Almost as soon as the polls closed (indeed, even before they did) there were news reports of confused voters who, seemingly through no fault of their own, had suddenly become disenfranchised.
It was November 2000, Palm Beach, Florida. And hundreds of retirees were claiming that the confusing, infamous and illegal "butterfly ballot" had caused them to vote for Pat Buchanan instead of Al Gore.
Reports on the "butterfly ballot" made for a an interesting series of anecdotes immediately in the wake of the 2000 election. But did it really have as great an impact like these people said it did?
The Florida Department of State maintains its election results data on its
website. It gives the official return for each county for each candidate in each race. People have anecdotally claimed that there was a problem with the Buchanan vote in Palm Beach County, so one would then expect to see those allegations reflected in the data.
First off, let's compare Bush's Vote share in each county with Buchanan's (Figure 1).
Figure 1
Well, there's a lot clustered in the middle, and Palm Beach is over to one side. Perhaps this is not the right way to look at this. Let's take into account the population of each county, and just plot the ratio of Buchanan's votes in a county to Bush's (Figure 2).
Figure 2
Now, this is incredible. Buchanan got more than three times as many votes in Palm Beach County as he did anywhere else. And, while there seemingly is a correlation between Buchanan votes to Bush votes, maybe this is just an artifact of county size.
Figure 3
Figure 3 plots the Buchanan vote vs. the total vote in the county. The Palm Beach data is still a huge outlier, but we can see now that the correlation between Bush and Buchanan votes in all the other counties in Figure 2 may be artificial, caused primarily by number of votes in each county. Let's conver Buchanan's share of the vote to Percent.
Figure 4
Two major things here. First of all, it seems that Bush's share of the vote is highly dependent on the size of the county. He did better in smaller counties and did worse, percentage-wise, in larger counties. Palm Beach County continues to be an egregious exception.
Figure 5
Figure 5 gives the Buchanan/Bush ratio as a function of county size. And this really informs us. Not only did Buchanan do better in smaller counties than in large counties, but he did better with respect to Bush in those counties as well. This is why Figure 1 looks as ugly as it does. But, as seen in every treatment of the data since Figure 2, Palm Beach County sticks out like a sore thumb.
So is this proof of the effect of the "butterfly ballot"? Well, here is a treatment of data that seems to corroborate the anecdotal evidence. It seems to me that Buchanan got about 2500-3000 more votes in Palm Beach County in 2000 than he should have, based on the trends in evidence in the rest of the state of Florida.
For those who have been searching for evidence of electoral irregularity, please demonstrate that your data has resulted in some outlying data point somewhere. Democracy would much appreciate it.