Saturday, November 08, 2008

The Ground Game -- Cause or Effect.

I am about to (gulp) open a semi debate with Nate Silver.

Silver argues that the data show the importance of the get out the vote ground game.

Indeed there is a strong correlation between the difference in the fraction of people who said they were directly contacted by the Obama campaign and the McCain Campaign and the difference between Obama's performance on election day and in the composite of polls.

If one interprets this as cause and effect one calculated "Roughly speaking, each marginal 10-point advantage in contact rate translated into a marginal 3-point gain in the popular vote in that state."

However, there is another explanation of the correlation. Many people are contacted by a campaign when the campaign has a lot of volunteers. The campaign has a lot of volunteers when the enthusiasm level of its supporters is high. If a large fraction with very high enthusiasm (enough to volunteer) is positively correlated with a large fraction with enough enthusiasm to actually vote, then the pattern is explained. Note this is possible even if one assumes that volunteers, GOTV etc have no effect on voting at all.

The pattern can be explained by enthusiasm causes a lot of people to volunteer which implies a large fraction of voters are contacted and also that enthusiasm causes high voting for the candidate compared to telling a pollster that one supports the candidate but then not bothering to vote.

A dumb idea about how to tell the direction of causation is to look at what happens with paid GOTV workers. Paying people to GOTV isn't a sign of enthusiasm. The conventional wisdom is that such an approach doesn't work and that you need volunteers. If one made the conventional wisdom shut up at that point and declined to think for a second, one might conclude that this shows that the causation is enthusiasm causes high GOTV and coincidentally votes not GOTV wins candidates votes.

However, the standard explanation also makes sense. Paid GOTV workers are not convincing but arther irritating. GOTV with volunteers work, because it is clear that the volunteers really believe what they are saying.

I don't see any reliable way to determine true causation.

No comments: