Dave M asks:
You say that they poll randomly through the day and when the final precinct vote (not survey) counts are available, they get incorporated. If exit poll interpretation isn’t done until the voting is done, then why use the survey (a limited sample with the inherent sampling errors) at all? Why not just use the official vote tallies?
Good question. The exit poll incorporates hard data on turnout first then data on actual voting as it becomes available precinct by precinct. My knowledge on the specific procedures they use is weak, but the interviewer is responsible for getting hard data on turnout, especially at the end of the day (either from their own tally, from the precinct officials or both). This turnout data is critical to weighting the end-of-day poll. Then as the precinct vote becomes available, they integrate the two to make the poll more and more accurate as the night wears on. Once the complete results are known, the exit polls are weighted to match the vote.
Mean Dean asks:
Do incidence rates play any significant part in the quality of data? If so, which way? I’d suspect if I was on my way (back) to work, or rushing home for dinner afterwards, or toting copious kiddage I’d find myself declining an exit poll as well.
The one thing exits polls can do is track the gender, race and (approximate) age of those who refuse to participate and adjust the resulting sample to correct for any flaw. Because the exit poll result can be compared to the actual vote in each of hundreds of individual precincts, methodologists can check whether the error rate increases in precincts with lower rates of response. This short answer is higher response rates do not lead to higher error in exit polls.
One last link recommendation: the folks who conduct the exit polls have a great guide on things to look for tonight (hat tip: Andrew Sullivan)
Now, time to watch this thing…I’ll post more later
Thanks for answering my question regarding “incident rates.” One more if you don’t mind. What is the difference between “exit polls” and “early results.”
Meaning, can I project accurate results with a small percentage from either?
How is it that news organizations can routinely declare a winner in a race that is merely (let’s say) five points apart, with only (let’s say) 25% of districts reporting?
This info on exit polling is really helpful in the midst of these nervous moments. It’s almost a syllogism:
– Exit polls are weighted to include early voters
– Early voters skewed to the Dems in the battlegrounds
– Early results today don’t include early votes and absentees (which will be counted last)
These facts explain much of the discrepancy between the exit polls and the early results. The pundits are being driven by the 5-min news agenda.
Mystery Pollster, tell me please why the numbers in Ohio and Florida seem so, to me, oddly skewed relative to what was expected (according to tracking polls, exit polls). Why does Bush have such a healthy lead?
Please set my mind to rest, I can’t get it out of my head the whisperings about ES&S and Diebold voting machines…
I want to believe that it isn’t possible. Could they?.. Would they?..
I believe I’ve put my finger on why the exit polls suggested a Kerry win. Bush voters voted, but were intimidated. They are afraid of the angry left, including the chance that their friends or co-workers will shun them if word gets out they voted for Bush. They may even be afraid to tell pollsters the truth.
What to do? Lie to the exit pollster.
There is a history of lying to political pollsters. There is strong evidence that in that in 1974 polls sandbagged British Prime Minister Edward Heath into calling an early election that his party subsequently lost.
How well are the poll takers trained? How experienced are they?
IOW, Even if you tell the poll taker to talk to every 100th person, has anyone checked on how well they actually do that?
IOOW, you get a report on what people refused to respond. You don’t get a report on when the poll taker took the 101st person (a cheerful, slow walking woman) rather than the 100th (a cranky, quick walking guy).
Since the later exit polls incorporate official polling data, if the voting machines were programmed to randomly change 5% of Kerry votes to Bush votes (like Bill Moyers reported had happened to a friend of his in Texas), wouldn’t that explain how the exit polling (adjusted by official votes) would converge with official voting reports as the official numbers came in? And one blogger noted that the exit polling numbers were matching final official returns everywhere but in Ohio and Florida. Shouldn’t that suggest impounding and examining the machines in those two states?