Analyzing the Nasdaq Streak

When I saw this piece from Barry, and this piece from Jason Goepfert on the eight-week Nasdaq streak, and read some of the questions, I said, “Hey, maybe I can help.” After struggling with what defines a week (close of business for the week, Friday, or Thursday if the market was closed on Friday, or Monday in the second week of September 2001), I ran the numbers, and here is what I found:

Up Streaks

I used data from the Nasdaq Composite Index from inception in February 1971 through last Friday. (Dividends not included in performance.) Streaks longer than seven weeks are rare, but they tend to be associated with good performance in the next twelve weeks. Again, the momentum effect is showing its face. Interesting that intermediate length streaks of five and six weeks have done poorly over the next 12 weeks, whereas shorter streaks are just noise. The frequency of streaks seems to follow an exponential decay pattern that is essentially coin-flip random, decaying at a rate of around 50%.

Down Streaks

Hey, if I have the data, shouldn’t I do the other side even if it is not immediately relevant? The market was in a bull phase from 1971 until the present, so it doesn’t surprise me that after streaks downward that the market tends to rally, and after streaks upward the market meanders? But long down streaks tend to bounce back hard (few observations, be careful), while the results after middling streaks are weak, and short downward streaks are stronger. Again, there is exponential decay of streaks, near coin flip levels here as well. Not surprising.

What does this tell about the current eight-week streak upwards? With weak confidence it tells us that there is more room to move up. Perhaps the Nasdaq Composite could be over 1900 by the end of July. Given the lack of confidence in the rally, that is a genuine possibility.

Whether you run out and buy a bunch of QQQQs is your own business, but momentum tends to persist. I don’t plan on buying the QQQQ.

Comments

[Andrew Butter:]

So is that a decision then the.....cat ain't dead?

May 06 10:39 AM

[Oliver Nguyen:]

Intriguing analysis! Can you do the same for the Dow?

May 06 01:01 PM

[greensleeves:]

I noticed that there are in your data five 11 week streaks and only two 10 week streaks. This means that your methodology is flawed if the question you are trying to address is: What is the subsequent performance of the NASDAQ after a winning streak of n weeks. An 11 week streak INCLUDES a 10 wk streak, a 9 wk streak, etc. At the end of a 7 week streak, we MAY be in a streak that ultimately ends after 7, 8, 9, or more weeks. You have eliminated all of the seven week streaks that were followed by more up weeks, therefore skewing the results toward the negative. This makes all of the numbers other than the one 15 week streak incorrect, and particularly the shorter week streaks. Maybe this was a typo, but the numbers to me (negative returns following the 5-6 week streaks make sense if you have a lot of longer positive subsequent performance eliminated)

May 06 04:55 PM

[greensleeves:]

Same deal with the down streaks, except the error is in the opposite direction. To get the right result for a seven week up streak you need truncate all of the streaks more than seven weeks at the end of week seven, include them with the seven week streaks, and run the subsequent performance for all of them. You could go through this for all periods, but it sounds like work. At least you should test it for one of the mid length up streaks - because I think your numbers are very wrong and the conclusion is incorrect.

May 06 05:03 PM

[Fred Voetsch:]

Interesting perspective...

Someone please correct me if you have more official data to override this:

www.bullandbearwise.co...

...shows the NASDAQ 100 P/E at 62

May 07 02:23 PM

[ANTS:]

I think Greensleeves shows that working on "streak analysis" is hard work.

I have found that a dynamic valuation model is a better way to go. You can use any formula that appears appropriate but it must contain an earnings revision and market premium component. As a stock (or index) is going up,it is a combined reaction to earnings revisions (either disclosed or about to be disclosed) and changes in premium (or market sentiment).

For example analysts may be upgrading earnings, but the market is not reacting, and therefore the premium is high. At some point, the market accepts these revisions and the premium shrinks. The stock then goes for a run because the new valuation will have both higher earnings and lower premium. There are many market dynamics of this nature which are worth following.

It is the dynamic interaction between these two components, once identified and quantified, that leads to superior forecasting.

May 07 04:45 PM