"Will My Risk Parity Strategy Outperform?"
Thorough and subtle description of a backtest comparison featuring two risk parity portfolios (one levered and the other unlevered) and two traditional portfolios (a 60/40 and a market portfolio). Results are mixed, but slightly positive towards RP portfolios: 1) levered risk parity had the best performance if transaction costs are disregarded, but in their estimation, those costs would have negated the gains. 2) The unlevered RP portfolio, meanwhile, was best from a risk-adjusted return standpoint. One strength of the paper is that it is backtesting a very long sample (since 1926).
Read the original:
Important Points for the RP Investor:
Page 75: One main takeaway is that results of backtests can differ based on start dates/end dates, assumptions for costs, the rate used to calculate the cost for leverage, or other choices made by the researcher. As such, it is wise to be circumspect about arguments for (or against) the empirical superiority of any particular backtest. Yes - true, and good to keep in mind.
Page 76: Their study is quite complex, and is fully described here. One thing I noticed is that they use higher transaction costs for the past (they estimate them at 1% for the period 1926-1955, and then .1% for the period since 1971). This makes earlier versions of RP look worse (since it does typically involve more transactions, especially if you have rules for frequent changes in allocation to keep up with shifting patterns of risk), but I wonder how suitable this is for current investors. No one is hopping in a time machine to put RP into practice in 1926, so shouldn’t we only be concerned with the transaction costs of now? That would lead us to merely using the raw returns of asset classes. I only mention this because one of their conclusions seems to be that RP works if you don’t include transaction costs but doesn’t if you do, so that choice about transaction costs seems important.
Page 77: Figure 1 shows the levered RP portfolio outperformed the other three. The unlevered portfolio was lowest, but its path was noticeably smoother than the others (showing up in the best Sharpe ratio of the four).
Pages 78 and 79: discussion of the various assumptions used in the backtests and how they impacted performance. Worth reading if you are really into understanding the parameters of this particular backtest.
Pages 80 - 82: Discussion of statistical significance of findings. Authors generally find that results are close enough so as not to be statistically significant in any direction (also see the chart on page 84).
Page 85: Figure 8 shows triumph of unlevered in terms of risk-adjusted return, in three separate scenarios regarding borrowing and trading costs.
Pages 85-86: Conclusion. Discussion of main points as noted above, and ending with the observation that “risk parity may be a preferred strategy under certain market conditions or with respect to certain yardsticks.” Reading between the lines, the take home for the DIY investor is that RP works best when borrowing costs are low (obviously) and if you can keep transaction costs low, perhaps by not transacting all that often.