Hi, Tipsters!
I have been looking at the "unofficial" results of the Florida recount, and
they seem odd from a statistical point of view. But, I typically don't do
statistical analyses on political data, so perhaps someone can tell me
where my reasoning is wrong (if indeed it is).
The total number of votes after the recount is about 3500 higher than from
the original count. In other words, about 3500 votes not counted the first
time were "found" during the recount.
Because the statewide Bush/Gore split is essentially even, then one would
expect that the undercounted ballots would be evenly split between Bush and
Gore. Thus, our expected number of undercounted ballots would be about
1750 for each candidate.
But, that isn't what happenned. Gore had 2500 undercounted ballots, and
Bush only 1000. Using a Chi Square and testing against an expected
frequency of 1750 each, that is a whoppingly significant deviation from
chance. In other words, it seems like something may have caused Gore votes
to be undercounted more frequently than Bush ballots.
Note that there are ways to explain this without claiming fraud - for
example, there might be less experienced election volunteers in Gore
precincts, or more first-time voters in Gore precincts. Of course, the
results are also consistent with an allegation of fraud.
Can anyone spot a hole in my reasoning? Note that this analysis should not
depend on the fact that the number of undercounted votes is small compared
to the 6 million votes cast. The distribution of undercounted ballots
should still be randomly distributed in proportion to Bush/Gore.
-- Jim Dougan
