Precipitation efficiency (PE) relates cloud condensation to precipitation and thus reflects how much of the total atmospheric condensate reaches the surface as precipitation. Because the PE in convective storms is directly linked to their updraft- and downdraft dynamics, it is a helpful metric to identify convective processes that influence precipitation. However, km-scale model simulations do not properly resolve convective processes such as individual updrafts and entrainment, which raises the question if such simulations can accurately represent PE. Here, we present two methods to derive PE from standard model output. The first method estimates PE from the state variables vertical velocity, temperature and pressure, whereas the second method estimates PE from ice water path (IWP) and precipitation. We validate the proposed methods with the explicitly calculated PE using a set of idealized Weather Research and Forecast model simulations of organized midlatitude convective storms at different horizontal grid spacings. We show that PE can be reliably estimated from state variables with an error of less than 5%, partly due to error cancellation effects. Additionally, PE can be simulated by km-scale models within ~15% accuracy compared to large-eddy simulations (LESs). The IWP-method is slightly less accurate with a stronger grid spacing dependency of the error, but since it is based on observable quantities, it allows for a validation of simulated PE with satellite observations. Finally, we analyze the grid spacing dependency of the climate change signal of PE and find that future decreases in PE in LESs are robustly captured by km-scale models.