The link: One Cohort at a Time: A New Perspective on the Declining Gender Pay Gap (NBER).

The abstract:

This paper studies the interaction between the decrease in the gender pay gap and the stagnation in the careers of younger workers, analyzing data from the United States, Italy, Canada, and the United Kingdom. We propose a model of the labor market in which a larger supply of older workers can crowd out younger workers from top-paying positions. These negative career spillovers disproportionately affect the career trajectories of younger men because they are more likely than younger women to hold higher-paying jobs at baseline. The data strongly support this cohort-driven interpretation of the shrinking gender pay gap. The whole decline in the gap originates from (i) newer worker cohorts who enter the labor market with smaller-than-average gender pay gaps and (ii) older worker cohorts who exit with higher-than-average gender pay gaps. As predicted by the model, the gender pay convergence at labor-market entry stems from younger men's larger positional losses in the wage distribution. Younger men experience the largest positional losses within higher-paying firms, in which they become less represented over time at a faster rate than younger women. Finally, we document that labor-market exit is the sole contributor to the decline in the gender pay gap after the mid-1990s, which implies no full gender pay convergence for the foreseeable future. Consistent with our framework, we find evidence that most of the remaining gender pay gap at entry depends on predetermined educational choices.

Shorter version: The payment gap between men and women fell because it fell for the younger workers, and it fell for them not because more (young) women get high-paying jobs but because older people keep working in high-paying jobs for longer than before, so there are less of those for the younger workers. Those would have gone to men, so it's men who perceive a loss relative to women, although it's actually the entire cohort having less high-paying jobs available.

Some thoughts: A model and some data does not a full explanation make, but it's an interesting way to look at wage dynamics I hadn't considered before. Cohort effects can be tricky to think about and lead to unintuitive effects if you don't model the situation explicitly. In particular,

• You can have a narrowed gender gap without any actual cultural improvements in gender equality in hiring and wage decisions, simply as the side effect of there being less high-paying jobs available for the younger.
• As the paper indicates, insofar as the model is right we shouldn't expect the gender gap to get better overall than it is for the younger workers now (i.e., not great).
• AI and automation in general (I had to mention this, of course) deepen this dynamic. as it's often used to augment the productivity of highly-paid workers, reducing the availability of those positions for younger workers: Why hire more young coders if you think you can just give Copilot to your existing team of ancient forty-something engineers? The economic impact on the younger cohort of workers is equally dire for everybody, but because high-paying engineering jobs went/go disproportionally to men, the loss in expected employment is higher for them.

The scenario makes doubly inaccurate the occasional male complaint about "women making it impossible to get jobs." Any reduction in the bias of a process will of course look like a loss to the group that was favored by the bias, but that might not even be what's happening. Instead of "women are taking all the good jobs," the authors of the paper suggest — oversimplifying — that "older men are keeping all the good jobs for themselves."

All models are wrong, say statisticians; some are useful. However right or wrong this paper might be, at the very least it's useful to expand the toolset of potential mechanisms we use to think about this.