Demand forecasts are an essential tool for planning capacity andformulating policy. Traffic estimates are becoming increasinglyunreliable, however, as accelerating rates of use and newcommunications applications invalidate conventional forecastingassumptions.This paper presents an alternative approach to the study oftelecommunications demand: build aggregate estimates for demandbased on the elasticity of demand for bandwidth.We argue that price elasticity models are necessary to grasp theinteraction between Moore-type technological progress and non-lineardemand functions.Traditional marketing models are premised on existing or, at best,foreseeable services. But in a period of sustained price declines,applications-based forecasts will be unreliable. Dramatically lowerprices can cause fundamental changes in the mix of applications and,hence, the nature of demand.We consider the option of posing demand theoretically in terms ofservice attributes. Our conclusion is that the positive feedback loop oftechnology-driven price decreases and high-elasticity demand willquickly make it possible to base forecasts on bandwidth elasticityalone.Industry analysts and policymakers need models of consumer demandapplicable under dynamic conditions. We conclude by drawingimplications of our demand model for network planning, universalservice policies, and the commoditization of communications carriage.
