Simulations are made of the long-term mean clear-sky and total-scene components of the earth's radiation budget (ERB), using detailed solar and thermal infrared radiative transfer schemes. These employ as input a wide range of multi-year observations of atmospheric, surface and cloud parameters. For clear skies, differences between the simulations and results from the Earth Radiation Budget Experiment (ERBE) are generally within the expected accuracy of the observations and calculations, with the most notable disagreement at high latitudes. For cloudy skies, the agreement is less good, but still generally within the systematic error bounds. the absorbed shortwave radiation (ASR) is simulated least well over oceanic regions, with the pattern of disagreement resembling the pattern of response when low cloud cover or liquid water path is perturbed; this suggests that the representation of low cloud cover or liquid water path dominates the ASR discrepancy between the simulation and ERBE. the simulated outgoing longwave radiation (OLR), which is sensitive to the high-cloud emissivity, differs most from the ERBE OLR in deep convective regions; whilst assuming the high cloud to be black lessens agreement between the simulation and observations in some areas, agreement is substantially improved in others. This indicates that better knowledge of high-cloud emissivity remains an important issue. the spatial and temporal variations of the sensitivity of the ERB to perturbations to low, mid and high cloud properties are investigated. Principal results include (a) the dominant importance of cloud amount, as compared to cloud temperature, in determining the sensitivity of the ERB to high-cloud properties, and (b) the greater sensitivity of the net ERB to low-cloud properties than to those of mid and high cloud. A major uncertainty in the calculations is the degree of cloud overlap; the inclusion of overlapping clouds is found to be capable of reversing the net high-cloud sensitivities in certain locations. Global mean sensitivity to low cloud perturbations are compared with previous results obtained using a GCM. the simulated sensitivity of the ERB to low cloud cover is substantially larger than in the previous results, although the effects of changes in liquid water path are similar.