Baseline Remittance Level Heterogeneity
Financial behavioral interventions can generate divergent outcomes depending on pre-intervention baseline activity levels. Evaluating program impact solely across an aggregated population can mask substantial positive effects occurring within specific sub-groups.
Picture this
Imagine offering a fuel discount voucher to two drivers: one who rarely drives due to high fuel costs, and another who already drives every single day. The driver who rarely drove increases their trips significantly when given the discount, whereas the daily driver makes no additional trips because their driving behavior was already at capacity.
What the evidence says
Among migrants with below-median baseline remittances, labeling access significantly increased the probability of remitting (+1.0 percentage point, p < 0.01), weekly remittance frequency (+0.012 transfers, p < 0.01), winsorized weekly amount (+174.94 PhP, p < 0.05), and log weekly remittance amount (+0.083, p < 0.01), while showing no statistically significant effect on high-baseline remitters.
- Who
- 2,408 low-baseline remitters (below median pre-treatment remittances) and 2,043 high-baseline remitters (above median) out of 4,451 analyzed migrant workers in the United Arab Emirates [19, 20].
- How
- Ordinary least-squares (OLS) panel regression with individual and week fixed effects across subsamples split by baseline remittance volume during weeks 1–5 [17, 19-21].
What to do
Segment financial policy evaluations by pre-treatment transaction volume to detect intervention impacts specific to under-performing user segments.
From the source
"The estimated treatment coefficients indicate a consistent positive effect of labeling on all remitting dimensions when the sample included migrants that were sending below the median in the first 6-9 weeks before the labeling feature was activated in Padalapp."
A Field Experiment among Filipino Migrant Workers in the UAE