Hearing through the noise
Sep. 19th, 2008 11:03 am![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
barking_iguanaThe state of the Presidential race is nearly unchanged from last week.
But, but, last week...
Now (and probably increasingly in the next few days)...
The reality is there are a large number of people who consistently tell pollsters they support McCain, there are an ever so slightly larger number of people who consistently tell pollsters they will vote for Obama, and there are a relatively small number of people who switch back and forth, depending on the last thing to move them.
Last week, the thing that had most recently moved them was McCain's speech at the GOP convention and the novelty of Palin (though not her speech). Now the last thing to move them is the mess in the financial sector of the economy, so they think some combination of 1) Democrats are better to deal with economic problems in a way won't hurt them and 2) the GOP got us into this mess, so it's time to change horses.
But to swing voters, September's economic news will also be history by the time the election rolls around. Something else—several something elses, actually, probably including the debates—will move them again between now and Election Day.
So we don't even have a strong idea who will be leading in opinion polls on November 3. And for a variety of reasons, the apparent state of the race that Monday may be quite different from the actual returns on Tuesday the 4th. There are very significant factors that push the actual results in either direction from what folks who put a lot of stock in the polls think will happen.
First, is that the GOP has been better in recent elections at getting its voters to the polls. Also favoring McCain is what's being called the Bradley or Wilder Effect. But the traditional Wilder Effect doesn't seem to exist anymore, judging by results within the past decade. That's where voters tell exit pollsters that they voted for a Black candidate when, in fact, they did not, presumably because they don't want to be perceived as racist.
There is no longer a measurable Wilder Effect, at least for offices lower than President, where we have no general election data. And there's no data from the primaries to support it, either. The names of the effect are now casually applied to the possibility that people are telling opinion pollsters the same thing they used to tell exit pollsters, presumably for the same reason. It's possible, but I'm skeptical. I think racism continues to gradually decline, but I also think certain kinds of racism, even while becoming less ubiquitous, carry less social sanction than they did 15 or 30 years ago. Those who don't trust Obama because he's Black may not tell pollsters the reason they won't vote for him, but I think they are saying they won't vote for him—and that's what counts when interpreting horse-race numbers.
On Obama's side, there is the fact that his supporters are disproportionately unreachable by land-line telephones, which is how almost all pollsters reach voters. Bigger, I think, is that Obama is putting a greater proportion of his money into field offices than any campaign since the ones I grew up with in the 1970s. I've been saying Democrats should do that for a long time; now we'll get an indicator of how well it works in the current era. Will the GOP advantage in GOTV be reversed? Lessened? Will the field offices have no effect? Ask me in mid-November.
There's also the fact that nobody really knows what voter turnout among the young will be. Obama's chances in close primary states were routinely denigrated because he was relying on that demographic which notoriously doesn't show up at the polls. But for Obama, they did. Every pollster has their own guess about how much of that will carry over into the general election, but nobody has better than a guess.
So overall, there's uncertainty about who will be leading the late opinion polls and also great uncertainty about how those polls do and don't reflect what will happen in the election.
But, but, last week...
- McCain was up in most polls
- Democrats trying to insulate themselves from despair in November were despairing in September instead
- the noise box was talking about how effective the Palin pick was at winning swing voters.
Now (and probably increasingly in the next few days)...
- Obama is leading in most polls (by a slightly larger margin than McCain was last week)
- Republicans are saying McCain is Bob Dole
- the rube tube is noting that Palin has energized the theocrats (charitably referred to as 'the base', of which they are only a part, but which phrase translated into Arabic shows their true nature) but has no perceptible effect on those who might vote for Obama and might not.
The reality is there are a large number of people who consistently tell pollsters they support McCain, there are an ever so slightly larger number of people who consistently tell pollsters they will vote for Obama, and there are a relatively small number of people who switch back and forth, depending on the last thing to move them.
Last week, the thing that had most recently moved them was McCain's speech at the GOP convention and the novelty of Palin (though not her speech). Now the last thing to move them is the mess in the financial sector of the economy, so they think some combination of 1) Democrats are better to deal with economic problems in a way won't hurt them and 2) the GOP got us into this mess, so it's time to change horses.
But to swing voters, September's economic news will also be history by the time the election rolls around. Something else—several something elses, actually, probably including the debates—will move them again between now and Election Day.
So we don't even have a strong idea who will be leading in opinion polls on November 3. And for a variety of reasons, the apparent state of the race that Monday may be quite different from the actual returns on Tuesday the 4th. There are very significant factors that push the actual results in either direction from what folks who put a lot of stock in the polls think will happen.
First, is that the GOP has been better in recent elections at getting its voters to the polls. Also favoring McCain is what's being called the Bradley or Wilder Effect. But the traditional Wilder Effect doesn't seem to exist anymore, judging by results within the past decade. That's where voters tell exit pollsters that they voted for a Black candidate when, in fact, they did not, presumably because they don't want to be perceived as racist.
There is no longer a measurable Wilder Effect, at least for offices lower than President, where we have no general election data. And there's no data from the primaries to support it, either. The names of the effect are now casually applied to the possibility that people are telling opinion pollsters the same thing they used to tell exit pollsters, presumably for the same reason. It's possible, but I'm skeptical. I think racism continues to gradually decline, but I also think certain kinds of racism, even while becoming less ubiquitous, carry less social sanction than they did 15 or 30 years ago. Those who don't trust Obama because he's Black may not tell pollsters the reason they won't vote for him, but I think they are saying they won't vote for him—and that's what counts when interpreting horse-race numbers.
On Obama's side, there is the fact that his supporters are disproportionately unreachable by land-line telephones, which is how almost all pollsters reach voters. Bigger, I think, is that Obama is putting a greater proportion of his money into field offices than any campaign since the ones I grew up with in the 1970s. I've been saying Democrats should do that for a long time; now we'll get an indicator of how well it works in the current era. Will the GOP advantage in GOTV be reversed? Lessened? Will the field offices have no effect? Ask me in mid-November.
There's also the fact that nobody really knows what voter turnout among the young will be. Obama's chances in close primary states were routinely denigrated because he was relying on that demographic which notoriously doesn't show up at the polls. But for Obama, they did. Every pollster has their own guess about how much of that will carry over into the general election, but nobody has better than a guess.
So overall, there's uncertainty about who will be leading the late opinion polls and also great uncertainty about how those polls do and don't reflect what will happen in the election.
![[identity profile]](https://www.dreamwidth.org/img/silk/identity/openid.png){kind=small}
no subject
Date: 2008-09-19 05:37 pm (UTC)no subject
Date: 2008-09-19 05:43 pm (UTC)Wow. This meme just seems to be unquashable.
Most pollsters now use 'random number dialing' which reaches cell phone
users (it is not illegal for political campaigns; I believe it is for
commercial apps).
Even when they weren't (or aren't) there seems to be only a tiny difference
between those of a given demographic (like 18-25 year olds) who have cell
phones vs. those who don't.
If there is any effect along these lines, I would suspect it might be
amongst people who are basically unreachable -- they don't have a phone
or a place to live perhaps any longer because of the economy. This is
presumably usually not a significant set of people, because they probably
don't vote much. But right now, it might be a lot bigger, and those
people new to this state of woe are more likely to vote, since presumably
they have in the past. This could maybe 1/2% or something in Obama's
favor.
no subject
Date: 2008-09-20 01:28 pm (UTC)http://www.fivethirtyeight.com/2008/09/estimating-cellphone-effect-22-points.html
no subject
Date: 2008-09-20 03:26 pm (UTC)I'm not sure I understand why he's doing what he's doing.
It seems like he's taking 184 polls by Gallup and giving
them collectively the same weight as the 3 polls, collectively,
by TIME/SRBI.
I thought I had read that almost all pollsters were accessing
cellphones now, but I guess not.
So insofar as Silver's methodology makes sense, and the
data on which pollsters do and don't access cell phones is accurate,
I stand corrected.
I do recall reading just lately that one of the reasons the Kerry
campaign initially thought they had won in 2004 was they had factored
in additional votes due to this factor, and then it turned out
not to be the case (or something else was screwy).
no subject
Date: 2008-09-20 06:16 pm (UTC)Weighting by sqrt(n), I get a factor of 1.5.
p.s.
Date: 2008-09-19 05:44 pm (UTC)Re: p.s.
Date: 2008-09-20 01:29 pm (UTC)