The meridional transport of salt is computed in a global eddy-resolving numerical model (1/12° resolution) in order to improve our understanding of the ocean salinity budget. A methodology is proposed that allows a global analysis of the salinity balance in relation to surface water fluxes, without defining a “freshwater anomaly” based on an arbitrary reference salinity. The method consists of a decomposition of the meridional transport into (i) the transport by the time–longitude–depth mean velocity, (ii) time–mean velocity recirculations and (iii) transient eddy perturbations. Water is added (rainfall and rivers) or removed (evaporation) at the ocean surface at different latitudes, which creates convergences and divergences of mass transport with maximum and minimum values close to ±1 Sv. The resulting meridional velocity effects a net transport of salt at each latitude (±30 Sv PSU), which is balanced by the time–mean recirculations and by the net effect of eddy salinity–velocity correlations. This balance ensures that the total meridional transport of salt is close to zero, a necessary condition for maintaining a quasi-stationary salinity distribution. Our model confirms that the eddy salt transport cannot be neglected: it is comparable to the transport by the time–mean recirculation (up to 15 Sv PSU) at the poleward and equatorial boundaries of the subtropical gyres. Two different mechanisms are found: eddy contributions are localized in intense currents such as the Kuroshio at the poleward boundary of the subtropical gyres, while they are distributed across the basins at the equatorward boundaries. Closer to the Equator, salinity–velocity correlations are mainly due to the seasonal cycle and large-scale perturbations such as tropical instability waves.