Interregional and International Telephone Communication. Aggregate Traffic Model and Empirical Evidence for Austria

The explosion of activities and requirements associated with the production, processing and transfer of information is increasingly being matched by a profileration and diversification of new telecommunication media for transmitting information, including text processing and transmission services such as facsimile transmission, videotex, teleconference services, electronic mail etc. Nevertheless, the telephone is still - by far - the most important telecommunication service. The paper results from ESF-research undertaken within the Network on European Communication and Transport Activity Research (NECTAR) and relates to telephone communication undertaken for the Netherlands by Rietveld and Jansen (1990) and Switzerland by Rossera (1990). The current study focuses on the Austrian case and relies on data measured by the Austrian PTT in 1991, in terms of erlangs, an internationally widely used and reliable measure of telecommunication contact intensity. The data refer to the total telecommunication traffic on the public network. Due to technical reasons oral communication can not be distinguished from other services such as data transmission, transfer of documents and text (facsimile) etc. But the demand for such new telecommunication services is still at a very modest level in Austria. The paper addresses two major issues. First an attempt is made to explore the factors influencing the spatial pattern of domestic telephone traffic in Austria. The econometric approach applied to this problem belongs to the class of spatial interaction models explaining a telephone communication flow from a region i to a region j by three types of factors, factors associated with the region of origin, factors associated with the region of destination and factors associated with origin-destination pairs (separation factors). In using the spatial interaction modelling approach in telephone traffic analysis various choices need to be made about how the above mentioned factors should be defined. To have confidence in the model results it is desirable that its interpretation is insensitive to the particular choices made. Whether this is so, will be investigated for several potential sources of variation in model performance.