We investigate the behavioural responses of individual taxpayers to changes in marginal personal income tax rates applying empirical bunching methodology to tax administrative data from Zambia over the period from 2014 to 2021. We find evidence for excess bunching at the first kink in the tax schedule for all years but less evidence of bunching at the second and third thresholds. While bunching is considerable and behavioural responses are observed to changes in the location of the kinks over time, bunching at reference points (‘round-number bunching’) also appears large. Implied elasticities of taxable income are however not remarkable, and comparing actual and estimated counterfactual wage distributions reveal that missed tax revenue arising from the excess bunching is limited. This is consistent with the observed bunching reacting sharply and immediately to changes in the location of the kink points over time, suggesting that observed behavioural change is driven by reporting behaviour rather than real economic responses.