Comments on FiveThirtyEight.com: Electoral Projections Done Right: North Carolina Prediction: Obama by Double Digits

Great analysis. I applaud your willingness to mak...

2008-05-06T05:15:00.000-05:00

Great analysis. I applaud your willingness to make predictions and lay out your methodology before the voting starts. We'll be following your story at Babbledog.

"rcp had it at 49.5 clinton, obama 43.4-final #? c...

2008-05-06T03:13:00.000-05:00

"rcp had it at 49.5 clinton, obama 43.4-final #? clinton gained 5.1 to bo's 2% gain."

referencing PA

rcp has it at 49.7-42.7.Given the trend of late br...

2008-05-06T03:11:00.000-05:00

rcp has it at 49.7-42.7.

Given the trend of late breakers favoring clinton, a double digit victory suggests that the late breakers will go 2 to 1 obama. rcp had it at 49.5 clinton, obama 43.4-final #? clinton gained 5.1 to bo's 2% gain.

I don't see= as I have yet to see it.

Obama's base is maxed at this point(aa/higher ed), while hillary has a lot broader demographic to draw from.

If those poor white hicks take time to stop acting 'clingy', put down their guns and bibles long enough to vote, hillary will still lose, but by 6 or less.

Indiana is more likely to go double digit, the other way.

CD-01 in NC will go at the very least 75-25 for Ob...

2008-05-06T02:44:00.000-05:00

CD-01 in NC will go at the very least 75-25 for Obama and likely it will be something like 80-20. You will see. :)

Has this model been used ex-post-facto on prior st...

2008-05-06T02:37:00.000-05:00

Has this model been used ex-post-facto on prior states to judge how it would have fared in prior contests?

If so, in what circumstances has it tended to over estimate which candidate, etc?

Obviously there are some issues in applying the model to previous states, such as changing perceptions of the candidates over time, but it would still seem useful.

Just wondering: could the counter-Bradley effect b...

2008-05-05T23:35:00.000-05:00

Just wondering: could the counter-Bradley effect be manifesting itself indirectly via the turn-out model (and not directly via the expressed preference), i.e. pollsters stick to a low AA turn-out model because they often hear from AA respondents "won't vote", meaning "I rather not have this conversation with *you*, thank you".

Empirical question: can we go back to the data and see how much of the Obama underprediction in Southern polls is due to AA turn-out models, how much due to wrong reading of AA preferences, and how much to wrong reading of White preferences?

Anon@5:04PM: NW IN being in Chicago suburbs coul...

2008-05-05T22:56:00.000-05:00

Anon@5:04PM:

NW IN being in Chicago suburbs could be a double-edged sword. According to some talking head on MSNBC, that area has been inundated with stories about Rev. Wright from the chicago media. Implication was that it was actually helping Clinton.

Polls, and statics, and predictions and pundits, a...

2008-05-05T22:52:00.000-05:00

Polls, and statics, and predictions and pundits, are always wrong. And if they are once in a blue moon right it's not because they were right, it was because reality happened to coincide with their guess. Zogby used to be spot on. No more. Rasmussen is the go-to guy now, but for awhile. I stopped believing the polls and stats after drinking the kool aid too many prior elections in the past.

So the pollsters are estimating only a 32% to 33% ...

2008-05-05T21:35:00.000-05:00

So the pollsters are estimating only a 32% to 33% AAQ turnout. Dio you supposed they forgot to include the effects of the significant voter supression that Rendell's minions pulled off in Philly? While we're talking about disinformation (or non-information) promogated (or covered up) by the privatized propaganda ministry, how many noticed that the Gallup poll that shows Clinton +7 CHANGED to a "likely voter" scam from another class of algorythim (sp). Wouldn't we like to know what the results would be if they'd treated their traw data consistently? Bet it wouldn't favor Clinton, thoough.

I honestly think Obama will overperform in Indiana...

2008-05-05T17:04:00.000-05:00

I honestly think Obama will overperform in Indiana (especially North west Indiana), as well as NC. The main reason here is that Indiana borders Illinois, and NW IN is in the Chicago Suburbs / Media Market. I know this isn't new news, and the disproportionate support in the region is reflected to a certain extent in the in the polls, but The home state advantage thing here throws the turnout projections out of whack. Not only are they more likely to vote for someone to whom they feel a home state connection, it MAKES THEM MORE LIKELY TO VOTE AT ALL than they might be otherwise. The people who don't feel the home state connection, on the other hand, feel less obligated to vote. So He'll overperform state wide due to high turnout (mostly due to Obama supporters) in NW IN.

My Indiana prediction is Clinton 52.3Obama 47.7Tho...

2008-05-05T15:22:00.000-05:00

My Indiana prediction is

Clinton 52.3

Obama 47.7

Though I'd much prefer that it were flipped around.

Any prediction for Indiana?

2008-05-05T15:02:00.000-05:00

Any prediction for Indiana?

Pretty much agrees!? Those results are nearly ide...

2008-05-05T13:58:00.000-05:00

Pretty much agrees!? Those results are nearly identical:

538: 58.6% O, 41.4% C; 1.15 million
SSS: 58% O, 42% C; 1.2 million

Another demographic-based regression pretty much a...

2008-05-05T13:20:00.000-05:00

Another demographic-based regression pretty much agrees with you. The NC election will offer a very interesting test case on the accuracy of polling methods in this election cycle.

anonymous,Independents can vote in NC.I would gues...

2008-05-05T12:58:00.000-05:00

anonymous,

Independents can vote in NC.

I would guess that 72% of NC Dems
and Indep are quite close to 72%
of PA Dems.

The one problem I see with using the Kerry vote i...

2008-05-05T12:38:00.000-05:00

The one problem I see with using the Kerry vote in NC is that NC weird in that it votes for Dems at the state level and Reps at the federal level. NC primary voters will be more conservative than you would expect from the Presidential election, so a better comparator would be the results from the 2004 governors race, in which blacks comprised around 41% of the Dem vote.

OoI realize.51,9-48,1.Sorry, and thanks Anon, I lo...

2008-05-05T12:09:00.000-05:00

Oo
I realize.
51,9-48,1.
Sorry, and thanks Anon, I love your sense of humor.

Anon @ 10:59 - ROFLRasmus...read your post again t...

2008-05-05T11:47:00.000-05:00

Anon @ 10:59 - ROFL

Rasmus...read your post again to realize what Anon was talkin about.

Also interesting: plugging 43% into your spreadshe...

2008-05-05T11:21:00.000-05:00

Also interesting: plugging 43% into your spreadsheet below gives a 17% Obama margin of victory.

Very interesting. If, instead of a multiple regres...

2008-05-05T11:09:00.001-05:00

Very interesting. If, instead of a multiple regression, you run two linear regressions against (1) Kerry vote, and (2) black population, you get two distinctly different predictions in NC: Kerry vote predicts 53%, while population predicts 33%. So I think you're right, the pollsters are using population only. The mean of these numbers is 43%, which is close to what one would expect base on early voting, and matches your multiple regression too (as expected).

In Indiana, single-regression on Kerry gives 16%, and on population gives 9%, with a mean of 13%.

I am also, but nevertheless that´s my prediction. ...

2008-05-05T11:09:00.000-05:00

I am also, but nevertheless that´s my prediction. And I got Pennsylvania within 0,5% with that ( I predicted a 9,6% margin, but that numbers changed, since the Rasmussen and Quinnipac polls from 24 and 26. April were good for Obama. Now I would predict a 6,2% margin. )- and Texas with the polls that were out on 4/3 also within 2%, even if I would now predict a 50,1-49,9 Clinton split.

And for Ohio I don´t want to calculate the prediction for 3/4, that would be too complicated, but for today I´d predict a 9,8% Clinton margin- that is also close to the election results.

You see, that method works not bad, and the NC numbers look reasonable also- I don´t know why I get so strange numbers for Indiana. Maybe the Republicans there love him- or we will see a big upset tomorrow.

I am 101% certain that your Indiana numbers are in...

2008-05-05T10:59:00.000-05:00

I am 101% certain that your Indiana numbers are in error, Rasmus.

Interesting. My- very simple- prediction of NC fit...

2008-05-05T10:41:00.000-05:00

Interesting.
My- very simple- prediction of NC fits not so bad with yours.
I predict a 57,5-42,5 win for Obama, based on my formula margin=((McCain-Obama Average)-(McCain-Clinton Average)*3)

For Indiana, it would be 52,7 Obama- 47,3 Clinton based on your numbers, I use a different weighting system and list the SUSA/Downs Center poll as Downs Center, not as SUSA polls, so I get a prediction of

51,9 Obama
49,1 Clinton

A question for you, jpm:Is the population of NC De...

2008-05-05T10:30:00.000-05:00

A question for you, jpm:

Is the population of NC Democrats 72% of the population of PA Democrats?

Anecdotally, not only African-Americans but Hispan...

2008-05-05T10:04:00.000-05:00

Anecdotally, not only African-Americans but Hispanics are turning strongly in support for Obama.

For some, there is a strong convergence between race-baiting and immigrant-bashing.

I can only guess how this is going to play out in places like Texas...and Arizona.

- cskendrick