Tuesday, January 3, 2012

Guessing Time

Who's going to win tonight in Iowa? No one knows. Certainly including me. But I do have a bit of value added here, although only after a couple of paragraphs of digressions & links.

Allow me first a bit of self-indulgence (or skip down to the next paragraph), but, well, here's what happened. I was included (thanks to my posting over at Greg Sargent's Plum Line) in a WaPo list of pundit predictions about the Iowa results. And while I try hard not to complain about editing, I'm slightly miffed about this one. They took out the first words of my explanation for my selection of Rick Santorum, which were "I really have no idea who is going to win." Ah well; next time, I'll emulate Dana Milbank, who wisely and safely predicted a Gary Johnson victory -- after a late-afternoon Huntsman surge fizzles out.

Anyway, for analysis of the polls, I'll direct you to Mark Blumenthal, who thinks that Romney will probably win, and Nate Silver, who also has Romney as the most likely winner but puts more stress on uncertainty. Those are both "polls of polls" types of analysis. Want to go to the next level? Harry Enten takes the next step, and does a poll of "polls of polls"; that is, he looks at all the models or averaging methods out there. His numbers come up Romney, too (yes, yes: by 2020 there will probably be a dozen or more Nate Silvers, and a handful of Entens, and therefore there will be a market opening some current sixth-grader will fill for someone to average the polls of polls of polls. That's the kind of US ingenuity that won the Cold War, folks).

I think it's time for that value added I promised up top. Here's the thing: all of the quantitative stuff found in the links in the previous paragraph are averages or models based only on what's in the top-line numbers available in the various polls out there. But we know of other factors that may cause candidates to over- or under-perform their poll numbers. So I figured it's worth it to go through all of them I can think of and note how they (may) affect each candidate. Note that if we actually knew how this stuff would work, we could incorporate it into the polling; the problem whether any of these effects will matter at all is unknown. Ready?

Organization: campaigns identify supporters and do whatever they can to get them to their caucuses. That's normal; what's different in caucuses is that each GOP meeting will feature speakers on behalf of each candidate -- if, that is, the campaigns have found someone to do it in that precinct. Helps: Romney, Paul, and Perry are said to have the most precincts organized. Hurts: Gingrich is almost certainly the least organized of the six

Strategic Voting: If you're a social conservative and don't trust Mitt Romney on the issues you care about, you may be relatively indifferent about several possible anti-Romneys, and shift your vote to whoever is doing best. Other shifts for other groups are possible, too. Helps: Santorum, perhaps Romney. Hurts: Gingrich, Bachmann, possibly Perry.

Iowa Conservative Turnout Bias: Nate Silver talked about this one. Candidates perceived as conservative tend to overachieve compared to polling in Iowa; moderates tend to underachieve. Helps: Santorum, Perry, Bachmann, and perhaps Gingrich. Hurts: Romney.

Enthusiasm: An obvious one; candidates with enthusiastic supporters should tend to overperform. Helps: Paul, of course. Hurts: harder to tell, but probably everyone else.

Momentum: Let's define this (as Silver's model does), for the purpose of this exercise, as simply extending recent trends to account for the time between the last polls and actual voting. In other words, if people were moving towards a candidate over the weekend, it makes sense that whatever was causing that would probably still be operating yesterday and today. Note that this one overlaps with some of the others, especially strategic voting, but it isn't quite the same thing. Helps: Santorum. Hurts: Paul, Bachmann. Gingrich was losing supporters until recently, but he's been pretty flat over the last week.

Late Campaigning: This one should almost completely overlap the previous one, but it's worth noting that some candidates were probably throwing a lot more money at TV screens across Iowa over the last few days than others. Helps: Perry, Paul, Romney. Hurts: Gingrich, Santorum, Bachmann.

Regional Weather Effects: Caucus attendance is presumably more sensitive to bad weather than primary voting would be, and to the extent that neither weather nor candidate support is uniform across the state, it's easy to imagine fairly significant effects here. Helps and hurts? No idea. Ask reporters for this one.

Am I missing anything? Important caveats: in addition to what I said above about not knowing the magnitudes of any of these, or even whether they exist, my "helps" and "hurts" categories are just my speculative impressions of what's going on there. It's certainly possible, and perhaps likely, that anyone who has dug more deeply into the polls and the detailed campaign reporting might have a convincing argument for why I got one of these wrong.

My general sense of this -- and again, it's just guesswork at this point -- is that Santorum was on the good side of a lot of these effects, and Gingrich was on the bad side. But what that means for tonight...we're just guessing.


  1. "Caucus attendance is presumably more sensitive to bad weather than primary voting would be." This one kind of surprises me. Intuitively, I would think that caucus-goers are more dedicated to the task than primary voters, and that they see their presence -- which includes not only their own votes but their possible influence on others' -- as relatively more important than a primary vote. Hence they would be less likely to "bag it" if they looked out the window and saw a little snow. (I've often skipped primary votes myself, whereas if I were in a jurisdiction with caucuses I would feel obligated to trudge there regardless.)

  2. You missed the biggest one: targeting.

    You've got to model how many people are going to show up -- without that you can't figure out if you're hitting targets.

    I suspect Paul's people are doing the same thing as the dean people did in 04. They were working off bad targeting numbers.

  3. The last PPP poll has Paul gaining steam on Sunday (the second day of a two-day poll, with a very large sample size): http://twitter.com/ppppolls

    In 2008, Paul performed 25% better than the RCP average on election night (a better bump than anyone else):


  4. Jeff -- Unless the weather is so bad that everything closes down, people will vote in a primary on their way to work, or wherever they're going. The caucus generally requires a second trip and nighttime driving.

  5. Jeff & Couves: both good points.

  6. A quick check on my i-phone shows that the weather in various Iowa cities is pretty moderate today--suggesting that turnout won't be depressed. I suppose that slightly decreases the chances of Paul stealing the night, since his supporters are probably the least likely to be stymied by bad weather.

  7. Kylopod - Polls show that Paul's supporters are the most loyal, but also the least likely to actually vote. Yes, his core supporters will show up no matter what, but he also depends on support from people who normally would never attend a caucus. Weather is mostly important to the post-caucus spin -- no one will dismiss a strong Paul showing based on bad weather.

  8. @Couves, Paul supporters have a history of showing up--at CPAC, at the Aimes straw poll. I think they'll show up tonight.

    But are you suggesting that the Paul camp could spin lower numbers, if that's happens, on non-turn-out by some milder supporters? That would be a delicious irony. (I doubt it will happen, so I'll have to get my chuckles some other way.)


Who links to my website?