07 January, 2008

Vacuous Political Speculation

Matt Yglesias makes a very important point about political science/journalism/speculation, though he may not know it:

I think it bears mentioning that it's always worth trying to not overread the trends. A month ago, it looked like Hillary Clinton would probably win the nomination. At the same time, it was clear back then that Obama wasn't prohibitively far behind in Iowa or anything. And it was clear that winning Iowa would give him a big edge in winning in New Hampshire. And it was also clear that if Obama swept Iowa and New Hampshire, the powerful culinary workers union was unlikely to back Clinton. And it was also clear that if Obama won in two lily-white states, that the odds favored African-Americans flocking to his banner in South Carolina. And it was clear that all that would create a lot of momentum working against Clinton moving into the big states.

We knew all that, but because it looked like Obama probably wouldn't win Iowa, that pro-Obama cascade looked unlikely. Then Obama did win Iowa, making the cascade look likely. And I think it is the most probable outcome. But just as Obama once looked like he would lose in Iowa and then won, just because Clinton looks down now hardly makes it impossible for her to recover.
The point is that while Clinton looked inevitable, she clearly was not. And everyone overestimated her chances of winning. Now, everyone is, I'll bet, overestimating Obama's chances of winning. I think that there are two problems.

The first problem is that political bloviating of the kind that we and Yglesias and others do is based on extremely unscientific methods. A popular tactic is to sketch a plausible "path to victory" for a candidate. This is all well and good and it's fun to imagine what it would be like if that happened, but what does it actually tell us? Narratives are useful for organizing history, but very very bad at predicting the future.

This is because you can come up with a narrative for every set of occurences and there is no standard by which you can choose one that is the most accurate. I could flip a coin five times, come up with a story about why I got Heads, Tails, Heads, Heads, Tails, and use that story to predict the next flip. But it is still going to be decided by randomness. We may have a bit more information about elections, but not much. The narratives that are bandied about distract us from that fact.

The second problem is that political pundits underestimate the amount of randomness that affects the election process and politics in general. Another way of saying this is that they overestimate the amount of information they have about the end result. This is partially because narratives play such a large role in our political thinking. It's very easy to be convinced by a story.

It's also important to keep in mind, that, looking back, the story that got everything right will appear to be prescient--but it is not. With a narrative for each set of outcomes, at least one of them will look good in hindsight even if the outcomes are totally random.


Elliot said...

It seems to me that the point of political narratives is not to predict the future but to influence it.

spencer said...

So are you saying that self-styled political pundits such as Yglesias, Sullivan and others are pushing narratives calculated to drive the political process? They certainly sell them as having some predictive power. A case in point:


Elliot said...

Perhaps I should not have said "the point"; rather, I should have said "the effect". With the exception of campaigns (which is where most of the "paths to victory" narratives originate, from the strategy itself), no, most political commentators do not sit around thinking up a narrative to push - although some do, if you read Yglesias' posts on Kristol today. Yet their acceptance of a narrative, to the extent that they are influential, does have the effect of driving the political process, rather than objectively predicting it.

The relationship between "facts" and "narrative" is by no means clean and neat, or one-way. The distinction between the two is very fuzzy. Most narratives attempt to be self-fulfilling prophecies, and thus, to become fact, not predict fact. If the narrative about Clinton's invincibility turns out to be false, then it is not because it lacked predictive power so much as because it lacked the power to convince enough people or the right people that it should be true.

Or is all that too postmodern?