A final unskewing
Mother Nature is our best teacher and the only one who is always right. – Viktor Hamburger, biologist In yesterday’s Los Angeles Times...
Romney post-moocher watch, Day 2
A continuing look at how specific people, followed over time with repeat polling, are resp0nding after the release of the “moocher” tape from the Romney fundraiser. In RAND’s own words, this survey “may be more accurately capturing the likely votes of a greater number of voters in the crucial ‘middle’ (i.e., not closely aligned with either candidate).
The gray zone is the confidence band. As you can see, since the Democratic convention something has been brewing. After this, six more news cycles to go.
Topics:
The Gallup Obama-Romney tracking poll (7-day rolling avg) shows movement back toward a near tie (currently 47 O – 46 R).
I had expected at least a transient 47%-gaffe-driven bump in those numbers, as well.
I do like RAND’s approach a lot.
Speculation alert:
If the effect is mainly that the Dem base gets energized, would we see a bigger spread between robopolls and cell/internet polls? As we know, pollsters try their best to control for that. Still, this may be plausible if the population of landlined respondents is different from cell-onlies on a really unprecedented scale… and it’s only really showing up now. Would be kinda weird.
It would be at least partially represented in the likely voter screen, so I don’t think that the polls would badly miss the effect on turnout.
Relevant:
https://twitter.com/fivethirtyeight/status/248473421372080128
and
https://twitter.com/fivethirtyeight/status/248474257049395202
Do we know if RAND is collecting other information from the switchers, which might give insight into when/why they switch, or who is more/less likely to do so? These folks, along with those who make a decision to vote or not vote (in contrast to those who will always vote or never vote regardless of circumstances) seem to be the ones who decide the election with an electorate generally evenly divided.
Perhaps I should have said, more accurately, that those decisions (to switch candidate, or to vote/not vote) are what decide an election in an evenly divided electorate. So factors contributing to those two decisions are of intense interest, apart from the measurement of the magnitude of those shifts on the panel.
Probability distribution has an interesting bimodal shape today, with a gap exactly the size of Florida.
Yes, looks like an NMR spectrum. That split tells you FL is more up for grabs than nearly any other state.
It’s a stark effect (or is it a zeeman effect?)
But it stands out because there’s not a lot else going on of that magnitude. If a bunch of other big states were more in play we’d see it all smeared out. Though I do wonder what the heck is going on up the street in New Hampshire.
>wishes he could like this comment
>realizes the carnage that a like feature might bring
>unwishes quickly
JamesinCA: you mean like this? (https://mmicdata.rand.org/alp/index.php?page=election#results-by-respondents-characteristics)
Nice catch. Sorry I missed your comment before posting in the next thread. Florida is up for grabs, but doesn’t seem to matter at this point in the election, much like in 2008.
“Each day we present the data broken out by a different characteristic, such as age, education, income, position in the labor market, race, sex, and state (Pennsylvania, Florida, Ohio). Anyone interested to know more about the results broken down by respondent characteristics, or about other ways data from the survey can be accessed and used, is invited to contact Arie Kapteyn, director of RAND Labor and Population, at kapteyn@rand.org.”
Thanks! Look what I would have found by clicking on that other tab. It’s pretty much what I had in mind, though it’d be interesting to explore what other characteristics map to shifting votes. Which it looks like they do offer to make available.
Looking a little deeper into the RAND data – specifically the responses cut by education is very insightful. A majority of the move toward Obama is in High School and below category, with the lowest impact being seen in College Educated voters.
1) it suggests those most likely to be offended by the “47%” comment were indeed offended.
2) it suggests that this might be a less sustainable surge for Obama as this cohort has traditionally been less motivated voters and may jump out of the voter pool once the emotions of the “47%” comment fades
3) it suggests that more educated voters are more locked in to their preference. I’m sure the campaigns understand this as well which might explain both campaigns focusing less on wonky policy and more on high level principles and emotion. (At least I hope there is some explanation for the appalling lack of details from either side).
This means that we are much less likely to have “the event” that turns the election – the college educated bloc is going to disproportionately be represented part of the population eventually voting in November. (No matter what we political junkies hope – Ryan! Chair! Clinton! Libya! 47%!)
I like your discussion of point #3, I’d never considered that.
Chris, what a great point.
New 538 post shows a huge spread btwn cellphone and no cellphone polls: http://fivethirtyeight.blogs.nytimes.com/2012/09/19/obamas-lead-looks-stronger-in-polls-that-include-cellphones/
This matches a lot of what we’ve been seeing I think. An argument for bias ~=0 perhaps…
So….it looks like Rasmussen and Gravis are the outliers…..would Mr. Freeland say their stratification adjustment is wrong? Both are robopoll shops.