Several of you are asking why the EV estimator took a sharp dive, from >320 EV yesterday to 303 EV today. Under current conditions, the Median EV estimator is prone to making some big jumps. Two large states, Florida (29 EV) and North Carolina (15 EV), were on the edge, to generate four major possibilities. Then North Carolina dropped out of contention, leaving one tie, and therefore pairs of large peaks that are 29 EV apart. This “lumpiness” makes the median inherently unstable.
For a smoothly-varying number, watch the Meta-Margin, which uses the information found in state polling margins and is better behaved:
As you can see, the Meta-Margin, defined as how much swing it would take to make the EV distribution a perfect tossup, has been moving up steadily. It’s almost at Obama +3.0%, the center of the likely range that I proposed back in July/August.
Don’t (just) watch the EV estimator
November 5th, 2012, 1:36am by Sam Wang
Tags: 2012 Election · President
and isn’t O+3.0 also the congressional house “tie” MM? I’ve been watching that with great interest, in the hopes it gets comfortably past 3.0 just for that reason…
I was thinking about that too. From year to year, the Presidential and House national popular vote are only loosely correlated. Within year, it’s less clear because the generic Congressional numbers are so flaky.
But what you are suggesting is still possible. If Democrats gain significant seats in the House, this late rise may be connected.
For a house majority, I’ll eat a bug.
Yup, bumpy lately.
yu like us to try and figure it out first, dont you?
All outcomes graph went from 6 tall stalks to 2
The previously-observed correlation between the Senate and Electoral College trends suggests to me that this year there could be a similar correlation between the House and Electoral College–call it the “nationalized Congressional election” hypothesis. I sure wish there was more data available on the House.
I’ll join the “I’ll eat a bug” in solidarity with Sam group (with sadness on my face), and the I’ll eat a bug if the Dems take the House (but it will be with glee).
However, I’ve just got to say as someone who works in entomology – if you like shrimp or lobster, you’re inhibitions against what you are imagining as bugs are entirely mental. For a site community that pats itself on the back about its objectivity so much, this is funny. I guess cool rationality about the numbers cannot be expected to necessarily lead to cool rationality regarding things in the sphere of biology.
(And yes, there are many insects and arachnids that taste great, and are perfectly safe to eat.)
Thank you for the reminder that the outcomes distribution is as it is; while we should all have remembered that, having the graph pointed out clarifies and re-focuses.
More interesting than the long-foregone presidential outcome, though, is the Congress. Is the Senate still looking like 53-54 D+I? Any late thoughts on the House?
Is there any reason not to report both the median EV estimator and a mean EV estimator?
The mean presumably would be less jumpy, with the median hopping back and forth across it. People might then be a bit less spooked or elated by the jumps.
Unnecessary since MM handles that. At some point there are too many numbers.
florida counts 29ev not 27
I really hope you’re right, Sam.
This country will be unfathomably better off if the racist, homophobic, flailing GOP is sent to a time-out to think about what they’ve done.
Hi Sam,
When will you post the finals? Thanks.
Hey, that’s _my_ question!
After all this excitement, it’s the velociRapture
I guess most agregators would produce their final numbers once all polling has been included. But the question is, is there a time when polling really stops? Can we expect the final polls to be published tonight? or will some firms still be publishing poll results on Tuesday morning? afternoon?
MOST polling firms have wrapped up their final polls, I know PPP is done, same with NBC/Marist. Might be a few firms that come out with some new numbers today, but to go back to the old metaphor, this election is in the oven.
yeah, but i remember in 2008, everytime we thought all the numbers were in, there would be some new polls released. Just wondering if anybody remembers when agregators usually were publishing their final predictions.
This is funny – Sabato’s final prediction gives Obama 290 EVs. He (Sabato) says he can’t pick CO, NH or VA so he’s splitting their total EVs (26) down the middle by awarding VA (13 EVs) to Romney and the other two to Obama. “It’s not very scientific, ” he admits. Er… no…
http://www.centerforpolitics.org/crystalball/articles/projection-obama-will-likely-win-second-term/
Sabato’s gut is almost as spectacular as mine. Since we really don’t know anything about how the electorate in a given state feels — state polls are all within the margin of error and thus tied according to the Wolf Blitzer Principle — we have to rely on the guts of savvy political prognosticators and mine tells me that the Old Dominion will go to Romney because I granted the other states to Obama. That’s pure political science, my friend, and I’m not sure you’re ready for it.
Stick with this nerd brainiac and don’t wear dickish bow ties to cocktail parties and see where that gets you!
Kind of surprised he didn’t say anything about the Washington Redskins losing to Carolina yesterday….he may as well add that consideration to his prediction, it’s about as scientific.
Sam, what is impressive to me, and will be deeply disheartening to political operatives if you and others nail this election, is how stable the race has been on a longer scale. Your mentioning of the meta-margin being about what you suggested in July reminds that that 38 is trending about where it was when it debuted in July (or June, don’t have time to look it up right now).
So, all of the bluster and handwringing notwithstanding, this race may well be marked more by a stubborn persistence more than by volatility. No political consultant anywhere will want to hear that, I suspect.
Sandy will prove to be a Godsend for the political grifters and punditerati. Trend lines be damned, an Obama win *will* be due to the storm disrupting Romney’s message, allowing Obama to look “Presidential,” and creating Christie’s “Come to Jesus” moment.
Their jobs depend upon it.
Nate just posted an article speculating about the Sandy story dominating retrospective coverage of the campaign… and pointing out that that would be wrong.
I think Sandy did change the pundit narrative in one underrecognized way: it knocked out the Gallup daily tracker.
Courtesy of Frum, Dick Morris, who predicted a landslide win for Romney, is already hedging by talking up Sandy.
http://www.dickmorris.com/in-the-last-few-hours-sudden-danger-signs-in-polling/
re: “landslide”
http://www.youtube.com/watch?v=YIP6EwqMEoE
I have been watching the debacle unfold in Florida and Ohio with the efforts to suppress voting. The incredibly long lines of people waiting to vote is just indefensible in the United States. In Florida yesterday they were towing people’s cars while they waited for hours in line to vote. It’s embarrassing. With the closeness of the race in Florida, this effort to suppress voting there might be a factor. My guess is the effects in Ohio may be somewhat less.
I don’t anticipate it being a huge problem in Ohio. Ohio has had a long time to vote, and voting continues today. While Husted fought tooth and nail to limit early voting, there hasn’t been a huge suppression issue since. the court ruling was handed down.
I spent yesterday afternoon in downtown Cleveland taking our middle tadpole to vote in his first general election. (I guess I should stop referring to him as a tadpole now that he’s in college.) Lines were long, in excess of 2 hours, but people were in good spirits, determined to vote, and primed for an Obama victory. When we finished up, about 30 minutes before the four-hour Sunday early voting was supposed to end, the lines were nearly as long, so I expect that voting continued for at least an extra hour.
I don’t know how many people voted in Ohio yesterday, but we can thank the Federal courts for the opportunity. Ohio’s government fought tooth and nail against voting the weekend before the election, only acquiescing after the Supreme Court declined to act. The actions to restrict voting here and in other states (Florida, Pennsylvania, Texas) drive home the need to take state government control away from Republicans.
It seems to me that the Republican party is trying out two strategies for stealing the vote. In Florida, the strategy is to block people from casting a vote. In Ohio, the strategy is to mess with the results reported by the voting machines. Either way, they disenfranchise honest voters. In my view, the Republicans are committing treason.
@Froggie: He’s your tadpole forever. :)
Eeep. I misspelled Froggy. Sorry, Froggy.
Sam, do you ever sleep?
I was going to mention that, too. Between the stress of the storm and not sleeping because he’s responding to us, I’m worried that Sam will get sick.
So get some sleep, Sam.
Ohio Voter:
Some people are making a big deal about the item below – what do you think?
“COLUMBUS, Ohio — Voter advocates are criticizing an order by Ohio’s elections chief dealing with the casting of provisional ballots. Advocates are saying on Saturday that the order by Ohio Secretary of State Jon Husted late Friday wrongly puts the burden of recording the form of ID used on a provisional ballot on voters, not pollworkers ….”
As a poll-worker, it’s a pain in the rear all ready.
I hold in my hand a provisional ballot affirmation sample, and it’s fairly straightforward. Personally, I think there is a bigger deal being made about it than necessary. It requires a little voter forethought, but it’s not exactly confusing. The reasons for making a voter cast a provisional are usually because the voter didn’t do their due diligence.
I don’t think it’s going to be as big of deal as it’s being made out to be right now.
Here are the reasons for issuing a provisional ballot:
1) Voter’s name does not appear in the poll book (likely because they are in the wrong precinct)
2) The voter changed their name and didn’t update it
3) The voter has moved into the precinct, but didn’t update their registration
4) Voter requested an absentee ballot
5) Notice of Election card was returned to sender AND the voter doesn’t have valid ID
6) Voter refuses or is unable to show valid ID
7) Poll worker challenges to eligibility of a voter.
A thing that concerns me, though, is that politicians and judges might be increasingly taking the need for “voter forethought” and “due diligence” as ends in themselves, rather than possibly necessary evils. It’s the notion of voting as an IQ test, in which, if you’re too dumb to jump through hoops X, Y and Z, you’re too dumb to vote.
Its like Christmas eve.
Im still not convinced that averaging is sufficient to remove systematic error in the legacy polling methodology.
Guess we shall see.
;)
I agree. I think, if anything, this might be Drew Linzer’s year. I wonder if FL, for example, is being torqued by bad polls (like Mason Dixon) and inability to capture what is going on with Latino and cell phone vote.
My gut “Larry Sabato” prediction – the post election story might be “Latino vote surges Obama to unexpected big victory.”
But what do I know. I will not eat a bug if I am wrong, just some (metaphorical) crow.
I had the same nagging fears, but yesterday’s national polls from Pew and YouGov and even the Rasmussen tie reduced my fears to 1%. I can live with that for 48 hours. My other 1.5% fear is that the Republican governor of Ohio, Wisconsin, or Nevada or a combination thereof refuse to sign the Certificate of Ascertainment because he/they claim voter fraud or some other irregularities tainted the vote. Walker is crazy enough. Kasich is borderline. Now that would be interesting. Once again SCOTUS would swoop in. Grab your women and children and head for the hills.
lojo – just eat a shrimp or lobster in solidarity. You’ll essentially be eating a water bug. There’s not really any phylogenetic reason to pretend otherwise (if we’re using “bug” in its expansive form, not to refer narrowly to Hemiptera).
Hello Dr. Wang,
My apologies if you have already addressed this on elsewhere, but I wonder if you had two cents on this: http://www.huffingtonpost.com/eric-zuesse/revolutionary-study-shows_b_2074356.html?utm_hp_ref=tw. It’s a study which suggests that because polls measure voter intentions rather than expectations, that even some of the polls forecasting the election for Obama may be overrating Romney’s chances in the polls. Something tells me that the results from voter expectation polls would match PEC’s forecasts, since the study specifically critiques Nate Silver’s blog. It’s from HuffPo, so it’s not an objective source, but nevertheless interesting. Your thoughts?
Best Regards,
Sebastian Jackson
Sam,
I think the data is suspect as we should treat these Polls as two groups, a) about 35 States with early voting and b) the balance without….
In my view the National Polls are misleading as the early voting in majority of the States and those without behave differently in Polls and if the States that do not have early voting are small, then the Polls by the virtue of not capturing early votes are producing an even contest….
SO if the early votes favor Obama 2:1, my bet is that Obama is way ahead in National Polls when we capture the effects of early voting….
http://www.270towin.com/early-voting-2012-election/
My guess is we are at 54-46 favoring the President.
National polls are still about 2 points below the Meta-Margin… but now that is O+1, so it doesn’t seem so dramatic.
check out the nat’l polls at 538 this am.
blue.
Yeah, those are the polls I was looking at. Median, mode and mean are all about Obama +1.
Nate says this is more in line with state polls, but based on PEC I’m not sure it is; the weird systematic gap’s still there, it’s just that it’s all on Obama’s side now.
The justification for believing that some massive systematic is biasing all the polls from a Romney win is disappearing, for now, and there’s not a lot of time left.
Behold Matt!
State and nat’l polls align!
according to Nate this am.
all blue, blue, blue.
I’m feeling those blue vibrations.
http://www.youtube.com/watch?v=XL2od1AF_Cs
Matt, I know you guys all think im just another cyber-chicklet with a jones for loop gravity and mathematical modelling, but…im actually your culture farmer for this site.
My nonparametrics professor was a big fan of what he called the “farmer method”, that is, how farmers make the weather predictions you see in the farmer’s almanac. But you see…it isnt “gut feeling”– its a thousand thousand subtle parameters; the hotness of the summer, the length of wheat stalks, the coats on horses and caterpillars, the kinds and amounts of insects, the quality of the milk…and their livelihood depends on their forecasting.
Thats how im reading culture this election season. we are all immersed in it, swimming in it and theres a thousand thousand subtle signs for us to read, social networks, twitter traffic, face bookers, pundits, and best of all, the Rise of the New Numeratocracy.
And very very best of all….PEC. Where we can discuss it all.
So then Wheeler’s Cat, what is your Election Almanac telling you?
I would say that here in Northern Virginia, despite a noticeable increase of invasive species in the form of Romney yard signs, the “coats” of the flocks of GOTV targets that I’ve hit, using the Obama campaign’s amazing micro-targeting lists, are unusually blue, not to mention enthusiastic and determined, with lots of Latino, Middle Eastern and African American, as well as single women/men — you know, the kind that don’t represent Politico’s “broad mandate.”
My personal weather remains calm and cautiously optimistic, with bouts of obsessive poll checking; tomorrow’s forecast involves taking the day off to GOTV.
Ross C:
I know what you mean about those lists. My wife and I canvassed in Springfield last night, and three separate times we were invited in for supper by the MENA families whose meal we’d just interrupted. The targeting lists are amazing.
haha, yes Ross! i did gotv here on campus and i was able to micro-target possible GaJo defectors. My persuasion leverage was his position on gay rights, the anti-choice law he supported, and his desire for overturning child labor laws.
piece of cake.
;)
From +2.84 to +2.28 overnight?
Ugh.
Yeah, I noticed that too. Ruh roh.
At that rate it’ll be back down to +1.08. No worries.
I wonder what happens to Neil Newhouse, Michael Barone, George Will, etc, after the election and they are profoundly wrong?
http://andrewsullivan.thedailybeast.com/2012/11/the-reality-check-election-ctd.html
Will there be any punishment for their widely exaggerated claims of landslide victory?
No and yes. No in the sense that these people largely won’t lose their jobs or access to media outlets, yes in the sense that even fewer people will credit their punditry in the future.
I think that it’ll be spun into Sandy having changed everything at the last moment and that they had the pulse of the electorate until then. The people who credit them will swallow that and keep crediting them. The confusion and misinformation never stops. :(
To Sam Wang, great site!
Why did the meta margin go down so much?
Medians went down in a few states just based on some polls dropping off by date.
PA went from 4.5 to 3 based on a single poll. Iowa moved down to 2 (very few polls being conducted in Iowa). Virginia ticked down as well.
Just some noise.
I think we’re seeing a little bit of regression to the mean after a spectacular week for Obama. It showed up in the RAND switcher graph a few days in advance. I doubt this presages any kind of collapse.
Here’s an example of the worst kind of political “reporting”…
http://harndenblog.dailymail.co.uk/2012/11/how-mitt-romney-can-win-the-white-house-on-tuesday.html
…I did a spit-take when I got to this part:
Can so many polls be wrong? The short answer is yes. It is worth remembering that in January 2008 virtually no one in the political world believed that Hillary Clinton could win the New Hampshire primary over Obama, fresh off his Iowa victory. But win it she did.
No one in my country with any sense believes a word in the Daily Mail. That article seems to be a combination of sources from the Romney campaign and selective citing of polls. The usual rubbish, but nothing I wouldn’t expect from the Mail (which is the closest we come in the UK to a Tea Party house journal).
Wow, I didn’t know people even bothered to read the Daily Mail.
I stick to the BBC, the Guardian — and if need be the Times and the Telegraph.
@Olav – actually, the Mail has a high circulation – but I don’t know how many people read it for the ‘alternative’ health tips, the house price scares, the personal sob stories and/or the articles criticising women for working, and how many people that leaves who get their political news from it.
Yes…I know. I just had to laugh that the ’08 NH primary was the example used to justify Romneyworld’s narrative. It’s probably too late for him to choke back tears, though.
Of course it has high circulation. It has the page 3 girl.
I see that Va lost 1 and Pa lost one. Why was this? Is this what brought the MM down so much?
Wow seeing the MM plummet this morning was NOT the comforting news I needed to get through the workday. I might have ulcers by this time tomorrow.
What happened to hurt Obama’s chances so much?
RE your comments on the M-M being close to 3. … and back down to 2.28 this AM. Don’t know what contributes to it, but the M-M has been rather twitchy the past week or so …. general trend up, but at this point, absent some late breaking big margins, looks like it will finish between 2.o and 2.5, and not so close to the 3.o
The images shown in the article changed after this morning’s update.
You should probably perma-link them to the graphs you were writing about at the time!
Please try not to get jittery over the mm going down. The polls would need to plummet for Romney to get back into this race. As a matter of fact, the overall polling for Obama has been good.
Yep. There’s relatively little polling in places like Pennsylvania and Iowa.
In Pennsylvania, for instance, just the introduction of an Obama +3% sent the median down from 4.5% to 3%.
Just remember:
OH +3
WI +4
NV +5
Total EV = Obama wins.
All the rest is icing on the already baked cake.
Ding ding ding.
I think 303 EVs is right in the wheelhouse. No matter where the MM goes today.
NJF, okay. So did it take a major dive because some older polls fell off? And if so, a lot of new polls in Va and Pa were good. So still not sure about the drop. I expected some, but not a 5.6% one.
Yes, some older polks fell off and I believe that NC became wider for Romney. PA got closer because if the median. The point is to see the bigger picture and not to fret about individual polls.
I do think, though, that Ohio should be O +3.5% now?
OV is right. It should be.
You’re right…
Looking at pollster, there are 12 polls in the last 7 days with a median of Obama +3.5%.
However, notice the poll from Pulse Opinion Research listed as 10/29 – 10/29. It’s marked as “new”, despite the fact that several newer polls are not. Perhaps Sam’s script looks for this sort of “new” tag (in the actual data file, of course)? Adding this Obama +2 poll (inadvertently) would certainly shift the median to Obama +3…
Is this the reason?
Take a look at their CSV entry:
Pollster,Start Date,End Date,Release Date/Time, […]
Pulse Opinion Research/Let Freedom Ring (R),2012-10-29,2012-10-29,2012-11-04 19:55:23, […]
We can see that it was indeed just released last night, so perhaps it was mistakenly included?
I agree w/ OV. Here are all the changes from last night’s table:
IA: O+3 => O+2
VA: O+2.5 => O+1
WI: O+4.5 => O+4
PA: O+4.5 => O+3
MN: O+7 => O+7.5
NC: R+2 => R+1
Ohio looks good to me at O+3. The latest average polling date is 11/3 (PPP and Ipsos). There are 22 polls having an average polling date of 10/27 or later:
O+4 or better: 9 polls
O+3: 5 polls
O+2 or worse: 8 polls
See, I’m using 16 polls (since 10/29). That Pulse poll being the ‘last’ one included.
Pulse Opinion Research, fwiw, is Rasmussen’s robopolling outfit.
Oh, I see. Yeah, I guess it depends on whether you count today or not…
If you go to 10/29, then perhaps Grove Insight on 10/28 – 10/29 with a median date of “10/28.5” is being treated as the 29th?
That extra Obama +3 would also shift the median to +3… lol.
The model drops fractions on average polling dates, so the Grove 10/28-10/29 counts as 10/28. However the seven days is not counted back from today, but from the latest average polling date.
The twitchy-ness from update to update is just something you have to live with given Sam’s methodology. Big picture wise, it is a consequence of using a method that does a good job rejecting extreme results, but doesn’t otherwise attempt much in the way of smoothing or correction for house effects. The upside is that this method will respond more quickly to real changes in voter sentiment, and the downside is it will briefly indicate many possible changes that turn out to be just noise.
Sam even pointed that out in his previous post: if you want a smoothly varying predictor, 538 does better at that. The price being that the top-line EV number will necessarily say unrealistic things like 307.2 electoral votes.
I wonder if there is a way to predict an expected amount of twitchy-ness based on some fairly realistic assumptions about how polling will tend to flow for the key states, including the fact that the same pollster (with the same house effect) will often release new swing state polls at the same time.
My guess is you would predict that even with unchanging voter sentiment, the MM would experience a sort of random oscillation through at least around a 1-point range (so some value +/- 0.5 points). But I would love to see this done rigorously.
Matt:
That’s only unrealistic if you discount the possibility that, tomorrow, Palm Beach and Broward Counties will secede from Florida, taking 4.2 of the state’s electoral votes with them… and then voting for Obama.
You heard it here first!
On the subject of the British press, I thought it interesting that both the Economist and the Financial Times endorsed Obama. On the other hand, this weekend’s Barron’s sees “momentum” as favoring Romney. Let’s see, Barron’s is a News Corp. publication…
Note too that in the nine elections from 1980 to today, the Economist has never, ever, endorsed the incumbent party. Until now.
Great job. I was going thru 2008 results and congratulations on hitting right on targets. I was wondering when would be your last prediction (to compare the results of your models with the actual results of 2012 elections). Would it be end of day today (5th Nov) or tomorrow morning (6th Nov). Good luck!
In 2008, it was ‘handed down’ at 5:06 AM. I hope that holds true this year, I have to be at my precinct at 5:30 AM!
Holy mackerel: http://spectator.org/archives/2012/11/05/fear-and-loathing-in-ohio
If he hates Nate Silver this much I can’t imagine what he would think of Sam Wang.
Thanks, Ohio Voter.
Here is list of models and markets I am following
O 307 R 230 Nate/NYT
O 305 R 233 Princeton
O 303 R 235 UI Election Analytics
O 332 R 206 Votamatic
O 290 R 248 UVU
O 67% R 33% Intrade
O 80% R 20% Betfair
If I had accounts at Intrade, Betfair – what an easy arbitrage trade !! Unfortunately, I don’t :-)
On a more ironic note:
Remember, remember, the 5th of November
The Gunpowder Treason and plot ;
I know of no reason why Gunpowder Treason
Should ever be forgot.
Today is Guy Fawkes’ Day.
Gallup sort-of returns: http://themoderatevoice.com/166574/usa-todaygallup-poll-swing-state-voters-split-48-48-between-obama-romney-and-obama-bounce-since-october/
Your meta-margin swings like 0.7 in one day while Nate Silver’s puts Obama at almost 9-1 in Obama’s favor?
I think you’re taking some of the outlier polls too seriously.
You will lose your audience after November 6.
Try not to speak for the rest of us. We aren’t going anywhere.
You sound silly talking to a computer. The computer that runs the numbers does not “take things seriously” or have any other emotions. It’s just pure number crunching. The computer script automatically takes the median of the last 7 days worth of polls.
If a favorable poll drops off the end, or an unfavorable poll appears, then the numbers will change. It’s as simple as that.
If you could read, Tony, you would notice that Sam’s method minimizes outliers by using the median… and still gives Obama a 99.8% chance of victory. If you could do math, you would see that is 499:1 in Obama’s favor.
Nate gives Obama 307.2 electoral votes, Sam 303. The EC is of course unlikely to deliver a fractional result. So exactly how is Sam going to be embarrassed come November 6?
The Florida ballot has pages and pages of constitutional amendments – mainly as a result of this, the lines are S … L … O … W and in crowded, Dem-leaning counties like Broward are backed up as much as 5 hours for early voting. I expect they’re not nearly as bad in the more R-leaning smaller counties. Expect this to depress turnout in an R-leaning direction in FL. Let’s see what Tuesday looks like.
FYI, I early voted on Sat, Oct 27 in Miami-Dade. Three-hour wait then, the first day of early voting. A lot of people showed up and left, saying they’d come back later for early voting or wait until Tuesday.
Rick, do you think there’s a strong likelihood of voter suppression tomorrow?
Do the rules state that the precincts obligated to let everyone in line vote?
Or is there a chance they may close their doors, run out of ballots, not have enough voting machines or printers or staff to service those in line…
That really sucks, but… does your state not send out voter guides to registered voters?
Here in CA, we receive a sample ballot and voter guide a few weeks in advance. Why would people stand around on Election Day reading about these things for the first time? It’s so much more relaxing to learn about them in advance by simply reading the voter guide…
I’m not sure just how common it is to send out voter guides to registered voters. I know that we do not get one in Virginia. Additionally, when we go to the SBE website, it tells us a little something about who is running for what office, but the other ballot issues are a lot harder to ferret out. What good is telling us there are two consititutional amendments being proposed, if we can’t tell what those are?
Rick in Miami Maddow did a good piece on that last night,She was as mad as i have ever seen her as we all should be..What is this a 3rd world country?
I didn’t get any voter guides on the amendments. There was a pro-Romney group giving out “how to vote on the amendments” on site for voting, but nothing beforehand.
I would not at all be surprised to hear conjecture that Obama will lose many thousands of votes from people in heavily-populated, Dem-leaning counties who show up, see the lines or wait a little while, then leave instead of voting. There are some R-leaning spots here too with long waits (wealthy Cuban-Americans in Coral Gables, for example), but the impact is definitely going to introduce a bias for Romney. It’ll all be conjecture though – we’ll never know for sure without post-election polling to see how many showed up to vote but didn’t, and who they intended to vote for. That would be a *very* interesting poll if FL ends up as close as it currently looks like it will. Hopefully, unlike 2000, we won’t make the difference.
Here’s the ballot we have in my corner of South Florida. Scroll to the bottom to see why the ballot is page after page of legalese. All that legal language is fully there on the ballot, in three languages (English, Spanish, Creole):
http://vote-fl.org/Ballot.aspx?State=FL&Election=FL20121106GA&Congress=027&StateSenate=040&StateHouse=114&County=086
Dr. Wang, could you ask Andrew to change the vertical 270 cutoff line on the distribution of all possible EV outcomes to some color other than red? That color has a political interpretation, it’s very tall, and it’s next to Romney’s name, which almost gave me a coronary a second ago. :-P
Well, only one more day to see how well the polls did and, thus, your model. Good luck! And thanks for the entertaining entries. May the storm recovery be smooth for you all.
Obama 303 I like the feel of it!
303 is the same number Truman and Kennedy got. Larry Sabato says that’s a lucky number for Dems.
Another pro heard from.
Mark Blumenthal: “When combined, these estimates add up to Obama winning 303 electoral votes to 235 for Romney. With the historical patterns of polling error factored in, the Pollster model currently gives Obama an 90 percent probability of winning the election.”
Sounds like PEC?
Linzer has a new article that shows that pollsters may be herding by looking at other pollsters results and weighting their own to match.
http://votamatic.org/pollsters-may-be-herding/
Any thoughts about this? Perhaps O’s Ohio numbers have become a little inflated over the last few days, and perhaps should be around 2-2.5%?
Appreciate all of the effort/work Sam,I have also kept a keen eye on Kos and was happy to see you there…With that said the thing that matters the most is that everyone who wants to vote gets to vote and gets that vote Counted right.
All of this news from Florida and Ohio about making voting harder is just unreal and Sad..So lets all hope for a happy ending within 48 hours!
thanks |Sam
so will it be 303 or 332? Not too sure about CO and VA either…
PPP has a new poll for CO with Obama up 6.
VA will be close but looks blue.
Sam,
I am analysing how poll aggregators differ in the way they aggregate the polls. I found out that your state by state numbers are quite “lumpy”. I suppose that, for any given state, you take the five most recent polls and take their mean. It can make your margin calculation for that state jump up or down by a full 1% and even more when a new poll replaces the “oldest”one. And that even if that particular poll hardly change the average at all.
Because of this use of the mean on a limited number of polls, you rate high on the number of “outliers”, although in an unbiased fashion (RCP rates as high as you do, but their outliers are almost always red!).
Let me suggest the following remedy which, IMHO, improves things while conflicting in no way with your general philosophy:
1. Take more than the five last polls, I suggest seven or eight is a good balance between responsiveness and curve smoothing.
2. From these seven or eight polls, don’t count the most republican as well as the most democrat leaning ones, and then AVERAGE the five or six remaining ones.
Just suggestion…
@Francois – Prof. Wang’s calculation takes the *median* of recent polls, not the mean. This significantly reduces the impact of outliers.
Methodology is detailed here: http://election.princeton.edu/methods/
Any prediction can only be as good as the underlying data on which it is based. Given the inherent uncertainty in the polls, which are not even conducted on the actual voting population but rather some combination of people who will vote and some who will not vote, how can one claim to be 99.9% certain of a given outcome?
@JoshuaE73 – by aggregating multiple polls per state, the effective margin of error is reduced as the effective sample size is greatly increased.
Secondly, if the question is “who will win” rather than “what percent of the vote will each candidate receive,” the confidence level increases substantially. For this latter question, if you perform enough polls and keep getting the same result, eventually you do reach a very high probability of that outcome occurring.
Thanks JamesInCA. Your “who will win” argument is somewhat confidence-building. And I agree more data is generally better, although one could argue that aggregating polls of varying accuracy results in an aggregate that is of average accuracy, especially in the presence of systematic error. Here are a few examples of the systematic error I’m talking about: 1. Polls collect data from non-voters (a segment that likely leans toward Democrats), 2) Polls often exclude voters who don’t use the internet (people who are more likely to be conservative and traditional), and 3) polls exclude voters who don’t answer due to privacy concerns… a strongly conservative trait. I’m just not sure I believe the data on which the prediction is based.
Thanks for all your hard work! Very interesting.