There's been an update today in the Alaska Senate race, with 11,612 more votes counted. Mark Begich now leads incumbent Ted Stevens by 1,061 votes.
Later update: 1,022. An additional 3,333 votes have been counted and Stevens pulled back 39 votes.
Per Sean Cockerham at the Anchorage Daily News, the ballots counted today:About 510 questioned ballots from Southeast, the Peninsula and Southwest Alaska
About 5,180 absentee and questioned ballots from Mat-Su
Questioned, absentee ballots from Richardson Highway and the Interior
About 3,600 absentee and questioned ballots from Western and Northwest Alaska, and North Slope
Monday the extra Richardson Highway ballots will be counted, and Tuesday another 24,000 or so ballots will be counted from Anchorage, Southeast, Kenai Peninsula and Southwest Alaska.
11.14.2008
Alaska Update: Begich Leads By 1,061, Then 1,022
@ 7:14 PM
Subscribe to:
Post Comments (Atom)
169 comments
does the 95% chance for Begich still hold true?
Given what's left to count, if Begich doesn't win people should be EXTREMELY suspicious.
Nate needs to hone his interviewing skills to become a rich media celeb. He's got the stats bonafides and is decent on the tube. If he can master the medium, he could be a multi-talented guru. Somebody needs to back him with funding for training on that. Take 10% as his agent. It's pretty obvious everyone respects him for his analysis and passion for the game. All he needs is a little better presentation and he's a rich superstar. OF course, he's a million times further along than any of us or most of us anyway. Props Nate. Hire an agent. And don't call Minnesota Alaska next time and I'll try to quit the typos.
Drop the glasses.
Aren't the areas that were counted today quite Stevens-friendly? I thought that Mat-Su and the North Slope were GOP stronghold areas. If this is the case and Begich still increased his lead, then Stevens doesn't really stand a chance, right?
Now where are the guys who were bashing Nate for having Begich at 100% chance of winning?
Whoa! I knew Nate would be successful in finding a book publisher but . . . Penguin giving him $700,000! That's a pretty big advance for a book about polling!
Whew!
Nate is just fine on the tube. It's refreshing to see a real person and not the overly coached media whores we usually see. Keep up the great work Nate!
Begich is almost certainly going to win. The only question is if his margin of victory will be greater than 0.5% of the total vote*, which is the threshold for a state-sponsored recount in Alaska. Begich gained about one vote for every 12.5 counted on wednesday, but only one out of every 47 counted today. The final tally will depend on how many provisional ballots get counted.
*He needs to take less than 500 extra votes to meet that criteria.
I don't understand why they can't just count the ballots now. They act like it's a stressful job or something, and have to break it up over 3 weeks? wtf?
Good news- this means a certain Begich win, outside recount range.
Bad news- Checking the MN SOS site for post-election ballot audit, it's clear that there is zero chance that Franken will make significant gains in a straight recount. It seems all our analysis of unintentional undervotes was nonsense, the rate is essentially zero.Any win chance rests on courts allowing some excluded absentee votes, etc, with small signature discrepancies etc. / mbw
It looks as though it is pretty secured now.
Come see my new probably awful blog:
http://youngerpolitics.blogspot.com/
i agree. i never hear of a state dragging out the counting process so much. i mean taking every other day off. wierd.
Link, michael?
Well, Sean, are you about to score a huge $700,000 payday for your reminiscences about the campaign trail?
I think your book would be far more interesting to read than one about the numbers!
According to an earlier report today by KTUU, Anchorage's NBC affiliate, today's count was supposed to be from Mat-Su, Juneau, Fairbanks, & Nome. These results indicated that 11,285 new ballots were counted, but supposedly there should be another 5,370 that they have yet to count today, so we may get a later report today.
According to KTUU's most recent story:
Nearly 16,000 absentee and questioned ballots from Anchorage need to be counted. There will also be another 8,300 ballots in the Juneau, Kenai and Kodiak district to be counted.
Those results are supposed to be reported no later than 10 AM Alaska time on Tuesday. That would be 2:00 PM Eastern time.
Michael
There has not been a recout yet in MN. I have no idea, therefore, what you are talking about. Does anyone else understand what means?
terrondt wrote: i agree. i never hear of a state dragging out the counting process so much. i mean taking every other day off. wierd.
I'm just guessing, but I think it might have to do with different Alaska Division of Elections offices around the state being at different stages of the process in checking ballots against registration rolls & whatnot.
For Tuesday, the Anchorage & Juneau votes should trend Begich, the Kenai & Kodiak votes will likely trend Stevens.
Link for MN vote counts:
http://electionresults.sos.state.mn.us/20081104/ElecRslts.asp?M=S&R=all&P=A&Races=%27%27
I make the difference between the presidential and senate races about 24,600 votes or 0.85% of the presidential tally.
It's strange how a state with one of the lowest populations in the country, (and the second lowest number of ballots cast after Wyoming), takes the longest to count its votes.
A bit off topic, but of interest, I hope. Rasmussen is excited about its (premature) Fordham #1 ranking (the Fordham study ranked national tracking polls with an assumed 6.15 point spread result, but the final difference will be closer to 6.8 or so). Even so, their national tracking seems good. However, they missed the spread in each state by an average of 4.5 points (c. 28% off of the actual margin on average).
For McCain, they missed high in 19 states (off by an average of 2.4 points), were on the money in 10, and missed low in 21 states (by an average of 3.0 points), for an overall average of 2.2 points off on McCain.
For Obama, Rasmussen was low in 32 states by an average of 3.4 points, on the money in 10 (though not all the same as for McCain), and high in 8 states by an average of 2.6 points, for an overall average of 2.6 points off.
Some observations: They did better with McCain that with Obama by about 17%. They underestimated Obama 4 times as often as overestimating him, while coming out low and high for McCain about equally often.
And finally, Rasmussen's "accurate" national result masks fairly big misses that canceled each other out. Wyoming's margin was 13 points bigger for McCain than predicted, Idaho's was 14, and Arkansas's was 10, but on the other side Obama's margin was 13 points over prediction in Vermont, 10 in Delaware, 9 in Rhode Island, and 8 in Connecticut and Nevada.
Dave Yandell
@reelgeist
My understanding is this: They hand-verified a sample of the machine-counted ballots. This is apparently standard practice and is used to validate the performance of the machines.
I can't recall where I read this, unfortunately. So, treat it with skepticism.
Why would that necessarily prove what Michael thinks it proves? I thought undercount, etc were not a simply matter of machine malfunction? That malfunction was only one option.
Get a look at the vote count for Alaska's House District 7:
Karl Kassel (D): 4999
Mike Kelly (R): 5000
I'm guessing that Kassel will be asking for a recount if it stays this close.
This is an incumbent GOP seat, so if the Dems win this one, they could narrow the AK House to a 21-19 Republican majority. Fingers crossed!
Eric said...
Nate needs to hone his interviewing skills to become a rich media celeb. He's got the stats bonafides and is decent on the tube. If he can master the medium, he could be a multi-talented guru. Somebody needs to back him with funding for training on that. Take 10% as his agent. It's pretty obvious everyone respects him for his analysis and passion for the game. All he needs is a little better presentation and he's a rich superstar. OF course, he's a million times further along than any of us or most of us anyway. Props Nate. Hire an agent. And don't call Minnesota Alaska next time and I'll try to quit the typos.
I completely agree...go for it Nate!
Its now down to 1022...2 updates in 1 day is quite amazing...
Reelgist: Michael doesn't know what he's talking about. The MN Sec of state has indicated that their machine recount is 99.99% accurate (in 2006). But this says NOTHING about the undervote, about which NOTHING seems to have been said on the Secretary of State's web-site unless I missed it.
In fact, we know the numbers haven't be totally updated there because, according to the news reports, during a short test of the machine count of around 5,000 votes, Franken picked up one vote, so the total should be 205, not 206 as reported (unless they're not counting it for some reason I can't figure out).
All in all, they are saying they aren't likely to have a manual recount of all the ballots because the machines were better than 99.50% percent accurate. That says NOTHING about the undervote of course.
The machines spit those ballots out and can't count them at all. They have to be ALL reviewed by hand to determine whether the person INTENDED to leave the Senate race blank, or if they did something wrong like circle one of the names instead of filling in the oval.
That sort of mistake will be most of the INTENTIONAL "undervotes" as Nate has explained in exhaustive detail that I don't have to repeat here.
Bottom line: Franken is very likely to win when they count all the undervotes unless for some reason it's determined that everybody INTENTIONALLY left the Senate race blank.
There will be thousands of unintentional undervotes, mostly from counties which Franken won handily. Whether they will all be counted and if so whether they will be enough to overcome a 205 vote Coleman margin remains to be seen, but it's where the smart money is betting right now.
The bummer results from MN are available form here: http://www.sos.state.mn.us/home/index.asp?page=1156
/mbw
thanks cugel-
for MN info
At 03:56:18 PM Alaska time official results show Begich's lead at 1022 votes (138959 to 137937).
see results here
Cugel
Thanks. This is not my area. But what you say makes more sense on a practical level.
Nate should totally become a rich media celeb. He tends to underwhelm his readers who are expecting an messiah to appear. But if his image improves, he would so totally and completely become a total celeb. Obama type celeb, not britney celeb of course
We'll soon see celebrity ads being released about Nate from pollsters such as Zogby and gang.
@Andy JS
Actually, it's not that strange that Alaska hasn't finished counting yet. As of right now, California hasn't either (http://www.sos.ca.gov/elections/c-status08/total_unprocessed_ballots08.pdf). It's just that the number of uncounted ballots in California aren't enough to make a difference in any major races.
Nate should NOT polish his media image, NOT lose the glasses, and NOT lose sight of what got him here.
TV is already full of talking haircuts. Stick with: fidgety, unassuming, nerdy number cruncher. It's working.
First time / long time. Love 538! Thanks Guys. Really.
looking at that MN site, it includes things about 1-vote corrections due to a Franken ballot found in a tray, x mark in oval, lightly filled oval...
That sure sounds like in these precincts, which have a lot more than 5000 votes, they were doing a full hand recount of everything. There were a significant number of new 'no votes' recorded, but only tiny shifts in F and C. Unless there's some entirely different vote pile lying around, we're screwed.
I could be wrong about this, but those of you asserting it most emphatically might try reading the .pdf first./mbw
Kathy A:
I'm not surprised that California is still counting, since it has such a massive population. Alaska should have been able to count around 95% of the relatively small number of votes it has by 11 days after the election, although I know that the weather in Alaska may be a factor.
@michael,
That link you sent does not discuss the recount situation, but rather a standard post-election review process. The recount is an entirely different process and will in fact pick up undervotes.
Be careful posting things like that... You needlessly scared people!
Also...
Franken needs to pick up LESS than 1 vote per 10,000 votes recounted to take the lead!
There's also the fact that there are no roads in huge swaths of rural Alaska -- what we call the Bush (no relation to the current lame duck) -- and mail service even in good weather isn't necessarily daily. Mail service is by small planes which are affected by weather conditions, & yes, it's winter here.
There was also the issue of checking absentees against the precinct rolls from the Election Day vote, in order to prevent double-voting as happened in the primaries. I did in fact read one report indicating that some bozos attempted to double-vote in the general election (first vote absentee, then vote at the precinct on Election Day).
Michael, et al.
I think the point is that these are corrections pertaining to the set of ballots that were already "successfully" machine-counted to see how accurate the scanning machines were for the ballots that they seemed to have counted. The postelection review (see the statute pdf at the cite you linked to) is about checking machine counts (via hand-counting), not about hand-counting ballots the machines didn't "report" counting.
BTW, does anyone know of a handy website for nationwide state-by-state presidential results that includes all 3rd party candidates?
Thanks!
Sorry guys, here's what the MN SOS site says about this post-election audit
"The postelection review must consist of a manual count of the ballots used in the precincts
selected and must be performed in the manner provided by section 204C.21. The postelection review must be conducted in the manner provided for recounts under section
204C.361 to the extent practicable."
Bottom line- we've got to fight GA as hard as we can. Don't leave another one the table. My $2300 is sent, and we'll be scrounging volunteers via MoveOn soon./ mbw
mel,
Every other state in the union has absentee or early votes, and NONE of them need to check lists and signatures AFTER the election. That is actually the part of this process that makes AK look so bad.
AK's solution to this double voting problem is, frankly, poorly thought out and a ridiculous waste of time!
Good news for the Dems. Can't believe a convicted felon has put up this kind of fight.
matt w -- that may well be true, but all the same -- that's what happened.
People have been asking why is the vote taking so long -- I'm simply trying to answer them based on reading my hometown newspaper every day. As it is, I still think there might be something fishy with the vote tabulation from Election Day itself (re: Berkowitz' numbers, & Don Young's smug non-worry about the election).
Does anyone else want to waste time telling Michael he's wrong about what he's reading?
Michael actually has a better point than I originally thought he had. However, I completely disagree with his conclusion.
Michael remember that Franken needs to pick up ONLY 1 vote for every 14,000 or so votes recounted! He picked up a lot more than that (relative to Coleman) in the sample I looked through in that .PDF
I am in fact very encoraged by this data!
My favorite:
"originally counted as overvote, voter intent was for Franken by voter statement"
Just think about what that ballot actually looks like!
Why is Alaska rushing this vote count? The new Congress doesn't get sworn in till January. They've got almost two months!
I guess they are all efficiency experts there and have nothing better to do between moose hunts and snow machine races.
wv - yantylap
I would appreciate someone telling me why Michael's point isn't valid. Most of the discussion has to do with undervoting, which would have been caught by this audit. The very low rate of errors is not promising. I haven't run through the whole pdf to see what the statistics show, but it's clear that the undervote where voter intent could be determined is quite low.
because voting is about voter intent. there will be some the machine will not read, but that may have been marked showing intent. Unless you are saying that the machines were changed to somehow read these markings, I am not sure I am following this discussion at all.
ecarlson,
This is a good set of data to test quantity of undervotes and reclassifiable overvotes.
Again, I think Michael came across something valuable here, but let's interpret it carefully.
Franken needs to pick up about 1 vote more than Coleman per 14,000 ballots reviewed.
As I go through this data he seems to be doing much better than that. Remember...there are few reclassifiable ballots, but these are really small samples. There are actually a fair amount if you look at it in a per 14,000 ballots sort of way.
Thanks, Michael this is a good find! - But I still think you are wrong about your conclusion.
From watching these results on the Alaska Elections board site, they update until about 20:00 local time, so they may not be done yet.
in other words, is the point that the machines will be accurate as to votes cast properly so that the machine can read them or is the point that there is a percentage of voters who may have intended a vote that the machine was not able to read? Again I maybe mistaken, but I find the analysis that we can figure this out by having the machine check seems well not to make sense
matt- I dont underswtand how this address votes were not readable by the machine?
They did a hand-count in MN, same as in the recount. A quick glance at a few dozen precincts suggests that about 150,000 votes were recounted (very rough). We picked up a margin of 1, and that's not an accidental cancellation. The changes are tiny. I'm not thinking that any of you are having luck talking me down.
/mbw
Nate Silver's 700k book deal!
Hey NATE TERRIFIC NEWS!!! I think you are str8 but in case you are interested I am hot male model available living in Europe! ( Str8/bi but for part of 700k I can adjust!!!)
michael
but huh?
did they do a hand count of 150,000 votes? also, how does that address the undercount issue of the nearly 25k where we know that there was no vote for senator, but there was one for prez? are you saying that they have a rule that says that if the sample isn't below a certain numbers, they will ignore the 25,000 known undervotes in a situation with a spread of 205 difference?
jesus everyone is a ho now a days. zip your pants back up pink.
No way has 150,000 Minnesota votes already been hand-recounted. There was only 1200 votes hand-counted in Hennepin County alone, which is about 20% of the state's population. The number of votes that have benn hand recounted must be less than 10,000 at this point.
OK, I took a somewhat random (not very random since results are grouped by precinct) sample of 14,111 ballots and Franken gained exactly 1 vote relative to Coleman. There were 4 votes reclassified for Franken and 3 votes reclassified for Coleman (which is not that many!). There are Machine errors, voter errors and perhaps the most important type of error are Poll worker errors (jammed ballots fed through twice).
I encourage others to take a sample of about 14,000 votes as determined by ERS votes column, Office Totals line. And track changes in Franken and Coleman votes.
http://www.sos.state.mn.us/docs/post_election_audit_review.pdf
I think that if the whole state mimicks this sample and the process used is in fact the same then It doesn't look great, but there is of course a chance! 206 is not that many!
Yikes. Sorry, I was wrong. I only counted one precinct in Hennepin County. The total votes that have been hand-recounted there is 18,934, which means there has been about 94,600 votes hand-recounted in the state. That's still not 150,000, but it's a lot higher than I earlier said.
1,022 vote Begich lead is GREAT NEWS!!!! FOR CONVICTED FELON TED STEVENS!!!!!!!!!!
for the simple minded here-
they have already hand counted 95,000
votes in MN? and franken picked up one?
well i guess we write off franken-thanks MN...
As I continue to go through this data I do not think this is going to exactly mimick the recount results. There seem to be inconsistencies like in Hennepin Precinct 1685 (page 197). It is the only precinct that disquallified ballots that were "questionable" but that the machine counted!
In MN, I'll take Nate's last forecast: " Given the total number of ballots cast in Minnesota's senate race, this translates to 4,835 ballots that will in fact be reclassified during the hand recount."
If Nate is right, and he always is, Franken, Coleman, and Barkley will be picking up another 4,800 votes between them. Given that nearly half of the undervotes are from the Twin Cities, I like our chances a lot.
I think that the recount will have much more uniform standards and that there will not be any pressure to supress vote changes. I imagine in this review most precincts wanted to "pass" the review and get that "acceptable" stamp.
Also back to Precint 1685... how is a change of 3 out of 1350 equal to 0.00% ???
My chief problem with this conversation matt is that I does not sound like a hand recout. Not the least of which because I know the rule is that during a hand recount both parties are to have representatives there to address issues of voter intent, etc. I am not sure what you or the other people are talking about. But I do know that your argument does not pass the "what the law says" test. I don't pretend to know the numbers. But I do remember the rules.
again- are you sure you are discussing hand recouts in the critical areas matt? since the under count occured in places like heneprin or whatever it is called?
reelgeist,
Yes! The column is actually labeled "hand-counted votes" and clearly they are hand counting because of the occasional comments they list. There are a few "undervotes" that are picked up, but that is also an area with a lot of inconsistency between precincts. I am not convinced that this review will really resemble the full recount.
From my analysis, Franken was about +8 on 1% of the vote. This is very encouraging.
reelgeist,
Go look at the data yourself...
http://www.sos.state.mn.us/docs/post_election_audit_review.pdf
Travis,
+8 against Coleman? on 1% of what? That doesn't seem like the same numbers I am looking at.
From my understanding of Minnesota:
1. All the ballots that so far have been hand counted are the ones that the machines kicked out for some reason, but the majority of the ones kicked out were NOT for undercount (the machines mostly did NOT kick out undercounts). Undercount is where most of Franken's votes will be picked up, if he picks up a significant number of votes in a recount.
2. The state will finish the initial count and certify the election by sometime next week.
3. If the initial count shows a difference of less than .5% between the top two candidates, an automatic, state-paid recount of ALL ballots will be made, and that recount will be by hand, NOT with machines. That means all approx 3,000,000 million ballot cast on November 4 will be hand counted.
During the hand count recount, each and every ballot will be observed by several people to determine the intent of the voter. Was there an intentional overvote? Was there an unintentional overvote? An intentional undervote? An unintentional undervote, where it can be determined which candidate the voter intended?
And remember, Minnesota has a very liberal law on intent of the voter - it must be very UNCLEAR as to intent of the voter before the ballot is thrown out and not counted.
As an example: Voter fills in circle for one candidate, then marks through that candidate's name and fills in the bubble for another candidate - interpretation is that the voter intended to vote for the candidate NOT crossed off.
Example: Voter circles a candidate's name, or places a check mark in the bubble, or marks through all but one candidate's name - interpretation is that the voter intended to vote for the 'uniquely marked' candidate.
For Matt W and everyone, I have posted up a plain-text extract of just the precinct ID, Coleman, and Franken lines from the MN audit review.
Incidentally, to make live links, type a '<' and an 'a' together, then a space, then href=" and mouse-copy the URL, then immediately type " and '>' Then enter the words for the link, then '<' then /a then '>'
WV "naterst" (no kidding!!): This is the Naterst thing I've done to help research on this site.
Congratulations on the book deals Nate, you deserve the attention, you did a wonderful job over the course of this election cycle.
let me try this once more- none of this helps us understand what if anything the difference between the numbers for president and the numbers for senate. you aren't going to be able to do that right now. the only way you could if to have went as i said to heneprin or whatever it is spelled and went through the ballots in a hand recout to determine voter intent.unless you are arguing you have a way to know this that no one including darmouth was able to figure it out- i think what you are more likely doing is assuming that what you found is representative of where the 25,000 will be found. this is where we keep going in circles. I haven't seen you write anything addressing the 25,000 votes portion.
Mike in Maryland et al.
That is not what we are discussing. MN does a "post-election review" of every ballot in certain precincts, not just "kicked out" ballots. This review is a hand count that is supposed to resemble the rules of the recount. Therefore it serves as a sample of what the recount might end up looking like.
Ironically there were very few or no "voter circled name" comments (oh yes, for the most part there are comments when a vote is changed/added/removed), but there are things like -ballot counted as overvote, but voter intent clear because of a note by voter-.
Some, but not most precincts adjusted the totals number to reflect true undervotes.
Any chance Palin "listens to God" on this one and gets the message she shouldn't be in national politics? Poor Jon Stewart.
matt
I note with interest you are now ingoring the point that this tells us nothing about where the 25,000 under count are. are you saying that they are located in the places you are already discussing? saying that they found few tells me nothing if they aren't the precints and areas in which we expect to find the higher number of such ballots.
reelgeist,
the 25,000 undervotes are either true undervotes, or will be picked up by a hand recount. Since we are discussing a hand recount that picked up very few undervotes the only conclusion I can draw is that the majority of them are actually true undervotes (meaning the voter did not vote). Of course this could be a bad sample and there are a lot of undervotes clustered in precincts that were not reviewed in this review. I am not sure what I could say to satisfy you
While this is not the "official" hand recount I do believe it does reflect what you will see during the "official" recount. Adjustments were made in cases where where the voter intent was clear (see comments)
I do not have the time or the inclination to total up the change in votes in this sample but remember this is a sample of about 200 precincts, there are 4130 precincts in Minnesota. Also not all precincts are created equal. In rural counties there may only be a few dozen voters, in Metro precincts of 3000-4000 thousand. Since each county has to have at least 2 precincts audited the sample does skew twords rural Minnesota.
FWIW I would be suprised if Franken won but I think it is possible.
kwregan,
First, Thanks... If you are able to do so quickly can you give us a final Franken Coleman count for all these precincts?
Second, Since most people read these comments in the little pop-up window I don't make links live since just clicking on a live link in the pop-up window makes problems.
A 'sample hand count' is NOT the same as a total recount by hand.
5,000 or 15,000 or 150,000 ballots counted is NOT the same as recounting the approximately 3,000,000 ballots cast.
In addition, if you were an election worker, and knew that a recount was upcoming, would you REALLY be as diligent during the 'sample hand count' as during the recount? Even if you consciously tried to be extremely fair, your subconscious would be saying "There's a recount coming up, you can slack off just a bit" and the 'sample hand count' could be off by 0.05% (or 1500 votes over the full 3,000,000 to be recounted).
Were there observers from each party watching the sample hand count, able to challenge whether a ballot should be counted or not? From what I've read, the answer is no. If 0.05% of ballots were not counted in the 'sample hand count' were adjudged to be countable in the full hand recount, that would be approximately 1500 votes over the full 3,000,000 to be recounted.
In other words, a very tiny variance in procedures, and adjudication of 'unclear' in a slightly different manner could make a very (relatively) large difference when the full hand recount is made starting next week.
Mike in MD,
I agree procedures will be somewhat different even though the review claims to use the same procedures as a recount. Even a tiny change would benefit Franken, however, this data is pertenent especially in so far as we can see certain problems that we had not been anticipating that could really hurt Franken! One of the biggest errors was the "stuck ballot fed through twice" problem. In the hand count this meant that votes were subtracted from candidates. I am guessing that this would happen most in urban precincts and disproportionately hurt Franken's vote total.
using kwregan's link I came up with Franken gaining 20 votes, Coleman 17 for a net of 3 votes. That is just a quick once through so I do not guarantee the numbers.
Matt W said... "My favorite:
"originally counted as overvote, voter intent was for Franken by voter statement"
Just think about what that ballot actually looks like!"
It looks like a quite typical case, actually: Franken bubble filled in, and also a write-in, "Franken". People who want to make really really sure that it's clear who they want often do this, not understanding it will actually make it UNclear to the machine. There were thousands of such Gore votes in Florida 2000 (quite sufficient to put him over the top).
PINKTUBE.TV: Back off! He's mine! (I claim him on grounds that I was the only one here to forecast Omaha for Obama, and did so weeks in advance.)
And just like that Nate breaks into the big time. See the comments - just like here, big fans of Mr. Silver.
Good Luck and Please, Please try to make time to help crunch the numbers for the folks fighting for marriage equality - we'd appreciate it so much.
Bliss,
Here's the link to the story - sorry folks.
Third time's a charm:
http://www.observer.com/2008/media/nate-silver-signs-penguin-two-book-deal-worth-sum-high-six-figures#comment-1153041
Thanks for the link Mr. Fielding
Because of the book contract, Barack Obama will raise Nate Silver's taxes! John McCain will soon say that "Nate the Blogger" is an American hero.
Now for some musings---without my having time to decide between the Star Tribune's estimate of a recount net of only 21 to Franken and the estimate of 234 by "tmess2" here in the earlier thread, though it seems to reflect Matt W's "bummer" conclusion.
(1) The main question to piggyback on Nate's previous analyses is, How much of the recount has effectively already happened? We do know that (273 - 27) + (124 - 24) = 346 of the original 725 gap closed owing to missing-digit typo fixes, leaving 379 from smaller-scale causes. (Were there any other "typo" cases?) With the current official gap at 206, perhaps 173 votes of de-facto (albeit not de-jure)recount have happened? Since comparisons to Wash State and other similar-sized races pegged the Dem. swing in the 200s, this may accord with the other evidence here that Franken may fall 100-odd short...
(2)...absent getting more absentees, of course. The second question to ask in the same line as (1) is, Did the Wash. State Gov. and other comparable races have similar equipment and a similar pre-recount review and adjustment policy?
(3) Third question: Are cases where "exceptions" resulting in valid votes were NOT made subject to further (judicial) review?
As my text file says, some of the "Reason"s got truncated if they spilled onto a second line---I extracted just lines with "Precinct", "Coleman", and "Franken". One case worth noting is that
1 ballot found under ballot counter
doesn't mean a human counter who wore "Undervote underwear" as suggested by an early commenter---the reason in 2 cases continues, "above ballot box".
About live links in the blue boxes, try right-click to open a new tab? Then you won't get a "navigate away?" message.
I trust that the deal will be signed before the end of the year and that the advance will be considered income in 2008. No Obama tax for Nate...yet
kwregan said...
"About live links in the blue boxes, try right-click to open a new tab? Then you won't get a "navigate away?" message"
I know that... I don't follow my own links anyways... Furthermore, I don't follow comments in the pop-up (most of the time). I open the comments in a new tab so that I get the full page view. Much better! except when you need to float the comment box above another web page.
I imagine that the recanvassing corrections would not have been part of the WA recount adjustments. I think that recanvassing is a standard procedure and it would seem very odd for those mistakes to have been included in WA's certified results (certification happens before recount at least in MN).
I think though that the analogies are somewhat off anyways. Each recount is unique it terms of specific numbers and only classes of errors and party and demographic tendencies are illuminating. I think we are in rough shape if the whole recount goes like this sample review went, but I think there are a lot of potential reasons it may not.
Ah, I completely forgot the 4th and most impt musing:
(4) What may be even more important than the net change to Franken is the # of correctable errors total, which affects the resulting estimate of the correctable error rate. I.e. the data may help us pry apart the "two dimensions" in Nate's posts. And
(5) Per Mike in Maryland's comment here above, this sample itself has aspects of being a "poll of 200" [precincts] and a "poll of 14,000" [votes]. The latter kind of poll gives much more reliable estimates, of course. Can we estimate a confidence in error bars on the estimate of both the correctable-rate and the Dem lean? The Dem-lean part seems to reflect the smaller poll population more than the other.
Two other matters that came up in other threads: One commenter thought Franken was home-free on election night (am) with just Hennepin & St. Louis precincts to come, but the answer I saw (can't locate link) is that several late-reporting Hennepin precincts came in heavily for Coleman. Hennepin Co. overlaps GOP suburbs as well as the city (my in-laws live there). Second, this quote from RedBlueRichPoor's analysis bears on the adjustment Nate's model made down from 6.9 to 6.3(6.4?) on the PV margin owing to previous cases where strong leaders underperformed their final polls:
"I see some systematic patterns: Obama underperformed where the polls had him way down, and he outperformed where the polls had him up. We should go back and look at these patterns from earlier elections and see if this is consistent. If so, it suggests a way to improve forecasts for next time."
Thus Obama did quite the opposite---!---and this may highlight the effect of his GOTV operation!
On MN: The point is not that the F-C difference is 1 or 3. That's not statistically stable. It's that the total new vote is under 40, which extrapolates to well under 1000 total for the state. We had been guessing more like 5000. At 1000, the fractional difference would have to be much bigger than any of us guessed to give 200 extra F-C. /mbw
Again: The lesson is to fight like hell for GA.
Minnesota Mike said: "this is a sample of about 200 precincts, there are 4130 precincts in Minnesota"
Mike hits the nail on the head here.
The result on Michael's pdf is only the SAMPLE hand-recount normally used in the election review in each county. A tiny fraction of the votes are counted to test the system. Normally, that's enough because it's not close. The machines prove accurate to within .05%, the result is certified and away we go.
Only 200 of 4130 precincts = about 4.8% of all Minnesota precincts. That's all they've done in the sample, and you can see why! As long as the vote isn't close, and as long as the sample error rate is within the legal parameters (.05%) then they certify the result, and the election is accurate.
You would not expect a very large error rate unless the machines weren't very good (and it seems they are).
Franken still has to pick up about 1 vote for every 14,000 or so. If he does, he will win, if not, then he loses.
Clearly going by this sample, it's going to be very close as to whether he can do it -- just as Nate said a day or two ago.
If some of the precincts have higher error rates and IF those precincts are urban precincts that Franken won, he'll pick up enough votes.
But, if NOT, or if the error rate in all the precincts closely mirrors the sample then he might fall short.
Either way, we're not going to know for an entire month until they do the manual recount of the entire state!
I am begining to get annoyed mostly because you ignore the fact you can not extrapolate from the numbers you are using to figure out where the 25,000 votes are going to be. I am not good at math, but even I know that. You are NOT addressing distribution of where the undervote is at all. Just random guesses based on a sample.
Re: question about recount
I think you guys are missing the main point that Nate was trying to make, which is that the undervotes are *NOT* distributed randomly but are clustered in certain counties with a high number of low-income/old/minority/etc. voters. In such a situation, a full recount in a small number of precincts is unlikely to be statistically accurate. Now if they recounted 150,000 votes randomly distributed across all precincts, that would be different, but in this case they obviously aren't doing that. Furthermore, Nate already posted figures from previous elections concerning accuracy figures with optical-scanning machines and it's hard for me to accept that the overall figures for MN will really be that different.
Agreed, the two variables that Nate outlined. A. # of reclassifiable votes and B. Franken % of F+C reclassified votes are the two important variables for the recount. I think this is a bad sample to really address variable B, but it may be a good sample to address variable A, as Michael points out.
I actually think it might not be. I think the recount might get more reclassifiable votes than this review. There are clear inconsistencies with how this review was done in different precincts and that suggests that there are methodological problems with it.
As far as putting everything in Martin in GA, yeah that is great, but I really love Al Franken and would love to have him in the Senate.
Unlike one poster who got three, I get a net-net gain to Franken of 11 votes (from the text extract someone helpfully posted), although that's not a very careful count. (Franken's new votes minus Coleman's new votes, minus his newly lost votes, plus Coleman's newly lost votes = 11). It's simplistic but interesting to multiply that by Minnesota Mike's ratio of 4130/200. If you do, you get just enough of a projected statewide Franken gain to put him over the top, by under 2 dozen votes.
reelgist, you're the one p__ ing me off. The posted .pdf does show where the "21,000 undervotes" mainly go. Most of that is clearly ballots with the Senate race intentionally left blank. See the .pdf. E.g, page 5.
WV "stpulfac" -- St. Paul facts??
Reelgist: Your point is well taken.
The sample review is NOT designed to provide a kind of "poll" to see what the actual results would be from a full hand re-count, it's designed to certify the election machinery --- NOT the votes.
Normally none of this crap is necessary because the election just isn't that close. As long as everybody can see that the machines are accurately tabulating the results within the acceptable error rate in a number of precincts, and the result is well outside that error rate, then it doesn't matter!
They can choose the precincts at random and don't have to worry about whether the sampling of precincts mirrors the population distribution of voters.
If you lose by 5% with 2.4 million votes cast, that translates out to 120,000 votes! Normally nobody cares because a recount of a few undervotes here or there just won't matter.
But, here we're worrying about an error rate of 1/14,000 or .00714%
"I don't understand why they can't just count the ballots now. They act like it's a stressful job or something, and have to break it up over 3 weeks? wtf?"
ALASKANS, my friend. They elected Palin.
What with the new taxes, Nate will lose *ALL* motivation to write the book, just like Joe Plumber decided to be an unlicensed tax cheat rather than a plumber the very minute he found out his $250K was only going to be $239K.
> Any chance Palin "listens to God" on this one and gets the message she shouldn't be in national politics? Poor Jon Stewart.
God already told her that he was preparing the way for her, but only as an elaborate practical joke -- a role at which she shined.
I see now that my extrapolation was done earlier (with close but not identical results) by tmes2 on the earlier thread at
http://www.fivethirtyeight.com/2008/11/dartmouth-study-minnesota-undervotes.html#comment-3567648318180051288. Looks like he got the same net-net of 11 (16-5 is how he put it.) I get 11*4130/200=227. Then 227-205=22.
Of course, the geographically distributed selection here isn't randomly representative, though it's still large enough to be useful. IF we can assume that this selection over-represents rural precincts, then it suggests Franken has better than even chances in a full recount.
I should have read the Star article, that clears things up nicely:
Most Hennepin County results are not yet in."
"If the same rate of change were experienced statewide, the DFLer would gain 21 votes in all. The largest changes came in precincts in St. Louis and Ramsey counties."
If Hennepin county results are not in, then the sample is pretty much worthless to extrapolate to the entire state.
The error rate in Hennepin county will be key -- that will tell us whether Franken can pick up enough votes because for every additional ballot counted in Hennepin, Franken will have close to a 60% chance of adding one vote.
Philosophical note: There are two main types of errors in thinking about these sorts of problems.
1. Forgetting prior probabilities. we saw a lot of that over the campaign as people freaked out over individual polls, as if there weren't many others.
2. Only considering priors, as if no new data had arrived. I'm afraid that's what we're seeing here, even among the very few (Matt W) who are actually following what this discussion is about rather than just spinning random bullshit.
Two cheerful notes.
1. Juneau is next week, so AK is in the bag. I'm guessing >3000 margin.
2. MN has at least shown that a very good election system exists, to our surprise. At least it's good in the hands of Minnesotans.
Still, I'm very bummed. Maybe the lawyers can scrounge some ballots that were improperly disqualified, but don't bet on it.
And send $ to GA..
/mbw
I'll post a spreadsheet soon and you can check my data, but here's what I got:
Coleman: 54988 + 28 - 13
Franken: 51959 + 30 - 16
First number: machine counts
Second number: Total adds
Third number: Total subtracts
Error rate (by eyeball) 1/12,000
If this is representative of the final recount, then it's pretty hopeless.
Coleman picked up 1 vote for every 3600 he already had
Franken picked up 1 vote for every 3700 he already had
I'm too tired to check my numbers, but will do so tomorrow.
Intrade has Franken at a 52% win percentage and Coleman at 48%. That's a slight gain for Franken who started the day at 48%. So, whatever the sample recount was or whatever it showed, it did not move the odds much according to those who have money on the line, and, if anything, was positive news for Franken.
"What with the new taxes, Nate will lose *ALL* motivation to write the book, just like Joe Plumber decided to be an unlicensed tax cheat rather than a plumber the very minute he found out his $250K was only going to be $239K."
Joe the Plumber actually only made about $45,000 or so, not $250,000. He just believes all the right-wing lies that Obama will increase taxes on people UNDER $50,000!
He doesn't make $250,000 and he's not actually going to be able to purchase a purchase a business, and he's not actually a licensed plumber and he doesn't pay taxes anyway. But aside from that, he's a real American icon of the small businessman!
Quite a guy, Joe!
This is way way off topic but if anyone wants to read a good article on the financial crisis you can HERE.
It's a long read but well worth it if you're interested in seeing what some of these guys were doing.
Also on Intrade, Begich is now up to 97%. Before today's votes came in, he was at 93%.
Spreadsheet posted at:
http://www.wfu.edu/~ecarlson/colemanvsfranken.xls
If you spot errors, email them to ecarlson@wfu.edu
Previous posting had wrong error rate.
Current best guess: error rate 1/1200.
No evidence that "corrections" favor one party over the other.
If you look at the data, it is clear that a large fraction of the corrections come from a couple of small precincts. Both had comments about the cause ("light markings" or "pencil problems" or something like that). This means that the changes could be dominated by a few rare events, in which case this sample is insufficient to conclude much of anything. A few democratic precincts with pencil problems could turn the tide . . .
More tomorrow to say, I hope, but have other work to do.
ecarlson said...
I'll post a spreadsheet soon and you can check my data, but here's what I got: . . .
Error rate (by eyeball) 1/12,000
If this is representative of the final recount, then it's pretty hopeless."
Did you pay ANY attention to what I just posted above? The Star reports that "Most of Hennpin County has NOT reported." That's Minneapolis-St. Paul.
Thus, the sample is NOT representative of the overall vote. It's not designed to be! It's only supposed to test the voting machine system.
It's like doing a poll by taking a random sample of all counties in the U.S. WITHOUT bothering to account for population centers. So, Mineral County in Colorado with about 4,000 voters has the same chance of being sampled as Cook County Illinois, with the City of Chicago!
Now, we wouldn't think much of a poll like that. But, if they both used the same voting machines, sampling Mineral county might tell us if the machines were working properly or not. And in an election that wasn't close that might be enough.
It tells us squat about what the VOTE results will be though!
Cugel:
Thanks. I find someone of the people here throwing around numbers frustrating. They give the illusion of know what they are talking about, but really the more I look as their analysis, and the fact they can not answer my basic questions, the more I realize they are talking out fo their ass. You are right- I get that they are trying to pretend as if they are doing a scientific poll. Thus, my problem with the discussion.
There are two factors I am looking guys in your analysis, and if you do not provide those answers you are wasting our time:
a) Whether random samples can extrapolate the outcome of undervotes considering the geographical distribution of the undervote? Its apparent these are not meant to be polling distributions. just samples to test the machine count. That's not the question. The question is the 25,000 undercount for the Senate race. In other words, geographic location in the stae of MN.
b) Whether the sample address the likelihood of determining what will be the breakdown of numbers for Franken and Coleman amongst the 25,000? This is a demographic question and what we know from this data from the past.
You aren't answering any of this. Just randomly telling me that across the state (which is not the question) there has been some movement although small to Franken. That's not the core, as I understand Nate's math, of what is seeking to be determined here.
Your argument assume uniform distribution of error across geography and demographics.
Cugel:
Again thanks. You pretty much independently confirmed what I wrote. I only saw your new post after posting my own. I am not going crazy here. We essentially know nothing right now. Well, we know a little bit. Even with out the distrubtion advantages and where the error rate advantage in terms of demographics is likely to occur, Franken still managed to gain a vote. So that means in addition to the other advantages, there may be some slight tweaks to either candidates with regard to the machine. But thats about it.
I don't have experience blogging, so I don't know how to quote. I'm mostly responding to Cugel.
No, I didn't read your post. According to the blog, it was posted three minutes before mine - probably while I was typing.
I didn't have any reason to think this is a representative sample. Nonetheless . . .
You can see that the sample has a higher fraction of Coleman supporters than Franken supporters. This can be compensated for by simply dividing by the number of supporters in each case.
Of course, this doesn't guarantee anything. It may be that Franken supporters from one county have a high rate of misvoting, while those from another county have a very low rate. But that's pretty much simply wishful thinking.
But the main point I want to make right now is my use of the subjunctive. I said "IF" this is representative. Pointing out that some areas were more sampled than others doesn't prove it isn't representative, it simply brings it in doubt - a claim I agree with. But I haven't the knowledge or motivation to analyze the data beyond what is presented in the document.
My main contribution isn't my analysis, but the spreadsheet, which allows you to make your own analysis. If you feel the data is too biased, you don't have to use it. But there ARE ways to compensate for the fact that it is republican-heavy.
Another quick response to Cugel:
My limited data doesn't tell you where the 25k undervote goes, but the full data tells a great deal. The vast majority of them are intentional. Hardly surprising - there are lots of people here in NC where I live who followed the presidential race and didn't vote on anything else.
Of course, it could be that in the counties surveyed, the people are interested in only presidential politics and not in the senate. But if I had to guess, I'd guess the other way - Urban areas tend to be more blacks, many of whom would be motivated to vote for Obama, but would not know who was running for the Senate.
I hope I'm wrong, and Franken wins. I'll admit my simple analysis is probably too naive, and there are definite potential flaws. It COULD still go either way. But to me it isn't promising.
Matt W said...
I really love Al Franken and would love to have him in the Senate.
Because he's good enough, he's smart enough, and doggone it, people like him!
@reelgeist
here's one of your questions e actually can answer approximately:
"what will be the breakdown of numbers for Franken and Coleman amongst the 25,000? "
Roughly speaking : 0 for F, 0 for C.
We knew ahead of time that at least 70% of the 30,000 + undervotes were deliberate. Now we know with good confidence that, unless there's some huge anomaly hidden away in a small number of precincts, well over 90% of those 30,000 were deliberate. So there are very few new votes to be had. There's also no reason to think that they will skew strongly to F, although it would be surprising if they didn't skew a little toward him, given that we D's are so good at screwing up.
@Another Mike: Good point about Intrade, which I'd also been wondering about. Maybe the gamblers are slow afoot. Or maybe they know something about the rejected absentee ballots and the courts. I don't think they'd be dumb enough to still expect many unintentional undervotes, after getting the audit data./mbw
Michael
There is no way you can be "confident" given the demogrpahic and geographic distribution issues. As Cugel points out that wasn't even the purpose of the test. It was to test the machines, not the voter intent. I am headed to bed. This has been a big waste of time.
Vote for Nate!
http://www.dailykos.com/storyonly/2008/11/14/194243/92/359/660693
Here's the $64,000 question though: Begich is looking a lot more solid AK, he believes Franken will pick up undervotes in MN, and now there's a new election for Saxby/Chambliss with Obama volunteers heading down there to push GOTV efforts in GA... Is this enough to get Nate to change the odds of 60 blue seats in the senate?
Michael - There is no reason to believe that well over 90% of the undervotes were deliberate. As reelgeist pointed out, the purpose of the post election review is to check the machines and not voter intent. As part of this process, they specifically do not look for ballots with votes marked outside the range the scanners check. From the Minnesota SOS page you linked above I went to the full list of regulations for the post election review of electronic vorting machines(sec 206.89). Here is the key quote:
Subd. 4. Standard of acceptable performance by voting system. A comparison of the results compiled by the voting system with the postelection review described in this section must show that the results of the electronic voting system differed by no more than one-half of one percent from the manual count of the offices reviewed. Valid votes that have been marked by the voter outside the vote targets or using a manual marking device that cannot be read by the voting system must not be included in making the determination whether the voting system has met the standard of acceptable performance for any precinct.
Looking through the PDF of results of the post election review and there are plenty of "blank for office" votes that could potentially be real votes once the full recount (which is more thorough and inclusive in determining voter intent) is conducted.
The Minnesota SOS was on radio the other day. If I understood him correctly, he mentioned that the error rate of the machine was 0.2%, meaning that potentially 0.2% of otherwise unacceptable ballots could actually be accepted by the machine as if there's nothing wrong.
when the machine rejects a ballot, it is not considered a machine error, because that's precisely what the machine is designed to do. It is considered a voter error that is correctly detected by the machine.
On top of the detected voter error vis a vis ballots spat back out, there is the 0.2% possible undetected error.
I suppose that's probably one of the reasons why in MN a margin of less than 0.5% would trigger a hand recount.
Which means that over votes could be erroneously accepted, at the rate of up to 2 in a thousand.
This is a response to MattR
MattR points out that in evaluating the success of the voting machines, hand tabulators are not supposed to count it against the machine if the voter made an error which made the ballot unreadable by the machine.
The implication is that the tabulators did not record such errors by the voters, and therefore we have no data on such undervotes.
If you read through the data, however, there are clearly some cases in which the tabulators DID make such determinations. For example, comments were made on write-in ballots where the name was written in but the bubble was not filled in.
There is also a separate column where the tabulators are supposed to fill in how many of the miscounted votes are "Exceptions", meaning they are valid votes, but unreadable by the machine. This seems to have been used inconsistently.
In some cases the tabulators definitely counted cases where the voter indicated their intention in a non-machine readable format. Maybe they always counted them, we can't tell unless someone knows the instructions the tabulators were given. Were they told to count all such cases, but then subtract them as exceptions? I can't tell.
on MN: State law, quoted above, specifies that the audit is supposed to count exactly the same way as a full manual recount. there aren't any more votes in the sampled precincts, which lean only very slightly more to C than would a pure random voter sample over the state. Again, we can't prove from these data the non-existence of some small number of precincts (i.e. less than about 10) in which a huge F undervote is sitting waiting to be counted. We also can't prove that the tooth fairy won't bring us some votes. Seriously, maybe the lawyers have some.
Reminds me of Brecht, where the cook says (my rough memory) "We're in the hands of God." and the priest replies "I didn't know it was that bad."
/mbw
"About 5,180 absentee and questioned ballots from Mat-Su."
I believe that includes Wasilla.
"Now where are the guys who were bashing Nate for having Begich at 100% chance of winning?"
A 100% chance of winning means this race is over on election night. We don't have one of these long recounts.
Win or lose this 100% call of Begich was absolutely terrible, and Nate should learn a lesson from this.
Updated numbers:
I found two errors in my transcription of the raw data into Excel. New numbers:
Coleman 54988 -14 +28
Franken 51959 -16 +32
These numbers are more favorable for Franken.
Some analysis:
The numbers are too small to tell, so you can't extract any extrapolations with confidence.
The subtractions: Note that Franken has more subtractions than Coleman. Most subtractions are not labeled; those that are usually represent votes that got jammed part way in, and were subsequently refed, and they got counted twice. If you believe democrats "cause" these, then the subtractions will be even more important when we look at heavily democratic regions. If, on the other hand, they are random occurrences, then the tiny difference is just a statistical fluctuation. There were also other causes, such as republicans who mismarked their ballot such that it was recorded as a vote for Franken. So let's hope there is no bias among these subtractions.
The additions: Again, there is a preponderance of Franken additions, even though there were fewer Franken voters. This could be small numbers, or it might represent the oft-claimed democrat propensity for spoiling their ballot.
If you assume that the democrats are representative of democrats statewide, and republicans are representative of republicans statewide, and that subtractions occur equally on both sides, and additions are proportional to this sample, you predict a net change of 129 votes in Franken's favor. However, the error bars on this number are easily big enough to pass 200.
Finally, it is easy to look at the numbers and discover that two of the 123 precincts account for about half of the "additions." This suggests rare precinct-specific problems. A few such "problem precincts" could cause large changes in the tallies, and we haven't sufficient data to determine the frequency or distribution of problem precincts. If you look at these two precincts (on pages 473 and 482) you will see that they both occur in precincts that preferentially went to Franken. However, the rate of "discovered" votes in each case was greater for Coleman than for Franken supporters, which could be small statistics.
@ecarlson
Excellent analysis. Thanks. It still looks like F is at best near the edge of the plausible error bars, way under 50% chance, as far as the recount results are concerned. I wonder what's up with Intrade?/ mbw
OK, we all have to read this...
http://the-uptake.groups.theuptake.org/en/videogalleryView/id/1300/
Apparently the review turned up 6 net votes for Franken and the difference is now exactly 200 votes. Further there are apparently more votes that will be uncovered in a true recount than were uncovered in this review. Further still, there was a poor standardization of methodology in the review, as many of us suspected from examining the pdf.
Franken Lives!
I think many of you are missing the point of the MN SOS machine check. As has been stated by a couple of other posters, the purpose of the count is not to determine voter intent--it's to determine machine accuracy; i.e., did the machine count what it was supposed to count and not count what it wasn't supposed to? Undervotes will not be picked up here. The machine will rightly not count ballots where the voter did anything other than fill in the oval with one choice. Circling, crossing out, checking, x-ing, or some combination of the above will not be counted against the accuracy of the machine, even if it is in the comments of the .pdf data. However, because of MN's law that errs on the side of inclusion in determining voter intent, the above examples of alternative vote-marking methods that are machine-unfriendly *will* be picked up in a full hand recount with canvassers from both sides present. In conclusion, extrapolating anything (other than how many votes candidates will pick up from the hand recount correcting the tiny machine error rate)from the machine check is impossible, and Franken supporters shouldn't be disheartened by the report or any of the posts here because the full hand recount is still to come and there is likely to be a (relatively) sizable Franken pickup.
"In Cottonwood County, a ballot was lightly marked for Norm Coleman but the electronic vote scanner did not record a vote for Coleman. This was discovered when Cottonwood election officials counted the ballots by hand for the audit. However, election official Jan Johnson sent in the audit indicating there had been no change in the vote for Coleman. Johnson told The UpTake that "voter intent" was not to be determined during the audit, but the ballot would definitely be in the Coleman column when t here was a recount."
@ecarlson
Actually, in one of those 2 precincts, more votes were discovered for F.
Check the discovered votes in the other elections. There are also spikes in them for POTUS and House in those two precincts. But there's a consistent pattern (which I'd predicted ahead of time) that there are more of these unintentional light-mark undervotes down ballot. People just get hastier and leave lighter marks. Thus the standard claim that POTUS undervotes represent a hard upper bound on unintentional undervotes in any race is not quite right.
It still is quite unlikely that enough magic precincts will show up for a win, but these data provide an excellent lab.
/mbw
@kl'
"The Secretary of State's office disagreed, telling The UpTake that voter intent was to be a factor in the audit as well as the recount."
ALSO!
"Nonpartisan observers warn not to extrapolate stats statewide"
http://twitter.com/theuptake/statuses/1007063502
Nate, Sean, Brent....thanks for continuing the site, and all the great info. And congrats, Nate, on the book deal, if that indeed does come to pass. You deserve it. And Nate, I guess since Prop 8 passed, we will have to meet in Massachusetts instead of California to get married. And it's not about the money...really...
From what I'm reading of those of you who've run stats, there is cause for optimism, as you're projecting Franken to pick up quire a few votes just from machine error alone, before any voter intent is analyzed...
@matt w., I read the news article and the links to the SOS as well, but I interpret that to mean within the context of the audit, how likely was the machine to pick up ballots where voter intent was clear, for example like pencil marks that were too light. The reason I'd interpret it like that is because if you're auditing how highly the machine is functioning mechanically, it's designed to throw out ballots with 'stray marks' such as circles and check marks. I think if you included stuff like that in a machine audit, it sets an impossible standard and you'd have a lot more 'failed' machines. I am in WA, work for the State, have a friend who worked in the SOS's office here during the 2004 recount in Rossi v. Gregoire, and she told me the process of certifying how well the machine counts is separate from any recount process - here, it happened before the official recount which Gregoire won. Granted, the state law here is different, but I couldn't imagine an audit which made the machine responsible for picking up things for which it's not designed and of which it is incapable of discerning.
P.S. (Sorry, Brett...not Brent...it will NEVER happen again).
Nate, wow! Just wanted to say you're doing a great job. Saw you on Hardball yesterday.
Congrats on the book deal Nate!
I went to the Uptake to figure out what is going on.
They have data at the bottom, and a link to the source of their data.
Their data lists:
Washington Cottage Grove P3 Coleman: -2
Their source lists:
Washington Cottage Grove P3 Franken: -2
Their source is:
http://www.sos.state.mn.us/docs/post_election_audit_review.pdf
Relevant data is on page 575.
This brings their numbers to match mine (which doesn't mean I'm right)
It sounds like the numbers in the audit are not necessarily the same as would occur in a recount, in any case.
In response to kI' (hope I typed that name right)
It may be true that the purpose of the spot check is not to determine voter intent. But it is clear from the pdf that voter intent WAS determined in many cases. Apparently, it was not done consistently.
Sample ballot from the precinct that had the highest number of errors picked up in the review...
http://www.co.st-louis.mn.us/slcportal/Portals/0/Sample%20Ballot%20-State%20General%20Election%20101608.pdf
Congratulations Nate on new book deals:
http://www.observer.com/2008/media/nate-silver-signs-penguin-two-book-deal-worth-sum-high-six-figures
If The Uptake is right about the gap having closed from 206 to 200, this is not reflected in the official numbers on the SOS elections website, which, despite having been updated as recently as this morning, still show the 206 vote gap.
http://electionresults.sos.state.mn.us/20081104/ElecRslts.asp?M=S&R=all&P=A&Races=''
Does anyone know, for us elections junkies (formerly: poll junkies) how frequently they will update the numbers in MN? Will it all just come in the end, or, as in the case of FL 2000, in dribs and drabs?
wv "shban": the sound the door made as God slammed it in Sarah Palin's face with Begich's lead increasing to 1,022 votes.
You told us to stick around after the election Nate, but what are you gonna do when the recounts are over? My thought is you have a good eye for news, not just numbers, and you should expand your focus a little. I'm worried that you're losing all the people that checked your site five times a day towards the end of the election, because you are potentially a great independent voice. I'm not saying you should go starting up The Silver Post, and write commentaries. Just saying these recount updates are a little lacking.
Franken is up to 56% on Intrade. The only news I can think of is the machine audit. So, since Franken's win percent has gone up about 10% in the last 24 hours, the bettors must think it is somewhat good news for Franken.
@ecarlson,
As you said, voter intent is not used in verifying the machines. But the tabulators have to determine voter intent so that they can classify valid votes missed by the machine because they are valid by voter intent, put them in the "exceptions" column, and not count them against the machine in their test.
That means that these numbers should be useful in predicting how much change the full manual recount is likely to produce.
Unfortunately the numbers are too small and the sample is not random enough (undersampling the large counties) for that purpose. Are those two precincts in St Louis County with light mark/pencil problems representative of that county, which voted 2:1 for Franken? Is it meaningful that in one of those two precincts the light mark errors resolved in favor of Coleman in much higher percentage than his percentage of the vote in that precinct? The numbers do not answer those questions, as there are only two precincts with that many and those kinds of errors.
The net change in this partial recount is +2 for Franken. Does that mean that counting 20 as many precincts, or 40 times as many ballots will result in +400 or +800 votes for Franken? Maybe. Or maybe not, as there isn't enough data in this sample to extrapolate that way. It does mean that a Franken win is possible. Or a Franken loss.
@bugstomper
20x2=40.
Of course, it's not even close to statistically reliable to try extrapolating the net vote shift from this little sample, even if we knew how to multiply.
More seriously, these data indicate a much lower rate of correctable errors (new votes) than we had been expecting. Even allowing for some undersampling of big cities, and the large statistical error in estimating the number of problem precincts when only two of these were found, it's very hard to see ending up with more than 1000 or 2000 of these new votes. All our calculations suggesting a good chance of a win for F were based on the belief that there would be some 5000 such new votes.
The odds look terrible, so it would be interesting to know what the gamblers see tht makes them think otherwise.
/mbw
Well I see Michael is continuing with whatever agenda he has. Gone for a day, and still the same disproven claims being made. At this point you are either a concern troll or a partisan.
Please ignore comments by "Michael." They have no basis in data or fact and are intended to drive a Republican narrative. He has provided nothing but negative speculation. Don't feed the concern troll, peeps.
Let's be clear here: the GOP is throwing major resources into poisoning the media environment prior to the mandated recount, starting with Coleman's whine the day after the election that a recount would be too costly.
It is likely that "Michael" is a Republican shill or even on the payroll, a blogger tradition around here since PowerLine got started receiving secret GOP money. Hell, "Michael" might BE Hinderacker or one of his toadies.
There is absolutely no way to know what the recount will turn up on the pencil-marked ballot sheets. As an absentee voter, it pisses me off that the GOP is working hard to disqualify every absentee ballot they can. Choke on that fact, "Michael."
(Verification code "loser!")
@ Nicole.
"As an absentee voter, it pisses me off that the GOP is working hard to disqualify every absentee ballot they can."
Me too, which is why I just sent Franken another $100 to help fight back. I suggest you do the same. You can indulge your vivid fantasy life after the GA vote and the MN recount are done./mbw
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,情趣,情趣商品,情趣網站,跳蛋,按摩棒,充氣娃娃,自慰套,G點,性感內衣,情趣內衣,角色扮演,生日禮物,生日精品,自慰,打手槍,潮吹,高潮,後庭,情色論譠,影片下載,遊戲下載,手機鈴聲,音樂下載,開獎號碼,統一發票號碼,夜市,統一發票對獎,保險套,做愛,減肥,美容,瘦身,當舖,軟體下載,汽車,機車,手機,來電答鈴,週年慶,美食,徵信社,網頁設計,網站設計,室內設計,靈異照片,同志,聊天室,運動彩券,大樂透,威力彩,搬家公司,除蟲,偷拍,自拍,無名破解,av女優,小說,民宿,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,懷孕,美女交友,交友,日本av,日本,機票,香水,股市,股市行情, 股市分析,租房子,成人影片,免費影片,醫學美容,免費算命,算命,姓名配對,姓名學,姓名學免費,遊戲,好玩遊戲,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,桌布下載,電視節目表,線上電視,線上a片,線上掃毒,線上翻譯,購物車,身分證製造機,身分證產生器,手機,二手車,中古車,法拍屋,歌詞,音樂,音樂網,火車,房屋,情趣用品,情趣,情趣商品,情趣網站,跳蛋,按摩棒,充氣娃娃,自慰套, G點,性感內衣,情趣內衣,角色扮演,生日禮物,精品,禮品,自慰,打手槍,潮吹,高潮,後庭,情色論譠,影片下載,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,減肥,美容,瘦身,當舖,軟體下載,汽車,機車,手機,來電答鈴,週年慶,美食,徵信社,網頁設計,網站設計,室內設計,靈異照片,同志,聊天室,運動彩券,,大樂透,威力彩,搬家公司,除蟲,偷拍,自拍,無名破解, av女優,小說,民宿,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,懷孕,美女交友,交友,日本av ,日本,機票,香水,股市,股市行情,股市分析,租房子,成人影片,免費影片,醫學美容,免費算命,算命,姓名配對,姓名學,姓名學免費,遊戲,好玩遊戲,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,桌布下載,電視節目表,線上電視,線上a片,線上a片,線上翻譯,購物車,身分證製造機,身分證產生器,手機,二手車,中古車,法拍屋,歌詞,音樂,音樂網,借錢,房屋,街頭籃球,找工作,旅行社,六合彩,婚友,單身聯誼,未婚聯誼,婚友,交友,交友,婚友社,婚友社,婚友社,大陸新娘,大陸新娘,越南新娘,越南新娘,外籍新娘,外籍新娘,台中坐月子中心,搬家公司,搬家公司,中和搬家,台北搬家,板橋搬家,新店搬家,線上客服,網頁設計,線上客服,網頁設計,植牙,關鍵字,關鍵字,seo,seo,網路排名,自然排序,網路排名軟體,交友,越南新娘,婚友社,外籍新娘,大陸新娘,越南新娘,交友,外籍新娘,視訊聊天,大陸新娘,婚友社,婚友,越南新娘,大陸新娘,越南新娘,視訊交友,外籍新娘,網路排名,網路排名軟體,網站排名優化大師,關鍵字排名大師,網站排名seo大師,關鍵字行銷專家,關鍵字,seo,關鍵字行銷,網頁排序,網頁排名,關鍵字大師,seo大,自然排名,網站排序,網路行銷創業,汽車借款,汽車借錢,汽車貸款,汽車貸款,拉皮,抽脂,近視雷射,隆乳,隆鼻,變性,雙眼皮,眼袋,牙齒,下巴,植牙,人工植牙,植髮,雷射美容,膠原蛋白,皮膚科,醫學美容,玻尿酸,肉毒桿菌,微晶瓷,電波拉皮,脈衝光,關鍵字,關鍵字,seo,seo,網路排名,自然排序,網路排名軟體,英語演講,托福,Toastmaster,
^^ 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,
^^ 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,
外遇外遇外遇外遇外遇外遇外遇外遇外遇 外遇
外遇 外遇外遇 外遇 外遇
外遇 外遇 外遇
外遇
外遇 外遇
外遇
外遇 外遇外遇
外遇
外遇 外遇外遇 外遇 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 ,
外遇 外遇 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇外遇 外遇外遇 外遇 外遇
外遇 外遇
外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇 , 外遇劈腿劈腿劈腿劈腿劈腿劈腿劈腿劈腿喜帖囍帖卡片外遇外遇 外遇 外遇外遇 外遇
外遇 外遇 外遇 外遇剖析 外遇調查 外遇案例 外遇諮詢 偷情 第三者外遇話題 外遇發洩 感情挽回 徵信社 外遇心態 外遇 通姦 通姦罪 外遇徵信社徵信社外遇 外遇 抓姦徵信協會徵信公司 包二奶 徵信社 徵信 徵信社 徵信社 徵信社 徵信社 徵信 徵信 婚姻 婚前徵信 前科 個人資料 外遇 第三者 徵信社 偵探社 抓姦 偵探社 偵探社婚 偵探社 偵探社偵探家事服務家事服務家電維修家事服務家事服務家事服務家事服務家事服務持久持久持久持久持久持久持久離婚網頁設計徵信社徵信社徵信徵信社外遇離婚協議書劈腿持久持久持久持久持久劈腿剖析徵信徵信社外遇外遇外遇外遇徵信社徵信社徵信社徵信社徵信社徵信社徵信徵信社徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信徵信社徵信社徵信社徵信社徵信社徵信社徵信社徵信社
艾葳酒店經紀公司提供專業的酒店經紀, 飯局小姐,領檯人員,領台,傳播妹,或者想要到台北酒店、林森北路酒店,私人招待所,或者八大行業當酒店PT,酒店公關,酒店兼職,想去酒店上班, 日式酒店,制服酒店,ktv酒店,禮服店,整天穿得水水漂漂的禮服酒店,鋼琴酒吧當酒店領檯,酒店小姐,公關小姐??,還是想去制服店當上班小姐,水水們如果想要擁有打工工作、晚上兼差工作、兼差打工、假日兼職、兼職工作、學生兼差、兼差、打工兼差、日領工作、晚上兼差工作、酒店工作、酒店上班、酒店打工、兼職、兼差、兼差工作、酒店上班等,想了解酒店相關工作和特種行業內容,想找打工、假日兼職、兼差打工、或晚班兼職想擁有快速賺錢又有保障的工作嗎???又可以現領請找專業又有保障的艾葳酒店經紀公司!
艾葳酒店經紀是合法的公司工作環境高雅時尚，無業績壓力，無脫秀無喝酒壓力，高層次會員制客源，工作輕鬆，可日領、現領。
一般的酒店經紀只會在水水們第一次上班和領薪水時出現而已，對水水們的上班安全一點保障都沒有！艾葳酒店經紀公司的水水們上班時全程媽咪作陪，不需擔心！只提供最優質的酒店打工,酒店上班,酒店打工環境、上班條件給水水們。
水水們妳有缺現金、有卡債、缺錢卡奴的煩腦嗎?想到日本留學日本打工嗎?妳是工讀生找工作??想要擁有高時薪又輕鬆的夜間兼職工作,打工機會和,假日打工,兼職工作日領假日打工的機會嗎??想實現夢想卻又缺錢沒錢嗎!??整天還在煩腦如何賺錢有什麼賺錢方法,和賺錢最快方法!?,想要打工,日領工作,短期打工,兼差工作,打工兼差工作嗎!?,
請加入我們艾葳酒店經紀公司工作單純輕鬆”高時薪”又可日領徵想要當傳播妹,上班小姐,酒店兼差,酒店兼職,歡迎學生打工,!!!
加入我們實現夢想就從現在開始^__^
酒店經紀人,
菲梵酒店經紀,
酒店經紀,
禮服酒店上班,
酒店小姐兼職,
便服酒店經紀,
酒店打工經紀,
制服酒店工作,
專業酒店經紀,
合法酒店經紀,
酒店暑假打工,
酒店寒假打工,
酒店經紀人,
菲梵酒店經紀,
酒店經紀,
禮服酒店上班,
酒店經紀人,
菲梵酒店經紀,
酒店經紀,
禮服酒店上班,
酒店小姐兼職,
便服酒店工作,
酒店打工經紀,
制服酒店經紀,
專業酒店經紀,
合法酒店經紀,
酒店暑假打工,
酒店寒假打工,
酒店經紀人,
菲梵酒店經紀,
酒店經紀,
禮服酒店上班,
酒店小姐兼職,
便服酒店工作,
酒店打工經紀,
制服酒店經紀,
酒店經紀,
菲
梵,
Post a Comment