Interesting. But this is pointing out a classic statistical result that cohort, calendar time, and age effects can’t be nonparametically separated. So we cant know, based on just observational data, what is society getting more liberal, people getting more liberal, or less liberal cohorts dying (unless you’re willing to assume one of these trends takes a low dimensional functional form, but that’s sort of assuming the answer).
One data point that won't fit nicely: a qualitative pattern I've seen is when some of the G.I. Generation and Boomers who watched a great deal of Fox News tended to shift their views to right of Tea Party-type beliefs. In a number of US nursing homes, the tendency to play Fox News 24/7 was also unkind to the presumption of increasing liberalization.
Would be interesting to see if the slight trend towards liberalism as people age persist if you split into male and female. Part of it might be men dying younger.
