Thursday, June 18, 2009

Negative Divergences: Not What You Think

As mentioned in a prior commentary this week, the weekly charts of many of the ETF's and indices that I follow are showing negative divergences between price and an oscillator that measures price momentum. I had done some research a long time ago on negative divergences, and the results are not what you think.

The article was written for in January, 2007 and you can access it (without subscription) by clicking on this link. I will summarize some of the results:

1) The study was done on weekly data and the NASDAQ Composite going back 35 years;

2) I used the same momentum oscillator I always use (value chart) and if anything, I define the parameters of the negative divergence by setting things like the look back period and closing price versus high price, etc., so this skews the results;

3) In 35 years, there were 62 unique negative divergence bars;

4) 55% of the time (0r 34 out of 62 times), the NASDAQ made a weekly close below the low of the negative divergence bar; of these 34 times, 19 were still lower than the low of the negative divergence bar 8 weeks later;

5) So we can say that the mere presence of a negative divergence bar led to a sustained downward movement 19 out of 62 times or about 30% of the time;

6) About one third of the time, there was a weekly close below the low of the negative divergence and then a reversal where prices closed above the high of the negative divergence bar;

So this covers the downside or weekly close below the low of the negative divergence bar. Now let's highlight some of the results when prices close above the highs of the negative divergence bar.

1) Closes above the high occurred 32 times (including reversals) out of 62 unique occurrences

2) In 20 instances, the NASDAQ remained higher 8 weeks later, and the average gain over that time period was 7%. As quoted from the article:

This instance shows that a negative divergence actually leads to higher prices, and those gains occur in a somewhat accelerated fashion, earning almost 1% per week of market exposure.... The underlying cause for this is probably related to short-covering. Traders generally bet against the market when prices are at extremes, which is when negative divergences are most likely to occur. Closes above those highs and negative divergences lead to acceleration in prices as traders cover their losing positions.

Lastly, I had this to say about market research, which holds true today as it did 2 years ago when I wrote it:

...this study should serve to illustrate that what we believe about the markets often isn't true; for instance, negative divergences can be both bullish and bearish. And it's only through a process of continually questioning the dogma that we can achieve a true understanding of the action.


Dave Narby said...

Thanks for your excellent blog.

From your RealMoney article: "But there were four negative divergences in the 2003 bull market (area 4), and none led to appreciable moves lower. "

That was during a bull market - Did you check this against a bear?

Guy M. Lerner said...

At some time in the past, I am sure I must have used some sort of filter or application of how divergences behave in certain market environments....

They probably behave as you expect; in a bull market negative divergences often are ignored or don't lead to lower prices and bear they do....because my time frames have gotten longer and longer over the past years, I tend to be more interested in the sum of a number of divergences occurring over time....see the last post on Treasury yields

thanks for the nice compliment too :}