The guffaws from the right were almost audible in cyberspace when King Executive Ron Sims claimed their electoral process rivaled any bank's. Daniel at On the Road to 2008 did the best job of supporting that seemingly outrageous claim, and Goldy at Horse's Ass and I both covered it as well. Further support was provided by the elections team of Los Angeles County, a jurisdiction much larger but generally comparable to King:
[Registrar Conny] McCormack said she was impressed with King County's inaccuracy rate of only 0.2 percent. "We've never hit 0.2 of a percent (in Los Angeles County). I think that is excellent," she added. Los Angeles County's error rate in its voter-crediting process for the 2004 election was about 1 percent of its 3.1 million votes. [emph mine]
Goobernatorial Don Quixote Stefan Sharkansky discounted that praise, generally on the basis of LA being even more screwed up than King. But we turn back towards Cali--at least their newspapers--to receive even more confirmation that perhaps Sims was not simply blowing sweet air up our collective skirts:
Although it is a deeply held American principle that every vote counts, research has found that 1 out of every 100 ballots cast in the 2004 election for president weren't counted.
...
In an analysis released earlier this month, [CalTech/MIT] researchers concluded that the presidential election of 2004, in which 1 in 91 votes didn't count, was run "much better" than the 2000 election, in which 1 in 53 ballots didn't count.
"In a close election, no level of accuracy is quite enough, is it, short of perfection?" asks Kirk Wolter, a University of Chicago statistics professor who supervised the team hired by the media to review Florida ballots after the Bush-Gore fiasco of 2000. "There are huge potential error rates, and the reason we've lived with it so long is because very few people knew about it, and elections weren't that close. ... Now we know better."
There are two salient points to take from this analysis: first, electoral error rates can run to at LEAST 1%--and I say at least, because the figure cited only refers to votes not counted, leaving out invalid votes that are accidentally counted. In fact, as the article points out, "The National
Commission on Election Reform has recommended that states reduce their error
rates below 2 percent no matter what mechanism they use [emph mine, again]." Below two percent? So what does it say about a county with a new database management system and record numbers of voters, when they manage to achieve an "error rate" about one-TENTH of the national standard?
Secondly--and this has great relevance to any case the Rossi team might advance as to the uncommon irregularity of their election loss--the only reason anyone can profess to be shocked at the level of inaccuracy is because they haven't been paying attention, or the races have never been razor thin as the governor's race was. Mary Lane, John Carlson, Sharkansky and the rest of the echo chamber want Washingtonians to believe that an unsurpassed level of sloppiness and blithe inexactitude is responsible for a complete mistrust in the results. One thousand voting felons!!! Accepting that at face value for the moment, is that good or bad? How many felons voted in 2000? How well do other rights-restorative states do? Was the process fixable under current law?
These questions all speak to the concept of irregularity, and the prospect that setting aside the election would yield a stronger, more trustworthy result. Would it? The staffer for the Chron nails the answer:
Statistics has a name for such outcomes: ties. Nobody will ever know who really won because a recount is no more scientific than a coin flip. Yet inevitably, someone must win. The loser is left to complain about voting irregularities, suggest the election was stolen, perhaps even take it to court.
A national accuracy goal of 98%. A 2000 national reality of 99%. King's reality of 99.8%. Chew on that for a while.
Update 1130 2/28
Thanks to Goldy for linking over here; it drew a few good comments before devolving into the typical HA comment thread. (No offense Dave; I think you like it that way). I think it's prudent to point out that I shouldn't imply a direct comparison between the two "error rates" on votes, and the voter-credit shortfall in King. (The comparison in LA is direct, however--King's was one fifth of LA's.)
What the MIT/CalTech study and the NCER goals are referring to specifically are "residual votes," mostly those that go through a machine and aren't counted because not enough circles are colored in or too many or there are smudges or it got bent or even that they just didn't vote that race. The point being made there is that no more than 2% of votes should be lost due to counting error, expressed here in terms of machine sensitivity.
The larger implication is obvious: regardless of the type of error, whether residual or unreconcilability or voter-credit--the FEC is telling us that losing 2 votes out of 100 isn't necessarily that bad. Furthermore, King's figure of 1,860 is indeed just a voter credit shorfall, but it does set a rough upper bound for total vote error. I don't think there's much question at all that the precinct reconciliation discrepancies total up to a good bit less than that; one former King election official cited at least once by SoundPolitics indicated he believed that would be true. But assume it's just less, by enough to contain all the non-discrepancy errors--the provisionals, the felons, the dead people. You're still looking at a final error rate of .2%.
What about the actual residual error in King's original count? Well, here King benefits from the manual recount. Because the county was able to review tens of thousands of optical scan ballots by hand and determine voter intent, they reduced their final certification residual error down to negligibility. That's what the process was for, of course--to catch any ballots that could be recognized as votes, but were missed by the machine. And that's the definition of "residual error" right there. That's an interesting question to ask the King people next time I talk to them, what their total spoiled ballot rate was.
--TJ
This seems to be the key point to me: Gregoire is just as legitimate a governor as would be Rossi had he maintained his recount leads. It's not that difficult a principle to understand, unless you write for Unsound Politics.
Posted by: eugene | March 01, 2005 at 00:19