Since Sandmo (1981), many articles have analyzed optimal fiscal policies in economies with tax evasion. All share a feature: they assume that the cost of enforcing the tax law is exogenous. However, governments often invest resources to reduce these enforcement costs. In a very simple model, we incorporate such investments in the analysis of an optimal fiscal policy. We characterize their optimal level and we show numerically how they interact with the other dimensions of the optimal fiscal policy. Finally, we highlight the differences between our results and those obtained in a model without investment in the tax administration.