Monday, November 10, 2008

Another precinct heard from #mnrecount 

John Lott writes today about the Minnesota "pre-count".
This all has occurred even though there hasn�t even yet been a recount. Just local election officials correcting claimed typos in how the numbers were reported. Counties will certify their results today, and their final results will be sent to the secretary of state by Friday. The actual recount won�t even start until November 19.

Correcting these typos was claimed to add 435 votes to Franken and take 69 votes from Coleman. Corrections were posted in other races, but they were only a fraction of those for the Senate. The Senate gains for Franken were 2.5 times the gain for Obama in the presidential race count, 2.9 times the total gain that Democrats got across all Minnesota congressional races, and 5 times the net loss that Democrats suffered for all state House races.

Virtually all of Franken�s new votes came from just three out of 4130 precincts, and almost half the gain (246 votes) occurred in one precinct -- Two Harbors, a small town north of Duluth along Lake Superior -- a heavily Democratic precinct where Obama received 64 percent of the vote. None of the other races had any changes in their vote totals in that precinct.

The Two Harbors point is new to me, and I'm seeking clarification from Lott. So too is the relative size of the gains and losses. Political Animal tweets an update: "Coleman has lost 65 votes; Franken has gained 456 votes and Barkley had gained 82 votes since Wednesday; my guess is that the difference is those 32 votes in the Hennepin County car. It may be that the Senate data is getting closer scrutiny (which reduces some of your faith in other results, yes?) and we again should note that there are basically two typos and a misreport that account for most of this gain. Typos are not subject to statistical analysis; most of my post on Friday regards the one or two vote changes we see. An error in the hundreds place is one random change, not one hundred changes. Still, they are rather large errors.
Indeed, the 504 total new votes for Franken from all the precincts is greater than adding together all the changes for all the precincts in the entire state for the presidential, congressional, and state house races combined (a sum of 482). It was also true that precincts that gave Obama a larger percentage of the vote were statistically more likely to make a correction that helped Franken.
Curiouser and curiouser. Just got another change, putting Coleman at +238. This is becoming like New England weather -- if you don't like the vote count, wait a minute! (Oddly, the SOS website doesn't seem to have that number.) Crap, read the time but not the date, that was last Friday. My mistake, sorry.

People are guarding the ballots now in a volunteer effort (Mrs. S has volunteered, though the Stearns County ballots appear to be secured.) One election official I spoke to called this the longest week of his life; he's going to have a few more.

Lott also says this about voter intent:
Voters themselves insert their ballot into the machine that reads and records their votes, and if the machine finds that a vote isn�t recorded, voters can either mark the race that they forgot to mark or didn�t mark clearly. Or if voters �overvoted� and accidentally marked too many candidates, voters can also get a fresh ballot. There should be no role to divine voters� intentions. If a voter wanted a vote recorded for a particular race, the machine tells him whether his vote in all the races was counted.
The law now allows for manual recount, which sounds great but allows for those squinting eyes "divining voter intent".

UPDATE: Ironman has a new tool, useful when we get to the recount process.

Now we can ask something about those absentee ballots. Using the numbers above I'm going to surmise that of the 32 absentee ballots found under some HennCo worker's McDonalds' wrappers, 21 went for Franken and four for Coleman (using the difference between John's numbers and the update.) Franken took just about half of the votes in the county. What are the odds that a batch of ballots drawn randomly -- or rather, sitting randomly in someone's car for a several day joy-ride -- are of the same type as the rest of the county? About 2.5%. More interesting, what are the odds that there would be only four Coleman votes in the batch of 32? About 0.3%. Curiouser and curiouser.

Labels: ,