3
$\begingroup$

The "common man" version of a leveraged Risk Parity portolio (40% stocks and 60% long term bond, Quarterly Rebalanced) can now be easily implemented using the 3X leveraged ETF's (UPRO=stocks, TMF=bonds).

The last ~10 years of performance seems to have benefited from:

  1. A period of rising stock and bond prices
  2. A period of negative correlation of daily stock/bond returns
  3. A period of friendly volatility regime that didn't affect the returns of the daily 3x leveraged ETF's

Given that the predictions of doom for long term holders of leveraged ETF's didn't pan out, naive investors might extrapolate the last 10 years of this particular implementation into the future.....

QUESTION: How can we properly test and asses the likelihood of future long term success of this portfolio?

enter image description here

enter image description here

$\endgroup$
3
$\begingroup$

Apologies in advance for being hyper-critical. I have somewhat strong feelings about this =P

  1. The purpose of risk parity is to improve portfolio efficiency via achieving better diversification. (We won't delve into philosophical debates about whether or not this is true here...) Mechanically, by leveraging up low risk assets (e.g., US Treasuries) and combining them with higher risk assets, you are spreading out the risks, so that the total portfolio risk isn't dominated by equities anymore.

  2. 40% SPY + 60% TLT is a simple risk parity portfolio (although far too simplistic to withstand many shocks; more on this later). Using UPRO and TMF just levers up a risk parity portfolio to boost returns, with associated increase in risks. It's perfectly ok if that's your objective and if you can tolerate the higher risks, but recognize that you haven't improved portfolio efficiency beyond the SPY/TLT combo at all. In fact, your backtest shows Sharpe ratio declined; i.e., for the same unit of risk, you're getting less return. Generally speaking, we should avoid directly comparing two portfolios with very different risk profiles.

  3. The backtest sample is quite biased. It covers just one economic regime – an extended economic recovery. It does not show how the portfolio will perform in adverse environments. The bulk of the sample happens to be a period of zero interest rate & strong monetary stimulations, which boosted stock and bond returns alike. In fact, you can see that the performance has been more subdued over the past few years as the Fed began tightening and there were multiple large drawdowns during a period when a simple 60/40 portfolio did fine.

  4. There are at least two ways to think through how this portfolio might behave going forward:

    • This portfolio intrinsically bets on lower inflation and lower discount rates. If inflation rises and/or discount rate rises, the portfolio will suffer, as both stocks and bonds will struggle. Given that this is likely to be a long-term strategic asset allocation plan, such strong views should probably be avoided, particularly on a levered basis...
    • Predicting the future is hard, but we could at least look at history. The chart below shows the historical drawdowns, based on simple simulations. The light line toward the end of the sample is based on levered ETFs; the darker line is based on 40% 3x levered stocks + 60% 3x levered bonds. We can clearly see some really ugly outcomes and extended periods of drawdowns that the portfolio just struggled to recover from. This is actually a little generous since since it doesn't assume there's a chance of being wiped out completely in some extreme months...

enter image description here

$\endgroup$
  • 1
    $\begingroup$ Does your simulation use daily prices to simulate the ETF's (which reset daily)? if yes how did you build the long term bond series?....would you be able to share it? $\endgroup$ – hernanavella Feb 10 at 15:15
  • 1
    $\begingroup$ Hi @hernanavella. The simulation is based on month end total return index levels. I was focusing more on the structural weakness of the portfolio (which is equally true for the unlevered version), rather than the intricacies of 3x levered ETFs. Unfortunately the data can't be redistributed. If you search this SE, you can find posts on how to create pretty good proxy time series from yields. $\endgroup$ – Helin Feb 11 at 0:27

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.