Monday, September 24, 2012

Early Voting and Polls/Predictors

I put this up on twitter, but I might as well write a post about it: I feel that there's pretty good information out there about the rules and extent of early voting (see Michael McDonald's excellent work tracking voting-so-far, plus follow links to his other work on this), but a lot less about how the various public opinion polls handle early voting and how the various predictor models out there that are primarily based on polls handle it.

Or maybe it's out there and I just haven't seen it. But for example Nate Silver today, in a very nice state-of-the-race post, noted in a short paragraph that there are lots of people voting over the next few weeks...but as far as I know, he hasn't detailed how his model handles it. Nate Cohn's similar post didn't mention it at all.

I vaguely remember that pollsters deal with it, but I don't really remember what I'm supposed to know about it. Hasn't mattered yet -- there are just a handful of people who have voted so far -- but in two or three weeks, it's going to start becoming a huge deal. When I brought this up on twitter, Josh Putnam pointed out that part of the problem is that there's limited reporting on all of this by some of the states. So there's that, too.

I don't know; I guess if there is something out there about it, I'd appreciate a pointer from someone; if not, I hope that one of the polling aggregator/polling-based prediction folks can give us a thorough discussion of the issue. Presumably, early voting makes the election a whole lot easier to "predict" from this point on than if there was no early voting, right? But I don't have any idea how much easier, or what kinds of error we should be looking out for.


  1. On early voting:

    When my state held the last election for governor, a lot of Democratic voters voted early. As polls began showing a shift away from the Democratic candidate, people on election voted for the independent candidate. There was a lot of complaining that early voting gave the election to the Republican candidate.

    I don't know if this has played out in other states, and it will be interesting to see how it plays out in this Senate election for Olympia Snowe's seat in the Senate.

    So I do think better analysis and understanding of early voting and polling it are crucial.

    1. zic, I'm not sure I follow the argument on the governor's race. If the early-voting Democrats were solid Democratic votes (as Matt suggests below), then their outcome would be the same if they early or if they voted on the day. If they were people inclined to vote depending on which way the wind was blowing, then the Democratic candidate should have benefited from the fact that they voted before the polls began to shift. Do you mean that the Independent would have won if they had waited?

    2. Exactly. Nearly everyone I spoke with who voted early and voted for the Democratic candidate wished, in retrospect, that they'd waited; they would have voted for the independent candidate.

  2. My GUT feeling, backed up by nothing but guesswork:

    As the non-election-day share of the electorate has gone up, this has a few implications.
    1) as it goes up, it is likely to look more and more like the overall population. When a group is 10% of the whole, it could be substantially different. When it's 25%, that would be tougher, as whatever process leads to one being part of the 25% is very likely to be having a more general impact than before.
    2) I would think voting early would actually require more political information/connectedness. You have to look into voting early, and there isn't a whole media/social world reminding you that "Tuesday is the day." Early voting also often takes place in places you're going to have to find, as opposed to the elementary school or two blocks over from your house. Plus, its very, very skippable; if you don't early-vote today, you can early-vote tomorrow. Thus, I would expect that early voters would be more likely to be partisans (and thus, that their votes were already spoken for). If this is right, they only matter if they're skewed, but see #1. Also, they were more likely to vote without these procedures anyway, so I don't think it's changing turnout much.
    3) If the concern is over late-breakers, see #1 & #2. I would doubt that early voters are undecideds, so whether they vote today or in 6 weeks isn't going to change their minds. In fact, I would expect that folks who are persuadable are going to select away from early voting, because, while liking Obama today, they're still open to voting Romney (or vice versa), and they would want to wait until they have more info. People who are undecided don't buy a car on the spot. (granted, for many people, this decision won't impact their lives as much, but the principle is generalizable, I think)

    So, I'm not sure that it matters. However, this is ALL just educated guessing. I'd be open to persuasion, so I'm not going to cast my vote on early voting effects yet!

    1. It is my experience that early voting in CA makes voting easier and increases turnout.

      In CA when you register to vote you can choose to be a vote by mail voter, and what this means is that they mail you the ballot some time before the election (maybe a month?). As long as you submit it postmarked the day of the election it is supposed to count. So you have far more days to remember to vote and can vote at a time that is convenient.

      For general elections this might not matter much, but it makes it a lot harder to forget to vote during primaries / midterms. That is, among people I know, no one who is vote by mail has missed an election, but some high information voters have missed primaries / special elections because of forgetting to vote.

    2. I suppose I should have distinguished absentee from in-person early voting. Anon's points are well-taken, and I know there is research on the effects of absentee voting on turnout (but I also know that I haven't kept up with that literature to know the answer).

  3. It should make the "likely voter" screens a bit more reliable. One of the questions is "have you already voted?" and they are, of course, the most likely of voters.

  4. Early voting is a really interesting and apparently under-researched. I believe there was some data that it worked well for Obama in 2008? I wonder if the logical thing--that since Obama is ahead in the swing state polls, this will be reflected in early voting--has any merit?

    By the way, this is one of the reasons that the voter ID law in Pennsylvania is so important--the Commonwealth has no early voting, same day registration etc.


Note: Only a member of this blog may post a comment.

Who links to my website?