Within-Firm Spillover Dilution Control
RCTClinical Trial
Randomizing individual branch locations of multi-establishment corporations risks information sharing and cross-branch treatment contamination that dilutes estimated intervention effects.
Picture this
Imagine testing a new internal recipe at one fast-food branch while attempting to keep a sister branch in the same city as an unexposed control group. Store managers routinely talk and share techniques, causing the new recipe to spread to the control store and ruining the experiment.
What the evidence says
Removing multi-establishment corporate branches prevented internal information spillovers, maintaining orthogonal treatment variation across 7,438 independent establishments.
- Who was studied
- N = 8,232 initially randomized establishments pruned to N = 7,438 unique firms across mainland France.
- How
- Ex-post experimental sample restriction excluding multi-branch corporate entities prior to treatment effect estimation.
What to do
Exclude multi-branch corporate establishments from organizational field trials to prevent internal knowledge spillovers from contaminating control units.
From the source
"We thus exclude firms that had multiple branches included in the experiment due to the possibility of within-company spillover effects that would dilute treatment effects. Our final sample comprises 7,438 unique firms."
Are_Active_Labor_Market_Policies_Directed_at_Firms_Effective_Evidence.pdf