PBOND 1.0 - Predictor of the Bond One Month Ahead - Version 1.0
Program Manual

Copyright 1994 by William L. Berggren
All rights reserved.

CONTENTS
Files Included
User Agreement
Warranty
Disclaimer
Summary of the Treasury Bond
Bond Model Predictors
Program Usage
Payment


*** Files Included ***

pbond10.exe -> Predictor of the Bond One Month Ahead 1.0 Executable
holt1.txt  -> Holt Stock Market Report, Present
holt2.txt  -> Holt Stock Market Report, 1 month old
holt3.txt  -> Holt Stock Market Report, 2 months old
pbond.dat  -> Economic Data File
pbond.bak  -> Economic Data File Backup
table1.dat -> MODEL 1 tables and results
table2.dat -> MODEL 2 tables and results
manual.txt -> This Manual

*** User Agreement ***

Read this text file first before using the program.  IN NO WAY CONSIDER
THIS INVESTMENT ADVICE. By using the software, you are bound by the
terms herein.  ONLY use the program in not for profit games and contests
between friends and office workers.  IN NO WAY WILL I BE RESPONSIBLE, OR
WILL YOU HOLD ME LIABLE, FOR ANY LOSS IN SAVINGS OR PROFIT which RESULT
FROM THE USE OF THIS PROGRAM. If you use the results as investment
advice, you are on your own. USE AT YOUR OWN RISK. The past may not be
predictive of the future. The software, software output, manual, and
results, are copyrighted and shall not be used without my written
permission. In no way shall you hold me responsible or liable for tort
or negligence. Furthermore, I have the right to refuse to respond to any
question about this program, installation, or its results. Generally, I
would be honored to answer any question.

*** Warranty ***

I warrant that the provided disk(s) are to be free of any defects and
that the programs are operational on a standard IBM PC with EGA/VGA
graphics capability for a reasonable period of time. All other liability
towards me is zero and zero dollars.  You agree I will not be liable for
more than zero dollars.  If there is a major error or bug in the program
I will fix it and issue a corrected version.

*** Disclaimer ***

I performed the program regressions truthfully with my limited knowledge
of statistics.  It is possible that I made errors in the numerous model
statistics, mathematics, program code, and file readers.  Long-term and
present accuracy of this model is suspect since more people these days
are using computer models.  Past returns may not equal that of the
future, and may be negative.

** Summary of the Treasury Bond ***

The 30-Year Treasury Bond (bond) is used by the Federal Government to
finance a portion of the federal debt as long-term debt.  You can
purchase bonds directly from the government, from a stock broker, or
indirectly through a mutual fund.  You can usually sell them at any time
with a commission like stocks.  The interest rate of the Treasury bond
is highly correlated with the mortgage rates, municipal bonds, and
corporate bonds because they all compete for long term capital.  No
State or local tax is paid on earned interest from T-Bonds.

The interest payments are fixed at the time of issue.  Interest
is paid at one half the issued coupon interest rate divided by two every
six months for 30 years.  The face value is returned at the end of 30
years.  $20,000 of bonds at one percent coupon will give you $100
interest every six months and $20,000 at the end of 30 years.

The 30-year bond is highly leveraged.  If the interest rates drop a
little, bonds will increase in value a lot; and vice versa.  This is
because of the long duration and fixed coupon rate.  If the yield on
bonds were to raise to 6 percent, the $20,000 of bonds above would be
worth a lot less than $20,000, probably around $4,000.

*** Bond Model Predictors ***

Two models were developed for this program and both are very SIMPLE
considering the potential use of over 150 data sets.  The models use
linear multi-variate regression of monthly government economic data
against the T-Bond at T+1.  Although the predictions seem based on
momentum, the variables used were some of the most accurate ones.

When interest rates were predicted to fall, bonds were bought, and
"Money" was calculated adding bond monthly interest to the capital gain,
determined by:

        Capital Gain ($) = Past Yield (%) times Past Price ($)
                           divided by Current Yield (%)

And, when rates were predicted to rise, bonds were out-of-the-market or
sold and the "money" accrued interest at the T-Bill discount rate.

The first model, MODEL #1, returned an average 18.3% annually from July
1974 through June 1994.  During this time $1000 increased to $28,834
through the power of compounding and predicting the movement in the Long
Term 30-Year Treasury Bond. The parameters used for MODEL 1 were the
Long-Term Treasury Bond yield, 91-Day Treasury Bill discount rate,
Average Prime Rate, and the Standard and Poors 500 Index. 588 data
points from July 1945 through June 1994 were used in the regression.
Average error squared was calculated at 0.043.  The R Squared was 0.994.

The second model, MODEL #2, returned 18.7% average annually from July
1974 through May 1994. $1000 increased to $30,836. The parameters used
for MODEL 2 were the Long-Term Treasury Bond, the Treasury Bill, the S&P
500 Index, the Japan Exchange Rate, the Japan Stock Market, Canada
Exchange Rate, and the Canada Stock Market. 276 data points from July
1971 through June 1994 were used in the regression.  Average error
square was calculated at 0.078 and the R squared value was 0.982.

If Treasury zero bonds or treasury strips were used, returns would be
over 25% per year and $1,000 would return $150,000 in twenty years.
Zero bonds pay no interest until maturity.  If you bought bonds at the
beginning, holding them would have returned $5,745 or 9.1 % average
annually over 20 years. Buying and holding Bills would have returned
$4,334 or 7.6 % average annually.


*** Program Requirements and Usage ***

The program requires the use of an IBM PC with EGA/VGA graphics card,
and should work with monochrome monitors and many laptops.  Some
extended ascii or ansi characters are used.  Both the automatic and
manual models produce identical results.  The required data is always
separated by one month which includes that of the present date (t), the
month previous data (t-1), and two month's previous data (t-2).

To RUN the program type:

"PBOND10"

at the prompt.  A menu will be placed on the screen with the following
choices described below:

1) The EXIT option exits the program.

2) The AUTOMATIC option runs the automatic version, if the Holt Files
are in the directory.  The Holt Report is an E-mail message which is
posted daily in the misc.invest newsgroup of the Internet.  Historic
issues are available via gopher.  Automatic E-mail service may also be
available.  You will eliminate the need to enter data manually using
this option.

You must copy three separate Holt files onto your computer, as either
Unix or DOS data files.  Then, rename them as follows:

holt1.txt        Present Date             e.g.  September 22, 1994
holt2.txt        Minus One Month          e.g.  August 22, 1994
holt3.txt        Minus Two Months         e.g.  July 22, 1994


3) The MANUAL option runs the manual version using "pbond.dat" as the
data file. You must keep the data file current manually.  Enter stock
market data each month into the data file "pbond.dat" with a text
editor.  This data is available from most daily newspapers.  Data fields
are described in the text data file.  For the Treasury Bill, remember it
is the "discount rate" to be entered and not "yield", they are different
values and will lead to errors.

4) The MAKE DATA FILE option creates the data file "pbond.bak".  If you
accidentally erase or mess up the data file, you can create another.
PBOMA will create "pbond.bak" which must be renamed to "pbond.dat" in
DOS before the program will use it.

5) The LIST MODEL1 RESULTS option displays the regression data tables
for MODEL1 on the screen.

6) The LIST MODEL2 RESULTS option displays the regression data tables
for MODEL2 on the screen.

7) The READ MANUAL option displays this manual on the screen.


*** Payment ***

This program is shareware, so please post the file on BBSs or give it to
friends.  The fee for use, past the 10 day trial period, is $30. Write
me on scratch paper the type of disk you want (3.5 or 5.25) and indicate
whether or not it should be high or low density (1.44MB to 360K).  Also
include your address.  You will be sent a users manual.  If you don't
want the paperwork and disk for Version 1.0 please send $25.

Send Check or Money Order To:

William L. Berggren (Bill)
9145 Balcom Avenue
Northridge, CA  91325

If you are interested in buying the rights to this program, or want to
give me an environmental or any job please contact me.  Your required
contribution will enable me to make a better program.

All registered users will get the new Version 2.0 sent to them when and
if available for FREE (if there are more than 20 registrations to this
program, I may not make it otherwise and you will not get it!).  It
should be completed about January 30, 1995.  It will be much improved,
more accurate, with more models.  One model, I just discovered, returned
$35,588 from $1000 in twenty years or 19.7%.  I will also add corporate
bond, zero bond, and stock models.  I may also add some statistics and
bond calculators.  I will also correct this small bias as follows. The
Prime Rate, the T-Bill discount rate, and the T-Bond yield will be
converted to the same unit, such as 6-month annual yield to give more
accurate results. I may have to rename it to BMC, Bond Market
Calculator.  Version 2.0 will not be shareware!

Why do I think these models work?  When I increased the length of the
data set in MODEL1 from 300 to 588, accuracy increased.  I presently
believe, economic data such as inflation and M2 are predictors of the
T-Bond, T-Bill, etc., which are predictors of the T-Bond at t+1 in the
SHORT TERM.
