The Experts Ponder Their Polls: GOTV, Big Mo
In the wake of Tuesday's election in which Jari Askins upset Drew Edmondson in a close finish that all the polls failed to detect, those involved in taking the polls are examining what may have happened. (See story below.)
Analysis
The most likely answer, some agree, is an aggressive "get out the vote" effort that can't be measured in a pre-balloting poll and can skew even the most accurate poll result.
Keith Gaddie of SoonerPoll says of what may have happened, "To that end, let's talk theories: (1) The poll aged. A likely culprit that you bring up, and correctly so. (2) The fact that all the polls tracked together, the wrong way, means that they either all shared an aging flaw, a sampling flaw, or that some other factor rendered them inaccurate. (3) The third question is whether or not the sampling flaw was a function of a bad sample (and therefore afflicting all three polls I know of); a screening flaw (leading to an unrepresentative sample across all polls); or, was it a product of campaign activity that broke the assumptions of the likely voter models used by all the pollsters in play in Oklahoma.
"This last theory is probably the most powerful. As we increase propensity to vote in our screen, Edmondson gets stronger. So, the Askins vote (actual) has to either be a function of conversion by Askins; demobilization of Edmondson voters (given the turnout, possible); and aggressive GOTV by Askins, thereby breaking the sample frame like happened in the NY gubernatorial election in 1994.
"I like this last theory, and am moving to test all three. Part of what I need are the sign-in data, so this will take a several days (maybe a couple of weeks) to get done."
Keith Gaddie of SoonerPoll says of what may have happened, "To that end, let's talk theories: (1) The poll aged. A likely culprit that you bring up, and correctly so. (2) The fact that all the polls tracked together, the wrong way, means that they either all shared an aging flaw, a sampling flaw, or that some other factor rendered them inaccurate. (3) The third question is whether or not the sampling flaw was a function of a bad sample (and therefore afflicting all three polls I know of); a screening flaw (leading to an unrepresentative sample across all polls); or, was it a product of campaign activity that broke the assumptions of the likely voter models used by all the pollsters in play in Oklahoma.
"This last theory is probably the most powerful. As we increase propensity to vote in our screen, Edmondson gets stronger. So, the Askins vote (actual) has to either be a function of conversion by Askins; demobilization of Edmondson voters (given the turnout, possible); and aggressive GOTV by Askins, thereby breaking the sample frame like happened in the NY gubernatorial election in 1994.
"I like this last theory, and am moving to test all three. Part of what I need are the sign-in data, so this will take a several days (maybe a couple of weeks) to get done."
Bill Shapard, also of SoonerPoll, said three factors are important: The voter turnout, the women's vote and the Barry Switzer endorsement of Askins.
Pat McFerron of the Sooner Survey cites several factors that could have resulted in poll results that weren't reflected in the actual voting. He notes, rightly, that voter turnout among Democrats in Tulsa (a perceived Edmondson area of strength) was below that of other regions, especially including Askins' perceived area of strength, southwestern Oklahoma.
McFerron said, "1) It was a snapshot in time. 2) Looks like SW really over-performed (turnout wise) and Tulsa was below. 3) Momentum matters, and Jari had it during the last week."
posted by Mike McCarville @ 6:50 PM
<< Home