While on the surface the market last week did nothing all that exciting, below it things were in abrupt turmoil - driven by the decoupling between stocks and bonds and the volatile, countertrend move in commodities and oil in particular - which was nowhere more evident than in the world of Risk-Parity funds and CTA, which suffered their worst two-week plunge since 2003.
A subsequent report from Bloomberg revealed that the damage among trend-following CTA was especially severe, "by some measures, commodity trading advisers are on track to post the worst yearly return since 1987, when data were first collected on the group." It also prompted the WSJ to write "Oil Up? Oil Down? Blame the Algorithms."
But what really happened last week, and will it happen again?
For the answer we go to one of the foremost vol experts on Wall Street, the team of Chintan Kotecha, Ben Bowler et al at Bank of America, who today described what took place last week “Quant quake”, and who continues a long trend of pointing out just how "weird" and fragile the market is (no really, in late May he wrote "While not obvious on the surface, these Markets Are Very Weird") by noting that markets continue to set long-term records for price instability or “fragility”, with a five standard deviation (5-sigma) sell-off in the S&P 500 on 17-May, a 3-sigma drop in the Nasdaq 100 on 9-Jun, and most recently a sharp rise in the bank's cross-asset Fragility Indicator.
The latest examples? A 3-sigma rally on average across commodities last week (chart 7) that was even more acute in agricultural commodities like wheat, soybeans, and soybean meal, and a 3-sigma sell-off on average across global bonds (chart 8) that was most acute in longer-duration Bunds. But the worst news for the systematic funds is that the average CTA (Commodity Trading Advisor) was caught offside by both moves, with a popular CTA benchmark experiencing a nearly 5-sigma drawdown last week (chart 9) its worst since 2000 (outside of the Feb-07 shock).
Focusing on CTAs in particular, BofA notes that trend followers were particularly hurt by outsized reversals in bonds and commodities. In a prior report, BofA showed that a significant portion of assets in CTAsappear to be in cross-asset, risk controlled, trend following strategies and, more specifically, CTA allocations across multiple asset classes are primarily a function of two factors: (1) the asset’s current price trend and (2) the asset’s prevailing volatility. Consequently, BofA writes that "the largest downside risk for CTAs comes from reversals in price trends that are driven by ‘high-sigma’ moves. By high-sigma, we mean moves that are sufficiently large in absolute terms relative to prevailing volatility."
Just like what happened last week. And, as BofA confirms, the recent historical decline for CTAs appears to have been driven in large part by high-sigma moves in Fixed Income and Commodity asset class futures.
To see this, first start with a universe of futures investments across Equity, Fixed Income, Commodity, and Currency asset classes (refer to footnote in Chart 10 for more information). Then, for each cross-asset futures investment, quantify its trend strength (from -1 (short) to 1 (long)) and measure its volatility. Calculate the percentile of the ratio of trend strength to volatility over a long horizon. High percentiles are indicative of more long futures positions via CTAs and vice-a-versa for low percentiles and shorts. Next compare a given day’s sigma-move (daily move over prevailing volatility) to the prior day’s percentile of the ratio of trend strength to volatility. The upshot: large negative sigma moves for high percentile ratios indicate longs that were subject to losses and vice-aversa for large positive sigma moves and shorts
The above chart presents this analysis for cross-asset futures daily moves from 26-Jun-2017 through 7-Jul-2017. The boxes in the lower right and upper left corners indicate positions that may have driven declines for CTAs over the last two weeks. Specifically, it looks like high-sigma declines in Fixed Income as well as high sigma increases in Commodities could have been a large driver of recent underperformance. BofA also adds that equity futures may have also contributed to losses as they appear in the lower right corner but in this case via less outsized sigma-moves (that is, the recent grind lower in equities may have also been a driver of poor CTA performance).
* * *
What happens next? Well, if last week was any indication, anywhere CTAs are overexposed is suddenly risky. In the aftermath of last week's commodity and fixed income mauling, two such leftover spaces are equities, where CTA remain long, and the USD, where they are short.
As Kotecha writes, the reversal in trend alongside a pickup in volatility for Fixed Income and Commodity asset classes could indicate that CTAs have now limited their exposure in those areas. However, the percentile of the ratio of trend strength to volatility remains elevated for a collection of global equity index futures as well as for certain currency pairs.
Furthermore, despite the recent sharp reversals in US 10Yr Bonds and Oil, their percentiles remain stretched, which could be indicative of a long bond and short oil position. However, CTAs do appear to have risk control mechanisms that include stop losses which may have been triggered in these two asset classes (hence why in Chart 11 BofA has depicted the US 10Yr Bond and Oil with hollow bars).
In summary: while the risk of a CTA unwind in oil and bonds is now negligible, it is quite high for equities and the USD.
* * *
What about other systematic strategies, of which the most notable is of course, risk-parity. As a reminder, risk parity strategies are often also considered alongside CTAs as they both (1) use rules-based models that can at times make them price-insensitive buyers or sellers, (2) typically increase leverage when volatility is lower, and (3) can deleverage in response to a shock from low vol levels.
As we also showed last Friday, risk-parity strategies were performing well year-to-date but like CTAs were also subject to sharp declines over the last two weeks. The just released BofA chart show the same.
However, while price trends and vols indicate CTAs may have unwound Fixed Income and Commodity positions, according to BofA risk parity portfolios have yet to adjust their leverage. In other words, despite the volatile market gyrations, risk-par did not deleverage. For risk parity, it is important to distinguish between changes arising from slower-moving shifts in cross-asset allocation versus dynamic adjustments in leverage due to target volatility overlays. Typically the larger and more significant deleveraging comes from vol control overlays. The amount of deleveraging is a function of a risk parity strategy’s target volatility and maximum leverage allowed. But more significantly, the deleveraging is a function of the prevailing volatility prior to a large move and the specific magnitude of those large moves.
Furthermore, as we repeatedly pointed out during last week's coordinated selloff between equities and stocks, i.e. surging correlation, the risk of a rapid deleveraging has spiked. Here's BofA:
Equity/bond correl has risen sharply, but asset volatilities simply remain too low Negative Equity and Fixed Income correlation (correlation between equity and bond price returns) has provided diversification for risk parity portfolios year-to-date and helped drive unlevered risk parity vol to multi-decade lows. However, the recent rise in equity/bond correlation has increased focus on the potential deleveraging pressure from risk parity funds.
Despite the changing equity/bond correlation dynamic, unlevered risk parity vol remains muted in part because Equity, Fixed Income, and Commodity component vols have not risen sufficiently through this recent downturn. To see this, using a target volatility overlay (10% target vol, max leverage 2x) on a hypothetical risk parity investment, we tallied in Table 2 3-day changes in leverage and ranked the 10 largest deleveraging events since the early 1970s. For each event, we also compiled component vols, pairwise correlations, and unlevered risk parity portfolio vol. Then, we listed the current component vols, correlations, and portfolio vol.
Indeed, despite the rise in equity/bond correlation (see red box in Table 2), component and unlevered risk parity portfolio vol still remain well below the levels observed during the most significant historical episodes that could have led to a theoretical deleveraging in vol controlled risk parity portfolios.
In plain English, the above means that while pairwise correlations between the two key risk-parity asset classes have spiked, overall asset volatility still remains low enough. However, another steep cross-asset selloff, coupled with the VIX spiking above the "Kolanovic Line" of 15 or so, and things get interesting.
* * *
Putting the above together, with the "quant" space now almost exclusively still long equities, does this suggest that the crash risk for the asset class is higher?
Here is one calculation from BofA according to which while risks are higher, an imminent crash is probably unlikely, absent some gating factor (for more on that see the recent Harley Bassman note we posted).
By our estimates, global equity index futures volume is currently about $300bn while fixed income and commodity futures turnover another $300bn and $100bn daily (Chart 13). Next assume that longer-term average equal risk weighted allocations to Equity, Fixed Income, and Commodity asset classes are near 22%, 54%, and 22% respectively. Or assume that an equal risk weighted exposure that includes FX would allocate on average 20% to Currencies and 15%, 50%, and 15% to Equity, Fixed Income, and Commodity asset classes. In either case, global equity futures markets are at least twice as liquid as both bond and commodity futures when accounting for the typical risk-weighted exposures applied by CTAs and risk parity. Moreover, once spread out over multiple days to reflect realistic model diversity, even estimates on the order of $125-225bn of global equities potentially for sale would amount to only 4%-8% of weekly futures volumes in high stress periods.
While is good news for the bulls, unless of course BofA's concentration and leverage assumptions are off. In any, the bank does warn that "both strategies remain at risk for selling material equity positions." Incidentally, nobody predicted last week's CTA crash, and yet it happened. It is logical that as more assets are allocated to a smaller universe of asset classes (i.e. stocks), the higher the delta to underlying vol shifts, especially when these become self-sustaining, and lead to an avalanche of selling once vega thresholds for short-vix funds are broken.
And while nobody can predict the next crash, we'll close with what BofA wrote two months ago for anyone who still harbors the false belief that these "markets" make sense.