A little more than two months ago, I ran through a scenario that predicted what might happen if half of the Hillary Clinton Democrats who said they were going to vote for John McCain instead gravitated back to Barack Obama. The prediction was that the so-called "unity bounce" would be worth 4-5 points to Obama in the popular vote, bringing him northward of 320 electoral votes and making him roughly a 3 to 1 favorite to win the election.
That is almost exactly where we have Barack Obama's numbers after a series of new polls from Quinnipiac. In Pennsylvania, Obama leads by 12 points -- up from 6 last month. His Ohio lead is 6 points -- he had trailed McCain by 4 points before. And then there is Florida, where Quinnipiac has Obama ahead by 4 points. Barack Obama has never before led a Florida poll -- not against John McCain, nor against Hillary Clinton -- so this is something of a watershed moment.
If Florida is in play, then John McCain's defense is completely broken; it was the one traditional swing state that always had looked off-limits to Obama. More frustratingly for McCain, he had spent the better part of three days in Florida earlier this month, hoping to raise doubts about Obama among Jewish voters. Although Quinnipiac does not break out the Jewish vote, Obama holds a 61-31 lead in Southeast Florida, where most of the state's Jewish population is concentrated.
Obama's surprisingly strong lead in Ohio isn't any better news for McCain. As recently as a week ago, McCain's strategy looked pretty simple: target Ohio and Michigan, and hope to win one if not both. But now, Ohio looks tough for him, and even if McCain can steal Michigan, Obama has so many other places he can pick up electoral votes -- Virginia, the Mountain West, Iowa, Missouri and now possibly Florida -- that McCain still might have trouble winning a close election.
Obama's lead nationally is still relatively small -- we have it at somewhere between 4 and 5 points -- but looks to be an unusually robust one in terms of the Electoral College. Rick Davis can probably give these numbers about another week or two to settle down before he has to start thinking about some fundamental changes in his candidate's messaging.
6.18.2008
Today's Polls, 6/18
@ 8:35 AM
Subscribe to:
Post Comments (Atom)
91 comments
Why isn't the EV graph a bell curve anymore? Your recent two big changes have made me very nervous. It gave Obama a huge, sudden boost and it seems to have destroyed the curve. I'm having trouble believing the rosiness.
It's still a bell curve. It's just that it's shifted considerably to the right on the graph, as Obama's now projected to get more electoral votes than he was before.
It can't be a perfect bellcurve unless it's virtually centered.
It's now a lopsided bellcurve - which is what it SHOULD look like if you believe Obama has a strong advantage.
If you look at the numbers, it's about 50% on each side of the peak - just that the peak is now around 360-370.
Obama will win in a landslide in November.
The Obama army is actually in the states for registered new voters in the swing states.
Easy win.
I can see the peak has shifted. But I'm also looking at a big bump around 450-510 and a big deficit around 90-150.
Maybe this is just how the combinatorics of how the EVs can add up, though.
Here's the thing: I've been burned before. I would prefer a system that I can be happily surprised was too low over a system that I will be disappointed was too high.
Curious that more likely for McCain to lose popular vote and win electoral vote than same for Obama. Seems counterintuitive - more likely that Obama piles up high margins in safe states, say NY, but narrowly loses say MI/OH/PA/FL
I'm guessing that the big bumps are scenarios where the wheels really start falling off for the GOP; something like Obama pulling a surprise upset in Texas. That'd be a 34-EV shift and might produce some peaks to the right of the main curve.
I'm more confused by the big spike on the far right of the graph. Dunno what that's about.
Hey Anonymous, NEVER the GOP will win in PA.
Every election, the GOP hope win here...lol.
Every poll Obama has a command lead in PA.
With the new registration, PA is strong DEM.
What are the values on the y axis of the EV distribution? Take for example the spike on the far right where Obama wins absolutely everything (except Utah?) - how many simulations out of 10,000 have this result?
can Texas go blue ?
highly unlikely but...??
PA will go for Obama
A bell curve does not have to be averaged at 50% to be a bell curve. (Only artificially derived standards such as IQ have that kind of distribution) It appears to be a standard gaussian distribution, with the mean result indicating that Obama will obtain 360-390.
I wonder, if a result like this actually happened, can they call the election before California closes?
Did the Pacific states disappear from the left margin for anyone else besides me?
Sure, a Texas victory is very unlikely, but the model shows it happening 18% of the time. And the bumps on the right side of the graph seem to be pretty evenly spaced out at about a 30-EV margin. I'd bet those are scenarios where Obama wins Texas, or does something equally improbable, like cutting a wide swath through the South.
""Curious that more likely for McCain to lose popular vote and win electoral vote than same for Obama. Seems counterintuitive - more likely that Obama piles up high margins in safe states""
I agree, especially when combined with a good chance that he'll pile up larger margins among blacks in southern states, but likely not enough to win many, if any of the states themselves.
"Here's the thing: I've been burned before. I would prefer a system that I can be happily surprised was too low over a system that I will be disappointed was too high."
72-90.
"I wonder, if a result like this actually happened, can they call the election before California closes?"
That's an interesting thought. Here's the currently projected vote totals by time zone:
Eastern: Obama 194, McCain 43
Central: Obama 70, McCain 90
Mountain: Obama 14, McCain 38
Pacific: Obama 78, McCain 4
Alaska/Hawaii: Obama 4, McCain 3
Where a state occupies two time zones, I put it in the westernmost column.
If Obama performs this well (which I have to admit would be a little surprising, but it seems within the realm of possibility, given the fundamentals of the election) then yes, he's at 270 EV before we get to the west coast and California.
Du, you poor lamb! Have some faith in your guy winning!
"If Florida is in play, then John McCain's defense is completely broken..."
To extend the sports metaphor, it might be time for the Belichick solution... start videotaping the opponent's moves. (Or maybe we swing back to politics and just call it a Nixon move.)
I'll just use a video game metaphor to describe today's polls, if it's okay:
C-C-C-C-COMBO BREAKER!
Um... don't look now, but the state-by-state chart on the left is missing the "Pacific" states.
I think I can sum today's update up in two words, one of which stutters:
C-C-C-C-COMBO BREAKER!
How does one poll result prove that Obama is suddenly leading in all the swing states?
Are the results so irrefutable as to warrant changing electoral vote projections? Wouldn't it make more sense to wait until a few more polls came out before making such sweeping claims about the electoral map?
That EV distribution looks pretty sweet -- and also pretty "normal" in shape to me.
But it's only June. There's a campaign going on.
I've lost all the blog entries before 6/17. Does anyone else have this problem? Is there a workaround?
Thanks
There's however one interesting thing concerning a "unity bounce" about the new polls:
In FL, 19% of declared clinton backers in the primary season say they'll vote mccain. In OH and PA, these numbers are even 25% and 24%. Nate's previous post (that he refers to) says nationwide 28% defection rate, and then he calculates what will happen if this defection rate reduces to 14%.
So at first view, those numbers tell a quite limited unity bounce, since they pretend a rather limited decrease in the defection rate. However, it might be that asking retrospecitvely now what people did in the primary might not be the right thing to do, so the quinnipac numbers on defection rates might not be relevant. moreover, recent polling in other states quite strongly supports the unity bounce hypothesis.
The unity bounce isn't as strong as the one you predicted in your earlier post. There is some unity bounce, since he is getting about 80% of the Dem vote, but there's clearly room for improvement. In FL, OH, and PA, the percentage of Clinton supporters who are backing McCain stands at 19%, 25%, and 24%, respectively. That's not too different from the 28% of Clinton supporters during the primaries who said that they would vote for McCain if she lost.
The thing that surprised me most is Obama's consolidation of black support. We're all expecting him to win it by at least 90%, but so many recent polls have shown him underperforming that particular demographic.
Actually, I think the proper video game metaphor is Obama saying to McCain (upon seeing the Ohio and Florida polls):
"ALL YOUR BASE ARE BELONG TO US!"
Poblano - Please make an archive of your map graphic. It would be wonderful to see the country turning blue. An animation would be even better. Pretty please?
Anon @8:22,
Obama already was leading the other major swing states. The Florida shift just moves the last one into his column, which really breaks any reasonable chance for McCain to win by just getting lucky and carrying the right states. Now he needs a big nationwide polling swing just to make a victory possible.
Of course, big nationwide polling swings happen all the time - which is why McCain is at 25% and not 2%.
"Rick Davis can probably give these numbers about another week or two to settle down before he has to start thinking about some fundamental changes in his candidate's messaging."
Agreed, unless their plan is to rely on 527 attack ads and/or an "October surprise" to swing votes at the end. That said, McCain simply can't bother playing defense in places like Florida or Missouri. The reason is simple - if he's losing those states, then he's pretty much dead in the water anyway. He basically has to bank on his overall numbers improving, and stick to playing offense in Michigan and Ohio.
Obama, on the other hand, has the money and volunteers to play offense in a lot of places at once, so that if his numbers DO get worse, he isn't stuck with all his eggs in one basket like Kerry was.
Before video games talked, the message would be more simple:
KABOOM!!!
EarBucket,
it's not really right to project by time zone, because polls don't close at the same local time in every state.
What's really needed is to break things out by poll closing times, which can be found here.
Here are the vote totals by poll closing time (given in Eastern time), again assigning each state to the latest time at which any of its polls close:
7:00: Obama 27, McCain 31
7:30: Obama 20, McCain 5
8:00: Obama 138, McCain 39
8:30: Obama 0, McCain 21
9:00: Obama 86, McCain 64
10:00: Obama 12, McCain 8
11:00: Obama 77, McCain 7
1:00: Obama 0, McCain 3
The total: Obama 360, McCain 178. This comes out more in Obama's favor than the "electoral vote" prediction, basically because right now there are more states leaning Obama than leaning McCain.
Of those, Obama gets 283, and McCain 168, before ten o'clock.
So it's possible that this could be decided before the polls close in California at eleven. But that's not a foregone conclusion, since 283 is just one big state away from 270; Michigan, for example, closing at 9, might not be settled by 11. (I'm not predicting anything here; it's just a large, currently Obama-leaning swing state.)
But what if Obama gets to, say, 230 in "called" states by, say, 10:30? Then it'll be interesting to watch the networks try to fill half an hour before the 11:00 closings pretending that McCain still could mount some miraculous comeback in the interests of not appearing to call California before it closes.
While the FL poll is certainly great news (unless you like McCain), I'm somewhat skeptical of how fast it turned.
Looking at the list, it's easy to see why... the new Q poll has a 1.73 weight and the regression is at its standard 0.63, then the next most recent are 0.50 and 0.38.
One thing I picked up on during the primary is that one poll with an unexpected result is more or less meaningless, regardless whence it comes -- you need at least two to confirm a trend.
I would think that the modeling should exhibit a similar sort of skepticism. If a new state poll varies from the previous average (probably excluding the regression) by a significant amount -- 10 points or more, say, and there's no roughly corroborating poll within the last month, it should be down-weighted.
I think some people are reading too much into the "bumps". With only ten thousand simulations one should expect to see them even if the underlying distribution is smooth. Let's say the probability of getting a certain number N of electoral votes is .0036, so on average it should turn up thirty-six times in ten thousand simulations. If we repeatedly ran sets of ten thousand simulations, the distribution of the number of simulations showing N electoral votes would be very close to Poisson with mean 36. The standard deviation of Poisson(36) is sqrt(36), or 6 -- which is pretty large compared to 36.
I'm not saying the bumps aren't real. But I'm saying that they might not be.
Isabel: that's a good summary. Keep in mind also that polls in the Florida panhandle close an hour later than those in the peninsula because those counties are in the Central Time Zone. I recall that lesson from the 2000 election. So effectively, Florida as a state is in the Central Time Zone.
Juris,
you're right. The polls at Florida close at 7 local time (ET or CT), so Florida is included at 8:00 (ET) in my analysis.
There are some other surprises here, too, in which states end up in which bins -- for example, North Dakota for the purposes of this analysis ends up being a "west coast" state. I say this because people probably think the states closing at 11 ET are West Coast states closing at 8 PT -- and most of them are -- but North Dakota closes at 9 local time, and the southwestern part of the state is in the Mountain time zone. (The situation is similar to Florida, except much smaller.)
What is the unity bounce under the old methodology?
Nate,
After your "We know less than we think" adjustments increased the variance of the results, perhaps you should increase the simulation size to smooth the distribution out a bit more. It won't really change the EV/win/vote numbers, but it will make the distribution itself a bit more aesthetically appealing.
Larger number of simulations? Methinks Nate needs more RAM and a faster CPU.
There's still going to be some 'rounding error' in what he can show in his graphics.
So Obama has never had a lead in a Florida poll before, but now just one comes out where he does, and he takes the lead in the projection? I'm sceptical of this.
This Florida trend makes it all the more likely that McCain taps Charlie Crist as his VP candidate. Not that Crist isn't making a blatant play for the slot as yesterday, one day after McCain's "let the states decide if they want to drill" pronouncement, he did a 180 on offshore Florida drilling, now coming down in favor of it.
Rorgg: FL polls are weighted as they are because all the other polls are a month or more old, when Clinton was still portraying Obama as the antithesis of hard-working white voters. The initial change to the 538 regression showed a FL win even though all the polls were showing him behind, but Nate refined his methodology since then. What I'm trying to say, is that Florida's move didn't happen suddenly, it happened over the last month. We just haven't had polling the to see it.
How can it be, that Florida suddenly jumps to 59% for Obama although there is just one single poll out giving him a lead?
You say your site is not about where the race is now but about predicting it, but it seems you have more volatility in your predictions than a traditional polling average. How come?
A larger number of simulations might help. But to smooth the bumps by a factor of two you have to do four times as many simulations, so it may not be worth the trouble.
For what it's worth, the "toy" election model where each state votes independently of all others and has a 50% chance of going to each candidate shows no visible bumps when you use the theory to find the probability of each number of electoral votes. But when a simulation is done of 10,000 such "elections" it comes out pretty "spiky" -- maybe a bit less bumpy than the graphs here, but it's hard to tell. I'm not sure how this carries over to the model here, though.
On Indiana, the same caveat as was just offered for FL: all the available polls are quite old, and as you can see by their weights they are fading in relevance to the model because of that. Conversely, the regression model is playing a relatively heavy role for now. What's needed, of course, is new polling for Indiana, which could move it onto the red side.
rorgg, we HAVE more than one poll - that's what the new trend analysis thing is all about. We have polls for other states with similar demographics, and national polling. All these things hint towards a strong bump in FL, so when a poll comes out in FL that shows a strong bump, the model believes it because it was already predisposed to do so based on the other evidence.
If the overall trends were showing Obama losing ground in states demographically similar to Florida, and nationally, and this poll came out, the model would be skeptical about it.
There is more to be sceptical: All three extremely positive polls are from one pollster, so I really wonder how your prediction can make such a jump after three state polls from one polling firm.
They did warn us that legalizing gay marriage would be the end of the world. I don't see why everyone's so surprised that the Pacific states have been erased from the sidebar. Probably crumbling into the ocean as I type...
A historical note:
Back when Reagan won his first election (1980) we were living in Los Angleles. We came home from work and prepared to go vote. I put the TV on to watch the evening news and they were already proclaiming Reagan had won the election. That was probably 6-7pm PST. We never went on to vote.
So it is possible Obama could win the election before the west coast gets to vote! Its happened before.
To have an idea of popular opinion, have a look at comments that people are leaving at at abc news website. For example:
http://abcnews.go.com/Politics/TheNote/comments?type=story&id=3105288
I found it astounding how many there are saying they will vote mccain... Is abc site an anamoly? i didnt think it was supposed to be a "conservative" site. At any rate, the opinions expressed there dont seem to jibe with the polls... it seems that majority of comments on abc website are anti-obama.
To extend the sports metaphor, it might be time for the Belichick solution... start videotaping the opponent's moves. (Or maybe we swing back to politics and just call it a Nixon move.)
But strangely, Lieberman wasn't on Plouffe's strategy call with Democratic senators, according to TPM!
Concerned,
Comments left on a news site are NOT representative of popular opinion.
It shows little in the end, except for the fact that the McCain voters happen to be more vocal on that site. And that's not even getting into the fact that it could all be coming from one or two people for all you know.
To add to the comments on FL and IN for those who are new to this site.
Nate's state projections are produced by doing a regression using state polls (adjusted, under the new methodology, for national polling trends). He treats his "538 Regression" estimate (which is not based on polls but instead on historical and demographic voting patterns) as if it were one state poll when he produces his state projections. The weighting of the individual state polls is adjusted in the regression according to the sample size, recency of the poll, and reliability of the pollster (calculation of the weights has been explained in his FAQ and his articles on the site).
What this means is that if there are few state polls, or if they are not recent, or if they are done by less reliable pollsters or with small samples, then the "538 regression" dominates the state projection. In addition, the adjustment for national polling trends in effect updates each state's polls, but would have less impact if there were more, and more recent state polls.
The remedy for relying on the 538 regression and national trend adjustments is more polling within each state. As the election season progresses, we can expect the density of state polling to increase. In addition, the latest state estimate will be less likely to be affected by a single very recent poll by a single pollster. In the meantime, in the last adjustment to his estimates, Nate dampened the effect of the "last poll."
To the many skeptics,
Firstly, a poll with almost 1500 respondents is always going to be taken more seriously than most.
Secondly, you are missing the point of what's happening here. This poll did take Florida over the 50% threshold, it's true. But the popular vote numbers, and even the expected electoral vote totals, have not drastically changed from the last few days. What has happened is we've reached a tipping point in winning percentage. It's an effect of the electoral college system. Before today, it was a lot more likely for McCain to win just by getting "lucky" in a couple swing states. But the shift of the last key swing state over the edge toward blue has drastically reduced the probability of McCain winning "swing state bingo". Now he is basically riding on the chance (which the model does consider) that there will be a big national move toward McCain before the election.
New PPP Poll in Virginia.
Obama lead by 2 points.
47-45
Great numbers.
With the hard work from the liberal groups and the Obama army for registered new voters, Virginia will be blue in November.
The recent state polling is merely confirming what we already know from the national polling that shows Obama with about a 4% lead. As Nate's graph shows, a 4% lead translates to roughly between 315-330 electoral votes, depending on the states. His current projection of 334 is proportionate to this.
I don't think McCain is particularly concerned about Obama's lead however. His strategy has ALWAYS been to have Republican 527's launch an endless series of slanderous attacks on Obama while he pretends it has nothing to do with him, aided and abetted by the MSM. The Rev. Wright "scandal" was just the trial balloon for this.
He fully intends to "Swift-Boat" Obama just like Bush did to Kerry. This campaign is going to get a LOT uglier and a lot closer. The intensity of the vicious attacks is going to startle millions. The real reason for the attacks is to give voters cover for racist doubts about Obama so they can conclude he's "just too inexperienced."
I.E. Electing a black man is "too risky."
That's McCain's entire strategy.
This election has shades of 1980 in reverse. That too was a nip-and-tuck election in what should have been automatically awful for the incumbent party from the beginning; the dam did not finally break for Reagan until the last week of the campaign, and even then he carried state after state after state by just one or two percentage points. So if you want the Democrats to have our own "1980" and stop the Republicans crowing about theirs, remember the lessons of that year and don't get complacent -- get to work for Obama.
And the memes are like 1980 too; Democrats denigrated Reagan for being inexperienced -- Republicans now do the same to Obama. It's important to counter that by showing undecided voters Obama's qualifications and skills and judgment for the job; his remarkably active legislative record in the US Senate and in Illinois, his terrific success rate with getting meaningful legislation passed, and the administrative and organizational ability he has shown both in his current campaign and his past jobs.
Luuucy! You have some 'splainin' to do!
OK, I get the new system (I think), but today my eye was drawn toward the newly-pink West Virginia, a lightly polled state, and it's clear that the weighted average from the raw poll numbers (McCain +4.7) is about 3 points different from the projection (McCain +1.7). So it occurred to me that it would be useful, in addition to the raw poll numbers, to have one more column for "Demographically Adjusted National Trend Since Poll" (you might want to abbreviate that) to give the casual reader a better idea of how these projections are done, and why they're more accurate than any snapshot poll. The weighted mean of DANTSP+Margin should equal the projection.
seandalai said...
"What are the values on the y axis of the EV distribution? Take for example the spike on the far right where Obama wins absolutely everything (except Utah?) - how many simulations out of 10,000 have this result?"
Nate has said he runs 10,000 scenarios. If you round down to 500 individual bars on the x-axis (from 538 minus a few that have no occurrences) and eyeball estimate the average height of each bar at 2 lines, then each line would equal exactly 10 events (500 * 20). So, the line on the far right end looks to be about 2 of 10,000.
Andy said...
"State Colors: I think the colors of the states in the left hand column of the blog page need to be updated. "
State colors are based on the 2004 election results. Personally, I like seeing how Obama is doing relative to Kerry. Changing the colors to match current projections seems kind of pointless, since the win percentage for the state is listed right there as well.
Joachim Rang said...
"You say your site is not about where the race is now but about predicting it, but it seems you have more volatility in your predictions than a traditional polling average. How come?"
Why do you conflate amount of volatility and accuracy of predictions? All McCain-Obama polling in Florida is somewhat old. Thus, the recent Quinnipiac poll has a lot of weight and changes the result significantly. Since the site is all about trying to make the most accurate predictions, it is entirely appropriate to make big changes when warranted by new data.
As a practical matter, it would be super-helpful to return to this theme of whether the election could be called before the polls close on the West Coast closer to the actual day, so that those who are trying to get Dems elected downticket in those states can plan for it. Thank God I live in Oregon, where we vote by mail. But I imagine in California, Washington, Alaska and Hawaii there will be a lot of consternation if people just don't go to the polls because it seems to be all over, leaving downticket candidates and ballot measures high and dry.
Juris said...
He treats his "538 Regression" estimate (which is not based on polls but instead on historical and demographic voting patterns) . . . ."
I think this is not entirely correct, if I understand how Nate is doing it. I believe the 538 Regression is calculated from the state polls, not historical voting patterns. In other words, Nate looks at all the state polls and then, based on the demographics in those states, calculates the relative importance of particular demographic factors. He then calculate the 538 Regression for a state based on those demographics. The interesting part is that the 538 Regression changes as new state polls are published. Maybe Nate can clarify.
I encourage everyone upset or disconcerted about the "volatility" or rapid change in the Obama win % to wait a week or two. If the projection is accurate, the polls will stabilize and the win% will seem to make more sense just as a polling average. If the bounce falters and the polls fall, the adjustments should reverse themselves also. Obama supporters, sit tight before breaking out the champagne.
Anonymous said...
"As a practical matter, it would be super-helpful to return to this theme of whether the election could be called before the polls close on the West Coast closer to the actual day, so that those who are trying to get Dems elected downticket in those states can plan for it."
I've always been curious about this issue: if an election looks to be a blowout based on polling or because it's called early, who benefits? Do more of the apparent winner's voters stay home, more of the apparent loser's, or proportionately equal numbers? In other words, does it change either candidate's share of the vote or just depress turnout without making any difference in the margin of victory?
Do not be concerned about the crazy anti-Obama comments left on news blogs and insane sites like NoQuarterUSA.NET
These people are not representative of average Americans. They are nut cases who have lost all right to be taken seriously with their insane anti-Obama BS.
NoQuarterUSA.NET is the best example of just absolute out to lunch nutcases I've ever seen. Those folks are "out there." They are less than 1% of the electorate and are not nearly as influential or as important as they like to think. They live in fantasy land.
If you think the No Quarter crowd is nutzo, you should take a look at the Hillaryis44 bunch sometime. Lots of overlap between them, but I'd say the latter is even crazier than the former.
BTW, for those skeptical of Nate's Florida projection, there's another poll out today for Florida: Obama 49 - McCain 44. It is, however, from Nate's good friend Bennett, so take it for whatever it's worth. ARG is also out with a NH poll showing more good news: Obama 51 McCain 39. http://www.americanresearchgroup.com/
Imperfections in the bell curve are due to the fact that there's an imposed floor and ceiling on the curve (i.e. you can't go lower than 0, higher than 538). That's why you get the bump at the extremes, because all the results that would show either candidate stronger than just being able to eke out that 50th state get clumped at the end. Plateaus and the sort are probably due to the fact that the data is not continuous (i.e. states are won in increments of 3 or more EVs, not by fractional number amounts like would be seen with the popular vote). Thus, you either win a state or lose it, and the EV outcome shifts in those chunks.
PPP has Obama up 2 in virginia. SUSA has him up 9 in Wisconsin.
Another Mike: You're right. My statement about the "538 regression" not using polling applies to the variables on the right side of the equation (the predictor variables -- which are demographic and historic vote outcome variables), not the dependent variable in the regression equation. If, as you note, the 538 regression changes over time, then I think the dependent variable in the 538 regression must be the (adjusted) state polls. The regression coefficients are then used to derive the 538 regression prediction for each state, based on the particular configuration of values on the predictor variables in that state.
If somebody sees this differently, please chime in. The FAQ still refers to Clinton and Obama and hasn't been updated; even there, Nate did not explicitly mention what the dependent variable is nor how it would be updated.
Another Mike said...
Anonymous said...
"As a practical matter, it would be super-helpful to return to this theme of whether the election could be called before the polls close on the West Coast closer to the actual day, so that those who are trying to get Dems elected downticket in those states can plan for it."
I've always been curious about this issue: if an election looks to be a blowout based on polling or because it's called early, who benefits? Do more of the apparent winner's voters stay home, more of the apparent loser's, or proportionately equal numbers? In other words, does it change either candidate's share of the vote or just depress turnout without making any difference in the margin of victory?
I have heard, though never with much evidence, that the early calling of the 1980 election (combined with Carter's early concession) lead to Dems sitting home and the GOP picking up a lot of the senate seats out West. Would they have gotten them in a clsoe election? Remains to be seen?
But I tend to believe that people like to vote for the winner, so if McCain falls early, it'll help dems not hurt them.
Thanks for the clarification on the state colors, Another Mike. Makes sense now.
Interesting. A lot of people claimed they'd vote for McCain if Clinton weren't the nominee, but 1/2 of them are voting Democrat. It's probably similar to the polls of the Obama supporters claiming they won't choose Obama any more if Clinton is VP. They'll mostly still choose Obama if it is Obama-Clinton.
Electoral vote distribution, and the graph.
That spike at the right end is essentially the right-hand "tail" all bunched up in the case of a popvote swing to something like 10% or more. I get that.
But one effect that's having is that the mean EV, 333.9, is now significantly different from the median and mode EV's, which are about 360 and 390 respectively. In other words, since the distribution is no longer normal, the mean is no longer the best measure of central tendancy.
Therefore, I'd like to suggest that you begin to use both mean and median, or better yet, median only, as the central tendancy measure in the piecharts, and drop the mean.
Anon & Another Mike - as to calling the election before polls close on the West Coast. My recollection is that after the dust up in 1980 over early calls of states for Reagan, Congress passed legislation to prohibit calling any states until the polls close in all the West Coast states. So unless you find something current from an authoritative source that says otherwise, don't worry about projected winners in East Coast states affecting the voting behavior of voters on the West Coast.
Juris & Another Mike - as to the 538 regression. I'm another lay person like the two of you, but my understanding is different.
What's in the 538 regression as independent variables?
The factors listed in the FAQ, with a few adjustments. Nate explained in a post at some point which of the factors he includes now as independent variables, but that explanation has not yet made it into a revised FAQ explanation.
What's the 538 dependent variable?
I believe the dependent variable in the regression equation is just the percentage margin for Obama (+values showing as +Obama, - values showing as +McCain).
How is the national trend (plotted on the Supertracker graph) factored in?
If I understood Nate's explanation correctly, he computes a time adjustment factor (or factors) for each state by comparing the changes in voter preference in polls by the same pollster in that state.
Then he compares the state changes over time to the national trend changes over the same time period or periods as the changes computed for that state.
(That is, the Supertracker graph values for the dates when the two polls were taken in that state.)
I'm not sure that I understand how Nate uses the time lags to adjust in each state for the national trends. My brain was hurting too much by the time I got that far! But I do recall that Nate introduced the individual state adjustments to respond to comments on an earlier model: many of those comments asked him to find a rigorous way to adjust for the national trend by a unique per cent in each state rather than adjusting every state's figures by the same amount.
Overall, I think Nate computes a unique time/change factor for each state; then adjusts the margins in each poll in a state by the time lag factor for that state (also using the weights for each poll that are shown in the Poll Detail on the right of the webpage); then averages the poll margins (as adjusted for the time/change factor(s) for that state) with the result of the 538 regression. The average so computed becomes the margin shown as the "Projection" as the last colored line for each state.
Of the independent variables for the 538 regression, only two of the "political" factors appear to have any potential for fluctuating:
$_Obama - fundraising in the state compared to Kerry may change to a larger or smaller ratio each time that there is a campaign $ reporting date for the FEC. Same for $_McCain - comparing his fundraising to Bush's.
The values for each of the other independent variables appear to be defined by 2004 exit poll numbers, 2000 census data, or perhaps some updated census estimate.
Hope I've (1) got the explanation basically right and (2) explained the concepts clearly enough for all or most of the lay readers to understand. I'd guess one of the statistical wonks will correct me if I'm wrong.
HHliberal wrote: "What's the 538 dependent variable? I believe the dependent variable in the regression equation is just the percentage margin for Obama (+values showing as +Obama, - values showing as +McCain)."
Yes, that's pretty obvious, but what's not clear is what his unit of analysis is and what data he uses for this purpose? All state polls, aggregated (combined, weighted) state polls, the latest state polls, or what?
Any analysis that turns Florida blue based on a single poll and declares McCain is doomed based on this "watershed" event is laughably partisan. If the analyst had combined his statistical skills with a passion for McCain instead of Obama, I have no doubt this site would be showing Florida in solid red and McCain with 300 EV.
To anonymous at 10:53, it's bovious that you either didn't read the methodological explanations or, if you did so, failed to understand them.
The analysis did not declare McCain "doomed." (Although some posters may have) The model still shows a 25% chance for him to win. It is also clear that changes in public attitudes could swing things back in his direction, and this is made clear in the the explanations. As Bob Dylan sang, "Don't criticize what you can't understand."
With respect to the issue of the impact of early broadcast of projections, I have the following anecdotal information, based on the 1980 election. I was managing a campaign in a western state, and the projections (and Carter's concession) came before the closing of the polls. Our own poll, taken about 3 weeks before the election, projected a 60-40 margin for the incumbent (my candidate). In the then leading public poll, taken roughly 10 days before the election, we had a 57-43lead. The absentee ballots, when tablulated, had us down 53-47, and the average gap between absentee and regular balloting was about 10%. In the end, we lost 51-49. A quick and dirty post-election analysis showed a significant correlation between change in turn-out from four years earlier and the the "republicanness" of precincts. A Republican friend of mine, who upset an incumbent Democratic State Senator by 500 votes, told me that he estimated, by looking at turnout, that the early projections cost his opponent almost 700 votes.
BTW, there is no law prohibiting projections before all polls have closed - it was talked about but would clearly have been unconstitutional. What the networkds do is not project a state until all the polls have lcosed in that state. (A productof the 2000 debacle in Florida).
In the absence of a Democratic landslide, it is likely (but not certain) that it the vote count will not permit a projection until the polls in California are about to close. Look how long it took to count the votes in Virginia last year in the Webb Senate race.
If Obama is up enough in Virginia that the state can be called relatively early, then a "premature" projection won't make any difference.
Here's the thing: I've been burned before.
Welcome to life.
I would prefer a system that I can be happily surprised was too low over a system that I will be disappointed was too high.
How bizarre. What I would prefer is the best, most accurate information available, but you want the data to be shaped according to the psychology of your expectations. Have you ever considered the possibility that you're rather self-centered and a bit dim?
What the networkds do is not project a state until all the polls have lcosed in that state. (A productof the 2000 debacle in Florida).
What's their position on a cousin of a candidate making the call?
Any analysis that turns Florida blue based on a single poll and declares McCain is doomed based on this "watershed" event is laughably partisan. If the analyst had combined his statistical skills with a passion for McCain instead of Obama, I have no doubt this site would be showing Florida in solid red and McCain with 300 EV.
That's because you're a laughably partisan and dishonest fool. Your mischaracterizations and denials won't make your candidate any better off.
My recollection is that after the dust up in 1980 over early calls of states for Reagan, Congress passed legislation to prohibit calling any states until the polls close in all the West Coast states.
Perhaps if you knew anything about the founding principles and laws of the country you live in, you wouldn't "recollect" such absurd things.
But I tend to believe that people like to vote for the winner, so if McCain falls early, it'll help dems not hurt them.
Apparently you "tend to believe" rather ridiculous things based on no fact or reason.
Um, as for the unity bounce, have you guys seen Scott Rasmussen's latest numbers on Florida and Ohio? They're quite different from what Quinnipiac released and I don't think we have a really compelling reason to believe Q.
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,水噹噹,台中隆鼻,果凍隆乳,改運整型,自體脂肪移植,新娘造型,婚禮顧問,下川島,常平,常平,珠海,澳門機票,香港機票,貸款,貸款,信用貸款,宜蘭民宿,花蓮民宿,未婚聯誼,網路購物,婚友,婚友社,未婚聯誼,交友,婚友,婚友社,單身聯誼,未婚聯誼,未婚聯誼, 婚友社,婚友,婚友社,單身聯誼,婚友,未婚聯誼,婚友社,未婚聯誼,單身聯誼,單身聯誼,白蟻,白蟻,除蟲,老鼠,減肥,減肥,在家工作,在家工作,水噹噹,台中隆鼻,果凍隆乳,改運整型,自體脂肪移植,新娘造型,婚禮顧問,下川島,常平,常平,珠海,澳門機票,香港機票,貸款,貸款,信用貸款,宜蘭民宿,花蓮民宿,未婚聯誼,網路購物,婚友,婚友社,未婚聯誼,交友,婚友,婚友社,單身聯誼,未婚聯誼,未婚聯誼, 婚友社,婚友,婚友社,單身聯誼,婚友,未婚聯誼,婚友社,未婚聯誼,單身聯誼,單身聯誼,白蟻,白蟻,除蟲,老鼠,減肥,減肥,在家工作,在家工作,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,汽車借款,汽車借款,汽車借款,汽車貸款,汽車貸款,借錢,借貸,當舖,借款,借貸,借錢,週轉,
^^ nice blog!! ^@^
徵信, 徵信, 徵信, 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, 徵信社,
^^ nice blog!! thanks a lot! ^^
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,
牙醫,植牙,矯正,紋身,刺青,創業,批發,皮膚科,痘痘,中醫,飛梭雷射,毛孔粗大,醫學美容,肉毒桿菌,seo,關鍵字行銷,關鍵字、自然排序,網路行銷,關鍵字、自然排序,關鍵字行銷、seo,關鍵字廣告,部落格行銷,網路行銷,seo,關鍵字行銷,關鍵字廣告,關鍵字,自然排序,部落格行銷,網路行銷,網路爆紅,牛舌餅，婚紗，台中婚紗,腳臭,腳臭,腳臭,腳臭,腳臭,中古車,二手車,中古車,二手車,高雄婚紗。
Post a Comment