 Feature
A Look at
Stock Data
By
Van
K. Tharp Ph.D.
I�ve
just spent two days with Bob Spear, the developer of the software Trading Recipes which I used in the 1980s and early 1990s. I really
liked Trading Recipes for
a couple of reasons: 1)
the programming was fairly simple; and 2) it allowed many forms of
position sizing. IITM
used to sell it, but we stopped in the mid 1990s because it was a
DOS based program in a Windows world and it looked like it would be
a long, long time before a Windows' version would appear.
So, I was happy when I recently discovered that Bob now has a
Windows' version called Mechanica
that does many of the things I was hoping for in terms of position
sizing.
Anyway, I went to
visit Bob with the idea that I might be able to use Mechanica in the
monthly portfolio research and perhaps even do some testing with it.
My goal is to develop rules for a long portfolio based upon
the efficiency screen and using 25% trailing stop and for a short
portfolio, using a breakdown of an efficient stock.
Bob has agreed to help me with both tasks and I'll be keeping
you up to date on how that progresses.
That�s the good
news. The bad news is
that our initial research won�t be done in time for this
newsletter. As a result,
I thought I would touch on some of the data problems that we are
having with stock data.
Data
Can be a Real Problem
Most
people think that if you buy the right stock, you can hold it
forever and make a fortune. But
that is not true. Just
to illustrate that let�s look at the most important US stocks that
make up the Dow Jones Industrial Average.
The average was developed by Charles Dow in 1896.
At the time, it only consisted of 12 stocks.
In 1916, the average was changed to 20 and in 1928, during
the roaring 20s bull market, it was changed to 28.
But let�s take a look at the original 12 and see what�s
happened to them in a little over 100 years.
Could you have bought any of them and just held them for 100
years? For many, you�d
have done fairly well. But
some of them have disappeared. Table
1 shows the 12 original Dow stocks and what�s become of them over
the last 111 years.
Notice
that North American, U.S. Leather, and American Tobacco are totally
dead. Thus, an investment in 3 of the original 12 stocks in
the Dow Jones would have disappeared on a buy and hold decision.
Six of the
12 were absorbed by other companies and three (General Electric,
National Lead, and Laclede) are still around in some form.
This simply illustrates how many changes occur to major
companies.
|
Table 1: The 12 Original Dow Stocks
|
|
Original Dow Stock
|
What�s Happened in 111 Years
|
|
General Electric
|
Still part of the DOW 30, although it was
taken out and then replaced.
|
|
American Cotton Oil Company
|
Distant cousin to Best Foods, which was
bought out by Unilever. Traces
also remain in CPC International.
|
|
American Sugar Company
|
Now part of Amstar Holdings.
|
|
Chicago Gas Company
|
Absorbed by People�s Gas and now part
of Integrys Energy Group.
|
|
Distilling & Cattle Feeding Company
|
Now part of Lyondell Chemical Company (I
was also told it was part of Quantum Chemical).
|
|
National Lead Company
|
Now NL Industries.
|
|
North American (Group of utilities)
|
Broken up in the 1940s.
|
|
Tennessee Coal Iron and Railroad Company
|
U.S. Steel
|
|
U.S. Leather Company
|
Dissolved in 1952.
|
|
U.S. Rubber Company
|
Became Uniroyal, which merged with
Goodrich and was bought out by Michelin.
|
|
American Tobacco
|
Killed by antitrust legislation in 1911.
|
|
Laclede Gas Light Company
|
The Laclede Group
|
The
big averages are simply like mutual funds.
The bad ones get dropped out and new, powerful stocks are
added. For example, in
1999 Chevron, Goodyear, Sears, and Union Carbide were dropped from
the DOW and replaced by Microsoft, Intel, Home Depot, and SBC
Communications. In 2004,
International Paper, ATT, and Eastman Kodak were removed and Pfizer,
Verizon and AIG were added. And
then, ironically, in 2005, SBC communications absorbed ATT and
changed its name to ATT, so ATT is still part of today�s DOW, but
it is different.
Between
1999 and 2004, several stocks in the index merged and/or changed
names: Exxon became Exxon-Mobil after their merger; Allied-Signal
merged with Honeywell and kept the Honeywell name; JP Morgan became
JP Morgan Chase after their merger; Minnesota Mining and
Manufacturing officially became 3M Corp; and Philip Morris renamed
itself Altria. In some
cases the company changes dramatically (e.g., Exxon-Mobil) and in
some cases the name just changes (e.g., Atria).
But all of this can wreak havoc on anyone trying to do
historical testing.
Bob and I were
originally going to do our study of efficiency stocks on the S&P
500 data. However, the
S&P 500 changes several times each year and there can be as many
as 50 additions/deletions each year.
For example, Microsoft was added to the S&P 500 in 2000.
Our database, which included S&P 500 stocks from 1985
through 2005, included Microsoft from its inception as a listed
company in 1986. Thus,
we could say we were only trading the S&P 500 as efficiency
stocks, but we weren�t. Instead,
we were trading the 2005 version of the S&P 500.
Thus, we might catch this great stock like Microsoft in the
late 1980s and make a fortune with it.
The only problem is that Microsoft was not a part of the
S&P 500 until 2000. And
many of the stocks that were part of the S&P 500 are no longer
part of the database.
This
is called the survivorship bias in data.
Only those stocks that survive are contained in the data.
Those that go bankrupt or are absorbed by other companies are
not included. There are
databases that include the data of companies that disappear up to
the time they disappear, but those databases cost a small fortune.
My understanding is that a database of fundamental stock data
(with no survivorship bias) can cost you well over $100,000.
Even
our portfolio from July 2006 through August 2007, which included 44
positions, has 3 stocks that disappeared.
Those were FAL, AETH, and MRCY.
The survivorship bias with stock data is huge, which is
another reason that most people can only test the way I did with the
model portfolio.
And
look what happens when a company gets absorbed by another company.
Its chart looks like the one below of Myogen.
It showed up as a very efficient stock because of the huge data
gap and then the very quiet steady movement after the gap.
This happened because it was announced that Myogen was to be
acquired by another company. As
a result, the price gapped to near the acquisition price and then
stayed there until it was acquired.
It stayed at the acquisition price until it disappeared and
it also looked very efficient. But
that�s not what I want in an efficient stock.
Other Data Problems
Other
gaps sometimes occur because 1) stock splits are not accounted for
in the data or 2) perhaps dividends are not adjusted for in the
stock data. All of that
also wreaks havoc on any sort of historical testing, which was one
reason that I chose to test the portfolio as I did in Tharp�s
Thoughts.
There
are basically two types of stock data.
One type is acceptable for screening.
If you want to screen all listed stocks for those making new
highs or those above their 200-day moving averages, then it�s
possible to get data that will do that.
However, many of these data sets are replete with errors.
They are fine for regular screening, but they are useless for
historical testing.
If
you want to do historical testing, you must get data that has
adjustments built in for dividends and for stock splits.
And even when you find this data, most of it still has the
major problem that I mentioned earlier with the survivorship bias.
Another
problem you face is huge gaps in stock data that appear
historically. For
example, I found the following gap in Alnylam Pharmaceuticals (ALNY)
that occurred in mid July. Chances
are there was a major news announcement that produced it, but I
couldn�t find the source of what was happening easily by looking
at the companies news � and that was only a few months ago.
What if it happened 15 years ago?
These
are just a few of the data problems that one will have trying to
test stock market
systems. And we
haven�t even touched upon programming software issues, problems
trying to program what you want (e.g., stocks going up in a smooth
line) into the software, etc. So
most of the time, testing by making trades the way I did in the
model portfolio is the only way to test.
[Editors
note: To review past model portfolio updates go to our newsletter
back issues list and scroll through the list. They are usually
titled Efficiency Portfolio Update or Market Efficiency Portfolio. Click
here to see back issues.]
About Van Tharp: Trading
coach, and author Dr. Van K. Tharp, is widely recognized for his
best-selling book Trade Your Way to Financial Fre-edom and
his outstanding Peak Performance Home Study program - a highly
regarded classic that is suitable for all levels of traders and
investors. You can learn more about Van Tharp at www.iitm.com.
|
