There are a series of arguments and counter-arguments for this one, so let's start at the beginning...
In Florida, the figures show that, in counties with one type of voting machine, voters with no Democrat or Republican party affiliation appeared to split their votes roughly 50/50 between Bush and Kerry, which was to be expected; yet in counties with another type of voting machine, unaffiliated voters seemed to vote nearly 100% for Bush!
Click for a table that shows, county by county, how many registered Republicans there were, how many registered Democrats there were, and what percentage of registered voters turned out to vote. From those figures, they arrive at an "expected" minimum vote for each candidate. However, since many more people vote besides those who are registered Democrats or Republicans (i.e. Independents, unaffiliated, or members of some other party), the actual vote for each candidate is naturally going to be higher than the "expected" figure which is a projection based only on registered Democrats and Republicans. The table also then shows the actual vote, as of 98.6% of the vote tallied.
Some Florida counties used an "E-Touch" voting machine, and some used an "optical scan" machine. Let's look at these two groups:
These two Florida populations who seem to have voted so differently are roughly equivalent in size and relative party strength. Yet the counties with the "optical scan" machines went disproportionately, overwhelmingly, and seemingly illogically for Bush. What could account for the difference in results? Did the machine itself make the difference, or is there another explanation?
VARIABLES THAT HAVE BEEN CONSIDERED:
Geographic distribution of machines
E-Vote machines tended to be used in larger/urban ares whereas optical scan machines tended to be used in smaller/rural areas, so without a random distribution of the type of machine used, the discrepancy could be based on the different perspectives of those two populations rather than on the machine itself. And that makes sense, considering the "Dixiecrat" phenomenon which we'll discuss shortly. However, it is also worth noting that further analysis showed that, in mid-size counties (some of which used each type of machine), there remained a significant correlation between unusual strength for Bush and the use of optical scan machines even when comparing counties of similar population size.
It has also been pointed out that the rural counties have been trending more heavily Republican for some time, out of proportion to what would be expected from party affiliation. That data from the last few presidential elections has been posted . While past results do provide additional support for believing this year's numbers, there continue to be arguments on the other side, discussed below. In addition, if there is, by some chance, a flaw in these machines leading to a higher result for the Republican candidate, can we be sure that this same flaw did not exist in prior recent elections?
Regardless of whether or not you accept the exact numbers from the historical data discussed in the previous paragraph, it does seem clear that there is a regional rural phenomenon that many registered Democrats do indeed tend to vote Republican. And that certainly provides some explanation, as the estimated base figures discussed above were based on an assumption that, while there are crossover voters in each party, they would probably balance themselves out in about equal numbers. So does this mean we have a fully suitable explanation? Not quite. According to a CNN poll, while 14% of Florida Democrats did indeed vote for Bush, 7% of Florida Republicans in turn voted for Kerry. The net difference is not nearly enough to account for the discrepancy. It wouldn't even have been enough if there were no Republican cross voters, or if all the cross-voters were in counties with one kind of machine, and none in counties with the other. However, on the other side, one reader suggests that, since the CNN data is based on an exit poll, perhaps there were registered Democrats who voted for Bush who did not identify themselves as such when polled. However, while exit polls may be flawed for a variety of reasons (including reasons like that), these polls would have to have been off by a tremendous amount to yield the results necessary to use the Dixiecrat phenomenon as a full explanation for the final results. And even bad exit polls aren't that bad.