I like good arguments, and I hate bad arguments. When people offer bad arguments on election integrity issues, it makes me want to scream. Do bad arguments actually undermine the cause of election integrity? Honestly, I don't know. I think so, but unfortunately, we don't get to rerun the last four years without the bad arguments to see if things go any better.
As time permits, I hope to post relatively terse rebuttals to some bad election fraud arguments. Inevitably some folks will interpret (or portray) these diaries as attempts to dismiss election integrity issues. Anyone who has been here long enough has read some variant of: "If you don't believe the exit polls, then you don't care about democracy." It's sort of like, "If you question the war, then you don't support our troops," or "Evolutionists are out to destroy morality." I think we can do better. More on that -- and my bullet points on exit polls -- beyond the jump.
Let's be clear: electronic voting machines are ridiculously insecure and generally unreliable, registration lists are far too easily gamed, and there is solid evidence of various kinds of election problems. Midwest Millian has written over a dozen fine diaries about e-voting; Project Vote produces consistently informative diaries on voting rights almost every week (or more). If half the people who complain about inattention to election integrity at DKos recommended diaries like these –- problem solved!
So if you aren't interested in rehashing the fraud war or this part of it –- please don't. Think of this as a reference that may be useful someday, or ignore it entirely. I'm writing about exit polls today because I've seen lots of quirky claims in recent comments, and because a new movie features similar claims.
The statements in bold are close paraphrases of arguments I've repeatedly encountered here on DKos and in other conversations.
Myth: Exit polls were always accurate until 2004, or, until 2000.
Nope. The 1992 exit poll was almost as far from the official returns as the 2004 exit poll -- and much more so than the 2000 exit poll. But almost no one noticed, because the election wasn't as close, and because the exit poll results weren't instantly posted on the Web. Many other U.S. exit polls have had smaller discrepancies.
Myth: Exit polls are so accurate that networks almost never blow a call.
Networks almost never blow a call because they don't make a call right away if the exit poll results are inconclusive.
Myth: Exit polls are so accurate that they are used around the world to confirm election outcomes.
It's true that exit polls are used around the world, but they are by no means universally accepted as a means of election verification. For instance, in a report leading up to the Mexican election in 1994, the Carter Center recommended, "Avoid exit polls, which are unreliable in a climate of suspicion and which will create a negative atmosphere if the voters feel they are being watched."
Myth: The United States relied on an exit poll to help overturn the result of the Ukraine election in November 2004.
It's true that the U.S. government helped to fund a major exit poll that indicated that challenger Viktor Yushchenko had won the first run-off. The results of this and other exit polls helped to inspire the "Orange Revolution." It's not true that the U.S. relied on the results. Instead, U.S. spokespeople and OSCE observers emphasized direct evidence of election fraud. At any rate, Ukraine exit polls gave Yushchenko as much as a 12-point lead, so one didn't need to trust the polls very much in order to doubt the official returns.
Myth: An expert calculated that there was one chance in 662,000 that Kerry really lost the election.
Steve Freeman estimated one chance in 662,000 that the exit poll discrepancies in Ohio, Florida, and Pennsylvania were due to random error. The calculation doesn't consider the possibility of non-random error!
Myth: The biggest discrepancies in the U.S. exit polls were in swing states.
Probably the most common measure of exit poll discrepancies is "Within Precinct Error," or WPE. (For instance, if the exit poll sample indicates that Kerry won a precinct by 10 percentage points, but he loses the precinct by 10 points in the official count, the Within Precinct Error would be -20.) Using raw average WPE ("IM WPE") as the standard, the biggest discrepancies were in Mississippi (-18.5), Connecticut (-16.0), Delaware (-15.9), Vermont (-15.2), New Hampshire (-14.0), and New York (-12.2). [See pages 32-33 of the Edison/Mitofsky evaluation report.] Different measures produce slightly different lists, but New Hampshire is the only swing state in the top five.
It is true, however, that the average discrepancy in swing state precincts was about two points larger than the average discrepancy in all other precincts. [See page 42 of the report.]
Myth: The exit poll predicted that Kerry would win Ohio.
The projection based on exit poll interviews only gave Kerry a 6.5-point lead in Ohio. The composite projection, incorporating pre-election expectations, have him just a 3.4-point lead. [See page 22.] Both these leads were well within a typical "margin of error," even assuming that the exit poll was unbiased. Ohio was too close to call. (But the "red shift" in Ohio was statistically significant.)
Myth: The exit poll results dovetail with other evidence of election fraud.
Not so much. I have seen little evidence of, or even speculation about, election fraud in Mississippi, Connecticut, Delaware, Vermont, or New York. In New Hampshire, a projection based on exit poll interviews indicated that Kerry would win the state by 15 points [see page 22] -- but pre-election polls showed a very close race. (Also, a partial recount focused on New Hampshire results believed to be anomalous found no significant miscounts.) Kerry's 31-point lead in the New York exit poll similarly made no sense.
In fact, the exit pollsters convened a conference call at 4:30 Eastern on election day afternoon to warn the networks that they didn't trust their own numbers in nine states [see page 17]. Interestingly, New York didn't make the list at that point.
Myth: Edison/Mitofsky simply assumed that the exit poll had to be wrong.
Edison/Mitofsky probably looked at exit poll results like the 15-point Kerry margin in New Hampshire, or the 31-point margin in New York, and concluded that these results were facially unlikely to be correct. Edison/Mitofsky also looked at the discrepancies in precincts that used Direct Recording Electronic (DRE) machines, and found that they were similar on average to the discrepancies in punch card precincts, and smaller on average than the discrepancies in lever machine precincts. [See page 40.]
Myth: The exit poll was only accurate in hand-count precincts, which indicates that all the machine counts were corrupted.
There were only 40 hand-count precincts in the sample, including only 5 in cities. The discrepancies in the 35 non-urban hand-count precincts were indistinguishable from the discrepancies in other non-urban precincts. [See page 40 again.]
Myth: Edison/Mitofsky grasped at straws to account for the discrepancy.
When a survey provides facially odd results, one likely cause is non-response bias -- in this case, that Kerry supporters were more likely to participate than Bush supporters. Unfortunately, the exit pollsters don't have extensive data on the people they didn't interview (obviously!), so the evidence is necessarily indirect. For instance, young interviewers tended to obtain larger discrepancies [see page 43], which could mean that some Bush voters steered around interviewers whom they expected to be Kerry supporters. But even if that particular argument is misleading, it wouldn't follow that the exit poll results were accurate. (Other evidence for non-response bias is discussed in the sources below.)
Myth: The higher average completion rates in the most Republican precincts prove that, if anything, Bush voters -- not Kerry voters -- were more likely to participate.
It's true that the completion rate in the 40 precincts where Bush did best was 56% compared with 53% overall [see page 37]. This doesn't at all prove that Bush voters were more likely to participate. These precincts likely were disproportionately rural, so probably both Bush voters and Kerry voters participated at higher rates in these precincts than elsewhere.
A crude analogy may help. Washington, D.C. has a far larger proportion of African-Americans than any state in the country, and its per capita income is about one-third larger than the national average. This result does not mean that African-Americans earn more money than whites nationwide, or even in D.C. -- quite the contrary. African-Americans in D.C. do have higher per capita incomes than African-Americans elsewhere, but the racial gap is even larger in D.C. than it is nationwide.
Myth: It makes at least as much sense to attribute the exit poll discrepancies to fraud (or miscount), as opposed to error in the poll.
If exit poll discrepancies favoring Bush ("red shift") evince fraud or miscount, and if this fraud or miscount varied from place to place, what might we expect? Bush would probably do better than pre-election expectations in states with lots of red shift. And he would probably do better than he did in 2000 or 2004 in red-shift precincts and states. He didn't. (Here is an analysis of red shift vis-a-vis pre-election polls -- using "TruthIsAll"'s pre-election measure, which most people consider biased in favor of Kerry. And here is an analysis comparing 2004 to 2000.)
It's perfectly possible that some exit poll discrepancies are due to fraud or miscount. But these results strongly suggest that poll error is a large part of the story. Moreover, some of the known flaws in the 2004 election -- such as the egregious misallocation of voting machines in Franklin County, Ohio -- could easily not show up in the exit poll at all. A fixation on "red shift" is misguided.
--
Considering how long this piece is, it's scary to think how much longer it could be. For further elaboration on the topics here, you might check out Mark Blumenthal's four-part series on RFK Jr.'s Rolling Stone article, or my (now slightly dated) paper "Beyond Exit Poll Fundamentalism". Febble wrote several classic diaries on this subject.