This is the website of Abulsme Noibatno Itramne (also known as Sam Minter). Posts here are rare these days. For current stuff, follow me on Mastodon



June 2008

Curmudgeonly Note #2

We talked a bit on the podcast about how to interpret general election polls. With Ivan saying they were meaningless at this stage because they weren’t good predictors of the election results, and me saying they did matter because they reflect the state of the race NOW.

The analogy I should have used was that of a sports event. The score at any given time in the game does NOT tell you with any certainty who will win. It is an indicator. If one team is way ahead, then the other team would have to do something really extraordinary to win… and they probably won’t, but they might… but fundamentally the score in the middle of the game does not tell you who will win, it tells you the state of the game NOW. And that is not irrelevant or meaningless… it just tells you something different.

This is how general election polls should be interpreted. Not as predictors of the winner at the end of the game (the election) but rather as a snapshot of where things are RIGHT NOW. Which is different.

And of course, looking at national popular vote polls *is* irrelevant. That is not the game that is being played. The game being played is the electoral college. You need to look at state by state polls only. The national popular vote DOES NOT MATTER.

6 comments to Curmudgeonly Note #2

  • gregh

    Neither of you are right.

    The sports analogy fails, because polls aren’t scores. Ascending sports scores (where higher scores win) represent affirmative movement toward the goal: to win the game. Polls, on the other hand, have nothing to do with the game. They’re scouting reports, at best. Scores tell you who would win the game now, if it ended now. Polls tell you… nothing. The only number that matters in the election is the vote count, and you don’t have a vote count from a poll.

    Polls are useful for the teams, however. A football team needs to know how its corner matches the opponents receiver, and by knowing that, it can make appropriate adjustments. To the observer, however, it’s really more an indicator of how the eventual gameplay might proceed, rather than a measure of how the game itself might turn out.

  • ivanbou

    Let me clarify that the reasons I thought the polls were BS were the following: 1. Without the candidates settled the polls had way too much noise as to see how the candidates really were in a head to head. 2. Obama had a bump from being selected the nominee that also affected results. In a few weeks when those two effects have waned I think a poll will give a clearer picture of what the real situation would be at the moment.

  • Abulsme

    OK, it is a loose analogy, but I still think it is a fine one. If you must, it is analogous to a sport in which you can lose points as well as gain them, in a close to zero sum fashion, and you can’t directly see the score, you can only get a rough sense of the score within a significant margin of error.


    The main point I was making about the score, and the relevant part of the analogy, is that polls at this point in the game are no more relevant as predictors of the final outcome of the game than the score of a football game is at halftime. If it is extremely lopsided, it can be a good indicator of the final result. But otherwise, it is just an indicator of the state of the game at that time. Which is still an interesting thing to look at… especially for those planning the strategy for the rest of the game.

  • Abulsme

    I also reject Ivan’s argument that the polls before the nomination was settled, or during the “bounce” are BS. Again, if you are looking at them as predictors of the final result, sure, of course they are BS. All polls before the end of October will probably fit in that category. But if you look at them as getting a snapshot of the current state of the race, then they do mean something even with Ivan’s two points, because those two points reflect part of what the reality has been these last few months.

    That of course does NOT mean that the situation in July will even slightly resemble the situation in June, which does not resemble the situation as it was in May. This does not mean that looking at where things were in these months is bogus. You just have to take it for what it is… the current state of things… rather than a prediction of the final results.

    It is also useful to look at the trends over time. Don’t just look at where things were in May or June or what they will be in July, but look for the direction things are moving.

  • gregh

    That of course does NOT mean that the situation in July will even slightly resemble the situation in June, which does not resemble the situation as it was in May. This does not mean that looking at where things were in these months is bogus. You just have to take it for what it is… the current state of things… rather than a prediction of the final results.

    Polls are not votes.

    They don’t represent the state of much of anything other than current opinion. So, they provide a decent indicator of the current state of electoral opinion. Polls alone, as we’ve seen in the primaries, do little to tell us about the state of the vote.

  • Abulsme

    I’m not sure why you’re arguing a semantic issue that is completely irrelevant to the main point. Too much studying for the Bar perhaps?

    Of course most polls are not votes. But votes are polls. And yes, polls represent the state of current opinion… where things are now… not the result of the vote at the end. Which is exactly what I’ve been saying all along. And isn’t “a decent indicator of the current state of electoral opinion” exactly what you want if you are trying to gauge the state of a race?

    Polls are valuable as they are the best currently available indicator of the current state of the race… a race that is dynamic and changes week by week… just as the state of a sporting event changes minute by minute… as we get close to the end of “the game” these results DO tend to converge to the final results (albeit still with a significant margin of error)… but that convergence doesn’t really tend to happen until just a few weeks out from the end. Because, well, things change week by week and while there is a core of people that make up their minds early and stick to it, there is a decent percentage who week by week are unsure and are influenced by what has happened recently. And, of course, there is a very significant chunk that won’t really pay attention at all until right before the election.

    Oh, and the “as we’ve seen in the primaries” I think implies something that is not true. For the most part when you look at the LAST POLLS before primaries, they have gotten pretty close to the actual results. There are exceptions of course, some of them very visible (New Hampshire for instance) but you expect that with any method that is a random sample with a fairly big margin of error. Not to mention that when things are very dynamic (as they have been in the primary campaign) things have often changed on a dime in a matter of days, and the polls take awhile to catch up with that… usually at least a week… When you are talking about events which change the dynamic happening within a day or two of the election (as with New Hampshire) the polls just don’t have time to catch it.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.