I responded: Just one of those things, I think.
Just to be clear, I am more than willing to believe the central point about the share of the population dropping while the share of the electorate holds relatively steady, but having dealt with more than my share of bad data, I get really nervous when I see a number like that hold absolutely steady.
I did a quick check of some of those 26% numbers online and they seem to be from actual exit polls. Them all being equal just seems like a coincidence. The part of the graph I really don’t believe is the sharp decline in % evangelical Christian. I’m guessing that the survey question on the exit polls is different than the survey question on whatever poll they’re using to estimate % evangelical.
And, since I have you on the line, here are some graphs from chapter 6 of Red State Blue State:
It seems that religious affiliation is becoming more of a political thing. Or maybe political affiliation is becoming more of a religious thing.
In any case, be careful about comparing time trends of survey questions that are asked in different ways.