Sunday, January 27, 2008

Does A Disjointed Nasdaq / Russell Relationship Mean Anything?

Below are the returns for the Nasdaq 100 and Russell 2000 for the last 7 days:


The correlation between the Nasdaq 100 and the Russell 2000 in 2007 was about 0.65. Over the last seven days the correlation is -0.14. While it’s a mildly unusual for the correlation of the two to become negative over seven day time spans, it’s not terribly notable. What is notable is something my father-in-law, a very smart and observant trader who is always full of ideas, pointed out to me yesterday. The percent change spread between the two indices has been greater than 1% each of the last seven days. Even days when they are going in the same direction, the disparity between the returns has been large. It’s not because one is clearly outperforming the other. They’ve taken turns having the best/worst performance.

I decided to run some tests to see if I could glean anything from this.

My Russell 2000 data went back to the fall of 1987, so that is how far back I ran the tests. Over the last 20 years I found 10 other instances of 1% or greater divergences of seven days or longer. The most recent occurred in 2001. I then looked to see how the Nasdaq performed over the 20 trading days following an event. The 10 trades are listed in the table below.




With the exception of the 1996 trade, the remaining instances all occurred between 1999 and 2001. The first thing I noticed was that while half the trades occurred during the recent bear market, 9 of the 10 instances led to positive returns over the next 20 days. This would indicate a decent chance of a bullish bias. This is not what really caught my eye though.

What most stood out to me was the size of the returns and the range ((high of 20 days-low of 20 days)/close of day 1) the Nasdaq 100 traded in over the subsequent 20 days. The average winning trade gained 12.8% in the next 20 days. The lone loser lost 18.1%. The average range for the 10 instances was 24.25%, with the smallest range being 14.9% and the largest a whopping 36.5%. Truly eye-catching numbers.

While it’s difficult to draw any concrete conclusions from this study, it appears to me that when major indices diverge this much, something is possibly disjointed in the market. This disjointedness has often corrected itself and led to a multi-week rally. In all cases though, it seems to indicate the market is volatile and is likely to remain that way for a while. This agrees with several of the other studies I have discussed recently. Recall my summary in CBI, Reversal Studies and Bear Market Rallies…

Rob

P.S. Per request, I’ve posted a CBI reading on the top right hand corner of the page. I will update there each day until the reading drops back to 3 or below. I will also update intraday should a significant move be apparent.

6 comments:

Jeff said...

Good stuff. I considering opening a basket of shorts tomorrow, and this gives me pause.

Anonymous said...

This reminds me of Gary Smith's book "How I trade for a Living" as he would often look for divergences in the indices for tells as to which way market may go...though somewhat differently than your method here.

Alex said...

Great analysis. I'm wondering - did it matter which index outperformed the other by 1%? How did the Russell do after each disjointed week? I'm wondering if the index that underperformed (outperfromed) subsequently outperformed (underperformed) the other....

Alex

Rob Hanna said...

Thanks for the thoughts guys.

To address a few questions, the Russell had somewhat similar performance. 8 of 10 instances were bullish. Volatility was also very high but generally not as high as the Nasdaq during the event (7 days) or the subsequent 20 - day period. I was unable to find a solid edge in comparing one to the other that could be taken advantage of as a spread trade.

Rob

Joshua Ulrich said...

A couple thoughts:

1) Since the Nasdaq 100 is large-cap tech, while the Russell 2000 is broad small-cap, do you have any theories why this relationship exists; or is it simply a result of data mining (not to be confused with data dredging)?

2) Which historical instances you list are similar to the current situation, i.e. one index isn't clearly outperforming the other?

Thanks for all your analysis; I've found them interesting and thought-provoking.

Best,
Josh

Rob Hanna said...

Josh -

1) I don't have a good explanation for the results. Many times when I run a test, I can anticipate the results somewhat because I have a theory going in. In this case I simply found the observation to be odd and then looked to see what had happened given like events in the past. I was surprised by the lopsided results. Normally if I see a 9.7% expected return and 90% win rate I'm all over the long side. In this case I wasn't quite sure what to make of it but felt the results were worth sharing.

2)Very few had strong outperformance during the event by one index or the other. Exceptions were 12/20/00, 8/1/00, and 4/19/99.