A-D Data: Composite, or Operating Companies Only
In your A-D data do you use operating companies only, or the composite data which of course includes ETF, Bond funds, REITs, etc. I would like to see your data with only operating companies. Lowry's I know uses the clean version in their work
The question you pose concerns a very interesting issue within the science of technical analysis. The short answer to your question is that we do not have data exactly like Lowry's, but we do have several series of A-D data which we think are useful.
The challenge of doing as Lowry's has done, and culling the list of issues down to 1600-1700 "operating companies" is a rather Herculean task from a computing standpoint, and one must grapple with several problems in doing that. One of the biggest philosophical challenges lies in defining an "operating company".
For example, a lot of people would say that closed end mutual funds should not be counted in that list. If that is the case, should we also exclude GE? It is a widely diversified company consisting of several different operating subsidiaries, and its trading is much more like that of a financial company than a manufacturer. What about Danaher? It has 74 (last we checked) operating subsidiaries, which is more than some mutual funds have in their portfolios. There are other companies like this, we do not know how Lowry's has approached issues like these.
Most of our work is done with the composite NYSE A-D and Volume numbers which are commonly reported, and more easily available. We have taken the position that the numbers reported in Barron's (same as WSJ) are the official data for use in our analysis. They tend to be cleaner than other sources. We may use other sources' data intraday, but we update all of our files with the final numbers after trading is over. Not many people know that A-D numbers differ quite a bit from one source to the next, because these data do not originate with the NYSE. Any data provider who publishes values for the number of advances and declines must either calculate these values themselves, or retransmit calculations made by other agencies. To do that, one must have an accurate list of which stocks are traded, and what their last and current prices are, plus make adjustments for splits. It is a cumbersome job.
We also work with composite Nasdaq A-D and Volume data. We track AMEX data, but it is such a strange little exchange that is now heavily populated by ETFs, mining and oil companies, and other unique situations, that drawing conclusions from the data about market direction is a bit iffy.
We have created our own database to track A-D and other information about the 30 stocks in the DJIA, as well as the 100 stocks in the Nasdaq 100 Index (NDX). The latter is in a 45MB spreadsheet file which tracks data back to 1988, so you can see that it is quite a big computing effort to magnify that to 1600-1700 issues.
We also track the "Common Only" data provided by Barron's, and this data is actually generated by the NYSE. It used to be reported a day later on the NYSE's web site, but is not any longer. But this is not a complete filtration of the data.
Prior to February 9, 2005, the "Common Only" numbers generated by the NYSE filtered out any symbols greater than 3 letters long, which catches all of the warrants, rights, and preferred stocks. That filter did not exclude closed end funds and country funds. Commencing on February 9, 2005, the NYSE's Common Only data that is published each week in Barron's was changed to include only those issues which make up the new NYSE Composite Index. That list non excludes closed end funds, and also limits component companies to only one class of stock. That way, companies which have more than one class of stock do not get to have multiple votes.
This common only data is generally weaker than the composite A-D data, because the "Uncommon" stocks (our name) have a fairly positive bias, at least over the years that we have been following this data. Those years, it should be noted, have had generally falling interest rates, which helps "stocks" that are tied to bond prices. In fact, when we subtract the Common Only A-D numbers from the composite, we get Uncommon A-D numbers, and those data make a pretty good tool for tracking corporate bond price movements.
In an attempt to measure how much "pollution" of the data is being caused by closed end funds, we created a spreadsheet using historical data to track the advancing and declining issues among the closed end funds that trade on the NYSE. That took some work, we can tell you. What we found, however, was that the closed end funds tend to have very little effect on the A-D numbers, partly because there are only 355 of them as identified by the NYSE (as of 2001). That's just over 10% of all the issues traded. We also found that they tend to move in sympathy with the other A-D numbers, going up when the market goes up and going down, in general, when the market goes down. From October 4, 1988 to June 20, 2003, the closed end funds we have tracked have added a total of only 9426 net advancing issues, or an average of 2.54 net advances per trading day. That is not a very big factor to have to worry about from a statistical standpoint.
One other reason why we continue to follow the composite A-D data is that is what is easily available during trading time in order for us to make decisions. I have tried to describe some of the difficulties that one must go through to try to manufacture "cleaner" data, but if we are going to make trading decision based on any piece of data, we need to have it available at decision time. Our focus, therefore, has been less on trying to find ways to purify the data that exists, and more on trying to find ways that we reliably use the data that we can get and use it to our advantage.
One final comment about A-D data: We have done some studies of A-D data for specific market sectors, e.g. gold stocks, high tech stocks, insurance companies, etc. What we found was that on any given day, these companies tend to move up or down pretty much together, so one does not see the types of divergences between prices and breadth numbers that one may see when looking at the whole market. The chief value of looking at A-D data for the overall market is to measure the health of the liquidity pool. If a few big cap stocks push the indices higher, but the rest of the market is not able to get enough oxygen to breathe, then that shows up in the A-D data. This is what we saw all during 1999, and you probably remember what came after that period for the price indices. But when liquidity is so plentiful that it can push up the majority of stocks, then that says that money supply for the stock market is tremendously healthy and likely to continue that way for a while. That is the whole reason we pay attention to these numbers.
On our web site, we publish a spreadsheet file that is updated each day, containing a limited amount of data for the NYSE A-D and Volume Numbers. That can be accessed at Daily Oscillator Data . We also sell a longer version of that file, with data back to 1990, for $30. The order form is on sheet B of that same spreadsheet file.
If you want even more data, then we have another file that takes the NYSE back to 1960, and the NASDAQ back to inception. That is in a larger file than we can email, so we have to send it on a CD, and the cost is $45. It too can be ordered using the form on Sheet B of the OSC-DATA file.