Three Fast Facts About Slow Markets

From Nicholas Colas of Convergex

Today we continue our recent exploration of the ongoing low volatility in US equity markets, but with a novel twist.  Instead of looking at the CBOE VIX Index, we look at the historical trends in actual S&P 500 price volatility.  Not only is the dataset longer here (starting in the 1950s rather than 1990), but it is also “Stickier” than the twitchy VIX. 

Three takeaways. 

  • First: actual price volatility is running at 48% of its long run average at the moment and has been stuck around these levels for 77 days.  The last recent periods of similar low-vol were in mid-2014 (75 days in duration) and late 2006/early 2007 (79 days).  The 1990s had 2 long runs of low volatility: 1995/1996 (254 days) and 1992 (179 days) but the 1980s had only one (80 days in 1985). 
  • Second: actual price volatility does correlate negatively to current returns (-44% from 1959 to today). 
  • Third: more volatility now/soon does not have any forecasting power on returns over the next year.

Consider the following facts:

  • On the day before the assassination of John F. Kennedy (November 21st, 1963), the S&P 500 fell 1.3%. It had not made a similar move in over a month, and fell a further 2.8% on November 22, when the US President was shot in Dallas.
  • In the 10 trading days before the surprise invasion of Kuwait by Iraq on August 2, 1990, the S&P 500 saw 2 declines over 1% (1.0% loss on July 20, and a 1.7% decline on July 23) and a 0.7% decline on July 27th. Once markets saw Iraqi tanks in Kuwait City, the S&P 500 fell over 5% between August 2 and August 6.
  • In the three trading days before the 9/11 terror attacks in 2001, the S&P 500 fell 2.2% (September 6th) and 1.9% (September 7th) before a slight bounceback (+0.6% on September 10). When markets reopened on September 17th, the S&P fell 4.9% that day and another +5% over the subsequent 3 days.

Is it the Illuminati trading their book with advance knowledge of geopolitical events?  Probably not…

The reality is that 1% moves in the S&P 500 are not that uncommon.  In these notes we’ve chronicled that in the typical year the index has 54 days where its prices moves by 1% or more.  Since there are about 240 trading days in a year, that’s better than a one-in-five chance (22.5%, to be precise) that there will be a 1% move in the week before a large and unforeseen event.

US equity markets at the moment are distinctly strange, given that track record.  There have only been 3 days in 2017 with +1% moves: March 1st (+1.4%), March 21st (-1.2%), and March 24th (+1.1%).  We are 87 trading days into the year, so we should have seen 19 days with a +1% move.  Instead, we have three.

We got to wondering – what is the historical track record for actual volatility in the S&P 500?  Most commentary (including 2 Convergex notes last week) center on the CBOE VIX Index as a measure of volatility.  That, of course, is options-based expected volatility.  What about the real thing – the price levels we see and trade?

We pulled the daily price record back to 1957 for the S&P 500 and here is what we found:

Point #1: Current actual price volatility is running at 48% of its historical average.  It’s not your imagination.  US equity markets are much calmer than usual.

The math behind this analysis is straightforward.  We took the daily price returns for the S&P 500 back to January 1 1958 (1957 was a partial year) and ran 90 day historical standard deviations on those returns.  For the detail oriented reader: the average daily return on the S&P 500 is +0.032%, and the standard deviation of that return is 0.883%.

Looking at the time series back to 1958 you see that volatility moves in cycles, with long periods of below average price movements and occasional spikes of increased market churn.  We are, of course, in one of those low periods for market volatility which started on January 19th and has run for 76 days since.  Other periods include:

  • Mid 2014 (June to October), with 77 days of abnormally low volatility.
  • Late 2006 (November) to early 2007 (March), for a total of 79 straight days.
  • Early 2006 (March to May), with 55 straight days.
  • The 1990s had 2 long runs of low volatility: 1995/1996 (254 days) and 1992 (179 days) but the 1980s had only one (80 days in 1985).
  • The 1970s had long stretches with low actual S&P 500 volatility, such as in 1977 (178 days).
  • A statistical word of caution: the S&P 500 was below 100 in 1977, however, meaning that a one point move would be 1%. Now, that one point is only 0.04% change to the index.  Yes, there was plenty of volatility in the early 1970s, so smaller numbers don’t always dampen measured volatility.  But it is a point to remember when looking at long time series.

The upshot is that the current period of ultra-low volatility may be coming to an end, if history is any guide.  At 76 days of low actual volatility, we are right up against the 2006/2007 and 2014 experiences.  No, volatility didn’t jackknife immediately higher either of those times, but it did increase notably over the next year.

Point #2: The correlation between trailing 250 day (an approximate trading year) returns and concurrent price volatility is negative 43%.  That’s a pretty tight relationship for a one-variable model over 5 decades.  Essentially, if you think volatility is moving materially higher then you need to worry about equity price levels.

Point #3: At the same time, rising volatility does not say anything about future returns for the S&P 500 (correlation of 1.1%).  Put another way: if I told you actual price volatility was going to double by the end of 2017, you would not be able to estimate how the S&P 500 was going to in 2018.

One aside: as with all our statistical work, if you want a copy of our spreadsheet just shoot us an email.  After all, who doesn’t want the daily closing prices for the S&P 500 back to the 1950s?

The upshot of this quick analysis is that actual US equity market volatility should begin to rise (at least modesty) in the coming weeks.  If you buy our construct that volatility follows cycles, then we should be at the trough of one right now given the math we’ve presented here.  That should cause a pullback in stocks, but it does not portend a subsequent meltdown (or melt-up, for that matter)