Properties of Data Sets that Conform to Benford's Law

The concept and application of Benford's Law have been examined a lot in the last 10 years or so, especially with regard to accounting forensics. There have been many papers written as to why Benford's Law is so prevalent and the concomitant reasons why(proofs). There are, unfortunately, many misconceptions such as the newly coined phrase "the Summation theorem", which states that if a data set conforms to Benford's Law then the sum of all numbers that begin with a particular digit (1,2,3,4,5,6,7,8,9) should be equal. Such is usually not the case. For exponential functions (y=aexp(x) it is but not for most other functions. I will show as to why this is the case. The distribution tends to be a Benford instead of a Uniform distribution. Also, I will show that if the probability density function (pdf) of the logarithm of a data set begins and ends on the x axis and if the the values of the pdf between all integral powers of ten can be approximated with a straight line then the data set will tend to conform to Benford's Law.