Volatility During Crises

[The following first appeared in the August 2011 edition of Expiring Monthly: The Option Traders Journal. I thought I would share it because it might help some readers put the current fiscal cliff crisis in historical context.]

The events of the last three weeks are a reminder that financial crises and stock market volatility can appear almost instantaneously and mushroom out of control before some investors even have a chance to ask what is happening. A case in point: on August 3^rd investors were breathing a sigh of relief after the United States had finalized an agreement to raise the debt ceiling; at that time, the VIX stood at 23.38, reflecting a relative sense of calm, yet just three days later, the VIX jumped to 48.00 as two new crises displaced the debt ceiling issue.

Spanning the globe from Northern Africa, Japan, Europe and the United States, 2011 has seen no shortage of crises in the first eight months of the year. Given this pervasive crisis atmosphere, it is reasonable for investors to consider how much volatility they should anticipate during a crisis. In this article I will attempt to put crises and volatility in some historical perspective and address a variety of factors that affect the magnitude and duration of volatility during a crisis, drawing upon fundamental, technical and psychological causes.

Volatility in the Twentieth Century

Every generation likes to think that the issues of their time are more daunting and more complex than those faced by prior generations. No doubt investors fall prey to this kind of thinking as well. With a highly interconnected global economy, a news cycle that races around the globe at the speed of light and high-frequency and algorithmic trading systems that have transferred the task of trading from humans to machines, there is a lot to be said for the current batch of concerns. Looking at just the first half of the twentieth century, however, investors had to cope with the Great Depression, two world wars and the dawn of the nuclear age.

Given that the CBOE Volatility Index (VIX) was not launched until 1993, any evaluation of the volatility component of various crises prior to the VIX must rely on measures of historical volatility (HV) rather than implied volatility. As the S&P 500 index on which the VIX is based only dates back to 1957, I have elected to use historical data for the Dow Jones Industrial Average dating back to before the Great Depression. In Figure 1 below, I have collected peak 20-day historical volatility readings for selected crises from 1929 to the present.

Before studying the table, readers may wish to perform a quick exercise by making a mental list of some of the events of the 20^th century that constituted immediate or deferred threats to the United States, then compare the magnitude of that threat with the peak historical volatility observed in the Dow Jones Industrial Average. If you are like most historians and investors, after looking at the data you will probably conclude that the magnitude of the crisis and the magnitude of the stock market volatility have at best a very weak correlation.

[source(s): Yahoo]

Any ranking of crises in which the Cuban Missile Crisis and the attack on Pearl Harbor rank in the lower half of the list is certain to raise some eyebrows. Frankly I would have been surprised if even one of these events failed to trigger a historical volatility reading of 25, but seeing that was the case for half the crises on this list certainly provides a fair amount of food for thought.

Volatility in the VIX Era

With the launch of the VIX it became possible not only to evaluate historical volatility, but implied volatility as well. With only 18 years of data to draw upon, there is a limited universe of crises to examine, so in the table in Figure 2 below, I have highlighted the seven crises in the VIX era in which intraday volatility has reached at least 48. Additionally, I have included five other crises with smaller VIX spikes for comparison purposes.

[source(s): CBOE, Yahoo]

[Some brief explanatory notes will probably make the data easier to interpret. First, the crises are ranked by maximum VIX value, with the maximum historical volatility in an adjacent column for an easy comparison. The column immediately to the right of the MAX HV data captures the number of days from the peak VIX reading to the maximum 20-day HV reading, with negative numbers (LTCM and Y2K) indicating that HV peaked before the VIX did. The VIX vs. HV column calculates the amount in percentage terms that the peak VIX exceeded the peak HV. The VIX>10%10d… column reflects how many days transpired from the first VIX close above its 10-day moving average to the peak VIX reading. The SPX Drawdown column calculates the maximum peak to trough drawdown in the S&P 500 index during the crisis period, not from any pre-crisis peak. The VIX:SPX drawdown ratio calculates the percentage change in the VIX from the SPX crisis high to the SPX crisis low relative the percentage change in the SPX during the same period (of course these are not necessarily the VIX highs and lows during the period.) The SPX low relative to the 200-day moving average is the maximum amount the SPX fell below its 200-day moving average during the crisis. Finally, the last two columns capture the number of consecutive days the VIX closed at or above 30 during the crisis and the number of days the SPX closed at least 4% above or below the previous day’s close during the crisis.]

Looking at the VIX era numbers, it is not surprising that the financial crisis of 2008 dominates in many of the categories. Reading across the rows, one can get an interesting cross-section of each crisis in terms of various volatility metrics, but I think some of the more interesting analysis comes from examining the columns, where we can learn something not just about the nature of the crises, but also about volatility as well. One important caveat is that the limited number of data points does not allow for this to be a statistically valid sample, but that does not preclude the possibility of drawing some potentially valuable and actionable conclusions.

Looking at the peak VIX reading relative to the peak HV reading I note that in all instances the VIX was ultimately higher than the maximum 20-day historical volatility reading. In the five lesser crises, the VIX was generally 50-80% higher than peak HV. In the seven major crises, not surprisingly HV did approach the VIX in several instances, but in the case of the 9/11 attack and the 2010 European sovereign debt crisis the VIX readings grossly overestimated future realized volatility.

One of my hypotheses about the time between the first VIX close above its 10-day moving average and the ultimate maximum VIX reading was that the longer the period between the initial VIX breakout and the maximum VIX, the higher the VIX spike would be. In this case the Long-Term Capital Management (LTCM) and 2008 crises support the hypothesis, but the data is spotty elsewhere. The current European debt crisis, Asian Currency Crisis of 1997 and 9/11 attack all reflect a very rapid escalation of the VIX to its crisis high. In the case of the May 2010 ‘Flash Crash’ and the Fukushima Nuclear Meltdown, the maximum VIX reading happened just one day after the initial VIX breakout. As many traders use the level of the VIX relative to its 10-day moving averages as a trading trigger, the data in this column could be of assistance to those looking to fine-tune entries or better understand the time component of the risk management equation.

Turing to the SPX drawdown data, the Asian Currency Crisis stands out as one instance where the VIX spike seems in retrospect to be out of proportion to the SPX peak to trough drawdown during the crisis. On the other side of the ledger, the drawdown during the Dotcom Crash appears to be consistent with a much higher VIX reading. Here the fact that it took some 2 ½ years for stocks to find a bottom meant that when the market finally bottomed, investors were somewhat desensitized and some of the fear and panic had already left the market, which is similar to what happened at the time of the March 2009 bottom. Note that the median VIX:SPX drawdown ratio for all twelve crises is 10.0, which is about 2 ½ times the movement in the VIX that one would expect during more normal market conditions.

The data for the SPX Low vs. 200-day Moving Average is similar to that of the SPX drawdown. For the most part, any drawdown of 10% or more is likely to take the index below its 200-day moving average. In the seven major crises profiled above, all but the Asian Currency Crisis dragged the index below its 200-day moving average; on the other hand, in all but one of the lesser crises the SPX never dropped below its 200-day moving average. Based on this data at least, one might be inclined to include the 200-day moving average breach as one aspect which helps to differentiate between major and minor crises.

As I see it, the last two columns – consecutive days of VIX closes over 30 and number of days in which the SPX has a 4% move – are central to the essence of the crisis volatility equation. Since the dawn of the VIX, the SPX has experienced a 2% move in about 80% of its calendar years, the VIX has spiked over 30 about 60% of the years, and the SPX has seen at least one 4% move in about 40% of those years. Those 4% moves are rare enough so that they almost always occur in the context of some sort of major crisis. In fact, one could argue that a 4% move in the SPX is a necessary condition for a financial crisis and/or a significant volatility event.

Fundamental, Technical and Psychological Factors in Crisis Volatility

Crises have many different causes. In the pre-VIX era, we saw a mix of geopolitical crises and stock market crashes, where the driving forces were largely fundamental ones. During the VIX era, I would argue that technical and psychological factors become increasingly important. The rise of quantitative trading has given birth to algorithmic trading, high-frequency trading and related approaches which place more emphasis on technical data than fundamental data. At the same time, retail investing has been revolutionized by a new class of online traders and the concomitant explosion in self-directed traders. This increased activity at the retail level has added a new layer of psychology to the market.

In terms of fundamental factors, one could easily argue that the top nine VIX spikes from the list of VIX era crises all arise from just two meta-crises, whose causes and imperfect resolution has created an interconnectedness in which subsequent crises are to a large extent just downstream manifestations of the ripple effect of the original crisis.

The first example of the meta-crisis effect was the 1997 Asian Currency Crisis, which migrated to Russia in the form of the 1998 Russian Ruble Crisis, which played a major role in the collapse of Long-Term Capital Management.

The second example of meta-crisis ripples begins with the Dotcom Crash and the efforts of Alan Greenspan to stimulate the economy with ultra-low interest rates. From here it is easy to draw a direct line of causation to the housing bubble, the collapse of Bear Stearns, the 2008 Financial Crisis and the recurring European Sovereign Debt Crisis. In each case, the remedial action for one crisis helped to sow the seeds for the next crisis.

In addition to the fundamental interconnectedness of these recent crises, it is also worth noting that the lower volatility crises were largely point or one-time-only events. There was, for instance, only one Hurricane Katrina, one turn of the clock for Y2K and one earthquake plus tsunami in Japan. As a result, the volatility associated with these events was compressed in time and accordingly the contagion potential was limited. By contrast, the major volatility events are more accurately thought of as systemic threats that ebbed and flowed over the course of an extended period, typically with multiple volatility spikes. In the same vein, the attempted resolution of these events generally included a complex government policy cocktail, whose effects were gradual and of largely indeterminate effectiveness.

Apart from the fundamental thread running through these crises, I also believe there is a psychological thread that sometimes spans multiple crises. Specifically, I am referring to the shadow that one crisis casts on future crises that follow it closely in time. I call this phenomenon ‘disaster imprinting’ and psychologists characterize something similar as availability bias. Simply stated, disaster imprinting refers to a phenomenon in which the threats of financial and psychological disaster are so severe that they leave a permanent or semi-permanent scar in one’s psyche. Another way to describe disaster imprinting might be to liken it to a low-level financial post-traumatic stress disorder. Following the 2008 Financial Crisis, most investors were prone to overestimating future risk, which is why the VIX was consistently much higher than realized volatility in 2009 and 2010.

While it is impossible to prove, my sense is that if the events of 2008 were not imprinted in the minds of investors, the current crisis atmosphere might be characterized by a much lower degree of volatility and anxiety.

Conclusion

As this goes to press, the current volatility storm is drawing energy from concerns about the European Sovereign Debt Crisis as well as fears of a slowdown in global economic activity. The rise in volatility has coincided with a swift and violent selloff in stocks that has seen six days in which the S&P 500 index has moved at least 4% either up or down – a rate that is unprecedented outside of the 2008 Financial Crisis.

Ultimately, the severity of a volatility storm is a function of both the magnitude and the duration of the crisis, as well as the risk of contagion to other geographies, sectors and institutions. Act I of the European Sovereign Debt Crisis, in which Greece played the starring role, can trace its origins back to December 2009. In the intervening period, it has spread across Europe and has sent shockwaves across the globe.

By historical standards the volatility aspect of the current crisis is more severe than at any time during World War II, the Cuban Missile Crisis and just about any crisis other than the Great Depression, Black Monday of 1987 and the 2008 Financial Crisis.

In the data and commentary above, I have attempted to establish some historical context for volatility during various crises extending back to 1929 and in the process give investors some metrics for evaluating current and future volatility spikes. In addition, it is my hope that concepts such as meta-crises and disaster imprinting can help to bolster the interpretive framework for investors who are seeking a deeper understanding of volatility storms and the crises from which they arise.

Related posts:

• The Biggest VIX Spike Ever: A Retrospective
• VIX Sets Some New Records, Suggesting Volatility Near Peak
• Highest Intraday VIX Readings
• Short-Term and Long-Term Implications of the 30% VIX Spike
• VIX Spike of 35% in Four Days Is Short-Term Buy Signal
• Lessons from the Post-2/27 VIX Price Action
• VIX Approaches Pre-2008 Record Highs
• VXO Chart from 1987-1988 and Explanation of VIX vs. VXO
• Volatility History Lesson: 1987
• Chart of the Week: VXV and Systemic Failure
• Forces Acting on the VIX
• A Conceptual Framework for Volatility Events

Disclosure(s): none

VIX Spike of 35% in Four Days Is Short-Term Buy Signal

It has been awhile since I posted one of my VIX studies and given the recent spike in the VIX, today’s action seemed like a good excuse to revisit the idea of VIX spikes as contrarian bullish mean reversion buying opportunities.

Today the VIX closed at 27.91, which is up 34.9% in the four trading days since Thursday’s close of 20.69. Over the course of the 20 year history of the VIX, the volatility index has posted close-to-close four day gains of 35% on 42 occasions. If you strip out the consecutive instances of +35% days, this leaves 27 instances in which the VIX crossed above +35% in four days. I have reproduced the full table of these 27 instances below for several reasons. First, the key takeaway is that from a timing perspective, a long SPX position entered after a 35% spike will generally perform best over the course of a five day time horizon. In the graphic below, the 27 instances average a five day gain of 1.99% vs. a typical five day SPX return of 0.14%, for a 1.85% net differential. While the net differential peaks at five days, it is apparent in just one day and persists for at least fifty trading days.

Not surprisingly, since we are talking about extremely volatile periods, quite a few of the returns are at the tail end of the distribution and are highlighted in green and red. In particular, this contrarian long strategy did an excellent job of timing the bounces off of the lows during the Asian financial crisis in 1997, the Russian financial crisis and Long-Term Capital Management crisis in 1998, the 9/11 World Trade Center attacks in 2001, as well as the technology bottom following the WorldCom bankruptcy filing in 2002.

In contrast to the excellent market timing above, it should also come as no surprise that the four long signals from September and October 2008 turned out to be disasters in the 20-100 day time frame. Over the course of a 3-10 day time horizon, however, these were excellent short-term trading opportunities from the long side. I do wish to point out, however, that if one strips out the last four rows of the tables, suddenly the 100 day time frame has a return of 6.86%, which is 2 ½ times that of the baseline (“census”) return. The obvious conclusion is that the VIX spike buy signal is quite reliable for the short-term, but not as reliable for over longer time frames. This is the key take away from the table and the reason I included the full data set. The secondary conclusion is that VIX spikes are generally good long setups as well, but here the risk is that it precedes a once in a generation or two meltdown that erases a decade or more of returns.

Finally, while it is nice to throw statistics at analogous historical situations, it is important to consider that all it takes is one rogue GDP number to throw a monkey wrench into an attractive set of statistics. Tomorrow should be interesting – and I expect the bulls will be putting a great deal of capital to work no matter how the GDP data falls.

For posts on related subjects, readers are encouraged to check out:

• VIX Jumps 10% on Consecutive Days
• Long the SPX on New VIX Record Closes
• Lessons from the Post-2/27 VIX Price Action
  

Disclosure: Short VIX at time of writing.

The Significance of Double Tops in the VIX

The following is adapted from a subscriber newsletter segment that appeared in the November 26^th newsletter.

I have never seen any research on the subject of VIX double tops, but given the recent double top formation in the VIX, I thought this might be a good time to share some of my thinking on the subject. First, the chart at the right shows the recent VIX levels from mid-August to the present. The first spike in the double top comes on October 24th and the most recent spike comes on November 19th.

Double tops are fairly common in the 19 year history of VIX data and frequently coincide with extreme readings in the VIX. By contrast, VIX triple tops are relatively rare; and while single VIX spikes are common in more mundane market conditions, they are less likely to be found at VIX extremes than double tops.

In fact, prior to this year, the three crises with the most extreme VIX readings were the 1997 Asian Financial Crisis, the 1998 Long-Term Capital Management Crisis, and the 2002 bottom of the technology bust that accompanied the WorldCom bankruptcy filing. In the graphs below, I have recorded the history of these VIX spikes. You can clearly see a double top pattern in the VIX in all three instances. Interestingly in each instance, the spikes were separated by approximately 2-3 weeks and signaled major turning points in the markets.

Recent events have shown that extrapolating from past chart patterns to the current market is fraught with danger, but I would argue that the presence of a VIX double top reinforces the case that the recent stock market bottom will prove to be an important market inflection point.

[source: Yahoo, VIX and More]

Long the SPX on New VIX Record Closes

In light of yesterday’s record close in the VIX, a reader asked about the historical track record if one were to buy the market when the VIX made a new high.

First, let’s detail the history of new end of day highs in the VIX. Working backward, the five most recent new highs came during a span of 8/27/98 to 10/8/98 as the Russian financial crisis morphed into the Long-Term Capital Management crisis. Prior to 1998, the most recent VIX high was from 10/30/97, during the Asian financial crisis.

Looking at data from the CBOE’s historical database, which goes back to the beginning of 1990, the next three VIX highs are from August 1990. Scrolling back before the August highs, there are 11 highs in January 1990, including eight of the first ten trading days. In the chart below, I have omitted the noise from the first ten trading days and included only the last three VIX highs from the initial month of VIX data. While this leaves only ten data points, the pattern is clear: going long the SPX when the VIX makes a new high is an effective strategy, at least when the holding period is from 1 to 100 days. The mean return for the SPX following a VIX high, for instance, is about 2% ten days after the VIX high, while the mean return during all ten day periods for the SPX during the past 19 years has been only 0.3%.

For some additional mean reversion data that may be relevant to yesterday’s 34.5% VIX spike, check out a previous post, One Day 30% (!) Spikes in the VIX.

VWSI at -10

With the VIX Weekly Sentiment Indicator (VWSI) ending the week at -10 for the first time since 9/11, the prognosis for the VIX over the next month or so is as bearish as it has ever been.

The current extreme readings in the VWSI have only been approached on three previous weeks since 1990:

• Following 9/11
• In the wake of the Asian Financial Crisis and a 554 point decline in the Dow on 10/27/97
• On the heels of a second Fed rate hike in consecutive months in March 1994, after a period of five years without any Fed rate hikes

In each of the three instances above, the broader markets were characterized by considerable turbulence and uncertainty for at least a year following the crisis.

As far as the VIX is concerned, if the history above is any guide, expect a sharp reversion to the mean. The three previous -10 VWSI readings resulted in the following changes in the VIX:

• 3 days: mean of -23% (-17%, -24%, -27%)
• 5 days: mean of -32% (-27%, -33%, -36%)
• 10 days: mean of -33% (-18%, -38%, -42%)
• 20 days: mean of -43% (-37%, -46%, -47%)

As is the case with most VIX mean-reversion plays, most of the gains in these instances were limited to the first 20 trading days.

Keeping in mind the history above, there are many possible investment approaches if one expects history to repeat itself. Being short volatility or short the VIX should be a central portion of that strategy. Neutral calendar spreads are a relatively conservative approach; put back spreads would be more appropriate for an aggressive investor. Those wishing to strictly limit risk should probably also be looking at iron butterfly and iron condor strategies.

Finally, I would be remiss if I didn’t add that in periods of elevated volatility even more so than in more ‘normal’ markets, one should always plan exits before placing any trade and use stops wisely. Better yet, if you are not used to trading options, this is not the time to start experimenting.

(Note that in the above temperature gauge, the "bullish" and "bearish" labels apply to the VIX, not to the broader markets, which are usually negatively correlated with the VIX.)

The SPX:VIX Relationship

After giving it fairly prominent play over the past few months at Technically Speaking, Ron Sen asks whether the SPX:VIX ratio chart works effectively as a risk metric.

Many people seem to be giving up on the VIX lately, but the SPX:VIX ratio is a special case. Part of the problem with this ratio is that it compares one trending number with another one that oscillates. Using Ibbotson data as a guide, we can assume -- over the long term at least -- that the SPX should return about 10% per year, while the VIX should oscillate around a stable mean. Interestingly, if you add a 10% trend line to the SPX:VIX ratio chart going back to 1991 in order to compensate for this, you will discover that 15 ½ years later, the trend line almost perfectly bisects the current Bollinger bands.

If you study the chart for a little longer, some interesting conclusions emerge:

• except for the latter half of the dot com crash and the Asian financial crisis, the SPX:VIX ratio has rarely strayed far from the values predicted by the 10% trend line
• since 2004, the SPX:VIX ratio has hugged the 10% trend line very closely
• the current deviation above the trend line is matched only by mid-2000, early 1994 and late 1995 – three periods that turned out to be the beginning of a deep bear, a mild bear and a strong bull market

I think the SPX:VIX ratio is indeed a useful risk metric, but I recommend using it in a manner that compensates for the long-term bullish bias in stocks and/or that focuses largely on the relative peaks and valleys.

At the moment, I think the SPX:VIX ratio is flashing a mild warning sign, but ultimately where you come down on the ratio is probably more dependent upon what you think about reasons for the historically low VIX than the historically high SPX.

The VIX and 3% SPX Drops

What happens to the VIX when the SPX drops 3% in one day? Does the VIX provide any advance warning of a selloff or merely serve as a contrary indicator after the fact?

Since 1990, there have been 25 instances in which the SPX has fallen 3% or more in one day. In analyzing those movements, it is important to remember that volatility has a strong tendency to cluster; as a result, a majority of the data points I looked at happened to fall within the same month and many were during the same week as another 3% drop. This should not be surprising, as almost all of the SPX plunges from the last decade have been triggered by the Asian financial crisis (including the spillover into Russia and Brazil) and the unwinding of the dot com bubble.

The following graphic depicts a composite view of changes in the closing price of the VIX on the 25 occasions from 1990 to the present in which the SPX has dropped 3%:


Note that the scales fixes the close on the day before the 3% drop at 100, so that percentage moves to and from that point are easier to calculate. It is worth highlighting that while the actual values of the VIX are not included here, only twice during this period did the SPX drop 3% with VIX values already below 20. In fact, the median value for the VIX prior to the SPX drop is 31.21, again partly reflecting the clustering of volatility that results in multiple sharp drops in the SPX.

On average, a drop in the SPX of 3% or more translates to a 14.3% spike in the VIX. More than half of the time, the day of the SPX 3% drop turned out to be the high in the VIX over this 11 day period. On average, half of that spike is retraced within two trading days and over 70% of that move is retraced in five days.

In future posts, I will discuss in greater detail the issue of the VIX providing advance warning of a selloff in the broader market. Suffice it to say that while the graph above shows that the VIX does jump approximately 10% in the four days before a 3% SPX move, there is not the obvious telltale movement in the day or two before the selloff that might get the attention of the casual observer.