-Run if it really aligns with your personal philosophy. Specifically, if you believe that:
All asset class Sharpe ratios are equal in the long-run
Growth and inflation are the two primary risk premia to which you desire a balanced amount of exposure
Nothing can be forecasted about asset class returns
2) Like all approaches, Risk Parity’s performance cycles in and out of favor compared to a traditional 60/40 approach. If Risk Parity’s philosophy and features were appealing to you in 2012, they should still be appealing to you now. Yes, it’s much less ‘exciting’ now because it has under-performed for 6 years, but constant out-performance was never its promise - just an unsubstantiated hope back in 2012, extrapolated from a recent rally and subtly advocated by the product manufacturers.
3) Looking at deeper history always has the effect of ‘calming down the excitement’ that shorter history can artificially create. If you looked at the Risk Parity backtest only from the 1970’s, you would see a line trending up vs a 60/40 - creating a false expectations that Risk Parity is superior. Yet the longer history says that Risk Parity is not a magic bullet, but also, perhaps not worse than a 60/40, but only if you stick with it.