While I’m sure Farhad Manjoo believes everything he says, his article is problematic. But before going into details, here are some opening observations.

• A number of Manjoo’s rebuttals invoke the official conclusions of the Democratic Party, which, while often laying out the same material as Kennedy, always concludes that the election turned out correctly. Is this evidence that folks like Robert Kennedy Jr. are crazy? Not really. Let’s say the Democrats did conclude that their own 2004 research constituted a case for election-swinging fraud. That would mean that Kerry and the party – both of whom refused to challenge the election when it counted – let their own voters down in the worst of ways. Which means regardless of their official conclusions, Democrats can’t cross the “stolen” line (in fact, they must defend that line). Consequently, the party's conclusions about Ohio do not constitute evidence for either side.
  
  
• I read RFK Jr.’s article before Manjoo’s. If Kennedy portrayed his article as “new and earth-shattering” dirt on the election, as Manjoo suggests, I didn’t pick up on it. Manjoo here is pulling a slight of hand, dismissing all past questions before explaining himself (his tone suggests it is beneath him to confront Kennedy’s "old" arguments, setting them in a tawdry light before feebly engaging them).
  
  
• Most revealing, a close read shows Farhad proves or disproves nothing. Indeed, all he does is patch holes in the official story using rhetoric that suggests he has debunked serious questions. This is a particularly pernicious tactic in DC – don’t prove your explanation, just produce statistically unlikely scenarios to obfuscate holes in that explanation (Republican machinations defending their tax cuts as help for the poor come to mind).

Now let’s look into the details:

CLAIM / REBUTTAL:

There were anomalous voting patterns, based on down-ballot results.

Manjoo refutes questions raised by Bush voters’ apparent support for a down-ballot liberal judge by comparing the unusual correlation to a similar example in the 2000 race (in some counties in 2000, voters went more for a Democratic judge than for Gore). First, using 2000 as a control given what happened in Florida demands at least a caveat, which Manjoo fails to provide. But even if we assume there was no chicanery in Ohio in 2000, the comparison may still be apples to oranges. In the examples he gives from 2000, Majoo does not include analysis of the ideological proximity of the 2000 down-ballot judge to Gore, which is part of Kennedy’s argument in his 2004 example. Therefore, Manjoo must be assuming that a lack of voter knowledge about down-ballot candidates just happened to run upstream with Bush in the official results (and pretty much says so). This makes for a weak case. Even if we accept Manjoo’s explanation as plausible, it in no way disproves Kennedy. Manjoo has simply come up with a statistically unlikely alternative to Kennedy’s interpretations.

CLAIM / REBUTTAL:

Blackwell scrubbed voters.

Manjoo defends against the acknowledged voter scrub by pointing out that current Blackwell-backed Ohio law scrubs at a higher rate legal voters who choose to participate less often or move more often – a.k.a. Democratic and minority voters. Well my goodness, if the law is used to suppress voters, then it must not be suppression, right? Manjoo is historically thick here – or contemporarily callous.

CLAIM / REBUTTAL:

Long Lines.

Here Manjoo blames documented long lines on the ineptitude of Ohio government officials. This is a stretch given that the “ineptitude” came from employees of a partisan Secretary of State and benefited that Secretary’s party. Oh, but Manjoo tells us that Bush and Kerry voters were equally affected by long lines. This is flat out wrong, and it is not clear on what Manjoo rests this claim. There is a mountain of evidence showing the disparity of lines to be radical depending on the likely disposition of the voter.

CLAIM / REBUTTAL:

Exit polling means something.

Manjoo claims that disparities between exit polls and official results are cause for further scrutiny in the Ukraine but not in the US; the disparities in the US were closer to the polling margin of error, he argues. But what Manjoo largely ignores is the more salient point that exit polls were wrong in one direction – favoring Bush – in not just one US state, but in multiple large battleground states across the nation. Statisticians would call these nonrandom results. Moreover, the fact that polling discrepancies were concentrated in certain counties and precincts within battleground states, while largely absent in others, seriously undermines the hypothesis that Bush voters avoided or lied to pollsters. (That Republicans lied where it mattered and didn’t where it didn’t would require a conspiracy to rival all conspiracies – and to what end?)

Here it is important to note that Majoo’s argument, at root, absolutely requires buying into the hypothesis that Republicans in key electoral locations were more likely to lie than Republicans in non-contested areas. Of course, there is no evidence for this. This major pillar of Manjoo’s argument is simply an untested hypothesis that is, statistically speaking, highly unlikely.

For Manjoo, long lines didn’t disenfranchise enough people to swing the election. Lines and voter purges, together, didn’t disenfranchise enough people to swing the election. Lines, purges and voter challenges didn’t sway the election. Lines, purges, challenges and seemingly partisan machine failure didn’t sway the election. Where does it stop? All of these things happened to one degree or another – that is beyond dispute even for the Democratic Party, behind which Manjoo frequently stands. Manjoo is just arguing degree, and his method is to chip away at the corners of massive, throbbing outrages by way of strategic Republican lying and other crackpot flights of fancy.

I challenge Manjoo to prove his hypothesis about selective Republican lying. The outcome of his proof will tell us whether or not we live in a democracy. The burden is on Manjoo, not Kennedy.

...to God, Joan.

Can't you just show Farhad the door?

I never thought I'd see the science of statistics receive such interest. What's next? The Chi-squared camp start fighting the maximum-entropy camp in a bid to see who really controls the normal distribution?

It seems unlikely that the mathematical parts of the arguments are going to be decided on the Salon letter pages. It's notoriously difficult to build a good intuition of margins of error and what happens when data lies outside of them. For starters, the margins of error are themselves prone to large errors. It's not uncommon to find results that are massively outside of the error estimates- for the simple reason that the error estimates themselves are often wrong.

In cases like these, the best that us normal Joes and Janes can do is often to try and understand what the expert concensus is amongst professional statisticians who have studied the matter.

It's true that you can find reputable statisticians on either side of this debate- but Manjoo indicates that the majority of recognized experts have not found significant proof to doubt the result of the election.

Does that mean that the election wasn't stolen? No, but extroadinary claims require extraordinary evidence. Kennedy is certainly making the extraordinary claims- but unfortunately he doesn't appear to have the cast-iron evidence required to substantiate those claims.

Others are welcome to continue investigating this matter- but they haven't yet produced the corpse.