The Yale Free Press Blog

A More Thorough Analysis of Tin Foil Hats

In the post below, I partially debunk Keith Olbermann's claims about counties in Florida that have 2-1 Democratic to Republican registrations going for Bush (or Bush outperforming significantly there relative to registrations).

Here is a PDF file of a table I compiled of the counties with what seem to me to be the biggest disparities between votes for Bush in 2004 and Republican Party registrations along with their 1996 and 2000 vote totals totals. I don't think that Olbermann's claim of fraud holds up.

The 28 counties I analyzed all voted for the same party's candidate as they had voted for in 2000. Of the 28 counties I looked at, only two voted for Gore in 2000 (the same ones went for Kerry in 2004). It is certainly true that in most of them, Bush increased his totals over 2000. But that is likely due to the fact that Gore was a Southerner (having been a Senator from Tennessee) while Kerry is a New Englander through and through or to the fact that this time terrorism was on the table. (UPDATE: On a second reading of the data, it looks like in 21 of the 26 counties Bush won this time, Kerry actually increased his vote totals. It's just that Bush increased his more, likely the result of a better Get Out The Vote effort by the Republicans. This is completely in line with Ken Mehlman's strategy of focusing on suburban and rural voters - which most of these are since the counties have tiny populations).

Also, in 1996, in an election that was largely a landslide for Clinton and with Perot in the race, Dole won 12 of those counties. Of the 16 Clinton won, he won 5 by less than 2%. And let me remind you that 1996 was a three-way race with about 8-15% of the vote going for Perot, that ended up being a landslide for the Southerner, Bill Clinton. So, it seems that party registrations played very little role. Just as an example, Liberty County has 10-to-1 Democrat to Republican registration. But, it went for Dole in 1996. Similarly, Calhoun County has 7-to-1 Democrat to Republican registration, but only went for Clinton by less than 2%. I think that this concludes that Olbermann's analysis is bunk.

One interesting bit: all of these are small counties, likely rural. Is it possible that maybe a local Democratic official like the County Sheriff or County Commissioner endorsed Bush? Does anyone have that information? If so, I'd appreciate it if you could share it with me.

UPDATE: I have been reading the comments on various hard Left discussion boards and noticed that when my analysis is brought up to debunk the myth, someone always says that this doesn't prove that fraud didn't happen (see the Democratic Underground discussion I linked to below). But in a similar fashion, I cannot disprove that we're all brains in a vat, that reality is a Cartesian dream, that there are pink elephants who fly and give birth to unicorns. The point is that voter registration was proposed by Olbermann and other conspiracy theorists as a good predictor of who people will vote for. They then looked at the data and said, "wait, the people actually voted for someone different than what our model predicts." Instead of making sure that their model predicts even past data, they decide to claim that some kind of fraud had occurred. I am merely saying that their model is clearly flawed since it can't even accurately predict the outcome of the election of 1996. Unless they are willing to claim that there was fraud even then too. But we must recall that many of these counties who voted for Dole in 1996, voted for Carter in 1980. So, the claim then has to become that sometime between 1980 and 1996, some organ was installed in the local county government that has defrauded every single election. I think this is starting to get into the pink flying elephants category of epistemology.

Another reason why any kind of "statistical" techniques for analyzing the election seem problematic is that statistical techniques assume that you have independent trials with essentially the same variables. But from election to election, you have all kinds of different issues on the table, different personalities running, etc. For example, here is an argument that claims to justify this conspiracy theory by saying that the fact that these counties were rural can't explain the disparity. Because if you compare it to rural counties in Pennsylvania, you get different results. Right, but that's why those counties are in Pennsylvania and these are in Florida. Pennsylvania is not a Southern state. Florida is. There are too many variables to list. But I am just saying that this is terrible methodology since it doesn't even attempt to control for the relevant factors involved. As for the argument that the only variable in common with all of these counties is optical scanning machines, I am not sure how this is relevant. What other characteristics do these counties have? Without extensive demographic information there is no way to conclude that there isn't a lurking variable in there somewhere. Look, I'm a Ph.D. student in math. I like numbers and statistics and boo-ya's. But, certain methodologies are just not suitable for election analysis because no two elections are even remotely similar. Period.

UPDATE: The discussion in the comments section in the earlier post requires me to stress this point. NO ONE CLAIMS THAT VOTER FRAUD DOESN'T OCCUR. By God, any Republican from Chicago can attest the falsity of that claim. But, to claim that Republicans do it more than the Democrats is disingenuous. I've seen no evidence that this is the case. Both parties seem to do it about as much as the other. Furthermore, it seems that if you really do care about voter fraud (rather than just making the other side look bad), it's rather imprudent and ineffective to yell "voter fraud!" without thoroughly researching what happened first. The Keith Olberman claim I discuss here is the prime example. The reason is that if you make wild accusations of voter fraud without a reasonable understanding of what is going on, then no one will ever take claims of voter fraud seriously. People will just shrug and say, "Well, both sides do it. I don't know which claim is true and which isn't and I don't have the energy to investigate." If anyone had just looked at the results from the previous two, three, four, etc elections and thought about these counties (all of which are tiny in population), they would have realized that this isn't all that weird. Where is people's common sense? I think voter fraud is very serious and undermines our democracy. But that does not mean that people ought be making spurious claims of fraud before thoroughly researching the subject.

UPDATE: Many people have objected here in the comments and elsewhere about the fact that the counties where this surge occured most (in support for Bush) were all using opti-scan technology rather than e-voting technology. In case you can't see the post I made later, addressing this exact issue, here it is. I hope that this answers your questions. I am now officially done updating this thread. I have a life outside of debunking false claims of fraud/irregularities, you know. So, hopefully others will take up the mantle, as Mickey Kaus said and either debunk the irregularities theories/stories or provide *proof* that something bad did occur on a significant scale that helped one party more than another. I would like to caution people against only posting stories on irregularities that hurt Kerry. I'm sure there were some that hurt Bush. It's just for some reason in no one's interest to investigate them. Ok, I've said my bit.

UPDATE: The transcript for Olbermann's Monday night show is here.

Mobius Strip | 3:13 AM | 28 comments|

Here's an e-mail I sent to Olbermann. Tried to send it to you as well, but accidentally sent it to your editor and the e-mail I found for you via google bounced.

That Cuyahoga County recorded more votes than it has registered voters is interesting, if true.

This source says there were over a million registered voters in that county, while this count doesn't come close to that number. Your claim on MSNBC is that certain precincts voted more than they should have, not that the entire county did, however, so let's look at what we can.

Right now, the county records the county as a big Kerry win. With 67% of that county's vote, it's his biggest win in the state by more than 10%, I believe. Republicans win little counties by big percentages, Democrat win populous counties by small percentages. That's how we get to be a 50-50 nation.

If we assume Bush stole (created out of thin air) 90,000 votes in that county and reduce his total accordingly, then recalculate based on the new numbers, Kerry would then capture 77% of the votes cast in that county, making Cuyahoga an astronomical win for him. This is a plausible occurrence, perhaps, but it seems unlikely. Cleveland isn't that powerful a Democratic stronghold.

Keep in mind, however, that our assumption requires that the precincts with 'more votes than voters' had 100% real turnout, and the only fraudulent votes are these 90,000. Turnout in Cuyahoga was 65%. (Without the 90,000 extra votes, it drops to 55%) How many more votes would Bush have had to steal in order to drive up turnout in these precincts to 100% of the registered voters and then add an additional 90,000 votes?

In the MSNBC scenario, it seems to me that Cuyahoga would really be a 90% Kerry / 10% Bush county. Perhaps you believe this is plausible, but we're skirting the edges of woo-wooism, I think. Party registration goes 3:1 for the Democrats. The MSNBC scenario seems to imply that Bush won only the 70,000 registered Republicans (215,000 votes for Bush counted subtract 90,000 he stole and then subtract whatever number he would have had to have stolen to drive up the turnout in the precincts in question to over 100% ) plus maybe a piddling number of independents, and Kerry won all the 200,000 Democrats and 95% or so of independents who voted.

I, for one, would like to see how the county voted in 2000. I suspect you should check it out.

There is an alternate explanation for where these votes came from (and who they went to), however. This alternate explanation would fit better since it doesn't require Kerry to have 'really won' the county by with approximately 90% of the vote, and it doesn't require a heavily Democratic county with Democratic elections officials to have conspired with Bush's people to steal the election.

I encourage you and MSNBC to keep digging.

Okay, I was reading your post, and I have a couple of objections to how you presented your position regarding the Dixiecrat phenomenon. I am not saying that the phenomenon doesn't exist, I'm just saying that you didn't evaluate the percentage gains in popularity in Bush in 2004 vs 2000. I have also done an analysis, separate from yours (I didn't know of the existence of yours until recently). You mentioned that Kerry only won two of the counties which concurs with my findings. You mention Bush and the Democratic candidate (Gore/Kerry) both increased their votes this time versus 2000. But I think it would have been better to point out that the percent margin of Bush's vote increased in all but one of these counties. This could certainly be due to greater popularity, but it also does not discount the theory of an extralegal skewing of the votes.

I took the 25 counties with the greatest degree of discrepancy between party registration and voter outcome according to the Dopp analysis. These were all the counties with a greater than 100% switch. (You mentioned 28 counties, I'm not certain how you selected them. Some are obviously overlapping with mine.)

There were 11 counties that had a greater than 200% switch. In order of greatest percentage switch to least these were: Liberty, Lafayette, Calhoun, Dixie (with 358 to 712% crossover) and Hamilton, Franklin, Holmes, Madison, Taylor, and Baker (with 220 to 297% crossover). As has been pointed out, all of these switches were in favor of Bush and all took place in rural counties using optical scanners. Each of these counties have a wide majority of Democrats registered over Republicans. This makes sense: if there weren't large Democratic majorities it would be hard to have such large seemingly anomalous crossovers. For these 11 counties, in the 2000 election, Bush collected more votes than Gore. However, in all of the eleven counties, Bush's margin was larger.

How much better did Bush do in these counties? Alphabetically: Baker, 9%; Calhoun, 8%, Dixie 11%, Franklin 6%, Hamilton 1%, Holmes 9%, Lafayette 8%, Liberty 9%, Madison 1%, Taylor 4%, and Union 11%. (mean: 7%)

Two possible explanations for this. Bush was 7% more popular in these counties this time or there was an extralegal skewing of the votes by an average of 7%. The manufacturers of the machines in these counties are: 6 ES&S, 4 Diebold, and 1 Sequoia. (No seeming pattern)

Only two of these are among the ten Florida counties in the Central Time Zone, which some have put forward had skewed year 2000 results.

I then ran the numbers on the next 14 counties in order of discrepancies between voter registration and actual votes (Washington, Jackson, Suwannee, Gadsden, Bradford, Gulf, Wakulla, Levy, Glades, Gilchrist, DeSoto, Putnam, Jefferson, and Columbia with between 111% to 180% discrepancy). All but one had an increase for Bush between 2000 and 2004. That makes for 24 of 25 counties I've checked. All twenty five of these counties are optical scan. The increase in Bush votes between the two elections were in alphabetical order: Bradford 7.2%, Columbia, 7.9%, DeSoto 3.5%, Gadsden -2.6%, Gilchrist 9.2%, Glades 3.0%, Gulf 8.3%, Jackson 5.1%, Jefferson 0.2%, Levy 8.6%, Putnam 7.8%, Suwannee 6.3%, Wakulla 5.1%, and Washington 8.9%. Mean 5.6% increase. (I jotted down two significant figures for this second run) Gadsden and Jefferson counties with the two smallest changes were won by Kerry, the rest were won by Bush.

Nine of these counties had Diebold machines and five ES&S.

Conclusions? Hypothesis #1: Bush was 6% more popular in these 25 counties. The theory that there could have been a subtle extralegal skewing of the vote (by approximately 6%) cannot be dismissed.

Perhaps the more important point for many would be the bottom line. These counties were small in population. Their contribution to the total Florida vote was modest. If the theory of vote fraud is the correct explanation for Bush's increased percentages, then he would have gained 15,000 votes over Kerry from these, the 25 "most discrepant" counties.

mdhill

Seems like folks are jumping off on a few high profile incidents. Any good investigation requires that you first define the problem you want to solve. Agreed?

We can start by asking some questions that will tell us the general health of our election system:
Does our system enable every citizen who is eligible and wants to vote, get to vote, and is every vote recorded as the voter intended, and subsequently counted? That would broadly define a healthy system.

Some non-healthy situations:
How many people were removed from the registration roles? How many people were denied the opportunity to register? How many registration applications were thrown out? (Ex: Ohio, not on the correct weight of paper) How many people requested absentee ballots but failed to get them? (Ex: Fla didnt mail a large # on time to be returned by deadline.) How many registered voters were harassed outside/inside the polls and left or were given wrong directions to the precinct? (Madison, Wisconsin) How many left because lines were too long to wait? (i.e. had to get back to work)
How many voting places ran out of paper ballots, or had machine failures such that people didn't get to vote? How many people believing they were registered were forced to complete provisional ballots? How many absentee and/or provisional ballots were thrown out?
Probably more questions to ask on that part. LOL

People: Then we have to ask questions about the proper conduct of the election officials and poll workers: Were they trained? Did they give bad advice? Did they give the help they are to give like to disabled persons? Did they conduct the election and ballot counting in keeping with the law(s)?

Equipment: Is the operation of the equipment such that security is ensured and that it's intrusion proof? Has the source code of software been evaluated and certified? (Categorically this one is NO!) Have all possibilities of hacking/intrusion been identified and eliminated?

There are more good questions, those are just the ones that come to my mind after reading many reports of irregularities across the nation. One Secy of State is in jail for accepting bribes... etc.

Six Congressmen are asking for a GAO investigation. We don't know yet if there was fraud, but we won't know if there is no investigation. Without this type of analysis, we won't get the things fixed that need to be. Nadar and another candidate are asking for some recounts. They believe something wrong occurred.
There were over 25,000 trained volunteers out there observing, filming, and recording data. They are still reporting their findings. So, we must await more data.
Let's work in a multi-partisan manner and bring our minds and actions to bear on this election problem. ok? ....Sherlock

Anonymous: Equipment: Is the operation of the equipment such that security is ensured and that it's intrusion proof? Has the source code of software been evaluated and certified? (Categorically this one is NO!) Have all possibilities of hacking/intrusion been identified and eliminated?

Please explain the categorical NO. All source code in these machines is rigorously tested, first by the manufacturer, then by an independent 3rd party of the state's/county's choosing. I've been in IT for 25 years and have first-hand experience on the stringent requirements established for vendors in the public sector.

1) I lived in Florida for over 25 years of my life. I can attest to the fact that northern and central Florida is part of "the South" and that southern Florida, particularly Southeast Florida is part of either "the North" or Cuba/Haiti. It is not at all surprising that the rural counties in the north would vote heavily for Bush - I'd be shocked if they didn't. Also, Jeb Bush is a very popular Governor and, no doubt, influenced many to vote for his brother.

2) I personally witnessed a prominent African-American Democrat activist in Miami say she was urging people to vote for Bush back in 2000 because "the Democratic Party has taken the black vote for granted."

3) Maybe the OptiScan machines actually resulted in more votes being counted and fewer ballots being spoiled? Anyone have any figures on this?

4) Midwesterners migrate down I-75 toward the north and west-central portion of Florida (most of these, but not all, admittedly, tend to be conservative and vote Republican). Northeasterners migrate down I-95 toward the South Florida counties of Palm Beach, Broward, and Miami-Dade, and tend to vote Democrat.

5) Many northern Florida counties border on Georgia. It is not surprising that these would have voted for Carter, nor is it surprising that these would, like Zell Miller, have voted for Bush.

Kerry wins in Palm Beach and Broward are predictable, with Miami-Dade being much closer due to the heavily Republican voting by the Cuban population.

Scott