This presentation will discuss the climate elasticity of streamflow, i.e. the sensitivity of the changes in streamflow related to changes in a climate variable (Schaake and Liu, 1989). Elasticity e can be defined as e = DeltaQ / DeltaX , where DeltaQ represent the streamflow anomalies and DeltaX the clim atic variable anomalies. The past has seen a wealth of elasticity studies based on models (i.e. based on flows simulated by a hydrological model fed with different inputs). But elasticity can be directly derived from data (see e.g. Sankarasubramanian et al., 2001; Coron et al., 2014), provided that long-term records are available. We wish here to discuss the differences between the two estimates, by using models of increasing complexity over a range of catchments. The fact that all the models we tested had a tendency to underestimate the observed elasticity leads to questioning whether hydrological models do not systematically underestimate the hydrological impact of climate change. We discuss possible solutions to “force” models to better represent data-based elasticity.