Conjoint analysis of demand for IP telephony: the case of Japan

Abstract

With the recent diffusion of broadband (BB) services, Internet protocol (IP) telephony is expected to spread significantly in Japan. This article investigates the demand for IP telephony by using conjoint analysis. Projecting IP telephony demand also contributes to Japanese info-communication policy discussions. Two points are made. First, IP telephony is still currently considered an optional supplement or an add-on service option of high-speed BB Internet access services in Japan rather than a close substitute of existing plain old telephone service (POTS). At this point, we find little evidence that many households will promptly forsake their fixed line service for IP telephony. Second, we conclude that the key condition for the proliferation of IP telephony is the complete guarantee of quality of service (QoS), including voice quality, number portability, fax usage and emergency access, comparable to or exceeding that of existing POTS.

Notes

¹ In addition, since interconnections among voice over IP (VoIP) circuits are not complete, 050-type IP telephony may not be used if different Internet service providers (ISPs) are involved.

² For previous research using conjoint analysis, e.g. see Madden and Simpson (%1997), San Miguel et al. (%2000), Bryan and Parry (%2002), Kim et al. (%2005) and Roe et al. (%2005).

³ Field pioneers McKnight and Leide (%1998) have analysed the costs, pricing and policy problems of Internet telephony, arguing for a comprehensive regulatory framework concerning the convergence of fixed telephone and Internet telephony. Ono and Aoki (%1998) have also compared how policy makers and regulators in the US, the EU, Japan and Singapore responded to the emergence of IP telephony; they insist that existing regulatory frameworks will not be aligned within dynamically changing environments. Jaiswal and Raghav (%2004) have suggested that IP telephony is a potential key enabler for service delivery in developing countries.

⁴ A sample survey by MIC on 26 January 2005 entitled ‘Study on the Development of the Telecommunications Service Demand in Japan’ also reports the importance for ISPs to guarantee QoS for further diffusion of IP telephony, since consumers do not want to move to IP telephony until they can use IP telephony for emergencies.

⁵ For references, Ida and Kuroda (%2005) investigated the Japanese demand for broadband Internet connection using RPM while Madden and Simpson (%1997), Ida and Sato (%2007) and Savage and Waldman (%2005) studied broadband demands using SPM.

⁶ Note that Fields and Kumar (%2002) indicated a potential defect of SPM (including conjoint analysis): it only models current consumer attitudes for market forecasting; their current intentions will become accurate reflections of future choices if the current environment represents a future in which the choices are to be realized. We believe that this condition is satisfied in the example of the Japanese IP telephony service.

⁷ Miller (%1956) points out that it is too complex for human beings to simultaneously process more than six attributes.

⁸ Here fixed telephone is defined to cover the fixed telephone services supplied by NTT and other companies renting NTT's facilities, or constructing their own telephone networks, as well as NTT's subscriber telephone services. On the other hand, IP telephony indicates phone-to-phone VoIP service for end users who have high-speed access to the Internet or dedicated VoIP networks. In this article, we do not deal with PC-to-PC, phone to PC, or PC to phone VoIPs.

⁹ If a user opts to preselect a specific carrier called MYLINE, calling charges are decreased according to a discounted fee schedule.

¹⁰ Since we conducted the survey online, the data may have a bias because the monitors who participated may be more interested in the Internet than average people. We believe, however, that such sampling bias is allowable because the purpose of this article is to study the future usage of such new technology as the Internet.

¹¹ We should have included interaction effects. However, Dawes and Corrigan (%1974) suggest that aliasing interaction effects with linear effects are justified because linear effects typically account for 70% to 90% of explained variance, while two-way interaction effects account for 5% to 15%, and high-order interaction effects account for the remaining explained variance.

¹² Having assumed ϵ as the IIDEV I, the CL model possesses ‘independence from irrelevant alternatives (IIA)’. This IIA property denotes that the ratio of choice probabilities is only determined by the two relevant alternatives and is completely independent of the other alternatives.

¹³ Here the scale parameter is normalized to one.

¹⁴ Due to space constraints, we avoid detailed analysis of each Internet connection preference (ADSL, CATV Internet, FTTH and narrowband). However, note that choice ratios for Alternatives 1, 2 and 3 differ depending on the type of Internet connection. For example, the IP telephony ratio is higher among broadband than narrowband users. This suggests that the type of Internet connection influences consumer preferences for the adoption of IP telephony. We will investigate whether parameter estimates (namely, consumer preferences) are equal among user groups based on their Internet connection type by using a likelihood ratio (LR) test.

¹⁵ Since conjoint analysis is essentially based on substitutability between attributes, it cannot deal with noncompensatory decision making, which results in biased WTP estimation. Therefore, it might be wise to exclude such noncompensatory responses from the data for estimation.

¹⁶ We should carry out the Hausman test and investigate whether the independence from IIA assumption holds, since we adopt a CL model. Accordingly, we carried out the Hausman test, excluding Alternative 2, for the six groups (Hausman and McFadden, %1984). Consequently, the IIA assumption is rejected in some cases, but not rejected in others. Concretely, the results for (d), the ADSL user group, are 6.76 and, therefore, the IIA assumption is not rejected, given that the critical value is χ² (p = 0.05, d.f = 6) = 12.59. On the other hand, the results for (f), the NB user group, are 166.91 and, therefore, the IIA assumption is rejected. In this article, since we tried complicated simulation analysis for predicting IP telephony penetration based on the estimation results, we adopt a simple CL model. However, we understand that the IIA assumption is rather strict, and we should try more general models including a nested logit (NL) model, a mixed logit (ML) model, or a probit model to allow for the variety of individual preferences or flexible demand substitution patterns. We consider this problem a subject for future research.

¹⁷ Socio-demographic differences may lie behind the differences in the preferences of BB and NB users. Therefore, we investigate such demographic characteristics of both groups as age, gender, living in the three largest metropolitan areas (namely, Tokyo, Osaka and Nagoya), education levels, employment history and household income. Consequently, we see that only the ratio of living in the three largest metropolitan areas (Tokyo, Osaka and Nagoya) differs between BB and NB users, whereas the other characteristics are almost identical. It is more likely that BB users live in metropolitan areas and that NB users live in rural areas.

¹⁸ See Shin et al. (%2004) who theoretically demonstrate that guaranteeing QoS could become a progressive equilibrium strategy in the future Internet market, which includes IP telephony or video conferences.

¹⁹ In a simulation analysis it may be natural to assume that a service of low quality is inexpensive while one of high quality is expensive. However, we set the same price for the 050-IP and 0ABJ-IP phones because we wanted to separate the effects of functional improvements on choice probabilities from the effects of price differences. Note here that the estimated penetration rate of 0ABJ-IP phone may be rather overestimated.

²⁰ Examples in line with this include Islam and Meads (%1996) and Ahn and Lee (%1999).

²¹ In November 2004, NTT announced a ‘Middle-term Management Strategy’ that aims to gain 30 FTTH million lines by 2010. If satisfied, IP telephony will increase market penetration rates much faster than in our projections.