Tag
heterogeneous_treatment_effects
10 findings
Economics (general)Agnostic Machine Learning Heterogeneity Detection in Field TrialsWhile overall average treatment effects were small and statistically insignificant (-1.4 to -2.7 percentage points, p > 0.10), generic ML inference uncovered significant treatment effect gaps between the top and bottom quintiles of 20.9 percentage points for social stigma (p = 0.009) and 12.6 percentage points for salient stigma outreach (p = 0.022).RCTLabour & employmentBaseline Work HeterogeneityLabor force participation increases across public and private sectors were largest among women who had not previously worked for MGNREGS and whose husbands held conservative gender views.RCTEconomics (general)Compositional Selection Shift Under Zero Average Treatment EffectsProfessional stigma framing shifted applicant composition significantly, increasing the share of rich applicants by 62% (p < 0.01) and currently working applicants by 28% relative to control, despite near-zero net application growth.RCTLabour & employmentEducational Heterogeneity in Information FrictionsHigh-school educated jobseekers doubled their permanent employment rate from 5.8% to 11.6% (ITT = +5.76 percentage points, p < 0.05) and increased formal employment by 45% (+4.82 percentage points, p < 0.10), driven by an 8.7% reduction in reservation wages (p < 0.10); post-secondary jobseekers showed no statistically significant change in permanent employment (-2.6 percentage points, p > 0.10) or reservation wages (-3.5%, p > 0.10).RCTMethods & evidenceHonest Split-Sample ML Heterogeneity InferenceIdentified statistically significant treatment effect divergence between the top and bottom predicted individual treatment effect (ITE) quintiles for social stigma (p = 0.009 in Exp 1, p = 0.022 in Exp 2) and professional stigma (p = 0.015 in Exp 1), uncovering hidden heterogeneity where average treatment effects were near zero.RCTDevelopmentHuman Capital Moderation of Risk HedgingAmong literate household heads, insurance provision increased the likelihood of investing in risky cash crops by 15 percentage points (p < 0.05), whereas among illiterate household heads, the treatment effect was indistinguishable from zero.RCTLabour & employmentLow-Ability Foot-in-the-Door EffectCandidates at the 25th percentile of baseline ability gained a 25 percentage point increase in post-intervention employment probability and earned $11.01 more per day. In contrast, candidates at the 75th percentile experienced a 1.5 percentage point decrease in employment probability and earned only $2.20 more per day.RCTLabour & employmentPrior Research Experience Return MagnificationAcquiring short-term work experience yielded a statistically significant positive wage interaction for candidates with prior research experience (interaction coefficient = 1.079, standard error = 0.618, p = 0.06), demonstrating magnified wage gains for research-experienced individuals.RCTEducationPrognosis-Score Stratified Intervention ResponsivenessParticipants in the top third of baseline degree propensity achieved a 65.7 percentage point increase in six-semester college enrollment (TOT, p < 0.05) and a 41.9 percentage point increase in degree completion (TOT, p < 0.10), whereas lower propensity terciles exhibited no statistically significant gains.RCTLabour & employmentSelective Responsiveness of Unmotivated Workers to SanctionsThe interaction between accountability treatments and baseline reciprocity was consistently negative (e.g., Monetary Penalty x Baseline Reciprocity coefficient = -0.269, p = 0.022), demonstrating that accountability sanctions selectively increase prosocial performance among unmotivated workers who have higher baseline absenteeism.RCT