Forecasting the Diffusion of Innovations: Implications for Time-Series Extrapolation

Abstract

The selection of an S-shaped trend model is a common step in attempts to model and forecast the diffusion of innovations. From the innovationdiffusion literature on model selection, forecasting, and the uncertainties associated with forecasts, we derive four principles. 1. No single diffusion model is best for all processes. 2. Unconditional forecasts based on a data-based estimate of a fixed saturation level form a difficult benchmark to beat. 3. Simpler diffusion models tend to forecast better than more complex ones. 4. Short-term forecasts are good indicators of the appropriateness of diffusion models. We describe the evidence for each principle in the literature and discuss the implications for practitioners and researchers.