Human-induced global climate change has emerged as the single most important environmental and societal concern of the 21st century. We have made considerable progress in understanding the responses of terrestrial ecosystems to elevated CO₂ and atmospheric warming. However, less is known about how ecosystems will respond to changes in the amount and timing of precipitation. Here we use meta-analysis to synthesize results from over 40 world-wide ecosystem-scale precipitation manipulation experiments. Effects sizes were calculated as a Hedges' d, or standardized mean difference. The ‘treatment' plots refer to the ‘dry' plots. In rainfall exclusion experiments, these were the true treatments; in rainfall addition experiments, these were the controls. Results showed that treatments significantly altered site moisture regime, with significant reductions in both soil and leaf water potential in response to reductions in rainfall. Reduced rainfall also significantly altered ecosystem carbon and nitrogen cycling, with significantly lower aboveground biomass and net primary productivity (NPP), total soil respiration, soil solution nitrogen export, and foliar nitrogen concentrations in the ‘dry' treatments compared to the ‘controls'. The magnitude of the response was generally similar for all response variables (Hedges' d ranged from -0.24 for leaf water potential to -0.80 for aboverground NPP), and generally increased over time. The magnitude of the response was generally insensitive to site factors such as species composition, biome, site nutrient status and climate, and to interactions with other ambient and experimental global change drivers such as warming, elevated CO₂ and N additions. This could be due to a fundamental resilience of ecosystems, or to an insufficient number of experiments for inclusion in a robust meta-analysis. The authors recommend that the global change community develop a new generation of replicated precipitation change manipulation experiments that use common protocols across as wide a range of biomes, ecosystems and locations as possible.