Date: Thu, 10 Jan 2008 22:37:21 +0530
Reply-To: ajay ohri <ohri2007@GMAIL.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: ajay ohri <ohri2007@GMAIL.COM>
Subject: Re: Slightly OT: New Hampshire Polling Fiasco
Content-Type: text/plain; charset=ISO-8859-1
Now that Richardson has dropped out , I was just wondering how it will
statistically affect the results.
Also could you pin a weightage to a specific event based on pre event and
post event voting probability (like Senator crying).
On 1/10/08, Swank, Paul R <Paul.R.Swank@uth.tmc.edu> wrote:
> There was an op-ed piece on this in the New York Times this morning. The
> writer, a former pollster, claims that it was nonresponse bias in that more
> poor whites voted for Clinton than Obama and that they are more likely to
> refuse to respond to a poll.
> Paul R. Swank, Ph.D.
> Professor and Director of Research
> Children's Learning Institute
> University of Texas Health Science Center - Houston
> -----Original Message-----
> From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of Peter
> Sent: Thursday, January 10, 2008 10:26 AM
> To: SAS-L@LISTSERV.UGA.EDU
> Subject: Re: Slightly OT: New Hampshire Polling Fiasco
> I think it is largely due to nonresponse bias.
> A huge proportion of people are simply refusing to answer polls (according
> to Pollster.com, a 20% response rate is GOOD (yeah, real good).
> So, if Hillary supporters are more likely to screen out calls.....that
> would do it. There is reason to think they might be more likely. For one
> thing, more of Hlllary's support was 'soft', and I would guess that soft
> supporters are more likely to screen calls
> -----Original Message-----
> >From: Mary <mlhoward@AVALON.NET>
> >Sent: Jan 10, 2008 11:17 AM
> >To: SAS-L@LISTSERV.UGA.EDU
> >Subject: Slightly OT: New Hampshire Polling Fiasco
> >Was wondering if any of you had any comments about how the polling was so
> wrong in New Hampshire- polling firms such as Zogby had the results as 42%
> Obama and 29% Clinton when it wound up being 39% Clinton and 36% Obama.
> >Some suggestions have been:
> >1. The weightings used to correct for undersampling did not adequately
> do so, particularly for women over 40 of lower income.
> >2. The anticipation that large numbers of college students would turn
> out did not materialize, perhaps because there aren't as many college
> students in New Hampshire as there were in Iowa, or because estimates of
> turnout among them were overstated.
> >3. The "Iowa bounce" was picked up in the polls, but not the deflation
> of that bounce that tends to happen about 4 days later.
> >4. People lied and told pollsters that they were going to vote for Sen.
> Obama when they really voted for Sen. Clinton.
> >5. People based their decisions on who to vote for based on the polls
> themselves; that they may have decided to vote for Sen. Clinton or go to the
> Republican side to vote for Sen. McCain after seeing that Sen. Obama had a
> comfortable lead in the polls.
> >Coming from Iowa, I tend to believe the last scenerio; I had been polled
> several times before the election and had said I would be voting for Bill
> Richardson, but BECAUSE of his low poll numbers I actually did not vote for
> him, but for Sen. Obama instead.
> >Any comments? The idea that voters actually make decisions based on
> polls could have implications in many areas of sample survey research, so I
> am wondering if the statisticians out there or others have any comments on
> what went so wrong with the polling.
> > Univ. of Iowa