The wet-gets-wetter, dry-gets-drier paradigm (WWDD) is widely used to summarize the expected response of the hydrological cycle to global warming. While some studies find that changes in observations and climate models support the WWDD paradigm, others find that it is more complicated at local scales and over land. This discrepancy is partly explained by differences in model climatologies and by movement of the wet and dry regions. Here we show that by tracking changes in wet and dry regions as they shift over the tropics and vary in models, mean precipitation changes follow the WWDD pattern in observations and models over land and ocean. However, this signal is reduced and disappears in model dry regions, when these factors are not accounted for. Accounting for seasonal and interannual shifts of the regions and climatological differences between models reduces uncertainty in predictions of future precipitation changes and makes these changes detectable earlier.