financial calculations

[validgh!uunet!research.att.com!sesvaSun.COM]

Financial calculations can give results which vary between
implementations (e.g., internal computations in decimal
versus floating point, rounding versus truncating).  This can lead
the situation where the corporate mainframe's results
differs by a few pennies from the PC's spreadsheet.  
Some institutions may be sensitive to this.  How
do people in ``the real world'' handle this?  

I asked David Hough about the topic & he responded:
>For instance, in mortgage and bond calculations, each financial
>institution adopts its own rules about rounding, so they're not
>consistent.  Beyond that I don't know much.  An inquiry 
>to numeric-interest might net you more info.

Here are some specific questions for numeric-interest folks:

1) Are there standards for financial calculations which
are as explicit about computations as IEEE-754 is for floating point?

2) What are the elementary operations in financial calculations?
Add, Subtract, Multiply to be sure.  Divide?  Net Present Value?
others?

3) Is David right?  Does this mean 9.75% interest on a loan may be
computed differently by two banks (all other details being equal)?

4) Do users expect ``off by a few cents'' differences?  When is
a ``few cents'' too large?

5) Do any PC-based applications implement decimal arithmetic
for financial computations?

Steve Sommars
sesvaresearch.att.com