Recently it is rare to hear news of Japan's success, but the broadband business might offer an exception. The number of DSL (Digital Subscriber Line) subscribers tripled within a year to more than 6 million in Japan(refer to MPHPT's table). This is the fastest growth of DSL users in the world, leading to a number of its users second only to South Korea.

On the other hand, the United States is becoming "the Bangladesh of broadband," according to Gordon Cook, the publisher of the Cook Report on Internet. Hundreds of CLECs (Competitive Local Exchange Carriers) invested in DSL during the dot-com boom, but most of them went bust when the bubble burst. Why did Japan's policy work while the American 1996 Telecommunications Act resulted in the massacre of CLECs? Here are my tentative answers.

1. Unbundling and co-location are necessary but not sufficient conditions for the proliferation of broadband.

At the same time the U.S. mandated the unbundling of local loop and co-location of competitor's equipment in telephone offices under the 1996 Act, Japan's MPHPT (Ministry of Public Management, Home Affairs, Posts and Telecommunications) enacted similar rules, to no avail. A few startups tried to begin DSL services, but NTT (Nippon Telegraph and Telephone) delayed the opening of their lines by "testing" the DSL equipment for more than a year.

Thus, in 2001, when Son Masayoshi, the president of SoftBank, began a DSL service called "Yahoo! BB" at the surprising price of 990 yen (2830 yen including ISP services) per month for a connection of 8 Mbps (Megabit per second), most people were skeptical about its profitability. However, Son invested more than 100 billion yen in Yahoo! BB, selling most of the assets that he bought during the bubble years.

At first NTT resisted the co-location, but Son urged NTT to open the lines in the governmental IT Strategy Council, where he and Miyazu Jun-ichiro, the president of NTT, were members. Furthermore, the Fair Trade Commission of Japan accused NTT-East of unfair treatment of DSL carriers, which was a shocking event for NTT, a half-national company. So it lessened its resistance and opened its facilities nationwide.

As a result, in two years, Yahoo! BB became the biggest broadband service provider in Japan with 2 million subscribers. Because no DSL carrier is making money in Japan, the prospect of broadband business is uncertain, but one thing is obvious: if the unbundling policy were not enforced, DSL could not prevail. Moreover, aggressive entrants and honest incumbents are indispensable for the proliferation of broadband.

Theoretically, the net effect of unbundling is ambiguous because it encourages entrant's investment while it reduces incumbent's incentives. However, from Japan's experience, we can say that unbundling facilitates investment as a whole if it is enforced adequately. Unbundling invites entry of new competitors and, eventually, increases incumbent's investment. Incumbents would not cannibalize their telephone business unless competitors threatens them to do so.

Optical fiber is expensive and useless for residential users, because 95% of Internet uses are mail and Web. So the "unregulation" policy of the FCC (Federal Communications Commission) that encourages facility-based competition would not work until competitors can arbitrage wired and wireless connections as a result of opening the spectrum for wireless Internet. Therefore the FCC's new policy to put an end to unbundling is likely to make things even worse.

2. The transition to IP networks can be painful for incumbent carriers.

Now Softbank is deploying a VoIP (Voice over Internet Protocol) service called "BB Phone" for every subscriber to Yahoo! BB. It costs only 7.5 yen per 3 minutes to call from Tokyo to New York, and it is free if both sides are using BB Phone. This could become a fatal blow to NTT, because BB phone is an end-to-end (E2E) system of VoIP that connects telephone terminals to IP networks through DSL without passing through telephone switches of NTT.

Part of the problem was created by strategic mistakes on the part of NTT. The "overlay price" to send data through the access lines of NTT has fallen to 173 yen per month. The wholesale price of dark fiber is 3.9 yen per meter per month. These are arguably the cheapest wholesale prices offered by incumbent carriers. Moreover, NTT should interconnect its lines with those of competitors in all areas in Japan. These bargain prices are the results of adoption of forward-looking cost estimation that the USTR (US Trade Representative) demanded in trade negotiations.

In 2000, when the USTR attacked NTT that its interconnection fees for access lines were too high, NTT was embarrassed because this was a domestic problem. NTT became preoccupied with this strange trade talks and tried to avoid the drastic price reduction demanded by the USTR (22.5% in two years and 40% in four years) at any cost. As a bargaining chip for the negotiation, NTT agreed to lower the wholesale prices for DSL as requested by the MPHPT. NTT did not take DSL seriously because it was planning to migrate from ISDN to FTTH (Fiber To The Home).

This mistake was expensive for NTT because it prevented NTT from making money with broadband. Yahoo! BB made the most of this opportunity by connecting their access line directly to the core network of 10-Gigabit Ethernet over the cheap dark fiber leased from NTT. Yahoo! BB "disintermediated" PSTN using NTT's pipes.

This year NTT had to raise the interconnection fees for the first time in its history. However, this measure is far from a cure, because revenue from PSTN is falling 10 per cent annually. Traditional rivals such as KDDI are arguing against the price hike, but this can be an incentive for ISPs (Internet Service Providers) that are going to invest in E2E VoIP services. Now it is cheaper to interconnect to NTT's switches than to invest in E2E, but the latter will be attractive if telephone fees go on rising.

3. In addition to facilities, the organization of incumbent carriers must be unbundled to realize broadband networks.

Kevin Werbach, an Internet analyst, commented to me that the wholesale prices were reasonable because incumbents had already amortized most of their assets. He argued that if such prices put incumbents out of business, that would imply that the incumbents were so inefficient and hidebound that they must die. That might be true, but they die hard; thus the regulatory nightmare seen in the U.S.

The unbundling of facilities would not be successful unless it is accompanied by the unbundling of organizations, because vertically integrated incumbent carriers would resist the "regulatory takings" that infringe their property rights. Even if their claim is justified, deregulation should be limited to new investment in IP networks. Old telephone swicthes, together with most of their employees, must be unbundled from pipes and scrapped as soon as possible.

As for NTT, I recommended that the whole NTT group should be reorganized into three companies; NTT-IP, NTT DoCoMo, and NTT-Telephone. NTT-IP (or NTT-Verio) would be a facility-based ISP that operates internationally. NTT-Telephone would be a re-nationalized company whose stocks can be converted from the 46 per cent stake that Japanese government has in NTT. Other companies must be completely private.

This scheme divides the problem into two parts: one is the transition from PSTN to IP networks; another is the liquidation of telephone switches with as few social costs as possible. The latter problem is far more complicated and politically difficult. If these two problems are bundled, they are usually brought to a deadlock as a result of strong resistance from incumbents with tight grip on facilities, as evidenced in the U.S.

NTT unintentionally made loopholes for others to "skim the cream" which hastened the end of PSTN. It is partly a result of its honest observance of unbudling obligation, but partly a result of miscalculation as described above. This is indeed bad news for NTT, but it might be a blessing in disguise for Japan, because it will accelerate the migration to the open broadband networks into which many startups can enter. Such migration is technically possible but barred by incumbents who monopolize local loops in other countries. From the banking crisis in 1990s, we learned that, if Doomsday is inevitable, it is better if it comes earlier.