Gentlemen:

We have had favorable responses regarding the use of our Volatility
Reduction Method (Roger, I've attached a copy of our article in case you
hadn't seen it).  However, there continued to be a quibble about how to
create the set of data points that would be inputs into the testing process.

Last week the consulting arm of a "Big Five" accounting firm indicated that
the following method proposed by us would be acceptable.  We believe this
method overcomes the statistical problems that arise from using interest
rate differences from overlapping ("rolling") quarters.

Method:

1)Calculate daily yield curve changes expressed as ratios, using historical
rates from the most recent, say, two years. (Note: no overlap).  This
results in a set of around 494 vectors of ratios (approximately 247 trading
days per
year).

Example:
If the first three yield curves in the historical set look like this:
19980801   6.5  6.6  6.7  .........  7.2
19980802   6.3  6.3  6.6  .........  6.9
19980803   6.6  6.8  6.9  .........  7.1

Then the change from 8/1/98 to 8/2/98 is:
6.3/6.5   6.3/6.6   6.6/6.7 .......... 6.9/7.1

And the change from 8/2/98 to 8/3/98 is:
6.6/6.3  6.8/6.3  6.9/6.6  ......... 7.1/6.9

2)Randomly select 62 of these "ratio" vectors (approx. 62 trading days in a
quarter).

3)Multiply these ratio vectors together to get a single vector (ie, the 62
6mo ratios are multiplied together, the 62 1yr ratios are multiplied
togeter, etc.).  The result represents a single quarterly yield curve
transformation.  Apply it to "today's" yield curve.  The resulting yield
curve represents one simulated quarterly change in interest rates

4)Repeat steps 2 and 3 until an adequate number of yield curves are
generated, say 100.

5) Proceed with testing process.


I would be interested in your comments.


Leslie Abreo
Andrew Kalotay Associates, Inc.
61 Broadway, Ste 3025
New York NY 10006
Phone: (212) 482 0900
Fax: (212) 482 0529
email: leslie.abreo@kalotay.com

Visit AKA's website at http://www.kalotay.com




 - FAS133 article.pdf