Welfare effects of quality change and new products in the Japanese mobile telecommunications market: 1995–2001

Abstract

This paper presents a simple framework based on a discrete choice model to assess the welfare effects of quality change and new products. Such a framework is shown to be useful where the hedonic approach is impracticable. This framework is applied to the Japanese mobile telecommunications market during the period 1995–2001. The estimated qualities for mobile telecommunications services in Japan are shown to have improved rapidly during this period, indicating the importance of quality change in the welfare gains of consumers relative to price change, and the need to adjust quality change in the construction of price indexes. However, it is also proven, at least in the earliest stages, that the effect of new services on consumer welfare was limited.

Keywords:

• Quality change
• Discrete choice model
• Cost of living index
• Telecommunications

Acknowledgements

I express my sincere gratitude to the editor, Professor Cristiano Antonelli, and two anonymous referees, for their valuable comments. I am also grateful to Professor Takanobu Nakajima, Professor Makoto Ohta, Mr Nobuyasu Atago, Ms Kuniko Moriya, Mr Shigenori Shiratsuka, Mr Kiyohito Utsunomiya, Mr Masato Noguchi and the other attendees at the 2003 Spring Meeting of the Japanese Economic Association and participants at a Bank of Japan seminar on 15 July 2003 for their helpful comments. Financial support from the 21st Century COE Program at Keio University (the Grant-in-Aid from the Ministry of Education, Culture, Sports, Science and Technology, Japan: I-1) is gratefully acknowledged. I am, of course, responsible for errors, if any, in this paper.

Notes

¹PHS services made their debut in 1995. They provide voice and data transmission services that are similar to cellular phone services. At first, the quality of voice transmissions of PHS services was better and the average data transmission speed of PHS services was faster than cellular phone services. However, PHS services sometimes cannot send and receive transmissions in high-speed moving vehicles. For more details, see Okada and Hatta (1999), Moriya and Kunihiro (2001) and each edition of ‘Joho Tsushin Handobukky (Information & Communications in Japan).’

²All carriers discontinued analog services in March 1999.

³Of course, some of them charged additional costs.

⁴The first of mobile Internet services in Japan is ‘i-mode’ service, which was introduced in February 1999 by the NTT DoCoMo group. During the year 1999, ‘ezweb’ service provided by the KDDI group and ‘J-Sky’ service by J-PHONE also entered this market.

⁵For the details of regression, please see the footnote of Table I.

⁶While Bresnahan’s study provides an assessment of market conduct and its consequences in the American automobile industry, Greenstein’s study attempts to measure the economic benefits that accrued to buyers from technological innovation in the American mainframe computer market.

⁷Feenstra (1995) uses the framework to construct exact hedonic price indexes for the case of linear random utility model, CES model and general utility function model.

⁸This assumption regarding the consumers’ decision-making process can appear strong and restrictive. However, it is consistent with the definition of the Japanese official price index. The implications of this in regard to the price index are discussed later.

⁹Although this assumption makes it easier to estimate quality change and construct the COL index, its robustness should be checked. The Appendix provides such a chock in the form of a certain skewed distribution, which includes the uniform distribution as a special case.

¹⁰The NTT DoCoMo group is chosen as it is the most dominant entity in this market and other carriers’ service programs are typically similar.

¹¹For estimation of these demand systems see Berry (1994) and Nevo (2000).

¹²In this study, the potential market size approximates the population over 15 years old. In an aging society with a low birth rate such as Japan, this assumption produces little estimation bias.

¹³To decompose compensating variation, one has to calculate the hypothetical marginal consumer with p ⁰ and q ¹. However, the calculated marginal consumers are not necessarily to stand in the order of their qualities and those of the highest quality network are greater than unity. The author manages such cases as follows. If those of the highest quality network would be over one, the marginal consumer set one, so that its market share is zero. If b _j would exceeded b _j+1, it is assumed that the network j with p ⁰ and q ¹ could not hold any market share. Then, the marginal consumer between j−1 and j+1 is re-computed.

¹⁴One has to pay attention to total expenditure, which is given as total hypothetical expenditure in this particular study. It is assumed that all subscribers use their network, for only 3 min and its movement may be different from that of the actual expenditure. The movement of total hypothetical expenditure was compared with that of carriers’ revenues, constructing respective indexes (1995=100). They show a similar pattern. These two series were depicted in Figure 10 of Sunada (2004).

¹⁵There is a large amount of existing research concerning the consumer gains resulting from the introduction of specific products. For example, Trajtenberg (1989) concerning medical imaging devices, Hausman (1997a) on breakfast cereals, Hausman (1997b, 1999) about cellular phone services, Hausman and Leonard (2002) on bath tissues and Petrin (2002) in regards to minivans.

¹⁶Such services are also available through PHS phones, but this possibility is not considered in the current study.

¹⁷The number of subscribers to mobile Internet services is that in December of each year.

¹⁸For more details, refer to Ugai (2001). Nevo (2003) attempted to construct a price index based on an estimated brand-level demand system for US ready-to-eat cereal market, taking account of the introduction of new goods and quality change.

¹⁹Nakajima (2002) defines r>−1. However, in the case with r<0, one cannot define the distribution function f (·) at v=1.

Additional information

Notes on contributors

Mitsuru Sunada

Private address for offprints: 2-7-21, Higashi-Himemiya, Miyashiro, Saitama 345-0813, Japan; E-mail: [email protected]