I use firm-level data to show that neither the Log-normal nor the Pareto distribution can approximate the shape of the productivity distribution along the entire support. While the former under predicts the thickness of the right tail, the latter does not capture the shape of the left one. Using...