No CrossRef data available.
Published online by Cambridge University Press: 01 September 2004
It has now been 30 years since Jeffrey Rohlfs developed the concept of network externalities to explore how demand for a telecommunications service depends on the number of participants in the relevant network—such a service is more valuable to any user if there is a larger number of other users (J. Rohlfs. “A Theory of Interdependent Demand for a Communications Service.” Bell Journal of Economics and Management Science 5, no. 1 [1974]: 16–37). Starting about ten years later, economists generally came to appreciate how network externalities give rise both to bandwagon effects in the adoption of specific techniques and to persisting lock-in of these techniques. Economists also noted how this analysis applies not only to networks based on a physical infrastructure but also to virtual networks, for which network effects arise because a larger number of users of some product (e.g., the Microsoft Windows operating system for personal computers, in contrast to Linux or the Apple Macintosh operating system) induce the greater provision of complementary goods or services (e.g., application software), which make the product more attractive. Some of the resulting literature has been synthesized in a three-article symposium in the Journal of Economic Perspectives (Spring 1994) and in several books, perhaps most notably Carl Shapiro and Hal R. Varians Information Rules: A Strategic Guide to the Network Economy (Boston: Harvard Business School Press, 1999). Citations to Rohlfs's seminal article have rebounded more than once as a result.
Rohlfs's new contribution to this literature breaks little new ground, but it offers an excellent, accessible introduction to the concepts, basic analysis, and empirical studies related to network externalities. Two chapters introducing concepts are followed by three more that delve into demand- and supply-side issues, using a few simple diagrams and no equations. The bulk of the book is then taken up by case studies of fax technology, voice telephony before about 1917, the abortive picturephone of the early 1970s, compact-disc players, videocassette recorders (VCRs), personal computers, television from 1945 to the present, and the internet. Two concluding chapters summarize the book's themes, and a mathematical appendix uses equations and diagrams to offer more rigorous, though still relatively simple and accessible, results.
Rohlfs tells good stories in a straightforward way, without a heavy analytical apparatus but with considerable wit and a touch of irony. His case studies are nonetheless thoroughly informed by economic analysis. Moreover, he is attentive to institutional and historical context, so that the reader is given a good understanding of how the analysis applies in the real world. The three chapters on telecommunications technologies display the author's expertise as a longtime economic researcher for Bell Labs. The chapters on recent developments in other facets of information technology show that he has been a keen, well-informed observer from the sidelines.
Rohlfs's case studies focus on his theme of bandwagon effects—the take-off in demand that develops in response to a larger user base for a product or service. He faults Alexander Graham Bell for greedily setting early telephone service prices too high to develop a bandwagon effect, leaving Bell companies unnecessarily vulnerable to competition when the Bell patents expired (p. 80). Similarly, he faults AT&T for introducing picturephone service without taking account of the start-up problem—how to reach a critical mass of customers, so that prospective adopters would have someone to talk to. Repeating the analysis of others, Rohlfs notes that VCRs reached critical mass as a result of their stand-alone value before the emergence of a videocassette rental market gave rise to a bandwagon effect, and he considers how this bandwagon effect gave the competition between JVC's VHS format and Sony's Beta format a winner-take-all character. Also in line with previous studies, Rohlfs credits Microsoft, and Bill Gates in particular, with a much more skillful exploitation of bandwagon effects in personal computers than either IBM or Apple Computer.
The book has much to offer to economic historians, to applied economists generally, to economics students at all levels, and indeed to anyone interested in how economic analysis illuminates the development of markets in information technology. A mild shortcoming is that Rohlfs does not explicitly address the implications of his analysis for path dependence, that is, for how history has a continuing impact on economic choices and outcomes.
Gottinger's Economies of Network Industries introduces a broader range of analytical concepts than Rohlfs's book and expresses them somewhat more in the language of mathematics; but Gottinger does little to explore the application of these concepts to concrete empirical cases. Although Gottinger often introduces concepts by referring to empirical cases, the reader is not left with a rounded picture of any of these cases. Furthermore, Gottinger's explanations are sometimes incomplete or ambiguous, so that his presentation is not readily accessible to nonspecialists.
The book does, however, offer analytical ideas and methods from near the cutting edge of recent research, including some of Gottinger's own. Thus the book will interest readers who already have a good grasp both of basic concepts and of empirical cases but who would like to consider how the analysis might be extended. Gottinger does not, however, claim to offer a general survey of current research topics, and he does not make clear how he has chosen which topics to include.
One notable omission is any discussion of how technology adoption is affected by the purposeful internalization of externalities, a theme developed extensively by Stan J. Liebowitz and Stephen E. Margolis (collected articles: Peter Lewin, ed., The Economics of QWERTY: History, Theory and Policy, New York: New York University Press, 2002). To be sure, the Liebowitz-Margolis analysis is marred by their lack of appreciation of limits to internalization of externalities, as I have argued elsewhere, but any advanced treatment of network externalities should address their argument. By contrast, Rohlfs addresses the issue of internalization explicitly and helpfully at several points, and he does so without the problematic baggage introduced by Liebowitz and Margolis. (Rohlfs would have done well, however, to cite the work of Liebowitz and Margolis in this connection.)
A further comparison of the two books arises in their treatment of Metcalfe's Law. Robert Metcalfe, the inventor of the Ethernet computer networking technology, has suggested that the value of a network for each of N users is proportional to N1, the number of other users, so that the total value of a network for all users is proportional to N(N1) or, approximately, N2. Both Rohlfs and Gottinger note that Metcalfe's formula must be modified if, as N increases, the value of each connection to the marginal user declines. Rohlfs's mathematical appendix offers straightforward numerical examples as well as a rigorous calculus treatment of this point. Gottinger's discussion of the same point amounts to impressionistic hand-waving leading to ambiguous diagrams and then an econometric test. Most distressingly, Gottinger's discussion confuses total network benefits with benefits per user. For example, he asserts that the influential model of W. Brian Arthur (“Competing Technologies, Increasing Returns, and Lock-in by Historical Events.” Economic Journal 99 [1989]: 116–31) assumes a linear relationship between network size and network value. In fact Arthur, just like Metcalfe, assumes a linear relationship between network size and benefits for each user, so that total network value (for all users together) varies exponentially with network size. Gottinger also errs in his econometric test of total network value by expressing data in per-capita terms.
I did not find any further outright errors in Gottinger's analysis, but his analysis at other points does sometimes suffer from a lack of clarity. This is true not only of the text but also of the figures. Axes and their units are often unlabeled, incompletely labeled, or wrongly labeled, as when a figure of a learning curve in semiconductor wafer fabrication uses time of production (rather than cumulative production quantity) on the horizontal axis and wafers feed [sic] into systems (rather than proportion of good wafers among those fabricated) on the vertical axis (Figure 1.5, p. 6). In short, Gottinger's book is stimulating for those with some background in the topic but spotty in coverage and not fully reliable.