Gyroscopic Investing

Normally I wouldn't give an actively managed ETF a second glance, but the new Canadian Global Risk Parity ETF (HRA.TO) has caught my eye.

It is a Horizon ETF, but the management of the fund is sub-contracted out to 'ReSolve Asset Management Inc.', or as some of us might know them better, the guys that run the excellent GestaltU blog. There is some great articles on the blog about portfolio theory (and even a couple on their take on the Permanent Portfolio, which they are fans of).

HRA is basically the product of the 'Risk Parity' approach to asset allocation, in which assets are weighted based on their relative contribution of risk to the portfolio in order to optimize the risk/return performance of the portfolio. This ETF basically exposes itself to the entire global investment market, on a risk-adjusted basis, and adjusts the asset allocation based on changes in volatility and risk. That's the 'active' part of the ETF, the fact that the asset allocation is not static, and asset weights evolve based on a pre-determined risk/volatility benchmark for the portfolio.

I think there is a lot of merit in the Risk Parity approach to asset allocation, and I know a lot of Gyros around here have even adapted Risky Parity modifications to the traditional HBPP.

I think what I find attractive about this ETF is that the "problem" with the Risk Parity approach is that it is more difficult for the individual investor to replicate in self-managed portfolios. This ETF appears to possibly be a viable way of contracting the math and trading out in order to benefit from global market exposure, on a risk-adjusted basis (something that my primary domestic HBPP does not provide).

Some downsides:
- 85% management fee, and that doesn't even include the MER of the underlying ETFs.\
- huge NAV spread ($9.5 bid, $10.5 spread)
- ETF is new, and therefore has no track record or proven history of stability/performance.

Thoughts?

TLDR: I guess this thread is another way of asking, is it possible to easily achieve a DIY Risk Parity, globally exposed portfolio without the middle man, OR, given the complexity involved in running such a portfolio, is this ETF a good candidate for achieving such exposure?

I like GestaultU's research a lot. They post some very interesting stuff. My only thought for consideration is that there is some pretty good evidence/research that a 1/N portfolio that is well constructed will do as well or better than higher math forms of the same assets. Looking for lack of correlation is conceptually the right thing to do as is equalizing risk. Practically speaking though, risk is highest when volatility is lowest which is exactly the worst place to be when that asset goes off the inevitable periodic cliff and in major SHTF events most correlations spike.

The same guys have leveraged private account versions that are on my watch list as potential future investments.

Kbg wrote:Practically speaking though, risk is highest when volatility is lowest which is exactly the worst place to be when that asset goes off the inevitable periodic cliff and in major SHTF events most correlations spike.

Can you elaborate on this?

I could but there is plenty to read by doing some google searches. Equal weight vs risk parity or mean variance portfolios.

Kbg wrote:I could but there is plenty to read by doing some google searches. Equal weight vs risk parity or mean variance portfolios.

I did, and I read this paper that I found (How Inefficient are Simple Asset-Allocation Strategies?) The TLDR conclusion that the paper makes is:

Our main finding is that the 1/N allocation rule (with or without rebalancing at each
trading date) is not very inefficient. In fact, it performs quite well out-of-sample: it often
has a higher Sharpe ratio and lower turnover than the policies suggested by both the static
and the dynamic models of optimal asset allocation.

Our analysis suggests that if one knew the true moments of asset returns, then it would
be best to use the portfolio weights suggested by an optimizing model. But, when the mo-
ments of asset returns are not known and have to be estimated, then the error in estimating
these moments is significant – at least for the data sets considered and using the estima-
tion methods that are standard in this literature.

Consequently, in many cases, the large estimation error overwhelms the gains from optimization and so simple allocation strategies
outperform those from optimizing models of asset allocation.

However, I'm still unclear about exactly what you are describing when you say, "Risk is highest when volatility is lowest which is exactly the worst place to be when that asset goes off the inevitable periodic cliff and in major SHTF events most correlations spike." Are you saying that re-balancing asset allocation into downward-trend volatility is dangerous because unnaturally low volatility acts like a coiled spring that suddenly explodes with force?

blue_ruin17 wrote:However, I'm still unclear about exactly what you are describing when you say, "Risk is highest when volatility is lowest which is exactly the worst place to be when that asset goes off the inevitable periodic cliff and in major SHTF events most correlations spike." Are you saying that re-balancing asset allocation into downward-trend volatility is dangerous because unnaturally low volatility acts like a coiled spring that suddenly explodes with force?

I think that's what he means, that volatility is mean-reverting, and you may observe a quiet period (like now in US equities) where the vol seems low, so the algo would weight stocks more, but also when everyone wants cash, all assets may move together, so who cares about your design?

Thanks Och.

As a generalization bull markets before they end tend to be low volatility so a volatility weighting approach will more heavily weight the bullish asset. The effect is you are more heavily weighted in the thing when it goes down. You see this happen all the time in the popular cap weighted indexes. And of course as the bear gets going volatility goes up and you pare back your weighting. There is some thought that the automation of this whole process is a contributing factor for the tendency of some markets to go straight down and then back up in a matter of days.

I think there may be some utility to weighting using very long duration historical measurements but the past month/week doesn't make much sense to me...unless you are doing something very short term in nature and/or using a good dose of leverage.