How the Bells built a case against dial-up internet

Between June and October 1996, four regional Bell companies and their shared research arm filed papers with the FCC warning that modem traffic was overloading local switches. The Internet Access Coalition hired economists to check the claim.

Five filings and one alarming speech

The breakup of AT&T in 1984 left seven regional holding companies owning the local wires that ran into American homes. Ameritech, Bell Atlantic, BellSouth, NYNEX, Pacific Telesis, SBC and US West were known collectively as the Baby Bells, and their operating subsidiaries were the Bell Operating Companies, or BOCs. Over five months in 1996, four of them plus Bellcore, the research organization they jointly owned, put a shared argument in front of federal regulators: the growth of consumer online services was tying up their central office switches and would soon force expensive rebuilding that nobody was paying for.

The paper trail is specific. Bell Atlantic filed a report on internet traffic dated June 28, 1996. Pacific Bell followed with its ESP Impact Study on July 2. NYNEX wrote to the chief of the FCC's Competitive Pricing Division on July 10. US West delivered the final results of its ESP Network Study on October 1. Bellcore researchers Amir Atai and James Gordon circulated a modeling paper on the impact of internet traffic on local carrier networks and switching systems that same year. Together these five documents are what the coalition's economists were hired to examine.

A meltdown that quietly went away

The public version of the argument ran hotter than the filings. In early October 1996 the San Francisco Examiner reported that Philip J. Quigley, chairman and chief executive of Pacific Telesis, had told an audience that surging online activity was piling equipment and service costs onto Pacific Bell and risked bringing the local network down altogether. The word he used was meltdown, and it traveled. For anyone reading a newspaper rather than a docket, that single image was the whole story: hobbyists with modems were about to break the telephone.

The retreat was quieter. On November 8, 1996 a Pacific Bell spokesman, John Britton, gave the San Jose Mercury News a considerably milder account. Six days later, in comments to the California Public Utilities Commission, the company stated plainly that its network was not degrading. The walk-back drew a fraction of the attention the speech had, and it changed nothing about the filings, which stayed on the federal record and continued to shape the policy debate in Washington through the winter.

"meltdown in the [local telephone] network"

Philip J. Quigley, chairman and CEO of Pacific Telesis, quoted in the San Francisco Examiner, October 4, 1996

Why online services paid flat rates

To see what the Bells actually wanted, you need one piece of regulatory plumbing. When a long distance carrier handed a call to a local phone company for delivery, it paid that company a per-minute access charge, governed by Part 69 of the FCC's rules at 47 CFR. Online services never fell into that bucket. A 1983 FCC order, MTS and WATS Market Structure, reported at 97 FCC 2d 682, treated them as end users rather than carriers. They bought ordinary business lines under state tariffs and paid the flat monthly prices those tariffs set.

The regulator's term for these companies was enhanced service provider, or ESP: a business that processed information rather than merely transporting it. CompuServe and America Online were ESPs, and so were the internet service providers that followed, such as EarthLink. The 1997 study treats ESP, ISP and online service provider as one category. The practical result was that a subscriber dialing a local modem bank produced a call the switch could not distinguish from a call to a neighbor. Reclassifying those companies would have turned every dial-up minute into a billable event.

The Bells were not making it up

The engineering complaint underneath the rhetoric was genuine, and it deserves stating at full strength. The circuit-switched network was designed around measured voice behavior: calls lasting a few minutes, with equipment sized on the statistical bet that only a small share of subscribers would be off-hook at any instant. A central office switch does not hand every line its own dedicated path. Lines share concentrating equipment, and the sharing ratio assumes most of them are idle most of the time. A modem session holding a circuit open for two hours breaks that assumption, and where many such lines landed on the same shared components, other customers on those components really could wait for dial tone.

What the coalition's economists disputed was the scale of the problem and the remedy proposed for it. Bell Atlantic's own filing reported that roughly half the ESP lines in its territory were already connected in a way that bypassed the contended components, the same treatment routinely given to other heavy users such as PBX trunks. That reframes the congestion as a provisioning decision confined to particular offices rather than a systemic threat. The BOC studies also examined 127 central offices out of roughly 5,200 in their own territories, about 2.4 percent, against 23,686 switches nationally, then generalized from that slice to the whole network.

"no deterioration of service in our network"

Pacific Bell, comments to the California Public Utilities Commission, November 14, 1996

What the coalition asked ETI to test

The Internet Access Coalition was a 24-member alliance of technology companies and trade groups, including America Online, Apple, Compaq, Dell, IBM, Intel, Microsoft, Netscape, Oracle and Sun Microsystems, chaired by steering committee head Paul Misener. It retained Economics and Technology, Inc., a Boston consultancy at One Washington Mall founded in 1972 by Lee L. Selwyn, an MIT-trained economist who had appeared as an expert witness before some forty state commissions, the FCC, Canada's CRTC and the United Kingdom's Oftel. Selwyn and his colleague Joseph W. Laszlo wrote the report, which was released on January 22, 1997.

The introduction is notable for what it concedes. It accepts without argument that whoever causes a cost should bear it, then turns the dispute into two factual questions a regulator could actually resolve: whether online users imposed costs on local networks out of proportion to other heavy users, and whether the traffic they generated was genuinely uncompensated. The timing was deliberate. The FCC had issued its access reform notice on December 24, 1996 in CC Docket No. 96-262, with a companion inquiry on data-friendly connections in Docket 96-263 and comments due March 24, 1997.

Three findings and the consumer stakes

The study reached three conclusions. First, data traffic posed no meaningful threat to network integrity, because dial-up sessions were dispersed across the national switch base rather than concentrated, and much of the usage fell in off-peak hours when capacity sat idle anyway. Second, the revenue side had been left out of the Bells' arithmetic entirely. Additional residential line penetration had climbed from roughly 2.5 to 3 percent before 1990 to 14.7 percent by 1995, with about 6 million residential lines used primarily for online access that year, producing $1.4 billion in 1995 alone and some $3.5 billion cumulatively from 1990 through 1995. Bellcore had estimated network reinforcement at $35 million per BOC annually, or $245 million nationally, a figure those revenues exceeded roughly sixfold.

Third, per-minute access charges were the wrong instrument. Money collected at the ISP's end of a call carried no obligation to be spent on data-capable infrastructure, and competitive firms finance new plant through debt or equity against expected future revenue rather than by raising prices on services they already sell. The alternative the study favored was local competition plus data-friendly technology, ADSL and HDSL lines and packet switching, at a moment when US Robotics had only just announced its X2 56k modem on January 9, 1997. For households the stakes were concrete: flat-rate accounts ran about $19 a month, and access charges of up to 60 cents an hour would have added roughly $12 to a 20-hour month. The FCC's tentative conclusion held, the exemption survived, and flat-rate dial-up carried American consumers onto the internet.