The Flathead Valley’s Leading Independent Journal of Observation, Analysis, & Opinion. © James R. Conner.


11 & 19 November 2012

Sandy Welch probably cannot win a recount


Updated 19 November. After the provisional ballots were counted, Juneau’s lead increased from 1,374 to 2,209 votes. I’ve inserted the new numbers in the paragraphs below.

Republican Superintendent of Public Instruction candidate Sandy Welch faces a huge decision later this month, after the official canvass of votes is completed: whether to ask for a recount in her race to unseat incumbent Democrat Denise Juneau, whom she trails by 2,209 votes, a 0.472 percent difference.

Montana allows a recount if the difference is 0.5 percent or less, and pays for the recount if the difference is equal to or less than 0.25 percent. Although Welch qualifies for a recount, she’ll have to post a $115,000 bond for it to happen. She’ll have no trouble raising the money given the ruling — which I consider correct — by Montana’s Commissioner of Political Practices.

Welch’s campaign has already announced it will seek a recount, but that’s more a gut reaction than a considered decision. Right now, her camp is thinking, “Wow. Just zero-point-three percent. Man, that’s a tiny margin.”

But that’s the wrong way to look at it. Percentages matter only in determining whether a recount can be sought, and who pays for it. Otherwise, it’s the absolute number of votes separating the candidates. The key question for Welch is, “Where can I find 2,210 more votes?”

There are only two sources of additional votes for Welch: (1) ballots mistakenly counted for Juneau, and (2) ballots for Welch that should have been counted but were not. So Welch, a former math teacher, faces a problem in probability: what is the likelihood of finding 2,210 votes by a combination of flipping votes for Juneau and finding uncounted votes for herself?

The most likely source of votes erroneously counted for Juneau is optical scan ballots incorrectly read by ballot counting machines. The most likely source of ballots that should have been counted but were not is absentee ballots wrongly rejected by elections officials (that’s where Sen. Al Franken found the votes he needed in Minnesota’s recount of the 2008 election; see Jay Weiner’s This Is Not Florida: How Al Franken Won the Minnesota Senate Recount).

Optical scan counting machines are not perfect. But the error rate is very low, on the order of one in 10,000 or less according to the limited literature on the subject I’ve been able to find. And unless there’s some kind of systemic error — a machine fault, or a programming error — the errors will be largely random, favoring neither candidate.

Counting machines will not record votes when the voter’s intent is clear, but the ballot is not marked so that a machine can read it. For example, a candidate’s name might be circled or underlined, but the oval the machine can read is left blank. In the argot of election administration, that’s an undervote. A hand count of the ballots will recover these votes, which is why hand counting optical scan ballots is the only recount method that makes sense. But the number of false undervotes will be very small.

That leaves ballots cast that never made it to the hand counters in Montana’s paper ballot counties or to the counting machines in Montana’s machine counting counties; ballots cast that were counted but should not have been counted; and ballots not cast because elections officials wrongly refused to let registered voters vote (West Yellowstone?). These are not null sets, but the probability they comprise a class comprising enough Welch votes for her to win a recount seems low. Perhaps Welch and her recount wizards can put a number on it.

As matters now stand, I think Welch’s chances of winning a recount are perishingly small. Eventually, I think she will come to the same conclusion and not seek a recount.