Targeting in Tax Compliance Interventions : Experimental Evidence from Honduras

Tax authorities often use low-cost communication with taxpayers to encourage voluntary compliance and avoid other costly interventions. This paper reports findings from an experiment with more than 30,000 taxpayers in Honduras, designed to assess how taxpayers with different risk scores respond to a communication intervention. Across several outcomes, the average effect of the intervention on compliance was 0. Contrary to the expectation of experts surveyed, only taxpayers considered to be at low risk of noncompliance increase their filing and reported income. Using rich administrative data and a causal forest algorithm, the paper finds that ex-ante predicted risk and responsiveness to the intervention are negatively correlated. These findings can inform the design of targeted interventions by tax authorities