Thank you for your help, Steve. I have confirmed it's a numeric precision
issue. Now I'm trying to find the best way of avoiding it.
My work situation involves training actuaries to use SAS for their
analyses, and to help program and trouble-shoot for them. Granted, $5
billion (down to the penny) is a big number, but it should be well within
the largest number 8 bytes can hold (according to the numeric precision
table for SAS under Windows, the biggest integer would be 9 quadrillion or
so. To represent a billion and change would take fewer significant digits,
and the mantissa would be within the right range, too). So is there a non-
manual way to have SAS cope with this level of precision (rather than sub-
dividing the number, summing the parts separately, and manually combining
them at the end)?