Friday, October 27, 2006

When Markets Don't Correct, Are We Due For A Correction?

The most recent post on evidence-based trading emphasized the importance of our knowing what we know.

This was illustrated for me quite recently when I read that we were due for a meaningful correction because we hadn't had one for a while.

Let's see if that reasoning is valid.

It turns out that we haven't seen a daily 1% or greater decline in the cash S&P 500 Index ($SPX) in over 60 trading sessions. Since 1990 (N = 4183 trading days), this has only occurred on 177 occasions. When we look at what happens 60 trading days later, it turns out that the S&P 500 Index is up by an average of 4.61% (153 up, 24 down), quite a bullish edge compared to the average 60-day change of 2.21% (2845 up, 1338 down). Over essentially every time frame leading up to 60 days, moreover, we see above average returns following from periods in which we haven't had a 1% daily correction.

In short, the absence of a drop does not make a fall more likely. Indeed, it has led to superior returns over the intermediate term.

Why is this? The majority of periods when we haven't had a large drop are periods in which we haven't had large moves of *any* kind. These have been low volatility periods in the market, from 1993-1995. Recall from the recent post that low volatility periods have actually shown superior returns over the intermediate term.

Are we overdue for a drop? Yes. We don't typically have runs of no 1% declines for months at a time. Does that mean the market is headed lower? Not at all. Only six of the 177 periods in which we have not had a 1% drop led to a decline of 2% or more in the following 60 trading sessions.

It helps to know what we know: that's the appeal of an evidence-based approach. Tomorrow we'll explore the limitations of the approach.

2 comments:

Paolo Pezzutti said...

Your approach to testing is of great interest. Most of your posts are based on a sound statistical basis, which I like very much. What software do you use build your statistics and, if you know, what packages are on the market that make this approach easy?

Brett Steenbarger, Ph.D. said...

Hi Paolo,

Thanks for the note and the interest. Actually, most of my posts are simply identifying patterns in market data. I do this with archived historical data in Excel. Some of my data sources are linked on the Trader Development page of my personal site. For more elaborate data mining and testing of significance (developing trading systems), I've used the CART tool from Salford Systems (www.salford-systems.com). But, overall, I find the database functions of Excel to be adequate for my needs.

Brett