Forecasting diffusion model considering eompetitive dynamics

Abstract

The purpose of this paper is to suggest a forecasting model for service subscribers which can reflect competitive circumstances. Our model incorporates the competitive dynamics which reveals that; (1) when a new rival enters the market, an impact to the demand occurs and it accelerate the diffusion speed, and (2) the demand for the service is affected by the number of subscribers for the rival’s service. In other words, we developed a forecasting diffusion model, naming the competitive impact diffusion model under the following four hypotheses; (1) the number of subscribers for each company is affected by the other firm``s subscribers; (2) the impact of entry by new competitors accelerates the diffusion rates of existing and new services; (3) The competitive impact includes the effects of all marketing mix variables; and (4) The competitive impact is slowing down with respect to time. We empirically verified the model by using out-of-sample estimation, following comparison with other models (e.g. Logistic model, Lotka-Volterra model, Bass model, Mahajan, Sharma, and Buzzell model, and Parker and Gatignon). The performances for the forecasts were measured through MAD (Mean Absolute Deviation), MAPE (Mean Absolute Percent Error) and the encompassing test which have been widely used for the assessment for forecasting models. In this setting, we used existing cellular and PCS cumulative subscribers`` monthly data from January 1998 to December 2005 to show the model``s predictability. In the first test, we investigated the competitive impact by using the data from January 1998 to December 2000. The last six observations were used for an out-of-sample. In the second test, we examined the model``s fitting ability for the longer horizon from January 1998 to October 2005. In comparison with the fundamental models and the other Bass-type models, by analyzing MAD and MAPE from one-step- to six-step-ahead forecast, our model shows superior fitting and forecasting...