In this paper we present a general framework for forecasting the smoothed maximum level of solar activity in a given cycle, based on a simple understanding of the solar dynamo. This type of forecasting requires knowledge of the Sun's polar magnetic field strength at the preceeding activity minimum. Because direct measurements of this quantity are difficult to obtain, we evaluate the quality of a number of proxy indicators already used by other authors which are physically related to the Sun's polar field. We subject these indicators to a rigorous statistical analysis, and specify in detail the analysis technique for each indicator in order to simplify and systematize reanalysis for future use. We find that several of these proxies are in fact poorly correlated or uncorrelated with solar activity, and thus are of little value for predicting activity maxima. We also present a scheme in which the predictions of the individual proxies are combined via an appropriately weighted mean to produce a compound prediction. We then apply the scheme to the current cycle 22, and estimate a maximum smoothed International sunspot number of 171 ± 26, which can be expressed alternatively as a smoothed 2800 MHz radio flux (F 10.7) of 211 ± 23 × (10-22 Wm-2Hz-1), or as a smoothed sunspot area of 2660 ± 430 millionths of a solar disk. Once the actual maximum for cycle 22 has been established, we will have both additional statistics for all the proxy indicators, and a clearer indication of how accurately the present scheme can predict solar activity levels.