Sometimes a single demographic variable is very predictive, but often you don't get much deterministic information until you combine two or more cross-tabs. The big three R's are Religion, Race and Region. Wealth alone doesn't seem to have a huge effect on voting, except at the really high levels. However, the combination of wealth with Religion or wealth with Region (rural) is very significant.
One of my favorite blogs is by Andrew Gelman, a bayesian over at "Statistical Modeling, Causal Inference, and Social Science". I haven't been there for a while, but found these two posts really interesting:
So religion matters. For whom does it matter? Does it matter for the frustrated masses, seduced by emotional issues, or for the less economically-pressed elites? We can answer the question by measuring the religious/secular gap among voters at different income levels.
Fig 2. Support for George W. Bush, as a function of income, plotted separately for frequent church attenders, moderate attenders, and non-attenders. The difference between rich and poor is large for religious people but disappears among the non-religious.
STATE BLOGS
