Entry and Competition in the Canadian Telecommunications Industry: The Case of Alberta Government Telephones
Gentzoglanis, Anastassios, Atlantic Economic Journal
In the last few years, the telecommunications industry has undergone tremendous changes throughout the world. Important technological changes have modified and greatly transformed the structure of the telecommunications market and called into question the traditional role of price and entry regulation. Leading the world in the deregulatory movement is the U.S. Public policymakers in the U.S. have determined that regulation can no longer be used as a substitute for competition and the latter should be introduced into both the transmission and customer premises equipment (CPE) markets.
The liberalization of the CPE market has been pursued on the almost unanimous belief that the introduction of competition would not sacrifice either economies of scale or economies of scope. By contrast, no unanimity exists concerning the deregulation of the telecommunications long distance transmission and switching markets. The econometric studies carried out in Canada provide little information on the important question of cost subadditivity, the qualitative property of the cost function that determines whether an industry is effectively a natural monopoly.
This paper examines the question of cost subadditivity by estimating multi-product, multi-input translog cost functions for a government-owned company and a relatively small one, AGT. The "Evans and Heckman" (E&H) test of natural monopoly is applied to determine whether AGT's cost structure is subadditive. This is the first attempt to examine the cost structure of AGT empirically and apply the E&H test to this firm.
Section II briefly describes the model and data used to estimate the cost function of AGT. Section III reports the estimations and subadditivity tests. Section IV concludes with some policy recommendations.
II. Conceptual Framework, Data, and Variables
A) Conceptual Framework
AGT produces a variety of services using factors of production grouped under the headings of capital (K), labor (L), and materials (M). For purposes of the empirical investigation, the outputs can also be grouped into local (|Q.sub.1~) and long-distance (|Q.sub.2~) service outputs. Generally, AGT's cost function can be written as:
C = f(|Q.sub.1~,|Q.sub.2~,|P.sub.K~,|P.sub.L~,|P.sub.M~,T), (1)
where |P.sub.K~ is the capital rental rate, |P.sub.L~ is the wage rate, |P.sub.M~ is the price for materials, and T is an index of technological change.
According to Christensen, Jorgenson, and Lau |1973~, a multi-product translog cost function provides a useful second-order Taylor series approximation to any twice differential cost function. Since their introduction, the translog cost functions have been frequently used in empirical studies. Their main attractiveness is that they are easy to estimate and impose fewer restrictions on the characteristics of the production structure than other commonly used functional specifications. Moreover, it is possible to obtain directly from the cost function, by simple differentiation, factor demands (Shepherd's lemma).
AGT's cost function can be approximated by the following second-order translog functional form:
|Mathematical Expression Omitted~
where i, m = |Q.sub.1~, |Q.sub.2~ and j, n = K, L, M. The cost share equations are:
|S.sub.j~ = ||Alpha~.sub.j~ + |summation of~ ||Gamma~.sub.nj~ ln (|P.sub.j~) where j = 1 to 3 + |summation of~ ||Delta~.sub.ij~ ln |Q.sub.i~ + ||Mu~.sub.jt~ ln T. (3)
Appending normally distributed error terms to the translog cost function (2) and to each of the cost share equations (3) allows one to estimate this multivariate regression system by using Zellner's |1962~ non-linear iterative seemingly unrelated regression (ITSUR) method. Given that the covariance matrix of full-system disturbances is singular (the error terms are not independent since the cost shares sum to unity), a cost share equation (the material share equation) is dropped. …