Betsy's Page
HOME
Comments from an AP history
and government teacher in Raleigh, NC.
e-mail betsynewmark AT gmail.com
Commissions earned from selling items through Amazon will go towards buying materials for my classes. Thank you.
Site Feed
Betsy neither necessarily uses,
nor endorses,
the products advertised on this site.
The truth about Avesil
Cheap Hosting
Atlanta Bankruptcy Attorney
Dallas Bankruptcy Attorney
Wikio
Get exclusive travel deals and book discount cheap flights
Online Bachelors Degree
Comments from an AP history
and government teacher in Raleigh, NC.
e-mail betsynewmark AT gmail.com
Commissions earned from selling items through Amazon will go towards buying materials for my classes. Thank you.
Site Feed
Monday, October 13, 2008
Polling is an art, not a science
Mark Blumenthal has a column up looking at three great unknowns in polling this year. Pollsters are still trying to figure out how to account for those who use only cell phones. Then there is the so-called Bradley Effect hypothesizing that white respondents tell pollsters that they'll vote for the African-American candidate but then, in the privacy of the voting booth, they vote for the white candidate. And then there are all the questions about what the turnout will be this year. Will there be the great increase in turnout among young voters inspired to go vote by Obama? As Blumenthal concludes, perhaps some of those uncertainties will cancel each other out?Meanwhile, we can probably dispense with the "Perfect Storm" analogy. In the movie of the same name, three different weather-related phenomena combined to produce a storm of exceptional severity. In this case, as Democratic strategist Joe Trippi pointed out in September, the potential polling foibles may work in opposite directions and "cancel each other out." A return of the Bradley-Wilder effect would work to McCain's benefit, while an underrepresentation of younger, African American or "cell-phone-only" voters will likely benefit Obama.So how do pollsters account for all these uncertainties? They will only know if their models are correct after we actually vote. All this seeming exactitude gives the impression that they're measuring in a very scientific manner. But that is just an illusion based on how they designed their model and what assumptions they pour into it.
Gallup has basically thrown up its hands and are now giving out two different results depending on a model that predicts likely voters based on the questions and modeling that they have usually used basing the likelihood of someone turning out partly on their past behavior. By that model Obama leads by 4%. Then they posit another model accepting the responses of people who say that they are likely to vote even if they have not previously voted. By that model, Obama leads by 6%. Throw in a little margin of error and it's all about the same. They don't really know, but it all sounds very precise.
I'm not one of these people who cling to hopes that the polls are all off because of assumptions on the partisan turnout vastly overestimating Democratic turnout in relation to Republican turnout. Though this poll by Newsweek seems particularly egregious as Jim Geraghty points out. I remember Republicans hopefully clinging to that argument in the weeks leading up to the 2006 election only to find out that things were really as bad as the polls were showing. So Rasmussen may well be correct when they posit a 6.3% gap between Democratic and Republican turnout for their model. But who knows? It's fun to think that we can detect the ups and downs of the campaigns with such precision, just as long as we remember that, as Donald Rumsfeld might say, there are unknowns that we don't know.
UPDATE: And V. Lance Tarrance, who was the pollster for George Deukmejian when he defeated Tom Bradley for governor of California thus initiating the whole theory of a "Bradley Effect" argues that there never was such an effect. Most of this mythology grew up out of overly optimistic polling done by Mervin Field who had predicted a Bradley win and then when Bradley didn't win came up with the theory of white voters not telling the truth to pollsters, particularly in his exit polls that showed Bradley winning. Tarrance says that Bradley actually won on election day, but Deukmejian won on the absentee ballots and that ended up determining the election. We were living in Los Angeles at that time and have ever since referred to the Field poll with ironic air quotes as "the respected Field poll" since it seemed that almost ever mention of the Field poll had that adjective in front of it.Labels: Politics
0 comments
Comments:
Mark Blumenthal has a column up looking at three great unknowns in polling this year. Pollsters are still trying to figure out how to account for those who use only cell phones. Then there is the so-called Bradley Effect hypothesizing that white respondents tell pollsters that they'll vote for the African-American candidate but then, in the privacy of the voting booth, they vote for the white candidate. And then there are all the questions about what the turnout will be this year. Will there be the great increase in turnout among young voters inspired to go vote by Obama? As Blumenthal concludes, perhaps some of those uncertainties will cancel each other out?
Gallup has basically thrown up its hands and are now giving out two different results depending on a model that predicts likely voters based on the questions and modeling that they have usually used basing the likelihood of someone turning out partly on their past behavior. By that model Obama leads by 4%. Then they posit another model accepting the responses of people who say that they are likely to vote even if they have not previously voted. By that model, Obama leads by 6%. Throw in a little margin of error and it's all about the same. They don't really know, but it all sounds very precise.
I'm not one of these people who cling to hopes that the polls are all off because of assumptions on the partisan turnout vastly overestimating Democratic turnout in relation to Republican turnout. Though this poll by Newsweek seems particularly egregious as Jim Geraghty points out. I remember Republicans hopefully clinging to that argument in the weeks leading up to the 2006 election only to find out that things were really as bad as the polls were showing. So Rasmussen may well be correct when they posit a 6.3% gap between Democratic and Republican turnout for their model. But who knows? It's fun to think that we can detect the ups and downs of the campaigns with such precision, just as long as we remember that, as Donald Rumsfeld might say, there are unknowns that we don't know.
UPDATE: And V. Lance Tarrance, who was the pollster for George Deukmejian when he defeated Tom Bradley for governor of California thus initiating the whole theory of a "Bradley Effect" argues that there never was such an effect. Most of this mythology grew up out of overly optimistic polling done by Mervin Field who had predicted a Bradley win and then when Bradley didn't win came up with the theory of white voters not telling the truth to pollsters, particularly in his exit polls that showed Bradley winning. Tarrance says that Bradley actually won on election day, but Deukmejian won on the absentee ballots and that ended up determining the election. We were living in Los Angeles at that time and have ever since referred to the Field poll with ironic air quotes as "the respected Field poll" since it seemed that almost ever mention of the Field poll had that adjective in front of it.
0 comments
Mark Blumenthal has a column up looking at three great unknowns in polling this year. Pollsters are still trying to figure out how to account for those who use only cell phones. Then there is the so-called Bradley Effect hypothesizing that white respondents tell pollsters that they'll vote for the African-American candidate but then, in the privacy of the voting booth, they vote for the white candidate. And then there are all the questions about what the turnout will be this year. Will there be the great increase in turnout among young voters inspired to go vote by Obama? As Blumenthal concludes, perhaps some of those uncertainties will cancel each other out?
Meanwhile, we can probably dispense with the "Perfect Storm" analogy. In the movie of the same name, three different weather-related phenomena combined to produce a storm of exceptional severity. In this case, as Democratic strategist Joe Trippi pointed out in September, the potential polling foibles may work in opposite directions and "cancel each other out." A return of the Bradley-Wilder effect would work to McCain's benefit, while an underrepresentation of younger, African American or "cell-phone-only" voters will likely benefit Obama.So how do pollsters account for all these uncertainties? They will only know if their models are correct after we actually vote. All this seeming exactitude gives the impression that they're measuring in a very scientific manner. But that is just an illusion based on how they designed their model and what assumptions they pour into it.
Gallup has basically thrown up its hands and are now giving out two different results depending on a model that predicts likely voters based on the questions and modeling that they have usually used basing the likelihood of someone turning out partly on their past behavior. By that model Obama leads by 4%. Then they posit another model accepting the responses of people who say that they are likely to vote even if they have not previously voted. By that model, Obama leads by 6%. Throw in a little margin of error and it's all about the same. They don't really know, but it all sounds very precise.
I'm not one of these people who cling to hopes that the polls are all off because of assumptions on the partisan turnout vastly overestimating Democratic turnout in relation to Republican turnout. Though this poll by Newsweek seems particularly egregious as Jim Geraghty points out. I remember Republicans hopefully clinging to that argument in the weeks leading up to the 2006 election only to find out that things were really as bad as the polls were showing. So Rasmussen may well be correct when they posit a 6.3% gap between Democratic and Republican turnout for their model. But who knows? It's fun to think that we can detect the ups and downs of the campaigns with such precision, just as long as we remember that, as Donald Rumsfeld might say, there are unknowns that we don't know.
UPDATE: And V. Lance Tarrance, who was the pollster for George Deukmejian when he defeated Tom Bradley for governor of California thus initiating the whole theory of a "Bradley Effect" argues that there never was such an effect. Most of this mythology grew up out of overly optimistic polling done by Mervin Field who had predicted a Bradley win and then when Bradley didn't win came up with the theory of white voters not telling the truth to pollsters, particularly in his exit polls that showed Bradley winning. Tarrance says that Bradley actually won on election day, but Deukmejian won on the absentee ballots and that ended up determining the election. We were living in Los Angeles at that time and have ever since referred to the Field poll with ironic air quotes as "the respected Field poll" since it seemed that almost ever mention of the Field poll had that adjective in front of it.
Labels: Politics
Comments:
Post a Comment
