The Causal Impact of Taking Parental Leave on Wages: Evidence from 2005 to 2015

Within the context of Luxembourg, we analyze the causal effect of parental leave take up on post-birth hourly wages of an important subpopulation of parental-leave-eligible mothers, i.e., first-time mothers that are always employed regardless of having taken parental leave. In our analysis, we simultaneously address selection of eligible mothers in taking parental leave, and selection of eligible mothers into employment. To deal with the first complication, we assume that conditional on observed pre-intervention covariates there are no unobserved factors associated to both the assignment to parental leave and the potential post-birth hourly wages. To this end, we control for a rich set of pre-intervention characteristics that were obtained from Social Security administrative data from 2005 to 2010. The second complication arises since the outcome of hourly wages is only defined for the (post-birth) employed subpopulation. We deal with selection into employment by utilizing a Principal Stratification framework and recent non-parametric bounds. We argue that the monotonicity-type assumptions employed for bounding causal parameters are plausible in the context analyzed and potentially weaker than conventional alternatives. Our estimated bounds allow us to undertake inference for a subpopulation of parental-leave-eligible mothers that are always employed regardless of having taken parental leave. This subpopulation accounts for about 80 percent of all eligible mothers in our dataset. Our estimated bounds on average effects of parental leave take-up on hourly wages are consistent with important, albeit statistically insignificant, reductions in all periods analyzed (i.e., 2, 3, 4 and 5 years after birth). Interestingly, we find evidence of heterogeneous impacts of parental leave take-up across the distribution of post-birth wages. Our estimated bounds, in general, show that the quantile treatment effects of parental leave take-up on post-birth wages of always employed mothers are negative and significant in most quantiles above the median.