Using a database of 13 solar cycles of geomagnetic aa data, we obtained cor
relations between cycle averages of geomagnetic activity (and sunspot numbe
r) and the numbers of days with disturbance levels above certain aa thresho
lds. We then used a precursor-type relation to predict an average aa index
of 23.1 nT for cycle 23 and inserted this average aa value into the above c
orrelations to forecast the integral size distribution of geomagnetic activ
ity for the new cycle. The predicted size distribution is similar to that o
bserved for cycles 21 and 22 but most closely resembles that of solar cycle
18 (1944-1954), which was slightly smaller than cycles 21 and 22. Our pred
iction agrees reasonably well with the "climatology-based" forecast made by
the intergovernmental panel tasked to predict geomagnetic activity far the
coming solar cycle and is significantly different from their "precursor-ba
sed" prediction.