This paper provides evidence that a tax on mobile money transactions shrinks its base by an order of magnitude more than taxes on income or turnover do, that the base responds to whether it is taxed rather than to how much, and that the damage spreads through the payment network to users and payments the law never taxed. First, using the complete transaction records of Tanzania’s three largest providers around the introduction of a notched levy on person-to-person transfers and cash withdrawals in July 2021, we show that the levy reduces the volume of taxed transactions by 23 percent and their average value by 13 percent relative to untaxed transactions on the same platform, an elasticity with respect to the net-of-tax price of about 10. Second, we show that a cut of about 30 percent seven weeks later brings none of it back, nor do two further cuts, while the removal of the levy on transfers in 2023 does, slowly. Third, we document that relationships between senders and receivers survive at two-thirds the normal rate, that new ones form at less than half the pre-levy pace, and that senders stop rerouting activity around recipients who go quiet; none of these margins recovers when the tax rate falls. Fourth, we find that the levy reaches activity it does not tax, as wages paid into wallets fall by a fifth, and that senders of every size cut back by a similar share, whether or not they have a bank account. Fifth, we estimate that the levy destroys 47 cents of surplus for every shilling it raises, several times the cost of a value-added tax in comparable settings.