Scatterplot dump: Colorado caucus edition

Just to follow up on my previous post about Tuesday's Colorado Republican caucus, I've played with a bit more county-level data and found a few interesting correlations. None of these are perfect, and a lot of things are moving simultaneously, so to find the one thing that caused Santorum's upset is a fool's errand. Nonetheless, these are suggestive.

I also encourage you to read Sean Trende's county-level analysis of New Hampshire's and Florida's primaries (he finds some similar things to what I found about Evangelical voters) and Nate Silver's interesting essay today.

Anyway, some graphs after the jump. All bivariate relationships are statistically significant at the p ≤ .05 level.

White Men Can Count

First of all, I'd like to commend C-SPAN for its coverage of Iowa Caucus events. If you watched last night, you saw regular Iowa citizens getting together to try to pick a president. They gave unpolished speeches, they voted, they counted the votes, they called in the results, then a few stuck around to talk about platform positions and to decide who would represent the candidates at the county conventions. People often compare caucusing with primary voting, describing the former as involving more conversation and taking more time. But it some ways it's closer to jury duty, only a lot less depressing. It's regular people coming together to perform an important civic duty. It was great to watch. Okay, maybe watching Iowans count green sheets of paper isn't your idea of a good time on a Tuesday night, but if so, you're probably not reading this blog.

Second, I want to address this article about Ron Paul's post-caucus strategy. The gist is that the Paul people are very organized and made sure that their supporters stuck around after the initial counts to run for delegates to the county conventions. This is how you ultimately end up with a greater share of national convention delegates than your caucus-night showing would predict. The naive campaign treats a caucus like a primary and leaves as soon as the voting is done. The smart campaign realizes that the caucus is just the first step in the selection of delegates and sticks around to try to control the post-caucus selections. (I wrote about this with regards to Obama and Clinton back in 2008). Anyway, I think Josh Putnam is right that Paul could end up with significantly more delegates than expected thanks to this level of organization, although I agree with Jon Bernstein that this won't make a difference for the nomination.

One point I'd like to add: In the Business Insider article linked to above, the author writes:

Iowa's Republican caucuses are non-binding — they are technically just a straw poll, so once selected, delegates are free to vote for whichever presidential candidate they choose.

I think the whole binding/non-binding thing is a bit of a red herring with regards to caucuses. I hope Josh or Jon will correct me here if I'm wrong, but the way that caucuses get to choose their delegates to the next levels (county, district, state, national) all but ensures that those delegates will be extremely loyal to their preferred candidate. After the Iowa caucuses last night, Romney supporters in each precinct gathered together to pick the people among them who would best support Romney at the county conventions. Now, there'd be no official sanction if one of them defected to Paul or Santorum, but that person would be a fool to do it. She'd be immediately distrusted and despised among local Iowa Republicans, and if she cared enough about her reputation in local politics to get involved in the county convention in the first place, that's an outcome she'd like to avoid. Now, if Romney somehow dropped out of the race and encouraged his supporters to back Santorum (this won't happen, but stick with me), that delegate would be happy to follow his request. But short of that, he can expect a great deal of loyalty from his "non-binding" delegates.

What an activist wants, what a party needs...

I was up in Broomfield for the Colorado state Democratic Assembly yesterday.  The Republicans held their event simultaneously.  In case you missed the big news, of the four major statewide races (governor and U.S. Senate in both parties), three of them were won yesterday by people who were not their party's presumed frontrunners.  On the Democratic side, Andrew Romanoff beat out Michael Bennet for the Senate nod by about 20 points.  On the Republican side, Ken Buck won the Senate contest (presumed nominee Jane Norton decided not to compete) and Dan Maes narrowly beat out presumed gubernatorial nominee Scott McInnis.

What does this mean?  In practical terms, not all that much.  Winning at the state assembly means you get top line on the primary ballot in August.  The research on the effect of ballot order is mixed, but generally leans toward the conclusion that the vote impact of being first on the ballot as opposed to second is either small or nonexistent.  (This paper from Michael Alvarez, Betsy Sinclair, and Richard Hasen does a nice job reviewing the legal and political science research on the topic.)

Now, if a candidate doesn't get at least 30% of the assembly vote, she does not get to appear on the primary ballot at all, although she can later petition onto it.  Only if a candidate fails to achieve 10% of the vote is she actually prevented from continuing as a candidate.

What does winning at the assembly mean politically?  Well, it provides a short-term media story -- party activists reject presumed frontrunner -- but that quickly fades.  And really, at this point, it's odd that it even generates that much media interest.  A thorough scholarly review of the Colorado assembly/convention system has yet to be done, but the available evidence suggests that it often ends like it did yesterday, with the party activists rejecting the presumed frontrunner.

Why is that?  Well, the party insiders who anoint a frontrunner don't wait until May of an election year to go about their business.  Those who are really influential at this level -- and here I'm talking about the businesspeople who picked McInnis and helped pressure Josh Penry out of the race, and of course President Obama and his involvement on Bennet's behalf -- did their work back in 2009.  They made judgments based on what they knew about the candidates and the state and on what they expected the political environment to be like by the fall of 2010.  These sorts of insiders are ideological but have a pragmatic streak; they want to win.  So, to some extent, they err on the side of caution, looking for candidates who stand for something but are still moderate enough to win with some margin of safety.  So they pick a Scott McInnis, who has solid GOP credentials but whose commitments to the pro-life and Tea Party causes are less than certain.  They pick a Michael Bennet, who is well within the mainstream of his party in the Senate but is hardly a rabid left-winger.  And so on.

And then we get to the caucus-assembly process, a three-stage procedure involving ever-more-hardcore party activists with each successive stage.  And they're basically asked, Do you like the relatively centrist presumed nominee you've been handed?  Of course they're going to say no.  They're activists.  They're way toward the ideological extremes.  They believe a) that a more ideologically extreme candidate can still get elected if given half a chance; and, b) that losing an election once in a while isn't as bad as selling out everything you believe in.  So they stand up and say, We want a different nominee!  We should expect to see this by now.  Really, it would be remarkable to find the party activists embracing the presumed nominee.  That sort of happened among Democrats with Mark Udall in 2008, but even he had a challenger within the party.  It's rare.

So, in sum, the whole assembly/convention thing doesn't actually matter that much.  It rarely disqualifies a candidate with any sort of serious backing, and it provides the winner little more than bragging rights.  Which begs the question, why do we do it?  I mean, reasonable people may disagree over whether party activists should have the power to pick or reject nominees like they do in Utah -- but at least there it matters.  The Colorado system seems more like a chance for activists to express preferences but with little actual power to see them enacted.  At this point, based on everything I know about how primary elections actually work, I would say the odds still strongly favor Bennet, McInnis, and Norton winning the August primaries.  I'd love to be proven wrong, of course (I am a Romanoff delegate, after all), but given where the money and endorsements are, that's the likely outcome.

So why are we doing this?

Caucus analysis, with maps!

Well, the results of last night's caucuses haven't been completely tabulated, but that shouldn't stop me from making inferences, should it?  Don't answer that.

Okay, on the Democratic side, here's what the county map looks like.  I have calculated Romanoff's share of the Romanoff + Bennet vote, ignoring the uncommitted for now.  Darker blue counties are those in which Romanoff did better.  The data come from here.

Now, the completely white counties are ones in which we don't yet have numbers (including, notably, Pueblo), so ignore those for now.  I'll update later.  But what we do see is that Romanoff did well in the Denver metro area and even better in some of the central and southern counties.  Bennet's areas of strength appear to be the northern plains areas and the western slope counties near Grand Junction.  Regression analysis shows that Bennet did better in counties with a higher percentage of college-educated residents, while Romanoff did better in wealthier counties.

Over on the Republican side, I calculated Jane Norton's share of the Norton + Buck vote, ignoring the uncommitted and supporters of other candidates.  Darker red counties are those in which Norton did better.  The data come from here.

This map shows a strong regional pattern, with Norton doing much better than Buck in the western counties and Buck dominating the eastern plains.  This may partially represent the candidates' upbringings: Norton is from Grand Junction originally, and Buck is the Weld County D.A. and has a long family history in Greeley.  So she's the western slope candidate and he's the eastern plains one (whereas we see less of a geographic pattern on the Democratic side with both candidates coming from Denver).  Regression analysis shows that Buck did significantly better in counties with a higher percentage of Evangelical Christians, although no other major demographic distinction between the candidates' supporters turns up.

Update: Dem map updated with Pueblo County data.

Later update: Maps changed with newer results as of 3/19.  Color breaks now 0-40, 40-50, 50-60, and 60-100.

(Cross-posted at ColoradoPols and Huffington Post.)

The caucuses

I'm just back from my precinct caucus.  My initial impression is that the Bennet campaign seemed a lot like the Clinton campaign back during the 2008 county and state conventions -- much better organized and funded than their opponents, but still outnumbered.  The Bennet folks had t-shirts, water bottles, cookies, and volunteer greeters.  They seemed in every way a more professional organization than Team Romanoff.  Nonetheless, Romanoff just had far more people in the room.

Organization can help make sure that your supporters show up, which is obviously very important.  Organization can also make sure that your team wins close calls, like disputes in rules or providing alternates for missing delegates.  And organization can help win over undecided voters.  But this was the precinct caucus -- undecideds mostly don't come.  (Well, so far it looks like about 7% of the Democratic caucus goers are uncommited, compared to the 26% of Democratic primary voters who are undecided.)  So the end result looks unsurprising, with Romanoff on top but both candidates easily qualifying for the ballot.

Meanwhile, the GOP caucus is looking really interesting, and it's not clear yet whether Norton will prevail. So I'm guessing the headlines tomorrow will say something to the effect of "insider candidates take a beating" or something.

Just how representative are caucus attendees?

Via roguestaffer, here's a new PPP poll (PDF) of likely Democratic primary voters.  Among that group, 40% say they prefer Bennet for the Senate nomination, 34% prefer Romanoff, and 26% remain undecided.  We'll know more tomorrow, but my guess is that the outcome of tonight's precinct caucuses will look almost nothing like that poll.  For one thing, the number of undecided people who will actually participate in the caucuses could probably fit in a Honda Civic.  For another, I'm guessing Romanoff will at least tie.

Caucus primer

I was recently asked by a reporter to explain how important Colorado's caucuses are and why party "frontrunners" aren't necessarily favored to win them this year.  I'm not sure how much of my response will make it into print, so I figured I'd just jot it down here in case it's of some value to my obscenely large readership.

How important are the caucuses?  That's a tricky question.  There have been a number of high profile candidates in recent years who lost the caucuses and went on to win their party's nomination anyway (e.g.: Ken Salazar and Pete Coors in '04).  Winning the caucus is neither necessary nor sufficient for getting nominated.  However, it can be influential.  The candidate that wins the caucus will get top ballot position in the primary.  And if a candidate gets too small a caucus vote, he or she may be eliminated from the primary altogether.  Beyond that, the caucuses are mainly an opportunity for longstanding party activists to make their voices heard.

One of the real reasons that "frontrunners" aren't favored in the caucuses is because caucus participation is so limited.  Even in the 2008 presidential election, only about 6 percent of Colorado's eligible voters turned out to participate in the precinct caucuses -- and that was historically high.  Participation this week will be substantially below that.

The people who actually participate in caucuses tend to be the most active partisans.  They're highly informed and passionate on behalf of their party and particular candidates.  Unlike most other voters, they tend to follow politics all year round, rather than just prior to a major election.  As a result, their preferences tend to be somewhat different from those of the larger population.

This year in the U.S. Senate race, we have an interesting dynamic.  In both parties, party elites appear to have anointed a particular candidate, and party activists are rebelling against that choice somewhat.  On the Democratic side, Gov. Ritter appointed Michael Bennet to the U.S. Senate last year, a move that caught many Democratic activists by surprise, and Democratic leaders in DC, including President Obama, have rallied to support Bennet.  Many state Democratic activists were bothered by this decision being forced upon them, even if they have no particular problem with Bennet's behavior in the Senate.  These activists have largely rallied behind Andrew Romanoff, who has spent years working for Democratic causes in Denver and across the state.  Since these are the folks who will dominate the caucuses this week, Romanoff is likely to do very well in that contest, even though Bennet has raised far more money thus far.

On the Republican side, many party activists have chafed at the apparent anointment of Jane Norton.  Although they haven't quite settled on an alternative candidate for the nomination, those who are resisting her nomination will likely turn out in high numbers.