The Limits of Behavioural Tax Nudges: Experimental Evidence from Tanzania

Sub-Saharan African countries face persistent challenges in mobilizing domestic tax revenue, while behavioural interventions have been promoted as a low-cost way to improve tax compliance. This paper reports a pre-registered randomized controlled trial involving 63,896 registered income taxpayers in Tanzania, assigned to six behavioural SMS treatment arms and a control group. The primary analysis estimates effects separately across six arms using cross-sectional ANCOVA, while a secondary McKenzie (2012) panel specification consolidates the treatments into three intervention families to improve statistical precision. The results show no statistically significant positive effects on aggregate tax payment activity or payment amounts. The findings are robust to alternative specifications, multiple-testing adjustments, randomization inference, and alternative assumptions about message delivery. The more efficient panel specification provides precise extensive-margin evidence across all four groups, though the confidence interval for corporate income taxpayers is slightly wider than the 2.7 percentage-point benchmark from the tax-compliance literature. On the intensive margin, confidence intervals rule out sizeable increases in payment amounts for pooled and presumptive taxpayers and increases of approximately 5.1% for personal income taxpayers, while the evidence for corporate income taxpayers remains inconclusive. The main exception is the stronger deterrence message, which increased the probability of making a payment among micropayers by approximately 6.7 percentage points during the intervention. This effect did not extend to payment amounts or persist beyond the treatment period. Overall, the findings suggest that generic behavioural SMS messages are unlikely to constitute an effective stand-alone revenue-mobilization strategy in this setting. Limited enforcement credibility is a plausible explanation for the aggregate null, although the experiment does not directly measure taxpayers’ perceptions of enforcement. The results suggest that behavioural communication may be more effective when combined with taxpayer-specific information and observable administrative follow-up that strengthens the credibility of deterrence.