Do Rich Voters Really Vote Democrat? (from Marginal Revolution)

Alex Tabarrok (at Marginal Revolution) provides an interesting example of the "ecological fallacy". He mentions a very nice blog called Statistical Modeling, Causal Inference, and Social Science, done by Andrew Gelman, a professor in the Departments of Statistics and Political Science at Columbia University (definitely a new one on my own list). This latest post examines the supposed finding from the last election that low-income voters surprisingly seemed to vote Republican, while high-income voters went Democrat:

A recent post looks at the difference between red and blue states and red and blue individuals. We all know that in the recent election poorer states tended to vote Republican while richer states tended to vote Democrat. On the basis of the famous maps many people jumped to the conclusion that poorer individuals were voting Republican (Nascar Republicans) while richer individuals were voting Democrat (trust fund Democrats). But the inference is a fallacy, the ecological fallacy. In fact, high-income individuals, as opposed to high-income states, vote Republican with greater likelihood than low-income individuals (the effect is not huge and it may be declining but it is significant).

Click here for the whole article.

Within states, higher income counties vote republican, but when you pool the data across states, the reverse seems true. It's a great example of the weird (and incorrect) inferences you can come up with when you try to analyze data with pooled samples.

Very nice.