Urban vs. rural voting patterns in Florida - comments by Andy Royle and others

Date: Thu, 04 Nov 2004 12:54:50 -0800 (PST)

From: J.Andy Royle

To: sswang@Princeton.EDU

Subject: plot


The attached figure is a plot of log(observed votes) - log(expected votes) vs. log(registered voters) (let's say that is proportional to population size) for both Democrats and Republicans. They show clearly the pattern of rural/low population counties favoring republicans and this advantage diminishing as the number of registered voters increases up to about log(R)=11. The red dots are the optical scan machines and the black dots are the electronic machines. The black dotes are roughly where you would expect them to be based on population size. If anything, they slightly favor democrats (consider drawing an eyeball "smooth" through all of the dots).

In summary, the GOP message seems really to have paid off in these low-population counties, which happen to all be optical-scan. Of course, one could argue that the fraud effect just happens to be confounded with voting machine type (fortunate if fraud=true), but I would be skeptical about that argument.



Comparison of last-week polls with real outcomes

Date: Thu, 04 Nov 2004 16:37:41 -0500

From: Andy McCord

To: 'Samuel Wang'

Cc: kathy@directell.com

Subject: Optical Scan Results

It seems to me the "expected Republicans" figures may be badly skewed by the definitions that base expectations on party registrations. The counties involved seem most or all from the panhandle or other north Florida areas less affected by recent developments and demographic changes. I don't know Florida but do know a bit about the South. These voters may be descendants of the white Dixiecrats, who still vote Dem in local elections but go Republican in national or even statewide elections. The extraordinary turnouts could be explained by the parental notification for underage abortions measure. It might be more accurate to compare to votes for Republicans in the 2000 election and earlier, though in 1992 and 1996, Bill Clinton would have had more appeal to white legacy Democrats than most Democratic candidates.

In any case, I think people need to pay very close attention to every county where extraordinary increases in Republican votes and turnout occurred. This could be a clue to fraud, but probably more likely would be valuable in analyzing how the values appeal was executed.

Needed a password or something for the public thread, otherwise I'd have posted there.


Date: Thu, 04 Nov 2004 13:29:28 -0800

From: Chris Mundt

To: sswang@Princeton.EDU

Cc: kathy@directell.com

Subject: explanation for touchscreen versus optical

Dear Sam:

I enjoyed following your website during the election - thanks for providing the service. I just wish that the election had turned out differently! But, hopefully, none of us the ability to control that directly.

There seems to be a simple explanation for the touchscreen versus optical results in Florida precincts. I would guess that it has nothing to do with voting medium but, rather, was influenced by community size. I'm in the middle of a couple of grant propsals right now so time is limited, but have quickly scribbled my reasoning below:

1) Precincts with smaller total votes generally represent smaller communities.

2) The touchscreen machines were used in large cities, but usually not in small towns.

3) If you look at the % change data, there is a negative number on the democratic side every time the total vote count is 35,000 or less - NO EXCEPTION. This is true for both touchscreen and optical scan.

4) There are only two cases of precincts with total votes 35,000 or less for touchscreen voting. In contrast, precincts with less than 35,000 votes are very common for optical scanning. Again, I assume that this is because optical scans were used in small towns and rural areas, while touchscreens were used in large cities.

5) Conclusion: there was a larger percentage of democratic voters who went for Bush in smaller communities than in larger cities. Sociologically, this makes some sense.

Hope this helps.

Chris Mundt

Date: Thu, 04 Nov 2004 17:32:25 -0500

From: Peter Nau

To: sswang@Princeton.EDU

Subject: Florida voting "anomalies"

Prof. Wang --

I believe there is a simple, common sense explanation for the so-called anomalies found in the different predicted results of electronic vs. optical scan machines.

The deviation from the "expected" result is not, I believe, due to a difference in accuracy of the equipment. Nor do I think any widespread fraud has been exposed.

The difference is likely due to the fact that scan machines were used primarily in rural counties where Bush has much of his support. The fact that these counties have many more registered Democrats than Republicans is nothing more than a legacy of "Old Florida," where voters tended to be registered Democrats. If you want to know the true character of these counties, you need look no further than the percentage change columns, which showed dramatic increases in registered Republicans and significant decreases in registered Democrats. While this still leaves many more registered Democrats than Republicans, these Democrats were about as committed to John Kerry as Zell Miller was. They are Democrats in name only, who simply have not bothered to reregister as Republicans. Look at the counties where scan machines were used and had similar percentage increases in registrations of both parties. Voila! Your "expected" results resemble your actual votes.

Still not convinced? How about the fact that these discrepancies appeared in the 2000 election? For example, Lafayette County, which voted overwhelmingly for Bush in 2004 despite having vastly more registered Democrats ALSO voted overwhelmingly for Bush in 2000. This was BEFORE a 459.3% increase in registered Republicans and a 69.3% decrease in registered Democrats.

If you are STILL not convinced, I think a quick trip through these counties to meet the voters would do the trick!

As a Kerry supporter and fairly new Florida resident, I understand the temptation to find something nefarious in the results shown on the ustogether.org web page. The fact is, there is nothing sinister about them.


Peter Nau

Date: Thu, 04 Nov 2004 14:52:49 -0800

From: Kevin Malone

To: sswang@Princeton.EDU

Subject: The Florida Numbers

It is very interesting to see the large difference in the outcome for different voting systems. I hope you are going to look into these numbers in more detail. My quick review of the numbers indicates that the optical scanners were used alot more in small counties (total number of registered voters) that are likely heavily rural. The question is are these counties also heavily protestant? If so, exit polling indicates that white protestants voted overwhelmingly for Bush and this would explain some of the diffference (See http://www.cnn.com/ELECTION/2004/pages/results/states/FL/P/00/epolls.0.html for details). However, I agree the difference seems almost unbelievable given that democrats made up > 70% of the registered voters in these areas. This would imply that the overwhelming majority of the rural democrats voted for Bush!

Date: Thu, 04 Nov 2004 15:29:33 -0800

From: jcschank

To: sswang@Princeton.EDU

Subject: Florida voting patterns

Hello Sam,

When I saw your posting on the Florida results I was really concerned by the difference in party registration and voting. So, I looked up the election results for the questionable counties for 2000 and although, for example, Lafayette county is almost entirely registered democrats, they voted overwhelmingly republican in 2000 as well. These counties are almost entirely white, so I would guess that they are old-south democrats who vote republican.

Jeff Schank

Date: Thu, 04 Nov 2004 18:45:09 -0500

From: David Fry

To: sswang@Princeton.EDU

Subject: Voter results in Florida


Thanks for developing and updating your powerful election Web site. I'm just as disappointed with the result as you are.

But isn't the issue with the optical scan machines in Florida obvious? Those are from the less populated, more rural counties. And we already know those populations are where Rove's 'value message' and GOTV push was more successful. The smaller and more rural a precinct was, the more likely Bush was to exceed his 2000 performance. Look at Hamilton country in Ohio, for instance.

This is borne out by looking at the individual optical scan counties in Florida. The smaller ones (e.g. Baker, Calhoun, Dixie, etc.) had the large Republican outliers. The larger counties with optical scan machines (e.g. Brevard, Duval, Orange, etc.) had percentages that fit inside the same range as the electric machines counties.

Unfortunately this election was determined not by voter fraud but by several million people who thought it was more disgusting to watch two men kiss than to watch thousands of people die needlessly.

David Fry

Date: Thu, 04 Nov 2004 16:24:11 -0800

From: Karl Rimbach

To: sswang@Princeton.EDU

Subject: Possible explanation for Florida numbers

Hi Sam,

I was very intrigued by the linked table of Florida counties by voting method. It certainly looks suspicious, with many heavily Democratic counties (at least by registration) voting for Bush. However, I did some further investigation to compare the numbers to the 2000 election, and in fact those same counties also voted heavily for Bush four years ago. The turnout then was not as high, but nevertheless it proves either that:

a) the fraud has been going on for awhile


b) registered Democrats in small, rural counties in Florida often vote Republican

Probably the second choice, but it still seems strange to me.

Thanks for your terrific site, and I'm sorry your predictions (the ones including undecideds and turnout) didn't come to pass.