There's a rather strange split in the polls in New Jersey's gubernatorial race: those conducted by live interviewers have Jon Corzine leading by an average of 4.2 points, but those conducted via IVR (automated scripts or "robopolls") have Chris Christie leading by an average of 3.0 points. The difference is statistically significant with probability t=3.26, i.e. about a 98 percent chance that it isn't the result of chance alone.
This is probably not a phenomenon local to New Jersey; PPP and Rasmussen have generally shown pessimistic numbers for Democratic numbers in other states thus far this cycle. An automated poll tends to be associated with lower response rates, since an automated script can't do as much a human to coax someone into an interview, and therefore sometimes tends to reach a more enthusiastic set of respondents (in effect, it may serve some of the same functions as a very tight likely voter screen).
Since Republicans tend to be more enthusiastic right now, that may be what's causing the automated polls to be more favorable to them. But since none of us yet know how the enthusiasm gap is going to play out in practice, it would be premature to come to any conclusion about whether the voter universe that Rasmussen and PPP are coming up with is "too tight" or "just right". For that matter, I'm not yet ready to make a forecast for New Jersey; I certainly don't see it as self-evident that Jon Corzine ought to be as much as a 2:1 favorite, which is where Intrade has him now.
78 comments
I think you want to have Corzine as +4.2 in Blue
Latest SUSA poll--tied at 43%.
Also noting that Corzine leads by 8
among those who already voted.
Maybe not a 2-1 favorite, but I think Corzine has a 55% chance of winning this.
Nate, I love the way you talk about Rasmussen's results. Scotty provides us with good data. Of course, it has to be adjusted just a little, but it's good data. (Not gospel or b.s., as the wingnuts will soon claim.)
Rasumssen has a pretty good record on verifiable polls. It will be interesting to see how he does on this one.
Anyone have any details of the early voting? Just from a glance of 2008 early voting results, you would think Corzine might not be that happy with 8% (for example, PPP had the early voting in Indiana right before the election as 64-35 Obama). But then again, 2009 isn't 2008 and the situation is different.
I certainly don't see it as self-evident that John Corzine ought to be as much as a 2:1 favorite, which is where Intrade has him now.
~~~~~~~~~~
Intrade is obviously taking into consideration NJ is a state, as a rule, where Reps are usually, almost always disappointed by election results.
If Christie grasps defeat from the jaws of victory it will just prove the axiom once again ie NJ is a Dem state, despite corruption, performance, etc.
Yea, Christie was a very good candidate until his past was uncovered and he started opening his mouth lol, plus having Joe You lie! Wilson campaign for him was a real coup, lol.
Everyone has a plan until they get hit!
My conservative NJ friend has been disappointed many, many times and yes, shocking that I have a conservative friend! ;)
Small nitpicky detail: Jon Corzine, not John.
Anyway. Aside from the irony of GROG posting immediately after Just John's commentary re: Rasmussen, I do find that interesting. I also find it rather noticeable that - along with Ras in the "favorable to Christie" universe are the (Communist) Research 2000 and PPP (which - IIRC - is a Democratic pollster, right?)
I also think that the major outlier dragging up Corzine's numbers is Suffolk - which has him up by 9%. As a dyed in the wool Democrat, I find it very hard to believe that 9% of voters would prefer Corzine to Christie this year. It's like taking a poll on whether you'd prefer to have chlamydia or gonorrhea, and finding that 9% more people would prefer the chlamydia.
I don't know why Nate is having trouble understanding that automated polls tend to be more accurate. His own pollster rankings show it.
Rasmussen and SUSA dominate.
My guess us that he does not like the implications of that.
Remember who administers live polls...
Low-paid workers, mostly women.
Ever wonder why auto-polls on abortion always show greater pro-life tendencies?
How do you think a man likes to tell a woman that he us pro-life?
Don't Know is easier...
Auto polls are better.
I was curious about how old these polls are. Here's that information, along with sample sizes, for anyone else who wants it.
So why has Nate not updated his pollster rankings since the 2008 election?
Curious.
It would be terrible if Rasmussen was #1, wouldn't it?
MidPointMan said...
~~~~~~~~~~
Why so personal? As mentioned previously, if one is upset, start your own damn winger blog!
Winger blogs/sites are a hoot, or so I hear ...
I'm not usually one to blow this horn, but the New Jersey governor's election is a clear poster child for the problems with a two-party system.
As Nate himself said, both mainstream candidates are "deeply flawed," and the extent of negative campaigning and hyperbolic coverage by the media has made it virtually impossible to actually deduce much factual information about the candidates.
On the other hand, Chris Daggett, the independent candidate, tests extremely well with voters from both sides of the table who are familiar with his platform. However, both parties voters' fear that a vote for Daggett will take away from their own party's tallies — irony at its finest.
Of course, it's also a testament to our deeply flawed political culture that Christie is a competitive candidate at all, given his refusal to actually provide a substantive platform, apart from deregulation and sweeping tax cuts. Cutting taxes in a state that is nearly bankrupt would be a very neat trick indeed. If elected, I give virtually zero odds that Christie would be able to get elected to a second term -- Bush Sr. learned the same lesson in 1992.
Republicans in NJ are embarrassed to call themselves Republicans so they will give the politically correct answer at times. The only time a governor who was disliked was on the ballot was Florio in 1993 (although he was a decent governor not like Corzine who, on his best day, can be kindly described as incompetent). Anyway, Florio was up 9-10 points in the polls 2 days before the election when all of the undecided and supporters of third party candidates (none of them as good as Daggett) voted for Whitman.
It is hard to support the party as as Corzine is the worst of the three candidates (I think little of Christie but since I know that Corzine is a failure, the other two can't be worse). That is going to be the question .. will the people in NJ vote party .. or person.
@Andrew:
It's a classic example of the failure of Plurality Voting. Daggett is pretty clearly the Condorcet winner (that is, he could defeat any other candidate in a head-to-head match up), and probably the second choice of all but a tiny percentage of voters.
That means he would win under virtually every voting system except Plurality and IRV.
Who else here hangs up the phone as soon as they hear one of those animatronic voices? I know I do--I assume it's a scammer or garden-variety autotelemarketer trying to waste my time.
Yeah a lot of people on both sides are scared that Daggett is pulling votes. I have literally heard Corzine supporters say that they would support him, but dont want Christie to win. The conventional wisdom would argue otherwise.
Christie's train has derailed. He had no plan. All he says is "I'm not Corzine", "I fought corruption" (that's like a fish saying "I fought water"), and "NJ is too taxed".
Unfortunately Jabba, we have a budget gap and billions in debt your namesake left us a decade ago.
It's either Corzine or Daggett. I hope NJ can break from the fear-mongering parties and make the right choice.
Another note: Corzine's rise is not Daggett "stealing" from Christie. The Democraps have rallied their base very well. Many were going to stay home, but now will not. Christie also had no plan as I said.
Andrew pretty much nailed it. Corzine has the charisma of a fish. From the bottom of the Passaic. With three eyes. (I know. I repeat myself. But bear with me.)
Christie has more chins than planks. Yes, that was a cheap joke about how his waistline is far more substantial than his platform has ever been. Also, if I didn't mention it, Chris Christie is fat lol. And hasn't told us anything except NO NEW TAXES and CORZINE (and now DAGGETT) IS THE DEVIL.
(Interestingly enough, that's the bulk of his advertisement now. It's been mostly, "Yeah, Jon Corzine sucks, but CHRIS DAGGETT IS EVEN WORSE FOR NEW JERSEY.")
Daggett seems to be the adult in the room. But up until a couple of weeks ago, I was afraid of voting for him because a vote for him would be a vote for Christie, with redistricting coming up. (Corzine is my grudging second choice.) The one thing that actually changed my mind was a commenter here pointing out my assumptions about NJ's districting system were mistaken.
I also think that this would be a good place to put in IRV (though, admittedly, it's not a cure-all). It'd probably encourage third-party voting more often; a vote wouldn't feel quite as wasted if - say - Vinnie from Hackensack could vote for Daggett and then Chris Christie.
I give Corzine a 3 in 5 chance. I was never convinced that this contest was winnable for the GOP. It's a machine state. And it's entirely dominated by the Philly and NYC markets. And Corzine has bundles of cash.
I bet it's not even that close. I suspect Corzine wins something like 48-43-8.
@kerrick
I also hang up right away.
I think robocalls undercount people with brains, i.e., liberals and Dems
The differential between automated and live calls may be the "Shy Tory Factor" at work. Under this theory, polling undercounts conservatives because they do not like to participate in polls. Perhaps conservatives do not like to talk about their politics with strangers, but do not mind speaking with a computer.
Conservative Hoffman will win in NY-23. All of the enthusiasm is on the right and the Club for Growth essentially (stupidly, in my view, because of what may happen in future contests) disqualified Scozz.
No question that Republicans win big in VA. Probably they'll sweep all three contests. If they make big gains in the state legislature then the Dems are going to be cooked in the state for the rest of Obama's presidency.
I suspect it will be a 2 for 3 night for conservatives - but again I'm not entirely on board with what's going down in NY-23.
The difference is statistically significant with probability t=3.26, i.e. about a 98 percent chance that it isn't the result of chance alone.
SHAME, Nate. That's not what statistically significant means! It means *if* the two polling methods really were equivalent, there's a 2% chance of getting a difference as extreme as this one. Without a prior distribution on the probability of the two polling methods being equivalent, we can't say what the probability is now.
This is an elementary error in interpreting statistics, and while I'm pretty sure you actually know better, plenty of your readers probably don't.
ALEX
all you did was basically restate the same conclusion as Nate & thereby you confirm his statement...
your say "there's a 2% chance of getting a difference as extreme as this one"
and Nate states the mirror reverse conclusion with the same end result "a 98 percent chance that it isn't the result of chance alone".
You say tomato, he said tomatoe...
but you are talking about the same conclusion basically, especially since the forum we are in is not for peer review, but for mass consumption imho
Happy…
You illustrate perfectly why your party is such a mess. Not once have you outlined or in any other way developed a point of view—all your posts are about how in the future the GOP will have glorious success and the Democrats will suffer ignominious failure.
You neglect, as do all the strategists of your party, that the GOP has to offer an alternative to policies it doesn’t agree with. “No!” just doesn’t go very far; it can get some people turned off to the party in power, but doesn’t stimulate them to vote for the party that’s out.
But keep it up. Your formula is perfect for keeping things just as they are for the foreseeable future.
Pragmatus,
Our party is going to do very well in less than a week. It would behoove you and your party to examine why liberalism has reached its high water mark and is on the decline.
Your president has seen his approval ratings decline more precipitously in the last year than any president in recent history. Don't smugly lecture me and Republicans on what it takes for electoral success.
What was Creigh Deeds' great proposal to fix transportation and the economic situation in Virginia?
One would think that commenting about electoral horse races would be welcome on a site entitled "538.com"
Or are you just sore because you are starting to feel the wind shift?
We'll see how it goes on Tuesday. "No!" is not really a platform.
Virginia is pretty red, has been for a long time - a Republicant victory there won't mean much. It's been turing purple - which means I don't expect 100% Dem victories, especially with a weak Dem candidate.
NY is fun. It's a joy to watch the right wing eating itself. The Republicant candidate looks to be the weakest, and that's a happy place.
NJ will be a Dem victory. History goes that way.
Happy GOPer said...
Conservative Hoffman will win in NY-23. All of the enthusiasm is on the right and the Club for Growth essentially (stupidly, in my view, because of what may happen in future contests) disqualified Scozz.
The trends do appear to be going for Hoffman:
1) The new kos poll has Hoffman taking 9 points from Scozz and Owens dropping 3 points into a tie with Hoffman all in one week. This confirms the surge for Hoffman reported in the CFG poll and another conservative group's poll.
2) Rep. Pete Sessions just announced that the GOP would welcome Hoffman into their caucus with no real mention of Scozz.
3) The final debate was essentially between Owens and Hoffman with both men largely ignoring Scozz sitting between them.
If the automated polling is correct in NJ, it could be a threefer on Tuesday, which would be really interesting considering how badly Hoffman and Christie have been outspent.
a Republicant victory
Funny contradiction, that. Eh?
which means I don't expect 100% Dem victories, especially with a weak Dem candidate.
Don't kid yourself. McAulliffe and Moran would be in the same spot right now.
I'm really confidant now. BdP has been 100% wrong for as long as I've been following his posts.
Bart DePalma said...
Perhaps conservatives do not like to talk about their politics with strangers
~~~~~~~~~~
Indeed, conservatives like yourself, don't enjoy always losing face to face political debates as it can be depressing and disillusioning. ;)
Whereas it's easy for blowhard, party of No! yahoo, winger lemmings to hide on the net, eh. :)
Happy…
You still illustrate perfectly why your party is such a mess. All your posts—including the most recent—are about how in the future the GOP will have glorious success and the Democrats will suffer ignominious failure.
(By the way, I don’t have a party. I am registered in California as having “no party preference”. Both major parties commit major mistakes as far as governance, but since the GOP has become little more than a hate-mongering theocracy I feel it incumbent to fight them wherever I can.)
Keep it up though! You’re the best mouthpiece for illustrating why independents should shun the GOP.
ALEX
all you did was basically restate the same conclusion as Nate & thereby you confirm his statement...
your say "there's a 2% chance of getting a difference as extreme as this one"
and Nate states the mirror reverse conclusion with the same end result "a 98 percent chance that it isn't the result of chance alone".
You say tomato, he said tomatoe...
but you are talking about the same conclusion basically, especially since the forum we are in is not for peer review, but for mass consumption imho
DCM, they AREN'T the same thing. They aren't the same thing at all. The inference you drew is a huge, fundamental mistake in the interpretation of confidence intervals.
An example:
Let's say I have a box of coins, and I know a prior that exactly half of them are fair coins, and half of them always turn up heads. I draw a coin at random from the box and flip it 4 times, and every time it comes up heads. There is a 6.25% chance of that happening if the coin I drew was a fair coin; that's the statistic equivalent to the 2% Nate reports. However, that DOESN'T mean that there's a 93.75% chance of the coin being unfair; the actual probability is (2^4) / (2^4 + 1) = 94.11%.
"Pretty close," you might say. But that's only if we assume beforehand half of the coins are unfair. Let's say that we assume only 1% of coins are unfair. The probability that the coin is unfair based on it coming up with 4 heads is (.01*2^4) / (.01*2^4 + .99*1) = 13.91%.
And the fact is, we DON'T KNOW what the prior distribution is in the real case of these two polls, so we CAN'T find "the probability that the result isn't due to chance".
Nate's statement would flunk anyone out of any introductory statistics course, anywhere, in the first week.
Prag,
Fair enough. I'll come back on Tuesday and look backward.
Maybe Independents should shun the GOP. We'll see if they do.
May I remind everyone, Virginia was a presidential Red state from 1964 to 2008 and it took (8) years of cheney/bush and the best Dem campaigner of all time, Barack Hussein Obama to turn the state Blue.
and yes, just to agree w/Happy one time :) Don't kid yourself. McAulliffe and Moran would be in the same spot right now. True, as would Tim Kaine also ie an off/off year election is low voter turn out and usually good for the party not in power as they are more motivated.
Again, no thanx needed for my keen grasp of the obvious ie yes Virginia ;) Virginia is still a toss-up er purple state when it comes to politics.
carry on
p.s. 7 in 10 voters say Obama will have no effect on their VA vote.
we now return you to the state of NJ ...
Not sure how strong Daggett's support really is. I have seen some polls that list all of the third party candidates in which Daggett's numbers go down. At this point I think in NJ, you have about 20-25% that would rather somebody other than Corzine or Christie (maybe more) but they will probably not all vote for the same candidate in the end
ALEX
no, you fail in your analogy.
first, you are the one making 'assumptions' and then diddling with the data input.
Nate arrived at a generalized conclusion [which he supports] - you are just spinning your wheels.
btw - I took graduate level courses in statistical analysis as well as publishing the results of my fieldwork. but as I said before, Nate is just throwing a glaringly apparent trend thought out there that the final results will either prove or disprove next week. This is not a forum for academic peer review, but for msm public consumption in a coherent manner.
YOU are just full of yourself... but then I am making an ass-umption...
The 4th parties (at this rate) will collectively get 2-5%. Daggett will range 12-38%.
I usually am very orthodox when it comes to polls, but people may just show up at the polls and make a gut decision for Daggett. I'm not terribly confident in this and would like some insurance in the high 20s, but it is a possibility.
This post by Nate is now a feature & linked over at Political Wire as 'Which Polls to Believe in New Jersey?'
http://politicalwire.com/archives/2009/10/29/which_polls_to_believe_in_new_jersey.html#disqus_thread
Nate,
How about plugging all these numbers into your magic machine and give us some regression analysis predictions and see if you still got it?
DCM, my point was that "the probability the results are not due to chance" varies based on the underlying likelihood of of the different hypotheses (the prior distribution). Since we don't know the underlying likelihood, we CAN'T know the conditional probability.
Nate's error, and yours, is a basic misunderstanding of statistics that even has its own name: the Prosecutor's fallacy (http://en.wikipedia.org/wiki/Prosecutor%27s_fallacy).
The fact that this is for mainstream consumption is exactly why it's so important not to tell total falsehoods. Nate's claim wasn't "sort of" correct in any sense; it was completely false.
ALEX
you are still reading too much into Nate's broad 'conclusion' that this particular array of polling data has "a 98 percent chance that it isn't the result of chance alone".
In other words, all Nate is saying is that for this set of data, there must be something MORE than basic 'CHANCE' that IVR would deviate so much from LIVE polling - or as we like to call it 'Correlation'...\
which is why LDF requested a regression analysis. then you can play with your variables & plug ins for stat modeling all that you [or Nate] like...
in summary, Nate did not attempt or claim he analyzed OR 'proved' anything more than a casual generalized conclusion based on the data at hand... that these particular polling results are highly unlikely to be produced randomly or by 'chance' alone.
you doth continue to protest too much...
The "shy tory" syndrome is very much a real thing, though I am not sure it is a shy TORY syndrome exclusively.
The concept developed after the 1992 UK elections, when John Major unexpectedly won, despite all of the polls having him well behind Neil Kinnock (the guy whose speech Biden plagiarized).
The issue wasn't so much that Conservatives don't like talking to pollsters. In fact, if that were the case, then virtually all telephone polls in the 80's would have under-counted the right. Yet we don't see anything with that kind of consistency.
What drove the 1992 result was that people didn't want to declare support for an unpopular government to another human. You can see a similar effect in polls of Canada's Liberals when they were embroiled in a scandal.
Why does this matter now? If anything, I should think voters are less likely to voice support for the corrupt Corzine administration. The poll gap is probably picking up an enthusiasm gap. Given that this is an off year, that could be enough for Christie to win, but I wouldn't bank on it.
tmess2…
I agree with your point about Daggett’s strength. There has been so much spin of the polling data in the MSM that it’s hard to know what’s really going on. I expect Corzine to win however, and it will be due not only to Daggett taking Christie votes but to Christie’s rising negative impression as well.
Same is true of the NY-23 race. Arch conservatives are so eager to prove that they are relevant while the GOP no longer is that their spin apparatus has twisted the poll figures out of all proportion. At this point it is wait and see—my guess is that by election day the right will have successfully murdered each other’s chances, allowing the Democrat to take the prize.
@DCM
Everything Alex has said in his posts is 100% correct and he has fully justified his posts with logic and sources.
You don't seen to understand what 'conditional probability' means.
SCHOOM or whatever
nope, you are making an ass-umption just like ALEX...
and also drawing & projecting your own set of 'conclusions' apparently by over-analyzing Nate's inference in this post
you guys are 'over-conditioning' the probabilities. [look it up]
2 % chance v. 98 % not by chance remains a perfectly valid probability calculation for this particular data set occuring by random/chance alone [rounded] as presented in this format.
cuz you know what they say about stats & statisicians [or is it liars...]...lol
full stop
I take it back, you're just an idiot troll.
As a New Jersey voter, I'd like to say something to Washginton Republicans.
This race is not a referendum on the president, his policies, or the Democratic Party. If Corzine loses, it's because he's quite frankly an awful governor. If he wins, it's because Christie is that much worse.
For many in my state, this is a vote between the evil you know versus the evil you don't know. NO ONE I know on either side of the fence is particularly enthusiastic about either candidate.
I'm personally voting for Chris Daggett, but if I had to choose between the two major candidates, I'd reluctantly vote for Corzine. I was considering Christie, but as his campaign continued and I learned of his positions, I liked him less and less.
and that makes you superior SHAM ??? pathetic. get skooled...
Alex is 100% correct about the basic issues in probability. In many cases, the distinction he's making have huge quantitative consequences. In medical cases, they are matters of life and death.
On the other hand, his claim about how knowing these things is required in elementary stats courses is almost completely wrong. These courses and many of their texts are notorious for getting such things scrambled up.
p.s. I had offered to write a little tutorial on this Bayesian issue for this blog, but the offer has not been snapped up. Probably a wise ratings move.
Um, no, DCM.
We literally do not know the probability that the result would occur by chance.
We can't, because we don't know the relevant prior probability.
This is basic, basic stats.
@ Michael and Michael
Don't even bother. You'll just be called asses or the subject of some other juvenile insult.
"On the other hand, his claim about how knowing these things is required in elementary stats courses is almost completely wrong. These courses and many of their texts are notorious for getting such things scrambled up."
That's just wrong (ethically, I mean, not factually).
Nate's statement would flunk anyone out of any introductory statistics course, anywhere, in the first week.
Blazing laser beams of hyperbole! I don't think it's actually physically possible to flunk out of an introductory statistics course in the first week, for any reason.
That being said, the substantial parts of Alex's arguements are correct. "Statistical signifigance" is an exact term, and it shouldn't be misused.
"Visually interesting but statistically meaningless" might be a better way to put it.
Okay, okay, y'all are right about the hyperbole :P
@Alex etc.
There are so many examples of far more sloppy use of statistical language out there (let alone on this site) that to call this point out seems absurd.
We have a null hypothesis. The null hypothesis states that ceterus paribus, there is no difference between human and automated polling. The alternative hypothesis is that there is some non-zero difference.
Our t-test simply allows us to evaluate the strength of our null hypothesis. The level at which we reject the null hypothesis is a subjective choice Nate leaves up to the reader, but he is correct when he asserts that there is "about a 98 percent chance that it isn't the result of chance alone" in that 2% of random samples of this size where the null hypothesis was true would be expected to produce t scores this high or higher if we assume he is only referring to the t statistic itself which is entirely plausible.
Nate makes no assertions about how strong an effect Robo versus Human polling actually makes. You are correct that we are in danger of making a conclusion along the lines of the Prosecutor's Fallacy (see Schwarzenegger's "coded statement").
It is true that he only thought to ask the question after he had observed some of the data. If he were to really test this hypothesis he would have to do so going forward.
But the reader is free to draw their own conclusions, so it is a bit assinine to say that because Nate did us the favor of associating his calculated t statistic with its 98% confidence level that "Nate's statement would flunk anyone out of any introductory statistics course, anywhere, in the first week.
".
Is it me or does it seem that New Jersey always threatens to go Republican but never winds being close?
I'm still torn over NY23. A raving right wing lunatic who might inspire the GOP to become even more wacky and regional? A pathetic crooked wussy republican or a pathetic wussy blue dog?
I almost like the first option.
@Valpey- Yeah, there are worse examples, but this general issue comes up so much that Alex had a good point to make.
On your "null hypothesis": We agree that all polling techniques have some systematic errors, which are real numbers. Qualitatively different polling methods (automated vs. live) have no special reason to have the same systematic errors. The probability that your null hypothesis is right is exactly zero, even before we see data, since any point is a set of measure zero.
We're really asking an entirely different question: Do we think that much of the 7% gap between the methods is likely to be accidental? Nate's legitimate point is that accidents that big don't usually happen. The size of the random error bars is fortunately not significantly dependent on the accuracy of the (certainly false) null hypothesis.
My poly sci professor (who works at the Heritage Foundation) calls NJ "the big tease". He said that for Republicans NJ is like that cheerleader in HS that at the beginning of the night doesn't know your name but by the end of the night is about to go home with you until the football player comes into the room.
I always thought that was the best explanation for NJ consistently raising Republicans hope, only to crush them.
BDP said
'The trends do appear to be going for Hoffman:
1) The new kos poll has Hoffman taking 9 points from Scozz and Owens dropping 3 points into a tie with Hoffman all in one week. This confirms the surge for Hoffman reported in the CFG poll and another conservative group's poll.
2) Rep. Pete Sessions just announced that the GOP would welcome Hoffman into their caucus with no real mention of Scozz.
3) The final debate was essentially between Owens and Hoffman with both men largely ignoring Scozz sitting between them.
If the automated polling is correct in NJ, it could be a threefer on Tuesday, which would be really interesting considering how badly Hoffman and Christie have been outspent.'
------------------------------
An alternate spin might be that if Va goes GOP, and the other two elections are won by the Democrat (Corzine) and Conservative (Hoffman), the anti-GOP vote will have won 2 elections to 1. Of course from there further conclusions may prove difficult. Would a more conservative GOP contender in NJ have faired better? Why did the anti GOP vote not work out in Va? What does the conservative win in upstate New York mean for statewide races in 2010? Is it going to push Republicans rightwards? Might conservative wins reenergise Democrats and show them that they still need to turnout in 2010? I think its a very interesting little set of elections.
"a 98 percent chance that it isn't the result of chance alone" is not the same as "a 98 percent chance it would have produced a smaller difference than 7%-points from chance alone".
Of course, Nate meant the latter. But many people would not realize that.
Inferno: "I also think that this would be a good place to put in IRV (though, admittedly, it's not a cure-all). It'd probably encourage third-party voting more often; a vote wouldn't feel quite as wasted if - say - Vinnie from Hackensack could vote for Daggett and then Chris Christie."
Why use IRV when it's easier to do one better with Condorcet? But yes, IRV would probably help in this situation. Probably.
@Inferno, Drachefly:
I was so busy feeding the troll, I missed Inferno's post. Daggett would almost certainly lose under IRV. The only way I could see him winning under IRV is if polls are underestimating his real support by 20-30 points.
IRV is a very poor voting system in regards to helping third-party candidates.
A few things to remember....
Everyone knew the polls would tighten..New jersey is as blue as the sky on a clear day
Corzine has spent more of his own money on Ads for this race than the combined lifetime income of everyone who has posted here about the race
Christie, while not as strong a candidate as he first appeared(one wonders how Lonagan would have done) he has been defined by all of Corzines money.
The fact that A Rep. is even close in this state is a miracle akin to Mosas parting the Raritan River.
Corzine is the worst governor in this state in a line of terrible ones..what a motly group that includes (Tax em til they bleed Florio)..Christie(If i only had a brain Whitman) and JIm (if you sleep with me I'll make you head of Homeland Security) Mc Greevey and finally Jon (give the union contracts to my girl friends firm) Corzine.
The Democratic Machine In 2005 Decided that Corzine should be the governor and failed to put up the one candidate that would have been the best of all for our state Richard Codey(this is a republican saying this)
This state is ungovernable...and sadly we are going to re-elect the most corrupt and worst governor in the entire United Statesto make it worse. No wonder people are leaving in droves
...yeah, I was just pulling stuff out of my posterior other than our current plurality voting system. Condorcet would probably work better, though it would lead to the Condorcet paradox under fairly rare circumstances.
w/r/t Christie - I don't think a more conservative candidate would have done better. This is a state whose Republicans - the conservative party - are to the left of Louisiana's Democrats, remember.
Point is? Okay, so this doesn't really apply to most of the suburban state, but there are quite a few urban centers in NJ. (Small ones, but still urban nonetheless.) There are two huge billboards on Mt. Prospect in Newark, by the city line (with Belleville).
One's a Spanish-language billboard showing Obama with Corzine, and the Spanish-language version of his campaign slogan "Keep It Going." (Maybe. It's been ages since I've taken Spanish.)
The other is an English-language billboard zoomed in close on Chris Christie's face (pores and all), and a photograph of George W. Bush on the right. It says something to the effect of "Bush's ally. Bush's policies. Chris Christie: Bad For New Jersey."
I also think that part of Daggett's appeal is that he's to the left of Christie in most regards. (Both socially - one thing I can think of is gay rights - and fiscally (Daggett is proposing adding sales taxes to other services; when Christie was told of that, he turned red and started sweating because TAXES ARE ANATHEMA).) Which - like I've said - he's been attacked over quite a bit recently. ("You think Jon Corzine taxed you a lot? Chris Daggett will tax you even more! Ooga booga!")
The way it's panning out now with regards to who's been attacking who:
Corzine: CHRIS CHRISTIE WILL EAT YOUR CHILDREN. WITH EXTRA CREAM CHEESE. (I mean, seriously, look at the guy. He's probably thinking about pie right now.)
Chris Christie: OH MY GOD JON CORZINE CALLED ME A FATTY. AND HE HAS SINGLEHANDEDLY DESTROYED NEW JERSEY WITH MASSIVE TAXES. HE IS THE SPAWN OF SATAN.
Chris Daggett: SCREW THESE ASSHATS, VOTE FOR ME BECAUSE THEY'RE TARDS.
Granted, Chris Christie is better than Steve Lonegan (I was seriously going to vomit if he was the Republican nominee), but that's like saying getting herpes is better than getting AIDS.
In other news, the candidates for NJ-Gov '09 are best described in terms of STDs.
Tangential, but: Minneapolis is doing top-3 IRV voting for its election next week. There was no primary, and candidates get to choose their own affiliations, so someone could choose three DFL candidates for mayor.
The counting may take a while, since it will be done by hand, and there are a lot of candidates. If there's no first-round majority, there may be a few rounds of redistributing votes from, say, one socialist to the other, or from the "Is Awesome" guy to half a dozen other candidates.
George:
"The Democratic Machine In 2005 Decided that Corzine should be the governor and failed to put up the one candidate that would have been the best of all for our state Richard Codey(this is a republican saying this)"
Totally agree with you there (and I'm a registered Democrat although that was only for the primaries last year). I thought Codey was an excellent governor when he was acting in between McGreevey and Corzine. Though I think he simply didn't want to run for governor instead of being passed over for Corzine.
DCM, my point was that "the probability the results are not due to chance" varies based on the underlying likelihood of of the different hypotheses (the prior distribution). Since we don't know the underlying likelihood, we CAN'T know the conditional probability.
Andrew, you seem to be missing the entire point, which is that there are different prior distributions between robopolling and live interviewers. The alternate theory Nate is seeking to dismiss is that there are not different prior distributions, in which case there is only a 2% chance that you would have such a wide disparity in results. I'm not sure what you're yelling about here.
He said there's a "98% chance that it isn't the result of chance alone" and thus that it suggests that the prior distributions aren't equal, and your response is, "Hold on Nate! What if the prior distributions aren't equal?!" Well, that wouldn't be the "result of chance alone", that would be the result of having different prior distributions among the two different polling methods!
Would you prefer him to have said "about a 98% chance that it isn't the same prior distribution"? What do you want him to do? Complaining that we can't find the different prior distributions just seems to absolutely miss the entire point of the post. The question is whether they are different or not, not what the exact chance of getting these results if they are different is. That makes no sense.
On-topic: God, I just want the New Jersey race to be over already. Who cares? No matter who's in charge, it's still New Jersey, for God's sake! :P
PERSUTER
thanks for the post. glad to see that at least a few others grasp the concept.
my stated objection to ALEX's post claiming that Nate had abused 'statistically significant' was that he [ALEX] was over-analyzing and parsing the conclusion "about a 98 percent chance that it isn't the result of chance alone."
you just can't [or should I say SHOULDN'T] try to compare a very limited universe such as this particular polling data which has at LEAST one clearly identified variable [IVT v. live] to finite probabilities such as random coin flips [as ALEX used in his analogy], or card counting or lottery number projections...
just sayin'...
Too bad, Persuter. I have yet one more anecdote today.
I just came in the door from job training. So, I'm sorting through the mail (since the mailman arrived before I got in), and there's a mailer addressed to my 76-year-old aunt.
Accusing Corzine and Daggett of the HORROR of supporting GAY MARRIAGE (ONOES). So you better vote for Chris Christie if you believe in saving New Jersey families.
It was sent from some stupid "family values" PAC.
I think the neighbors heard me shout, "OH HELL NO," from across the courtyard. I may literally burn that thing - which may make the landlord shake his head, but fuck that shit I'm pissed.
...at least if I do burn it, I'm taking it outside.
Right down to the wire. Chris has a real shot at doing this. We'll see how the turnout and Daggett factors come into play on Tuesday.
There's a reason the robo-polls are more accurate. Some of you may hang up on them, but you're just another responder. Any serious pollster will weight their polls, so with or without you, the final response percentages effectively wont change.
The key is that people that do answer robo-polls have no reason to be less than truthful with their answers, but as others have pointed out, there are reasons to lie to a live person. (The pro-life question asked by a women is a good example of this.)
I can also give a personal example. I was polled on the right-track, wrong-track question. The words used were whether I thought the country was "going generally in the right direction" or if we had "gotten seriously off the rails!" It doesn't take a genius to see the wording problem. But that isn't the true problem; the way the question was asked was icing on the cake. The first half of the question was asked in monotone; the latter half sounded as if the lady on the other end of the telephone was about to explode.
(For what it's worth, I suspect the proper wording of the question wasn't so leading; I think the pollster added her own little twist.)
The point being, I lied in my answer, just to piss her off. She wasn't expecting my answer of "Actually, I think the country is in very good shape right now."
Now, ask yourself how many people (who lean to the right) lie to (obviously) low-income workers doing polling? (I've been polled across many topics several times; sometimes these people can barely put a coherent sentence together.)
Now ask yourself in races like the ones we're talking about now, is it more or less likely that they'll be truthful.
As for NY-23, there are now at least 3 polls that show Hoffman either in front or tied with Owens. It's a squeaker for sure, but I think with the polls showing Hoffman with a real chance of winning, the usual "abandon the 3rd party candidate at the last minute" rule of thumb wont apply in this election, and he could win easily when the dust settles.
Regardless of our individual politics though, we should all be rooting for Hoffman, because it will send a message to BOTH parties to listen to their constituents. Hoffman isn't going to win with just Republican support; if he does win, it will be with the support of many Democrats as well. The end result of which is we all win.
MIKE
you are truly delusional...
an example from your last laffable posting:
"Hoffman isn't going to win with just Republican support; if he does win, it will be with the support of many Democrats as well."
of course, I suppose it depends on how you define 'many Democrats'...
but where is your proof or anything factual to support this allegation that many DEMs will defect & actually VOTE for Hoffman ?
the polls do not support your thesis, or Hoffman would have a substantial lead.
Hoffman may in fact prevail in a 3 way race - but he would have lost in a 2 way most likely.
Add Scozz + Hoffman = GOP base #'s + new enraged tea-baggers with high motivation to be Palindrone & Beckclone sheeple...
Owens + Undecideds = DEM #'s
it will be close, but that alone is a major smudge on the GOP since this SHOULD be a SAFE RED seat & Owens never shoulda/woulda/coulda had a chance
I almost hope that Hoffman prevails since the GOP caucus in DC deserves what they sow... and Steele deserves the beatch-slap that is headed his way since Scozz ain't happenin'
DCM in FL said...
MIKE
you are truly delusional...
an example from your last laffable posting:
"Hoffman isn't going to win with just Republican support; if he does win, it will be with the support of many Democrats as well."
of course, I suppose it depends on how you define 'many Democrats'...
but where is your proof or anything factual to support this allegation that many DEMs will defect & actually VOTE for Hoffman ?
the polls do not support your thesis, or Hoffman would have a substantial lead.
-----------
Really?
This is a leans-GOP district. Scozza has, roughly, 20% of the vote. Hoffman and Owens each have, roughly, 30%, with 20% undecided. Simple math tells us that either (a) Hoffman + Scozza have locked up every single "R" vote between them, and have picked off 1/2 of the Independents, leaving all the decided Dems in Owens corner with the remaining Dems undecided (ouch! for Owens), with the rest of the Independents undecided; or, (b) Hoffman + Scozza have picked up several Dems, maybe 5-10% points.
Since it's unlikely that Scozza is picking up many Dems, it stands to reason that Hoffman is. More importantly, unless Scozza quits the race, for Hoffman to win he simply has to pick up 5-10% of Dems, unless all the current undecideds simply don't vote*.
Note, I didn't say "a majority" of Dems, I simply said "many" Dems.
(* Which is quite possible in an off-year, but unlikely.)
Nate -- what's your take on the latest Rasmussen poll, specifically the numbers related to those that are 100% committed to voting on Tuesday that shows Christie's lead increasing from 3% to 7% with those folks?
Isn't the reality of this election that if Corzine can actually get people out to vote he has it won, which, in an off year with a very unpopular Governor is highly unlikely?
Well, well, well, here we go.
Scozza has dropped out of the NY-23 race. I guess now we'll really get to see Dem vs Conservative.
And also politician (Owens) vs non-politician (Hoffman).
I guess we'll now get to see who "We the People" really want in D.C.
@Persuter- I think you misunderstood what Andrew meant in an awkwardly worded phrase. What he meant by equal priors was equal prior probability that either
1. The two methods were equivalent
or
2. that they were inequivalent.
You took as being a statement that maybe (2) was right.
Neither his way nor yours is logically coherent. The methods are certainly inequivalent, with some arguable prior distribution for how inequivalent. Zero inequivalence is possible, but like any other particular value has prior probability of zero.
test
электронная почта без регистрации
Post a Comment