Crossposted at PreemptiveKarma and also also
A reporting error in Friday's co-published article on county "discrepancies" in the contested Washington election, requires several things: a correction, a retraction, an apology, a restatement of the facts and a firming of the premise laid out in the original article.
(more after the jump)
The first county that provided information to TJ at AlsoAlso on the subject of ballot discrepancies was Spokane, whose Paul Brandt responded immediately and specifically. Noticing that the discrepancy was the more unusual "voters without ballots," TJ called back and confirmed the numbers which ended up in our original article. We then moved on to collect similar information from other counties across the state.
During today's presentation by Dean Logan in front of King County Council, a by-county comparison of discrepancies listed Spokane with what appeared to be a total of 77. In checking the information, it was discovered that what was initially reported as a 976-ballot shortage was an unadjusted number. The county recorded 1,052 absentee ballots mailed after the deadline, but awarded "voter credit" to those voters so as to maintain them as active registrants. Stripping these "courtesy credits" is necessary to attempt an apples to apples comparison of discrepancy. King County has not used the credit system since 2002. To repeat, the initially reported discrepancy for Spokane was accurate, but we failed to fully refine our question to them, so as to receive the appropriate information.
Spokane's adjusted variance ratio is .04%, much lower than the originally reported .47%. They are in fact one of the counties more able to fully reconcile their tables, particularly given their population size relative to the others. AlsoAlso and Preemptive Karma regret the error, and retract the conclusion that another county had a higher rate of comparable discrepancy than King. So far in our analysis, none has. Allegations of anyone ignoring data for partisan reasons have also been redacted from the original text.
Interestingly, when told over the phone of King's report about Spokane, the Spokane representative clucked a bit and said something to the effect of, "See,that's why we don't recommend trying to compare these things..." Logan's presentation to Council stressed the non-regulatory, unusual circumstance of a "reconciliation" conducted 1/3 on poll night, the rest several weeks later. We have discovered in this exercise the folly of an overly serious look at the ballot/voter variances. To do so is to admit that they are legally or administratively relevant. Comments to the originally posted article have in fact stressed this theme. We clearly have some smart folks here.
Another comment that came up was the request to run a linear regression on the relationship we discussed: the hypothetical that the number of total ballots has a material effect on ballot variance. In light of the changed information, King's total of both ballots and discrepancies were both so far ahead of the rest of the pack, that we knew it would distort our attempts to see an honest correlation, making the relationship appear much stronger than perhaps it is. So we re-ran the Pearson and two non-parametric correlations, and then added a simple linear regression, but without King County's numbers, to be more confident of the results. The adjusted r-squared value remained a very strong .909, meaning roughly speaking that the number of ballots is responsible for about 90% of the variance in ballot/voter discrepancy. The relationship was significant to less than 1 in 1,000, as the possibility the results were random. So despite the change in discrepancy for Spokane, our original conclusions still appear valid, strengthened now by extra research.
It's certainly unpleasant--to be lauded in print by one such as Orcinus as having followed journalist-worthy practices--and not 48 hours later have to print a retraction. We're not journalists, we're not paid, and we're not subject experts. We merely engage ourselves in a subject in order to present an educated review. But we hope if we are honest and forthright with our missteps, you will stick with us as we continue. Zap and Kevin will surely have some useful and illustrative Scripture here; please leave your own verses or other comment below.
--TJ and Carla