From time to time, Kossacks and others will cite Robert F. Kennedy Jr.'s summer 2006 article, "Was the 2004 Election Stolen?", as an examplar of investigative reporting into election integrity issues. When Markos recently opined on the front page that "no Diebold trickery was needed to steal" the 2004 election, at least four commenters cited the Rolling Stone article, or Kennedy himself, to assert that kos was wrong. Trouble is, Kennedy's article is riddled with hyperbole, non sequiturs, and outright errors -- some of them bizarre. Kennedy could have written a solid review of election irregularities in Ohio and elsewhere; he just blew it. Over two years later, many people are surprised to hear that there are criticisms of Kennedy's article, so I wanted to walk through some of them.
(If you're more interested in election integrity than you are in RFK Jr. -- good, stick with that, and you can safely skip this diary. Just be wary about recommending Kennedy's articles.)
Exit poll hype
I start with exit poll hype because Kennedy does, in the very first sentence: remembering how he spent election night "wondering how the exit polls, which predicted an overwhelming victory for John Kerry, had gotten it so wrong." Kennedy's discussion of exit polls makes up about one-seventh of his analysis. Kennedy asserts that "Over the past decades, exit polling has evolved into an exact science.... The results are exquisitely accurate: Exit polls in Germany, for example, have never missed the mark by more than three-tenths of one percent." And so on. From Kennedy, one might get the impression that experts agree that the exit polls were most likely accurate, while "the mainstream media" take the other side.
Not so much. The expert consensus on exit polls was summed up pretty well in an SSRC working group report in December 2004: "Discrepancies between early exit poll results and popular vote tallies in several states may be due to a variety of factors and do not constitute prima facie evidence for fraud in the current election." In several states where the exit poll results differed widely from the official returns -- for instance, New Hampshire, Minnesota, and Pennsylvania -- they also were far from pre-election polls. (Pre-election polls gave Kerry narrow leads in all three states; the exit polls gave him double-digit margins.) The U.S. presidential exit polls have been significantly "off" at least since 1988, and the discrepancy in 1992 was almost as large as the one in 2004. Kennedy isn't even right about Germany. And he offers this bizarre assertion:
On the evening of the vote, reporters at each of the major networks were briefed by pollsters at 7:54 p.m. Kerry, they were informed, had an insurmountable lead and would win by a rout: at least 309 electoral votes to Bush's 174, with fifty-five too close to call.
Umm, no. Kennedy's 'evidence' for this assertion is a page in Steve Freeman's book which provides Freeman's own reconstructions of the exit poll estimates. (Freeman's figures don't even match Kennedy's -- perhaps Kennedy was working from a draft.) In real life, the pollsters briefed the networks, around 4:30 in the afternoon, that they did not trust their own numbers. Neither the pollsters nor the networks even came close to calling the election for Kerry.
I could go on. Mark Blumenthal of pollster.com devoted four entire articles (starting here to Kennedy's errors about exit polls, and I've written extensively about it myself. Bottom line: whatever else happened, the exit polls were wrong. And we can expect them to be off again -- whether or not there is fraud. Deal with it.
80,000+ Kerry votes counted for Bush?
Kennedy alludes in his intro to evidence that "upwards of 80,000 votes for Kerry were counted instead for Bush." In Section VIII, he elaborates:
One key indicator of fraud is to look at counties where the presidential vote departs radically from other races on the ballot....
Take the case of Ellen Connally, a Democrat who lost her race for chief justice of the state Supreme Court. When the ballots were counted, Kerry should have drawn far more votes than Connally -- a liberal black judge who supports gay rights and campaigned on a shoestring budget. And that's exactly what happened statewide: Kerry tallied 667,000 more votes for president than Connally did for chief justice, outpolling her by a margin of thirty-two percent. Yet in these twelve off-the-radar counties, Connally somehow managed to outperform the best-funded Democrat in history, thumping Kerry by a grand total of 19,621 votes -- a margin of ten percent....
Kucinich, a veteran of elections in the state, puts it even more bluntly. "Down-ticket candidates shouldn't outperform presidential candidates like that," he says. "That just doesn't happen. The question is: Where did the votes for Kerry go?"
Kennedy goes on to extrapolate an estimate of 80,000 stolen votes. Trouble is, Kucinich's (and Kennedy's) premise is flat-out wrong. In 2000, one Democratic judicial candidate got more votes than Gore in 81 out of 88 counties, and the other outdrew Gore in 40 counties. In 1996, one Republican judicial candidate outdrew Dole in 81 counties, and the other in 58. There is no evidence that the so-called "Connally anomaly" exists at all. In all three years, Democratic judicial candidates tend to do disproportionately well in Republican counties, and vice versa. The likely reason is that judicial candidates are non-partisan on the ballot and are not well known, so some voters may accidentally vote for a candidate of the opposite party.
These counties' election returns actually aren't especially anomalous. Walter Mebane and Michael Herron, in a report for the Democratic National Committee which RFK Jr. himself cites (see below), concluded that precinct-level returns "present strong evidence against the claim that widespread fraud systematically misallocated votes from Kerry to Bush" in these counties or elsewhere. Mebane even provided evidence to Rolling Stone that the "Connally anomaly" was a canard -- yet it remained as a centerpiece of the story.
Ohio's 357,000 missing votes
Kennedy says,
A review of the available data reveals that in Ohio alone, at least 357,000 voters, the overwhelming majority of them Democratic, were prevented from casting ballots or did not have their votes counted in 2004.... [O]ne in every four Ohio citizens who registered to vote in 2004 showed up at the polls only to discover that they were not listed on the rolls, thanks to GOP efforts to stem the unprecedented flood of Democrats eager to cast ballots.
Kennedy points to some very real problems in this analysis, although he rarely gets them quite right. It is unlikely that these "missing votes" provided Bush's margin of victory. Let me take them point by point.
- "174,000 unable to vote because of long lines on Election Day."
That estimate is based on the DNC Voting Rights Institute's Voting Experience Survey (the VRI's full report is available here). The report isn't entirely consistent. At one point, it says that 3% of voters left the lines and did not return. At another point, it says that 2% left the lines and did not return, while about 0.83% never went to the polls because they did not receive absentee ballots or had heard about the long lines and other voting problems. Of those who left the lines, the report says, "these potential voters would have divided evenly between George Bush and John Kerry" (Section III, page 2). Talk about suppressing exculpatory evidence! Now, that's a survey estimate based on a small sample. In Franklin County -- which arguably had the worst lines in the state -- one peer-reviewed analysis (by Benjamin Highton) estimated that about 22,000 votes were lost and that about 64% of them would have gone to Kerry, costing Kerry about 6,000 net votes. Statewide, Kerry's share of these potential votes probably was somewhere in between the survey estimate and Highton's.
Long lines were a serious problem in Ohio, but there isn't much evidence tying them to deliberate fraud, and they probably contributed only fractionally to Bush's winning margin.
- "72,000 disenfranchised by avoidable registration errors."
This figure is based on a study by Norman Robbins of the Greater Cleveland Voter Registration Coalition. Doing a back-of-the-envelope calculation based on Cleveland, Robbins suggests that about 42,000 votes in big cities statewide may have been "lost," and another 30,000 "put at risk," by avoidable errors. The calculation is speculative -- but it's a good-faith estimate that I won't debate. I will point out that "put at risk" does not equal "disenfranchised."
Provisional ballots may have saved many tens of thousands of votes that would otherwise have been lost. About 158,000 people statewide ended up casting provisional ballots -- and all but about 35,000 of those ballots were counted. Even assuming that all those uncounted provisionals should have counted, it is hard to conclude that registration errors and purges combined amounted to as many as 72,000 votes statewide.
- One in four new registrants not listed on the rolls (not part of the 357,000 votes)
The DNC conducted a study of provisional ballots in Cuyahoga County, which estimated that in that county, 26% of newly registered voters were asked to vote provisionally. We do not know whether the statewide proportion was similar. Kennedy offers no evidence that this percentage was "thanks to GOP efforts to stem the unprecedented flood of Democrats eager to cast ballots."
The DNC study further found that overall, approximately equal proportions of Kerry and Bush voters in Cuyahoga County had to cast provisional ballots, and that approximately 69% of the provisional ballots for Kerry and 79% of the ballots for Bush were counted. Since there were a lot more Kerry voters than Bush voters in Cuyahoga County, we can infer that more Kerry votes than Bush votes on provisional ballots were lost in Cuyahoga County, although not necessarily statewide.
- "66,000 had ballots invalidated by faulty voting machines"
Kennedy notes that 95,000 Ohio ballots "recorded no vote for president at all -- most of them on punch-card machines." Based on an estimate that 0.5% of voters intentionally do not vote for president, he estimates that 66,000 of these "undervotes" were unintentional. (Walter Mebane offers a slightly smaller estimate based on the assumption that 0.75% of voters intentionally did not vote.) In Ohio as in other states -- famously Florida in 2000 -- punch cards have often gobbled votes.
Kennedy asserts, "If counted by hand instead of by automated tabulator, the vast majority of these votes would have been discernable." Pshaw. In Florida in 2000, there were about 175,000 undervotes and overvotes on punch cards, of which a few thousand could have been recovered in a recount. Kennedy offers no reason to believe otherwise of Ohio 2004, and Richard Hayes Phillips' analysis of thousands of punch card ballots offers no such evidence either.
- "30,000 purged from voter rolls for failing to vote in two previous elections"
Kennedy says that election officials in Cleveland, Cincinnati and Toledo "summarily expung[ed] the names of more than 300,000 voters." Kennedy says that if one-tenth of these voters showed up to vote -- "a conservative estimate, according to election scholars" (who remain unnamed and uncited) -- that would amount to 30,000 disenfranchised voters. He adds, "In Cleveland, which went five-to-one for Kerry, nearly one in four voters were wiped from the rolls between 2000 and 2004." Moreover, "In Cleveland's precinct 6C, where more than half the voters on the rolls were deleted, turnout was only 7.1 percent -- the lowest in the state."
Purges always pose the possibility of disenfranchising legitimate voters -- and some seem positively designed for that purpose. However, there is very little evidence that these particular purges affected many legitimate voters. As I mentioned above, only about 35,000 provisional ballots statewide, from all kinds of voters (including new registrants), went uncounted. In Cleveland at the time of the 2004 election, even after all those purges, there were almost as many registered voters in the county as there were adults!
The example of Cleveland precinct 6C is a strange one: Kennedy seems not to realize that if fewer voters had been purged, the percentage turnout would have been even lower. Turnout there has been similarly small in other recent elections. Why? Some map work shows that the Cleveland Clinic has expanded to overrun this precinct; it seems likely that most of the people registered there are long gone, but haven't (or hadn't) been purged yet. (See also this corroborating DKos comment from just after the 2004 election.)
- "10,000 had ballots discarded because they stood in the wrong line"
Norman Robbins extrapolates this figure from "inquiries to 18 counties representing over 60%" of rejected provisional ballots. I can't vouch for the figure, but the phenomenon is for real. According to Robbins, in Cuyahoga County, over half these rejected ballots were cast at the right polling place but in the wrong precinct! Others were cast at the wrong polling place in the right county. In some cases, voters were at least partly to blame for going to the wrong polling places -- but it seems clear that in many cases they were given incorrect information.
(Note that if 10,000 provisional ballots went uncounted for this reason, only 25,000 uncounted PBs are left to be attributed to registration errors and purges.)
- "5,000 turned away by GOP challengers"
The DNC survey description states, "A smaller group of potential voters (0.08 percent) were not given ballots at all due to registration challenges. These approximately 4,798 voters favored Kerry, according to the poll (extreme sample size caution)." Extreme sample size caution indeed: 0.08 percent equals one respondent out of the 1201 in the survey. The report does not state whether this one respondent was "turned away by GOP challengers," or by a poll worker. Surely some people were turned away who should at least have been encouraged to cast provisional ballots -- although many others did cast provisional ballots.
Conclusion
There's a lot more in Kennedy's article -- some of it fine, some of it skanky -- but this diary is already crazy long. Many things went wrong in the 2004 Ohio election, and some were intended. People can legitimately disagree about how to summarize the story, and what lessons to learn from it. But Kennedy seems to want to draw the lesson that Kerry pretty clearly won, and only Rolling Stone and a ragtag band of experts have the courage to tell us. Spare me. Here's what Mark Chu-Carroll had to say in his blog "Good Math, Bad Math":
RFK Jr's article tries to argue that the 2004 election was stolen. It does a wretched, sloppy, irresponsible job of making the argument. The shame of it is that I happen to believe, based on the information that I've seen, that the 2004 presidential election was stolen. But RFK Jr's argument is just plain bad: a classic case of how you can use bad math to support any argument you care to make. As a result, I think that the article does just about the worst thing that it could do: to utterly discredit anyone who questions the results of that disastrous election, and make it far more difficult for anyone who wanted to do a responsible job of looking into the question.
So, no matter what one thinks happened in 2004, one can find pretty good reasons to be unhappy with Kennedy's article. (Not that Kennedy really has it in his power to "utterly discredit anyone who questions the results," but why mess with it?)
[Some folks may be interested in Farhad Manjoo's rebuttal in Salon (which makes a few mistakes), plus the ensuing colloquy between Kennedy and Manjoo. Disclosure: I'm referenced as a source in Manjoo's article.]