Stve Cohen:Why polls tend to undercount democrats
Very good article that spells it out why the polling is the way it is favoring republicans most of the time http://www.nytimes.com/2014/10/30/upshot/why-polls-tend-to-undercount-democrats.html?abt=0002&abg=1
= new reply since forum marked as read
Highlight:
NoneDon't highlight anything
5 newestHighlight 5 most recent replies
tomm2thumbs
(13,297 posts)'No race illustrates the issue better than the Nevada Senate race in 2010. Senator Harry Reid easily won re-election, by 6 points, after trailing in every nonpartisan poll conducted over the final few weeks of the race. It was one of the biggest general-election upsets in recent memory, but it is perhaps less surprising when viewed in light of polling’s challenges today.'
yup yup!
NewJeffCT
(56,828 posts)would be nice if we could pick get the seat in Iowa and maybe one other...
1StrongBlackMan
(31,849 posts)in terms of "modeling problems" ... I will state for the record, this married, highly educated (higher than the national and local average), non-felony conviction having, 50-something, African-American male, who has voted in every election since turning 18 and has, both, a cellphone AND a landline, and that lives in a purplish segment of a red state, has NEVER been contacted to be a part of a national polling ... and, to my knowledge, neither has my demographicaly-similar wife.
JustAnotherGen
(31,811 posts)I live in a bright red district.
And I'm a 41 year old female.
I canvass, GOTV, take people to the polls . . . but they never poll me.
1StrongBlackMan
(31,849 posts)Who is this "they", they keep saying speaks for us/me.
JustAnotherGen
(31,811 posts)It's like - we are almost invisible - you know?
1StrongBlackMan
(31,849 posts)
BronxBoy
(2,286 posts)JustAnotherGen
(31,811 posts)Between 10 percent and 12 percent of people with cellphones live in a different state from the one suggested by their telephone number’s area code. Those people are unlikely to be surveyed in statewide polls in the state they live in using random-digit dialing – the method used in most polls sponsored by nonpartisan news media organizations.
Common sense suggests these young, mobile voters lean Democratic. The earliest evidence is also consistent with that hypothesis: Pew Research found that people with telephone numbers that did not match their state leaned Democratic by 14 points, in a new analysis based on 5,003 cellphone respondents, provided to The Upshot. Cellphone users with telephone numbers that matched their location leaned Democratic by only eight points
That was me in 2006 - when I still had a 585 cell phone. I was an outlier in Central NJ.
1StrongBlackMan
(31,849 posts)poll modeling problems.
I wrote this last week:
the REPORTING ON the Mid-term Polling.
I caught a segment of Morning Joe this morning (in between making the coffee and watching my dogs on their morning romp from the patio) and I noticed two somethings: First, the inconsistent use of populations ... in one graphic, the crew was talking up a 13 point "battleground" republican advantage among "VERY LIKELY VOTER"; whereas, most/all of the other races were reported as "LIKELY VOTERS", and these races were all within the MoE.
That brings me to my second something ... I have been pondering the inaccuracy of polling since 2008, and have concluded that the polling organizations merely (use historical voting patterns) and build their prediction on a guessimation of turn-out. The polling groups, in each year, were very close on white turn-out, not close on youth turn-out, and flat out wrong on Black turn-out. While I haven't run the numbers, my guess is these flawed turn-out models account for most of the inaccuracy.
Which brings me to my larger something point ... there is no way to predict a race using merely "LIKELY VOTERS", much less, VERY LIKELY VOTERS", because there is no way to do anything but guess as to how many "LIKELY VOTERS" will actually go out and vote; nor, is there anyway to predict how many "LESS THAN LIKELY VOTERS", will come out to vote.
We (and polling outfits) should understand, GOTV efforts really do work, as evidenced by the "surprises"; and ... PREDICTIONS BASED ON POLLING IS JUST B.S. COMMENTARY.
So, I am predicting (or more accurately, hoping like hell, i.e., MY B.S. commentary) that the MS, NC, GA and VA turn-out pattern will hold, and Democrats will hold the Senate and pick up a few (but no where near enough) seats in the House, with a few surprising pick-ups in Democratic Governorship. And it appears that the gop, suspects this is true, as well, thus explaining the active vote/voter suppression efforts.
Leith
(7,809 posts)About 20 years ago, I lived & worked in Arlington, VA. At the time, I smoked so I stood in a courtyard at the front of the building. In a nearby building, there was a polling company that sent young people out with clipboards to conduct quick interviews with people on the street who had 5 minutes to spare (a smoker on break is a good subject).
One poll was one where the pollster read a commercial ad from the copy, then ask questions to see how much the listener remembered. The interesting one was political. There was a series of questions with an obvious political bent. After a few questions, the available answers got more and more skewed until there was one that didn't have a response that I could agree with at all and I said so. The pollster said (I swear this is true), "well, let's just choose this one." It was as far from my opinion as a response could get and he checked it off before I could stop him.
That ticked me off. I told him not to decide my responses and I ended the interview immediately.
That is my experience with polling.
About the Harry Reid election: I live in Nevada and I was very pleasantly surprised by the outcome. The only thing I can think is that people who were all rah-rah for Sharron "Obtuse" Angle before the election had some serious second thoughts in the voting booth. That woman exposed herself to be a clueless idiot in every way; Harry Reid was the Senate Majority Leader of our small state. People came to the last second realization that Angle could and would do nothing for us while Give'em Hell Harry was a fighter for us.
Midnight Writer
(21,745 posts)then it will be a bad night for Republicans.
One possible reason for the skew may be that older folks (who trend Republican) take the time to answer their phone and talk to length with a pollster. When my phone rings, I check the caller ID and if it is not someone I know, I don't bother to answer. On the other hand, when my 78 year old brother's phone rings, he will rush to pick up and, even if it's a telemarketer, will happily discuss everything from the weather to his recent hernia operation.
Another "skew" point is the "likely voter" aspect. It is up to the caller to decide whether or not the respondent is a "likely voter" or not according to the responses and past voting record. Many new voters are not considered "likely" because they have no past record of voting (it's like trying to get credit when you have no credit history).