Civitas Review

Polling Early Voting

By | Posted in Polling |
none
Oct
15

I've started seeing a couple of post on some polling websites (here and in the comments here) attempt to infer election trends from Survey USA's poll of NC that was released last week.

SUSA asked a question of whether the respondents had voted already or not.  A couple of problems — their poll was conducted 10/5-10/6.  Early voting doesn't begin until tomorrow 10/16.

Even so, 5% of respondents said they voted early.  And Obama was winning those voters by a fairly large margin.

Here are the problems with drawing any trends or conclusions from that:

1. Early voting does not begin in NC until 10/16 (tomorrow), so the only people who could have “voted early” are people who had requested absentee ballots.

2. When SUSA released that poll, according to NC State Board of Elections less than 3% of registered voters had requested an absentee ballot, so we have some margin of error in the actual numbers (<3%) vs. SUSA’s 5%

3. If you look further down ballot at the poll results for the Governor’s race, 17% of those who claimed to have voted early were “Undecided” in the Governor’s race.  Ok, so you’ve voted already, but you’re undecided on who you are voting for…

4. SUSA’s sample size was 617 respondents, of that, 5% said they voted early.  So the number of respondents who claim to have voted early is roughly 30 people.  The margin of error for a sample size of 30 among a population of 6 million is +/- 18%.  Thus the results saying Obama was winning early voters 65%-31% could be anywhere in the range of +/- 18% from either of those values.  So Obama could be getting anywhere between 83 and 47% and McCain could be getting anywhere between 49 and 13%.  So there is an equal confidence that McCain is actually winning people who claim to be early voters 49-47.  (There is also and equal confidence that Obama is winning 83-13 — but does anyone really believe that?)

5. Thus,  we are to conclude from a random sample of 30 people with a margin of error of +/- 18%, who claim to have voted early (but don’t know who they were voting for), that Obama is winning so-called early voters by a large margin?  Yeah, I don’t think that passes the sniff test, and I would think most would agree.

In the next couple of weeks you're going to see many claims made by both sides about how the election is going to turn out based on how early voters are voting, I may even blog about it here, but just remember with all polls there is a margin of error and other factors that we just can't account for or control.  Polling is really more art than science.