A Midpoint Test: What the Evidence Has Actually Established

The Brazilian judge evidence now gives us more than a headline result. We have an event timeline in which recorded output falls during leave and returns afterward, alongside a conceptual distinction between employment penalties and productivity penalties. The midpoint question is whether these pieces fit together into a defensible interim judgment: did judges merely maintain a sentence count, or does the wider evidence support the narrower claim that observed work did not suffer a persistent post-leave decline? The answer requires connecting the timeline to quality, case allocation, and career outcomes without treating any one measure as conclusive.

The most useful way to proceed is to separate what the data directly record, what those records make plausible, and what remains outside their reach. This produces a stronger conclusion than either “nothing happened” or “motherhood reduces productivity.” It also shows why the credibility of the result will depend on measurement and comparison, questions that become more important once the outcome checks are assembled.

From a leave interruption to a post-leave result

The timeline supplies the first piece of the argument. A judge's sentence output is stable during pregnancy, drops sharply during the period of maternity leave, and recovers when the judge returns. The decline during leave is not evidence of a lasting productivity penalty because the worker is not in the ordinary production state. It is an expected response of the individual output measure to formal absence, even while substitute judges may keep the court's docket moving.

The post-leave period changes the question. Once the parent resumes work, a persistent productivity explanation predicts that sentence output should remain below the earlier comparison level. Instead, the study reports a quick rebound. Any decline is concentrated in the first transition month, when the judge is returning to a docket and routine that have been managed during the absence. After that transition, the average effect is close to zero and statistically insignificant for mothers. Fathers show the same broad pattern after their shorter paternity leave.

This is more informative than a single null coefficient because the whole sequence behaves as the institutional story predicts. Output is affected when a judge is away, then returns when the judge resumes work. Pregnancy does not show the kind of broad early collapse that would make the leave period unnecessary to explain. The pattern is therefore consistent with a temporary interruption rather than a new, lower post-birth level of measured work.

The confidence interval makes the interim conclusion more precise. The study can rule out post-leave losses larger than 10.9 percent over the full post-leave period, and larger than 6.5 percent after the first transition month is removed. These bounds do not establish that the true average effect is exactly zero. They leave room for small effects that the study cannot distinguish precisely from zero. They do, however, make a moderate continuing decline difficult to reconcile with the observed estimate for the matched judge sample.

A second piece comes from other official output measures. The number of settlement agreements ratified and hearings held also shows no evidence of a persistent change during pregnancy or after parents return. That does not turn three administrative counts into a complete measure of judicial work.