Follow the Many

The success of prediction markets in the electoral arena hinges on the idea that while talk is cheap, nothing sharpens the mind like real money. Combining the wisdom of the crowd with real money yields better predictions.

So, if how third-party observers choose to use their money tells us what they really think, can the same be said for campaign expenditures? (Key: Expenditures as opposed to campaign rhetoric.) Now, we could dissect the strategy of either presidential campaign (e.g., the famous McCain Strategy Briefing). But we have to contend with incomplete information, possible misdirection, and sample bias. It may be more instructive to analyze the financial actions of the most interested market actors: congressional candidates.

We want to know - where are congressional candidates up for election putting their hard-bought campaign dollars? Are Republican candidates, for instance, tying themselves to the GOP candidate? Or are they assuming a posture of independence from the party establishment? (It goes without saying that all candidates in the 2008 cycle are either attacking or ignoring President Bush.) How much of each candidate’s campaign expenditure co-opts the Obama theme of “change”, how much focuses on the candidate’s individual brand, how much on party affiliation, how much on the economy, etc.

From data on each candidate’s media expenditure - along a breakdown similar to that implied above - we can find regression coefficients that essentially mimic the market price for each “meme”. These market prices could be aggregated regionally or nationally or through district segmentation (”hard red”, “leans blue”, etc.). Extrapolating based on such prices at the state level should provide a better forecast of the electoral dominos than opinion survey-based polling.

An illustrative instance comes from the following ad released by Senator Gordon Smith (R-OR). He is not only allocating dollars to the “bipartisanship” meme, he is explicitly attempting to leverage the Obama brand.

Leave a Reply