(This post was co-researched by Carla of Preemptive Karma and Torrid Joe of Also-Also)
We want to begin this second part in our series by noting that we believe Washington State has room for improvement when it comes to having a well run and error free election. Unfortunately, this can also be said for many other states in the Union, and reflects a wider problem of accountability and attention to detail in our election systems that must be substantially addressed.
(more after the jump)
To briefly account for the King 1,860:
· 1,217 was the adjusted discrepancy initially reported by King (a figure which seems to no longer exist at the Elections website, but seems consensually accepted as the initial figure).
· Due to an identification of 636 duplicate entries in the voter registration files, the discrepancy was then boosted to 1,853.
· Of the 348 misfed provisional ballots cited by the County, 341 were conclusively identified as either valid or invalid registrations. 252 of those were deemed properly registered, credit for voting thus assigned. The seven unidentified provisionals are added to the discrepancy list, yielding 1,860.
We're certainly not going to claim that 1,860 is the precise final accounting of the voter discrepancy in King County. Our contacts with Bill Huennekens of King County Elections yielded "essential confirmation" of that number, however. While KC acknowledges that 348 provisionals were accidentally fed through the Accu-Vote system before verification, they also believe that their recognition of those 348 constitutes an accounting of those votes, to the extent that they explain why voter credits for those ballots were not found.However, Huennekens admitted it was fair to include the remaining seven unresolved provisional ballots as part of the total discrepancy, as we have done. The release of King County's post-mortem report today highlights the "provisional errors," among others. As they indicate in the report, they faced challenges and did not always meet them. But we're not convinced that King County's problems are outsized or endemic.
This is a key point for Rossi. His strategy seems to revolve around the "nobody knows who really won" argument. He said just that in his press conference on Feb. 8th: "I believe we have more than enough evidence to show, really, that we'll never know who won this election." His strategy relies on the abnormality of error to provoke an annulment of the election. We find it extremely difficult to believe that any jurist would set aside an election occurring within a normal rate of error even if the extent of the error seems surprising. Election issues are to be fixed by the legislatures in most cases, after all.
Rossi must therefore prove that this election was absurd, crazy and possibly a suspicious abrogation of both custom and practice. They've chosen to zero in on King because it easily contains the largest number of mistakes, discrepancies and messy bureaucracy. Stefan Sharkansky rhetorically hammers this point repeatedly: King can't be trusted.
Or can it? We wondered whether King was truly singular in its administration of the election. So we set out to compare errors as a ratio of all ballots, for 14 other counties featuring a cross-section of the Washington electorate. Respected Washington journalist David Neiwert was kind enough to suggest a list of 12 for us. We contacted each of the 12 counties for their numbers. We augmented those results with four counties that Sharkansy provided to his readers at Sound Politics (we hope he got them right).
View table here(Excel required)
*[These numbers were obtained directly from county election officials in all cases except Snohomish, Stevens, Jefferson and Pierce, where we provisionally accept and reprint the figures shown by Sharkansky at SP. In some cases the Secretary of State's listing of the manual recount ballot totals were used to extrapolate the number of registered voters. Since all variance counts the same, we chose to assume that the discrepancy was more ballots than voters when not otherwise known.]
Notice Spokane County. Rossi won this county easily. Total voters are fewer than one fourth of King County's. But they have twice as many errors as King per voter. Moreover, their discrepancy represents more voters than ballots, a condition which Sharkansky claimed lacked plausibility as a type of honest error. However Spokane is apparently a Republican leaning county--is it therefore not worth mentioning? Are their errors not so egregious?
Now notice the pattern within the table: Handling fewer than 150,000 ballots seems to make reconciliation a lot easier. The three counties with perfect accounting don't even total 75,000 ballots between them. This is one-twelfth the number of ballots King had to count. Problems of scale seem to crop up at around the 200,000 voter mark, as you examine the table.
Next we figured we'd better validate our visual clue that population size could be connected to error rate. We ran a Pearson's correlation on ballots cast and variance to voters credited. On a scale from 0 to 1, where 1 represents a perfect positive relationship (ie, as total ballots rise, so does discrepancy), the set of variance figures here yields a very strong score of .91. The margin of error--that is, the chance that this correlation is the result of random chance--is less than 1 in 1,000.
There are always caveats. Having one county too far away from the rest of the group can increase the appearance of correlation. To try to correct for that we ran two additional correlative types, both of which showed lesser but very strong relationships. To include all 39 counties would be ideal, but a sample of 16 represents more than a third of Washington's counties. Finally, the old standby: correlation is not causality.
We believe that the difficulties surrounding King's effort reflect poor planning and inadequate training. But a natural tendency for error is also reflected when handling large numbers of ballots. Problems are further exacerbated by a record number of registrations, a record number of absentees, a compressed general election cycle, a first-ever manual recount in King, and registration management technology that was rushed into service and brand new to users.
We wonder if Rossi will continue to make "nobody knows who really won" the focus of his contest. Especially given the relative harm of such unattributable error. Discrepancies do not necessarily produce illegal ballots. And now that the counties are all dismissed it's much more difficult to paint them (King, specifically) as having messed up the election.
The counties can and likely will be called as witnesses and will have to provide records for discovery. But the GOP has lost the technically adversarial relationship with King, something it might like to have kept.
We also believe Rossi is no longer finding Sharkansky and Sound Politics a net positive, except to keep the faithful in a froth. The state GOP already has the list of felons and deceased. They will most certainly bring that to the forefront should they need to show proof of serious irregularity.
If Rossi is truly finding SoundPolitics less useful these days, their habit of serial conclusion-jumping and lackluster fact checking are likely just some of the reasons.
Carla and TJ