Thursday, January 19, 2006

Overnight Moves and Next Day Volatility

I recently submitted an article to Trading Markets that looks at the relationship between the size of the overnight moves in the major indices and the trading range the next trading session. It turns out that for such averages as the S&P 500 Index (SPY), the NASDAQ 100 Index (QQQQ), and the Dow Jones Industrial Average (DIA), the correlation between the size of their move from close to open and the subsequent high-low range is in the order of .33. When the overnight move is greater than .50%, the market has a trading range that exceeds 1% over three-quarters of the time. When the overnight move is less than .05%, the market trades in a range greater than 1% less than half of the time.

Interestingly, the Russell 2000 correlation is only about half that of the Dow. Overnight events may well impact large cap issues more than small caps. It is also interesting to look at correlations for individual stocks. Many stocks, such as MSFT (.30), AAPL (.27), and IBM (.23) have correlations similar to the indices. Others, such as XOM (.07) show little relationship between the magnitude of overnight moves and trading range the next day. This may be because XOM is impacted more by contemporaneous movement in the oil market than by overnight events.

Using overnight movement and current volume to predict near-term volatility has been extremely helpful in my trading, as volatility is a predictor of whether markets will break out of ranges vs. revert to mean trading prices. This appears to be a strategy that can work for individual equities as well as indices.


Paulo de León said...

Excelent comment, i will incorporate that observation into an equation that i´m working in trying to predict today´s range. I found some explanatory variables: a) yesterday range (persistance) b) day of the week (for example: wednesday statiscally has the highest trading range, monday and friday the narrowest) c) the volume in ES in the first 30 mins of trading......i´ll check Opening Gap Size.....

Brett Steenbarger, Ph.D. said...

Thanks for the comment. I follow 5 minute volume in real time to update my assessment of the day's likely range/volatility. I wasn't aware of the day-of-week effect you noticed. Perhaps there are seasonal effects as well as well as end of month/beginning of month effects and effects on days that have scheduled economic reports (Fed days, etc.)

bryan wendon said...

Very useful numbers. Great stuff.

One thing I wondered was whether you had also looked at the relationship between the size of the Globex range and the next day's action? Maybe there is an interesting relationship there too?

Thank you for sharing your hard won research findings

Brett Steenbarger, Ph.D. said...

Thanks for the note. Whew! What a great trading day we're having.

Anyway, I like your idea of looking at the Globex range specifically. Another idea is looking at that range, but weighting the most recent data more highly to update volatility estimates. That's really what I'm doing by following each 5 min volume during the day. It alerted nicely to the huge volatility during today's session.