![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
osewalrusFor the last several special elections, we have seen the Democrats averaging +16 on voter turn out. That is to say, using the 538.com model, Democrats have, on average, performed 16 points better in the special election than one would predict based on the overall partisan lean of the district.
That's notable because the generic ballot gives Dems about 6.5-7.5 point advantage over the last few months (again, using 538.com).
The difference is important because it is the difference between a standard mid-term where the party in power loses seats and a wave election that would give Dems the House (and possibly the Senate). I am guessing that the difference reflects the difference in voter enthusiasm. Voter enthusiasm is always difficult to predict. If the difference is significant, it can play Hell with the projected voter formula. I think this last is at play.
Keep in mind that polls aren't just a random sample of the population. Every polling firm has its secret sauce of how it conducts its survey to try to accurately reflect the population likely to turn out to vote. This is true even for polls of total registered voters, not just polls of likely voters. Registered voters are polled in a way that reflects the registration pool. So if Dems have an advantage in registered voters, even if the state largely votes Republican (West VA is an example of this), the poll will contain a larger number of registered Democrats to reflect this reality.
I'm also seeing the occasional wild fluctuation in specific polls, which I suspect reflects polling firms tinkering with their formula. There is a lot of concern about getting it right and making reasonably accurate predictions in light of past polling failures and the increasing overall volatility of the electorate. (As I like to say, 'volatile electorates are volatile.')
All of which suggests that this will be much more like 2006 and 2014, where it was uncertain even up to election day how things would fall out, than like 2008 and 2010, when it was obvious there was a "wave election" and the question was how big. This matches my highly impressionistic sense of the electorate. In 2006 and 2014 overall economic numbers were good, but lots of people were hurting from wage stagnation and other issues. It wasn't that lots of people who had traditionally voted for one party flipped, but that lots of people who traditionally voted for the party in power stayed home. Which is why you get the split between the generic ballot and the actual election outcomes.
Another thing to note in the recent special elections, which is again consistent with 2006 and 2014 rather than 2008 and 2010, is that polls tightened considerably in the last few weeks before the election with lots of undecideds breaking at the last minute for the Democrat.
All of which is to say, it is going to be an interesting ride, with the outcome remaining largely unpredictable until the actual election.
That's notable because the generic ballot gives Dems about 6.5-7.5 point advantage over the last few months (again, using 538.com).
The difference is important because it is the difference between a standard mid-term where the party in power loses seats and a wave election that would give Dems the House (and possibly the Senate). I am guessing that the difference reflects the difference in voter enthusiasm. Voter enthusiasm is always difficult to predict. If the difference is significant, it can play Hell with the projected voter formula. I think this last is at play.
Keep in mind that polls aren't just a random sample of the population. Every polling firm has its secret sauce of how it conducts its survey to try to accurately reflect the population likely to turn out to vote. This is true even for polls of total registered voters, not just polls of likely voters. Registered voters are polled in a way that reflects the registration pool. So if Dems have an advantage in registered voters, even if the state largely votes Republican (West VA is an example of this), the poll will contain a larger number of registered Democrats to reflect this reality.
I'm also seeing the occasional wild fluctuation in specific polls, which I suspect reflects polling firms tinkering with their formula. There is a lot of concern about getting it right and making reasonably accurate predictions in light of past polling failures and the increasing overall volatility of the electorate. (As I like to say, 'volatile electorates are volatile.')
All of which suggests that this will be much more like 2006 and 2014, where it was uncertain even up to election day how things would fall out, than like 2008 and 2010, when it was obvious there was a "wave election" and the question was how big. This matches my highly impressionistic sense of the electorate. In 2006 and 2014 overall economic numbers were good, but lots of people were hurting from wage stagnation and other issues. It wasn't that lots of people who had traditionally voted for one party flipped, but that lots of people who traditionally voted for the party in power stayed home. Which is why you get the split between the generic ballot and the actual election outcomes.
Another thing to note in the recent special elections, which is again consistent with 2006 and 2014 rather than 2008 and 2010, is that polls tightened considerably in the last few weeks before the election with lots of undecideds breaking at the last minute for the Democrat.
All of which is to say, it is going to be an interesting ride, with the outcome remaining largely unpredictable until the actual election.
no subject
Date: 2018-08-13 04:31 pm (UTC)Do you think polls may be getting less accurate because fewer people have landlines and/or more people have caller ID and don't answer the calls?
no subject
Date: 2018-08-13 08:21 pm (UTC)