Future two-variable surveys like this should definitely consider taking Dave's Hacker Survey data (http://davelyon.net/mldata.html) into perspective. The 21-35 range seems to cover something like 1.5 sd on either side — around 85% of us if we assume normality.
Investing more of these poll choices to age resolution within that range would have been interesting.
Investing more of these poll choices to age resolution within that range would have been interesting.