Much has been written about the ever-so-close race for governor in Washington, still undecided almost a month after Election Day. As it happens, I contributed not a few of those words, writing a
series of daily updates during the protracted initial count as well as
a couple of reports about last week's mandatory machine recount.
As you know by now, after the 39 counties counted 2,883,499 ballots in the initial tally, Dino Rossi (R-wingnut hiding behind bland salesman facade) held a slim 261-vote lead over Christine Gregoire (D-wimpy state AG running weak and unfocused campaign). Libertarian Ruth Bennett, running on a platform of support for gay marriage, took about 2% of the vote. Following a mandatory recount, in which ballots were again fed through the tabulating machines, the margin had narrowed considerably, to a mere 42 votes.
Before the next acts of this saga -- certification of the result by SoS Sam Reed, followed by Gregoire's request for a hand recount -- unfold, I here present a wrapup of what changed between initial count and machine recount.
First, some background and definitions...
All county totals presented in this report come directly from the Washington Secretary of State website. I captured them from the SoS site at various points in the process, and most of the analysis here is based on spreadsheet comparisons between those snapshots of the data.
- Ballots counted refers to the total number reported by a county in the initial count
- Ballots recounted obviously is the number processed during the machine recount
- Margin is the difference between the county totals for Rossi and Gregoire at a specified point (either the initial tally or the recount)
- Votes gained/lost by a candidate is the change in the candidate's total tally from initial count to recount
- Margin increased/cut is the change between the initial margin and the margin after the recount
The recount increased Gregoire's total vote by 1289, Rossi's total by 1070, and Ruth Bennett's by 70. Looking only at the votes for the three of them, that's 53.1% for Gregoire, 44.1% for Rossi, 2.9% for Bennett. Gregoire and Bennett both did better, and Rossi did worse, in the recount than their overall percentages (48.9%, 48.9%, and 2.3%, respectively).
We can see that, for the state as a whole, the recount decreased Rossi's margin by 219 (=1289-1070) votes. In many counties, however, there was essentially no difference between the initial and recount results. Three counties (Skamania, Stevens, and Yakima) reported precisely the same numbers in both counts -- same totals for Gregoire, Rossi, and Bennett, same number of ballots tallied. Makes me wonder whether they did anything at all during their "recount".
Another 22 counties produced count-to-recount differences in single digits; changes in the counts for the three candidates and for total ballots tallied were all under 10. This group included some of Washington's largest counties. For instance, in Clark County (across the Columbia from Portland, fifth-largest in the state), Gregoire gained 3 votes while Rossi added 7, thereby increasing his margin by 4; the county tallied 9 additional ballots in the recount. Corresponding figures in Thurston County (Olympia, seventh-largest county) were Gregoire +2, Rossi +4, Rossi margin +2, total ballots +5; in Whatcom County (Bellingham, on the Canadian border, eighth-largest), it was Gregoire +3, Gregoire margin +3, total ballots +4.
Rather strange reports came from Douglas, Pend Oreille, and Whitman (all small, all east of the Cascades). Each showed little change in the individual candidate counts, with the margin changes coming to Rossi +2, Gregoire +4, and Rossi +5, respectively. But each of those counties reported counting far fewer ballots in the recount than in the initial count. In Whitman County, it's clear that there are multiple problems. It's not just that their count of recounted ballots exactly equals the sum of the tallies for the three named candidates (implying that there were no blanks, no write-ins, no overvotes, no undervotes ... in the original count, they had 519 such ballots). On top of that, the count of registered voter was shown as 21,092 throughout the initial count but displays as 27,310 in the recount. How did the voter pool suddenly jump by 29%? The county reported 85.7% turnout originally, but now is listed at 64.2% (almost 10% lower than any other county). In terms of not-for-the-three-candidates ballots, Pend Oreille (158 originally, 2 at recount) and Douglas (246 originally, 22 at recount) are similar to Whitman, though the rest of their numbers look right. Combined, those three counties reported processing 898 fewer ballots in the recount than in the first count, while the three major candidates combined to receive 1 more vote than in the initial tally.
That leaves 11 counties with relatively large gains or losses for individual candidates. Let's review them in ascending order of their effect on the Rossi-Gregoire margin.
Snohomish County recounted 223 more ballots than their initial total, and 271 more ballots were found to be votes for a named candidate than previously. Even with that much action, however, the overall margin changed by exactly 1 vote; Rossi gained 130 votes while Gregoire gained 131.
In Kittitas County, all three candidates came out of the recount with additional votes. Overall, however, adding 19 Gregoire votes and 26 Rossi votes resulted in an increase of 7 in Rossi's margin. The county counted 34 additional ballots in the recount, and the named candidates combined to gain 47 votes.
The oddest duck in the state was surely Cowlitz County. Gregoire netted an 11-vote gain on Rossi in the county, but that's because she lost fewer votes than her opponent did. The Democrat's total decreased by 29 votes while the Republican lost 40. The total ballots recounted came in at 99 fewer than on the initial count. This is not a misplaced column, as I suggest may have occurred in Douglas, Pend Oreille, and Whitman. Cowlitz still reported nearly 1400 ballots cast without choosing one of the three named candidates. Were I a Democratic honcho planning to request a partial hand count, I'd be looking very closely at Cowlitz County. In addition to the recount oddities, it's also one of only eight counties where Gregoire (barely) outpolled Rossi.
Recounts in five counties -- Spokane, Adams, Grant, Skagit, and Pierce -- each increased the Rossi margin by somewhere between 13 and 18 votes. In all of them, both candidates gained votes ... it's just that Rossi gained more. All five counties recounted more ballots than they had originally, with increases ranging from 7 to 147 ballots.
Walla Walla County may have been the biggest positive surprise in the recount. In this heavily Republican county (overall, Rossi leads 63%-35%), Gregoire gained 59 additional votes in the recount, to just 37 extras for Rossi. That's a 22-vote change in the margin, in the Democrat's favor! The county found 123 additional votes in the recount, even while processing 2 fewer ballots than previously.
Kitsap County, one of the last to report its recount results last Tuesday, was one of only four counties in which one major candidate gained votes while the other lost some in the recount. It's the only one of those four in which the change made an appreciable difference. Overall, Kitsap added 15 votes to Gregoire's count while decreasing Rossi's total by 19, resulting in a 34-vote change in the margin between them. These changes came while processing one more ballot than in the original count.
Then there's King County, with its nearly 900,000 ballots. Rerunning all of those ballots (the recount processed 336 more ballots than the initial count) through the op-scan readers resulted in the addition of 348 to Rossi's total, 593 to Gregoire's. That's a net +245 for Gregoire.
Combining all these county-by-county results, we see that King County, all by itself, nearly erased the entire initial margin. Had Gregoire been able to gain 17 votes in the other 38 counties -- that's less than half a vote per county -- she would have come out ahead in the recount. Instead, she had a net loss of 26 votes in those 38 counties (just over 2/3 of a vote per county). That's how microscopicly close this race is.
One would expect that a recount would result in at least as many ballots processed as in the original count. Yet four counties reported appreciably fewer votes on the recount. I've already remarked about those unexpected numbers, but must reiterate that I believe that three of them are simple, easily correctable, and probably innocent mistakes. Cowlitz County remains a mystery to me; I can't understand how they ended up with 99 fewer ballots on the recount. The overall effect of the odd results in those four counties is that the recount seemingly processed 158 fewer votes in the state than in the initial count.
Much speculation has centered on the possible effect of the Libertarian Bennett in the race. Her 2.3% may (or may not) have come from a different portion of the electorate than most of her party receives; she is, after all, a gay rights activist, a strong proponent of gay marriage, and a lesbian. While that will long be debated, I observe that she reached 3% of the vote in only three counties -- rural and conservative Ferry and Skamania, liberal San Juan. Make of that what you will. In most counties, however, none of the above (blanks and/or write-ins) outpolled Bennett. That includes almost 23,000 King County voters, as well as nearly 9500 (5.5%) in Clark County.
Other diaries here on dKos have discussed strategies for the recount request by Gregoire, so I won't go into that here. Nor have I undertaken sufficient study of the types of voting systems employed by the counties (paging RonK...). The great differences in numbers of added votes may be, at least in part, an artifact of op-scan vs. touch-screen vs. punchcard vs. whatever, to say nothing of the different product lines within each general form of ballot-tallying system.
Once the hand recount begins, expect to see me spouting off once again ... like it or not.