Auroral ocurrence frequency may be used as proxy data for solar activity. In recent years magnetic data and auroral data during the declining phase of one sunspot cycle have been used to predict the maximum sunspot number of the following cycle. This paper discusses appropriate use of auroral data for this purpose, and tests the validity of such a procedure for the period 1721-1943. The results show that the use of auroral sums for sunspot minimum an the preceding three years can predict the subsequent sunspot number maximum, but that the predictor involves two branches. One, fitting the majority of observations, may be considered the normal branch.The other, called here the deficit branch, gives fewer auroral occurrences than would be expected. Treatment of magnetic activity data in the same fashion gives similar, but not identical results.