Now take the parents of these two. The parents of the 18-year-old will have a child who can vote; the parents of the 17-year-364-days-old will not. Both pairs of parents may want to instill civic virtues in their child.

But only in the household with the young potential voter can the parents and the young voter go to the polls together. So are parents of newly enfranchised voters more likely to vote in elections, modeling good civic behavior, than parents of not-quite-enfranchised voters?

Any two pairs of parents are obviously different on a wide range a factors that influence voting, not just having a voting-age child or not. But for this research I was able to look at large groups of parents of eligible and ineligible teenagers, in which all the differences cancel out — except for whether their children are old enough to vote. I compared the turnout rates of parents whose children are barely old enough to vote to those of parents whose children are not old enough to vote by the same bare amount. I used an algorithm to pick an optimal window for each election, which means that the estimation windows varied from 38 to 56 days. As an alternative, I also compared turnout rates within windows of just one day, two days, three days, a week and a month.