A recent paper by a Harvard Ph.D. student illustrates how science can fail to answer the question (hat tip: Tyler Cowen). The paper's method: Use Google searches using the N-word as a proxy for racism; areas with the highest level of such searches (e.g., West Virginia) are about 5 points less for Obama than areas with the lowest level of such searches (e.g., Laredo, TX) when other factors are controlled for; conclude, "The estimates imply that racial animus in the United States cost Obama three to five percentage points in the national popular vote in the 2008 election."
Hardly. 100% of the work of the conclusion is based on one's priors; the method has no value whatsover in determining whether Obama was helped or hurt in 2008 by race-based voting. The author may be correct; so may a reader who thinks that Obama was helped overall by race-based voting from voters in places like the Virginia and North Carolina suburbs, places considerably more important than West Virginia where he was likely hurt.
Since this is an academic blog rather than a rant-blog, I will not rant here. I will instead say that the paper just noted is a good example of how science can be applied in a way that amounts to porn for red advocates, blue advocates, and those like myself who are emotional about the misuse of science as a tool for ideological advocacy.
Comments