FiveThirtyEight.com

[KSigkid]

I like fivethirtyeight.com a lot - as I've mentioned before on here, I'm a big fan of Nate Silver's baseball work, so I was happy to see him doing work on politics.

I also frequently check out the Rasmussen website.

[UGAalum94]

Although I don't know if McCain can win, I don't think he's going to lose by the margins showing.

I admit that I, of course, have no way of knowing what will happen in the future and how well pollsters are tracking in the present, but the couple of articles that I've read explaining that polls are trying to predict for changes in new registrations and likely voters make me think that there may be a lot bigger actual margin of error than is being accounted for when people report results.

I also think that people responding to polls may be less likely to admit to supporting McCain while there is so much negative coverage of McCain and Palin rallies and supporters. (This may or may not go beyond whatever Bradly/Wilder effect may exist.)

ETA: I was Wikipedia-ing Bradly Effect after I posted, and they mention the Shy Tory Factor and the Spiral of Silence to describe what I mean.)

[Munchkin03]

Originally Posted by UGAalum94

I also think that people responding to polls may be less likely to admit to supporting McCain while there is so much negative coverage of McCain and Palin rallies and supporters. (This may or may not go beyond whatever Bradly/Wilder effect may exist.)

Wouldn't those people just say they were "undecided," then?

[UGAalum94]

Originally Posted by Munchkin03

Wouldn't those people just say they were "undecided," then?

Apparently not in the cases where they've tracked this stuff in the past, but I don't know really. I think the deal is they give the answer to the pollster that they think makes them look good.

The polls all may be dead-on for all I know, but the next time you are looking at a particular polls results, look for the breakdown of who participated in the sample and see if the breakdown looks right to you. Maybe it is with new registrants, but we probably won't know until election day.

[KSig RC]

Originally Posted by UGAalum94

Although I don't know if McCain can win, I don't think he's going to lose by the margins showing.

I admit that I, of course, have no way of knowing what will happen in the future and how well pollsters are tracking in the present, but the couple of articles that I've read explaining that polls are trying to predict for changes in new registrations and likely voters make me think that there may be a lot bigger actual margin of error than is being accounted for when people report results.

I also think that people responding to polls may be less likely to admit to supporting McCain while there is so much negative coverage of McCain and Palin rallies and supporters. (This may or may not go beyond whatever Bradly/Wilder effect may exist.)

ETA: I was Wikipedia-ing Bradly Effect after I posted, and they mention the Shy Tory Factor and the Spiral of Silence to describe what I mean.)

All of this does occur, as does simple regression to the mean and backlash to the frontrunner - Silver's models use varying ways to account for this, mostly based on historical comparison, and it's still a 90%+ shot for an Obama victory.

Additionally, we can't account for racism, and that will certainly play a role in the actual lever-pull moment for some portion of the population, but unless there is something incredible that happens, you're looking at a massive edge for Obama and all of the momentum going his direction. That conclusion certainly passes the smell test for me, even accounting for psychological effects of polling and similar.

[UGAalum94]

Originally Posted by KSig RC

All of this does occur, as does simple regression to the mean and backlash to the frontrunner - Silver's models use varying ways to account for this, mostly based on historical comparison, and it's still a 90%+ shot for an Obama victory.

Additionally, we can't account for racism, and that will certainly play a role in the actual lever-pull moment for some portion of the population, but unless there is something incredible that happens, you're looking at a massive edge for Obama and all of the momentum going his direction. That conclusion certainly passes the smell test for me, even accounting for psychological effects of polling and similar.

Yeah, I don't know that McCain can win although I'm still hoping, but as I said, I don't think the 14 point lead from last week will match up with reality.

It was the size of the lead that seemed so far off. It's apparently narrowing this week.

And maybe it's because I'm supporting McCain but I don't think the reluctance to own up to supporting him is even predominately a reflection of racism on the part of McCain voters.

I think many people who plan to vote for McCain aren't talking about it because it's not worth feeling like you have to defend yourself to a bunch of people whose political opinions you may not value.

Even though race is clearly involved in the Bradley effect, I'm not sure that the reason that people vote the way they do (in elections where its affect has been assumed) is particularly attributable to racism by the voters who vote against the black candidate. It may have more to do with the fear of having your motivation judged to be racist when you are asked in advance if you support the black candidate. It's far easier and I think pretty common to in a lot of, even non-political, instances to offer public approval for figures who you might be judged to be a racist if you publicly disapprove of, even if you have specific and non-race based reasons for your disapproval.

There's absolutely no risk in showing approval and a big risk of seeming racist in disapproving.

[KSig RC]

Originally Posted by UGAalum94

Yeah, I don't know that McCain can win although I'm still hoping, but as I said, I don't think the 14 point lead from last week will match up with reality.

It was the size of the lead that seemed so far off. It's apparently narrowing this week.

I have no reason to doubt that it was an accurate reflection of reality as of last week - the race always tightens at the very end, that's a historical trend with many and varied explanations.

Originally Posted by UGAalum94

And maybe it's because I'm supporting McCain but I don't think the reluctance to own up to supporting him is even predominately a reflection of racism on the part of McCain voters.

Remember, I didn't make this connection - you did.

I'm saying that, in addition to the other polling biases that exist, there will be an unprecedented lever-pull moment effect related to race - we don't know how significant this will be, or how many it will affect, we simply know it is likely to exist.

Originally Posted by UGAalum94

I think many people who plan to vote for McCain aren't talking about it because it's not worth feeling like you have to defend yourself to a bunch of people whose political opinions you may not value.

Even though race is clearly involved in the Bradley effect, I'm not sure that the reason that people vote the way they do (in elections where its affect has been assumed) is particularly attributable to racism by the voters who vote against the black candidate. It may have more to do with the fear of having your motivation judged to be racist when you are asked in advance if you support the black candidate. It's far easier and I think pretty common to in a lot of, even non-political, instances to offer public approval for figures who you might be judged to be a racist if you publicly disapprove of, even if you have specific and non-race based reasons for your disapproval.

There's absolutely no risk in showing approval and a big risk of seeming racist in disapproving.

All of this may or may not be true - I'm not sure of any sort of substantive research on voter satisfaction with their own choices and how it is reflected in a willingness to answer an anonymous survey (at least beyond historical comparisons, a la 538.com) - but again, these effects should bear themselves out in the comparative and historical data, regardless, right?

[UGAalum94]

Originally Posted by KSig RC

I have no reason to doubt that it was an accurate reflection of reality as of last week - the race always tightens at the very end, that's a historical trend with many and varied explanations.



Remember, I didn't make this connection - you did.

I'm saying that, in addition to the other polling biases that exist, there will be an unprecedented lever-pull moment effect related to race - we don't know how significant this will be, or how many it will affect, we simply know it is likely to exist.



All of this may or may not be true - I'm not sure of any sort of substantive research on voter satisfaction with their own choices and how it is reflected in a willingness to answer an anonymous survey (at least beyond historical comparisons, a la 538.com) - but again, these effects should bear themselves out in the comparative and historical data, regardless, right?

I don't know. Again, I think there are too many new variables to know for sure. If we accept what you've said about an unprecedented amount of race related level-pulling effect, isn't it going to be particularly hard to measure in advance?

But I didn't even realize I had made the racism connection that you commented on, so what do I know. (Do you simply mean bringing up the Bradley/Wilder Effect? When did I make that connection before you did?)

[KSig RC]

Originally Posted by UGAalum94

I don't know. Again, I think there are too many new variables to know for sure. If we accept what you've said about an unprecedented amount of race related level-pulling effect, isn't it going to be particularly hard to measure in advance?

YES, which is why I noted it was "unprecedented" and "unmeasurable."

However, other effects that you noted are completely wrapped within other elections, and thus make for a worthy comparison study.

Remember - supposedly unmeasurable race/gender effects were rampant throughout the Democratic primaries, and Silver's model did a fantastic job mapping those (better than every mainstream media source). So . . . yeah. Feel free to wonder about the numbers, but there's a good chance your Spidey-Sense is tingling for reasons other than a lack of numerical or scientific accuracy of the assay.