In this work a model for estimation of the switching activity in ripple-carry adders is presented. The model is based on word-level statistics, such as mean, variance, and correlation, of the two input signals to be added. It is shown that the proposed model gives accurate results when the two-s-complement represented inputs are real world signals.