« No Honeymoon for Bush, No Parity on Party ID for Republicans | Main | Old Democrats, New Democrats, Newer Democrats »

Did Bush Really Benefit from E-voting in Florida?

By Alan Abramowitz

Perhaps you've seen or heard of an analysis by Michael Hout, a Berkeley sociologist, of the impact of e-voting in Florida. Hout and his associates claim that Bush did better than expected in the 15 Florida counties using e-voting. See the link below to their report.

I did my own analysis of their data. It does not support their conclusions. In fact, I find that Bush did slightly worse than expected in the 15 e-voting counties.

I did three things. First, I just compared the change in percent for Bush in Florida counties with and without e-voting. Contrary to their conclusion, Bush gained more support in counties without e-voting. Then I looked at a scatterplot of Bush2004% by Bush2000%. There is no indication at all here of any non-linearity in the relationship. Therefore, I cannot see why the Hout team added a quadratic term to the model. Then I did a regression analysis of Bush2004% with Bush2000% and a dummy variable for e-voting counties. The dummy variable had a negative but statistically insignificant effect. So if anything, Bush did slightly worse in 2004 in counties with e-voting when you control for his support in 2000. My guess is that this is because the e-voting counties tend to be in large metropolitan areas but Bush's gains were greater in smaller, rural and exurban counties.


"It's called bait and switch..."

We've been hoodwinked.

How to steal an election (plan 12-b):
1. Get opposition worked up about the dangers of voting method A.
2. Rig devices involved in voting method B.
3. As opposition is looking closely at method A, use method B, which the opposition agrees is safe to secure the election.
4. Leave opposition wonder what happened.

We need to watch all their hands.

Thank you for doing this. The steps in the analysis you did (checking whether there was an effect, comparing 2004 graphically with 2000, directly testing the e-voting effect) sound straight forward and simple. This may be a case in which less is more (e.g., avoiding the fancy but questionable quadratic term). What is sad is that the Hout's analyses distracts from the better substantiated and widespread reports of voter suppression in Ohio and elsewhere, such as the failure to properly allocate the voting machines to inner-city minority districts.

actually, its my understanding that the Berkeley study agrees with your findings -- when considering all 15 e-voting counties they don't find any bias. Only when you just compare just the 6 democratic counties with e-voting did they see it. And the bulk of that bias comes the 2 or 3 largest of those counties. And so the most cogent criticism of the study for me is that are many reasons those 2 or 3 democratic counties might have moved towards bush -- burying this fact in a complicated statistical analysis isn't very helpful.

It is good that this website finally addresses the issue of Votergate 2004, albeit only to pooh pooh it. It would be good to try to confront the evidence in greater particularity -- it seems basically like a mere assertion of conclusions.
I personally do NOT consider the Berkeley study the cutting edge on the issue -- a study, including the Coleman study that was used as the "basis" (face-saving rationale) for Brown v. Board can always be challenged ("refuted", when the media, as in the NY Times front page article of Nov 12, has the agenda to so assume). But the quickness with which efforts are made to REFUTE the theory of voter fraud here WITHOUT addressing the serious powerful allegations about the systematic deprivation of adequate voting machines, the hacking theory in Fla, which canNOT be dismissed as mere "Dixiecrat" counties, (and the problem that these counties, as comparisons, might be tainted data) and of course the exit polls are avoided here as in the press. I think that one of the reasons for the "civility" regarding the "New Democrats" is that the situation is so much broader than a 'new' v 'old' Democrat issue -- with a general media lockdown of Votergate 2004 following the less remarked-on lockdown on timely debunking by commentators of the flipflop spin and the Matt Bai terrorism distortion (not to be confused with his justifying of the lying on Ohio) showing that the problem is the whole machine, not just the New Democrats. Yet I am one UNCIVIL Democrat who sees the New Democrats, ALONG WITH A LOT OF OTHERS, as very much a part of the problem.
I don't (lacking the data) dispute the conclusion, but I do dispute the focus EXCLUSIVELY on debunking the fraud issues rather than addressing all of them FULLY. This is what points in the direction of the function of the "New" Democrats, and many others, within the system.

The interesting question about this is why did Kerry not substantially improve his showing versus Gore's showing in the primarily Dem counties that mostly used e-voting, while Bush increased his margins by about 5%, or so, in many other counties in Florida. Dems need to understand why they were unable to make more inroads into these groups, which you would think would be Seniors, Hispanics, and Jewish voters to name a couple of groups.

Thank you for the analysis. But still the fact that this sort of analysis is a necessary part of ascertaining whether the election results are accurate or not is absolutely crazy. What happens when there are no paper ballots to measure e-voting results against? Is America about to become the world's first faith-based democracy? (ie. faith in the Ahmanson brothers)

Compare e-voting in Canberra, Australia: open source coding and paper receipts to check and then drop into a ballot box for possible recounts. Simple. Logical. Verifiable.

When so many in the US fight against such a system you can't but help ask why.

This non-American hasn't yet heard any good answers.


I don't mean to nitpick, but I know how careful Mike Hout is and so I just read the report and your comment and there is something important to say. Your analysis makes sense and it is what I would have done to begin, but the Hout piece is much more subtle and cannot be ignored on the basis of your findings.

The key thing to note is the interaction term of Bush2000% with e-voting makes the thing complicated to interpret even if it is statistically superior. Perhaps it would be better to argue that they could have addressed the lack of evidence in the more simple specification and motivated their discussion better.

On non-linearity; just looking at the scatterplot shows no direct sign that there should be a non-linearity, but the model comes up significant with the quadratic and this suggests that the eye lies. There is curvature in the given specification and there must be good reason for them to include it. I assume they keep a quadratic on the interaction to complement the non e-voting curve.

The Hout model is complicated and it is troubling that your direct examination fails to provide a smoking gun but it strikes me as unfair to dismiss the findings with the argument you make. For example, you don't pull the killer move by claiming that you reproduced the model and they were wrong by committing some obvious error. This is a good and defendable model as far as I can tell.

If your regression with its simple story were to come up significant, I am sure we would have had a recount of some kind by now, but there is still a story in the Hout model. It is just a typical Michael Hout story that requires several hours of serious head scratching to fully take in.

Solon Simmons


I looked into this some more and here is the answer. Restrict your sample to only those counties that Bush lost to Gore in 2000. This is where the action in the Hout model is. Regress b_change on etouch if for example b00pc is less than some number. For example I ran the following command in Stata, "regress b_change etouch if b00pc<=.4" on the data from Hout's website and the results suggest a 4% difference in e-touch counties as compared to non e-touch counties. That is pretty strong. If you set the limit at .5, the difference is 2%. Under the scenario suggested here, some evil plotter only conspired to rig things in the counties that Bush lost last time around. In the others they let it ride. This is a simple story as you had tested and is completely consistent with the data.

I'm still waiting for someone to explain how most of the Florida counties ended up with the equivalent of 100% of registered Republican voters casting a ballot. That plus in some areas all the independents apparently voted republican as well as many Democrats.
I have lived in Florida for almost 40 yrs, most of my voting life, and this is the first time I have ever seen so many angry people going to vote. Yet the experts will have me to believe that these angry people cast their ballots for Bush? Angry people vote for change. They don't vote to keep the cause of their problems.

I live in one of the optical scan counties which had a 100% Republican turnout plus all the other miracles which benefited Bush. It reminds me of a comment written by a gentleman who did a study of the 2000 election:

"Ballots were destroyed because they were cast for Gore. The probability that Escambia's vote distribution occurred by chance is less than one chance in ten raised to the power 56 which means that the Escambia results are about as likely as Santa Claus, the Tooth Fairy and the Loch Ness monster running into one another at a Shriner's convention."

The Bush miracle that happend this time was even more astounding. Sorry guys, I've been watching this fraudulent panhandle for years and I quit believeing Santa Claus, the Easter Bunny, etc.,
many years ago.

I do believe that fraud took place on a massive scale in this recent election. I believe that the Republicans knew they could not count on another close call & the intervention of the SC, so they did what no one was counting on.... massive voter fraud. Most of it probably took place with the touch screen machines which cannot be checked but they went for optical scanners too, just like they did with the S-election of 2000 but on a larger scale. Here are some comments that a friend wrote about the Oklahoma Election which seems to have been another miraclous event for Bush:

"Recognizing most all newspapers print parcel voting results up to the time they go to press. In Tulsa this amounted to 70% of all the votes cast and the local conservative paper showed Kerry leading in 57 counties. That means with 70% of the vote total Kerry and bush had received a specific number of votes.

And when all 100% of the votes were counted 37,982 votes had been stolen from Kerry's total as they appeared at the 70% count.. AND...... in the entire 30% of the votes later counted............. Kerry had not received a single vote. In other words out of some 420,000 votes no one, not one person,, had chosen to vote for Kerry! And almost 400,000 had voted for bush"

Voting Machines Count Backwards in Okla.