This paper provides cross-country empirical evidence on the productivity of bank transaction taxes (BTTs). Our data set comprises six Latin American countries that have levied BTTs since the late 1980s: Argentina, Brazil, Colombia, Ecuador, Peru and Venezuela. We find that, for a given tax rate, revenue declines over time. Therefore, in order to meet a fixed revenue target in real terms, the tax rate needs to be raised repeatedly. However, we also find that successive increases in the tax rate erode the tax base by more than they raise revenue yield and that the higher the increase in the tax rate, the more and faster the tax base is eroded. We conclude that BTTs do not provide a reliable source of revenue, especially over the medium term.