The vertical transport by shallow nonprecipitating cumulus clouds of conserved variables, such as the total specific humidity or the liquid water potential temperature, can be well modeled by the massflux approach, in which the cloud field is represented by a top-hat distribution of clouds and its environment. The massflux budget is computed by conditionally sampling the prognostic vertical velocity equation by means of a Large-Eddy Simulation of shallow cumulus clouds. The model initialization is based on observations made during BOMEX. Several different sampling criteria are applied. The presence of liquid water is used to select clouds, whereas additional criteria are applied to sample cloud updraft, downdraft and core properties. The budget of the massflux is presented and is compared to the vertical velocity variance budget. The massflux and vertical velocity variance budgets appear to be qualitatively similar. The massflux is driven by buoyancy in the lower part of the cloud layer, whereas turbulent transport is important in generating massflux in the upper part of the cloud layer. Pressure and subgrid-scale effects typically act to dissipate massflux. The massflux approach is verified for non-conserved variables. The virtual potential temperature flux and the vertical velocity variance according to the the top-hat approximation do not correspond very well to the Reynolds-averaged turbulent flux. The top-hat structure for the virtual potential temperature is degraded by lateral mixing and the subsequent evaporative cooling of cloud droplets which support the development of negatively buoyant cloud downdrafts. Cloudy downdrafts occupy about 20% of the total cloud area in the upper part of the cumulus layer, and are the cause that the vertical velocity variance is not well represented by the massflux approach, either.