« Exit Polls: What You Should Know | Main | Exit Polls: More Questions »

November 02, 2004

Exit Polls: Answering Your Questions

Let me answer some questions that came up in the comments:

Frankly0 asks:

I don't get the notion that the exit polls have a 3% MOE nationally. I can see 4% MOE for a given state, but the sample set is so large nationally (150,000, given the stipulated numbers) that it's hard to believe that the MOE only goes down to 3%.

I don't know the specific numbers involved, but the reason is essentially this: The National Exit Poll does one sampling of precincts nationally (to look at the popular vote and the national electorate) and separate samplings in the various battleground states. I don't know exactly where they are polling, but it is safe to assume the networks did not see any reason to do exit polls in places like Utah or the District of Columbia.

Yes, they could roll everything together and weight down the value of the battleground states, but in terms of sampling error, the national sample is still not stronger than it's weakest link, the smaller sample in the non-battleground states.

Cableguy asked the following (Ted B had a similar question):

How do you deal with people who voted before today, whether in person or w/ absentee votes. I assume you do polls to supplement exit voting? Do people in media/campaigns already have this data?

I thought I had answered that one already, but perhaps I was not clear: To handle early and absentee voting, the networks have completed telephone surveys in 13 states and screened for voters who said they had already voted. Their final exit poll will combine results from phone surveys of early voters and in-person interviews with in-person voters.

The networks should have this data now, though whether the leaked exit polls include them is anyone's guess. Network coverage tonight will discuss this, I am sure.

P.S. Please fire away with questions in the comments section, but again, please refrain from posting leaked exit polls questions about numbers you may have seen. Thank you!

Related Entries - Exit Polls

Posted by Mark Blumenthal on November 2, 2004 at 04:35 PM in Exit Polls | Permalink


What about 37 states that they didn't poll? Do they just assume early voting = today's voters? Do the networks do their own polling for early voters, or share them like exit polls? What about military/overseas voters? I don't have too much confidence in the current methodology after both 2000 and 2002 debacles.

Posted by: Cableguy | Nov 2, 2004 4:52:45 PM

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.

Is this measured, and if so, is it equated into the 'scrubbed' data?

Posted by: Mean Dean | Nov 2, 2004 5:02:22 PM

To make sure I understand the methodology correctly, this year's exit polling will depend on the precinct turnout in the past: if precinct A and precinct B both had 1000 voters last time, then they have an equal probability of being chosen this time? However, if precinct A comes in with 1200 voters to 1000 for B, then the results for A are weighted more heavily?

Posted by: SC | Nov 2, 2004 5:20:15 PM

How does the heavy voter turnout affect the reliability of the exit polls?

Posted by: Pat | Nov 2, 2004 5:26:48 PM

Do exit polls overpoll urban voters over rural voters?

Posted by: Jeff R. | Nov 2, 2004 5:57:44 PM

Only 30 states allow some form of early voting. It may be that the networks only polled in states that both allowed early voting and seemed to be w/in the statistical margin of error (for instance, no reason to poll early voters in Texas, where Bush should win easily).

No reason for exit polls to "overpoll" urban voters (if by that you mean sample at higher than their proportion to the population). If you mean are more urban voters sampled, yes, because more people live in urban areas.

I'd think the higher voter turnout does not effect reliability one way or another. I suppose if there are more voters, than an estimate based on a sample of N over a population of P will have a larger standard error, as P increases.

But exit polls are so large that the impact of increasing P even 10% (no one is suggesting national turnout anywhere nearly 10% higher) will have a small impact on the standard error.

(Hey Mystery, I was a grad student at Michigan,teaching Poli Sci and working for NES, when you must have been there. Drop me a line.)

Posted by: The Prof | Nov 2, 2004 6:52:18 PM

After I voted today I was asked, for the first time, to participate in an exit poll. Out of habit with phone pollers, I just waved him on. Afterwards I wondered if maybe I had missed an opportunity.

What are the pros and cons of participating in an exit poll?

Posted by: Jack | Nov 2, 2004 7:22:34 PM

Did anyone else notice that CNN changed their Ohio exit poll number? I was pretty certain that they had Kerry taking ~52% (maybe 51%) of the vote in Ohio shortly after 8 PM EST. When I checked again around 4 AM EST, the numbers seemingly changed to Bush taking ~51%.

I don't think it would take very long (no more than few minutes) to adjust raw exit poll number to account for actual votes cast in different precincts, etc. Did CNN make a blunder and release some unprocessed number? (I'd like to think that I didn't simply willed myself to misread their graph although it's not impossible)

[note. A partial explanation of the times when I found those numbers is that I am in England.]

Posted by: Ho-Yon | Nov 3, 2004 11:42:06 AM

The comments to this entry are closed.