Water use efficiency (WUE) is a key indicator to assess ecosystem adaptation to water stress. Rain use efficiency (RUE) is usually used as a proxy for WUE due to lack of transpiration data. Furthermore, RUE based on aboveground primary productivity (RUEANPP) is used to evaluate whole plant water use because root production data is often missing as well. However, it is controversial as to whether RUE is a reliable parameter to elucidate transpiration efficiency (TE), and whether RUEANPP is a suitable proxy for RUE of the whole plant basis. The experiment was conducted at three differently managed sites in the Inner Mongolia steppe: a site fenced since 1979 (UG79), a winter grazing site (WG) and a heavily grazed site (HG). Site HG had consistent lowest RUEANPP and RUE based on total net primary productivity (RUENPP). RUEANPP is a relatively good proxy at sites UG79 and WG, but less reliable for site HG. Similarly, RUEANPP is good predictor of transpiration efficiency based on aboveground net primary productivity (TEANPP) at sites UG79 and WG but not for site HG. However, if total net primary productivity is considered, RUENPP is good predictor of transpiration efficiency based on total net primary productivity (TENPP) for all sites. Although our measurements indicate decreased plant transpiration and consequentially decreasing RUE under heavy grazing, productivity was relatively compensated for with a higher TE. This offset between RUE and TE was even enhanced under water limited conditions and more evident when belowground net primary productivity (BNNP) was included. These findings suggest that BNPP should be considered when studies fucus on WUE of more intensively used grasslands. The consideration of the whole plant perspective and "real" WUE would partially revise our picture of system performance and therefore might affect the discussion on the C-sequestration and resilience potential of ecosystems.