The paper investigates the relationship between tax revenues and literacy level, using a panel-model approach. The dataset covers the period 1996 to 2010 and includes 123 countries. The estimations suggest that the assumed function is nonlinear, with inverted-U and U-shaped curves. More precisely, a very low literacy level is associated with reduced tax revenues. Furthermore, the government inputs increase as the literacy level increases, reaching a maximum point. Beyond this level, the tax revenues decrease even if the literacy has an ascendant tendency, registering a minimum level. Finally, the tax revenues increase in a parallel manner with the literacy index.