We exploit the panel dimension of the Canadian Financial Monitor (CFM) data to estimate the impact of retail payment innovations on cash usage. We estimate a semiparametric panel data model that accounts for unobserved heterogeneity and allows for general forms of non-random attrition.