As I suspected, the analysis is based on aggregates, not a longitudinal panel, which limits the kinds of inference you can draw. It does show generational differences, but does not provide any tests of statistical significance, so we don't know how much confidence to place in the findings.
But we do know that historically, getting out the vote is more important than any other factor by far in determining who wins elections. We also know that Democrats have a worse turnout track record than Republicans in mid-terms. It doesn't have to be that way--it's a statistical fact, not a law of physics. Now would be a good time to change the pattern.