Forecasting in the Rapidly Changing Telecommunications Industry: AT&T Experience

Article excerpt

In the changing market dynamic of telecommunications industry, the traditional econometric models are inadequate for preparing forecasts ... the best way is to prepare a base line forecasts with traditional models and then overlay best judgments for events not embedded in the model... it is becoming more and more difficult to get high quality data in this industry.

The telecommunications industry is facing continuos technological and regulatory changes. While the distinctions between local and long distance companies have become less clear as each vies for the other's business, new competitors from the cable, electric and gas industries have emerged which have blurred industry definitions that were once well defined. Even jurisdictional differences are being removed as leading US carriers are forming joint ventures with foreign companies to enter markets that were previously inaccessible. As the industry becomes more competitive, consumers have benefited through lower prices, which have stimulated telecommunications demand to unprecedented levels.

These changes have necessitated the development of new forecasting techniques. Traditional forecasting methods such as time series and econometric modeling have become less accurate because the industry no longer has the stable historical relationships that these models rely upon. The forecaster therefore needs to incorporate the industry dynamics into the model. This article presents an approach to estimate the demand for long distance telephony by combining standard econometric techniques with market realities to capture the dynamics of the long distance telecommunications industry.

LIMITATIONS OF CONVENTIONAL MODELS

Econometric and time series modeling are the most popular methods for forecasting telecommunications demand. Often forecasters combine the two approaches by incorporating into the model both explanatory variables and past growth rates of the dependent variable. The most common explanatory variables used are price and income, which attempt to reveal the nature of demand for a product. It may show that the quantity demanded for telecommunications industry tends to increase when real prices decrease and real income increases. In addition to demand, it estimates price and income elasticities that have implications for further pricing policies. The inclusion of past growth rates into the model can reveal trend and seasonal patterns. Of course, the forecaster needs to check for multicollinearity which often appears when both income variable and trend term are included in the same model.

These conventional modeling approaches rely upon a stable historical relationship between the independent and dependent variables, as well as trend and seasonal patterns that are not expected to change dramatically. However, the realities of the telecommunications industry do not meet these requirements. To give just one example, the technological advances that made wireless telephony affordable to many people have at the same time cannibalized card services. There are some estimates which show that as much as 50% of the card market migrated to wireless telephony in 1999. This cannibalization would not show up in any traditional modeling approach, and the forecaster would seriously overstate the demand for card services.

Another factor limiting the usefulness of conventional econometric modeling is the lack of reliable and consistent data. While the FCC accumulates and makes available to the public telecommunications data free of charge, this data is often not of the highest quality. In most cases, the FCC relies solely on the cooperation of the reporting company and has no authority to audit the results. This often leads to data that is inconsistent from period to period, incomplete, and available only after a long time lag. Earnings release information filed with the SEC is timely and of very high quality but limited to publicly held companies. …