It is known that the semiannual variation in the Dst index is excessively large compared to all other indices of geomagnetic activity. This has been interpreted in terms of a separate non-storm component which forms roughly one half of the whole semiannual variation in the Dst index. Since this component is not related to storms or the ring current it should be removed from the Dst index. We show how the non-storm component arises from the seasonal variation of the magnetic field at the Dst stations and from the erroneous treatment of the quiet-time curve during the construction of the index. Moreover, we reconstruct a corrected Dst index which is purified from the non-storm component and show that then the semiannual variation indeed attains the same level as in other geomagnetic indices. This correction will greatly affect earlier estimates of the physical causes of semiannual variation based on the Dst index. Since the correction will reduce the power at periods even longer than the semiannual period, in particular the annual variation, it has important consequences also on other types of studies based on the Dst index.