Disappearing Daytime vs Overnight Volatility (and Why That Matters)
The chart below shows the 1-year standard deviation of changes in SPY’s daytime (red) versus overnight (grey) session, since 1994.
The daytime session is of course from SPY’s open to the close, and the overnight from the close to the open.
Note that for nearly all of the last 19+ years, the daytime session has been the more volatile of the two by a wide margin. But not any more.
The next chart makes this clearer. Here I’ve divided the std. dev. of changes in the daytime session by the std. dev. of the overnight. Values greater than 1 indicate the daytime session is more volatile.
Note how daytime versus overnight volatility has been steadily collapsing to 1.0 over the last few years. The two are now more or less equally volatile (1).
The last time that ratio came even close 1.0 was the crash of 1987 (not shown), and that was an anomaly – the market gapped down so far on one single day that it skewed the stats.
Put another way, what we’re seeing today is extremely abnormal in modern market history.
Why That Matters
I’m sure that I’m not the first to make this observation.
It only caught my attention because I’ve been doing research lately related to trading the overnight market, inspired by Rob Hanna’s excellent new OvernightEdges (2).
I’ve found that a lot of the obvious statistical edges in the overnight market (i.e. the low hanging fruit) have suffered in the last few years, and I wonder if this change in relative volatility is the footprint of the market closing those inefficiencies.
In other words, the off-the-cuff conclusion that all of this is the result of Europe and Asia having a greater impact on the US market, may not tell the whole story.
Further, I have to assume that like all shifts in the fundamental mechanics of the market, there will be other impacts on quantitative traders beyond what is immediately obvious. Other indicators may change (even those using, say, closing data only), and other strategies may need to adapt.
Food for thought…
. . . . .
(1) This is all a little oversimplified because it doesn’t take into account intra-session moves (ex. highs and lows of the session), but that’s a post for another day.
(2) This post is not at all an indictment of Rob’s work with overnight trading as he and I take very different approaches to wrangling the markets. In other words, the impact of this observation on my own research is not necessarily relevant to his.
Filed under: Stock Market Mechanics, VIX & Volatility | 4 Comments