Coverage area expansion, customer switching, and household profiles in the Japanese broadband access market

Abstract

This paper examines the diffusion process of Internet broadband access in Japan by modeling the household choice of access modes. Estimates reveal that the characteristics of users, rather than those of the access modes, play a significant role in demand substitution across the modes. Simulation exercises indicate that had optic fibers (fiber-to-the-home; FTTH) been made available to the whole country in 2005, only 10% of households would have switched to that mode. This result implies that once chosen, the household choice of access mode persists, and that indirect network effects are an important element in decisions to adopt the new technology of broadband. Finally, policy implications are discussed.

Introduction

Rapid expansion in the coverage of broadband Internet access worldwide has been noteworthy in the early 21st century. Although the dominant mode of access varies from one country to another, fiber-optic access is a promising mode of transmission in wired and wireless communications. According to OECD (2006), Korea and Japan were prominent among OECD countries in terms of the penetration of optic fibers, as of December 2005.

This paper examines the diffusion process of Internet broadband access in Japan by modeling household choice of broadband access modes, the choice set of which includes dial-up, integrated services digital network (ISDN), cable television systems (CATV), asymmetric digital subscriber lines (ADSL), and optic fibers (fiber-to-the-home; FTTH). Furthermore, we simulate the extent to which households switch between these modes when FTTH extends its coverage to households that were previously unable to access the technology. In order to investigate the most important factors affecting the adoption of a new technology, like FTTH, we then conduct simulations for subsamples divided by various household characteristics.

There are several studies estimating Internet access demand. Eisner and Waldon (2001) estimated a binomial probit model of US demand for Internet access. Rappoport et al., 2003, Cerno and Amaral, 2006, Goldfarb and Prince, 2008 estimated demand for both access and usage simultaneously. Madden and Simpson (1997) used a simple logit model to estimate Australian residential broadband access demand using stated-preference data. Crandall et al. (2002) estimated the price elasticities between DSL and CATV using a nested logit model. Regarding Japanese Internet access, Tanaka et al., 2004, Ida and Kuroda, 2006 estimated broadband access demand using nested logit models, and found that FTTH, ADSL, and CATV were close substitutes for each other. These previous studies were based on the strong assumption of restrictive substitution patterns among different access modes. In a recent study, Pereira and Ribeiro (2006) estimated price elasticities and marginal costs of broadband access in Portugal using a mixed logit model that allows flexible substitution patterns to evaluate the structural separation of Portugal Telcom.

According to Gaynor and Vogt (2003), however, a conventional fixed-coefficient linear random-utility model with rich microdata can be an effective substitute for a random-coefficient logit model in order to capture rich substitution patterns. Therefore since we have detailed information about household profiles and coverage areas of municipalities, we adopt the fixed-coefficient linear random-utility model promulgated by Gaynor and Vogt (2003).

Estimates from random-utility models of access demand reveal that the characteristics of users, rather than those of the access modes, play a significant role in demand substitution across the modes. Simulation exercises indicate that even if FTTH had been made available to the whole country in 2005, only about 11% of households would have switched to that mode. Moreover, heavy Internet users (14%) are more likely to switch to FTTH than nonheavy users (7%), and households with a student (18%) are more likely to change to FTTH than households without a student (9%).

These results imply that inertia in existing access modes is considerable, and the usage intensity of the Internet is an important factor determining an access mode. That is, demand-side conditions of broadband access could change drastically because of technological development and evolution of complementary markets that make the Internet more attractive and enhance indirect network effects. Therefore, government policy should focus not only on the expansion of the FTTH coverage area, but also on the enhancement of the quality and variety of software associated with FTTH.

The remainder of the paper is organized as follows. The subsequent section briefly reviews the Japanese broadband access market. Section 3 describes the data, and Section 4 explains the empirical specifications, variable definitions, and estimation results. Section 5 conducts simulations regarding the expansion of FTTH coverage. Section 6 concludes the paper.