Unconditional Quantile Earnings Gain
RCTReview
Average treatment effects on labor market earnings can be distorted by extreme high-earning outliers, masking zero or negative impacts for the majority of program beneficiaries.
Picture this
Imagine checking whether a new tutoring program improved class performance. Instead of calculating a single average test score that could be pulled up by one top student, you measure the test scores at the 10th, 50th, and 90th percentiles to prove that students at every level improved.
What the evidence says
The cumulative distribution of total earnings for treated youth shifted uniformly rightward across all quantiles (Mann-Whitney permutation p = 0.0005), confirming medium-term earnings gains (+15%, p = 0.0169) were distributed across low, middle, and high earners rather than driven by outliers.
- Who was studied
- N = 1,670 youth respondents surveyed at 4-year endline follow-up in Côte d'Ivoire.
- How
- Unconditional quantile treatment effect (UQTE) estimation and Mann-Whitney rank-sum permutation tests (10,000 replications) on 4-year total monthly earnings.
What to do
Estimate unconditional quantile treatment effects and run permutation-based rank-sum tests to confirm program earnings gains are distributed across the entire income distribution.
From the source
"The figure shows that the proportion of youths who earn less than a given amount is uniformly and significantly lower in the treatment group than in the control group. This suggests robust positive program impacts on earnings across the distribution."
Direct_and_Indirect_Effects_of_Subsidized_Dual_Apprenticeships.pdf