I am on the 2025-2026 job market! Here is the abstract of my JMP:
This paper revisits the identification of the gender gap in the effect of parenthood on earnings. I formalize the setting as a staggered adoption design and examine three identification frameworks: difference-in-differences (DID), triple differences (TD), and normalized event studies (NTD), the common empirical strategy in this literature. I show that NTD fails to identify the target causal estimand when parallel trends are violated. Drawing on predictions from human-capital models of fertility and earnings, I argue that parallel trends are likely violated, rendering DID invalid. I further show that TD is likely invalid due to life-cycle variation in counterfactual gender inequality. These insights motivate two alternative frameworks. First, I show that NTD can identify a new causal estimand, the change in the gender earnings ratio due to childbirth, without additional assumptions. Second, I show that adding a null-effect assumption for fathers yields a bias-corrected identification formula for NTD. Using Israeli administrative data, I document empirical patterns for both alternatives and discuss their relationship to the baseline NTD estimand. I finish with discussing aggregation across multiple treatment groups.
I also have a website (and package) for the paper — childpen. Please let me know if you find an issue