The variability of the Sun's luminosity is thought to have effects on the world climate. This variability is partly measured by the sunspot numbers. In this paper the statistical link between sunspot numbers and Addis Ababa rainfall is established. For this purpose a transfer function plus noise model (TFN) is constructed, with Box-Cox transformed annual sunspot numbers, X,, as input and annual Addis Ababa monsoon rainfall, x, as output. This model explains 50.2 per cent of the annual rainfall variability in the period . The time series of the sunspot numbers is simulated by an autoregressive model and then used to provide the input forecasts to the TFN model; it explains 88.9 per cent of the annual sunspot numbers variability.The TFN model is utilized for long-range annual rainfall forecasts. A check of the forecasting skill of the TFN model is performed by fitting the TFN model to the 1900-1970 data and forecasts are made for 1971-1991. The results show that 14 out of 21 forecasts generally track the observed rainfall. The correlation coefficient C between the forecast and observed rainfall is 0-58, whereas for the sunspot numbers C yields 087. The paired t-test shows that the forecasts and the observed rainfall are statistically similar at the 95 per cent confidence level.A general physical mechanism that may partly explain how the variability of the Sun's luminosity is related to the world climate is suggested. It is proposed that for those regions whose climatic variables exhibit the 11-year periodicity, the TFN model approach can be used to find a statistical link between the sunspot numbers and the climatic variables. Furthermore if a global pattern in such a link is found, it may be used for exploring the physical mechanisms of Sun-climate relationships.KEY WORDS Addis Ababa rainfall Sunspot numbers AR ARMA Multiplicative ARMA Transfer function plus noise analysis