The Big Advisors Perspective Debate: AQR’s QSPIX, A Tale of Two Return Streams?
Updated: Sep 19, 2019
For those of you who don’t have access to Advisors Perspective (AP), I am doing a cut on a hotly debated topic between advisors and academics at AP about a relatively new offering from alternative investment industry bellwether AQR. AQR has been around for 20+ years, carries $35B+ in assets and has many products to its credit. In fact, one of Cliff Asness’s articles (Momentum and Reversion Everywhere) is a favorite of mine. So, I was intrigued enough by the AP debate to have to extend my quant background and experience to look under the hood of AQR’s Style Premia Alternative I (QSPIX, “Fund”), which I am going to provide some clues related to how to assess its future potential from its past. Full disclosure, I have no interest in AQR or its products.
Let’s start with a simple picture:
Source: Data from AQR Funds
Of course, QSPIX ends up at approximately net +34% over 4 years of operation. This is not as important as the journey getting there as depicted by decomposing the Total Return (Green Line) into its two constituents, a Core Return (Blue line) and Special Distribution Return (Red Line). Here are some key takes from the picture:
1. QSPIX bills itself as “market neutral”, an overused, much maligned term that those in the alternatives, hedge fund industry, somewhat pride themselves that they can source alpha without taking on market risk. Keep in mind this idea was well propagated throughout the 90s and 2000s right up into the 2008 financial crisis that proved largely futile. And, AQR built a reputation in part on this idea. The nature of these rules-based definitions of factors and correlations will always lend themselves to liquidity-driven arbitrage unless those same definitions have some form of adaptive nature to them. I refer to this as “unnatural” risk taking because it centers on engineering return streams that essentially deviate from a more normal distribution within the asset class. Nature and, in this case the marketplace, tends to win these battles.
2. Every single current market neutral idea is a misnomer IF the strategy relies on simplistic ideas like correlation as the basis for market risk neutralization. Why? Because the market does not care how any of us defines such backward-looking ideas (read, back tested), it will eventually find and break that rule (via liquidity) and find a point in time when the risk neutral free ride ends up costing the fund/strategy sometimes more then it made when it was right vs when it goes wrong.
3. QSPIX paid out about 23% to investors since inception in Special Distributions. From their own copy, “Distributions represent the total of all capital gains and income paid to shareholders on the specified distribution date.” It goes on, “Adjusted NAV is an estimate of the fund NAV assuming that all distributions had been retained and not distributed by the fund. This is a purely hypothetical value.” Hmmm...so I get this occasional gain thrown off and have to reinvest it at the same rate of the return of the Fund otherwise the NAV return quoted is purely hypothetical or fictitious? Also, let’s not forget to pay current taxes on it too assuming you aren’t in a tax deferred setting like self-directed IRA, pension or endowment.
4. Since inception, 2/3rds of QSPIX’s Total NAV return came from special distributions, one of which in late 2014 was about 50% of the total Special Distributions of 23%. Meaning, hard to count on this item for future return expectations and even AQR states its Total Return is hypothetical leaving us to have to dissect return streams: one for the core return of about 11% and the other Special 23%.
Here is how the “liquidity cost” of this form of active management can manifest itself:
5. Volatility of gains is less than that of losses, so negatively skewed return volatility. This is a key attribute that feeds and governs whether an active portfolio construct will generate alpha over time. Take a look at the return stream in the picture below; it does not take a spreadsheet to figure out the return volatility is slightly negatively skewed measured on a daily basis. OK, I did break out a spreadsheet and calculated Total Annual Volatility (“Vol”) at 7.9%, Upside Annual Vol at 4.95% and Down Annual Vol at 5.04%, not bad, I have seen worse. AQR will likely focus on the Fund’s low Total Return volatility, but that is not as telling as the skew of return volatility. You want gain vol to be higher than loss vol in an active strategy otherwise the use of liquid provisioning (eg, trading, rebalancing, turnover) can lead to net degradation in fund results. QSPIX shows minor skew to the negative side so don’t expect upside long-term performance in excess of the known risks it is taking on.
6. Note that the Total Return stream drew down from a peak cumulative return of 18% (November 2014) to about 9%, or a total of drawdown of approximately 9%. This by itself is not a big deal but in light of the slightly negatively skewed upside vs downside volatility may be symptomatic of a susceptibility to sub-performance. Basically, the drawdowns are of greater severity than the drawups due to an overreliance on risk “controlling” to minimize volatility. Risk “control” is another illusion that can be quite costly over time and this symptom reflects that cost.
7. Next up, time in drawdowns. The Fund hit its high early on (good for marketing) but lousy for investors from about late 2015 to mid-2017. Time in drawdowns is a dead giveaway that the quantitative ideas underlying this engine are potentially misfiring or the factor that was driving early returns faded away while the approach waits for some other (or the same one) factor to go its way, a bias if you will. The marketplace is tough on those who think they can play market neutral games AND secure great risk-adjusted returns. So, dead money is costly long term.
8. Is the Fund uncorrelated, yes, but to what? So, for the fund to pick up any excess return, it has to take some form(s) of risk on, whether from stocks, bonds, currencies, commodities, regardless of geography (US vs International) otherwise, what’s the point? In this case, the only way for the Fund to extract even a modicum of gain is to leverage itself back up to 10% annual volatility target. Sounds good, except leverage is not a free ride either.
9. Hidden Levers stress testing and analysis reveals a strong inverse correlation between oil prices and this Fund’s returns to date but, worse yet, its correlation is skewed against Fund investors. So, it appears that the Special Distributions component of this Fund was baked on catching the oil market dislocation correctly. In fact, 66% of its Special Distributions were paid out by the end of 2014, when oil coincidentally hit massive lows and since then there have been no major market dislocations that this Fund has capitalized on until August 2017!
10. On Special Distributions again, with the big relative recent run-up in the fund, investors should get another pretty big distribution. Lumpiness in this component of the Fund’s return stream to investors has to be accounted for, tax implications and reinvestment rate returns notwithstanding.
11. AQR is using a Bank of America 3 month T-Bill index as a reference benchmark to this Fund. Seems a little aggressive to me in that if I had a choice between placing my free cash in 3 month T bills vs this Fund, its 7-8% annual vol, that 2015 to 2017 drawdown and lumpiness in Special Distributions – not sure where the strong connection is between them. And, it’s not as if Fund investors are principal protected by the Federal Government.
12. Now here is the final kicker I offer. This Fund is baked on over a thousand positions in multiple asset classes using standardized factor ideas and longs covered by shorts (market neutral, zero beta to stocks). Sounds slick until you realize that the Fund’s return distribution takes a big departure from normal distribution. This is usually a direct function to the market risk neutralization idea, which seeks to “control” the Fund’s return distribution to generate regular returns (to feed Special Distributions) and provide a sense of security to principal (ala T bill benchmark). In a Fund as well thought through as this, it might be tough to see but kurtosis is an obvious byproduct of unnatural engineering that affects both extreme ups and downs and what happens under normal circumstances. A comparison of its daily return distribution to that of the SP500 index and a mystery long/short alternative fund (XYZ ALT) can give you a better sense of this:
Note QSPIX distribution skewed more negatively based on slope and number of observations heavily clustered on either side of slightly positive mean return.
Note SP500 distribution skewed slightly negatively based on slope but observations more evenly clustered on either side of positive mean return.
Note XYZ ALT distribution skewed positively based on slope and number of observations heavily clustered right at a nicely positive mean return.
Keep in mind that all three represent daily return streams in the exact same time frame (November 2013 to December 2017) and represent vastly different methods to access return streams from liquid securities. So, when you are looking at return streams, this sets you up for a frame of reference as to what to expect going forward when the strategy is challenged by the marketplace. QSPIX has a return stream that could be described simplistically as robbing Peter to pay Paul, meaning sourcing returns from a marketplace within such a narrow confined space to prevent the return stream from looking more volatile. Looks great until that day when Peter (the broader liquid marketplace) figures it needs to get some back. Each investor has to decide which return distribution makes sense for their objectives and constraints.
13. Back to spreadsheets, another quick calculation in Excel shows excess kurtosis of QSPIX at .94, meaning you have very little kurtosis, which can lend itself to an odd distribution of returns in the “normal” periods centered around the mean return. This type of return distribution can cost you during the normal times versus the extreme. SP500 has excess kurtosis of 3.0 and XYZ ALT is 7.6 so both of these are more likely to provide some element of accessing normal asset class return streams to minimize performance surprises. Think of this as strategy behavior indication rather than a gain/loss indicator. Continuing this line of thought - consider, for example, Long Term Capital Management (LTCM) of late nineties fame had 43 seemingly independent strategies running simultaneously and never thought that they had a common thread connecting each of those seemingly uncorrelated strategies (I know this only because that's what their Chief Risk Officer told me in a post-mortem). But, the market ultimately found it and led to its wholesale collapse. While QSPIX is not levered like LTCM -- and you probably won’t encounter wholesale liquidation failure -- that kurtosis can show up in costly sub-performance however muted it may appear. An insidious little gremlin baked into these highly human engineered and potentially euphemistic return streams.
There’s more I could relate on funds like this but keep in mind that this analytical framework applies to all investment strategies (AQR is not alone and not trying to pick on them). If this type of Fund works for you vs investing in 3 month T Bills -- great, it has historically beat 3 month T bills right now presuming you buy in at the right time.
I guess if you made it this far in this rundown, one other oddity strikes me about AQR and ALL the funds I see reported on their website: not one out of 50+ funds has an inception date for reporting purposes prior to 2009. For a firm with a 20+ year history and $35B, I guess they had to reset their bars after the financial crisis of 2008. What does that tell you relative to what their numbers depict above?
Markets are way too efficient to let traditional and now fairly standardized active funds have persistent out-performance net of management fees, trading costs, leverage and the inefficiencies that humans can, unknowingly or otherwise, bake into their investment process. Don’t shoot me, I am just the messenger with a fairly efficient $35T market behind me. And, I only offer these points to help those who are open to next generation active investing to seek purer asset class return at lowered risk. It exists but you have to know what to look for when peering under the hood.