We use the grouped fixed-effect estimator by Bonhomme and Manresa (2015) to control for time-varying heterogeneity. Cannot use did approaches —> parallel-trends absolutely do not hold, all trend no treatment.

The paper is joint with @bsiflinger.bsky.social! A quick run-down of the main result: After controlling for time-varying heterogeneity, effects of an unplanned pregnancy followed by an abortion —> no effect on a woman’s mental health.