Two Responses to Paul Gronke

The two discussion-thread postings reproduced here were my contributions, dating from January 4 and 22, 2006, to a discussion under the heading “New Year's Resolutions?” initiated on December 30, 2005 at the website BlueOregon, One of the participants was Paul Gronke, professor and chair of the Department of Political Science at Reed College in Portland, Oregon, and the author of Settings, Campaigns, Institutions and the Vote: A Unified Approach to House and Senate Elections (University of Michigan Press, 2000), as well as of articles, book chapters, and policy reports on US elections. Professor Gronke intervened to rebuke a discussant whose New Year's resolution was “to notify as many people as possible that GW Bush has not been elected as our president, ever.” On January 1, he wrote:

With all due respect, the claims made on bradblog have been debunked extensively. They just don't stand up to scrutiny. Bush won more votes than Kerry in November 2004 and is the legitimate winner of the 2004 election. It is time to move on. Check the discussion on (click on the link to exit poll controversy); a forthcoming issue of Public Opinion Quarterly (unlike bradblog, a peer-reviewed scholarly publication); and the Democratic Party's own funded study of balloting in Ohio ( If you want to confirm this with other blogs, try Rick Hasen's electionlaw blog, Dan Tokaji's equal voting rights blog, and websites maintained by the CalTech/MIT Voting Technology Center, and Continuing to believe this urban legend only forestalls the serious work that Democrats need to do in order to regain the Congress and the Presidency.

My own work was alluded to on January 2 by a discussant who urged readers “to download a graphic developed by the Oregon Voters Rights Coalition, Timeline of National Exit Poll & Machine Tally Data in Presidential Race—Election Day 2004” which was based on an interpretation of “Dr. Michael Keefer's analysis” in the first of my election fraud essays, “Footprints of Electoral Fraud.” Professor Gronke's response on January 4 included the following:

Yes, and please read the extensive debunking of Keefer's analysis at the sources I posted above. These are not right wing radicals folks—I know every single one of the academics personally. If it matters (it should not but in these times it does), they are progressive Democrats as well as well-established statisticians and scholars of political analysis. And NONE of them, I repeat NONE have given any credibility to the verified voter claims. The claim made above about Keefer is ONLY acceptable if you take the exit polls as valid and accurate to a certain margin of error. If, as has been shown in many states, the implementation of the exit polls in 2004 had serious flaws, the claims made by Keefer collapse.]



Jan 4, 2006 11:20:58 PM

May I join this very interesting discussion to offer corrections to a couple of Paul Gronke's postings? To the best of my knowledge none of the people Paul mentions as experts have engaged in “extensive debunking” of my analysis. I've read Rick Hasen, Dan Tokaji, Mark Blumenthal ('mystery pollster') and the others at intervals—always with respect and interest—and have never found the least indication that they even know of my existence.

These are, as Paul says, highly reputable people. (So also, by the way, is Paul Gronke, whose work on certain aspects of US elections is well known among political scientists.) But with the exception of the scholars associated with the CalTech/MIT Voting Technology Center, who as 'BethP' noted lack any critical perspective on issues of electronic security, it's really not accurate to claim that any of them know very much about electronic voting mechanisms and vote counting methods. The real experts in that area—people like Rebecca Mercuri, Bruce Schneier, Doug Jones, and Aviel Rubin—were issuing a crescendo of warnings about the openness of new voting technologies to fraud during the years leading up to the 2004 election. (The first 50 or so entries in a bibliography on “Evidence of Fraud in the 2004 U.S. Presidential Election” that I posted at on December 5, 2004 are devoted to that subject.)

Paul's also wrong to claim that “the sole basis for fraud is the claim that we should rely on the exit polls rather than the precinct results.” 

The exit polls are important evidence, and those who want to argue that this particular set are mysteriously unreliable are in my opinion trying (at best) to protect themselves from a truth they find too painful to contemplate. To my mind, our best guides to what the state exit polls reveal are 'TruthIsAll' (who's posting now at the Progressive Independent site), and Steven F. Freeman (who humbled Warren Mitofsky in their debate at the University of Pennsylvania in October). As 'TIA' and Freeman have also both observed, the 2004 nationwide exit poll likewise shows that there's something very seriously wrong with the national vote tally in that presidential election. 

Let's run through the arithmetic. The number of votes cast for presidential candidates rose from 105,405,000 in 2000 to 120,255,000 in 2004, while votes cast for third-party candidates (chiefly Ralph Nader) declined from 3,959,000 in 2000 to 1,170,000 in 2004. As I noted in an article posted at on November 5, 2004, the exit poll data were corrupted early in the morning of November 3rd, while we all slept. 

But here's what the national exit poll data that was made available by CNN on the evening of November 2nd (and that was based on responses from 13,047 randomly selected voters) tells us. First, 83% of these people said they had also voted in 2000. Of the Gore 2000 voters, 91% voted for Kerry in 2004, while 8% voted for Bush. Of the Bush 2000 voters, 90% stuck with him, while 10% swung to Kerry. Of the people who'd voted for third parties in 2000, 64% voted for Kerry and 17% for Bush. People voting for the first time in 2004 went 57% to Kerry and 41% to Bush.

Gore, remember, won the popular vote in 2000 by almost 544,000 votes (50,999,897 votes to Bush's 50,456,002). On the assumption, generous to Bush, that the year-2000 voters who switched loyalties in both directions in 2004 balance each other out, we can take the base number of supporters for Bush and Kerry as amounting to nearly 95% of the Democratic and Republican presidential vote tallies in 2000: in round numbers, 48.4 million votes for Kerry and 47.9 million for Bush.

If a similar percentage of the 3,949,000 who voted for third-party candidates in 2000 also voted in 2004, then since this group went 64% to Kerry and 17% to Bush, that gives about 2.3 million more votes to Kerry, and about 600,000 to Bush. Their totals are now at 50.7 million votes for Kerry and 48.5 million for Bush.

Add in the 20.2 million new voters, 57% of whom voted for Kerry and 41% of whom supported Bush. That gives Kerry 11.5 million more votes, and 8.3 million for Bush. The final expected total? It's 62.2 million votes for Kerry, and 56.8 million for Bush.

Now compare that to the official results: 61,194,773 votes (51% of the total votes cast) for George W. Bush, and 57,890,314 (48% of the total) for John Kerry. Do you smell a fault? Or were the 13,000 people who were sampled systematically deceiving the pollsters?

But as I said, Paul's wrong in thinking that the exit polls are the sole indication of fraud in the 2004 election. In January 2005, I posted an article entitled “The Strange Death of American Democracy: Endgame in Ohio” at (I'm sorry if the title seems apocalyptic, but I think that's what's at stake.) In that article, I tried to summarize the evidence for electoral fraud in Ohio, the state whose Electoral College votes determined the national outcome.

That evidence is both massive and also, for anyone who cares about democracy, sickening. I'm not talking just about the disgusting vote-suppression tactics that forced long lines of African-Americans to stand for hours in the November rain outside precincts were the voting machines had been deliberately shorted.

There's lots more. It includes the 106,000 provisional ballots and punch-card undervotes that were never counted. It includes the touch-screen machines which 15% of Ohio's voters had to use, and which in Youngstown and elsewhere systematically flipped Kerry votes to Bush or into cyberspace. It includes the Democratic precincts in Cleveland where people lined up for hours in the rain in order to achieve, according to the certified vote tallies, precinct-level turnout figures such as 20.07% of the registered voters (precinct 13F), 14.59% of the registered voters (13D), and 7.85% of the registered voters (6C). And then there's Cleveland's precinct 10L, where the turnout figure magically rose from 24.72% in the initial report to 56.21% in the figure certified by Ohio Secretary of State Kenneth Blackwell. (Did someone forget to throw away the provisional and absentee ballots cast in this one inner-city precinct?)

In Republican Perry County, however, there was no problem with voter turnout. Blackwell (who was also co-chair of the Bush campaign in Ohio) certified turnouts of 124.4% and 124.0% of the registered voters in two precincts of that county. In the Concord South and Concord South West precincts of Republican Miami County he certified turnout figures of 94.27% and 98.55% of the registered voters. And there's more. There's Warren County, where the administration building in which the vote tabulation was carried out was locked down on the phony pretext of a terrorist threat. And guess what: in Warren and the two adjoining counties, Bush's margin of victory rose by more than 30,000 votes over what it had been in the 2000 election.

Is that enough to give just a taste of what went on?

Here's a challenge for you, Paul. You're a fine scholar: I know, because I've read some of your work. You love your country, and you're passionate about democracy. How is it then that I know these things and you don't? I'm a scholar too, but not a specialist on elections, or even a political scientist. Hell, I'm not even an American. My great-great-great-great grandfather died in 1780 defending Long Island from George Washington's army, and his widow was kicked off the family farm in Paulinskill, New Jersey, and had to walk north to Ontario with her two boys and a cow. (That's o.k.; we've learned to like it here.)

What's my challenge? To have a go at what the anthropologist Clifford Geertz called “thick description,” to get in close to the details of what people reported from that dirty election, and to assess it critically and unflinchingly—without forgetting for a moment what's at stake for your country and the world.


[Professor Gronke was not persuaded. On January 11 he wrote as follows:

Sigh. I got sent a link from where “The Bear” suggests that my lack of reply indicates that I have been convinced. Far from it—I just don't see any point in continuing to debate with people who are unwilling to listen to any contrary evidence.

Over at progressiveindependent, Keefer claims: 'Nor were any of the experts he cited—Mark Blumenthal, Rick Hasen, Dan Tokaji, the CalTech/MIT Voting Technology gang, and Old Uncle Tom Cobbley and all—people of statistical persuasion.'

This is complete hokum. The CalTech/MIT group contains some of the most accomplished political methodologists in the country. Blumenthal has been conducting political polls for twenty years. Hasen and Tokaji are experts in election law. Regardless, I cited these places as sites to find discussions and alternative links, not as “experts” on this particular topic (although the CalTech/MIT folk do qualify).

I'll only leave you with these links. Draw your own conclusions.

WalterMebane's series of reports on Ohio: Most important is this statement, pg. 7 of the executive summary: 1. The statistical study of precinct-level data does not suggest the occurrence of widespread fraud that systematically misallocated votes from Kerry to Bush. If anyone wants to question Walter's partisan credentials, here is his published paper which demonstrates that Bush lost the 2000 election:

Bruce O'Dell (founder of US Counts Votes, the organization which originally claimed fraud), believes the evidence does NOT support fraud. His paper, along with a variety of responses, is contained here:

Here is the CalTech/MIT Voting Technology Project report on 2004: Note this phrase from page 1: 4. We conclude that there is no evidence, based on exit polls, that electronic voting machines were used to steal the 2004 election for President Bush.

Here is a report from the Social Science Research Council, written by some of the most accomplished political scientists in the country:

Note that ALL of these reports indicate serious problems with our current election system, and support the reforms that have been advocated here. I support the work of Oregon VRC.

But I think it's a gross political error to continue to use the alleged stolen election of 2004 as their launching off point. Not only [has] this been extensively debunked, but it undermines their support among Republicans who might otherwise support election reform.

When folks say “move on,” that's what they mean.]


Jan 22, 2006, 10:35:46 PM

I hope it's not a violation of thread etiquette to be responding so belatedly to Paul Gronke's last posting.

Paul Gronke was probably right to detect (and call me on) a note of arrogance that crept into the remarks of mine he quoted from Mea culpa: I must have been still at my desk long past the hour when people who are or wish to remain sensible have gone to bed. 

And his reminders of important work by Walter Mebane, Bruce O'Dell, and the authors of the Social Science Research Council's 22 December Interim Report are of course helpful.

But shall we take a closer look at the work he thinks shuts the issue down?

First, Walter Mebane—a brilliant scholar who (unlike those Gronke previously referred to) has done innovative and heavyweight work in statistical modeling and data analysis. Mebane's work on the Ohio data is indispensable, and on some issues seems to me conclusive. (For example, his analysis of the shorting of voting machines in Franklin County demonstrates that on a conservative estimate the result was a more than four percent reduction in voter turnout in predominantly African-American precincts.)

But on the larger questions of whether, where, and with what results vote-count fraud may have occurred in Ohio in 2004, Mebane's analysis fails to incorporate and explain certain key anomalies. These include the irregularities highlighted by Richard Hayes Phillips, whose county-by-county studies are posted or linked at the site. Many of these same irregularities are discussed in the very detailed report on the Ohio election assembled by Congressman John Conyers (which to its shame the Democratic National Committee largely ignored in its report Democracy at Risk: The 2004 Election in Ohio).

One symptom of Mebane's insufficiently elaborated conceptual framework is the way in which he and Michael Herron, in their analysis of “Turnout, Residual Votes and Votes in Precincts and Wards” (Section 6 of the DNC report), find themselves speaking of a repeated (and unexplained) “weirdness” in the data they are analyzing (pp. 5, 7). Repeated “weirdness” in data sets and their correlations is—or should be—a clear sign that the researcher's conceptual apparatus needs further refinement: either the relations among causal factors have not been adequately theorized, or else further unrecognized factors are in play.

On the evidence of his postings in this thread, Paul Gronke doesn't seem interested in doing critical work of this kind. Other people who are will find some astute (and respectful) criticisms of Mebane's analysis in two postings by 'Time for change' at Democratic Underground [on 7 and 30 October 2005].1 'Time for change' notes that the within-precinct correlations among five variables that Mebane thought provided “strong evidence against the claim that widespread fraud systematically misallocated votes from Kerry to Bush” are significantly weaker in seven counties—among them Cuyahoga County (which includes Cleveland), where one of the correlations is actually negative. S/he notes that another correlation remarked on in Section 4 of the DNC report, between voter turnout and the ratio of voters to machines, also goes negative in Cuyahoga—which could of course be a sign that large numbers of votes in that county were being trashed by hackers working with the electronic tabulators.

As 'Time for change' also observes, there's no reason to think that Mebane's network of correlations would detect randomly conducted deletions from precincts in Democratic strongholds. I don't suppose, for that matter, it would detect ghost-votes added to the tally in the right proportions in Republican strongholds. (Remember Miami County, Ohio? It first announced a total of 31,620 votes, with 100% of the precincts reporting—and then, late on election night, brought in a second total of 50,235 votes, which was in two respects anomalous: Kerry's 33.92% share of the vote remained, to one-hundredth of one percent, what it had been in the first returns, and Bush was shown to have won the county by exactly 16,000 votes.)

'Time for change' reports correspondence with Professor Mebane in which the latter says: “ I don't know what went on in Cuyahoga County. As I wrote in several places in the DNC report, there were many anomalies in the data from Cuyahoga County that warrant further investigation.” Mebane isn't conceding anything in these words—but neither is he rudely telling people to “Move on.”

Let's do so anyway, and turn to Bruce O'Dell—who, despite what Gronke tells us, nowhere says that he “believes the evidence does NOT support fraud.” Can we try for some minimal accuracy here? O'Dell thinks “The case for fraud is still unproven,” and believes that it cannot be proven “through exit poll analysis alone.” But he insists in the paper to which Gronke refers that his work “should not be misinterpreted as an argument against the likelihood of vote fraud. Quite the opposite; I believe US voting equipment and vote counting processes are severely vulnerable to systematic insider manipulation and that is a clear and present danger to our democracy.”

In this light—given Gronke's stern disapproval of “people who are unwilling to listen to any contrary evidence”—it may be worth noting that Kathy Dopp thinks it possible to determine mathematically what patterns of exit poll discrepancy result from random sampling error, partisan exit poll completion differences, and vote miscounts. (See her recently published paper “Vote Miscounts or Exit Poll Error? New Mathematical Function for Analyzing Exit Poll Discrepancy” [16 January 2006],

Let's leave the mathematicians to work these matters out among themselves—though not without observing that Steven Freeman's presentation to the American Statistical Association on October 14, 2005, “Polling Bias or Corrupted Count?”, offers a persuasive account of what exactly the exit polls can tell us. Interesting, when you think about it, that a political scientist like Paul Gronke should prefer the glib sallies of “mystery pollster” Mark Blumenthal to the carefully weighed, scrupulously referenced—dare I say scientific?—analyses of a scholar like Dr. Freeman (available, by the way, at But there do seem to be leanings among some political scientists toward the patronizing sighing Paul Gronke indulges in at the beginning of his post: a bunch of them, in fact, hang out together at a blogspot they've named “Polysigh,” where they groan together over the follies of the uninitiated.

We needn't pause over the SSRC report Gronke recommends, unless to note that a great deal more data is available than that which was examined by this group in December 2004—and to observe that they think “continuing uncertainty over the extent of irregularities merits closer public scrutiny and full disclosure of relevant data.”

But I hope we can be forgiven, finally, for smiling over Gronke's recommendation of the CalTech/MIT paper “Voting Machines and the Underestimate of the Bush Vote.” This is the piece that made the embarrassingly elementary error of using the exit poll percentages published by CNN on November 3, 2004 as a basis for dismissing concerns about discrepancies between the exit polls and the vote tallies. As Gronke proudly tells us, the authors “found no evidence, based on exit polls, that electronic voting machines were used to steal the 2004 election for President Bush.” Of course they didn't: the exit poll 'data' they were using had been conflated or corrupted to move it toward conformity with the vote tallies.

So where, in the end, is all the debunking we were promised? Professor Gronke began by claiming that Blumenthal, Hasen, Tokaji, CalTech/MIT and others had shown that those who asserted (as I have done) that the 2004 election was stolen were peddling an “urban legend.” It now appears that some of these people are to be regarded rather as sources of links and discussions. But out comes another list of definitive debunkings—rather more to the point than before, but still not conclusive, unless one wants to lie down and roll over in the face of unexamined authority. Like the first list, it includes CalTech/MIT, whose authority on exit polling Paul Gronke so fancies.

There is, after all, something about this mode of conducting an argument that is irresistibly reminiscent of the old folk-song “Widdicombe Fair.” Here's the stanza I was half-remembering in my post at Progressive Independent:

Tom Pearse, Tom Pearse, lend me your grey mare, 
(All along, down along, out along, lee) 
I wants for to go down to Widdicombe Fair, 
Wi' Bill Brewer, Jan Stewer, Peter Gurney, Peter Davey, 
Dan'l Whiddon, Harry Hawk, 
Old Uncle Tom Cobbleigh and all, 
Old Uncle Tom Cobbleigh and all.

Substitute “Caltech/MIT” for “Uncle Tom Cobbleigh” in the refrain and the rhythm still works, more or less.




1  The details, given in my text in the original, are placed for convenience in this note: “Was Kerry Cheated out of Almost 100,000 Votes in Cleveland Alone?” (7 October 2005),; and “What Happened in Cleveland?—A Plausible Scenario for a Stolen Election” (30 October 2005),