Chapter 5: the conclusion, and what happened next

Chapter 5 closes the argument and then looks past it. The modem was near its physical limit, the study said, and the answer to rising data traffic was packet-switched networks and local competition, not a meter on internet calls.

The finding, stated plainly

The closing chapter is short because the arguing had already been done. Four earlier chapters had walked through the design of a Class 5 switch, the sampling behind the carrier filings, and the five Bell company submissions one at a time. Chapter 5 only had to say what all of that added up to. Selwyn and Laszlo characterized the carrier evidence as anecdote: observations from a limited set of central offices, generalized into a claim about the entire public switched network. Against 23,686 switches nationally, a sample of 127 offices could not carry that conclusion. The blocking the carriers found was real but local, concentrated in end offices serving large providers, and the study had shown traffic could be routed around the one component where it occurred.

The chapter also restated the money, because the money was the whole case for a charge. Second residential lines bought for online use had generated roughly 3.5 billion dollars in cumulative revenue between 1990 and 1995, about 1.4 billion of it in 1995 alone, from an estimated six million lines used mainly to get online. Bellcore's own reinforcement estimate ran to roughly 35 million dollars per operating company per year, about 245 million dollars nationally. Revenue exceeded that cost by a factor of about six. On those numbers a per-minute charge was not cost recovery. It was a second payment for something already paid for, and traffic filling idle capacity would tend to push per-minute costs down, not up.

Why 56k was the end of the road

Then the chapter turns forward, and this is where it stops being a rebuttal and becomes a forecast. Thirteen days before publication, on January 9, 1997, US Robotics had announced its X2 modem technology, putting the top voice-grade dial-up rate at 57.6 kbps. The authors treated that not as the latest step in a series but as close to the end of one. Modems had already outrun what anyone would have predicted a decade earlier, and compression could squeeze more effective throughput out of the same channel, but the constraint was no longer engineering effort. It was the shape of a circuit built to carry a human voice.

That was the sharpest call in the report, and it held. The dial-up modem had one more nominal step and then stopped, while applications kept getting heavier. Graphics, animation and video were already in view in early 1997, and no amount of cleverness inside a voice bandwidth was going to serve them. Every argument in the access charge fight assumed the internet would keep arriving through a telephone call. Chapter 5 said that assumption had an expiration date, and the interesting policy question was what replaced it.

"the laws of physics and mathematics will soon work to limit further data rate gains"

Lee L. Selwyn and Joseph W. Laszlo, Economics and Technology, Inc., for the Internet Access Coalition, January 22, 1997

Packet switching and unbundled loops

The diagnosis follows from the ceiling. The problem was never data on the network. It was data on a circuit-switched network, where every session ties up a path sized for speech regardless of what moves through it. The copper already running to American homes could carry far more than that, and the coalition's own advocacy put the potential gain from ADSL and HDSL at a factor of 100. None of it was speculative. It existed, and the incumbent local carriers were not deploying it at scale.

So the study named two remedies and neither was a price on internet calls. First, build data-friendly networks that move packets at speed, which would lift data traffic off the voice network entirely and dissolve the congestion question rather than answering it. Second, and this is the load-bearing part of the argument, let competing local carriers get at the individual loop components that incumbents controlled, so that someone other than the incumbent could deploy if the incumbent would not. The FCC's access reform proceeding, CC Docket No. 96-262, ran alongside a companion notice of inquiry on data-friendly connections, CC Docket No. 96-263, with comments due March 24, 1997. The coalition was arguing that the second docket, not the first, was where the real answer lived.

The exemption held, dialup boomed

The immediate fight went the coalition's way. The FCC had tentatively concluded, in the notice released December 24, 1996, that internet service providers should not be made to pay interstate access charges, and that tentative conclusion was not reversed. The exemption for enhanced service providers, rooted in the commission's 1983 MTS and WATS Market Structure decision, survived access reform. The flat monthly account stayed the normal way Americans bought internet access, and the arithmetic the coalition had put in front of the public, a 19 dollar plan plus as much as 60 cents an hour, never became a bill anyone had to pay.

Dial-up subscriptions grew through the rest of the decade, second lines kept selling, and the network meltdown that Pacific Telesis had warned about in the autumn of 1996 did not happen. Pacific Bell itself had told California regulators, before the study was even published, that its network showed no deterioration in service. Then the ceiling arrived on schedule. By 1999 the growth story had moved to broadband, and Economics and Technology, Inc. had moved with it, publishing work on building a broadband America and on bringing broadband to rural America that year. The firm that spent early 1997 defending the telephone line spent 1999 arguing about what would replace it.

"The future of ISP/ESP communications lies in the development of alternative, data-friendly networks that possess the capacity to route packetized data traffic."

Chapter 5, Conclusion, Economics and Technology, Inc., 1997

Where the forecast went sideways

The packet-switched future arrived. It did not arrive by the route the study mapped. Chapter 5's remedy was institutional as much as technical: open up the incumbent's loop, and competitors will deploy the data-friendly networks the incumbents were slow to build. That mechanism was the weakest link. The unbundling regime the argument leaned on was narrowed substantially over the following decade, particularly for the broadband-capable facilities that mattered most here, and most of the competitive local carriers meant to use it did not survive. Mass-market broadband ended up being built largely by incumbents and by the cable operators, two networks competing as platforms rather than many competitors sharing one loop.

So the ledger is mixed in a specific and instructive way. The measurement critique holds up completely: the sample really was too small, the revenue really did exceed the cost, the blocking really was confined to a handful of components. The technical forecast holds up: the voice channel was at its limit, and the future was packetized. The policy forecast is the one that did not survive contact with the next ten years. The study is more reliable when it counts things than when it predicts how a regulatory framework will behave, which is true of most advocacy economics, including today's.

Why the episode is still cited

This is why the 1997 report still gets pulled up. It is an early and unusually clean case of an incumbent industry framing a new technology as a threat to infrastructure, and of that framing being tested against measurement rather than argued about in the abstract. The claim was specific enough to check: how many switches were examined out of how many, what the upgrades cost against what the traffic earned, and where in the network blocking physically happens. Once those questions were asked in public, the meltdown story did not hold, and the companies telling it had already begun backing away in their own regulatory filings.

The shape of that dispute recurs. Every few years some class of traffic is described as a free rider straining a network that somebody else paid to build, and the proposed fix is a per-unit charge on whatever is new. The 1997 exchange is worth keeping because of how it was settled: not by deciding whose interests deserved more sympathy, but by checking the sample, the cost accounting and the physics, and by pointing out that the congestion everyone was arguing over would be engineered away within a few years anyway. The infrastructure threat was real in the sense that something did have to change. It just was not the internet that needed to pay for it.