As I began this diary, Supreme Court oral arguments about the disputed King County ballots had just ended. King County was due to announce its results,
sans the disputed ballots. And Democratic state chair Paul Berendt had leaked that King's results (without the disputed ballots) showed Christine Gregoire (D-ineffectual Locke-lite wimp) taking 154,766 more votes in King County -- up from a mere 154,709 in the first recount -- than Dino Rossi (R-smarmy, unctuous snake oil salesman), resulting in an
8-vote margin in her favor.
Is that enough drama for you, 3 days before Christmas?
Over the last 6 weeks(!), I've written extensively about this race, so I hope my analysis won't get lost in the blizzard of diaries about the final(hah!) outcome. Click here to link to my most recent previous report, which in turn leads to further links to the series.
While I'll discuss the day's new totals a bit, I'd rather continue with some deeper analysis. Let others exult ... while doing so, I'll also try to inform.
Analysis below the fold...
OK, so it's actually 10 votes, not 8. Probably not enough, in and of itself, to fire Berendt.
King County reported its preliminary hand recount results late this afternoon. Gregoire gained 47 new votes, while Rossi lost 12 and Libertarian Ruth Bennett lost 10. The county, which uses Global AccuVote opscan technology, tallied just 59 more ballots than in the machine recount. Coincidentally, 59 is also the margin by which Gregoire increased her lead in King County, overcoming the 42-vote edge for Rossi in the machine recount and the 7 additional votes he had gained in the hand recounts conducted in the other 38 counties. These results do not include approximately 723 ballots held up by Chris Vance and the state GOP until the Washington Supreme Court unanimously dissolved their TRO this morning.
For the state's hand recount as a whole, Gregoire added 609 new votes, Rossi's total increased by 557, and Bennett gained 24 additional votes. Reflecting, in no small measure, the efforts by Dean Logan and his staff to begin the process of ascertaining voter intent between the initial count and the machine recount, King County's new-found rate during the hand recount was extremely low, just 0.28 additional for-a-candidate votes per 10000 ballots.
In previous analyses, I've looked at the number of new-found for-a-candidate votes in comparison to the total number of votes cast in the county. Those votes are almost entirely ballots that the machine tabulating systems were unable to recognize as valid votes, whereas human beings (always at least one Republican, one Democrat, and one county official) could reach consensus that said ballot showed clear intent on the part of the voter in favor of one candidate. In his reports from the King County counting floor (here, here, and here), switzerblog has given us some great insights from the front lines.
The statewide rate of new-found votes in the hand recount is defined as:
[(3-candidate vote in hand recount)
minus
(3-candidate vote in machine recount)]
divided by
Ballots Counted in hand recount
For ease of visualization, I multiplied the result by 10000
Under this definition, the statewide new-found rate was 4.13 per 10000 ballots, representing 1190 new-found for-a-candidate votes out of 2,884,217 ballots tallied. The highest new-found rates were seen in Asotin (16.7/10000), Franklin (15.1/10000), Pierce (14.1/10000), Kitsap (13.5/10000), and Okanogan (11.4/10000); of these, only Pierce and Kitsap have large enough populations to mean much in the state totals. Six counties reported fewer for-a-candidate votes in the hand recount, led by Adams at -9.6/10000 (that's based on -5 votes out of 5204 ballots) and Skagit at -4.6/10000.
After thinking about it for a couple of days, I realized that there's another way to look at the effects of the hand recount as compared to the machine recount. Instead of using the total number of ballots tallied as the denominator, perhaps it would be preferable to base the analysis on the number of ballots that the machine recount had failed to assign to one of the three named candidates. In essence, this would remove the great bulk of ballots -- those where there was already a valid candidate-vote -- from the denominator of the fraction. Instead of a new-found vote rate, as described and reported above, we're now calculating the conversion rate ... the proportion of previously unassigned ballots that were identified as for-a-candidate votes during the hand recount. The equation for this quantity is as follows:
[(3-candidate vote in hand recount)
minus
(3-candidate vote in machine recount)]
divided by
[(Ballots Counted in hand recount)
minus
(3-candidate vote in machine recount)]
This approach assumes that previously-valid votes for a candidate were once again counted for that same candidate. While there's no way to tell whether that's always true (in fact, it certainly isn't, as we've already seen that candidates lost votes in a few counties), I believe the assumption must be correct for the vast majority of ballots.
That said, based on current statewide figures, 1195 new-found for-a-candidate votes were uncovered during the hand recount, out of an estimated 75,876 ballots not previously assigned to one of the three named candidates. That comes to 1.6% of the tallied ballots that hadn't been cast for Rossi, Gregoire, or Bennett ... ballots that machines couldn't recognize as votes for one of those three people, ballots on which a small group of human beings could reach consensus as to the voter's preference among the three candidates.
The rank-order of these two quantities is roughly similar (Pearson correlation = 0.734), there are some interesting anomalies. For instance, 2 new-found votes were recorded in Douglas County (1 each for Gregoire and Rossi), out of 13,263 ballots. Thus, their new-found rate was rather low, 1.5/10000. But, based on the two recounts, the county started with only 19 unassigned ballots, giving them a conversion rate equal to 10.5%, highest of any county in the state. Intriguingly, Douglas County's initial ballot tally (but not votes for the named candidates) was over 200 higher than was reported in either of the recounts.
Two other counties, Kitsap and Pierce, were able to convert 6.5% of their previously-unassigned ballots to votes for candidates. You'll recognize that they were also on the list of the highest new-found vote rates. Asotin County, with the highest new-found rate of any county, was the only other county over +4% in conversion rate. Adams County, where Rossi lost 5 votes, had a -4.4% conversion (dissipation?) rate. As might be expected, with few new votes in a very large county, King County's conversion rate was a very low 0.1%.
King County's final figures -- with the erroneously-untallied ballots reinstated by the state Supreme Court -- are due to appear tomorrow morning. More reports forthcoming...
Final personal aside -- I didn't really study the list of 573 untallied King County ballots, other than to confirm the presence of County Council chair Larry Phillips and black box voting activist Andy Stephenson. So it came as something of a surprise to learn today that my colleague in the cubicle right next to mine is also on the list! She is deeply offended that her vote hasn't (yet) been counted. She also told me that she got a note some weeks ago, vaguely suggesting that she should sign a form and mail it back to the county. The letter and form were so "generic" (her term) that she was quite reluctant to respond to something that seemed so non-official.
For the record, I have not asked her who she voted for in the gubernatorial race. That her ballot will be counted is far, far more important.