Prognostication

When making small talk with regular folks here (soup vendors, satay sellers, taxi drivers, etc.), the question of what I do often comes up.  When I say that I study Indonesian politics, almost without fail the next question is "can you tell me who’s going to win the election in 2009?"  The answer to this question is that I can sort of explain how different candidates are polling right now, but that’s not the same as actually knowing who’s going to win in 2009.  And everyone knows this information.  I’m not shedding any new light on that topic by recounting the same information that everyone sees in the newspaper already.

But there is a larger point, though, which has been made by Andrew Gelman quite elegantly in the US case.  It’s not clear that using poll numbers right now to say something about elections in the future makes any sense at all.  The question normally posed is something like "if the election were held today…"  But the election is not being held today.  It’s not clear what the response to that question signifies.  Sometimes they ask "in the upcoming elections…," but what this really means is something like "if you could cast your vote right now for the election in 4 months so that nothing else that happens between now and then will affect your vote."  It is probably correlated with what people think about their choices right now, but that may or may not be correlated with what is going to happen 4 months from now (or 9 months from now in Indonesia).  Note, of course, that this does not mean that the polls are meaningless, or that poll results are just noise.  Gelman links to a very important paper of his that argues that changes in polls as elections approach reflect how voters are learning about where the candidates stand; in another paper they find more evidence of this.  So as people learn that Obama is a liberal communist Muslim radical Christian, their answers will change accordingly, and we will basically be able to predict election results based on simple fundamentals that always work like age, race, income, and the national economy.

Incidentally, I am a big fan of Andrew Gelman’s blog.  Its appeal is probably mostly to academic types, but I like how crabby he is about people who make inscrutable graphs in their presentations of data.

Posted in Politics
On Twitter
Categories
Archives
%d bloggers like this: