This year of pain in liquid alts has continued. 1 1 Close As I did in my prior piece I’m using the jargon “liquid alt” in this piece as is common in some sectors of the market—but the discussion is really about lowly correlated alternative investments in general. Note that a lot of the themes that have impacted liquid alts this year have also impacted the active returns of long only portfolios. Such portfolios have underperformed their respective benchmarks, but the underperformance has been less dramatic as the long-only portfolios target less active risk than the standalone liquid alt portfolios. Even as systematic value has mildly come back, other factors have taken up its mantle, further turning the screw. A bad year is expected occasionally, and we all understand this even if nobody likes it. But, given that our alts (as defined below) are generally supposed to be market-neutral, and the equity market was down significantly in October, it was particularly disappointing to many that we also stunk (again) that month. What’s going on? Let’s dig in a bit.
Below is a scatter plot of the excess over cash returns of a representative liquid-alt portfolio using the same portfolios we examined in a prior blog
Our representative liquid-alts portfolio returns are comprised of our market-neutral Stock Selection Strategy (started in December of 2000), our DELTA Strategy (started in October of 2008), and our Style Premia Strategy (started in September of 2012). What I examine here are returns from the start of the Stock Selection Strategy. When the DELTA Strategy begins, it becomes an equal-weighted portfolio of the two. When the Style Premia Strategy begins, it becomes an equal-weighted portfolio of the three. If I only used the longest available Stock Selection Strategy (you want to see a lot of data points including other bear markets) you get very similar results to this scatter plot.
(y-axis) against world equity excess over cash returns from here
Note, as of the time of writing this piece, the data on Ken French’s website was not updated for the month of October 2018. To proxy for that one month I used the MSCI World Index return (USD) for October 2018.
This representative liquid alts portfolio has certainly been rocked this year, down, after fees, about 14% through October (the numbers plotted above are smaller as they’re monthly returns, not 10-month returns). This scatter-plot includes all the months since December of 2000 with the red dot being October of this year. Yep, it was a bad month in a down market. We’re disappointed in any bad month (or any bad hour or minute—I preach the long-term but like many others suffer each tick!). But as we stress all the time, being “market-neutral” is not the same thing as being a hedge.
If you plot, as I do above, a market-neutral investment against the market itself you should see a “cloud” and indeed that is what we see above. Now, that cloud should have a bias (a good bias). As both the market and the AQR market-neutral strategies in the representative liquid-alts portfolio make money more often than they lose, there should be (and are) more dots in the upper right quadrant. Since 2000, including this excruciating year, we see 34% of the months fall into this quadrant (when both the liquid-alts portfolio and the market make money over cash). This is the best of all worlds for investors and happily the most common occurrence (though only a plurality not a majority). The lower right quadrant where the market makes money, but we lose happens 27% of the time; the upper left quadrant where we make money, and the market loses happens 24% of the time; and the lower left quadrant where both we and the market lose, the worst of all worlds, happens 15% of the time (thankfully the lowest frequency but still not close to a rarity). So, while this October was the least enjoyable of all combinations (both down), it still doesn’t look anomalous at all. 4 4 Close We also draw your attention to the fact that the absolute (both good and bad) representative liquid-alt portfolio returns are generally larger when the market is down versus when it’s up (the vertical dispersion of the dots is higher on the left side of the graph than the right). This increased dispersion is entirely natural. Down markets are well-known to be more volatile than up ones, and while we attempt to control our volatility (to target as close to constant as we can) we know this process isn’t perfect. It’s always somewhat backwards looking, and thus when market volatility is higher ours will be too (though, again, without a directional bet, and hopefully a more muted increase than the market itself). Remember, we’re attempting to design a diversifying source of return not a direct hedge for the market; and the evidence, even including this year and this October of pain, shows we’ve (so far) succeeded.
Furthermore, note the dashed line in the scatter plot above. That’s the linear regression line. I’ll discuss this more below but two quick observations. One, it crosses the y-axis above zero. That implies we’ve generated positive realized alpha against a simple market model. Two, the regression line is mildly negatively sloped. That means, if anything, we have had a very small negative market beta (though I think that’s noise given that we target a zero beta, and it’s so close to zero). Both of these we believe are good things long-term.
OK, so October stunk but isn’t out of the range of what we should expect occasionally given our intended strategy design. But our analysis doesn’t end there. Let’s look at some other measures of “normality” (both in the geeky sense of the normal distribution and more colloquially as measures of how “well-behaved” our alts and the market are). Investment returns are generally not perfectly “normally” distributed. To some degree almost all investment returns seem to be somewhat “fat-tailed.” This is most extreme for, say, strategies that take big positions in options (betting on volatility), but even non-option exposures usually show some non-normality. I’m going to go through a bunch of measures that I think are interesting for determining how “well-behaved” an investment has been. To give some context, I’m also going to show the same measures for the world equity market portfolio itself. I will also study two hypothetical portfolios (both monthly rebalanced which I know is kind of unrealistic, but I’m doubtful matters much versus say annual rebalancing). One is 50% in the representative AQR liquid alts portfolio and 50% in the world stock market (also unrealistic as perhaps only we would go all the way to 50/50 in our own portfolios, but it’s more interesting to examine! 5 5 Close While 50/50 seems extreme our realized volatility has been about 1/3 of the world equity market’s volatility, so, in “risk-space”, this is still mostly a traditional market portfolio. ). As an alternative, and more realistic example, we will use a 60/40 global stocks / global bonds (Barclay’s Global Aggregate USD-Hedged) portfolio and then add in our representative liquid alts portfolio at a 25% weight. Here goes. Hypothetical results below 6 6 Close While 50/50 seems extreme our realized volatility has been about 1/3 of the world equity market’s volatility, so, in “risk-space”, this is still mostly a traditional market portfolio. :
• Over the full-period, including this painful year, our representative liquid-alts portfolio has returns of 4.5% over cash at 5.9% volatility for a 0.76 Sharpe ratio using monthly data. The world equity market has returns of 5.7% over cash, but at 15.1% annualized monthly volatility, for a Sharpe ratio of 0.38 over the same time. The 50/50 portfolio results in 5.1% over cash at 8.0% volatility for a Sharpe ratio of 0.64 — substantially better than the market for only a 60 basis point give-up in performance. 7 7 Close In real life you could source your investment in alts not 100% from stocks as implicitly done here, but also from lower risk assets and convert this gain in Sharpe ratio to higher expected returns not simply lower risk (or equal expected returns at lower risk). In fact, as we are about to show, that’s what our 60/40 plus alts portfolio begins to get at, though if we funded the alts out of bonds the expected returns would actually increase. This increase in Sharpe ratio is a key reason investors seek uncorrelated liquid-alts, and it has indeed shown up live—even including this annus horribilis.
• While global stocks have a 0.38 Sharpe ratio over this period, the 60/40 portfolio comes in at an improved 0.51 due to the diversifying power of bonds. The mean return of the 60/40 portfolio is, of course, lower than for 100% global stocks, coming in at 4.6% per annum over cash where global stocks returned 5.7% over cash. But the volatility of the 60/40 portfolio is way lower at 9.0% per annum vs. 15.1% for just stocks, leading to the considerably higher Sharpe ratio above. Now imagine you started with this improved (versus 100% equity) portfolio but only invested 75% in it and 25% in the representative AQR liquid alts portfolio (again, including the results of this terrible year). The mean annualized monthly return stays at 4.6% (the compounded return would actually rise slightly from better compounding at lower volatility), but the volatility falls considerably, down to 6.9% for a Sharpe ratio of 0.67 (vs. 0.51 for 60/40 and 0.38 for global stocks). The worst 6-months (additive excess over cash returns 8 8 Close To make them consistent with the Sharpe ratio calculations, all returns here are additive monthly excess, not compound returns. The story doesn’t really change at all using compound returns. ) for this portfolio is now a very ugly -23% vs. a horrendously ugly -30% for the simple 60/40 portfolio (sorry, 25% moved into an alternative can make things better but cannot cure all ills) and a ridiculously horrendous -52% for 100% equities. These are material improvements which, again, include this exceptionally bad year. If one wanted to use these results for still more expected return at volatility no higher than 60/40 itself, one would have to apply very modest leverage to the whole, or fund the investment more out of bonds, a topic for another day (and discussed here in the different context of risk parity). Bottom line, for realistic allocations, over our live period, and including 2018, we see that an allocation to our alternatives has significantly improved on a more basic allocation. Just not very recently!
• The realized monthly skewness (a measure of how fat-tailed returns have been on the downside) of our pure alts portfolio has been -0.2. That’s trivially non-normal (a normal distribution would be 0.0). The market is a bit more non-normal coming in with a realized skewness of -0.8, and the 50/50 portfolio (for simplicity I’m going to switch back to only looking at this 50/50 global stocks / AQR liquid alts portfolio going forward) also clocks in at -0.8.
• The realized monthly excess kurtosis (a measure of how fat-tailed returns have been in both directions) of our AQR liquid alts portfolio has been 1.2, a little more than trivially non-normal (a normal reading would be 0.0). 9 9 Close Remember we think most investing series show some degree of “non-normality.” We’d note that looking at rolling 10-month returns (to compare all similar length periods over the last 18 or so years to this year-to-date), our best such period is still about doubly as good as this year is horrible (which doesn’t make it doubly statistically anomalous given our positive average return, i.e., losses are generally more surprising than gains). The market however is, again, less well-behaved coming in with a realized kurtosis of 2.1. 10 10 Close I’m only examining monthly and longer-term data here. Both our market-neutral strategies (i.e., August 2007) and the long-only market itself (i.e., Oct 1987, The Flash Crash) are even more non-normal at very short horizons, as, again, we believe are most financial assets that are marked-to-market (the others are too… you just don’t see it!). On this measure, diversification hasn’t helped, but also has not been too bad, increasing kurtosis to 2.3. So, on monthly kurtosis we’re better behaved than long-only stocks, but the 50/50 portfolio is still “fatter tailed” than normal distributions would imply. It’s really hard to get rid of monthly kurtosis!
• Our portfolio is very slightly positively auto-correlated (good and bad months tend to follow each other a bit more than if they were truly fully independent) with a realized monthly auto-correlation of 0.08. However, again, we’re better behaved than the market whose auto-correlation over this period has been 0.17 (and the 50/50 portfolio is almost the same at 0.18).
• While monthly autocorrelation gets at the idea that portfolios can be “streaky,” it doesn’t always cover all possibilities. We could be streaky in a way than what would not show up just looking monthly. To get at this, I also examine rolling 6-month returns. Some results:
◦ Another way streakiness can show up is by causing bigger realized ups and downs over longer periods (i.e., a lot of small monthly returns in the same direction can show up as a big longer-term return—either positive or negative). Our annualized volatility calculated using rolling 6-month data is slightly bigger than that using monthly data (the rolling 6-month volatility is 1.03x the monthly volatility). That’s pretty tiny, but it is in the undesirable direction (bigger 6-month returns in both directions than you’d get if returns were independent month to month leads to slightly bigger 6-month vs. 1-month annualized volatility). However, again, the market is more extreme. The global stock market’s annual volatility calculated over this same period using rolling 6-month periods is 1.20x larger than just using monthly data. The 50/50 portfolio is again quite similar to the market (1.23x larger—though, of course, still much lower total volatility and worst cases than the market itself).
◦ The autocorrelation of 6-month returns (the correlation of each 6-month period with the prior 6-month period for the same investment) is -0.13 for us (so, while I wouldn’t trade on this, we find that good 6-month periods tend to follow bad and vice versa to a slight degree). The market comes in at a 0.00 and the 50/50 portfolio at 0.04.
◦ Looking at 6-month rolling returns, our skewness has been -0.7, again showing the slightly larger than “normal” left tail common to many, even most, investments. But, again, the market is worse-behaved with a 6-month skewness of -1.3 (the 50/50 portfolio is the same).
◦ Again, the worst 6-months for the representative portfolio of AQR liquid alts is -13% (additive monthly returns versus cash). For the global stock market it’s -52%. For the 50/50 portfolio it’s -28%. Basically cutting the worst 6-months almost in half for a very small reduction in top line return seems like a pretty good trade-off to me (and again if one didn’t allocate to alts just from high expected return equities, that average return hit can likely be reduced or reversed).
While we’re here we must do some regressions (would you expect anything different?). Let’s first regress the representative monthly AQR liquid alts portfolio on just the world equity market (that regression line is the dashed one in the scatter plot above). Using net of fees, excess over cash returns for both us and the world equity market (abbreviated WMKT), here are the results with the alpha annualized (and with t-statistics in parentheses under each coefficient):
See end of article for further disclosures related to this performance.
Even including this year (and October) that regression is quite satisfying. We have produced a trivial (and trivially negative) world equity beta and an economically and statistically significant alpha.
One more regression. Ken French's website, where we got the returns for the world market portfolio, also reports the returns for world-wide small minus large (the size effect or SMB), cheap minus expensive (the value effect or HML), and winner minus loser (the momentum effect or UMD) long-short portfolios.
If I made up the numbers (I did not do that!) I couldn’t do much better (except, of course, it’s always better to add more to the intercept!). We should be correlated to world HML and UMD, as it’s part of what we do. But it’s not all of what we do. 12 12 Close We measure value and momentum somewhat differently from simple price-to-book and one-year price momentum—we hope better but only a very long time will tell, and, as the DELTA and Style Premium portfolios get added, we add more asset classes than just individual stocks—where individual stock long-short returns are all that’s covered in the French data (though there may be some embedded country bets, I’m not 100% sure of his methodology). The evidence is that after these two intentional exposures, which are both economically and statistically significant, we still have alpha, which is even more impressive considering the right-hand side is all gross of fees and transactions costs and, of course, not crediting us with the two intentional bets. 13 13 Close It just misses the conventional measure of statistical significance (>2), which is not very shocking when we’re removing the returns from basic value and momentum investing which, again, are also gross of fees and transactions costs—but if we keep up the pace of the last 18 or so years it will be statistically significant quite soon as statistical significance is a function both of the level of returns and how long you’ve maintained that level of returns. Importantly, recall that some of this discussion of “alpha” is semantics. If we are providing efficient and reasonably priced exposure to two good long-term sources of return, HML and UMD, that is very valuable to an investor’s portfolio, even if we choose to not call it “alpha” in the above regression. Our market beta is now a bit positive (though economically very small) presumably offsetting some negative realized beta in the combination of world HML and UMD. Finally, our size exposure is quite small, consistent with our design decision not to bet on the simple size factor.
Let’s sum up. The original question of “what about October?” is answered above through an analysis of our representative liquid alts portfolio. It always sucks to lose, but long-term, and very much including this year through October itself, our investment process, as demonstrated through the liquid alts portfolio has behaved as intended, looking like an uncorrelated “cloud” when plotted against world markets (again, a cloud nicely biased to the upper right quadrant). Then, on a host of measures, both monthly and looking at rolling 6-month returns, we show that we, like the market, have been slightly “non-normal” but not excessively so, and we’ve generally been better behaved than the world equity market itself. Perhaps most importantly we confirm that, again even including the Job-like ordeal of 2018 14 14 Close Job-like might slightly exaggerate 2018 as nothing “biblical” has happened to us. But remember our discussion in Ragnarök about how these years feel much much worse than they rationally should and seem to go on much much longer than they really do (what we call “time dilation”). , a portfolio of our liquid alts has generally been a good addition to a diversified world equity portfolio, and regression analysis confirms that it has also acted exceptionally consistent with the exposures that we target. 15 15 Close Nor do we take for granted that the future might not look like the attractive past. Much of my Ragnarök post was devoted to examining, I hope with a very open mind, this very issue.
This is undeniably an absolutely excruciating year that (again, like most other investments do in their worst periods) goes somewhat beyond what we “normally” expect. So far it shows little sign of brightening. But it’s also a year, based on our analysis above, that leaves our long-term liquid alts record sound. Importantly, I don’t just mean our record of attractive returns and low correlation to traditional markets, but also our record of being reasonably, though of course imperfectly, well-behaved, and better behaved on many measures than the market portfolio itself. Viewed against our goals and historical record, October stunk but was not abnormal at all. Finally, it’s gratifying that we’ve added value over some of the basic exposures we intentionally take (themselves hopefully also adding value long-term).
I’d sign on for all that over the next 18ish years. Though when a year like 2018 once again rolls around one day in the (hopefully distant) future, it, sadly, probably won’t be any less excruciating!