A Proud Member of the Reality-Based Community

Acknowledgements

SEARCH eRiposte!

For the eRiposte Election 2004 home page, click here. For a database on vote fraud/suppression in Election 2004, click here.

[Note: In Part B of Election 2004 coverage, I provide a somewhat detailed post-mortem of this election. To send me feedback please email feedback-at-eriposte-dot-com (I can't promise a reply to every email I receive.)]

SECTIONS

I: Revised Exit Poll Shows Inconsistent Results - and Gets Revised Again!

II: Voting irregularitiessuppression//fraud/"glitches" and suspicious results

II-1: National

II-2: Florida: Select Optical Scan Machine Counties

II-3: Florida: Select E-Vote Counties

II-10: Ohio

III: Exit Polls v. Voting Results

IV: Was this election won by Bush because of vote fraud?  Click here for my position on this burning question

Part I: REVISED EXIT POLL SHOWS INCONSISTENT RESULTS AND GETS REVISED AGAIN!

One of the reasons the Kerry camp was disappointed with the Election 2004 results is that in two key states, Florida and Ohio, the exit polls initially showed Kerry leading. Now, one may attribute this discrepancy to the error in the "initial" exit polls themselves - and indeed, interestingly, AP decided to "revise" them the night of 11/2/04 (see this post comparing Ohio before and after, for example) to get them better in sync with the reports from the actual vote counts [very odd indeed]. 

When I first started to look at this, there were many things in the revised exit polls that raised questions. But one thing that caught my eye in particular was an obvious one - how Gore/Bush/Other 2000 voters voted this time. 

If you look at the exit poll results posted as of 11/3/04 ~ 7 am Pacific (I thought it was final but alas, it was not to be!) in terms of the national popular vote, here is what you got. 
Males were 46% of the electorate and Bush got 54% of the male vote and Kerry got 45%
Females were 54% of the electorate and Bush got 47% of the female vote and Kerry got 52%
Based on that:
Bush's share of the popular vote = 50.2% = 0.46*0.54+0.54*0.47
Kerry's share of the popular vote = 48.8% = 0.46*0.45+0.54*0.52

Thus, as of ~7 am on 11/3/04, CNN showed the actual (ostensibly) national tallied vote percentages as follows:
Bush 51%
Kerry 48%

These results were qualitatively consistent with the tallied vote results and quantitatively close to the tallied vote percentages. But when I looked at the Bush/Gore/Other vote from 2000, something looked strange - Kerry won the popular vote according to their own "revised" exit polls! 

ALERT!! 

The numbers in the table below (Table 1) were based on the data CNN had posted on their website as of ~ 8 am on 11/3/04 (click here for a screenshot - colors do not reproduce as in the original). 

I checked back there on 11/7/04 at ~ 7:45 pm and discovered CNN had changed the data again (Table 2) - click here for a screenshot! 

So, I have two versions of the revised exit poll now, with different results (Kerry led per Table 1, Bush leads per Table 2)! 

Why did CNN update the results and when did they do it - and why are exit poll numbers being updated long after the fact? Details below...

2000 vs. 2004 comparison on Bush/Gore/Nader vs. Bush/Kerry/Nader

TABLE 1: Numbers on CNN's website as of ~ 8 am on 11/3/04 [originally posted on 11/3/04]

[Note that the numbers in the first column (Presidential Vote in 2000) add up to 101%, which is fine since there is some rounding error involved].

This is a very important table. Why? 

1. What is interesting is that the popular vote totals according to this table are as follows:
Bush 49.80% [=0.17*0.44+0.38*0.1+0.42*0.9+0.04*0.18]
Kerry 49.88% [=0.17*0.54+0.38*0.9+0.42*0.09+0.04*0.68]
Nader 0.005% [=0.17*0.01+0.38*0+0.42*0+0.04*0.08]

In other words, according to this table Kerry leads the popular vote percentage, contradicting the popular vote percentage mentioned at the top of the exit poll page! Even if you adjust for rounding errors it is hard to explain how Bush could be 3% higher in the overall tally, when this table shows Bush is slightly behind Kerry. Something does not add up.

2. Let's also explore the implications of the data above - from the revised exit poll (Table 1):

(a) In Election 2004, voters who did not vote in 2000 were a whopping 17% of the electorate and they went for Kerry by a big margin (10%).

(b) Gore 2000 voters went for Bush 2004 by a margin of 1% more than Bush 2000 voters went for Kerry 2004.

(c) Those who voted neither for Bush nor Gore in 2000 went for Kerry by a massive margin (50%), and only a small fraction of the "Other" voters in 2000 went for Nader.

(d) There were 4% more Bush 2000 voters this time around compared to Gore 2000 voters! [Note that this means a huge percentage of Gore 2000 voters didn't bother to show up at the polls in 2004!] And in spite of this, according to Table 1, Kerry won the popular vote (because of the large Kerry gain from the new/other voters).

We need to keep in mind the following additional factors this year:

• The Democratic base was strongly mobilized this year, with an unprecedented GOTV operation that - at least before the election - showed significantly higher Democratic voter registrations than Republican in key battleground states

• We had an incumbent President who almost never got a re-elect rating above 50% in virtually every poll for months and months, and Republican moderates who were deeply unhappy with their candidate

There are a lot more factors/questions which I am skipping right now (click here for one example). But something smells fishy here and I don't know what. I am not trying to get into tinfoil hat territory here but someone should sit down and comb through all the data very carefully. Someone should look at the initial vs. revised exit polls very carefully state by state and compare them to the final tallied results. Someone should look at the tallied results to see if the demographics seem reasonable - and compare them to all the polls taken in the few days before the election to see where the big differences are. Another question: How many states showed a shift from Bush to Kerry or vice-versa when you compare (original) exit polls to the tallied votes? Which states are these? Is there anything in the models for those states or the circumstances of polling in those states that can rationally and reasonably explain it? Questions, questions.

TABLE 2: Numbers on CNN's website as of ~ 7:45 pm on 11/7/04 - with changes IN RED [posted on 11/7/04]

With these "latest" "revised" exit poll numbers (when will they stop revising them?), now:

Bush 51.10% [=0.17*0.45+0.37*0.1+0.43*0.91+0.03*0.21]
Kerry 48.48% [=0.17*0.54+0.37*0.9+0.43*0.09+0.03*0.71]
Nader 0.003% [=0.17*0.01+0.37*0+0.43*0+0.03*0.03]

How convenient! The national popular vote has also been "updated" in the "latest" exit poll as follows:
Males were 46% of the electorate and Bush got 55% of the male vote and Kerry got 44%
Females were 54% of the electorate and Bush got 48% of the female vote and Kerry got 51%
Based on that:
Bush's share of the popular vote = 51.2% = 0.46*0.55+0.54*0.48
Kerry's share of the popular vote = 47.8% = 0.46*0.44+0.54*0.51

Very interesting indeed.

Even with this change I find two things rather singular here:

1. Even though Gore won the popular vote in 2000 and there were more Gore voters than Bush voters in 2000, according to this "latest" version of the "revised" exit poll in 2004, there were 5% more Bush 2000 voters in the 2004 electorate than Gore 2000 voters! According to this, a huge number of Gore 2000 voters (several %) never bothered to show up at the polls in 2004!! This makes little or no sense to me - and I find it utterly implausible considering how strongly the Democratic base was motivated this year. 

2. The CW now is that Bush/Rove got many more Republican base voters out to the polls this year than they were able to in 2000 and that this contributed to Bush's win in 2004. Yet, the exit poll results show that those WHO DID NOT VOTE in 2000 favored Kerry by a 9% margin! 

The repeated revisions of the exit polls as well as the Bush/Gore data makes me even more suspicious of what is going on with the manipulation of the exit poll results. I hope there is a proper investigation of how the data is being manipulated and presented. 

[Reader JP sent in this link in an email that discusses the same issue and she concludes that:

So, of the 105 million who voted in 2000, only 98 million came to the polls this time. And, coincidentally, *all* of the missing 7 million were Gore voters (Gore got 51 million votes in 2000).
...
Conclusion: Either the original exit polls were right, because the tweaking leads us to absurd conclusions *or* 3.5 million people who switched from Gore to Bush wanted to lie about it to pollsters (extrapolating from the sample, obviously).

]

NOTE: I have no problem accepting a legitimate Bush election, but there are way too many things with the data and the results so far that raise serious questions. 

NOTE II: See Section III for more on the exit polls and their accuracy

Part II: VOTING IRREGULARITIES/SUPPRESSION/FRAUD/"GLITCHES" AND SUSPICIOUS RESULTS

II-1: National

The eRiposte sub-domain Vote Watch 2004 catalogs vote/election fraud, vote suppression, voting irregularities, and voter intimidation in Election 2004. 

Brad Blog has also been cataloguing emerging news on discoveries in different states that show the unreliability of voting machines (not necessarily just touch screen but also optical scan).

Some example cases (more on the websites above):

• NC PAPER: 11,238 More E-Votes Recorded Than Actually Cast

• 70,000 Votes 'Discovered' in Broward County!

• OH Precinct: 638 Votes Cast, 4258 of them for Bush! What a Difference a Day Makes.

• 10,000 Extra Votes Added in Nebraska County

• 50,000 Votes Lost in Dem Indiana County!

II-2: Florida: Select Optical Scan Machine Counties

Through BradBlog I came across this report.

Kathy Dopp has been crunching hard election result numbers and a strange -- yet very clear pattern -- is emerging.

Using the very rough gauge of comparing Republican and Democrat registrations county by county to the actual final voting results, she finds that the percentage change of "Expected Votes" versus "Actual Results" are fairly consistent in counties that used E-Voting machines.

But in counties that used Optical Scan Voting the "Expected Votes" versus "Actual Results" are wildly discordant to say the least.

Kathy Dopp's page (at the bottom) has a number of interesting charts.

Reader RP sent in some useful tables [corrected at ~7:40 pm Pacific on 11/7/04 to account for some calculation error] that takes Dopp's results and summarizes them in easy-to-understand form. This data is stated to be that with 98.6% of all votes counted (I've truncated the top of the image). Do review it (again, I have not independently verified the data myself and I'll be happy to post a correction if it is shown to be a data error - but I would like to put it out on the web so people can review and analyze the results themselves). 

These tables, on the face of it may suggest some irregularities. However, I also provide a link to a counter-argument from Bryan Pfaffenberger below these tables and rebuttals to that!), since he says "it is extremely misleading (and contrary to accepted practice) to define "expected results" in terms of the percentage of registered Republicans and Democrats in an election district. "Expected results" should be defined in terms of the way the district voted in the last general election."

[Note added on 11/20/04: Reader Susan Goya sent in an article that maps the counties that went to Bush and Kerry and how they should have gone if voting registration determined the final vote. I have posted it to allow readers/independent analysts to review it for accuracy/comment.]

Note added 7 pm Pacific, 11/8/04: I am only showing the most recent table that reader RP sent out and I removed the two original tables per RP's request - since the third table, which is the one shown below, essentially captures what he wanted to point out.

Bryan Pfaffenberger sent in an email saying that he is very skeptical of the claims that the above data prove fraud in any way. He addresses the data in his website, here - and here are some snippets (I have my comments in response below these snippets):

Those who dismiss the allegations of voter fraud note that the suspicious pattern is found in small, predominantly rural counties, many of which are in the Panhandle, in which Bush is known to have made significant gains. (The two counties mentioned above, Baker and Dixie Counties, are in the Panhandle.) According to Cornell University political scientist Walter Mebane, many (if not most) of the putative Democrats in these rural counties are Democrats in name only -- they just haven't gotten around to changing their registration. The seemingly disproportionate vote for Bush, in short, is part of a larger trend, seen throughout the rural and suburban South and personified by Zell Miller, of Democrats deserting their party.

Moreover, there's nothing new about this pattern. In 2000, Gore received only 2,392 votes from Baker County voters, 83.3 percent of whom registered as Democrats; Bush received 5,610. In Dixie County, the 2000 numbers are even more striking: even though 85.9% of voters registered as Democrats, Gore received only 1,826 votes while Bush received 2,697. In short, a tectonic shift is going on in Deep South voting patterns, and it started four years ago.

If there's an association between the "Bush up/Kerry down" pattern and optical scanners, therefore, it's because the smaller, rural counties -- including most of those in the Panhandle -- tend to use these machines; only the large, urban counties use e-voting machines.

But wait just a minute. A widely-linked Web page located at an activist Web site (ustogether.org) argues that, even after taking the rural/urban split into account, the numbers are still suspicious. One of the site authors (Elizabeth Liddle) tried to control for the rural-urban contrast by focusing on 26 mid-sized Florida counties (those with between 80,000 and 500,000 registered voters), in which both types of voting technologies were used. The point of this analysis was to reduce or eliminate the rural/urban factor as an explanatory variable.

...

Here's what this chart says, in plain English:

• In the counties that used electronic voting machines, both parties surpassed their 2000 turnout, but Democrats beat the Republicans.
  In the counties that used optical mark-sense voting machines, the Democrats surpassed their 2000 turnout by a small percentage. But the Republican increase is massive.

Having ruled out the rural/urban split, Liddle contends, voting machine type strongly correlates with the voting patterns.

Does this analysis prove that election fraud actually occurred? No, and for two reasons:

• Liddle does not disclose which counties she selected, so it is impossible to evaluate how successfully she controlled for geographic factors.
• The pattern could be explained by a third, unknown variable.

I think the pattern is indeed attributable to a third variable -- namely, the explosive growth of Republican voter turnout, not just in the Panhandle or rural areas, but throughout all of Florida except the major cities. Suburban as well as rural Democrats abandoned their party in record numbers -- thanks, in part, to a Democratic tactical mistake: while Democrats focused on cities and legal challenges, Republicans focused their efforts on fast-growing suburban counties.

A telling example is Pasco County, situated north of Tampa. Formerly dominated by Democratic-leaning retirees, the county is now home to growing numbers of suburban, middle-class families, for which Bush's campaign message resonated far more than Kerry's. Like many other fast-growing Florida counties, Pasco is very well characterized by precisely the <demographics that gave Bush huge gains across the American heartland:

Bush fared well among gun owners ( 30 over Kerry), weekly churchgoers ( 26), Protestants ( 24), military veterans ( 20), married people ( 20), non-Hispanic whites ( 14), Southerners ( 14), and men ( 12).

In contrast, Kerry appealed mainly to

blacks (+86 [over Bush]), those residing in union households (+34), unmarried women (+28, and +10 among unmarried men), 18- to 29-year-olds (+20), those who seldom or never attend church (+20), Easterners (+16), and urban residents (+12).

I think these figures explain quite plainly what happened in Florida on November 2nd. To be sure, I know some Democrats are still convinced that the exit poll numbers were indeed valid, which means that Bush could have won Florida only if hundreds of thousands of Kerry votes disappeared. CEO Walter Mitowsky of the exit polling organization (National Election Pool) explains that that Kerry voters were more eager to talk to interviewers than Bush voters. Pollster Mark Blumenthal says that, if Mitowsky's explanation is true, it is truly extraordinary:

It is the first time in his exit polls and one of the first times I can remember for any political survey, that those who refused to participate in the survey were different enough (more supportive of George Bush) than those who were interviewed to make a material difference in the overall results.

But there is at least one precedent (Crewe 1992). In the 1992 British general election, those who voted for the Conservatives were markedly less forthcoming to interviewers than Labour voters, resulting in the single greatest exit poll failure on the record.

...

One conclusion seems inescapable: Our country's election system is in need of reform. We need:

• Uniform, clearly stated standards for registration lists, voting technologies and the distribution of voting machines; and
• Non-partisan election administrators

Let me close with a personal note. Writing this essay has been a painful process for me -- frankly, I wanted desperately to believe that Floridians really did go out and vote for John Kerry in massive numbers.

Since Bryan's response, there has been more back-and-forth, with other critiques of Kathy Dopp's conclusions (including in the New York Times).

To make it simple, let me refer readers to the rebuttals from Kathy Dopp's group (which I became aware of via reader Brian Julin who has a useful catalog on links on his website).

Response to Mebane et al.: Discrepant Results appear in Mid-State, Mid-Size Counties (Kathy Dopp's response)

Cross-party voting in Florida seems to depend on the local voting technology (also from Dopp's website)

Response to the [Caltech/MIT] VTP criticism of the study by Dopp/Liddle

Response to MIT/Caltech (by David Dodge)

Additionally, Sam Parry at Consortium News joins the fray with this rebuttal report (bold text in the paragraphs is my emphasis) which I am including because it goes to the heart of the matter:

The Washington Post and the big media have spoken: Questions about Nov. 2 voting irregularities and George W. Bush’s unusual vote tallies are just the ravings of Internet conspiracy theorists.

In a Nov. 11 story on A2, the Post gave the back of its hand to our story about Bush’s statistically improbable vote totals in Florida and elsewhere. While agreeing with our analysis that Bush pulled off the difficult task of winning more votes in Florida than the number of registered Republicans, the Post accuses us of overlooking the obvious explanation that many independents, “Dixiecrats” and other Democrats voted for Bush.

Mocking us as “spreadsheet-wielding conspiracy theorists,” Post reporters Manuel Roig-Franzia and Dan Keating signaled their determination to put questions about Bush’s victory outside the bounds of responsible debate. Yet, if they hadn’t been so set in this agenda, they might have avoided sloppy mistakes and untrue assertions.

In an example of their slipshod reporting, Roig-Franzia and Keating state that we focused our data analysis on rural counties in Florida. They suggest that Bush’s gains in these rural counties might be explained by the greater appeal of son-of-the-South Al Gore in 2000 than Bostonian John Kerry in 2004.

But we didn’t focus on rural counties in Florida. Rather we looked at the vote tallies statewide and zeroed in on Bush’s performance in the larger, more metropolitan counties of southern and central Florida, where Bush got the vast majority of his new votes over his state totals in 2000.

It was in these large counties where Bush’s new totals compared most surprisingly with new voter registration because Democrats did a much better job in many of these counties of registering new voters. In other words, Bush outperformed Kerry among a relatively smaller ratio of Republicans to Democrats in many of these counties.

Also undermining the Post’s claims, Kerry actually improved on Gore’s total in the smallest 20 counties in Florida by 5,618 votes -- 50,883 votes for Kerry vs. 45,265 for Gore, a 12.5% increase. So, even the Post’s notion that Gore’s Southern heritage made him more attractive to rural Floridians doesn’t fit with the actual results.

Simple Question

We began our analysis of the vote totals with one simple question: Where did Bush earn his new votes? Since one of every nine new Bush voters nationwide came from Florida, we thought this battleground state was a good place to examine county-by-county tallies.

We also didn’t go into the analysis expecting to find statistical oddities. We were open to the possibility that Bush’s totals might have fit within statistical norms.

What we found, however, led us to report that Bush’s vote tallies were statistically improbable – though not impossible. Contrary to the Post’s claim, we did take into account the Dixiecrat element, which is why we didn’t focus on the Bush totals from Florida’s panhandle or the smaller, rural counties.

Our analysis found that of the 13 Florida counties where Bush’s vote total exceeded the number of registered Republicans for the first time, only two were counties with fewer than 100,000 registered voters. In 2000, Bush’s vote total exceeded the number of registered Republicans in 34 counties – not 32 as the Post inaccurately reported – but in 2004, this total shot up to 47 counties.

Rather than a rural surge of support, Bush actually earned more than seven out of 10 new votes in the 20 largest counties in Florida. Many of these counties are either Democratic strongholds – such as Miami-Dade, Broward, and Palm Beach – or they are swing counties, such as Orange, Hillsborough, and Duval.

Many of these large counties saw substantially more newly registered Democrats than Republicans. For example, in Orange County, a swing county home to Orlando, Democrats registered twice as many new voters than Republicans in the years since 2000. In Palm Beach and Broward combined, Democrats registered 111,000 new voters compared with fewer than 20,000 new Republicans.

However, in these three counties combined, Bush turned out about 10,000 more new voters than Kerry, a feat made all the more remarkable given that Kerry improved Democratic turnout in these counties by 21 percent.

No Landslide

Historically, increases like those Bush registered throughout Florida and across much of the country occur when there are huge swings in voting patterns caused by national landslides.

In 1972, for instance, Richard Nixon won millions of votes from Democrats who two elections earlier had supported Lyndon Johnson. But in 2004, the Democratic ticket didn’t suffer a hemorrhage of votes, actually turning out about 5 million more voters nationwide than in 2000.

Nor was that the case in Florida. In county after county in Florida, Bush achieved statistically stunning gains even as Kerry more than held his own. Bush earned nearly 35 percent more votes statewide than he did in 2000, which was already a huge surge for Bush over Bob Dole’s 1996 Florida turnout.

Contrary to assertions in the flawed Post article, the most surprising numbers actually don’t come from small rural counties in the state, but rather from large counties, including Orange county (mentioned above), Hillsborough (Tampa), Brevard (Cape Canaveral), Duval (Jacksonville), Polk (next to Orange county), and heavily Democratic Leon (Tallahassee) and Alachua (Gainesville). These are not tiny Dixiecrat counties with longtime registered Democrats who haven’t voted Democratic in years.

Rather, these seven counties have large, diverse populations that collectively saw Bush turn out 1,025,493 votes, exceeding the 946,420 registered Republicans. In these counties, Bush turned out nearly twice as many new votes than the number of newly registered Republicans. In these same counties, Kerry got more than 200,000 new votes, meaning that Bush’s tally can’t be attributed to crossover Democrats.

While Bush’s totals are not statistically impossible, they do raise eyebrows. Our question was: where did these gains come from? We are not claiming that the surprising numbers are evidence of fraud, but we do believe the tallies deserve an honest and independent review.

It also should be the job of journalists to probe questions as significant as the integrity of the U.S. voting system, not to simply belittle those who raise legitimate questions. The fact that Internet journals and blogs are doing more to examine these concerns than wealthy news organizations like the Washington Post is another indictment of the nation’s mainstream press.

At this point, I have to say I don't know what to think...

What I am happy about is that there is good debate and detailed analysis going on - this is critical in a democracy. I am going to wait and see how this transpires.  

Three points that I believe need further exploration:

1. Is there any reason to believe that Bush-supporters were less likely to speak to exit pollsters ONLY in those states where the exit polls predicted a result that was divergent from the final voting result? (I find it dubious that Democrats who waited for hours and hours on end would have been that motivated to spend more time talking to pollsters at the end). This is related to the question of exit poll accuracy by state. Why is it that polls in some states were close to the final result and those in others were not? (See Section III) 
[P.S. I wonder if the seemingly odd outcome that 5% more Bush 2000 voters ostensibly voted in 2004 than Gore 2004 voters arose in the massaged exit polls because of the exit pollsters trying to make up for their claim that Bush voters were less enthusiastic about answering questions this time. Yet, I still cannot fathom how 5% or more Gore 2000 voters simply dropped out of the electorate this year.] 

2. As I showed in Section I, the national level exit poll results, even after having been massaged again and again to provide Bush the popular vote, provide conclusions about who voted that directly contradict conventional wisdom. So, I am not sure how to interpret the conclusions on the demographics of who voted for Bush and who did not.
[P.S. I am also not convinced about an explanation that uses an experience in the United Kingdom where a lot of things could be different. I would be more convinced if we could explain these results using American voting history].

3. With the rebuttals from Kathy Dopp and co-workers and from Sam Parry, there is clearly more to analyze before calling it quits on this topic.

II-3: Florida: Select E-Vote Counties

The Election Verification Project has pointed out this: "On November 2 more voters cast paperless ballots than ever before in a U.S. election: roughly 30 percent compared to 12 percent four years ago. The Election Verification Project is not questioning the legitimacy of the presidential election results. Instead, we are questioning a voting process in which 30 percent of the ballots cannot be verified and, of the remaining 70 percent, most will not be verified."

Given that, we need to note something that Markusd at Dailykos has posted an alert on - a UC-Berkeley Study that questions the accuracy of electronic voting machine results in some highly Democratic counties in Florida. [Before someone gets confused about the Optical Scan study and this E-voting Study, read this note at the end of this section!].

The study's summary is here:

The Effect of Electronic Voting Machines on Change in Support for Bush in the 2004 Florida Elections

Summary:

- Irregularities associated with electronic voting machines may have awarded 130,000 excess votes or more to President George W. Bush in Florida.

- Compared to counties with paper ballots, counties with electronic voting machines were significantly more likely to show increases in support for President Bush between 2000 and 2004. This effect cannot be explained by differences between counties in income, number of voters, change in voter turnout, or size of Hispanic/Latino population.

- In Broward County alone, President Bush appears to have received approximately 72,000 excess votes.

- We can be 99.9% sure that these effects are not attributable to chance.

Details:

Because many factors impact voting results, statistical tools are necessary to see the effect of touch-screen voting. Multiple- regression analysis is a statistical technique widely used in the social and physical sciences to distinguish the individual effects of many variables.

This multiple-regression analysis takes account of the following variables by county: - number of voters - median income - Hispanic population - change in voter turnout between 2000 and 2004 - support for President Bush in 2000 election - support for Dole in 1996 election

When one controls for these factors, the association between electronic voting and increased support for President Bush is impossible to overlook. The data show with 99.0% certainty that a county’s use of electronic voting is associated with a disproportionate increase in votes for President Bush.

The data used in this study come from CNN.com, the 2000 US Census, the Florida Department of State, and the Verified Voting Foundation – all publicly available sources. This study was carried out by a group of doctoral students in the UC Berkeley sociology department in collaboration with Professor Michael Hout, a member of the National Academy of Sciences and the UC Berkeley Survey Research Center.

Wired News reported this as follows:

Electronic voting machines in Florida may have awarded George W. Bush up to 260,000 more votes than he should have received, according to statistical analysis conducted by University of California, Berkeley graduate students and a professor, who released a study on Thursday.

The researchers likened their report to a beeping smoke alarm and called on Florida officials to examine the data and the voting systems in counties that used touch-screen voting machines to provide an explanation for the anomalies. The researchers examined the same numbers and variables in Ohio, but found no discrepancies there.

Their aim in releasing the report, the researchers said, was not to attack the results of the 2004 election in Florida, where Bush won by 350,000 votes, but to prompt election officials and the public to examine the e-voting systems and address the fact that there is no way to conduct a meaningful recount on the paperless machines.

The analysis -- which hasn't been formally peer-reviewed, but was examined by seven professors -- showed a discrepancy in the number of votes Bush received in counties that used the touch-screen machines and counties that used other types of voting equipment. The researchers examined numerous variables that might have affected the vote outcome. These included the number of voters, their median income, racial and age makeup and the change in voter turnout between the 2000 and 2004 elections. Using this information, they examined election results for the Republican and Democratic presidential candidates in the state in 1996, 2000 and 2004 to see how support for those candidates and parties measured over eight years in Florida's 67 counties.

They discovered that in the 15 counties using touch-screen voting systems, the number of votes granted to Bush far exceeded the number of votes Bush should have received -- given all of the other variables -- while the number of votes that Bush received in counties using other types of voting equipment lined up perfectly with what the variables would have predicted for those counties.

The total number of excessive votes ranged between 130,000 and 260,000, depending on what kind of problem caused the excess votes. The counties most affected by the anomaly were heavily Democratic.

Prof. Sam Wang of Princeton University (via Katerina at Dailykos) believes that the Berkeley study holds water. For more than one reason:

There's an interesting report from a statistical analysis group at Berkeley about Florida voting patterns. After reading it over, I can't find any big problems with it. It is the best study yet done on the question of possible irregularities. This does not mean that irregularities occurred. However, it does mean that something unusual happened in Florida. Also, note that their estimate of voting discrepancy (130,000 to 260,000) is similar to my estimate based on pre-election polls (270,000).

What they have done is use voting patterns by country from 2000 and 1996, income by county, total population, and Hispanic population to try to explain voting patterns in 2004. These variables encompass many arguments made for why this year's results are surprising. For instance, it has been argued that rural voters liked Bush better this year. But since the Berkeley group used county population as a variable, this should stand as a proxy for population density. Note that because they are using regression analysis, they are not claiming that Bush should have gotten the same number of votes as previous years. This would not be true: we all know that Clinton-Dole 1996 turned out very differently from Bush-Gore 2000. Instead, they are using those variables to help account for variations among counties, for instance if one county tends to be more Republican than another.

Their analysis indicates that even when all these variables are accounted for, a significant difference remains between counties that used electronic voting and counties that used optical scanning or paper ballots (see also a previous analysis by Jeff Chambers).

But - in which group did something strange happen? They say electronic voting, but one could also imagine that optical scan and paper ballots were both off. This is less parsimonious, but can one rule it out? Maybe.

An additional piece of evidence is available: pre-election polls. In my own analysis (see below), I find that pre-election polls predicted a margin of Bush over Kerry by 1.4%, or about 105,000 votes. Instead the reported result was Bush over Kerry by 5.0%, or about 375,000 votes. The discrepancy is 270,000 votes, which is consistent with their model of misattribution, i.e. Kerry votes being counted as Bush votes, which predicts a discrepancy of 262,000 votes. The other model, "ghost votes," is also possible, though it only predicts a 131,000 vote discrepancy. Therefore, assuming that electronic voting had problems relative to other methods leads to an estimated discrepancy that is very similar to one made using pre-election polls.

A more innocuous explanation may exist, but their work (and my analysis) put bounds on what that explanation might be. A change in sentiments among rural voters seems unlikely, since the Berkeley team found no anomaly in Ohio (and also used county population as a variable). Another possibility is the fall hurricanes, since these did not happen in other states; however, how this would work is unclear. In any event, this story is not over.

Tuesday, November 16, 11:45PM: Below is a graph of all state polls in Florida, Ohio and Pennsylvania in the weeks preceding the election. (Comparisons for other states are available here.) The gray band indicates ±1 SEM surrounding the average for the last seven days; this is about half the width of the 95% confidence band. The arrowheads indicate the final announced outcomes. Ohio and Pennsylvania polls were consistent with the final outcome, but Florida polls were not consistent.

Pennsylvania polls indicated a Kerry win by 2.1 ± 0.7%; the final result was Kerry by 2.2%.

Ohio polls indicated a Bush win by 1.0 ± 0.7%; the final result was Bush by 2.5%. This could be accounted for by the trend toward Bush in the last five days.

Florida polls indicated a Bush win by 1.4 ± 0.9%; the final result was Bush by 5.0%. This final result is off by 4 SEM. It is also in the same direction as the claims of voting fraud made based on county-level data.

Although not definitive, these data are at least consistent with the suggestion that something unusual happened in Florida, either in the 33 polls conducted in the final weeks of the campaign, or in actual voting.

Via reader radtimes, this report in the Oakland Tribune indicates another group has independently verified the results:

The UC Berkeley report has not been peer reviewed, but a reputable MIT political scientist succeeded in replicating the analysis Thursday at the request of the Oakland Tribune and The Associated Press. He said an investigation is warranted.

"There is an interesting pattern here that I hope someone looks into," said MIT arts and social sciences Dean Charles Stewart III, a researcher in the MIT-Caltech Voting Technology Project.

Stewart isn't convinced the problem is electronic voting. It could be absentee voting or some quirk of election administration. But whatever the problem, it didn't show up in counties using optical scanning machines. Rather than offer evidence of fraud or voting problems, the UC Berkeley study infers they exist mathematically.

Frustrated at the lowbrow, data-poor nature of allegations of election fraud flooding the Internet, three Berkeley grad students decided to apply the tools of first-year statistics class.

"We decided, well, you might as well test it properly instead of sitting around speculating," said first-year sociology grad student Laura Mangels. She and two colleagues downloaded voting and demographic data, ran them through statistics software and in the first night had results that produced a collective "Wow" among the students, she said.

They shopped their results to faculty and finally to Hout, a well-known skeptic who is chairman of the university's graduate sociology and demography group.

"Seven professors later, nobody's been able to poke a hole in our model," Mangels said. "Our results still hold up."

Hout agreed. "Something went awry with the voting in Florida."

In terms of a partial rebuttal, Kevin Drum has a post reporting on others who looked at the same data and reached a less broad conclusion:

However, both Kieran Healy and Andrew Gelman are skeptical that Hout's results stand up. The chart on the right, adapted from Gelman's site, shows why.

The chart shows the size of the vote swing toward Bush in all Florida counties, and the basic pattern is simple: the more Republican a county was in 2000, the more heavily their vote swung even further toward Bush in 2004. At the bottom left are counties that are strongly Democratic, and their swing to Bush was actually negative compared to 2000. At the top right are counties that are heavily Republican, and their swing toward Bush was eight percentage points or more, far higher than the statewide average swing of 2.5 points.

Counties in red used electronic voting and counties in black didn't. Nearly all the counties fall within the two black lines, and there doesn't seem to be much difference between them. Both red and black circles show the same basic pattern, which means there's no special reason to think that anything funny was going on in counties with electronic voting.

In fact, it turns out that Hout's entire result is due to only two outliers: Broward County and Palm Beach County. This suggests several things:

• There was almost certainly not any systemic fraud. If there were, it would have showed up in more than just two counties.

• The results in Broward and Palm Beach are unusual, but it's hard to draw any conclusion from just two anomolies. As Kieran says, "it seems more likely that these results show the Republican Party Machine was really, really well-organized in Palm Beach and Broward, and they were able to mobilize their vote better than the Democrats."

• Anyone who wants to continue investigating possible fraud in Florida anyway should focus on Broward and Palm Beach.

CONFUSED BETWEEN THE KATHY DOPP STUDY AND THE UC-BERKELEY STUDY?

Brian Julin alerted me to the possible confusion that might arise in the media and supposed debunkers of this latest study. As he points out (bold text is my emphasis):

OK, the Berkeley study is about to hit. And this could go either way, media-coverage speaking. That's because the Berkeley study would appear at first glance to say the exact opposite of what Dopp's study says. The Berkeley study indicates that e-voting added to Bush's total while Dopp's study blames opscan.

Again, this is a matter of which counties were included, but this time on the other end of the scale. Dopp excluded counties over 500K. The Berkeley study says:

The impact of e-voting was not uniform, however. Its impact was 
proportional to the Democratic support in the county, i.e., it was 
especially large in Broward, Palm Beach, and Miami-Dade. The evidence 
for this is the statistical significance of terms in our model that 
gauge the average impact of e-voting across Florida's 67 counties and 
statistical interaction effects that gauge its larger-than-average 
effect in counties where Vice President Gore did the best in 2000 
and slightly negative effect in the counties where Mr. Bush did the 
best in 2000.

...and those three counties plus two other large democratic ones (Pinellus and HillsBorough), plus the republican-voting smaller counties of Nassau and Sumpter, were specifically excluded from Dopp's analysis. The "slightly negative effect" noted would, if anything, support Dopp's work, but I don't want to get into a math lecture.

As in the earlier case, let the debate begin anew about Florida...let's get all the data out to independent analysts and understand what is going on.

II-10: Ohio

Ohio has been a royal mess in this Election, with massive vote suppression and numerous irregularities. Details are available in the Vote Watch 2004 Ohio page.

Votersunite has an excellent single page graphic that shows some of the reasons why the problems in Ohio need to be investigated and why Ohio deserves a recount. For convenience I have captured their chart in JPEG form and reproduced it below:

Another recent example adding to the Ohio irregularities - rotten Denmark points a Dayton Daily News report (bold text is his emphasis):

Two precincts had high undercounts, analysis shows
By Ken McCall and Jim Bebbington

DAYTON | Two Montgomery County precincts had extraordinarily high numbers of ballots cast Nov. 2 with no presidential vote counted, and the county's overall rates of such undercounts were highest where Democratic hopeful John Kerry did best.

Undercounts are ballots that do not register a vote for a particular race, in this case for president. Two precincts — one in Kettering and another in Washington Twp. — had undercounts of more than 25 percent, according to a Dayton Daily News analysis of the county's unofficial results.

Overall in Montgomery County, 5,693 or 2 percent of the ballots cast registered no valid vote for president.

As predicted by political scientists, who say the poor and less-educated are more likely to have problems with punch card voting, the rate of so-called undercounted presidential ballots was higher in Democratic areas of the county than in Republican strongholds.

The undercount amounted to 2.8 percent of the ballots in the 231 precincts that supported Kerry, but only 1.6 percent of those cast in the 354 precincts that supported President Bush.

Across the state on Nov. 2, counties that used punch-card voting, as Montgomery County did, had a higher rate of undercounted ballots than counties that used optical scanning technology or electronic voting machines, which had the lowest undercount.

With punch cards, undercounts can occur when a voter:

•Inadvertently votes for two candidates in the same race.

•Decides not to vote in the race.

•Does not sufficiently puncture the punch card to eliminate a "hanging chad." Hanging chads can make it impossible for machines to read the punch cards.

The highest undercount rate in Montgomery County was in precinct Washington X, around Paragon Road and Spring Valley Pike in Washington Twp.

In the precinct, 168 or 27.5 percent of the 611 ballots cast did not have a good presidential vote. That was followed closely by Kettering 3-A, near Stroop Road and Far Hills Avenue, where 121 or 27.3 percent of the 444 ballots cast were undercounted.

Both of those precincts supported Bush, as did seven of the 10 precincts with the highest rate of undercounted presidential ballots. That's despite the county's overall trend, in which precincts where Kerry did well tended to have above-average undercounts, while precincts where Bush won had lower-than-average undercounts.

County elections officials said they have no reports of any problems at either Washington X or Kettering 3-A. The punch-card voting stands, checked Wednesday using demonstration ballots, appeared to work appropriately.

The presiding judge of Washington X, Shirley Wightman, a 40-year veteran of working polling places, said voters in her precinct encountered no problems.

"We checked the machines periodically and I could see nothing wrong with them," she said.

Wightman said turnout was high that day and there were 16 provisional voters at the precinct, a higher-than-normal number. But those provisionals do not account for the under-votes and won't be tabulated until next Monday, after officials confirm the registration of those voters.

One voter reported having trouble pushing her ballot into the slot in the voting machine, but she had not pushed the card in far enough and a poll worker helped her, Wightman said.

"Other than that things went pretty smooth," she said.

Two Washington X voters said they checked for hanging chads on their ballots before turning them in and found none.

"I personally checked mine and it punched the number I wanted," said Heather Baarlaer of Washington Twp.

Rates that high show something must have gone wrong, said Larry J. Sabato, a political scientist from the University of Virginia. Undercounts during presidential elections are typically between 1 percent and 2 percent, he said.

"It is very difficult to believe that a quarter of the people would not vote for president, especially in a year like this," Sabato said. "If I were the election officers in those areas I would be doing some very extensive checks of those machines."

Moraine had the highest presidential undercount of all the county's jurisdictions: 5.6 percent of the 2,557 votes cast in the city's seven precincts had no valid presidential vote. It was followed by Germantown with 3.6 percent undercount, Trotwood with 3.1 percent and Dayton with 2.8 percent. Both Moraine and Germantown supported Bush, with margins of 2 percent and 34 percent respectively, while Trotwood and Dayton went heavily for Kerry by margins of 60 percent and 45 percent.

Part III: EXIT POLLS v. VOTING RESULTS 

Brad Blog has a post comparing the exit poll results in states with complete paper trails and those that do not. I have not verified these results myself but I am providing a link here to those who are interested and want to explore this further for accuracy. The results seem to suggest that when paper trails existed, the exit poll results closely matched the final results. In the other case, the final voting results deviated consistently towards Bush, sometimes by wide margins. This post on the Left Coaster also suggests that states with paper ballots showed vote tallies comparable to exit poll results but those which had some level of electronic voting showed sharp gains for Bush and sometimes, reversals, compared to the exit polls. Yet another exit poll analysis here (via reader JD). Again, I don't know how accurate these results are and would like independent experts to analyze them and comment. 

However, on 11/12/04 I became aware of a more substantive analysis:

The Unexplained Exit Poll Discrepancy (Stephen F. Freeman, Ph.D., University of Pennsylvania)

Rather than reproduce entire section, I strongly encourage readers to take a look at his paper since it is the first systematic analysis of the exit polls that I am aware of, which also attempts to be statistically robust. Kevin Drum mentioned this on his blog Political Animal, with the following comments, which seem reasonable to me:

In fact, Bush won 10 out of 11 battleground states by more than the exit polls predicted. The odds of this happening by chance are essentially zero.

Now, Freeman's paper corrects a couple of the problems I've seen in earlier efforts along these same lines. First, he uses final (raw) exit poll data, not early afternoon data. Second, he provides a levelheaded discussion of the various reasons on offer for why the exit polls might be off — although in the end he finds them unconvincing.

However, Freeman doesn't allege voter fraud, and neither do I. There are several good reasons for this:

• All Freeman demonstrates is that these results couldn't have happened simply by chance. However, they still might have been produced by systematic problems of some kind in the polling methodology.

• Freeman knows this, so he then takes a look at possible sources of systematic error. He isn't persuaded that any of them stand up to scrutiny, but he also admits that there's just not enough data to say for sure. What's more, there might be some other source of error he's not aware of.

• I've seen several indications that there really are some weird anomolies with the exit poll data. I mentioned one last night, and there's another one you can see right in the table above: the total national sample was about 13,000 voters, but the state level polls all claim to include about 2,000 voters. Obviously something is screwy there.

• Finally, you have to posit a way in which the Bush campaign was able to rig the results in every single battleground state. Let's face it: it's a lot easier to think of ways in which the exit polls might be screwed up than to think up credible ways in which the Bushies could have pulled that off.

So why am I posting about this at all if I don't believe it? Mainly because I'm getting progressively more pissed off about the exit polls with every passing day. The folks who ran them have actually encouraged rumor mongering by refusing to publicly explain what happened to us benighted masses. So far, all they've done is write a confidential report that apparently didn't even acknowledge the systemic errors in the final results and instead laid the blame on those irresponsible bloggers who got everyone riled up by posting early results. Meanwhile, their defenders in the media were practically apoplectic about the gall of non-experts using data they can't possibly understand in order to advance their own bizarre conspiracy theories. Which would be fair enough if they'd step up to the plate and give us the benefit of their expertise.

But they haven't. They've declined to talk to reporters, they've released no serious findings of their own, and they haven't made their raw data available even to qualified researchers. Freeman's numbers might well be wrong, but there's no way for him to find out. Nature abhors a vacuum, and the experts are the ones who have created the vacuum in which rumors thrive.

So that's why I'm posting this: because I want to put some pressure on them to come out from their caves and tell us what they think — and to debunk guys like Freeman if they can. After all, there's probably a perfectly plausible explanation for all this. In fact, I'm sure there is. But until they tell us what it is, the conspiracy theories aren't going to go away.

I would like this analysis and debate to continue. 

And continue it does.

Kerry supporter and respected pollster Ruy Teixeira of Donkey Rising has a post challenging one of Freeman's contentions - that exit polls are intrinsically trustworthy. Read it. I am waiting for Freeman's response.

Part IV: WAS THIS ELECTION WON BY BUSH BECAUSE OF VOTE FRAUD? 

Now, now. This is a trick question :-).

Let me first quote John Belisarius at Donkey Rising since I am in complete agreement with this post of his (bold text, except the header, is my emphasis):

How Big A Role Did Fraud, Ballot Theft and Suppression of the Vote Play in The Election?

In the last few day's accusations of massive vote fraud, ballot theft and suppression of the Democratic vote during the 2004 elections have mushroomed to such a level that both the New York Times and the Washington Post have given the charges front page coverage.

Unfortunately, almost all the discussion of this issue has become focused on the specific question of whether a sufficient number of votes might have been stolen or suppressed to have changed the outcome of the election. In many cases, the unstated assumption seems to be that if such violations did not rise to the level where they changed the result then they can safely be ignored.

That's the wrong way to look at this issue. What the vast majority of Democrats find most disturbing about 2004 is that Bush's victory was based on a pervasive strategy of dishonesty--a dishonesty that included major distortions of Kerry's record by the Bush campaign's own television commercials, outright lies told by the Swift Boat Veterans, grotesque distortions circulated among rural or minority voters (such as the claim that Democrats would take away religious people's bibles or that Martin Luther King was a Republican), flyers listing false reasons why voters should believe themselves disqualified, leaflets and phone calls falsely announcing changes in polling places and phony voter registration groups that collected and then destroyed voter registration forms.

Layered on top of this were techniques for suppressing the vote in Democratic areas that included last minute changes in polling places, use of felon lists known to be inaccurate and the provision of inadequate numbers of voting machines and ballots. [eRiposte note: For a more comprehensive listing, go to my Vote Watch 2004 site]

It is this entire pattern of appallingly anti-democratic behavior that should be at the center of the national discussion today, and not just the specific question of whether these kinds of activities--along with any direct theft or alteration of votes by electronic or punch card voting machines--could have risen to a level sufficient to reverse Bush's victory.

Regarding the precise amount of voter fraud and suppression that actually occurred during the election, data are still trickling in. A widely quoted article by Harpers magazine writer Greg Palast pulled together a variety of issues to draw the conclusion that Kerry might actually have won the election. Follow-up articles in Salon and The Nation by Farhad Manjoo and David Corn, however, while entirely sympathetic to Democrats basic suspicions and complaints, reviewed Palast's evidence and reached the opposite conclusion.

The debate is not over. Two web sites that continue to collect and evaluate reports from around the country are the Election Incident Reporting System and the CalTech/MIT Voting Technology Project.

But the most important thing for Democrats to remember about this debate is that they should not allow it to be reduced simply to the question of whether or not the election was "stolen". What vast numbers of Democrats as well as many moderates and independent voters already believe and believe very strongly is that Bush's victory was based on a campaign that was deeply, deeply dishonest and profoundly unfair.

Folks, I am of the belief that before a claim of a "stolen election" is made, incontrovertible proof must be shown that it was indeed "stolen". Based on the data I have seen so far, I can't say there is incontrovertible evidence proving Bush was wrongly awarded millions of votes by voting machines - namely, by a margin sufficient to overturn the current verdict. 

Having said that, there is growing evidence of suspicious results - especially in Florida (and possibly in Ohio). There is significant evidence that vote suppression, especially in Ohio, was extensive. There is incontrovertible evidence that the Bush campaign and the GOP ran a sometimes fraudulent and a general vote-suppressing campaign. Just as there is no proof (yet) that Bush illegally got millions more votes than he did, there is also no proof that the irregularities, suspicious results and full range of dirty and fraudulent tactics employed by the Bush campaign and the GOP did NOT add up to suppress Democratic votes by a large margin. So, here is my conclusion (for now).

The only way we can sustain this democracy is by allowing free speech and allowing people to bring up all irregularities and fraud to the surface and investigating it to the end, with the full resources that such investigations require. Nothing short of the future of our democracy depends on it. It is from the results of these investigations that we can educate people about the country they live in, about who really is on their side and who really won and lost. It is the only way to find better means to run elections the next time. So, I say, keep the investigations and analysis going, but let's not get too carried away by the rhetoric just yet.