|
I apologize up front for the length, and only the most dedicated number cruncher might find it interesting, but I think it is important to get as much feedback as possible on this. I'm just a working guy trying to break down the election numbers on the side to ease (or confirm) my own doubts regarding the validity of the election.
I'm looking for anyone who may have access to or know where to find information regarding precinct level data for Broward County, FL. Maybe someone, somewhere has already begun this work, and I'd hate to duplicate. I'll start by briefly summarizing what I have done thus far / am planning on doing. Then I'll report on some interesting observations after only starting to wade through the mess.
I've gathered a lot of information from publicly accessible sites regarding precinct information in Broward. I have collected published results by precinct from both the 2000 and the 2004 elections. I have an Arcview precinct map of the county for both the 2000 election and post-2002 re-districting. I'm looking to consolidate registration stats, vote totals, and precinct location/naming changes into one data source.
I have been in contact with the Broward BOE and they have actually been extremely helpful thus far. I'm currently being forwarded hard copies of the official precinct maps for both elections, as well as a couple of other items which I will describe below.
First, let me describe the frustrating mess that collecting this data has become:
1) As stated above, the precincts have been modified. In some cases, it is just a slight boundary change, in others, surrounding precincts have been completely rearranged. Some have been combined, some have been split. Some are unchanged, and some have dissappeared. It truly is a mess. After getting into it, the first thought that came into my mind was: What better way to hide fraud? Make everything as chaotic as possible to forensically study, and many things can be hidden. I have spent many hours deep within zoomed in precinct map printouts comparing the 2000 and post 2002 maps, looking for the adjustments. I've created a method for identifying which ones have gone unchanged geographically, and which groups of 2000 precincts geographically coincide with which goups from the post 2002 map. And, yes, it is manual. This is very tedious. I'm about a third of the way through the county, and there are a number of green-highlighted precinct map sheets on my desk. I should be getting the offical hard copy maps from the 2000 election as well as the 2004 election this week from the Broward BOE to verify this work. I find it amazing that there was no available information at the BOE which mapped these changes or described the logic or criteria for the changes that were made. But, I'm making do with what is available.
2) There was a huge drive to early/absentee voting in 2004. On the surface, this might not seem like a problem. But for data analysis, it most certainly is. In 2000, there were 48,562 Broward votes cast in this manner, or in so-called "Absentee Precincts." In 2004, the number was 278,236, a 572% increase. 176,000 or so of these votes were from "Early Voting Locations." Like Absentee votes, these are tracked as consolidated "super-precincts". Absentee votes are grouped by "style", which I'm assuming refers to the available races on the particular ballot. I believe early votes are grouped similarly, i.e. a combination of particular sub-precincts into one large, super-precinct. It becomes practically impossible to extract the sub-precinct level data from the larger data sets to compare to past elections. I've asked the BOE if this extraction is possible, and they tell me "NO". Reason: it violates voter anonymity. Back when there were fewer absentee ballots cast, it might be possible to determine the choice of a particular elector by knowing from which precinct the absentee ballots were cast, and the candidate breakdown for the absentee sub-group, since they are tracked separately. This seems a little ridiculous. Especially now, with approximately 40% of the county's votes being cast before election day. My contact at the BOE totally agrees with me on this one and is doing everything he can to address it. He even said he asked the question publicly at the infamous Florida BOE conference (crashed by Bev Harris). He has agreed to get a me a copy of their "Recap" report for 2004, which in short, gives the total (not candidate breakdown) absentee ballots cast by registered precinct. He's not sure if this is available for 2000, but is working on it. Along with knowing which precincts are grouped together to form the larger early & absentee style "super-precincts", it is possible to pretty much (with just a few assumptions) nail down the absentees & early vote results by sub-precinct.
Two things worry me about this so far. First, the response to the question at the BOE conference was that there will be a push to go to and "election period" as opposed to an election day, and the need for the small precincts goes away. In other words, in future elections, ALL votes may be cast in these "super-precincts", thus rendering it even harder to forensically investigate. Especially if everything changes at each election cycle. The second thing that worries me is that there has been a week or two delay in getting me some of this information, specifically the 2004 Recap report, which my contact has said needs to be redone because the wrong precinct definitions were used. That concerns me. He had previously told me they would be ready in like a day or two.
3) I've finally collected and organized all of the registration statistics for both elections. The source for 2004 was the BOE's website, and the source for the 2000 #'s was the metadata in the Arcview file. But, because of the precinct adjustments, some of these numbers are worthless without solutions to 1) and 2) above. I'm specifically looking for registration increases / decreases in precincts, or groups of precincts, which have not changed since 2000. I think I have this data ready, but it needs to be consolidated with the absentee votes, early votes, and the re-districting.
Observations thus far: Out of the 200-300 precincts that I have looked at so far, I have identified 57 precincts which did not geographically change between the 2000 election map and the post-2002 map (to be verified with hard copies). Of these 57 precincts, 51 showed a net change in registration advantage to the Democrats between 2000 and 2004. Some precincts lost republican registration, some gained, some lost democratic registration, some gained. But, in 51 out of 57 counties the registration margin changed in favor of the Democrats. The Democratic registration margin in these 57 precincts was 5,833 people higher than it was in 2000. Conversely, the election day results show something completely different, and rather alarmingly. In the 57 precincts, the democratic election day vote margin decreased in 53 counties. The Democratic election day vote margin was 7,575 votes less than it was in 2000. And this is with significantly less people casting votes on election day this time around. Wouldn't you expect an increased Democratic registration margin to translate into an increased Democratic vote margin? I don't see how it could possibly go the other way, and go the other way so definitively. Here, the increased Democratic registration margin translated into a hugely decreased Democratic vote margin. Makes no sense. Precinct 5R is a typical example of this large anomaly. There are many, many other similar cases, and I'm only a third of the way through the county so far.
Precinct 5R: Republican Registration Change: -4 Democratic Registration Change: +69 (+73 Dem Margin)
Bush Election Day Vote Change: -107 Gore-Kerry Election Day Vote Change: -289 (-182 Dem Margin) (Remember, many less Election Day votes due to more Early voting)
Another example is Precinct 28E: Republican Registration Change: +43 Democratic Registration Change: +210 (+167 Dem Margin)
Bush Election Day Vote Change: -71 Gore-Kerry Election Day Vote Change: -406 (-335 Dem Margin)
Precinct 5R shows a 255 person swing between Change in Registration Margin and Change in Election Day Vote Margin. This swing represents 10.9% of all registered voters in this precinct (2348).
Precinct 28E shows a 502 person swing between Change in Registration Margin and Change in Election Day Vote Margin. This swing represents 16.6% of all registered voters in this precinct (3030).
Out of the 57 precincts, not one showed a swing between Change in Registration Margin and Change in Election Day Vote Margin that favored the Democrats. Not one. The swings ranged from 6 to 614 and represented anywhere from 1% of the precinct total registration to 30% of the precinct total registration. That is an alarming fact. One that shouts fraud on election day. Theoretically, I would expect a 50-50 split, in some cases the registration drive benefitted the vote, in other it didn't. But not 100% across the board. That's saying that in every single case, the Democratic registration drive (which was quite succesful, 51 out of 57 favored the Dems) failed to produce a higher margin of Kerry votes (and it consistently failed, 53 out of 57 precincts showed Bush gains against the Dems)
One possible explanation of this anomaly is that the absentee/early ballots were cast more in favor of Bush in 2004 when compared to 2000. The numbers will show that this is a far-fetched theory. The absentee/early votes would have to go so much more in favor of Bush than in 2000 that it is highly unlikely. After all of the absentee/early data information that I have requested is available, I will include these numbers in the analysis. However, as I've stated earlier, this data is difficult to come by.
For all who read completely through this, thank you. Please critique the method and offer suggestions. Please point me in the direction of others who may be doing similar work. Please give me the strength to finish what I've started. It's tough work.
|