Suppose there is an election in which all counting districts are chosen so that the expected vote total from each is 1000 votes, and the votes counted fall within 10% of this figure. Suppose also that the two candidates each poll close to 50% overall, and the variance between counting districts turns out to be within 20% of this figure (i.e. all districts report between 30/70 and 70/30 votes for all candidates). Based only on the above information, would it be possible to use Benford's Law on the first digits of the numbers of votes cast to determine whether the results were, or were not, suspect?
I'll supply the answer later, if anyone hasn't got it immediately.
Dave