12.03.2008

Missing Ballots in Minneapolis?

The latest drama in the Minnesota recount is taking place in the 1st Precinct of the 3rd Ward of the city of Minneapolis, where officials identified 133 fewer ballots in their re-count today than had been recorded on election day. As this particular precinct voted heavily for Franken, who nearly doubled Norm Coleman's vote total there on November 4, a reduction in the ballot count there would be a significant detriment to him, which the St. Paul Pioneer Press estimates as a net loss of 36 ballots.

City elections officials initially claimed that no ballots were in fact missing -- rather, they believed that some ballots were double-counted on election day. From the Pioneer Press article:

Elections officials in Minnesota's largest city today discovered that one precinct came up 133 ballots short of election day totals, resulting in a net loss for Democratic challenger Al Franken of 36 votes.

The development wipes away what had been a boon for Franken in his bid to overtake Republican U.S. Sen. Norm Coleman, after Ramsey County officials found an additional 37 votes for Franken from a Maplewood precinct on Tuesday.

Minneapolis elections director Cindy Reichert said she believes the error occurred when election judges at the precinct on election night mistakenly ran ballots with write-in candidates through a counting machine twice. There were 129 such ballots.

Reichert said although the numbers do not match exactly, she is confident that that's what happened and will report those numbers to the Secretary of State's Office
You may already notice one small discrepancy here. The recount identified 133 fewer ballots than had been processed on election day. Reichert thinks that the culprit is ballots with write-in candidates, of which there were 129. The number 133 is close to the number 129, but it isn't 129 exactly.

However, there is another way to test Reichert's hypothesis. Reichert appears to be claiming that any ballots with any write-in votes for any office were segregated, and that such ballots were mistakenly counted twice on Election Day. If this were the case, then it would have one distinct fingerprint. In particular, the number of write-in ballots for each office on the ballot would be an even number. (The algebra, in case this isn't intuitive: Let's say that of the 129 ballots that contain write-ins for any office, x of them contain write-ins for County Commissioner, where x is a positive integer. If you've double-counted those ballots, you'll wind up with 2x write-ins for that office. Since any positive integer multiplied by 2 results in an even number, the number of write-in ballots you record for that office must be even).

For quite a number of offices in this particular precinct, however, the number of write-in ballots recorded on Election Day was an odd number. For instance:
Office                                       Write-Ins
Soil & Water Supervisor, District 1 39
Associate Justice - Supreme Court 4 11
Judge - Court of Appeals 9 9
Judge - Court of Appeals 15 5
Judge - 4th District Court 1 5
Judge - 4th District Court 4 5
Judge - 4th District Court 9 7
Judge - 4th District Court 10 3
Judge - 4th District Court 19 3
Judge - 4th District Court 30 1
Judge - 4th District Court 36 3
Judge - 4th District Court 42 1
Judge - 4th District Court 46 3
Judge - 4th District Court 56 3
Judge - 4th District Court 58 7
Judge - 4th District Court 60 3
Judge - 4th District Court 61 3
Judge - 4th District Court 62 7
Note also that for a couple of offices, there was exactly one write-in vote recorded. If you've in fact counted all ballots containing write-ins twice, how can you end up with exactly one write-in for a particular office?

It would also stand to reason that if ballots containing write-ins were counted twice, there would be an unusually high number of write-ins in Precinct 1. However, this doesn't happen to be true. In Precinct 1, 10 of the 2026 votes tabulated for President were write-ins, or 0.49%. In the other eight precincts in Ward 3, a total of 58 of the 12,023 ballots tabulated for President were write-ins, or 0.48%. Or, we can look at an office with a higher rate of write-ins, such as "Soil & Water Supervisor, District 1". In Ward 1, 39 of the 1050 ballots cast for this office were write-ins, or 3.71%. In the other eight precincts, 200 of 6,360 ballots cast for this office were write-ins, or 3.14%. The number of write-in ballots in Precinct 1 appears to be very normal, not the high numbers we'd expect such ballots were counted twice.

To her credit, Reichert now appears to be backing off the double-counting theory. And it looks, frankly, like the ballots have in fact been misplaced. The Uptake got a chance to look at registration tables for Precinct 1, and found the following:
1,047 voters signed in on the roster.
932 additional voters registered in person on Election Day.
35 absentee ballots were accepted in this precinct by the city.
15 absentee ballots were accepted in this precinct by the county.
TOTAL: 2,029 voters cast legal ballots (2,028 votes are recorded on the machine tape).
TODAY: 1,896 ballots were included in the recount.
That is, a total of 2,029 voters either signed in on the registered voter roster in this precinct, registered in person on Election Day (Minnesota is one of the few states that allows you to do this), or sent in absentee ballots. This closely matches the 2,028 votes recorded in the precinct's November 4 count, but does not so closely match the 1,896 ballots that were identified in the recount today.

It looks more likely than not that 133 ballots have in fact gone missing; I have no idea what happens if they cannot be found.

204 comments

hill.tops said...

*
FIRST

[ tyler curtain ] said...

Can't be found? Sue.

Dmitry said...

Looks like the dead people got disqualified. Wah wah! The wambulance is coming!

Douglas said...

1,896 + 133 = 2,029

seems pretty simple, no?

Dominic said...

Wow. Random ballots going missing? This just keeps getting weirder and weirder.

wv: sphyst. an accurate description of the current status of the recount (random and chaotic).

Douglas said...

it would have to be one heck of a coincidence otherwise that 2,029 - 1,896 = 133...

Alexander K. said...

Maybe the missing ballots got up and drove to Maplewood, where they were found yesterday!

All these discrepancies bother me. Some ballot magically appear out of thin air (or random cabinets), some ballots disappear for no apparent reason. How is anyone supposed to have any confidence in the electoral system when ballots appear and disappear all the time?

RedHawksO4 said...

Missing? Dear God. This race will likely be decided by less than this amount, and these ballots are missing? This is beyond belief....

Just when I thought this thing was over (for Coleman)....

A Former First Lady & A Guy Named Tim

Sedi said...

Wow, if this race is decided on lost ballots, there is going to be some serious lawsuits coming our way.

Oooh, do we have comment moderation now?

Nate Silver said...

test

ABowers said...

Dear Nate,

Why does the Secretary of State site show Al Franken has a total of 1,188,736 votes to Norm Coleman's 1,177,465? This means that Franken is leading on the official site by 11,271 votes even though almost all the votes are counted. And it seems that the uncounted precints are almost all in Hennepin County which Franken won by a tidy margin. This is much bigger margin than even the total challenges. What gives?
CONFUSED

John said...

Nate:

I am skeptical about your even-number theory. I assume counting votes by scanning through a machine and counting writein votes are separate processes. I could easily see someone segregating the ballots with writeins after they were machine counted, then doing the manual counting of the writeins, and then mistakenly running these through the machine again. If that happened, then the machine counted twice, but the writeins were counted once.

That said, I think the misplaced (or stolen?) ballots is still highly likely because of the crosscheck of the number of people that signed in to vote. If the misplacement was intentional, then 133 being close to 129 might not be by chance.

Jeremy said...

The SoS's numbers have been updated: Link

The 8 has to be a typo. It looks like Franken is now leading by over 10,000 votes.

dcgenerals said...

I see. When Franken's camp finds ballots in someone's backseat, that's the way the cookie crumbles. When Franken loses votes, well, there has gotta be something fishy to report.

Noah S Kunin said...

Reports are that Reichert will go over all the ballots at Minneapolis City Hall tomorrow to figure out what's up before certifying.

Oy.

Noah Kunin
Senior Political Correspondent
www.theuptake.org

Sedi said...

"Looks like the dead people got disqualified. Wah wah! The wambulance is coming!"

Okay, so we've established that comment moderation bar for acceptability is very low.

ABowers,
Not all of the precincts have been re-counted thus far, and the ones that have been re-counted are in places where Franken received more votes (as counted on election day). The margin that folks discuss is what the final margin should be if all of the precincts that haven't yet been re-counted yield identical results to the election day count.

Mark Lawless said...

In Response to ABowers:


Check out Wright and Scott counties. Both started counting today, and both gave Coleman a net of somewhere around 10,000 votes on election night. (Figures are approximations, check out the Star Tribune website for more accurate numbers) In the remaining Minneapolis precincts there are something around 10,000 net votes for Franken yet to be counted. So once you add in the Wright, Scott and remaining Minneapolis ballots, you're essentially left with the tie that has been reported.

rej4sl said...

This isn't the first case of missing ballots in the recount - Franken has noted quite a few instances so far - seems quite suspect - and also with ballots being found !!!!

Mainer said...

The Secretary of State site ONLY shows ballots that have been recounted. It does not include ballot counts from election day from areas that were not recounted.

In contrast, the Star Tribune site keeps the original reported vote numbers intact for areas not counted and uses the recounted totals for areas counted.

Thus you get different gaps between the candidates in these two counts.

Statler N Waldorf said...
This post has been removed by the author.
Mainer said...

Nate - Your link to precinct data on the SoS site is not working properly.

Reid said...

@dcgenerals

The ballots in the back seat "story" has been thoroughly debunked. Absolutely no funny business there.

David said...

Was the Washington state governor's race recount in 2004 this much of a mess? Is this sort of thing par for the course, or is Minnesota having a particularly bad time of it?

Jeremy said...

"I have no idea what happens if they cannot be found."

Nate, I have always loved your honesty... even when it scares me!

Sedi said...

dcgenerals,
When additional ballots are added during a recount because they never got counted, that makes sense (and as Reid notes, the backseat story is universally acknowledged as utter rubbish). If there are FEWER ballots counted in a recount than in the original count, it requires an equally compelling explanation. Nate has suggested why the initial explanation didn't make sense. Do you have a good explanation for the discrepancy?

Joe said...

Micheal the Douchebag stole the ballots

petelutz said...

I worked for the Board of Elections office in Buncombe County, NC. We used very similar ballots/counting machines as MN. The machine doesn't actually count the write ins at all since it would have to decipher hand writing. It counts the rest of the ticket, and then the write ins are counted by eye. So running the ballots through the machine twice would only inflate all the races on the ticket that weren't write ins. So you can't tell anything by the write ins having odd or even results. The totals of each not matching up is a different story. They are definitely scrambling for answers.

ABowers said...

Dear Mark Lawless,

Assuming you are correct that both Coleman and Frankin each have a net of 10,000 votes yet to be counted, then they cancel each other out and leave Frankin with the 11,271 vote margin he has today. I understand that only recounted precints are included in the SOS site. But as of today, that is over 92 percent of the sites so this is a huge margin to overcome with net votes.
Something wrong here.
CONFUSED

David said...
This post has been removed by the author.
David said...

Upon consulting Wikipedia, it seemed the same sort of thing went on there, with ~1000 ballots found during the recount which were missed during the initial count. It certainly suggests that a better system might be desirable, but at least if there are laws in place requiring recounts in close elections, there is some comfort in knowing that a careful search will be conducted in those cases when it matters.

What is really troubling is when some judicial body steps in and prevents votes from being counted. I don't understand how anyone who cares about democracy can stomach such a thing, regardless which candidate they support.

C said...

The first actually unexplained discrepancy in the entire election has Coleman partisans smiling.

If the ballots aren't found before the state canvassing board meets and certifies the results, Minnesota law provides for an election challenge lawsuit to be filed in Ramsey County. It will be a short wait, I'm sure.

In the meantime, the Senate will justifiably refuse to seat the "winner."

Pretty straightforward.

RufusRules said...

Maybe they should look for them on the top shelf of that storage closet in Indiana.

Seriously, it's amazing that someone up there is ballsy enough to be playing hide-the-ballots under the scrutiny of a recount. I'm sure I'll get bashed for this, but when there's funny business afoot, it's uncanny how it somehow always seems to benefit the Republicans.

Jeffrey said...

I for one am pretty pleased with the way the recount has gone, especially since there seem to be a lot of safeguards in place to make sure ballots aren't lost, etc. Does anyone know if other states have similar safeguards, to make sure that the number of ballots counted during a recount is the same as the number of ballots cast?

wv saryedi: "A thousand pardons, Mr. Albert"

Jack-be-nimble said...

i'll tell you what happens when the ballots can't be found.....


Coleman wins......coleman wins....

The RNC has moles inside the state grabbing ballots to help Coleman......yeah.....

They had to make up for those phony ballots that were found in the other county.

Michael said...

The irony is this particular voting location (a church) in Minneapolis is the bluest spot in the bluest city in the State. According to the boys on this board it must be some vast Republican conspiracy! Franken and the DFL look like idiots for calling into question the very people that support them the most. The Coleman people are smiling from ear to ear...

Mark Lawless said...

Sorry, I should have made myself more clear. Wright County has (again, all of these numbers are very rough approximations) 10,000 net votes for Coleman. Scott county has an additional 10,000 net votes for Coleman. One of these counties cancels out the 10,000 net votes remaining in Minneapolis. The other 10,000 net votes cancels out the supposed lead on the SoS website.

The Star Tribune's website tracks only changes in votes from the totals reported from the election, and is a better way of keeping track of votes (although it's still crippled by the sheer number of ballots that have been challenged).

LDW said...

I am not much of a believer in conspiracy theories, and maybe this is all just coincidence.

But how strange can it be?

171 additional ballots are found in one area that leaves Franken up a net 37 votes. Another location is missing 133 ballots that happens to leave Franken down a net 36 votes.

It is amazing that the net impacts of events at one location almost exactly negates the net impact of events at another location. And the initial explanation, if accepted, of double-counting ballots at the second location (which Nate now clearly shows must be incorrect) would have made the events totally offsetting without need for further investigation.

I realize that highly improbable coincidences must occasionally occur in the real world. After all, there is a big diffence between the words "highly improbable" and "impossible".

But a thorough investigation of these 2 events (especially the last one) by unbiased investigators should take place. Otherwise there will be a cloud of suspicion about these findings that will shake the foundations of trust in our electoral process as much or even more than did the Florida 2000 election.

For this is not just about what happens in one state of the union. The interest level in this election is high accross the nation. It is as much or more "nationalized" than was the recent Geogia senatorial runnoff election.

Once again, all of America (and no doubt part of the rest of the world) is watching our democracy in action.

Craig said...

An interesting study in human nature. In this case (Hennepin County), the people in charge are almost certainly Democrats. And even though losing ballots certainly favors Coleman... what is the answer? "We didn't lose no stinking ballots."

Summary: CYA prevails over party affiliation or any other human instinct.

Also, forget fancy algebra... Out of 133 supposedly write-in votes, Franken got 80 and Coleman 34? So all the write-ins (rest for Barkley presumably) were for candidates on the actual ballot? I don't think so.

Paul Bradford said...

What if it's not a conspiracy? What if it's just human error? What if the error gets rectified? There are a lot of folks in Minnesota who want this recount to be squeaky clean and fair.

shma said...

ABowers said...

Dear Mark Lawless,

Assuming you are correct that both Coleman and Frankin each have a net of 10,000 votes yet to be counted, then they cancel each other out and leave Frankin with the 11,271 vote margin he has today.


That's not what he said. He said that Coleman has about 10,000 more votes in the two counties that are still counting. This makes the total a tie up to rounding.

And Hennepin finished counting, so there are no votes left there for Franken to pick up.

walt526 said...

171 additional ballots are found in one area that leaves Franken up a net 37 votes. Another location is missing 133 ballots that happens to leave Franken down a net 36 votes.

That was my read on the situation as well. But it's worth pointing out that it need not be the case that this is some part of a huge GOP conspiracy or that Coleman is directly involved. All it would take is one individual with access to the ballots who felt that it would be acceptable to "undo" the alleged "backseat ballot find". Such a person need not be an elected official--could be sheriff deputy or county administrative clerk, so partisanship leaning of the precinct is irrelevant.

I don't know how you correct for this. I assume that by ordering a hand recount, the initial tally is rendered unofficial. Hopefully the election is either decided in Franken's favor or decided in Coleman's favor by a margin in excess of the missing votes. Otherwise it will be a cat fight in the courts or Senate over how to award the "missing" votes... not a good situation.

LDW said...

To the highly improbable coincidences I should have added the temporal factor. The events not only negated each other in magnitude, but one event followed the other in a very short time.

For those who thrive on conspiracy theories, it might seem like the first event promptly led to a "causation" of the second event.

Cugel said...
This post has been removed by the author.
Cugel said...

The end result of this is going to be horribly ugly and leave a massive stain.

If the ballots cannot be found, then they cannot be counted. There can only be theories about what happened to them.

Were some ballots lost? Was this foul play or was it accidental? Were some ballots but not others counted twice some how?

Unless the ballots are found, there's no way to resolve this issue.

The Minneapolis elections official who tried to peddle the "they were double-counted" theory is full of crap.

Either the count EXACTLY matches the discrepancy or it doesn't!

If if does, at least there's some plausible theory. If not, then the entire theory is shot down as utterly without evidentiary foundation.

"It's sorta close" is not good enough!

You can't explain 133 missing votes by claiming some were counted twice, unless you can show EXACTLY which ballots were counted twice, and unless they EXACTLY ADD UP.

And they can't do that.

Anything else is just a big cover-your-ass for that elections official that is not going to fly.

This might ultimately come down to whether Senate Democrats have the courage to seat Franken, if he loses by less than 36 votes.

MNLatteLiberal said...

@ ABowers,
Here goes :). I will attempt to explain the difference.

At this point only 35 % of votes are counted at Wright County. The current totals are
14097 Coleman
7874 Franken
That gives you an edge of 6200 for Coleman, roughly.

Compare that with the Nov 4 totals
ca. 33K for Coleman and 18.6K for Franken for a difference of 14.4K for Coleman.

That means that in that county alone Coleman, all things being equal (which is a HUGE if in the last couple of days), will make up about 14,400 - 6200 = 8200 votes.

The situation in Scott County is somewhat similar with only 59.1% counted. But there too Coleman is projected to make up 3100 or so for the net total of about 11,300 which is Franken's current recount margin (11,266) on the SOS site that Nate is quoting.

The challenges will make up the rest of the difference, which is tiny by comparison.

Hope this helped.
~Latte,

who is more maddened than saddened by these W3 P1 missing 133 ballots. Those precinct judges deserve to be shot!

Who is doing the bookkeeping? This isn't rocket science, this is pretty straight forward stuff, no matter how busy it gets. As a Franken volunteer during the very organized recount, I cannot fathom how a discrepancy, nay, scratch that discrepancIES this large do not get reported until the 12th hour. Those precinct judges KNEW, they KNEW APRIORI that their count was way off! They were URGED by both SOS Mark Ritchie and by Mark Elias and Franken campaign to report the discrepancies up front. AND NADA!

Shame on them! These precinct judges from Maplewood and from esp. from Woodbury and particularly Ramsey Ward 3, precinct 1 ought to be shot behind the same barn as the GOP Coleman observers/site leaders from Fillmore county who challenged McCain/Palin/Franken ballots on the ground of "GOP voter intent". No mercy for these slackers. They give our democracy a black eye!

Man, am I pissed!

Ed said...

ABowers:

When reviewing the MN Secretary of State posted counts, you have to keep in mind the rules in place for challenged ballots. Under MN's rules, if either side challenges a ballot, it is counted as a "non-vote" until the challenge is resolved by review board in mid-December. Challenges may - and have - been directed at votes counted for one of the candidates on election day, as well as ballots originally counted as non-votes in the Senate race. Because of this, and in light of the very high rate of challenges issued by both campaigns, a huge number of perfectly valid Coleman and Franken votes are, provisionally, being treated as "non-votes," pending resolution of the challenges. Therefore, the Secretary of State's totals are essentially meaningless; as the outcome of this race is going to turn on a very small number of votes, the resolution of challenged ballots will provide that margin.

Michael (mbw) said...

(I wish Michael were not such a common name.)

Up to this point, it was really amazing how clean the whole business had been on both sides. There was the usual playing with numbers for spin, and entirely unjustified nasty paranoid claims from R's. But as far as any actual actions went, the system looked great.

This new bit looks very rotten, however. BTW, I believe that this DinkyTown precinct (Ward 3, precinct 1) is the same one that made it into the news on election day by sending away many students, claiming that the proof of residence that worked in the past was no longer adequate.
So actually, sorry to sound paranoid, but I think there was a bad apple at this one site. I mean a bad R apple.

Cugel said...

Well, here's a twist I didn't think of but should have:

"In their complaint, Franken campaign officials said that records indicated that 2,029 people voted in the precinct on Election Day and the recount recorded only 1,896 votes."

They keep a tally of who's voted and who hasn't don't they? And they can add up the actual names of everybody who voted, right? And if the number of names of people who voted does NOT match the number of ballots, but DOES match the number of votes tallied on election day then some ballots WERE LOST.

Q.E.D.

Well, those bastards will have a hell of a time trying to say "we lost the ballots so they don't count."

Apparently, they are backing off the "double-counted" crap theory now.

If they try and report a number of votes that's different than the number of people who showed up at the polls, showed their ID, and voted, then there's going to be hell to pay.

They better just certify the numbers after adjusting for the ballots they have, but WITHOUT deducting the missing votes from either party!

MNLatteLiberal said...

OK, help me out here with arithmetic because there is something I am missing with this "net loss of 36 for Franken" number as I try to deal with the mess.

If 80 (supposedly write-in) ballots had Franken, and 34 had Coleman, and the rest presumably had Barkley (29?), then the net loss for Al should be 80-34 = 44, right? And not 36. Where does 36 come from?

Please help. I am feeling obtuse.
~Latte

Dink said...

Nate, your analysis on the possibility of the write-in ballots being double counted is off-the-mark. The theory proposed by Ms. Reichert was that the ballot were run "through a counting machine twice". Counting machines don't count the write-ins just the votes for listed candidates, so the theory does not rest on the write-in votes themselves being counted twice. There may be an explanation for the discrepancy between 129 ballots red twice and 133 votes missing. Two ballots were challenged according to the Star-Tribune. If these two ballots contained write-in votes and were counted twice on election night that brings the figures into proof (133-(2*2)=129. This is not all that unlikely since many of the challenged ballots I have looked at have included write-ins for other offices. One of the grounds for excluding a ballot is that it contains identifying information and there seem to be cases where the side that wants to suppress a vote contends the name on the ballot is not actually a write-in but identifies the voter.

In response to folks who are shocked, amazed and bothered by "all the discrepancies". What has amazed me watching this recount is how accurate the original results were. Remember on election night the votes are handled by poll workers who have spent a very long difficult day. Does it amaze anyone that someone might have gotten confused and run a small stack of 129 ballots through a counting machine a second time or that someone in Maplewood forgot to run 171 ballots through a machine? The whole point of a recount is to catch these types of errors.

tkk13above said...

They tied, period. Just like Gore and Bush tied in FL in 2000. The system of vote counting is not precise enough to measure differences with such large numbers and such a tiny margin of error.

I truly hope someone finds those votes and counts them. It would be a shame if Coleman won by, say, 20 votes, which is probably less than the number of votes Franken would have netted over Coleman in the missing Minneapolis pile. At this point every single vote counts, because they really are that close...!

tkk13above said...

oh, and just to agree with Dink -- I would love to have a better system and I'm sorry it is not precise enough to accurately measure a winner in a case like this one, but on the other hand it is truly amazing just how good the system is! We notice the errors because the race is so close, but if you think about it the number of errors is exceedingly small given how many votes were cast.

Andy said...

If Coleman and his people are smart and very very VERY cynical, they would try and sow doubt about the legitimacy of the recount. Because if this all dissolves into a massive court case, and the court throws up its hands and orders a revote (which is possibly in most states, I believe) Coleman would have a huge advantage since Franken could not count on the Obama surge anymore (see Georgia runoff).

I rather suspect this is where we're heading. To a revote, which I'm going to guess the bad guys will win by 6-8 points.

Statler N Waldorf said...
This post has been removed by the author.
Shaun Martin said...

I'm no monster expert, having prosecuted only one election litigation (in CA, at that), but I think Nate's right that it's very very likely that these are "missing" rather than double counted. (But I don't think the "even number" theory holds at all; John at 8:34 is totally right about this).

The sign-in sheets are the best piece of evidence; if you sign in, you almost certainly cast a ballot, and the fact that the 2029 number is so close to the 2028 on the machine tape is pretty clear evidence that there was (a) not double-counting, but (b) rather lost ballots. (The 1-ballot discrepancy, Nate, is likely due to a voter having signed in but not, in the end, electing to turn in a ballot; it happens rarely, but it happens.) Indeed, given the numbers here, if I were a judge, I might well rely on the initial count in this unique setting rather than the recount. It's a hard doctrinal move to make, but it's doable, as well as equitable. And I say this as a pure legal view; I'd say the same even if the missing ballots were for Coleman.

Hopefully they find the missing ballots. (FWIW, I doubt they're deliberately stolen, but have no evidence one way or the other on this point.) If not, yeah, it's a tough legal call, especially if the 36 votes are dispositive, but
but given the evidence, if they're not found, I'd given Franken the 36.

P.S. - I'd suggest one more move by the folks in charge, which I don't see has been tried yet: compare the write-in totals on the first count to the write-in totals on the second count. Write-ins are hand-counted both times (plus, we know there's no double-counting on this front even the odd numbers). There's collectively (in all of the races) a sufficiently high percentage of write-ins so some of the "missing" 133 ballots almost assuredly had write-ins. If the first count, say, showed 39 write-ins for "John Smith" for Soil Guy, and the second count only shows 37, there's almost certainly missing ballots. (Even better if the first count shows 1 vote for Mickey Mouse and the second shows 0 votes for 'em; then you're sure.)

Now, this is a one-way rachet; even if the write-in totals are the same in both counts, you can't be sure there's no missing ballots. But if they don't match, you're pretty sure that there's missing ballots.

I'm not in MN, but it's worth a look by those with better access to the data than me.

Dink said...

Cugel,

While it is true that records are kept of people who voted, this was not where the numbers in the Franken complaint came from. The numbers in the complaint are the number of ballots counted in the Senate race the first time and in the recount. To reconcile these totals to the registration lists requires also considering under-votes, over-votes and other excluded ballots.

greg said...

Sure you know, Nate--absent an equally plausible explanation from the Coleman camp, the canvassing board (or a court) will rule that the tape count controls (rather than the ballots found) because the tape count much more closely matches the total number of votes cast. It is, therefore, the "best evidence" as we say in the law, of the actual vote and therefore Franken gets his 36 vote edge back--the same as he would if a fire destroyed all the ballots.

Michael (mbw) said...

@Dink- Sure it can easily happen that a batch of ballots gets run through twice. That happened on a small scale in many districts. It's not what happened in DInkyTown, where 133 votes of people who were recorded as showing up or having submitted absentees got counted on election night and then disappeared.
So far as I know only one of the 4000+ precincts was reported on election day to be doing something weird to stop voting. Only one has a large number of disappeared ballots. If my reading of Google finds is right (please check, I can get things wrong) these are the same precinct.

Statler N Waldorf said...

The now-moderated format is very much appreciated, BTW. The site has finally been reclaimed from the trolls.

MNLatteLiberal said...

@Shaun Martin
I like your analysis, but how do you get 36? How does everyone get 36 when I get 46? 80-34=46. What am I missing here?

Marcus said...

How the fuck do you not know how many ballots you're suppose to have? People should sign in when them come in, sing out when they leave. You should count how many ballots you have at the beginning of the day and at the end. There should be a count on the tabulator. ALL THESE NUMBERS SHOULD MATCH. It's not rocket surgery, but basic arithmetic. Not knowing how many people voted in your precinct is a sign of gross incompetence on the part of election officials.

merzbow said...

If they decide to go with the tape count for this district, they should also do it for the other district were ballots were found in favor of Franken. Only fair, no? It's as easy to stuff a bunch of newly made-up ballots into a machine in a warehouse as it is to steal legit ballots from a highly-secured area.

Kennyb said...

Marcus, the sign in and sign out numbers of voters rarely matches. Too many people leave without signing out, despite signs and the best intentions (and attention) of the poll checkers. Our New Hampshire precinct sign out count was off 17 from the sign in count, although the sign in count was only 2 more than the number of live ballots cast.

mediapost said...

@Latte

Your right, This is a 46 vote net loss for Franken, not a 36 vote loss. I looked on the SOS site and it shows 1090 F-before, 1010 F-after, 595 C-before, 561 C-after. 80 - 34 = 46.

2 questions:
1) Does the board have the ability to revert to the machine count in a given precinct if they find that ballots have been lost?
2) We have heard the Franken campaign talk about a lot of lost ballots all over the state. Are there any other precincts where the registration table totals don't match the paper ballots on hand?

Jake said...

There's moderation here now; how does that work?

Cris said...

This might be a dumb question, but what happens if the recount is completed and the difference in votes is something super close like less than 100 votes - does either side have a recourse? Can they object in any way? Can they force another recount? Or is it over once the recount is complete?

-- Cris
My site: Obama Wallpaper Archive

Michael (mbw) said...

@ Dink- You're a new poster. You have a lot of thoughts about how 133 people who signed in and had their votes counted, mostly for Franken, really didn't exist. In the same precinct where the SoS had to issue special orders to try to get students to be allowed to vote, very late in the day. Called Dinkytown.

I thought, just in case you hail from those parts, you could help spread the word there on a few facts.

1. The new U.S. attorney for that area will be appointed by President Barack Obama.
2. And approved by the Democratic U.S. Senate, with or without Al Franken.
3. and will have discretionary power to investigate and press charges for felony vote tampering.


On the cheery side, I was in a Federal Pen some 37 years ago, and it wasn't all that bad a place.

LDW said...

Latte,

I suspect the answer to your question about calculating 44 votes down for Franken (rather than 36) is driven by the original description of the problem as being caused by "double-counting write-in ballots".

If the double-counting of write-in ballots is not the answer, then it may be that the 36 down for Franken is just the difference in the original machine count Franken/Coleman votes vs. the recount (which does not take into consideration the missing 133 ballots and which are probably not exclusively write-in ballots).

In other words, the 44 vs. 36 Franken difference may be just another element of proof that eliminates the "double-count of write-in ballots" theory.

By the way, thanks for letting me know that it has been known for some time that there was a 133 vote discrepancy at W3 P1 between the machine counts of the election and of the recount. That points to a greater probability of human error rather than deliberate intent.

But it was indeed strange that the announcement of a "found" positive net gain of 37 for Franken in one location was followed on the very next day by an announcemnt of a "lost" 36 votes for Franken in a totally separate location.

My view now is that the timing of the announcements and similarity in vote gain and loss is strictly coincidental - strange as it may seem.

Michael (mbw) said...

@merzbow
No. To add ballots you'd also have to fake the sign-in sheets, recorded separately. Nice try, though.

Statler N Waldorf said...
This post has been removed by the author.
Matthew Wintergarden said...

Speaking of Minnesota, the city of Duluth had a city council rep retire. So how do they fill the position?

They ask for resumesfrom the residents of the city. Interesting form of Jeffersonian democracy.

merzbow said...

@mbw

Excellent. Then the next step is to meticulously examine all of the sign-ins, same-day registrations, and absentee ballots cast in this district to indeed make sure they all represent separate, legal voters. If indeed 2028 such people are found, then we have no problem.

Shaun Martin said...

@Latte - I think Nate (and I) just grabbed the 36 figure from the Pioneer Press story. It seems to me (like you) it's 46, and some stories now use this latter figure.

I've also confirmed that this precinct listed "1978 signatures" in addition to 42 regular military overseas ballots and 8 federal absentee ballots, for a total of (you guessed it) 2028 signatures & voters. This figure also ties in neatly with the number of total presidential votes recorded in the first count for this precinct: 2026. (I assume the remaining 2 voters over- or undervoted the POTUS race, a very plausible number).

So stuff looks missing. Maybe the Board can examine the ballots for, say, Supervisor District 1 (among others) to know for sure. The first count shows a 2% write-in rate (39 ballots) in this uncontested election. Are all 39 of these ballots still there? If not, oopsies. And if 133 ballots are missing, you'd expect around 2 (2% times 133) of these write-ins to be missing, assuming normal distribution. Add that chance to the chance you can find, say, a missing write-in ballot for Soil Supervisor 3 (36 write-ins) and lots of other races and you've got a decent shot of coming up with a missing write-in and hence your answer.

DCM in FL said...

That .39% cumulative 'lead' for Franken tonight on the MN SOS site sure looks sweet...

That makes 48 hours with Franken in the lead before it reverts to the virtual statistical 'tie' again tomorrow after the final #'s come in for Scott & Wright Counties...

unless, of course, more ballots appear or disappear again tomorrow that is...

WV - 'rot heda'

Statler N Waldorf said...

How long before the SoS looks at the disputed ballots?

Mr Z said...

Ok, there are 133 missing ballots, and Franken's vote total has dropped by 46 relative to Coleman's, but how have the vote totals overall changed in this precinct?

Here's what I'm getting at: We're comparing against the pre-recount totals, right? If the ballots hadn't gone missing, then we'd have a complete total. The complete total may not match the original total, and so there may have been some gain for Franken or for Coleman.

What was Franken's trend in this precinct for the ballots that didn't go missing?

obsessed said...

Nate - According to your calculations, what are the odds that Republican treachery is NOT involved in the "lost" votes from a precinct where Franken had doubled Coleman's total? 1 in 1000? 1 in 1,000,000? Zero?

THE NEW AMERICAN AURORA said...

I'm sorry. But I don't understand the issue.

Tonight, with 97.58% of the recount complete, Franken is ahead by 11,170 votes. Franken has 1,186,134, and Coleman has 1,174,964.

There are 6,326 contested ballots out there, 3085 by Franken and 3241 by Coleman, not nearly enough to fully close this gap in favor of either candidate.

Are you saying that the final 2.42% (55 precincts) + these 133 votes will not only close the gap but put Coleman ahead?

Where are these precincts? Are they died-in-the-wool Republican precincts?

It seems to me that, excluding the challenged ballots, Coleman needs to pull about 60% of all remaining votes to get even.

Or am I not correctly reading the SOS's web site?

http://electionresults.sos.state.mn.us/20081104/SenateRecount.asp

I think we can forget about trying to figure the margins based on the original count. Clearly, the machine count was profoundly screwed up. I have no doubt, based on the reporting about black box voting over the last eight years (Kembrew McLeod's work bothered me most of all), it's that way throughout the states.

But here, at last, we have clear proof that something is wrong with the machine-count system. And heaven help the manufacturers of the equipment and software.

obsessed said...

If there were a re-vote, I bet the third candidate's total would evaporate.

Would that help Franken? I mean ... MN is pretty blue - what the hell are they doing with a crooked Republican senator in the first place. Of course, they've also got a Republican governor don't they. But the other senator is Klobuchar.

Ay ... shades of Fargo.

Clay said...

I have a question for the people that have used these particular vote counting machines. I know write-in votes must be counted by hand to interpret the hand writing. I also know there is a bubble to fill in if you are writing a vote in. Does the count machine not count the number of write-ins, even if it doesn't count who those write-in votes are for?

Depending on the answer...Nate's analysis of looking at the number of write-in votes is very feasible evidence to consider.

Does any one know the process for tallying voters in MN? In Georgia we must keep certificates voters fill out (which we check to make sure match their registration before they are allowed to vote). We also keep an electronic check list of who voted, as well as, a count of voter certificates on election night, which is recorded separately. The count of ballots cast is also recorded separately, as well as the ballots being stored on memory cards and immediately sent via modem on election night (we use computers to vote). With our procedures if there was a discrepancy we would know about it election night. I would think most states would have similar procedures to keep multiple safeguards to prevent tampering with the number of ballots cast.



Btw, I am a liberal person living in a very conservative area. I was a poll official alongside conservative poll officials. I am sure you can find the same thing in many different precincts and party affiliations. Therefore, just because an area is mostly liberal does not mean any tampering by a poll official must have been caused by a liberal person. If anything if someone lacked the moral standards to help maintain fair elections, they might try to target precincts that typically vote for the other party, since there is probably more room to "lose" ballots than "add" ballots without people knowing who did it.

Hopefully, that is not what happened, but I don't think saying it was a liberal area debunks the notion the conspiracy theory a conservative person tampered with ballots. Additionally, since ballots were potentially lost the tampering could have occurred at a central office/warehouse where there is more room people with varying political ideologies to -lose- votes.

If the votes are not found. I think most courts would use the tape count unless there is sufficient evidence that ballots were counted twice. Ideally the votes from that precinct should be recounted, but if they can't be because some are lost, at least the initial count was a legal count of all the election day ballots.

obsessed said...

The New American Aurora: There's something very funky with the site that shows Franken way ahead. Even Franken's campaign only says he's ahead by 20 or 30.

R682770 said...

IMHO If the election is very close, and none of the missing ballots are found, the senate should force a new election.

Other then perhaps implementing a IRV system or Runoff elections I could see Minnesota taking a few ideas out of the shortcomings of this elections.

1. Making duplicate ballots for those which are not read by a machine is absurd. These should be read by hand, as the duplicates can obviously cause confusion. This alone could arguably cause enough of a justification for a revote.

2. Minnesota needs to standardize party position on ballots. as can be seen by a number of the partially filled in ballots where the voter actually caught this mistake, LOTS of Democrats accidentally voted for Coleman, and LOTS of Republicans accidentally voted for Barkley. In both of the states I have voted in, there are party lines which can be easily followed.

3. Perhaps in future elections, ballots could be numbered. that way, someone could know if ballots have been lost! You can still have a secret ballot as long as the number of the ballot you used was not revealed.

Clay said...

The New American Aurora:
Why is it so hard to believe that a few precincts in MN could go for Coleman by 60% to close the existing gap with Franken?

It doesn't seem that far fetched. In most states there are usually some very "red" areas and some very "blue" areas, even when the state is very blue or very red.

Clay said...

R68220,

ballots can not be numbered for the same reason MN law prevents people from leaving identifying marks on ballots. If ballots were publicly viewed post election it could leave to fraud via vote buying or peer pressure.

One of the problems with vote buying (besides being illegal) is the buyer can not check to make sure a person voted as they should.

The guy who doesn't want to admit to friends how he voted, but might be coerced socially into giving up their ballot number (or not not liked by not giving it up...causing social pressure to vote a certain way in the first place) could have their votes made public after the election.

Ballots need to be kept private and untraceable.

What there needs to be is an over redundant system that tracks the number of voters and the number of ballots cast on election day and replication of these results at the end of election day.

standardizing position on ballots is also not necessary. In the ballots from MN the party affiliation is clearly listed in a defined area next to the persons name. Anyone reading the ballot (instead of christmas treeing the bubbles) can know who and what party they are voting for.

I am curious what your source is that lots of democrats accidentally voted for Coleman and lots of republicans voted for Barclay?

George In Florida said...

From working the polls here in Fla:

To determine if the number of missing ballots is indeed missing or not, the answer is simple. Each poll is issued a certain number of blank ballots. We have to count the number of black ballots at the end of the day to determine the number of ballots issued. While the numbers may be off by a couple, certainly not 133. So it is very easy to determine if the ballots were indeed used or not.

It is very possible to accidently re-scan the write-in ballots. In our machine, they are scanned, and seperated out into two bins. A third bin is used for unscanned ballts. It is possible for the write in bin to accidently be re-scanned again, through human error. However, there is no way that the write ins would be counted twice, only the other races which were on the ballots with write ins, but for which there was not a write in.

Write ins are hand counted here later in the process, not on election night at the polls. However, the other races on the ballot which were not write ins, are counted by the scanner immediately.

WV: idstypti Idystppti my comment, you read it.

George In Florida said...

BTW:

Counting signatures IS NOT an acurate method to determine the actual number of voters. The cout is usually off by a few, since there are many things that can go wrong in the process (way to many to list) which may make it hard to get an accurate count from the book. Most precincts here are off in the signature count by a few (one book we had here was counted by three different people, and we had four different totals!). However, usually they are only off by up to a handfull, not 133.

wv:daydme If I told you the first election I voted in, it would probably daydme.

Clay said...

George in Florida,

Just curious... does that scanning machine scan that a write-in vote exists, even though it doesn't tally who that write-in vote is for?

Clay said...

George,

Counting signatures, certificates, or whatever a state uses can still work. Even if on election night poll workers get the count wrong, there is still a paper record of it. If there is a problem later on in the recount (like there is now) that record can be examined to determine the real number of voters.

In Georgia we use certificates instead of signatures. Since we have computerized voting, we also have a list that is marked on the computer. At the end of election night we have to balance everything. Getting the certificate count accurate sometimes takes awhile and sometimes dozens of times recounting it by multiple people to get it right, but in the end we get it balanced.

Even if we misplaced a certificate and got it wrong, we would know it election night and the difference between votes cast and voter certificates would be documented that night (rather than found out in a recount). I would assume the same thing would happen in Florida.

Btw, the counting blank ballots idea is good. It adds an extra layer of redundancy where information can be collected to make sure everything totals out correctly

George In Florida said...

@ R682770:

Here in Fla, the ballots are numbered.

George In Florida said...

Clay:

A total count of all write in votes is listed here on the tape. It lists how many people filled in the bubble for wite in for each race.

George In Florida said...

Counting signatures:

if a person signs the book, but the signature doesn't match, you have a signature without a bollot issued. The person at the poll is supposed to make a note, but the note can be missed after working the polls for 12 hours. Or the person forgot to make a note.

Ocassoinally, a voter will have a signature both in the normal books, and the add on books. It should not happen, but is does.

A person signs the book in the wrong place, then signs the book in the right place. The poll worker did not notice. Two signatures, one vote.

The person signs the book, and then says that his address is not correct in the book. A poll worker error, but a signature with no vote. Again, a note should have been made, but....

Some counts are made on the number of signatures, others on the numbers of initials of the verifyer. There is always some voters who the poll worker forgets to initial. The poll worker then goes back and initials where he forgot ( I found one poll worker initial next to a sticher saying that the voter could not vote because an absentee was issued, but there was a poll worker initial (no signature).

Hand counting signatures is not that easy. I'm usually very good at it, but I get a wrong count ocasionally. Other signature counters are not as good. I had to re-count one book, whre the two people who counted it were off by 15 signatures. And neither one of thier counts was right.

In my precint, I ended up by counting 5 of the 6 books. Believe me, the process is not as easy as it seems it would be.

Thewe are just some of the things that can make a hand count of signatures off. There are plenty more.

There are also some problems with counting blank ballots, and filled in ballots, but I won't go into them.

Tim said...

If there is a revote, don't be too surprised if Minnesotans say to hell with both these guys and elect Barkley. Hopefully he'd still be on the ballot.

Personally, I'll be going door to door in support of Lizard People.

Clay said...

Thanks..

That makes me wonder if the numbers Nate has are the total machine counts of write-in voters or the result of hand counts of write-ins

ahh.. I see your problem down there. In my precinct in GA everyone must fill out a certificate. We have to verify the information on the certificate is what is on the voter registration data base and file that certificate if we issue a ballot. Any other certificates (voided, incorrect, no ballet issued) are kept separately in another file envelope.

Mrs B said...

is it likely to be the case that election procedures will be changed in MN after this recount to avoid problems like not knowing how many people voted, precincts going home without phoning in their totals, poor ballot design, unreliable machines, rules for challenges etc etc?

ecarlson said...

I just wanted to comment that if it is agreed that there are some small number (ca. 133) of lost ballots, it is possible to restore those lost ballots. The method is simple. Run all the ballots from that precinct through machines. Subtract the resulting totals from the original machine count. This tells us how the MACHINES marked the missing ballots. Then add those back into each candidate's totals. This denies the partisans their right to have these 133 ballots examined by hand, but since the machines are better than 99 percent accurate, it would represent an error of about one vote. Not perfect, but far superior to dropping these votes, or any alternative I can think of. Of course, there's a high likelihood (certainty?) it would be challenged in the courts.

WV: comor

I hope these problems are done, but I'm sure there will comor in the next week or so.

Michael (mbw) said...

As usual, ecarlson has the rational analysis and solution.

There are a surprising number of other comments which go off on tangents which had already been resolved.

1. Yes, Aurora, the remaining counties were heavily for C, so the net count will end up very close to tied again.

2. No, Nate's analysis is irrelevant because, as always, we have a highly accurate, though not perfect, count of how many votes there should be.

There is no doubt that 133 or 134 ballots are missing, and there is circumstantial evidence to indicate foul play.

It's amazing that the election officials went off into fantasyland before doing the obvious first step, which they must all learn in training, of comparing the total with the sign-in number.

As a precaution, I hope this blog system archives all the information on the commenters, since in this case some of it may be useful for a future investigation. There are two very slick new characters arguing that there's nothing to see here folks (Dink) and then as a fallback (merzbow) that it's easy to resolve, just do a one-time demand for retrospective IDs from this one precinct, reverting to the reduced (-133) count unless the Herculanean task of tracking down over 2000 voters individually is accomplished. We see a Rep. legal talking point in gestation. They're just as ugly as newborns as they are grown-up.

HmblDog said...

In banking when the numbers don't want to add up it is usually because we forgot to include the NNTB value. I'm sure many of you were already familiar with this one. For those that aren't this would be the "Number Needed To Balance"

susan said...

@David

Judicial stepping in? You mean like how Sandra Day O'Connor elected Bush!

yeah yeah yeah

And isn't it cute that all that ACORN stuff was touted to the skies, to provide distraction from much larger disenfranchisement cheating going on all over the place? Typical.

With all these shenanigans it seems like a revote is the only fair answer. And despite the loss of Obama coattails, the fact that is is one on one would probably provided a clear result.

MNLatteLiberal said...

My thanks to mediapost and LDW for your responses to my "36" puzzle. LDW, using Occam's razor, I remain unconvinced, but I see your point why the math does not have to be as linear as I'd like.

@Dink, who said
"In response to folks who are shocked, amazed and bothered by "all the discrepancies". What has amazed me watching this recount is how accurate the original results were. Remember on election night the votes are handled by poll workers who have spent a very long difficult day. Does it amaze anyone that someone might have gotten confused and run a small stack of 129 ballots through a counting machine a second time or that someone in Maplewood forgot to run 171 ballots through a machine? The whole point of a recount is to catch these types of errors.
"

What amazes this liberal, as he desperately tries to visualize WTF happened on the election night at those precincts (let's take the P1 W3 one for instance) as the judges counted the ballots. They had their total from the registration logs for all those signing in. That figure was 2029. Then they sort and count the ballots. X for Coleman, Y for Franken, Z for Barkley, ? write ins and all else. The net sum of X + Y + Z + ? must add up to 2029. It adds up to 2028. They live with the one off. Fine. I am still with them.

Now, a month later they are trying to tell us, wait, wait. The number 2029 for who registered and signed in AND the number 2028 for those whose vote was recorded are actually something to be ignored because the total ballots on hand are 134 less = 1896. And what ACTUALLY happened is that a bunch of them (129 (+2 write-ins/challenged)) were run through the machine twice. And that the actual number of people registered and signed in at the precinct ought to be 1896, and all those "extra" folks who signed in on Nov. 4 are what? Ghosts?

I am sorry, but that part of the story does not wash for me. I still claim, that based on the story Reichert try to sell originally, on the election night they HAD to notice the discrepancy between the 2029 signing in and 1896 casting votes. And they HAD TO REPORT IT. NOW AND THEN.

Furthermore, they were urged, as I have already written above, by both the SOS Office and Franken Campaign REPEATEDLY as were ALL the precinct judges to immediately report the count discrepancies as they arise. And I repeat, acc. to the Reichert original version of the events, she HAD TO KNOW of that delta on the evening of Nov. 4. In fact, she was not to leave the precinct as the election judge until that issue was resolved.

So that theory is out the window now, and we have no new theory as of yet. The issue of NOT reporting the discrepancy in the count immediately remains.

How we go forward is problematic unless the ballots are found. I too agree with ecarlson's suggestion; the problem, I bet, will be with the Republican side embracing it. Something tells me they might feel an urge to challenge that. If only on the basis that those 133 missing and mathematically derived ballots cannot be physically challenged by their alert observers.

FWIW
~Latte

wv - sycie. I got this sycie feeling about these 133 running amiss.

Jason S said...

You made a slight error here Nate.

If I segregate the ballots with write in votes and then double count them by:

1. First feeding them into the machine

2. Counting them by hand

Then the actual write-in votes will only be counted once because the machine doesn't count write-in votes.

The machine will count each write in once when calculating the total number of write-ins (but will record no votes since it can't read) and the actual vote will be counted once (by the human who reads it).

The vote for Franken or Colman will be counted twice, once by each.

Zev said...

What is the margin of error in any given election? Any time you are dealing with a million of anything there is always the potential for "data loss." This occurs in the voters being unable to express themselves clearly (eg contested ballots) or simply because poll workers are not careful with the ballots. Do you have any sense regarding how many ballots can be expected to be lost in any given election? My first thought is that less than 100 in a million vote contest is within this margin.

aaaaaaas said...


What is the margin of error in any given election? Any time you are dealing with a million of anything there is always the potential for "data loss." This occurs in the voters being unable to express themselves clearly (eg contested ballots) or simply because poll workers are not careful with the ballots. Do you have any sense regarding how many ballots can be expected to be lost in any given election? My first thought is that less than 100 in a million vote contest is within this margin.


The margin of votes is far less than the marigin of error, given the factors you mentioned and others. It would be perfectly rational to decide the election with a coin flip, saving the cost of a hand recount.

Alternatively, legalize dueling.

Michael (mbw) said...

For anyone who thinks that the precinct (Ward 3, P-1) was in D hands but made a mistake, please read this article published well before the current missing ballot ruckus.

http://www.tcdailyplanet.net/article/2008/11/28/screw-u-students-dinkytown-turned-away-polls-election-day.html?mini=eventcalendar/2008/12/all

The people running this precinct were doing everything they could to keep students from voting. They probably felt, as suggested in one of the merzbow's posts above, that really it was ok to throw out some of those votes because they came from improper sources.

Unless I'm mistaken one or more of these people belong in prison and may end up there.

MNLatteLiberal said...

FWIW,
I just got a Franken campaign mailing stating that they were ahead by 10. It was dated late this morning. Based on those two facts, I am surmising that the 10 number includes whatever votes lost yesterday in P1 W3, unless of course Coleman made a sudden gain of 12 in Scott/Wright.

~Latte

Jon said...

Almost all the votes yet to be recounted are in Scott and Wright Coubties. They went very heavily for Coleman. Nominally Coleman has 34,000 votes left to be recounted and Franken has 22000.

Michael (mbw) said...

Latte- A gain for C or 12 in the most recent count is well within the usual range, since these are mainly C votes coming in. A gain of another 34 or so for F (needed to keep him at +10 despite the lost/stolen W3P1 votes) is much less likely.

So I think the latest F count is holding onto his W3P1 votes, which we hope is what the official decision will be.

The big question about the most recent F releases is whether they are straight summaries of the judges' counts (as were the '84' and '73' releases) or whether they also include a few no-brainer probable successful challenges. If it's the former, we're in good shape. If it's the latter, then it's hard to call.

Then there are the absentees. And the awful possibility that the loss/theft could be ratified by the officials. But maybe not by the Senate. oy

merzbow said...

@mbw

So now it's not OK to even think about rechecking the validity of the votes in this district when it could benefit Coleman, but it's absolutely required that we do so for all of those rejected absentees because it could benefit Franken. You can believe whatever you want to believe to make you feel better, but wishful thinking is not going to change how the courts are going to look at this - as a double standard.

Michael (mbw) said...

@merzbow

Here's the point. There is no more reason to go back and check the eligibility of the voters in W3P1 than in any other precinct. That someone lost/swiped 133 ballots does not place the eligibility of the voters in doubt- it's entirely irrelevant.

Since there are at least a few cases known of improperly rejected absentees (no surprise, given time pressures) it makes sense to check the rejected absentees for mistakes. From what I hear, there's no guarantee at all that the results would help F. It would be ironic if he wins the recount, but then loses from the absentees. That's not the result I want, but nonetheless it's clear what the fair procedure is.

Michael (mbw) said...
This post has been removed by the author.
Michael (mbw) said...

@merzbo (again)
I should slightly modify my last claim. The possible connection between the eligibility of the voters in W3P1 and the missing ballots is this:
We know that election officials there viewed many of the students voting as ineligible, and tried somewhat successfully to stop them. That subjective sense could be what was on the mind of the person or persons who seem to have feloniously removed 133 ballots.

Your read is that since someone swiped the ballots, then that's evidence that the voters were no good. That confirms my impression that certain people involved felt those voters 'had it coming'.

green libertarian said...

Here's a tiny url (link was broken) to that story about the students being turned away at the suspect precinct:

As of Tuesday night, Hennepin County officials had yet to recount ballots from Minneapolis’ Precinct 1, Ward 3. But when they do, there likely won’t be as many ballots to count as there were voters who tried to cast them.

http://tinyurl.com/64ds5v

Something smells in Dinkytown.

ecarlson said...

My understanding is that the stated purpose of contacting the voters in the area where the votes are missing is NOT to verify that they are qualified voters, but rather to either (a) ascertain EXACTLY how many of them actually voted, or (b) let them have a do-over on their vote, since some of their votes are apparently missing.

The reason I chose the words "stated purpose" is because, from a practical standpoint, you simply cannot contact 2000 people and find all of them. Statistically, some of them will have moved, or won't have phone numbers listed, or won't be home when you go to meet them, or may even be dead. Hence following this plan effectively means following the backup plan, throwing the votes out.

I think the evidence that there are MISSING votes (not double-counted) is pretty much overwhelming, and though I'm slow to accuse someone of hiding ballots, the evidence is disturbingly strong. Nonetheless, if we follow the procedure I recommend, namely, treating the original machine counts of these 133 ballots as correct, you would effectively reverse the effects of this apparent crime (or gross incompetence), so it would hardly matter.

Personally, I'm hoping they go back through the machines and say, "look at all these ballots in here - wonder how we missed those before?"

Some of the other suggestions, i.e., let them flip a coin or have a reelection, are not realistic. You don't change the rules in the middle of the contest. A revote (under EXACTLY what conditions?) doesn't solve the problem, since it can just as easily be the revote that requires a recount. Flipping a coin when it's close doesn't either, since you have to define "close," which might require a recount as well.

I don't think anyone believes we can determine the will of those who voted to an accuracy of 20 (or even 50) votes. Although a coin toss might be equally fair at this point, the law says you only do that in an exact tie.

iness - Politician's former dedication to the public good has been replaced by iness focusing only on what gets I reelected.

Michael (mbw) said...

I again agree with ecarlson.
My additional point is that if all the people working W3P1 understand how long the prison term is for stealing ballots, and that the prosecutor will be an Obama appointee, the chance that the ballots will suddenly turn up under a box (to everyone's surprise and relief) will be increased. I'm assuming they haven't already been shredded, which would have been a very bad move by someone from Dinkytown. Are they called 'Dinks"?

net-e said...

In Mishap in the Midwest, the true life story of the Minnesota recount, Cynthia Reichert will be played by Shelley Long.

@Michael: Who you calling "Dinks"? [Okay, I don't live there now, but I used to.]

mediapost said...

Does anyone know of other precincts with missing ballots like this? The Franken campaign has claimed there are many precincts with missing ballots, but I haven't heard of any other precincts with this level of detail.

MNLatteLiberal said...

@mediapost,
In Washington County we were missing 8 ballots from one Woodbury precinct and 19 ballots from the other. A runner was dispatched from the County's office to Woodbury City Library where the machines were being stored to recheck the machines. Our observer went to oversee the process, and a Coleman person went as well. Nothing was found for either precinct.

In other Washington County precincts the # of missing ballots was nowhere near that big. AFAIR, the next largest was 5 (-3 for Al) and singles here and there.

The funny thing is, the Coleman folks could care less about the 5 and the 8 instances, and just by some weird coincidence that count just happened to break for them, netting them +6 total. In the missing 19 case, all hell broke loose because that broke against them, resulting in Norm losing 3, (11/8 afair only again - this was 2 weeks ago now).

As I was leaving for the day at 5PM they had a "team of lawyers" on its way to Stillwater. I was not there the next day on Saturday AM when the recount ended to see the aftermath. But the missing ballots were never found, which boggles the mind. I've written about this here a couple of weeks ago when it happened.

So, these instances I witnessed firsthand and was almost dispatched to go to Woodbury, in fact. Anecdotally, I've heard of other counties with missing ballots like this, but nothing that I've been able to follow up on personally.

~Latte

green libertarian said...

If a zealous precinct worker decided to disappear the 133 ballots, they'd have to be AWFULLY stupid to have not destroyed them by now.

I would expect that Ms. Reichert, along with a police detective, or someone from the prosecuting attorney's office, is interviewing each and every person who worked in that precinct and had access to the ballots, and people anywhere else who were in the chain of custody.
-taconite12

MNLatteLiberal said...

@green libertarian,
a bit off topic for everyone else, but thanks for your Strib posts. :)

T.W. said...

It looks more likely than not that 133 ballots have in fact gone missing; I have no idea what happens if they cannot be found.

Surely, the original counts should be used. A pretty accurate but complete count (given the accuracy levels revealed by this recount) will be more accurate than a perfectly accurate but incomplete recount.

Michael (mbw) said...
This post has been removed by the author.
Michael (mbw) said...

@ecarlson & T.W.

Given that in the rest of the W3P1 ballots (more than 10x the 133) there were a grand total of two challenges (one from each side) and that the vast majority of challenges will be rejected, by far the most probable difference between the original count and the hypothetical full hand-count of those 133 ballots is zero

green libertarian said...

MNLatteLiberal:

Thank you and for your direct participation in this process, and spot-on and humorous comments over there at the Strib.

Haven't been able to comment there since this morning. It doesn't seem to recognize me anymore, and can't find the proper place to login.
:(

merzbow said...

@ecarlson -

Deciding to ignore the results of a recount and fall back upon the word of a machine is the definition of "changing the rules in the middle of the process". If you want to entertain the possibility someone destroyed those ballots, it's also possible those ballots never existed in the first place and something was falsified, something malfunctioned, who knows what the heck happened. You can't throw fudge numbers into a recount because something may have existed that doesn't now exist. One ballot, one vote, otherwise it's not a recount, it's literary contest of which the winner is the one who can spin the best narrative of how he thinks things were on election night.

mediapost said...

@Latte,

Did you see the same situation in Washington as we're seeing here where the registration table and the machine counts were nearly identical but the number of ballots was short, or were you just comparing the number of ballots with the machine counts?

Michael (mbw) said...

@merzbow
That's an astoundingly insincere argument. It's not just the machines but also the sign-in sheets that show those ballots existed. While the machines aren't perfect, they have an error raate far below 1%. You're proposing that it is only fair to change the margin by 46 from the true count in order to avoid a remote chance of a machine changing it by +/- 1.

As for the 'rules', MN has one over-riding rule, clearly expressed in a Supreme Court decision: making every effort to accurately count the intended votes of legitimate voters. That's the rule which you're proposing throwing out.

merzbow said...

@mbw -

There are many other possibilities aside from the low possibility of benign malfunction. Again you seem to be ignoring the hypothesis of tampering having occurred to artificially increase the vote count while assuming there must have been tampering in the other direction to destroy the ballots. Probabilities could be argued about endlessly in court, hundreds of scenarios examined, as could scenarios for many other districts (all the mysteriously discovered votes, plus a NYT editorial today mentioned several other districts with double-digit numbers of missing votes). This is a recipe for a never-ending election.

Michael (mbw) said...

@merzbow

Please check my math on this, but I think 133 is actually triple digits, not double digits.

And sure, the same poll workers who spent all day fighting to keep Franken-leaning students from voting could have gotten together to fake a bunch of sign-ins and then run up the Franken counts on machines. Maybe out of remorse.

Or (I know this is far-fetched) one of them (let's call him Dink) could have grabbed a handful of ballots and destroyed them.

Like you say, it's all just narratives (and in this post-modern world who's to privilege one story over another?), so let's just forget the whole thing and say the Republican wins.

See you in court. Literally, I suspect.

MNLatteLiberal said...

@mediapost,
The short answer is "I don't know."

A longer answer is that at the recounting site we did not have access to the registration rolls/tables. All we had to compare is the certified total on the seal of the box with what we* counted up.

It was brought up that we need to check our totals with the registration rolls, but while I was there that was not yet done. I do not know the outcome of that comparison and have grown more and more curious about how that went. I suspect the situation was similar to P1W3: the numbers added up/checked on the election night, and a bunch of ballots went missing between the recount and the precinct.

How and where? Well, in the case of 19 or 8, it is conceivable (to me at least) that most of those were absentees that ended up being mis-delivered to the wrong precinct from the central site. We see evidence of that in the rejected ballots on the Strib site: ballots meant for Washington popping up all over the place and visa versa.

Now, based on the 600 ballots that I saw on the Strib site, that number is not anywhere near the thirty-odd missing from Washington County alone. In the 600, there, like, 6 or so marked as wrong precinct, afair. Extrapolating, it would be impossible for more than half of all the misdelivered ballots to end up being associated with Washington County.

That especially does not wash with these missing 133 from Dinkytown.

~Latte


*When I say "we", I was not observing at those Woodbury tables, I just finished a stint and was waiting to be rotated in when our observer and the site coordinator came to the computer to report this.)

ecarlson said...

@merzbow:

You need to include the best estimate of the number of votes. The first thing you have to determine is whether votes are missing or not. You have suggested that there are multiple explanations. But you haven't named even one.

If the higher vote counts represent fraud, malfunction, or incompetence, you are going to have to explain the near perfect agreement between the number of people signing in and the actual number of machine counted ballots. If we imagine, for example, fraud, we have to imagine someone who goes to the trouble to sign 133 fictitious names onto the roles, then electronically (but not physically) inserts 133 virtual votes into the system. If we imagine double feeding some ballots into the machine, we have to simultaneously imagine that more than six percent of the voters inadvertently signed in twice, and this number, by an amazing coincidence, is nearly identical with the number of write-in ballots. Such explanations stretch credulity.

[Incidentally, a lot of not-necessarily correct things have been said about write-ins, like the machines won't count them. Since the forms have a place to mark a write-in vote, the machine CAN count them, though they won't know who the vote is for. I am unconvinced by some of Nate's arguments, but the coincidences named before make such arguments unnecessary.]

Although one could simply throw out all the hand counting and use the machine counts, that is NOT what I am proposing. I want to use the machine count ONLY on the ballots that have disappeared. We know the machine counts to high accuracy, because machines are very consistent - give them the same ballot twice, and they will give the same answer. We also know from a variety of evidence that hand counting agrees with machine counting more than 99 percent of the time. The expectation value of errors is probably one or less.

@mbw -

Just because the challenges are low doesn't mean the machines are accurate. For example, if both sides were being reasonable, one could imagine the judge saying, "well, he didn't mark it right, but clearly he meant to vote for Franken" and the Coleman rep saying "yeah, I guess you're right. Plus one for Franken." Nonetheless, the expectation value of "errors" by my method for the missing ballots will be probably one or less, probably significantly less. Such a "recovery" of a significant number of otherwise disenfranchised voters is really pretty good.

reptas - Those who vote for Lizard People.

Michael (mbw) said...

@ecarlson

But if I remember the numbers you yourself posted, there was only about 1 possible change per 1000 machine votes. So the odds of anything being off in those 133 are low.

MNLatteLiberal said...

Somethin's funky. 12-4 updated SOS site has Coleman down by 2578 and Franken down by 2568, meaning that the overall delta is now 205. The difference in challenges is 3311 - 3197 = 114. That brings the simplest margin to 91, the lowest it's been since the recount started.

Meanwhile the Strib count, which always runs ahead (of itself in part) has the vote delta at 251 and the challenge delta at 114.
Not as sexy as the SOS numbers, but still good.

Unless I am missing something, there's been a net pick up for Franken somewhere because the Strib vote difference fell by over 50 with no real change in the challenge differential. The situation is even better looking at the SOS site.

Now for the $64K question, what am I missing? *g*

~Latte

wv conis. What coneheads use to procreate.

ecarlson said...

@ Michael (mbw)

I don't remember what numbers I posted before, but I'll take your word for it. Assuming this is a representative sample, that means a 1/7 expectation value, a number I would describe as "significantly less than one" but not "tiny" or "negligible"

Still, given a choice between PROBABLY making no error, though possibly making an error of one, and DEFINITELY making an error in excess of 30, what reasonable person would choose the latter?

astese - Dangling spurious statistics in front of someone who is anal-retentive.

mediapost said...

@Latte,

The SOS counts do not include W-3 P-1, so that explains the difference between the SOS and Strib.

MNLatteLiberal said...

@mediapost,
Thanks, but the crux of my issue/question/happiness is that both Strib and SOS sites show net gains for Franken in the last 24 hours. And I am unaware of any major gains/losses for Al after yesterday's tally was posted. Perhaps I should have been clearer.

These changes in the last 24 hours would not include the Maplewood 37 gain because that happened before the most recent cycle.

So, if anyone knows what's up, I'd be curious. Strib has nada, which is not surprising. I am done with work, so free to surf uptake and dailykos and what not. But if anyone knows something new/breaking, I'd appreciate the update.


~Latte

KWRegan said...

@merzbow and everyone:

The author of that NYT editorial sincerely believes that there are a "kergillion" planets out there that are just like Earth, except that YOU are about to be devoured by a giant carnivorous alien wombat! If you can't get Science Magazine access to read his article there, you can find a summary here a quarter way down (page-search "wombat").

When he started off on "Lizard People" I thought he was going to bring in the wombats. And the NY Times itself this year has been seriously into the theory behind them. But the catch may be that they are not carnivorous but rather papyrivorous, and having been blasted into the Ward-3 P-1 church by an exorcism, simply grabbed the nearest food around. In some universe, anyway...

Michael (mbw) said...

@ecarlson
I always hate to say this, but we're in exact agreement, even on terminology.

More importantly, for Nate: Kahlil Greene to StL? Do the numbers say that's a good move?

homunq said...

When will the withdrawn Franken challenges show up in the SoS numbers? I'm getting tired of reading these numbers with no caveat (for instance the last paragraph of http://www.cnn.com/2008/POLITICS/12/04/minnesota.recount/ ) and I don't think the media can report a Coleman +900 number with a straight face, so I can't wait.

mediapost said...

@Latte,

My take on it using SOS numbers. Franken picked up 98 votes today, 46 are due to W3 - P1. It was posted on SOS site yesterday but withdrawn today. That leaves us with a gain of 52. Yesterday Coleman had 156 more challenges than Franken, today he has 114 more. A difference of 42, so it looks like Al had a good day, probably picking up 10 votes. Al has been over challenging in Wright county which accounts for the decrease in the challenge gap.

MNLatteLiberal said...

@mediapost,
ah, thanks! that's the piece I was missing, that the SOS had the Dinky votes in yesterday's tally and then withdrew them today. Thanks again.
~Latte

mediapost said...

I haven't been optimistic for Al for the past week but, here's some guarded optimism. I went through 500 ballots on the STrib site (the ones I saw were almost exclusively red counties BTW). I assumed that no challenges would be upheld. I found 26 'new' votes for Franken versus 16 'new' votes for Coleman. By 'new' votes, I mean that ballots that clearly would have registered as either a non-vote or an overvote when the machine read them were deemed by the local official to be a valid vote for one or the other candidate. If what I saw was correct Franken is gaining about 10votes per 500 challenges. That extrapolates to 120 new votes over 6000 challenges. The margin stands at 205 on the SOS site with 114 extra Coleman challenges. 205 - 114 = 91. 120 - 91 = a 29 vote win for Al. Assuming of course that they can find those lost ballots or just use the machine count for W3-P1. I'm keeping my fingers crossed. In addition, Franken also had a slightly higher rate of challenging ballots that were deemed non-votes by the local officials, so losing those challenges will not result in Coleman gaining a vote.

KWRegan said...

Now for my more-serious post. I see Latte has ack'ed a reply to his first question---I saw it said too at TheUptake's live blog. To try to answer his other question on whether there was any serious movement today, I'd need to know more than this:

() Did Franken's "lead of 10" assume the 133 ballots would be found, or work from their being missing? Not only does the former make more sense, but I agree with a commenter above who noted that it makes a smaller change and is thus more likely.

So I'll use the STrib's numbers. The other question is:

() Was Franken's "10" a morning figure not counting today, or a mid-day figure from his presser?

Maybe I don't need the answer to this---I might be able to compensate by assuming my calculations yesterday were correct, and catching the variable which I can't solve for individually in a meaningful "Net". So let me call the margin "X" instead of 10, and now here goes with the new STrib figures 3,312, 3,198, and 251:

N_C = 3,312 - V_C
N_F = 3,198 - V_F
V_C - V_F = 251 + X. Therefore:

N_F - N_C = (3,198 - V_C) - (3,312 - V_F)
= V_C - V_F - 114
= 251 + X - 114 = 137 + X.

If one assumes that the candidate's #s of undervote challenges didn't change much between yesterday and today---or at least that they increased by a similar amount---then 137 + X should be close to the value of 144 I got yesterday (which was based on the "22"). Thus I get X around 7, i.e. not much different from the "10". Hence I gather that there was no swing of 50 votes for Franken---and maybe mediapost's answer resolved that for you as well. (Note that the alleged increase in Franken's V_F from Wright County is factored out by this analysis.)

The fact that the SoS recounted margin went from +11,000 Franken to +4,000 Franken indicates (of course) that most of the recounted precincts favored Coleman handily, so it's not surprising to suppose Coleman would gain some back. The STrib does show Minneapolis as all done, and maybe it's disappointing that the 24% that was left from the "76% done" yesterday was not more fertile to Franken than Wright to Coleman.

Finally, Franken did succeed...in punting Coleman below 1,209,000! Before 6xx challenges apiece are withdrawn, that took cojones! From TheUptake.org, the STrib now has 1,794 (or more) ballots online, and Reachert is said to have asserted that the missing ballots are in an envelope marked "1/5". This is not an allusion to the Mad Hatter's tea party, because he wore the fraction "10/6".

Michael (mbw) said...

@KWRegan- That seems like a reasonable approximate calculation.

Is there any good evidence about whether the Franken releases switched from straight reporting (as in 'C+84', then 'C+73') to including some opinions about challenges (maybe in 'F+22', 'F+10')? If so we shouldn't be very confident. If not, things look fairly good.

KWRegan said...

I haven't taken the time to listen to the tape of Franken's press conferences, to parse the words myself. As I said in the other thread, that's the reporters' job---and IMHO they're falling down on it. They should clarify by direct question, rather than write their articles to say by "Induction" that Franken's behavior is continuing the same. That's how you get spun. I'm from a journalist family as well as being a math/scientist, and I've really noticed the loss in precision of reporting over 30 years...

In my analysis, if you suppose that Coleman has made Y more undervote challenges since yesterday morning than Franken, then the margin I get is 7 - Y. I think of Y as positive since the precincts involved favor Coleman markedly, but Y could be negative too. Anyway, it's arguing single digits when the error bars even of events alone are double-digits-and-higher.

Cubic said...

@Michael (mbw)
Franken's team today said again that their count assumes all Challenges are rejected. So it's safe to take them at their word. In fact, they said they anticipate gaining votes in the Challenge process, but those are NOT currently included in their tally.

Gaining 10 votes today or so (as suggested by the figures this evening) is very good for Franken, incidentally. Even if the 133 votes aren't found, there's at least an outside chance Franken could win without them with a strong pickup (26 votes) during the Challenge count.

Cubic said...

And yes, I did listen to the presser today; I've been listening to almost all of them and they've never changed from saying their count "assumes all challenges are rejected"

Cubic said...

Also, the "F+10" figure was based on the totals as of the end of counting Weds night and DOES assume the 133 ballots are found. This specific question was asked at the presser and the Franken camp directly answered it.

Michael (mbw) said...

@cubic- Thanks!

It looks highly likely F will win, unless the W3P1 loss/theft stands. The Senate could choose to fix that.

Cubic said...

@Michael(mbw)

You're welcome :) I think highly likely might be a stretch, but the odds are somewhat north of 50% if Franken's team is telling the truth. That said, who knows what the remaining ballots will hold tomorrow (I doubt Coleman will net 20ish though).

On the final 24% of Minneapolis being mediocre, I have to wonder whether the StarTrib was behind the campaign's knowledge - my hunch now is that it was that 24% that gave Franken his 36ish vote boost in addition to the 36 votes he netted from the 171 found ballots on Tuesday - but that the StarTrib didn't report them until Weds (late counting?)

Cubic said...

I suppose I should add that the Franken team often references 538.com specifically in its press conferences - it did today in particular, referencing this post by Nate!

green libertarian said...

Now for the $64K question, what am I missing? *g*

~Latte


The alignment of Mars, Venus, and the Crescent Moon, obviously.

heh heh heh

H. said...

Can anyone explain the discrepancy betw the stats on the official MN SoS site (found here: http://electionresults.sos.state.mn.us/20081104/SenateRecount.asp) and the figures quoted on CNN?

SoS site indicates Franken in front by 4000+, while CNN and others state Coleman leads by anywhere from 100 to approx 300.

Greatly appreciate clarification, thank you ~

Dr. H. Kampar

Michael (mbw) said...

@H- The answers to all your questions appear in the thread above.

On the 133: Coleman's lawyer is now openly denying that the votes exist.(In some literal sense he may now be right.) So it just happens that 134 more people signed in than voted, and that the machines somehow registered 133 more votes than they should have, and that the precinct workers, on a whim, stored the ballots in envelopes numbered 2-5. Maybe they had a superstition about 1. And that all this happened in the one precinct in the state in which the precinct workers had created a tremendous, partly successful, ruckus by trying to prevent Franken-leaning students from voting.

I had been assuming that some hot-shot kid (see 'Dink') thought he was being cute swiping that envelope of ballots and maybe swinging the election his way, making up for what he perceived (see 'merzbow') to be votes from people who shouldn't be eligible.

In light of the current official Coleman position, it now seems possible that the entire effort to disenfranchise Dinkytown was deliberate from start to finish.

MNLatteLiberal said...

@Dr. Kampar,
This has been discussed ad nauseam here, including in this thread, but I will repeat, as imho it bears repeating.

SOS site started their recount from scratch, assuming nothing. Every day the re-vote tally would come in from the counties in the form of completed precincts, and SOS would post that tally each night at 8 PM.

Now, since the counties run the recount through the precincts alphabetically as they recount, the SOS numbers at times showed Coleman in the lead, at other times, like for the last 3 dasy - they show Franken leading and leading by a significant amount.

The current bump is due to the Hennepin and Ramsey counties having wrapped up. Those are the largest in the state, and trend strong blue, giving Franken the bulk of his votes. In the last 2-3 days (Wed, Thur and today), as every other county had already wrapped up their recount, save for Wright and Scott (strongly red), the Franken lead has been closing from 11,000 + to whatever it is now.

Hope that explains the SOS number(s).

As to the Strib numbers that are picked up by the major outlets, including CNN afaik, the paper focuses on the initial delta between the two candidates before the recount started. That number was 215 in favor of Coleman. Every day, as the counties reported, the paper focused as to how that number was affected by both found and lost votes for each candidate.

Because a challenged ballot with a vote for either candidate is recorded as as lost for him in Strib tally, the initial delta is far more sensitive to the challenges than to the "new" votes found, as there are >6200 challenges and only a few hundred "new" or found undervotes. The vast majority of the 6200 challenges are deemed frivolous by all sides. And so, the Strib number is obscured by that temporary (until the Canvassing Board reviews them) subtraction.

I hope that explains the Strib method with the proviso that Strib tended to run ahead of SOS and ahead of itself quite a bit in the recount.

There is a third methodology, one employed by Mark Elias, which is perhaps the most accurate. That is discussed in this thread as well up above. He does not subtract the challenged votes from either candidate, as most of those will be tossed anyway, opting instead to stick with the initial judge's ruling. That number is viewed by most posters here as the most accurate for all the aforementioned reasons.

~Latte

mediapost said...

Looks like some good news for Al:

10 new absentee ballots were found while they were searching for the missing 133. Just more evidence that the margin of error in this race will well exceed the margin of victory. For some reason they seem unsure if the ballots will get counted.

Full Article:
http://www.startribune.com/politics/national/senate/35611679.html?elr=KArksLckD8EQDUoaEyqyP4O:DW3ckUiD3aPc:_Yyc:aULPQL7PQLanchO7DiUs

mediapost said...

Oops, make that 12 new ballots.

Michael (mbw) said...

BREAKING NEWS

Franken says that, following the election judges' calls (and not counting those 12 new ballots) he has won by 4 votes. It's unclear if that includes the 133 (+46 F margin), but I'm afraid it does. However. it seems most likely they will in fact be counted.

This leaves a few sources of possible error:

1. Maybe the board will toss the lost/stolen count, but I doubt it.
2. Some dozens of the challenges will probably be accepted, giving purely statistical error bars of about 10, in addition to uncertain asymmetry in the quality of the challenges.
3. Those last 12 military absentees, from Franken territory, will probably be counted, adding a tiny bit of uncertainty.
4. Whoever ends up down will probably try to get 500-1000 absentees counted, with both random and systematic uncertainties in the neighborhood of 25.

So at this well-defined point in the process it looks like F+4, +/- 30.

MNLatteLiberal said...

@Michael (mbw),
Acc. to SOS interview on MPR today at 11 AM and maybe even his press conference on uptake.org (listening to it now), Mark Ritchie stated the following:
1) Until 133 are found the recount isn't over.
2) They have until Dec. 16 to find them
3) Should the ballots not be found, they will count as the machine count went and that restores +46 to Franken.
4) Precedents for such actions exist and Ritchie sited them.
5) MPR and uptake broke the <20 military absentee ballots found right around 11 AM, scooping Strib. Strib then reported it was 12 and they were already opened.
(I wonder how they break - a lot of speculation on that all over the place)
~Latte

Michael (mbw) said...

@latte- Thanks.

Just checking on one key thing. F's +4 does already include the +46? I hope not, because +50 would be really hard to beat, but it sounds like it does. Is that clarified in what you've seen?

MNLatteLiberal said...

I have no idea, Michael. We are in the same boat there: hoping against hope that +4 does not include, but fearing that it does.

One way to check and I cannot do this till tonight:
- SOS site now shows "completed" recount numbers.
- The only thing missing is the Ward in question, missing altogether.
- You can prob. figure out what the original count in P1W3/Hennepin was using all the info on hand (Strib breakdown, SOS site, 2028 number breakdown, etc.)
- You can get an idea what that "lowest common denominator" (i.e. recount vote diff less challenge diff) difference is today and compare it to yesterday's.
- Contrast it with the Franken campaign count of +10 vs +4 yesterday.

For example, using that procedure yesterday, I got -91 lcd number (for Coleman) when Franken had +4.
That gives you a feel for the undervote/type (2 was it?) challenges (= 95 for Franken) difference between the two. Assuming that changed only minimally, if you compare today's numbers to that, it will give you an idea whether +4 includes the 46 or not.

Hope I am clear enough, but i know i am fuzzy. Gotta run,
~Latte

KWRegan said...

The STrib's numbers seemt o have stabilized, with a margin of 192. Weirdly, this is exactly 46 less than their previous margin of 238, which equals the recorded Franken net from the missing 133 W3P1 ballots, but I don't know if that's the adjustment.

On a similar note, for another infuriating press contradiction, The Minnesota Independent here emphasizes "(The Franken campaign’s count relies on the assumption that these ballots will ultimately be included in the recount.)" Whereas Kos himself writes here, "...the Franken campaign claims it would win by four votes. That number does not include the 133 missing ballots from a Minneapolis precinct that have likely cost Franken about 36 votes. A state election official has said that if the ballots aren't found, they would probably stick with the election night numbers, giving Franken a bit more of a cushion."

Reiterating my method here, let's see what the math says about all this, writing "X" for the margin can comparing with X=4 later, and going with the new STrib figures 3,375, 3,280, and margin 192:

N_C = 3,375 - V_C
N_F = 3,280 - V_F
V_C - V_F = 192 + X. Therefore:

N_F - N_C = (3,280 - V_C) - (3,375 - V_F)
= V_C - V_F - 95
= 192 + X - 95 = 97 + X.

This is indeed off by about 40-47 votes from my values of "144+X" and "137+X", in a way that ostensibly favors Franken. It seems to mean one of the following---probably not a combination since two of the possibilities are "deltas" of exactly 46, so exactly one:

() Kos is right, which means Franken gained 40-odd ultimately-real votes since mid-day yesterday.
() The STrib has (over-?)compensated for the missing ballots in going from a 238 to 192 margin.
() Coleman has made a few dozen more undervote challenges than Franken since mid-day yesterday, so my assumption of near-continuity in "N_F - N_C" is wrong.

The last possibility would be good news for Coleman, and will be objectively verifiable before the first is, but doesn't seem to square with the small increase in overall challenges by Coleman (only +63 from before) and overall decrease in the "challenge gap". So I tentatively conclude that either the STrib has overcompensated in Franken's favor (or undid an undercompensation from yesterday), or Franken had a really good final two days. Help checking this? (NB: The SoS tallies are behind, showing "99.93%" done. None of this includes the 12(?) military absentee ballots found today, which haven't been tallied yet.)

Ah, I see a few more comments from you all touching on exactly the same uncertainties---hope my math helps...

MNLatteLiberal said...

sorry, sloppy there:
yesterday's Franken number was +10, so my nominal undervote/type 2 difference should be +101 for Franken.

mediapost said...

Just noticed that STrib has removed W3-P1 from its counts. So SOS and STrib should match tonight.

PS I hope those military saw Al's USO show with the Patriot cheeleaders coming onstage in full burka and then stripping down to cheerleader uniforms. That would probably net him a few votes.

KWRegan said...

Ah, I see: that would mean yesterday's figure was using the recounted figures from W3-P1, and they've gone back to the Election Night figures? That would indeed reduce the margin by 46...

...and it could mean Kos is right. (Meant to say: the boldface emphasis was his, not mine.) Alternately it could mean my "N_F - N_C similar" assumption was wrong *yesterday*. I'll wait until seeing the numbers after 9pm ET tonight before looking again.

MNLatteLiberal said...

Nate desperately needs to start a new thread/post on the end of the recount and provide some insight into the role of the missing 133 in the Franken +4 tally. It's well past due. He's got a direct line to Elias; he's got the tools and there is nothing else going on in the world :).

On a related note, I would like to see some educated conjecture from him on
a) the anticipated outcome/breakdown of the "5th pile" (of the rejected absentee ballots), including, but not limited to
i) total anticipated number in the pile
ii) vote distribution projections.
and
b) the 12 military overseas absentee ballots from the warehouse, same analysis as in a), but instead of i) (for we know apriori there are 12), he ought to weigh in whether the State Canvassing Board is going to count them or have someone else count them first or refuse to have them counted altogether.

Is that so much to ask?
*g*
~Latte

wv: homarymp.
After sitting here and reading/typing all the posts I've developed this huge homarymp in my posterior.

mediapost said...

Does anyone know the scoop on withdrawn challenges? I've heard both campaigns make promises for several days but have not seen any challenge numbers drop. In fact the last 2 days saw another 500 added in Scott and Wright.

MNLatteLiberal said...

@mediapost,
half of Ritchie's news conference today on theuptake.org (they removed the feed for some reason) was dedicated to those two issues, so you are in sync with our SOS, FWIW :)

a) Ritchie said he was very disappointed by the very high number of challenges coming just today and just from one county. Then he rubbed his head,
and
b) He talked quite a bit about what a royal pain (not his words) it is to remove the challenges: the officials have to physically locate each challenges to be removed one by one, add the tally to the corresponding precinct/county, place the ballot back into the proper stack in the its storage facility somewhere in the great State of MN. Then they have to adjust all the "linked" documents to that challenge, one by one. Then they go to the next.
This is what's ahead for them. 633+650 times + more to be announced soon.

I do not envy them. The underbelly of the democracy is not a sight for the those with weak stomachs, I
~Latte.

mediapost said...

Hey Latte,

What is your take on this ballot? A Lake Of The Woods primary ballot. It doesn't even have the presidential race on it. It can't be legal, can it? It was a Coleman vote and Franken has challenged, so evidentally the local officials think it is valid.

http://senaterecount.startribune.com/media/ballotPDFs/Lake%20of%20the%20Woods_2A_challenged%20ballot%201.pdf

KWRegan said...

Here's the state of confusion over the "Franken +4" margin:

Franken's Recount Update - Dec. 5th says in its first item, "...And, at the end of the hand count, Al Franken leads Norm Coleman by 4 votes. Many media outlets are calculating the margin by a different method, relying on raw data from the Secretary of State's website to conclude that Coleman holds a lead of a few hundred votes. However, that calculation leaves out a Minneapolis precinct of 2,029 votes which is still being counted." A DailyKos commenter here deduces that since "...they are criticizing others for not using the outstanding Minneapolis numbers, [this] suggests Franken's camp is using those numbers (and since the hand recount is not finished, they must be using the Nov 4 numbers). [Emphasis in original]

However, Franken's entry links to this Politico item, which says: "The number from Franken's internal tally does not include the Minneapolis precinct where 133 ballots went missing. The Franken count includes the thousands of disputed ballots that both campaigns objected to during the recount process." This seems to say that Franken is not including the 133 ballots (which after all might be phantom), but look again: it says "does not include the precinct" which makes no sense, since if all of Dinkytown didn't exist Franken would be toast. Aaaargh! It doesn't take a doctorate in logic from Oxford to say that the language imprecision in the reporting is frustrating!

However, nothing has changed in the numbers since mid-afternoon, and I see the SoS tally differs from the STrib's in exactly the 1,090 Franken, 595 Coleman, reported by W3-P1 on election night. And I think the numbers show that the last two days were also good for Franken. A simpler way of putting it: Over any stretch of time where you hold my "N_F - N_C" constant, you're basically just tracking the difference of the reported margin and the "challenge gap", which TheUptake has been calling their ("crude") "Challenge Adjusted Recount Margin". And indeed this is listed as going from +137 Coleman this morning to +97 now.

This indicates a combination of () Franken gains, and () Coleman making more undervote challenges than Franken since my last check. But since Coleman made only 63 more challenges overall since that check, I find a lot of the latter unlikely. So until this is clarified, my tea leaves say Franken had a good two days even with the Dinkytown lapse. And if Dinkytown winds up costing Al the Senate, Bob Dylan can return to his roots and write a Ballad of Ballot Blues for him!

KWRegan said...

@mediapost: There must be a "back-story" to that ballot which doesn't show in the STrib's framework. Something like: voter flubs Senate race, doesn't want to fill out a whole new ballot, helpful precinct volunteer has the bright idea of pulling out an old Sep 9 primary ballot, since it has all the names. If so, then an argument that the ballot misrepresented the Senate race, and hence should be void, is blunted by the voter's intent having been formed from the original correctly-represented ballot.

MNLatteLiberal said...

@mediapost,
To KWRegan's explanation, I'd add that the ballot bears a handwritten "Original #3" on its top. That indicates to me that a duplicate ballot was ultimately used in its place to feed through the machine.

Let me explain why I think that. If you already know all this, I apologize. Duplicates are used (and the words "original #n" are inscribed) in two cases that I know of in a precinct. One is when an absentee vote is opened and it is cast in whatever form is unacceptable and cannot be fed through an optical reader. Sometimes these are overseas absentee ballots printed off electronically and thus are the wrong format. Other times the original is badly folded/chewed up.

The second instance is more rare. It is when a voter casts a ballot and leaves. And for whatever reason, the machine spits out the ballot, but the voter is gone and cannot be found/chased/whatever.

In both cases, two election judges together fill out this duplicate ballot, and if there is more than one, number them. The larger precincts with, say 2000 voters, often have a dozen or so duplicates casts.

In those cases, the duplicate is marked with a similar inscription "Duplicate #n" and is then found in the midst of the otherwise "normal" ballots in the sealed ballot boxes during/before the recount. The originals are sealed in a separate enveloped all together and are signed and yadi yadi yada. And kept separately. Then, during the recount, you have to fish out the dupes from the pile at the end of the count and make sure they match up with the originals. Then the fun really begins if they do not exactly match up. But that's another story.

Finally, the originals are resealed after the recount, the dupes are returned to the common pile (this time sorted for each candidate) and life goes on.

What looks odd about this ballot is that what election judge would accept an original when it excludes all other choices but the Republican. It's an invalid ballot, imho, and I could see Franken folks objecting to it. Voter intent is clear, but the voter for whatever reason was not given the choices and the Reps had an unfair advantage.

I agree with KWR *surprise! :)*, in that there must be a background story on this puppy. There must, and it must be a damn good one. I'd like to know what it is too. I have not seen this ballot before, I should add.

Finally, @KWRegan and mediapost and anyone else, fwiw, I came to the same conclusion as KWR after reading the politico story a couple of hours back. Even earlier this afternoon some miniurl or tinyurl (you have to forgive me here, I am not up to all the latest wigits/chatter methods on the web yet) popped up on the theuptake.org while Ritchie was speaking. And when I went there, the same story was there, stating basically that the Franken people are saying that the 4 number does not include the 46/133. Then after everybody else, dailykos ran it, as KWR (?) posted, afair.

I am off to bed. I have been playing on the Strib site all night and my wife is rightfully mad at me. Good night, all
~Latte

KWRegan said...

Thanks! Meanwhile, I went to bed and then started up with the realization that as a scientist I'm obliged to put prediction where my mouth is. In this case that means predicting the Predictor. I predict that Nate will re-regress tomorrow, will clarify the "+4" issue as you niftily summarized, and will likewise say he is not including the 133 ballots but does include the rest of the precinct. Since the 133 ballots stayed more in-proportion than Maplewood's find this makes even less difference to his parameters---you can just add the "+46" on top afterward. Now for the real prediction: I predict that 2 of his models will show Franken with an almost-50 vote lead (which would come true if Franken netted 40+ on the undervote challenges), 3 will show a Coleman lead between 50 and 75, and the remaining 3 will show Coleman at 20+, "but that becomes Franken 20+ if the election night totals in W3-P1 are used, as by precedent they should be." Then he'll note that with the 133 ballots included the results basically slightly favor Franken. Then there's still the 12 military absentees to count, and then the improperly-rejected absentees will be counted, because the equality usage in Bush v. Gore underscores the state's responsibility to treat the improper rejection as if it were a class action, given that some individual(s) supporting Franken have already filed suit as private citizens over this.

I have not taken time to run any code or even fiddle with the spreadsheets. G'night!

Michael (mbw) said...

Damn. Here's the ambiguity. "Doesn't include the 133..." can either mean "doesn't include those votes" or "doesn't include the correction to the original count caused by the loss/theft of those votes". It's hard to see, tracking the numbers over the days, that the +4 could really include the loss of 46. I hope I'm wrong. Fortunately, the SoS has made it clear those votes will still count. If they're already in the +4, the election is still a toss-up.

Michael (mbw) said...

Aha- The Franken +4 does include the original count from W3P1:

from Strib: "The Franken campaign says that its tally shows Franken ahead by four votes. Like the Star Tribune, the campaign includes the pre-recount tally from the disputed Minneapolis precinct."

So that's it. It's +4 now, with god knows what from accepted challenges, etc. to come.

green libertarian said...

PS I hope those military saw Al's USO show with the Patriot cheeleaders coming onstage in full burka and then stripping down to cheerleader uniforms. That would probably net him a few votes.

LMAO!!!!!!

wv rawylic

In the interest of chivalry, no comment.
-taconite12

KWRegan said...

< Michael (mbw), quoting this, which I now see comes up when you mouse-over "See more related items":

"The Franken campaign says that its tally shows Franken ahead by four votes. Like the Star Tribune, the campaign includes the pre-recount tally from the disputed Minneapolis precinct."

OK, that makes better a-priori sense: it means Coleman picked up 18 votes since Franken's claim of "+22", over 2-1/2 days dominated by Coleman-friendly Wright and Scott counties, which border on Hennepin. And if my figuring is correct, it must mean Coleman made 40-odd more undervote challenges ("Types 2 & 3" in Nate's scheme) than Franken over that period. Maybe my Thursday check from the STrib's figures was out-of-sync with when they happened. Still, looking at the two counties, it's hard to see. In Wright, Coleman was -216 from 244 Franken challenges, Franken -135 from 143 by Coleman. In Scott, Franken was -55 from 65 Coleman challenges, Coleman -72 from 91 Franken challenges. Franken made many more challenges, 335 to 208, while Franken went -190, Coleman -288. If we match up minuses to challenges (incorrect but maybe not far off), that leaves at most 18 undervote challenges by Coleman, to 47 for Franken. Probably 20+ of those 47 for Franken were also on Coleman votes, but my point is, those 20+ should be washed out of my figuring. So I'm stumped.

KWRegan said...

OK, just noticed another thing: at TheUptake's live blog, there's an entry at 8:13pm CST last night by Mike McIntee saying, "The Franken campaign says it is 4 votes ahead as of Thursday night." (emphasis mine). So maybe Friday was a good morning for Franken, but this wasn't known by his lunchtime presser. Too many permutations!

Anyway, let me explain what I'm trying to isolate by an example. Suppose these two things happen, say in a Wright or Scott precinct:

(a) A ballot voting Coleman that wasn't picked up by the machine, say because of light red ink. Election judge rules it valid; Franken challenges.
(b) A ballot with a drip of ink in Coleman's bubble. Judge says no, Coleman challenges. Now:

() Neither event changes the STrib's totals, not +, not -.
() Neither event changes the SoS totals.
() The two events do not change the "challenge gap", since both made 1 challenge.
() But Franken's internal margin "X" goes down by 1, since the judge ruled a vote for Coleman.
() And my N_F - N_C also goes down by 1, since Coleman made 1 undervote challenge, while Franken made a vote challenge.

Since midday Wednesday, based on Franken's "X", my N_F - N_C has gone from 144 to 97+"4" = 101. So I infer that after you sift out the STrib's vote and challenge figures, there must have been 43 such events pro-Coleman.

The other reason I'm posting this example is to ask, what is its effect on Nate's equations? If it contributed net-nothing to Nate's equation for that precinct (because the challenges are equal and neither shows a gain or loss), then his 8 models would all be unsound. So it must contribute something, and since both (a) and (b) are ostensibly favorable to Coleman, I suspect it would make two pushes---showing up as moving his outlier models by 2 to move their center by 1. So though I'm hedging my prevous prediction, I can still make a nontrivial one: If Kos is right, then even without the 133 W3-P1 ballots, no model will show more than Coleman +75. But if the STrib is right, then with those ballots, at least one model will top Coleman +100.

Michael (mbw) said...

@KWRegan
Kos is ambiguous and, if taken in the optimistic sense, has low prior probability.
Strib is unambiguous and has high prior probability.
Obviously, unfortunately, we should assume Strib is right.

Now it looks like this is the situation (prior to last 12 uncounted ballots, absentee changes...):
If all challenges are rejected, F wins by 4.
If all challenges are accepted, C wins by 192.

This sounds pretty bad, since if there's no systematic differences between the Board and the local election judges, the accepted challenges should be distributed randomly. That means that any accepted challenge rate greater than 2% shifts the expectation of a win to C. This already takes into account the differences in challenge types between F and C.

We're pretty sure that more than 2% will be accepted. That means that unless there's some systematic tilt, or some effect from absentees, etc., things look pretty poor. The Intrade numbers make sense, for a change.

One possible hope is that the Bpard will have a slight tendency to accept more votes than the local judges. There are probably more F ballots near the margins of countability, so that would produce a slight F shift. Of course if the Board is more restrictive it would go the other way.

Cubic said...

@mbw:
I actually disagree that things look pretty bad. Many Minn election workers have said that the odds of a challenge being accepted are in the range of 1% - I haven't heard anyone say they assume that more than 2% will be accepted. Joe Mansky himself managed to make that initial challenge disarmament agreement in St. Paul by stating that he was going to win all of the challenges made the first two days.

Further, not all challenges are created equal - as we have said numerous times, a lot depends on the quality of the challenges - Franken's team says they expect to net ballots during the challenge phase, and if they're actually up 4, all they need to do is NOT lose any of the challenges. Additionally, don't forget that Coleman has more challenges in this situation, so the true differential is not F+4->C+192, but more like F+4 to C+100, assuming (as I do) that not all challenges are equally likely to be accepted.

At base, there are a lot of unknowns in this process. I think Intrade is actually a relatively poor measure of success here because EVERYONE is operating under a poor amount of information. If you gave me a percentage number right now, I would say Franken's odds are just under 50%, but that's because of uncertainty in terms of how the Dinkytown ballots are addressed.

Incidentally, the current number of challenges i 3280 from Franken, 3375 from Coleman, and we have strong evidence that more of Franken's challenges are to non-votes than Coleman's are. That said, we expect that somewhere around 80% of the total challenges (giving the losses in votes of both candidates) are to each other. This does mean maybe Coleman has more to gain from his challenges being accepted (since those could both strike a Franken vote and create a Coleman vote), but those challenges are LESS likely to be accepted since they were already accepted by the machines! As a result, it's probably advantageous to have a higher percentage of challenged ballots be from non-votes, since there are many situations where a machine will fail to count an actual vote (outside bubbling, writing, etc).n

The canvassing board will be working from 5,372 challenged ballots. 1% Acceptance would be 54 ballots being accepted, 2% Acceptance would be 108 ballots being accepted, while 3% Acceptance would be 162 ballots being accepted.

In the 1% scenario, Coleman would need to win the 54 ballots by 29-25, or about 52.5%-47.5%. In the 2% scenario, Coleman would need to win the 102 ballots 56-52, or about 52%-48%. In the 3% scenario, Coleman would need to win the ballots 83-79, or about 51.2%-48.8%. Could those situations happen? Absolutely, but I think it's a near toss-up.

Finally, I expect given that the 12 new ballots are from Minn (and Franken's numerous USO shows) for him to net between 0-2 ballots out of that pool. If he gets lucky and wins them 9-2 or something, I think the odds he loses becomes moderately low, while if Coleman wins the ballots by 3 or more votes, he becomes the clear favorite.

This doesn't even begin to address the rejected absentees, mind you.

Cubic said...

One final point, Nate noted that about 10% required any kind of a "judgment call" but that doesn't mean 10% will be accepted. There's going to be some kind of natural preference for going with the call at the Table barring a preponderance of the evidence, so I still believe 1-2% is the accurate number of Challenges that will be accepted.

Also, not all election judges are created equal across the state. If there's one thing we've learned from this recount, it's that you can't assume uniformity across different counties, since the projections for victory have swung wildly as different places have been counted.

Cubic said...

Final final point:

The 5372 number comes from reducing both sides challenges by the amount they "unchallenged" 633 from Franken, 650 from Coleman. Of that 5372 count, 49.2% of the challenges are from Franken, 50.8 are from Coleman. Still, given that the challenged ballots aren't the same, I think Franken will do better than 50/50 on the challenges for the reasons listed above.

just_looking said...

Michael (mbw),

Wouldn't the break-even point be a 4% challenge success rate (assuming the +4 Franken figure for all challenges fail is accurate)?

Coleman must have 196 more challenges of the variety where the board said it was a vote (the difference between +4 and -192 Franken). Franken must therefore have 101 more challenges of the variety where the board said no vote.

At a 2% success rate, Coleman makes up 4 votes in the first variey, but loses 2 votes in the second.

KWRegan said...

Agreed, Politico (too) was ambiguous and the STrib note which I missed has more cred. Can't tell whether TheUpdate's McIntee mis-spoke in dating the "+4" to "as of Thu. night"---no source unambiguously says that Elias' 12:30pm CST use of that figure included Friday morning results.

Disagree on the ballot analysis: the STrib margin of 192 does not mean the result if all challenges are accepted. You would get a contribution from the undervote challenges as well, which is not reflected in the STrib's tally. Indeed, you can get this from my example above: Let's say the candidates were tied and those were the only two challenges in the entire recount. Since neither was picked up by the machine, neither was in the STrib's tally before or after. But when both challenges are accepted, Coleman nets a vote. And if both are denied, Coleman nets a vote. It cuts both ways. As an exercise, suppose Franken withdraws 780 not 633 truly-frivolous vote-challenges, and Coleman goes 95 not 17 past him to withdraw 875. That leaves both on 2,500 challenges, and a STrib gap of 97. Now assign a% likelihood to a "vote challenge" being accepted, and b% to an "undervote challenge". You'll find that whenever a = b, Franken wins by 4. But Nate basically asserted b > a in his Challenged Ballot Primer, and that favors Franken even more.

I do have to hedge something else, though: The net effect of my two hypothetical ballots is---and can only be---to increase the error bars in Nate's model. The reason is that if you reverse Coleman and Franken, none of the observable figures changes, while the underlying situation flips to favoring Franken. This sounds like Heisenberg's principle, but it's actually kind-of how the proof of the Pauli Exclusion Principle goes! What Nate would hope is that the two ballots came up in different precincts, where their effects could be observed separately. This underscores that his regression model is likewise trying to take advantage of "lumpiness" in the data---in the extreme of all the action being in just 1 precinct, it could say no more than my algebra above. By now of course it wouldn't be predicting how the rest of the hand-count would go, but it could still be predicting the distribution and disposition of the challenges.

Cubic said...

Agree with KWR; one other thing - counting was light on Friday and done a full hour before the Franken presser, and they said they were up 4 with the recount done, so I believe that includes Friday.

And I concur with KWR on the b>a analysis; the undervotes are inherently more likely to be accepted. I tend to think we've gotten so many challenges b/c while Franken was gaining for the first 2 days clearly, Coleman's team realized they could mask this and they were (per Nate's figures) the ones to start the escalation.

And again, Franken's team, who has a copy of ALL the challenged ballots and the record of initial rulings, says they expect to net ballots during this upcoming phase. Coleman's team simply expects to win the recount.

Both sides think it's close and I imagine both sides already have lawsuits brewing.

One thing is for sure, that 171 ballot find was clutch.

MNLatteLiberal said...

@KWRegan,
I am still a little behind due to some serious tennis playing this AM, but I think our math/analysis is fairly similar, though we name our variables differently.

From my yesterday's post(s) 3:13 and 3:16PM: I derived your "X" to be 101 for the Wednesday numbers using the challenge differential adjusted voted difference (91 that day) and the Franken campaign "internal/announced" number (10).

Yesterday at the end of the recount, that X was still 101: 97 for the chal. diff. adj. vote delta + 4 for the internal Franken number. 101.

The CDAVD (LOL) slipped 6 overnight and so did the Franken campaign announced lead.

So, I think the 4 figure has to include the 133 missing ballots and factors in the 46 they are not going to lose as a result of those ballots missing.
FWIW.

~Latte

PS: Nate seems to be all over the place, just not MN :(

just_looking said...

Yeah, KWR is right. Franken would lose by 91 (192-101) if all challenges were accepted.

Michael (mbw) said...

@everybody

my bad. My calculation was bullshit, which gradually dawned on me an hour later. Glad to see so many people who are thinking more clearly.

Michael (mbw) said...

The one thing I can't follow in the generally correct arguments above is KWR's remark about how it's like the proof of the Pauli exclusion principle. Spin-statistics? Expressing the non-commuting relativistic boost operators in terms of creation and annihilation operators in order to calculate the commutation relations? Maybe you skipped a step.

A more prosaic way to put that point might be that we have four major unknowns (two major types of challenge types from two sides) and only 3 equations (each side's total challenge number and the difference between the challenged 192 and Franken 4. The problem is under-constrained.

MNLatteLiberal said...

One last thing :)
Another challenge figure of merit is (afair) something Nate might have mentioned: look at how many opponent's votes party's total challenges knock out.

Using Strib numbers, For Coleman,
3375 challenges resulted in reducing Franken tally by 747.
For Franken camp that figure is
3280 - 2651 = 629.

That means that there are 747 net new votes found for Franken LESS any non-Franken count reducing challenges (such as if, for example, a Coleman supporter objected to a Barley vote, claiming it to a Coleman ballot). For Coleman that number is 629. So, despite launching 95 challenges more, Coleman managed to knock out 118 less Franken votes than visa versa.

I managed to convince myself that this 118 is a better figure of merit than the challenge differential of 95. I am very tempted to use it to subtract from the 192 number :). What say ye, math experts?

~Latte

MNLatteLiberal said...

@mwb Michael -
perhaps he meant pauli in the sense that no two senatorial candidates can occupy the same seat at the same time? But that assumes that they are fermions, when in fact they are clearly bosons. (For different reasons, each, obviously. And about Al I say it with love).
~Latte

Michael (mbw) said...

@latte- I hate to think what Pauli would say of your attempted justification of KWR. Say that a seat is an orbital state. You're assuming spinless politicians!

Michael (mbw) said...

On the odds: Maybe this is a way out of some of the numerology.

The only systematic term in the changes to the "4" (other than from the "12") comes from differences between local judges and the state board. We think those differences are very small. therefore only a very narrow range of ballot 'countability' will be affected. Within that range the density of potential F's and C's can be well approximated as a constant. (Most of the other numbers we're talking about are dominated by irrelevant far-from-threshold challenges.) Both the challenge types and prior probabilities say that the F density is higher. Therefore if the Board tends to be more inclusive, F tends to gain, and if less inclusive C tends to gain. Either way the systematic effect is likely to be smaller than the random term, of order 10.

next post: thoughts on absentees.

Michael (mbw) said...

I think we can safely say the 133 will be counted, since the case is black-and-white and the Senate has the ultimate say.

Now to absentees.

As Nate pointed out, absentees went roughly 5% more (margin) for F than the overall vote. Furthermore, there is a near universal tendency for more D's to mess up ballots or to have election officials screw up their vote. Both of these effects would suggest that any newly-counted absentees would tend fairly strongly to F.

However, news reports say that different counties behaved differently toward voters who submitted flawed absentee envelopes. I suspect that nice DFL counties were more helpful in getting those repaired, and tight-ass R counties less so. That could mean more remaining fixable ballots from R counties.

So we have no idea who will be helped by getting into '5th pile' absentees.

MNLatteLiberal said...

disclaimer/warning: bad puns ahead. no. really bad. ok, you've been warned.

this is still in ref to bozon/fermion debate, excluding pauli, senators notwithstanding.

still here? you have been warned! LOL

masochists.

mbw, don't get excited! given the virtual state of the recount, looking for the 133 is a little like searching for Schrodinger's cat, Dirk Gently-like.

And speaking of the old man Schrodinger, the state of the Senator-Elect imho is EXACTLY like that of the cat prior to collapsing the probability function via observation. This whole election is so finely parsed and is so within the tightest, most conceivable MOE, that it, for my money, has been quantized. There is no more continuum. There are only discrete levels. Every else about this election is virtual.

We have achieved 4 in 2.9 million. That's subatomic. Uncertainty Principle starts applying. Like we might know precisely the number of missing votes (133=80+36+17), but then we cannot ascertain their location. We might know the exact mass of the candidates, but not their momentum. Etc. Etc. Ad nauseum.

After all, you have been warned.
~Latte

KWRegan said...

LOL about the quantum allusions which are all spot-on :-), but maybe this can satisfy MBW about the proof. He's right that it's called

The (How To) Spin Statistics Theorem,

which makes at least its title relevant for this blog :-). The operative line in this Wikipedia summary is, "So that exchanging the order of two appropriately polarized operator insertions into the vacuum can be done by a rotation, at the cost of a sign in the half integer case."

In lay terms, suppose (for sake of contradiction) two Demopublican particles are identical in everything except their position, for which there's a difference x. Now interchange them by a 180-degree rotation. Because they have half-integer spins, the x changes the way every complex number does when you rotate it halfway around a circle: it gets its sign negated to -x. But since the rotation leaves the system the same, x must equal -x. Hence x must equal 0, so the particles' positions coincide (like true Demopublicans:-). Thus we must have been talking about just 1 particle all along.

I can add that the 192 is exactly what you get if all vote-challenges are accepted and all no-vote challenges are denied, which is actually perverse for the candidate who has made more undervote challenges. Since by Nate's cited studies of "vulnerable" voters showing them more likely to vote Democratic (as mbw says just above), the STrib's way of counting organically favors the GOP side as much as possible. And well that side knows it! Latte, I don't quite follow your numbers (do they re-derive theUptake's metric?), but agree with everything else. My "X" was the internal Franken number 4, then add 97 to get 101. @Cubic, the one remaining question is, was it "+4" both after Thursday night and after Friday noon? A tad unlikely, maybe? I still see a discrepancy pro-Franken, though maybe it's due to what Nate said about the STrib "sometimes getting ahead of itself", since I used their figures mid-day.

WV "pressend" (no kidding!): I just wish the press end were clearer.

KWRegan said...

Ah, I see where the "118" comes from---it's 97 taken from the original pre-recount margin of 215. So yes, we're talking in equavalent terms.

And one more fix: I should clarify "the extreme of all the action being in just 1 precinct" to be "the extreme of having just 1 precinct." If you had 4000+ precincts but all challenges were in just 1 precinct, that would actually be a good case because all the other precincts would give clear info on who picked up what. Indeed the no-challenge precincts are the base case of Nate's model. Put another way, Nate is utilizing properly much more information than we are, so () it's amazing that just the simple algebra is already interesting, and () because of the way they've been reported, each challenge destroys some information.

Anyway, my prediction that Nate would re-emerge with re-regression today has only 4 hours more...and I'm off to a tree-raising party...

Michael (mbw) said...

KWR- Half-integer spin changes x -1 on 360° rotation, not 180°.

KWRegan said...

Ah yes, I forgot (in making my simplification) that it's important that each individual particle rotates by 180° of-itself when the two particles are switched around.

信次 said...

情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,美國aneros,rudeboy,英國rudeboy,英國Rocksoff,德國Fun Factory,Fun Factory,英國甜筒造型按摩座,甜筒造型按摩座,英國Rock Chic ,瑞典 Lelo ,英國Emotional Bliss,英國 E.B,荷蘭 Natural Contours,荷蘭 N C,美國 OhMiBod,美國 OMB,Naughti Nano ,音樂按摩棒,ipod按摩棒,美國 The Screaming O,美國TSO,美國TOPCO,美國Doc Johnson,美國CA Exotic,美國CEN,美國Nasstoy,美國Tonguejoy,英國Je Joue,美國Pipe Dream,美國California Exotic,美國NassToys,美國Vibropod,美國Penthouse,仿真按摩棒,矽膠按摩棒,猛男倒模,真人倒模,仿真倒模,PJUR,Zestra,適趣液,穿戴套具,日本NPG,雙頭龍,FANCARNAL,日本NIPPORI,日本GEL,日本Aqua Style,美國WET,費洛蒙,費洛蒙香水,仿真名器,av女優,打炮,做愛,性愛,口交,吹喇叭,肛交,魔女訓練大師,無線跳蛋,有線跳蛋,震動棒,震動保險套,震動套,TOY-情趣用品,情趣用品網,情趣購物網,成人用品網,情趣用品討論,成人購物網,鎖精套,鎖精環,持久環,持久套,拉珠,逼真按摩棒,名器,超名器,逼真老二,電動自慰,自慰,打手槍,仿真女郎,SM道具,SM,性感內褲,仿真按摩棒,pornograph,hunter系列,h動畫,成人動畫,成人卡通,情色動畫,情色卡通,色情動畫,色情卡通,無修正,禁斷,人妻,極悪調教,姦淫,近親相姦,顏射,盜攝,偷拍,本土自拍,素人自拍,公園露出,街道露出,野外露出,誘姦,迷姦,輪姦,凌辱,痴漢,痴女,素人娘,中出,巨乳,調教,潮吹,av,a片,成人影片,成人影音,線上影片,成人光碟,成人無碼,成人dvd,情色影音,情色影片,情色dvd,情色光碟,航空版,薄碼,色情dvd,色情影音,色情光碟,線上A片,免費A片,A片下載,成人電影,色情電影,TOKYO HOT,SKY ANGEL,一本道,SOD,S1,ALICE JAPAN,皇冠系列,老虎系列,東京熱,亞熱,武士系列,新潮館,情趣用品,約定金生,約定金生,情趣,情趣商品,約定金生,情趣網站,跳蛋, 約定金生,按摩棒,充氣娃娃,約定金生,自慰套,G點,性感內衣,約定金生,情趣內衣,約定金生,角色扮演,生日禮物,生日精品,約定金生,自慰,打手槍,約定金生,潮吹,高潮,後庭,約定金生,情色論譠,影片下載,約定金生,遊戲下載,手機鈴聲,約定金生,音樂下載, 約定金生,約定金生,開獎號碼,統一發票號碼,夜市,統一發票對獎,保險套, 約定金生,約定金生,做愛,約定金生,減肥,美容,瘦身,約定金生,當舖,軟體下載,汽車,機車, 約定金生,手機,來電答鈴, 約定金生,週年慶,美食,約定金生,徵信社,網頁設計,網站設計, 約定金生,室內設計, 約定金生,靈異照片,約定金生,同志,約定金生,聊天室,運動彩券,大樂透,約定金生,威力彩,搬家公司,除蟲,偷拍,自拍, 約定金生,無名破解,av女優, 約定金生,小說,約定金生,民宿,大樂透開獎號碼,大樂透中獎號碼,威力彩開獎號碼,約定金生,討論區,痴漢,懷孕, 約定金生,約定金生,美女交友,約定金生,交友,日本av,日本,機票, 約定金生,香水,股市, 約定金生,股市行情, 股市分析,租房子,成人影片,約定金生,免費影片,醫學美容, 約定金生,免費算命,算命,約定金生,姓名配對,姓名學,約定金生,姓名學免費,遊戲, 約定金生,好玩遊戲,好玩遊戲區,約定金生,線上遊戲,新遊戲,漫畫,約定金生,線上漫畫,動畫,成人圖片, 約定金生,桌布,桌布下載,電視節目表, 約定金生,線上電視,約定金生,線上a片,約定金生,線上掃毒,線上翻譯,購物車,約定金生,身分證製造機,身分證產生器,手機,二手車,中古車, 約定金生,約定金生,法拍屋,約定金生,歌詞,音樂,音樂網,火車,房屋,情趣用品,約定金生,情趣,情趣商品,情趣網站,跳蛋,約定金生,按摩棒,充氣娃娃,自慰套, 約定金生, G點,性感內衣,約定金生,情趣內衣,約定金生,角色扮演,生日禮物,精品,禮品,約定金生,自慰,打手槍,潮吹,高潮,約定金生,後庭,情色論譠,約定金生,影片下載,約定金生,遊戲下載,手機鈴聲,音樂下載,開獎號碼,統一發票,夜市,保險套,做愛,約定金生,減肥,美容,瘦身,當舖,約定金生,軟體下載,約定金生,汽車,機車,手機,來電答鈴,約定金生,週年慶,美食,徵信社,網頁設計,網站設計,室內設計,靈異照片, 約定金生,同志,聊天室,約定金生,運動彩券,,大樂透,約定金生,威力彩,搬家公司,除蟲,偷拍,自拍, 約定金生,無名破解, av女優,小說,民宿,約定金生,大樂透開獎號碼,大樂透中獎號碼,威力彩開獎號碼,討論區,痴漢, 約定金生,懷孕,約定金生,美女交友,約定金生,交友,日本av ,日本,機票, 約定金生,香水,股市, 約定金生,股市行情,股市分析,租房子,約定金生,成人影片,免費影片,醫學美容,免費算命,算命, 約定金生,姓名配對,姓名學, 約定金生,姓名學免費,遊戲,約定金生,好玩遊戲,約定金生,好玩遊戲區,線上遊戲,新遊戲,漫畫,線上漫畫,動畫,成人圖片,桌布,約定金生,桌布下載,電視節目表,線上電視, 約定金生,線上a片,線上a片,線上翻譯, 約定金生,購物車,身分證製造機,約定金生,身分證產生器,手機,二手車,中古車,法拍屋,歌詞,音樂,音樂網, 約定金生,借錢,房屋,街頭籃球,找工作,旅行社,約定金生,六合彩,整型論壇,整型論壇,珠海,雷射溶脂,婚紗,網頁設計,水噹噹論壇,台中隆鼻,果凍隆乳,改運整型,自體脂肪移植,新娘造型,婚禮顧問,下川島,常平,常平,珠海,澳門機票,香港機票,貸款,貸款,信用貸款,宜蘭民宿,花蓮民宿,未婚聯誼,網路購物,婚友,婚友社,未婚聯誼,交友,婚友,婚友社,單身聯誼,未婚聯誼,未婚聯誼, 婚友社,婚友,婚友社,單身聯誼,婚友,未婚聯誼,婚友社,未婚聯誼,單身聯誼,單身聯誼,白蟻,白蟻,除蟲,老鼠,減肥,減肥,在家工作,在家工作,

平平 said...

^^ nice blog!! ^@^

徵信, 徵信網, 徵信社, 徵信社, 徵信社, 徵信社, 感情挽回, 婚姻挽回, 挽回婚姻, 挽回感情, 徵信, 徵信社, 徵信, 徵信, 捉姦, 徵信公司, 通姦, 通姦罪, 抓姦, 抓猴, 捉猴, 捉姦, 監聽, 調查跟蹤, 反跟蹤, 外遇問題, 徵信, 捉姦, 女人徵信, 女子徵信, 外遇問題, 女子徵信, 徵信社, 外遇, 徵信公司, 徵信網, 外遇蒐證, 抓姦, 抓猴, 捉猴, 調查跟蹤, 反跟蹤, 感情挽回, 挽回感情, 婚姻挽回, 挽回婚姻, 外遇沖開, 抓姦, 女子徵信, 外遇蒐證, 外遇, 通姦, 通姦罪, 贍養費, 徵信, 徵信社, 抓姦, 徵信社, 徵信, 徵信公司, 徵信社, 徵信, 徵信公司, 徵信社, 徵信公司, 女人徵信, 外遇

徵信, 徵信網, 徵信社, 徵信網, 外遇, 徵信, 徵信社, 抓姦, 徵信, 女人徵信, 徵信社, 女人徵信社, 外遇, 抓姦, 徵信公司, 徵信社, 徵信社, 徵信社, 徵信社, 徵信社, 徵信社, 女人徵信社, 徵信社, 徵信, 徵信社, 徵信, 女子徵信社, 女子徵信社, 女子徵信社, 女子徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信社,

徵信, 徵信社,徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 外遇, 抓姦, 離婚, 外遇,離婚,

徵信, 外遇, 離婚, 徵信社, 徵信, 外遇, 抓姦, 徵信社, 徵信, 徵信社, 徵信, 外遇, 徵信社, 徵信, 外遇, 抓姦, 徵信社, 征信, 征信, 徵信, 徵信社, 徵信, 徵信社, 征信, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社, 徵信社, 徵信社, 徵信, 外遇, 抓姦, 徵信, 徵信社, 徵信, 徵信社, 徵信,

平平 said...

^^ very nice

徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社,徵信, 徵信社, 徵信, 徵信社, 徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,

徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,徵信, 徵信社,