The influence of surface albedo on tropical precipitation is widely appreciated, but albedo bias over snow-free areas in climate models has been studied little. Here, historical CMIP5 simulations are shown to exhibit large multi-model mean bias and intermodel variability in boreal summer mean surface broadband shortwave albedo. Intermodel variability in this albedo is globally coherent over vegetated regions, and correlates with intermodel tropical precipitation variability. Evidence supports the hypothesis that these spatially coherent albedo variations cause precipitation variations. Specifically, the spatial structure of albedo and precipitation variations are distinct, suggesting the latter do not cause the former by darkening soil. Furthermore, simulated interannual albedo variance is small compared to intermodel albedo variance, while the ratio of interannual to intermodel precipitation variance is much larger. Finally, imposing the dominant pattern of intermodel albedo variability in one climate model causes a precipitation change with structure similar to that of the intermodel variability.