The Iowa Cascade

by publius

It’s true that a Hillary Clinton victory isn’t yet inevitable – but it’s sure getting close. She’s got strong, commanding leads in the polls. Lots of money and institutional support. Tons of experience, etc. In short, she’s got all the things that inevitable nominees tend to have. The catch, though, is Iowa.

Personally, I think Iowa is the only thing standing in her way. And so the million dollar question at this point is how the other states will react to Iowa. It’s impossible to predict, but I’m going to take a stab at it by using information cascades.

What makes Iowa such a wild card is that it can trigger instant information cascades – or “herd behavior.” The basic idea is that cascades happen when people rely on the visible actions of others rather than their own imperfect or limited information.

For instance, imagine you are at a fancy-pants dinner with lots of people and don’t know what fork to use. Let’s say you’re 55% sure that the small fork is for salad. But, you see two other people before you pick up the longer fork for the salad. You get nervous and discount your own private information, opting for the big fork instead. Imagine too that the guy next to you is even more confident that the small fork is for salad. But, with three people opting for the big fork, he does the same. And on it goes until everyone is using the wrong fork. Thus, the cascade. (For more formal statistical discussions, see here - pdf).

We see the same thing happen with bubbles, financial panics, etc. But we also see it following Iowa. And that’s the problem.

The idea of the Iowa Cascade is that voters in subsequent states ignore their own private information and follow the Iowa crowd (i.e., the first guy to pick up the wrong fork). In the article that inspired this post, Paul Waldman explains why the Iowa Cascade is bad -- namely, that Iowa voters are both few and uninformed. The question, then, is whether an anti-Clinton cascade will happen if she loses in Iowa. Frankly, I don’t think so.

Cascades depend on the strength of voters’ “private” information. If, for instance, no one knows much about any of the candidates, the first election arguably matters a lot more because it will have more informational benefits (cascades lower information “costs”). If, by contrast, the candidates are well-known, it’s less likely that people need to look to Iowa for guidance.

That’s the key difference between Howard Dean and Hillary Clinton. The 2004 slate of Dems simply wasn’t well-known by the casual, not-well-informed voter. Their “private signals” were probably fairly weak. Thus, when Kerry won, it was easier for everyone to abandon ship and join him.

Clinton, however, is arguably the best known non-incumbent candidate in recent history. She’s running a good campaign and all, but don’t underestimate the power of name recognition (especially when the name is Clinton). The point is that the people in New Hampshire and beyond probably have a firmer opinion of Clinton (for good or ill) than, say, Howard Dean. Their private signal is therefore less susceptible to the movements of the Iowa herd. That’s why Clinton’s commanding leads in New Hampshire (and nationally) should trouble Obama. An Iowa win will definitely bring him some points. But it simply can’t overcome a 20-point deficit against a candidate as well known as Clinton.

More broadly, Atrios is absolutely right that the press is largely to blame for what I call the Iowa Cascade. In a perfect world, the public would understand what “winning Iowa” really means -- a few thousand, very-white Iowa Democrats gave one candidate a plurality in a seriously-flawed caucus system. Instead, it becomes this mammoth media event that gives the impression that the “winner” has far more support than he or she does. Returning to the example above, the press is essentially adding thousands of imaginary people to the table all picking up the wrong salad fork before you do. In doing so, they undermine people’s private information.

That’s why it would be better to have “Super Tuesday” as the first major primary election. It would give far more people (more diverse people at that) the chance to make a decision without the corrupting influence of information cascades.

Comments

[rilkefan]

What a physicist probably thinks of in this context if not this.

[rilkefan]

I wonder if the great access to extensive national poll information will make the early primaries less important. Also the financial positions of at least Obama and Clinton.

[tsam]

One of the key foundations of Paul Waldman's article is that Iowa's participation is less than 10%, much less than typical primary elections. It discounts that the commitment of time and lack of privacy makes a precinct caucus more difficult to participate in than an election.

Also this article assumes that the degree of correlation between results in Iowa and New Hampshire is due to NH voters reacting to the Iowa results, not to both states' participants reacting to the same set of candidates.

The first assumption is questionable. Here in Washington this article suggests that 2% of our Democratic primary voters (20,000 vs. a million) participated in our caucus. That 5 times the fraction of Iowa's voters participate in their caucus — despite severe weather — is evidence that they are motivated by its unusual attention.

I do not find the process of a precinct caucus fundamentally flawed. The 15% threshold causes more voters to think deeply about which viable candidate they can support, effectively unifying splintered allegiances. In this, it is closer to instant runoff voting than your typical primary.

Aside from Waldman's bitter-toned and questionable article, the analogy of imitating formal dinner etiquette poorly fits the situation. You suggest that voters imitate the early states because they are told the voters there are "wiser". Their apparent imitation also can reflect a desire to vote for a viable candidate, to maximize their vote's influence.

I am more concerned that voters make decisions based on unfair negative TV ads than by looking closely and directly at candidates. I see this as a greater influence than pack-driven media coverage. While primaries have relatively less negative ads than general elections, most voters pay more attention to ads than they invest in watching debates or other direct appearances of candidates. I do not see a large primary as the first voter event as improving the real reasons that voters make uninformed choices. (Which they, alas, seem to do.)

[wrog]

The 15% threshold causes more voters to think deeply about which viable candidate they can support, effectively unifying splintered allegiances. In this, it is closer to instant runoff voting than your typical primary.

The first fallacy here is the idea that viability within a precinct has anything to do with viability statewide. Precincts are far too numerous and varied. Entire precincts can go for what turn out to be marginal candidates. Meanwhile other candiates can turn out to have significant, perhaps even majority support statewide, but if a given supporter doesn't happen to live in a precinct with sufficiently many likeminded folks then either his vote is lost or he is forced to lie about his preference in order gain entry to the delegate selection process.

The second fallacy is the idea that 15% is any kind of meaningful threshold. To be sure, if you're the only supporter of a given candidate in a precinct or if there simply aren't enough delegates to go around in that precinct for your candidate to be able to have one, then something has to give and you need to find people you can band together with and work out some kind of agreement (for Uncommitted or whatever). But you don't need a 15% rule for this to happen. It is enough that the precinct has a small number of delegates.

What the 15% rule actually does is set an extra arbitrary bar that is completely without purpose at the precinct level.
If 1/7 of the voters in a 7-delegate precinct want to vote for somebody -- and again, for all we know it could be a major candidate; we simply won't know who the major candidates are until the rest of the state's results come in -- there is absolutely no reason (in any moral/ethical sense) not to be giving them 1 of the 7 delegates. Any other allocation misrepresents their votes, but because 1/7=14.2%, they lose, and it gets worse for the higher-delegate-count precincts.

The 15% rule also creates discontinuities, situations where single votes can shift multiple delegates -- all you need for this is several candidates hovering near the threshold. This then causes intense pressures to be brought to bear on individual voters, because their one vote affects so much. Never mind that these situations are often resolved on the basis of "if you change your vote, we'll let you be the delegate" rather than anything having to do with the merits of the candidates.

Since you're evidently a Washington resident, you may be interested to know that the Washington state Democratic Party has eliminated the 15% rule for its precinct caucuses. For 2008, precinct delegates will be awarded strictly proportionately. The 15% rule will still be in effect at the higher caucus levels, in order to satisfy the DNC requirement that the state send no delegates to the national convention that don't have at least 15% statewide support.

In fact, I strongly suspect this last was the only real purpose of the 15% rule to begin with (i.e., to ensure that the national convention not be having to deal with marginal candidates), that the only reason it ever made its way down to the precinct level at all was because state party Rules Committees prefer having uniform rules -- that way they only have to write one set of rules to deal with all levels of caucuses at once. So when the 15% criterion was inserted at top level by the DNC, it was easier for them to just insert it everywhere. And then it became a (useless) tradition

[cleek]

in semi-related, but happy, news: the mullahs of the religious right refuse to back Rudy.

[Nathan]

Publius said:
"The point is that the people in New Hampshire and beyond probably have a firmer opinion of Clinton (for good or ill) than, say, Howard Dean. Their private signal is therefore less susceptible to the movements of the Iowa herd."

Ah, but your reasoning can work both ways. If the Iowa Cascade affects the views that New Hampshire voters have of less-well-known candidates, than an Obama win in Iowa would trigger exactly the kind of information cascade that you describe. Obama is still unknown to the average person, and a win in Iowa would cause other voters to re-evaluate him. Exactly the way that they re-evaluated Dean after he lost. But this time it would be in a positive direction.

I still believe that people aren't ecstatic about Hillary, and she continues to coast along because of name recognition and the coronation by the press. An Obama win in Iowa would pierce this illusion, and the Iowa Cascade would still apply. Your mistake is to apply it to Hillary and not to Obama.

[trilobite]

I still believe that people aren't ecstatic about Hillary, and she continues to coast along because of name recognition and the coronation by the press.

I'd like to believe this, but I don't. The press was pretty nice to Obama, too. Hillary is not coasting, she's winning what looked like an uphill fight at first, and she's winning because she has a good organization and good strategy, and because she manages to project a safe, "normal" feeling on TV that Obama does not. Obama is, I think, just a little too exotic for most voters -- not just his skin color, but his name, his ideas, his passion. Despite all the polls that say people want someone with ideals and passion, those candidates never seem to win the nomination.

Hillary DOES have incredible name recognition, but it seems to me that Edwards is fairly well known by likely voters, and he's nowhere.

[OCSteve]

cleek: in semi-related, but happy, news: the mullahs of the religious right refuse to back Rudy.

I keep saying you folks should be contributing to Rudy’s campaign. ;)

If Rudy Giuliani wins the Republican nomination and a third party campaign is backed by Christian conservative leaders, 27% of Republican voters say they’d vote for the third party option rather than Giuliani. A Rasmussen Reports national telephone survey found that a three-way race with Hillary Clinton would end up with the former First Lady getting 46% of the vote, Giuliani with 30% and the third-party option picking up 14%.

Man, I guessed 5-8%. 27%!

[Anarch]

Heh. I love the idea of a Republican splinter party. Their motto can be "When the fascist isn't conservative enough".

[liberal japonicus]

I keep saying you folks should be contributing to Rudy’s campaign. ;)

I started to reply to this a while back, but deleted it. I know that OCSteve isn't being entirely serious with that, but I'm thinking that one of the reasons public discourse is in the dumpster is this 'enemy of my enemy deserves support' approach. While I am sure it has been around forever, we see it cresting with Nixon working to get McGovern nominated, and, like a bad penny, it keeps coming up and we saw it in supporting the Nader on ballot initiatives. I'm trying to remember who support Perot, but I think it was unclear who Perot would harm the most, so this didn't occur.

It's probably unfair to be all serious with a tongue in cheek comment, but I worry that you'll get what Kurt Vonnegut said in Mother Night: "We are what we pretend to be, so we must be careful about what we pretend to be."

[KCinDC]

LJ, I think the most recent example is Republicans donating to the Green candidate for Senate in Pennsylvania in 2006, in the hope of siphoning off votes from Casey (who was viewed as too conservative by many Democrats) and thus saving Santorum. It came up this week when Darrell Issa was trying to show that Blackwater wasn't closely affiliated with the Republican Party because their owners and employees even donated to the Green Party.

[Katherine]

Obama is really impressing me lately--showing signs of running the campaign I hoped he would run all along. We'll see if it lasts, or if it's enough.

[tsam]

Responding to wrog...

The second fallacy is the idea that 15% is any kind of meaningful threshold.

I argued that the caucus process is representative enough to produce meaningful results. This general argument is not based on the precise parameters used.

The first fallacy here is the idea that viability within a precinct has anything to do with viability statewide.

Certainly stronger local support for a candidate is positively correlated with stronger overall support. There are exceptions, getting but more votes someplace never hurts a candidate.

Entire precincts can go for what turn out to be marginal candidates. Meanwhile other candidates can turn out to have significant, perhaps even majority support statewide, but if a given supporter doesn't happen to live in a precinct with sufficiently many likeminded folks then either his vote is lost or he is forced to lie about his preference in order gain entry to the delegate selection process.

The nonlinear nature of caucusing, with or without a threshold, adds measurement error, but strong overall support for a candidate can not be completely hidden.

Also, primary voters is a broad race typically can find more than one candidate who looks good. It's rarer than you suggest that a voter is forced to completely lie or be completely shut out.


By the way, caucusing is a hassle. It takes more time than mailing in a ballot, and it exposes participants to their neighbors' opinions. (One of the most amusing parts is the slanted questions prepared by the party committee. They make obvious how the event is structured to tell voters what they want, not to ask them.) But anything that gives the governed a share of real power contains some of the magic of democracy.

[dutchmarbel]

This is probabely a very dumb question - but why don't they have all the primaries on one day?*)

*) primaries, voting in primaries, having people vote for the preferred political candidate of their designated party - maybe the term is not correct but I assume you all know what I mean.

[Steve]

This is probabely a very dumb question - but why don't they have all the primaries on one day?

Historically, of course, travel was a lot harder. It made more sense for candidates to stump in one state, have a vote, and then move on to the next state.

Today, it would be more feasible to hold a national primary, but it would clearly be super-expensive. The best-financed candidates would probably blow everyone away. Under our present system, it's possible for a lesser-known candidate to campaign in a couple states, leverage success in those states into increased financing nationwide, and follow that path to the nomination. It would be really hard for anyone who isn't extremely well-known already, or super-rich and able to self-finance, to compete in a 50-state primary held on a single day.

[Jim Parish]

why don't they have all the primaries on one day?

The scheduling of primaries (and caucuses) is largely in the hands of the state legislatures, and - for example - New Hampshire highly values the cachet of being "first in the nation". It's not clear that there's any body with both the inclination and the power to organize things more sensibly.