Our good friends over at the Horse's Math blog (think of a circus horse that pretends to count to three by stomping its hoof) continue to perform math tricks for their readers.
I thought I’d take a look at a single number: 271. That’s the total number of additional or subtracted votes counties have reported thus far.It's true that if trends persist from the counties that have reported their recounts, then we can expect several hundred new ballots to be discovered in the remaining counties. But I'll try to do a better job of analyzing their significance than the Horse Crap Math people do with their specious assertion of "42 votes give or take two thousand".
...
With only about 14% of the counties reporting, this projects to about 2000 changed votes statewide.And so to all those who argued that Rossi won the first two counts, thus there was no need for a third, I ask you: are you really comfortable saying that Rossi won by 42 votes… give or take a couple a thousand?
As of this writing (and including the as yet unofficial report from Asotin county), Rossi's lead in the reported counties is +31,054. So far we have counted 259 more two-party votes than in the machine recount. In the as yet unreported counties, Gregoire enjoyed a lead of 30,979 votes in the machine recount, which translates into 50.6438% of the two-party vote in those counties (with Rossi's share 49.3562%). The counties that have reported represent 12.2% of the statewide vote.
Based on the 259 new two-party votes in the 12.2% of the state that has reported, we extrapolate to 1,860 new two-party votes in the rest of the state. (This extrapolation is debatable for many reasons, but let's stick with it for argument's sake). In order for these new ballots alone to flip the outcome of the election from a Rossi win to a Gregoire win (or tie), Gregoire would need a lead of at least 75 votes from these new ballots.
The Horse's Math people have a self-defeating habit of making fools of themselves by ridiculing those who actually know how to use Excel. At the risk of encouraging them to shoot themselves in the hindquarters yet again, I will now explain how to use Excel to calculate the probability that Gregoire could win a 75 vote lead out of 1,860 ballots.
If we assume that the newly discovered ballots are as likely to be for one candidate or the other as the previously counted ballots in the same counties (and there's no real reason why they shouldn't), then this is a cumulative binomial distribution problem, where we have 1,860 trials, the probability of success (a random ballot being for Rossi) is 49.3562%, and we have a scenario where at most 892 of these ballots are to fall to Rossi. This is calculated by =BINOMDIST(892,1860,.493562,1) = 11.8%.
So the probability that Gregoire picks up enough new votes is less than 12%. As I've written many times before, this is not a sure win for anybody, but the odds favor Dino Rossi. And of course this can change as new information presents itself and if the rules of the recount are to change.
Naturally, some will be troubled by any attempt to apply actual math to a math problem. But those who prefer to read specious platitudes and watch a circus horse stomp its hoof and count to three know where to find that sort of commentary.
UPDATE: Yikes! Speaking of new information presenting itself, King County has discovered a whole bunch of new ballots
Approximately 561 absentee ballots were mistakenly rejected because the signature on the ballot did not match the original voter registration records. In fact, these were signed ballots where a signature was not on file in the county's voter registration system. Original registration records should have been retrieved to verify the ballot signatures.It's not clear why signatures are not on file in the county's voter registration system and how the ballot was determined not to match the original voter registration records if the original voter registration records had not been retrieved. It will be very interesting to hear an explanation for all of this. Posted by Stefan Sharkansky at December 13, 2004 12:07 PM | Email This
Kind of like when Democrats were beating the "count every vote" mantra when they should have meant every legal vote, I'm finding the hair splitting to be interesting and figure I'll add to the punditry.
I think this is an issue of a "statistical tie" from a voting process that mechanical, manual or otherwise that has a margin of error that exceeds 42 votes, 216, or probably fair amount more. However, this election and law don't recognize the concept of a "election tie" other than the fairly small chance of the recount having the exact same number of votes. (I don't county anything out anymore) What's going on is we as a state are finding some way through this conflict.
Having been an observer in Franklin County anyway--I've been impressed with everybody (D, R, and other) to want to recount accurately.
What's going to be VERY interesting is the broadcast in 65 minutes.
Posted by: Matt on December 13, 2004 12:24 PMWe'll hear all kind of rhetoric from the MSM in Seattle about how "counting this many ballots is just a difficult process." And how, "it's really tough for us to keep track of all the signatures." And how, "mistakes are always bound to happen in an election of this size, etc."
Over and over again we've seen the arrival of ballots that were previously non-existent. And we've seen every sort of apology and explanation.
561 New Ballots? How come King County can't get this right? Fire that incompetent Dean Logan ASAP.
Posted by: Jeff B. on December 13, 2004 12:50 PMIf we were to hold a runoff or a recall, Rossi would win hands down because so many of the WA voters are simply disgusted with this woman and her "say anything, do anything necessary to win" mantra.
Posted by: Jeff B. on December 13, 2004 01:12 PMThanks for your feedback and please write again soon, if you can spare the time.
Posted by: Stefan Sharkansky on December 13, 2004 02:26 PMMy "42 votes, give or take a couple thousand" was intentionally vague, because I felt it better communicated my point than a pageful of calculations. There was a 42 vote margin, and now hundreds (maybe thousands) of votes are being added or subtracted from the totals. You can calculate all you want the probability that it will swing the election one way or the other, but the point it is... this election was still worthy of a hand recount.
Remember, those 561 King votes aren't the only new ballots being "recanvassed" in this recount. You simply cannot calculate reliable projections from crappy data. But if it makes you feel better about yourself to ridicule me for pointing this out, by all means, have your little fun. I rather enjoy being your foil.
Posted by: David Goldstein on December 13, 2004 03:22 PMAnother reason to be hopeful is that the Dems have been trying really hard in King County to contact rejected absentee and provisional ballot voters who were believed to favor Gregoire, and have them submit new signature documents. These 561 absentee voters would have been included on the list of people who could have been contacted, but for whatever reason, the Dems failed to or (more hopefully) chose not to. With any luck, and perhaps some divine intervention, these 561 absentee voters might even net Rossi a few votes.
One question I have is whether or not King County Elections bothered to contact these people by letter, as the WAC regulations clearly require whenever an absentee ballot is not signed or is signed with a signature that does not appear to match. If these voters were contacted, and did not respond, then excluding their votes would seem to be appropriate and at least not unfair. However, the RCW and the WAC clearly require signature matching, and if the auditor does not have a signature on file, or does not bother to check the signature that is on file, rejecting the ballots would not be appropriate.
Hopefully, my optimism (or at least hope) does not clue the Democrats into realizing that they somehow need to dredge up even more new ballots in King County to ensure that Gregoire comes out ahead ...
Posted by: Richard Pope on December 13, 2004 03:47 PMIn other words, when Stefan uses a more objective method to find the true answer to the very small differences between the results, such as using proper mathematical / statistical formulas, he makes it much harder for Goldy to argue in vague subjective terms that the election is a tie.
It is true that the data is still changing, but Stefan makes that disclaimer. If indeed, the Democrats continue to find ballots, reorganize databases to find matching signatures, and whatever else they need to do to buy the time to unravel the Dean Logan quagmire and end up with votes for Gregoire, the election indeed may swing to her.
Consider it a virtual miracle that Dino is in the race at all with 82+% turnout in this state, including 83% turnout in King (v. about 76% in King, if I recall correctly, in 2000 when Slade lost).
You're off base going after the King Co GOP on this issue. I saw GOTV efforts first hand this year as well as 1998 and 2000. Not that's it can't get better, but it's a lot better now than in the past.
I'm normally one for eagerly pointing out the flaws of local GOP party structures, but in this case the GOTV effort was driven by the RNC through the WSRP, via the 72-hour taskforce.
While WA wasn't a true battleground for Presidential purposes, the RNC kept the 72-hour taskforce rolling here to aide Dino and the races in the 5th & 8th CDs. Obviously, county parties provided support but it wasn't their gig.
What ifs are easy at this point, but at a macro level you're barking up the wrong tree.
Posted by: eric on December 13, 2004 06:32 PM"the point it is... this election was still worthy of a hand recount."
(Well... I guess it would have been clearer if I had left out the extraneous "it", but we're all accustomed to typos in blog comments, aren't we?)
Read the original blog entry that Stefan is criticizing. I had a simple premise:
And so to all those who argued that Rossi won the first two counts, thus there was no need for a third, I ask you: are you really comfortable saying that Rossi won by 42 votes… give or take a couple a thousand?
And rather than answering my question, Stefan chooses to ridicule my math skills -- even while accepting the one and only mathematical assumption I make -- before once again digressing into yet more calculations. The problem is... I wasn't posing a math question. To refute my straightforward premise with probability calculations would be like running a spreadsheet to refute my atheism.
Here is the reality: the hand recounts are turning a portion of the under vote into real votes, in numbers far exceeding the slim 42 vote margin. So let me rephrase the question... does that or does that not suggest that a hand recount was reasonable under these circumstances?
Posted by: David Goldstein on December 13, 2004 06:43 PMMr. Goldstein - Why is that exactly? A hand recount is LESS accurate than a machine count. Chirp all you want, but anyone with half a brain (or anyone whose job depends on accuracy) uses a calculator to do math and Excel spreadsheets to sum large amounts of numbers.
I have been observing the King County recount, and my good friends and relatives have been involved in the count. From everything I have witnessed and heard, it is a farce.
Each ballot has a bar code indicating the precinct from which it originated. This bar code can be read by the optical scanners along with the vote. Easy cheesy, right?
Counting hundreds of thousands of ballots by hand using hundreds of different counters after sorting by hand into precincts etc etc etc is not more accurate.
The point is: THIS ELECTION WAS, AND IS, STILL WORTHY OF AN ACCURATE COUNT! A hand count is less accurate than a machine count. The machine counts are as close as we'll ever get to the most accurate count.
Get the point?
I'm not ridiculing your math skills as such. I'm ridiculing your bogus arguments that you dress up as being grounded in a real world thought process ("42 votes give or take a couple of thousand", "statistical tie", "margin of error"), but don't actually have any substance behind them and where you employ phrases that you brag about not understanding. I'm also ridiculing your ridicule of me for actually doing the work to explain to others how to think about these concepts in a more meaningful way.
But don't take it personally and I'm glad you enjoy being my "foil". I take that to mean that you admit to wearing an aluminum foil hat!
Posted by: Stefan Sharkansky on December 13, 2004 10:16 PMThis analysis (like any other) has a few debatable
assumptions (like equal probabilities of finding new
ballots from any county), but overall, it seems very
reasonable.
If these are online somewhere, someone please post a link...
Posted by: Mike on December 14, 2004 08:09 AMMr. Goldstein - Why is that exactly? A hand recount is LESS accurate than a machine count. Chirp all you want, but anyone with half a brain (or anyone whose job depends on accuracy) uses a calculator to do math and Excel spreadsheets to sum large amounts of numbers.
That is NOT what the CalTech/MIT studies say. Show me the research that says hand counts are less accurate than machine counts.
Posted by: David Goldstein on December 14, 2004 02:01 PMSee my correspondence with him, here
Posted by: Stefan Sharkansky on December 14, 2004 02:20 PM