Electromagnetic transfer functions at coastal sites are known to be strongly distorted by the conductivity of the seawater. This ocean effect is generally considered to be small for periods greater than a few days. We revise this statement by detailed and systematic model studies in the period range from 1 to 64 days, with subsequent comparison of the modelled and observed electromagnetic responses. The conductivity model consists of a radial symmetric (1-D) section that is overlaid by a thin spherical surface shell, the conductance of which is compiled using the NOAA ETOPO topography/bathymetry and map of sediment thicknesses. The simulations were performed for spatial resolutions of the surface shell of 5¿ ¿ 5¿, 2¿ ¿ 2¿ and 1¿ ¿ 1¿, respectively, and for two, continental and oceanic, underlying 1-D conductivity models. The inducing source is described by the spherical harmonic P10 in dipole coordinates. Comparisons are made for the coastal geomagnetic observatories Apia, Hermanus, Kakioka, Kanoya, and Simosato where an anomalous behaviour of the local response has been previously detected. From the comparison of observed and modelled responses we conclude that peculiarities in the observed coastal responses in the period range from 1 to 20 days can be explained to a large amount by induction in the oceans. We show that correction for the ocean effect results in responses that are much better interpretable by 1-D conductivity models compared to the uncorrected responses.