Revisiting the breadwinner norm: Replicating Bertrand, Kamenica, and Pan (2015)

I perform a narrow and wide replication of the labor force participation analyses in Bertand, Pan, and Kamenica (2015), which finds a negative relationship between realized and predicted female breadwinning and wives' labor force participation. Their results replicate in a narrow sense, even when samples from the same data sources cannot be perfectly reproduced. In the broader replication, I test whether the results are robust to two standard adjustments from labor economics: using hourly wages rather than annual earnings to estimate potential relative income, and using predicted rather than observed earnings or wages for husbands just as for wives. Both adjustments decrease the magnitude of the estimated negative relationship substantially. When combined, they yield a positive coefficient in a cross-sectional analysis, and a precise null when couple fixed effects are incorporated in a longitudinal analysis. This replication casts doubt on the conclusion that wives leave the labor force when they become likely to outearn their husbands.