For four decades, American boys and girls finishing high school described their politics in almost the same terms. Then, around 2016, the two lines split — and never came back together.
Counting everyone who calls themselves liberal, very liberal, or radical, the share of female seniors on the left climbed from 23% in 2016 to 35% in 2023. Among male seniors it drifted the other way. A gap that was barely visible for forty years blew open to ~16 points.
“The election of 2016 reads in this data like a fault line.”
Each band is a self-description, stacked to 100% of seniors in each year. Toggle between the sexes and watch the blue (liberal) bands swell for girls while, for boys, it is the gray “none / don’t know” band that grows — disengagement rather than conservatism.
Here each sex is reduced to a single number — the left share minus the right share. Above the line means net-liberal; below means net-conservative. From 1976 to about 2015 the two sexes stayed within a few points of each other. After 2016 the female line launches upward while the male line sinks below zero.
The bars show the change in each label since 2016, in percentage points, on a shared zero axis. Girls poured into liberal, very liberal and radical. Boys mostly drained out of moderate — and the slack was absorbed not by the right but by “none / don’t know.”
1. The split is recent and abrupt. For roughly forty years the male and female trajectories were nearly indistinguishable. The widening is concentrated in the years after 2016, not a slow drift across the whole period.
2. It is asymmetric. Girls moved toward a position (the left). Boys moved away from one (the middle) without clearly adopting another — their largest gain is in the non-committal “none / don’t know” response.
3. The far edges grew for girls. The female “radical” share, near zero for decades, multiplied several times over after 2016, and “very liberal” rose alongside it — the female shift is not just more liberals but more intense ones.
A caution. These are self-labels from teenagers, smoothed annual estimates with sampling uncertainty. They describe identity, not turnout or vote choice — but the pattern is large, consistent, and mirrors trends seen in other young-adult surveys.