Or, Maybe Primary Polls Just Suck

by: Chris Bowers

Thu Jan 10, 2008 at 19:44


Final seven-poll New Hampshire average in 2000 (results in parenthesis):

McCain: 40.3% (48.5%)
Bush: 31.7% (30.3%)

Gore: 51.1% (49.7%)
Bradley: 41.9% (45.6%)

Eight years ago, the final polls were off by 9.6% in the Republican campaign, and 5.1% in the Democratic campaign. That is about as bad as the polls performed this time around. As the National Council on Public Polls wrote eight years ago:

Once again the polls published before the New Hampshire primary demonstrated the difficulty of forecasting a primary. For most of his career George Gallup shunned pre-primary polling. Wise man! The problems are almost insurmountable.

Mystery Pollster has more on how large polling error of this sort is not unprecedented, especially in the New Hampshire primary. Really, with the myriad of explanations flying around as to why the polls were wrong, including my own explanations, perhaps the simplest reason is the correct one: it is just as difficult to poll primaries as it ever was. Primaries are much more difficult to poll than general elections for many reasons, and as time goes on we don't appear any closer to solving the problems associated with polling primaries.

Chris Bowers :: Or, Maybe Primary Polls Just Suck

Tags: , , (All Tags)
Print Friendly View Send As Email
So why don't they re-poll? (0.00 / 0)
I was asking this over at pollster.com.  Why don't they re-survey the same people they polled before the election, and ask them how they in fact voted?  Whether the results match the election or the pre-election poll, either way that would shed a lot of light on the source of the error.  Why don't people do this?

it's called a panel-back (0.00 / 0)
Pollsters do run panel-backs, but it is expensive and there are privacy issues. I saw some public outfit had run a panel-back recently but I don't remember who it was. Private groups use panel-backs more often, they are a more precise than trends for measuring what contacts are effective.

I'm sure the campaigns have reviewed their repeat voter-ids, and probably done a follow up on a sample, to see where the motion was. They probably have a pretty clear idea of where the public polls messed up by now, but they have little incentive to share their work.


[ Parent ]
A couple of thoughts (0.00 / 0)
What about the undecideds?  What were these numbers in each of these polls?  With McCain (2000), Bradley and Clinton (2008), the bigger number that changed was a gain, so if undecideds broke in favor of that candidate, that isn't really any error in the previous polls, that would be simply people shifting from the "nobody" category to that particular candidate.

Also, what's the margin of error on these polls?  When the polls are published, they tell you that there is a chance that the poll could be off by such and such amount.  Maybe the problem wasn't with the polls, but with the way that pundits interpreted them.

I'm sure there were other factors, but these two strike me as pretty plausible.

http://flaprogressives.org


The problem with this analysis... (0.00 / 0)
is that the coalitions in NH represent the coalitions overall.

As Jay Cost notes, not only is this unlikely, but it's directly contradicted by the Iowa vote.  So only looking at NH and trying to extrapolate out to the country writ large seems flawed; it doesn't take into account Hispanics, it doesn't take into account African-Americans, and it doesn't take into regional cultural differences.

There's no reason to assume that, say, union and lower-income voters resemble culturally or in their political taste similar demographic voters in the mid-west.  The fact that Obama was raised by Kansan grandparents could have given him a distinctly mid-western appeal that worked in Iowa but didn't translate to NH.  Having lived in NH for 4 years, I can say with some level of confidence that they don't have much love for the mid-west up there. 

Cost's take on comparing cross-tabs for the two elections we've had so far:

The bottom line: Iowa and New Hampshire diverged in the fullest sense of the word. Not only did Iowa vote Obama and New Hampshire vote Clinton - identical demographic groups broke in opposite directions! This implies that the nomination contest is very much up in the air. The big question is who can put together more of the traditional Democratic voting coalition. Obama did in Iowa. Clinton did in New Hampshire.


Oh sorry! (0.00 / 0)
I posted this in the wrong thread!  My bad.

Will re-post in other thread, ignore here.


[ Parent ]
USER MENU

Open Left Campaigns

SEARCH

   

Advanced Search

QUICK HITS
STATE BLOGS
Powered by: SoapBlox