Killer Rabbits, the May 6 Flash Crash, and the Microstructure of the VIX Index
Contributor Since 2010
Note: The following article is an excerpt from the Second Quarter 2010 Letter to Investors from Artemis Capital Management LLC published on June 30, 2010.
In mythology the hero's journey often involves a chapter whereby he must use bravery and cunning to slay a dangerous supernatural beast that stands guard over a valuable treasure. In the storybooks the beast usually takes the form of a large dragon, minotaur, or demon - but in investing the beast can appear as less dramatic but equally destructive monsters like leverage and credit risk.
The classic film Monty Python and the Holy Grail depicts an epic battle with one such guardian beast, the supernatural Killer Rabbit of Caerbannog. In the film King Arthur and his Knights of the Round Table severely underestimate the lethal nature of their fluffy foe despite warnings from their sorcerer guide describing the animal as, "... the most foul, cruel, bad-tempered rodent you ever set eyes on!" A full frontal assault of the bunny amounts to a whirlwind of white fur, blood, chopped heads, and severed limbs and a quick retreat. To vanquish the rabbit, King Arthur orders the holiest of weapons, and the bunny is obliterated from afar by the sacred Holy Hand Grenade of Antioch. In the end, a creature as evil as the killer rabbit can only be destroyed by a weapon fashioned by God Himself.
The myth of the killer rabbit is allegory for all things that at first, are seemingly normal, but later reveal themselves to be extremely dangerous. We all probably believe that, if required, we would prevail in mortal combat with a bunny. We base this assessment entirely on our past experiences of the combat capabilities of rabbits. As King Arthur discovered, confidence in your past understanding of the world may be detrimental (and even lead to certain death) if your opponent is something entirely unique to your realm of experience.
As you will see, from a volatility perspective, the May 6th, 2010 Flash Crash was every bit as weird as a knight-decapitating killer rabbit. To fully understand why we must explore the microstructure of volatility, which describes how vol changes during the trading day (minute-by-minute, tick-by-tick).
I warned you, this market is no ordinary rabbit...
The Flash Crash and its aftermath emerged from nowhere as a one-of-a-kind nefarious volatility creature without strong historical precedent. Like many volatility traders, I came to the lair of the beast expecting a fire-breathing sovereign risk dragon whose roar would unleash a traditional self-reinforcing volatility surge. I was prepared for battle only to face a very different type of monster than expected. As widely reported, on Thursday, May 6th, a routine down day turned into a bloodbath when, for reasons still unknown, at 2:42pm EST the DJIA dropped 1,000 points in two minutes. Then, without rationale, by 2:47pm (less than 5 minutes later) the DJIA regained 500 points. Intra-day volatility (VIX index) climbed over 14 points in slightly over an hour and then dropped from 40+ to approximately 32. All-in-all between May 6th to May 10th we experienced the largest intra-day volatility move in history, the third largest two-day increase in history, followed by the single largest one-day decline in history.
For further evidence of how bizarre the Flash Crash was on a micro-level take note of the chart to the right that maps the minute-by-minute volatility of the VIX index (this graph used over a million data points to produce). The blue line measures how fast the VIX changed on average for every minute of the trading day between April 2004 and June 2010. The red line shows the volatility of volatility for each minute of the May 6th Flash Crash. As you can clearly see the average daily change in volatility is not even on the same magnitude scale as what was measured during the crash. What makes this even more exceptional is the fact that the data used to produce the daily average included the 2008 crisis, one of the most volatile periods in market history. If these volatility measurements could be compared to brainwaves this event was tantamount to an epileptic seizure.
A comparison of the Flash Crash against the other top-five highest ranked volatility of volatility days (since April 2004) shows that May 6th was an anomaly even among the outliers. On other extreme days high levels of volatility of volatility cluster around the opening of the market. This is logical, as market participants change risk perceptions based on new information since the prior close. In this sense volatility clustering at the open is not true minute-to-minute volatility but rather a rational adjustment to the fact markets are not continuous throughout time. In comparison, the volatility of volatility exhibited during the Flash Crash was relatively calm in the morning before exploding to extreme levels between 2:25pm and 3:30pm, and then calming without strong follow-through to the close. This is a true black swan pattern, even amongst a peer group of abnormal swans.
Historical Analysis of Intra-day Volatility of Volatility ("VOV")
Let's turn our attention to comparable days of high volatility of volatility across history. The tables above rank the days with the highest intra-day VOV with a simple formula: the daily volatility high, minus the volatility low, taken as as a percentage of the opening volatility. The first table demonstrates the highest ranked VOV days since inception of the new VIX index methodology in 2003. The second table is expanded to include the intra-day volatility measured with the VXO index (the predecessor to the VIX) from 1987 to 2003 and finishing with the VIX index from 2003 to 2010. While the Flash Crash registered as the largest intra-day movement of the VIX index in history, it ranks clearly behind the Black Monday Crash of October 19th, 1987 on the all- time list. It is important to note key differences between the two crashes: in the Flash Crash volatility surges were self-correcting (up movements followed by down movements and vice versa), but on Black Monday volatility surges were self-reinforcing (changes were consistently higher throughout the day). For example, on Monday October 19th, 1987 the VXO index opened up +34.96 points (to 71.33) above the previous closing level of 36.37 and continued to surge closing near highs of the day at above 150. Although the magnitude of the Black Monday crash was unprecedented, the pattern of the volatility of volatility movements was relatively conventional with heavy movements concentrated at the open and self-reinforcing into the close (exact minute-by-minute data for the VXO index is not available for the October 19th, 1987 crash). In contrast, the self-correcting volatility changes during May 6th, 2010 Flash Crash included eight +/- 10% movements in the VIX index in the span of an hour with volatility finishing much lower than its intra-day high.
The 1987 crash was the king of scary monster, but from a volatility behavior perspective, a traditional dragon compared to this bizarre zig-zagging killer rabbit Flash Crash. I am surprised you never hear about the February 27th, 2007 mini-crash that was sparked by a 9% sell-off in China. That day represented one of the largest intra-day volatility spikes in history and the sixth largest jump in 21-day realized volatility of the S&P 500 index since 1952. The 2007 spike was a multiple standard deviation event that foreshadowed the coming recession, but was largely ignored and forgotten after volatility fell back to previously low levels, at least until the credit crisis ramped up later that year.
Market phases of elevated intra-day volatility of volatility are not a new phenomenon and often foreshadow impending recessions. This point is further demonstrated by the chart on the next page which depicts the evolution of intra-day VOV since the late 1980s (while also providing a glance at changes in VOV microstructure since 2003).
In the media much has been made about the role of high-frequency trading strategies as a contributor to higher vol, and regulators are attempting to shine light into this opaque area of the trading landscape. I strongly believe the impact of high frequency trading on the May 6th Flash Crash should be fully examined, but to blame the higher volatility since 2007 on these funds is simply ignoring the powerful but fundamental role of human fear. The conspiracy theorists should be reminded that we experienced higher levels of intra-day VOV throughout 1987 and 1994, a time when many of today's high frequency programmers were playing Dungeons and Dragons in high school. Despite this fact, the daily high-low-open VOV measurement (represented by the bars) is less nuanced than the average minute-to-minute VOV (represented by the red line) and understanding the effects of high frequency trading will benefit from a longer microstructure history. In the future it will be interesting to further monitor trends in the relationship between microstructure volatility (minute-to-minute changes) and traditional volatility (close-to- close changes).
What is the meaning of the Killer Rabbit?
Was the Flash Crash the tipping point whereby microstructure volatility began driving macrostructure volatility? More importantly, will this represent any fundamental change in market dynamics going forward?
You may be familiar with Moore's Law that states that the rate of computing power (or theoretically technology) doubles approximately every eighteen months in an exponential growth curve. The follow-on to that theory is as humankind progresses along Moore's exponential growth curve the likelihood of extreme events (black swans) from the use of that technology also increases exponentially.
If this is true consider the Flash Crash a very scary development for the market. Many professional investors find it unsettling that regulators still have no definitive idea as to what caused this crash. In their initial findings the Securities Exchange Commission attributed the cause to an abrupt pullback in liquidity in both equity and futures markets. Although there is no evidence of either, they still cannot completely rule out cyber-terrorism or human error. For the next six months a new circuit breaker system is being tested that would temporarily pause trading for five minutes in any individual S&P 500 stock that moves 10% or more in a five-minute period. These changes to market infra-structure should at least provide the illusion of control.
The existential interpretation of the Flash Crash is that it represents a purely random event that it is unlikely to ever happen again. In this view, the killer rabbit doesn't have any greater meaning beyond being a freak bunny we just happen to have the bad luck of encountering in an expansive forest. Maybe we never see anything like it again, maybe we do, but as a glitch in the system it is ultimately irrelevant to our future view of the markets. To some this is a deeply unsatisfying conclusion, but potentially accurate.
If you prefer to assign fundamental meaning to the Flash Crash take note that other unexplained volatility shock events in history were bad omens foreshadowing future recessions. The most notorious intra-day volatility events (see earlier table) all occurred in years just prior to the downturns of 1990, 2001, and 2008. Volatility shocks may represent the temporary meltdown of the market simulacrum and pose the question whether economic growth, as we perceive it, is a fabrication. As the artificial glow of government stimulus wears off our economy looks increasingly susceptible to a second recession. For evidence look no further than the Economic Cycle Research Institute Weekly Leading Index which recently dropped to a 56 week low of -7.7% in late June. In the past anything below -10% has guaranteed a recession. It should be noted that the economists at the ECRI don't expect a double dip recession, but this fundamental weakness coupled with technical market shocks don't help my confidence. The market increasingly seems to be driven more by fear and short covering rather than fundamental trust in the bedrock of the economy. Either way, higher volatility is here to stay until the economy can prove one-way-or- another that a second recession is not in the cards. 
Sometimes it doesn't help to waste time wondering why the killer bunny exists when you just need to exterminate the damn thing (or at least contain it). The last quarter was a bruising reminder that sometimes the most treacherous monsters appear from seemingly harmless forms. Like King Arthur, many traders were caught off guard by an unnatural bi-polar beast of volatility that involved some of the largest and fastest VIX declines in recorded history. The Flash Crash and its aftermath was every bit as surreal as a killer rabbit - a supersonic white blur attacking with flash speed and razor sharp teeth. On this new quest into the volatility microstructure he (or she) whom is brave and cunning enough in the wake of volatility dragons, killer bunnies, and the other guardian beasts of the market will be find the treasure. God speed.
Sincerely,
Artemis Capital Investors, L.P.
Christopher R. Cole, CFA
Managing Partner and Portfolio Manager
Artemis Capital Management, L.L.C.
Disclosure: Artemis Capital Investors LP maintains volatility positions in an array of financial indices.