Blog - Investing Notes
July 17, 2005 – Reader Inputs: Regulation FD/Zacks and an ISEE Tactic
Readers have recently emailed a questions/comments regarding: the impact of Regulation FD on the Zacks system, and a specific tactic using the ISEE sentiment indicator. Without any illusions of gurudom, we post below the questions and responses in case other readers may be interested.
We are always open to questions, suggestions, challenges and corrections. Feel free to contact us at any time.
Regulation FD vs. the Zacks System - More on the ISEE Sentiment Indicator
 |
Charles in Summit NJ observes and suggests:
Your article on July 12, 2005 - Regulation FD is Working? confirms what I expected. However, what I cannot figure out is how, with the exception of 2001, Zacks Rank has continued to work so well. I just reviewed its performance on Zacks Advisor, and it seems to still be working well. Perhaps it is the earnings surprise component? Might be a topic for one of your studies.
CXOAG:
The effects of Regulation FD (Fair Disclosure) on the Zacks system should be small because this system uses relative, rather than absolute, earnings data. A surprise is still a surprise when analysts revise estimates or actual earnings exceed the consensus (average) estimate. Maybe revisions and surprises are even bigger on average since Regulation FD.
Also, in his 2003 book Ahead of the Market, Chapter 12, Mitch Zacks talks about earnings uncertainty (defining it as the standard deviation of the spread of analyst earnings estimates for a given company and quarter). He asserts that "stocks that exhibit a low degree of earnings uncertainty tend to outperform those stocks that exhibit a high degree of earnings uncertainty." Even if Regulation FD has made the standard deviations bigger, there are still some that are lower than others.
Verifying independently how well the Zacks system works would be very difficult, almost as hard as keeping track of all of Jim Cramer's Mad Money gyrations. The pure Zacks system involves a lot of stocks and a lot of churn to capture the value of new earnings information (both estimates and actuals) before the market fully reflects it. We have done a review of the general market commentary of Ben Zacks.
|
More on the ISEE Sentiment Indicator
Rich in Chicago IL (responding to our follow-up analysis of the ISEE sentiment indicator):
I look at the ISEE 10-day moving average and ignore the daily readings. When the 10-day average gets above 200, it gets my attention as a signal that retail investors are becoming unusually bullish. There have been only a handful of occurrences, and I suspect drawing any conclusions would only be torturing the data.
Finally, I'm not sure if the site has mentioned Nassim Taleb's Fooled by Randomness, but I would recommend it to anyone who is a regular reader of your blog. I got turned on to the book after reading this New Yorker piece by Malcolm Gladwell. While I'm at it, I'll throw out Kurt Eichenwald's Conspiracy of Fools on Enron.
[Coincidentally, the same day Adam Sussman in New York NY asked: "Have you ever read Fooled by Randomness by Nassim Taleb? It is my favorite book on trading... If not, check out Nassim Taleb's Home Page."]
CXOAG:
Although we are somewhat familiar with Nassim Taleb's work and have visited his web page, we have not read his book. It is now on our reading list and we will report after reading it.
Regarding Rich's approach to using the ISEE indicator, we have done some further analysis. The following graph shows the behavior of the ISEE 10-day moving average using all available data. There are 75 days with ISEE over 200, but these data involve only five distinct crossovers: 9/8/03, 11/28/03, 12/9/03, 1/20/04 and 11/29/04.
The next chart summarizes the returns for the "ISEE>200" tactic for the Russell 2000 index over the next 5, 10, 21 and 84 trading days. The "Over 200" sample uses all 75 data points. The "Crossover 200 Only" uses just the five crossover points (a more practical trading signal). There may be some value in "ISEE>200" as a warning to get out of the market for a little while.
However, both the "Over 200" and "Crossover 200 Only" results have very low statistical confidence because of small sample size or overlapping test intervals (a problem with all rare signals). Also, the results here are not consistent with those of the more frequent sampling approach used to calculate S&P 500 index returns in the prior analysis of the ISEE sentiment indicator. It may be that investor behavior is qualitatively different at extreme values of ISEE. For the short-term results, standard deviations are larger than the differences in average returns. The samples are so small that the standard deviations do not have much meaning.
We also reformulated total exchange CBOE put/call ratio data (by taking the inverse and multiplying by 100) for the same calendar period to construct a sentiment indicator similar to ISEE. The following graph shows the behavior of this indicator. A "CBOE>140" tactic generates 74 days with 12 distinct crossovers: 11/25/02, 1/7/03, 3/25/03, 6/19/03, 7/21/03, 8/21/03, 10/20/03, 11/10/03, 12/30/03, 2/13/04, 3/8/04 and 11/16/04.
.
The next chart summarizes the returns for the "CBOE>140" tactic for the Russell 2000 index over the next 5, 10, 21 and 84 trading days. The "Over 200" sample uses all 74 data points. The "Crossover 200 Only" uses just the 12 crossover points (again, a more likely trading signal). "CBOE>140" does not give much of a signal.
Again, both the "Over 140" and "Crossover 140 Only" results have very low statistical confidence because of small sample size or overlapping test intervals. For all results, standard deviations are large compared to the differences in average returns.
