2016 Postmortem
Related: About this forumCommon misperception about exit polls is driving me nuts!
Most people imagine the scenario where voters are asked a series of questions in a face-to-face interview. That is the exception. If you're ever involved in an exit poll, you are far more likely to be given an anonymous confidential questionnaire to fill out that you then drop in a collection box.
I don't know whether more Trump voters would be inclined to brag about their vote or not be able to bring themselves to admit it to a stranger, but either way it's largely irrelevant. Confidential questionnaires remove the primary incentive to lie.
Well designed and executed exit polls predict election results with a high degree of reliability. This is precisely why the US has funded exit polls in troubled states to detect fraud. Election results have been overturned in countries like Serbia and the Ukraine because of exit poll data.
To dismiss them as irrelevant because people are too ashamed to admit they voted for the Tangerine Nightmare is based on an erroneous image of voters being tackled by a stranger with a clipboard as they exit their polling place. Can we please dispense with this view and stop perpetuating it?
ETA: In another thread, someone linked to an exit poll that used online and telephone methods. That is a type of exit poll I would not consider well designed and executed as it does not eliminate the biases that surveys conducted immediately upon exiting a polling place address. So that brings up another great point. All exit polls are not the same and should be looked at critically in terms of their methodology as well as their data.
SickOfTheOnePct
(7,290 posts)In general, I've not given much weight to exit polls, because I've always felt the sample was too small. However, having done a great deal of research the past couple of days, I've gained not only a better understanding of how and where they're conducted, but also how the reliability grows throughout the day, and most especially as the models are updated once the actual vote counts begin.
PufPuf23
(8,776 posts)1) Poorly designed,
2) Poorly administered, and/or
3) Inherently biased or fraudulent.
Exit polls are essentially worthless or misleading if the sample statistics consistently differs in excess of calculated sampling errors from the actual vote.
You are correct in that all exit poll methods and data are not equivalent.
Pre-election polls in the USA seem to not be that predictive too often as well.
spin
(17,493 posts)Obviously I haven't been faced with an exit poll. However if someone were to ask me who or what I voted for I would just tell them that it was none of their damn business.
I wonder how many people feel the same way I do.
rickford66
(5,523 posts)Young lady from NY Times with clipboard and asked me questions.
stevebreeze
(1,877 posts)than being predictive of results.
CK_John
(10,005 posts)ucrdem
(15,512 posts)It's not a traditional exit poll because early voting takes place over several weeks, but it's a carefully designed poll conducted by professionals and scholars:
The TargetSmart/William & Mary Poll was conducted using a blended web/phone data collection approach. As weighted, 530 online interviews were conducted from October 25-28, 2016 among panelists from six leading online opt-in panel providers who were matched to the TargetSmart voter file. As weighted, 188 telephone interviews were conducted from October 27-30, 2016 to supplement the online sample, with telephone numbers for targeted populations being selected at random from the TargetSmart voter file.
The survey reached 311 respondents who voted early in the 2016 general election, either absentee, by mail, or in-person, according to data from the Florida Secretary of State that was matched back to the TargetSmart voter file. These early vote data were weighted to reflect population parameters on TargetSmart Partisanship Score, TargetSmart 2016 General Election Turnout Score, age, gender, party registration and media market of the overall population of early voters in Florida who had participated as of October 31.
The survey reached 407 respondents who had NOT voted early in the 2016 general election, either absentee, by mail, or in-person, according to data from the Florida Secretary of State that was matched back to the TargetSmart voter file. These non-early vote data were weighted using a two-step process to yield a likely voter model. First, the non-early vote sample was weighted to reflect population parameters on TargetSmart Partisanship Score, TargetSmart 2016 General Election Turnout Score, age, gender, party registration and media market of the overall population of all registered voters in Florida who had NOT participated as of October 31. Then, to derive a likely voter model, a second round of weighting was employed on the non-early vote sample only, in which weights were assigned to each respondent in proportion to their TargetSmart 2016 General Election Turnout score, with higher weights being assigned to respondents with a higher likelihood to vote, and lower weights being assigned to respondents with a lower likelihood to vote.
The TargetSmart 2016 General Election Turnout model is an ensemble method classifier model that was created to predict the likelihood that an individual will vote in the 2016 general election. In this second round of weighting all early voter respondents were assigned a weight of 1, to reflect the fact that they have already voted. No margins of sampling error are calculated for this survey, as this statistic is only applicable to fully randomly sampled surveys, which this survey is not due to its partial reliance on online interviews.
https://www.scribd.com/document/329698329/TargetSmart-William-Mary-Florida-Poll-of-Early-and-Likely-Voters
This might sound like ironic advice but ... don't believe everything you read on the internet.