Oh, no! Not another "Kerry won" diary!
by Mathwiz, Fri Apr 01, 2005 at 01:25:49 PM EST
Well, not really, but this report from USCountVotes has gotten surprisingly little attention, even in the left blogosphere where you would think it'd be most warmly received.
In this report, USCountVotes analyzes the claim that the discrepancy between the 2004 Presidential vote and the exit polls was caused by Bush voters' greater reluctance to be interviewed by exit pollsters. That claim is found severely wanting. Instead, the evidence suggests the worst discrepancies were in GOP strongholds, which is consistent with (but doesn't prove) the hypothesis of widespread GOP tampering in areas they control. It would also be consistent with the observation that GOP-leaning demographic groups, like the fundies, appeared to swing even more toward the GOP in 2004 than in 2000.
So why does it matter, especially at this late date? Well, for one thing, we Democrats have been running around like decapitated chickens since the election, trying to figure out where John Kerry went "wrong," so we can avoid the same fate in 2008.
However, this report would imply that Bush's margin of victory was much less than the official figures indicate, and that, had the vote been accurately tallied, Bush may well have lost the electoral college. If so, perhaps we've been too hard on Kerry (except, of course, for criticizing his quick concession). Instead, perhaps we need to concentrate more effort on insuring our elections can't be tampered with.
(On the other hand, the report supports the conclusion that the Democrats lost the Senate races fair and square. So our criticisms of the losing Senate Democrats, as well as the DNC's Senate strategy, were apparently untainted by inaccurate info.)
There was approximately a 5.5% spread between the weighted, but unadjusted, exit poll results, and the official results: the exit polls showed Kerry headed for a 3% popular-vote victory, while the official result was a 2.5% Bush win.
There are three possible explanations for this (note, however, that they are not mutually exclusive):
- Statistical Sampling Error (i.e., bad luck);
- Inaccurate Exit Polls; and
- Inaccurate Election Results (which could include everything from "spoiled" ballots to deliberate fraud)
My own view has been that both of the other explanations are partially valid, with exit poll inaccuracy accounting for somewhat more of the discrepancy than election shenanigans. However, the official position of the exit pollsters has been that exit poll inaccuracy accounts for the vast majority of the discrepancy. This report bolsters my view at the expense of the official one.
But how can we square the hypothesis that at least some of the error was in the official election results, with the failure to find any discrepancies where recounts were conducted in New Hampshire, New Mexico, and Ohio?
Well, if, and to the extent that, deliberate fraud was involved, there is one type of fraud that would likely go undetected by precinct recounts - especially if, as in Ohio, the recounted precincts were often hand-picked, rather than being selected at random (as was supposedly required by Ohio law): central tabulator fraud, in which county totals, rather than precinct totals, are manipulated.
Central tabulator fraud is obviously impossible if every precinct total in the county is accurately reported, because the numbers literally wouldn't add up. But if only some of the precinct totals are reported, a discrepancy wouldn't be obvious, and a recount obviously wouldn't show anything wrong.
The hypothesis of central tabulator fraud would also fit with the observation that hand-counted paper ballots were the only voting technology where the exit polls were relatively accurate.
If so, the 2004 Presidential election was, in fact, probably as close as the 2000 election was.
Very, very late update: About three weeks after this diary was posted, Febble at Daily Kos posted a very interesting diary, which goes a long way toward explaining the USCountVotes results without invoking fraud. Her diary is very technical and the results very counterintuitive, but suffice it to say that a non-obvious secondary statistical artifact can arise from an exit-poll bias which varies randomly from precinct to precinct, but which favors Kerry overall. This would account for the disproportionate exit-poll discrepancies in the heavily-GOP group of precincts even in the absence of fraud.
Unfortunately, by the time I became aware of Febble's contribution, I'd, um, forgotten I'd posted this diary on the USCountVotes study, so I didn't post this update for several months. My apologies.
I don't think Febble's results change my conclusions much. I do now think it unlikely that there was enough fraud in the 2004 election to have flipped Ohio, and with it the electoral college. Nevertheless, I remain convinced that our electoral system remains all-too-susceptible to fraud, and that this must change for Democrats to have a fair shake in close elections. Unfortunately, it's a catch-22: Democrats first have to win more elections before we can change the system.
As I post this, the GOP is self-destructing with the Abramoff scandal, Plamegate, the Katrina aftermath, and Bush's illegal wiretaps; and Dems held their ground in November 2005. So maybe we'll finally make up some lost ground in 2006. Let's hope so - 2008 may be our last chance to win a close Presidential election. After that, short of a landslide, we may not stand a chance.