Abstract: The "Votes" dataset contains the percent of the popular vote that was won by the Democratic presidential candidates in the 1980 and 1984 elections. Both candidates, Jimmy Carter in 1980 and Walter Mondale in 1984, were defeated by the Republican Ronald Reagan. (In 1980 an independent candidate, John Anderson, captured 6.7% of the national vote). Many states have persisting political traditions, so we expect similar behavior in two successive elections.
A scatterplot shows a strong positive linear relationship, lending support to this hypothesis. A cluster of states voted heavily Democratic in 1980 but not in 1984. A closer look at the data reveals that all of the states in this cluster are southern states. Creating a dummy variable for southern states and including this variable in a regression of 1984 percent on 1980 percent significantly improves the model.
Image: Scatterplot of 1984 percent vs. 1980 percent with southern states as red x's
