Daily and seasonal variability of long time series of magnetometer data from Dst stations is examined. Each station separately shows a local minimum of horizontal magnetic component near 18 local time (LT) and weakest activity near 06 LT. The stations were found to have different baselines such that the average levels of activity differed by about 10 nT. This effect was corrected for by introducing a new base method for the elimination of the secular variation. This changed the seasonal variability of the Dst index by about 3 nT. The hemispheric differences between the annual variation (larger activity during local winter and autumn solstice) were demonstrated and eliminated from the Dst index by addition of two Southern Hemisphere stations to a new index termed Dst6. Three external drivers of geomagnetic activity were considered: the heliographic latitude, the equinoctial effect, and the Russell--McPherron effect. Using the newly created Dst6 index, it is demonstrated that these three effects account for only about 50% of the daily and seasonal variability of the index. It is not clear what drives the other 50% of the daily and seasonal variability, but it is suggested that the station distribution may play a role.