Name of theory for fraud spotting by counting number of digits

From: Carlo (ckielstra_at_hotmail.com)
Date: 11/18/04 

Date: 18 Nov 2004 15:27:10 -0800

About a year ago I read an article on analyzing a random set of data
for possible frauds.

The theory was based on the digit '1' being the most used number in
the universe, then '2', '3', etc. Most humans tampering with data will
try to use random numbers, but in doing so they create an evenly
distributed set of digits.
The theory worked for any true random data set and any number system.
Funny side note was that shop prices aren't random, the number 9 tends
to be more common than in other data sets.

What was the name of this theory so I can look it up again?