Online Program

Friday, February 22
PS2 Poster Session 2 (with refreshments) Fri, Feb 22, 4:45 PM - 6:15 PM
Napoleon Ballroom

Modelling Seasonality in Innovation Diffusion

View Presentation View Presentation *Mariangela Guidolin, Department of Statistical Sciences, University of Padova, Italy 
Renato Guseo, Department of Statistical Sciences, University of Padova, Italy 

Keywords: Seasonality, Time Series Decomposition, Guseo-Guidolin model, Generalized Bass model, Nonlinear Least Squares

The ability to forecast new product growth is especially important for innovative firms that compete in the marketplace. Innovation diffusion research aims to describe the evolution of a new product in the market by modeling its life-cycle. Today many new products exhibit a very strong seasonal behavior, which may deserve a specific modeling, both for producing better forecasts in the short term and for better explaining special market dynamics. By considering seasonality as a deterministic component to be estimated jointly with the trend through Nonlinear Least Squares methods, we develop two extensions of the Guseo and Guidolin model (2009), that are able to describe simultaneously trend and seasonality. Such models are based on two different but equally reasonable approaches : in one case we consider a simple additive decomposition of a time series and design a model where seasonality is directly added to the trend and jointly estimated with it; in the other we design a more complex structure, miming that of a Generalized Bass model and embed a seasonal perturbation within the evolutionary trend, so that seasonality directly operates on it. Testing the models to two applied cases, the life-cycle of a pharmaceutical drug and the diffusion of a technological device produced by Apple, we find that the more complex modeling option is more suitable to the case of the pharmaceutical drug, while the additive solution performs better in the other case. Both models are quite simple to implement and to interpret from a business point of view.