This page is preliminary.
I welcome comments and more detailed mathematical analysis than I had time to do myself.
Formatting will be updated (including links) in the near future, though in the mean time, I expect that people reading this will already know where the data is coming from!
____________________________________________
Hey, I have my own reasonable explanation, too!
The people who have countered the Florida optical scan machine fraud charge with historical data, when confronted with the exit poll data that doesn't show the reported shifts, have generally replied that that exit poll data is just wrong, which to me, is not a sufficient explanation. Exit polls are usually right (or, more precisely, not THAT far off), and if you're going to use mathematical science to prove one side of the argument, you can't dismiss it on the other. However, I have a better answer for people on that side of the argument. Actually, the exit poll data can be RIGHT and still be consistent with the "skewed" results. You have to combine numerous elements, and yes, make some assumptions, but ones that are somewhat supportable, and in the end, you can account for the seeming contradictions without calling the exit poll data wrong. Here's the scenario.
First, a brief overview, for those without the time or inclination to read all the explanatory text:
• Bush's base expected vote in the optical scan counties: 1.34 million
• Exit poll indicated higher than expected Republican turnout.
Weighting that turnout 75% toward the counties in question, we can justify another 155k votes.
• Exit poll indicated 14% Dems for Bush, and 7% Repubs for Kerry.
Weighting both kinds of cross-voters 75% in Bush's favor in the counties in question, we justify another 250k votes.
• Exit poll indicated 41% of independents voted for Bush.
Weight that at 60% for these counties, we justify another 300k votes
This raises the Bush total to 1.95 million in these counties, which is what he received/
THIS DOES NOT PROVE the vote is legitimate. However it does show that, IF the Florida numbers are accurate, they don't have to necessarily be seen as at odds with the exit poll data regarding cross-voters.
Now the details:
Let's start with the "getting out the base" variable. The initial figures used in this entire discussion assumed that party members turned out roughly in proportion to the general population, yielding an expected 3 million total Democrats (1.57 million in the E-Touch counties plus 1.43 million in the optical scan counties), and 2.78 million total Republicans (again, E-Touch plus optical). But the exit poll tells us that Republican turnout was stronger, about the same amount ahead of Democrat turnout as Democrat had been anticipated to be ahead of Republican in that initial chart. So let's swap the numbers (not actually mathematically accurate, but in the ballpark), and say that there were 3 million Republican voters and 2.78 million Democrats. So now, we could assume roughly another 220k votes for Bush overall. (And the swing could conceivably be greater than that, since there is a margin of error to the data that said more Republicans showed up than Democrats, a margin of error which could, of course, swing these numbers in either direction. But let's leave the margin of error out for now.)
Since the optical scan counties represent about 47% of the overall vote, we would start by granting Bush 47% of that 220k, or about 103k additional votes in those counties. But let's not assume that this increased Republican turnout was evenly distributed throughout the state. It is not unreasonable to assume that the Republicans' increased, more motivated turnout was indeed in these rural Republican-leaning areas. By how much? I have seen no figures by which to hazard an educated guess. More than 47%? Quite probably. Less than 100%? Quite probably. Let's grant a not entirely unreasonable 75% of our "found" Bush votes to voters in these counties. So then increased turnout of the base can reasonably explain about 75% of 220k, or 155k additional Bush votes.
Now let's add the cross-voting (Dixiecrat) variable. The exit poll again says that 14% of the Democrats voted for Bush. Again, margin of error could further increase (or decrease) that number, but we'll leave that out. Out of 2.78 million Democrats, that 14% yields about 400k extra votes for Bush overall. Now while it is ridiculous to postulate that all 400k would be in the rural optical scan counties, it is reasonable to assume that that they disproportionately gravitated there. Let's say that 75% of those crossover votes (about 300k) were in those counties. Yes, I'm pulling that 75% figure out of the air, but again, it's not an intuitively unreasonable premise. And to be fair, we can't assume that ALL the Republicans who voted for Kerry are in the other counties, so let's put 25% of them here as well (about 50k), so that the cross-over voters of each area favor that area's winner by 3-to-1. High, perhaps, but not beyond reason. Subtract that 50k from the 300k and we still have 250k more Bush votes in the optical scan counties.
So where are we now? We originally assumed a base of 1.34 million votes for Bush. So let's add 155k that we account for by the heavier Republican turnout weighted reasonably towards the rural counties, and the 250k we account for by the net effect of each party's cross-over voters, weighted for who was stronger in which regions. We are already up to about 1.65 million Bush votes we can expect in the optical scan group without things looking suspicious.
Going back to the original chart of numbers, the optical scan counties had about 650k voters who were not registered Democrats or Republicans (that's calculated from 3.42 million votes cast, minus the 1.43 and 1.34 million they originally expected to go to Kerry and Bush respectively, based on party affiliation and turnout percentages). While in this discussion we have subsequently allowed that the Democrat/Republican mix was different, the total number of votes cast remains the same. We postulated 155k more Republicans showed up, but we can't assume that came at the expense of Democrat turnout. In fact, as people who are affiliated with a party are generally more heavily personally invested in an election, I am inclined to believe that heavier Republican turnout should be balanced by a reduction in the number of voters who were not affiliated with any party in order to keep the total constant, and we should not further reduce the Democrat figure (in fact, it's even possible that the Democrat turnout was ALSO higher than expected, just not by nearly as much as the Republicans in these areas, so there is really no case for diminishing the Democrat turnout numbers). So let's subtract those 155k additional Republicans from the 650k unaffiliated, and we have 495k independent voters to consider. The same CNN exit poll data we've been working with throughout this hypothesis also tells us that Bush got 41% of the independent vote. So let's give Bush 41% of that 495k, or roughly another 200k votes. We're now up to a total of 1.85 million.
But again, just because Bush got 41% of the independent vote overall doesn't mean it was even throughout the state. Maybe in these rural counties, he got more like 60% of the independent vote, which would still yield a 41% statewide average if he only got about 25% of the independent vote in the E-Vote counties. (If you're noticing that 60 and 25 don't average to 41, remember, the E-vote population is slightly larger, so you have to factor that in.) Well, with 60% of the 495k independents, we're talking about 300k rather than 200k votes from these voters. Yes, another 100k, bringing our magic total up to 1.95 million. That's how many votes Bush received, and now you see, there is a way to get there, reasonably, without dismissing the quality of the exit polls. Sure, maybe some of the estimates are a little generous. But they're not unreasonable, without even taking margin of error into account, which could conceivably also result in a some slightly higher Bush numbers while still not invalidating the polls.
But please remember, I'm NOT saying that "this is how it happened!" It is just a plausible scenario. Coming up with a plausible scenario does not render other scenarios inherently false. There is still only one way to be sure what happened, and that is, MANUALLY COUNT THE VOTES in some of the suspect areas. Until that is done, we are only arguing about whether something is more or less likely.
---
