On the combination of exponential smoothing and diffusion forecasts: An application to broadband diffusion in the OECD area

The innovation diffusion literature has established that the spread of a successful innovation over time typically follows a sigmoid curve. Therefore, the forecasting in this area has been monopolized by the use of well known aggregate diffusion models. Time series forecasting has been surprisingly neglected, as it provides mainly accurate short term forecasts. In this work, a method of exponential smoothing, the Holt's damped trend with a modification, is applied in recent broadband diffusion data of two large regions after the reach of the inflection point. As validated with holdback sample data ranging from 6 up to 30 months, the key for successful forecasting is the use of the estimated saturation level calculated from a diffusion model, in order to specify the appropriate trend. The results indicate improved predictions compared to two popular diffusion models, the Gompertz and the Linear Logistic model. The paper concludes with the application of the proposed method in a 48-month forecasting horizon, as well as the suggestions for further research.