Andrew Gelman’s graph of age effects:
Within the African-American community there was, of course, very little age effect but among whites and Hispanics it was quite pronounced.
Looking at this is a reminder that it would be nice if age cohorts were broken down into four-year blocs on the exit polls. If we could compare an 18-21 bloc from 2000 to the 22-25 bloc in 2004 and the 26-29 bloc in 2008 then we could look at cohorts evolving over time. But the way the data’s currently organized, it’s a bit difficult to interpret.