|Man Without Qualities|
Monday, January 19, 2004
An astute if, perhaps, somewhat mischievous, reader informs the Man Without Qualities that all of the complaints here about the Iowa polls being misleading predictors of the caucus results are all wrong-headed. He says that it's the polls that count - not the caucus results:
Nobody thinks the Iowa caucuses are about the incremental contribution of the delegates selected in Iowa to the probability of getting the nomination -- that number is close to zero no matter who "wins" the caucuses or by how much.
Instead, they are about the ability of candidates to gain traction among somebody other than pundits. The caucuses are almost pure signal. In that case, the raw Zogby-type poll numbers (to the extent they are accurate) actually measure what people are interested in. Indeed, reporting the outcome of the caucuses is probably the mistake here.
When results don't match polls there are three possibilities: a) the polls predicted the raw totals, but the process outcome doesn't mirror the raw totals (your case); b) the polls failed to predict the raw totals because the electorate changed their minds at the last minute -- or lied to pollsters -- or didn't show up as the pollsters predicted they would in some differential fashion; or c) the polls were really political and not scientific documents, i.e. the polls were calculated to misreport in order to skew perceptions.
It seems to me that b) and c) are both more common, and more interesting, than a) but I have no real evidence on relative frequency.
So put all that in your pipe and smoke it!
Comments: Post a Comment