Sunday, November 30, 2003


In statistics, “ecological” correlations have nothing to do with environmentalism. The term refers to correlations between grouped data. The correlation between national IQ levels and national income levels that Chris Brand often refers to is an ecological correlation and H.C. Lindgren’s finding of a .61 correlation between high income and voting against Richard Nixon in the the 1972 U.S. Presidential election is another example.

Because ecological correlations often seem surprisingly high, some statisticians and social scientists think they should be devalued in some way. I disagree. There is a brief discussion of the issue about half way through my article here -- where I note what is probably the most spectacular ecological correlation ever reported in the social science literature -- a correlation of .9 -- which showed that people who have most contact with Australian blacks like them the least. Which in turn shows that the racial dislike concerned is postjudice rather than prejudice -- i.e. the fruit of experience rather than ignorance.


No comments: