BENEATH THE ENGLISH CHANNEL — Having just reoriented myself to turning my head to the right when about to cross the street, I have boarded the train at St. Pancras with Maggie, as we leave London and head to the 8th Conference on Philosophy and Economics in Nancy, France — via Paris, of course. There is a stretch of this journey, an hour or so out of London and deep beneath the seabed, where the phone signal dies and there is nothing to do but think about infrastructure. The train I am on is doing the genuinely hard part right now: a fifty-kilometre tube bored under the sea, the most famous bottleneck in European travel. Except it is not the bottleneck. It is the easy part. The hard part was a rail yard in east London, and to explain why, I have to start with the map I left behind.
The previous dispatch in this series was about that map — Harry Beck’s diagram of the London Underground, and the honest way a good diagram treats position. This one is about something a map cannot fix.
The simple line
The London Underground is a genuinely complicated object: eleven lines and some two hundred and seventy stations, assembled over more than a century out of a pile of separate and once-competing private railways. And it coheres. One ticket carries you across all of it, one fare structure governs the whole, one diagram makes the tangle legible to a tourist who landed this morning. Now consider the network this train is part of. From St Pancras the cross-Channel railway reaches a short list of cities — Paris, Brussels, Lille, Amsterdam — through a single tunnel. By any measure of structural complexity it is a toy beside the Underground. And it is the toy that a national regulator has spent the past two years refereeing a four-way fight over.
If the difficulty of coordinating a transport system rose with its size — the number of lines, the number of firms once involved, the count of junctions where everything has to agree — then the Underground would be the hopeless case and the cross-Channel railway would run itself. The opposite is true. The enormous tangle was coordinated decades ago and has stayed coordinated. The small clean network is stuck. Either size is not the variable, or something else is doing the work. It is the second.
What 1933 actually did
The Underground was not designed. It accreted. The Metropolitan Railway opened in 1863, the first underground railway anywhere in the world; the District followed; and around the turn of the century a cluster of deep-level tube lines was dug, several of them financed through a single holding company assembled by the American streetcar magnate Charles Tyson Yerkes.1 These were rival businesses. They duplicated routes, sold their own tickets, printed their own maps, and the competition between the Metropolitan and the District around the Inner Circle was genuinely bitter. The thing a passenger faced was not a network. It was a quarrel with stations.
That quarrel was settled twice, and both times in living memory. From around 1908 the companies began to market themselves jointly under a single word — UNDERGROUND — on a shared map. And in 1933 the matter was closed outright: a public body, the London Passenger Transport Board, absorbed the tube companies, the Metropolitan, the buses and the trams into one organisation, with one set of accounts and one authority finally answerable for the whole.2 Beck’s diagram reached the public that same year.
Here is what is worth holding onto. The coordination problem the Underground posed — incompatible schedules, rival ticketing, a passenger expected to know whose train she wanted — was real, and for decades it went unsolved. It was not solved by making the network smaller or simpler. It was solved by absorbing the network into a single decision-maker. A coordination problem does not evaporate when you tidy it; it has to be housed somewhere. The year 1933 housed the Underground’s inside one organisation, and Beck, the same year, housed its complexity inside one diagram — a governor and a representation. After that the network could be as intricate as it pleased, because the intricacy finally had somewhere to live.
The Chunnel’s problem is small, and structural
Now the cross-Channel railway, and why being small did not save it. The line through the tunnel is governed by a regime called open access, which guarantees any qualified operator the right to run trains and exists precisely to stop a single company from owning the route. Open access forbids, as a matter of design, the move of 1933. You cannot fold the cross-Channel operators into one accountable authority, because the entire purpose of the regime is that there should be several of them, competing.3
So the residual problem has nowhere to live — and there is a residual problem. Continental trains are large, particular machines; they must be maintained and stabled at a depot, and in Britain exactly one depot can take them: Temple Mills, in east London, occupied by the incumbent, Eurostar, which for thirty years has been the tunnel’s only passenger operator. Four challengers — Virgin, the Spanish start-up Evolyn, Gemini Trains, and the Italian state operator Trenitalia — applied to the Office of Rail and Road for access to that depot. The regulator commissioned an independent study of how much room it had. The study found space for the incumbent’s own expansion, or for one new entrant, but not for both.4
So in October of last year the regulator did the only thing the arithmetic left it. It ran what was, in form, a neutral access regime and, in substance, a contest, and it named a winner: Virgin, judged to have the most financially and operationally robust plan. The other three were turned down. Notice what open access became under scarcity. A rule whose whole purpose was to ensure that nobody held discretionary power over the route resolved into a regulator choosing, by discretion, one firm out of four. The neutrality did not disappear. It relocated, to the body administering the rule.
And the scarcity itself persists for a reason worth stating plainly, because it has a textbook shape. The depot could be expanded; expanding it would dissolve the whole problem. But an enlarged depot is a benefit shared by every operator who would use it, and the operators have each said, in their various ways, that they will help pay for it once the others do. An improvement that everyone wants and no one will fund alone is a public good among competitors, and public goods among competitors do not get funded. So the bottleneck stays exactly where it is.
This is the pattern this blog calls conservation of impossibility: a hard problem is not abolished, only moved, and the question always worth asking is where it has been moved to. The Underground moved its coordination problem into a single organisation. Open access moved the cross-Channel railway’s into a depot. One regime parks the difficulty in an accountable governor; the other parks it in a rail yard the incumbent happens to occupy — and a bottleneck occupied by an incumbent is the object the previous dispatch was really about. Temple Mills is Warwick the Kingmaker rendered in steel and overhead wire: the single point every path must run through, conferring on whoever holds it a power that has nothing to do with running good trains and everything to do with sitting at the interchange. Open access, in the end, resolved into the incumbent and its one approved rival sitting down to negotiate the terms of the bottleneck between themselves.
There is one more turn, and it is the one that connects to where this train is taking me. The entire contest rested on a single number: the depot’s spare capacity. The incumbent, which holds the depot, said the number was essentially zero. The challengers said it was not. The regulator could not simply look the number up, because a depot’s capacity is not painted on a wall — it depends on shift patterns, maintenance schedules, how trains are stabled overnight, and what one is willing to spend. It is a measurement, and measurement, as Maggie and I have argued, is never the innocent reading-off of a fact; it is a construction, and often a contested one.5 This casts no aspersion on the incumbent: an operator that owns the only depot has every ordinary incentive to experience it as full, and would do so in perfect sincerity. An “independent assessment” is simply what an institution reaches for when the number it needs would otherwise be supplied by the party with the largest stake in its value.
Which is, more or less, where I am headed. Trenitalia, shut out of Temple Mills, has not given up; it is building its own depot near Paris and will challenge from the French side instead — the bottleneck not removed but routed around, rebuilt on the far shore.6 I am, in my small way, doing the same thing. This train comes up into the light near Calais, and a few hours after that I will be in Nancy, where Maggie gives a keynote on three paradoxes of optimal evaluation. An independent assessment of a depot is an evaluation. So, it turns out, is a great deal else. More on that from France.
With that, I leave you with this.
Notes
1 The Metropolitan Railway, opened in January 1863 between Paddington and Farringdon, was the world’s first underground passenger railway. The deep-level tubes that followed a generation later were largely the work of the Underground Electric Railways Company of London, a holding company assembled around 1902 by the Chicago transit financier Charles Tyson Yerkes. The competing lines were private, profit-seeking and frequently at each other’s throats; the shared roundel and the common “Underground” branding, from about 1908, were a marketing truce well before they were an institutional one.
2 The London Passenger Transport Board was created by statute in 1933 and absorbed the Underground Group, the Metropolitan Railway, and a long list of municipal and independent bus and tram operators into a single public corporation. That Beck’s diagram first reached the travelling public the same year is a coincidence of timing, but a telling one: the network acquired a single governor and a single legible representation in the same twelve months. On the diagram itself, see the previous dispatch in this series, “Two Doctrines of Position.”
3 Under the open-access regime, infrastructure deemed an “essential facility” — including maintenance depots — must be made available to operators on fair and non-discriminatory terms; the relevant applications here were made under section 17 of the Railways Act 1993. It is worth noting that the tunnel itself is not the constraint: it is owned by Getlink and runs at roughly half the traffic it was built to carry. The scarce thing is not the crossing but the servicing.
4 The four applicants were Virgin, Evolyn, Gemini Trains and Trenitalia. The Office of Rail and Road’s independent assessment concluded that Temple Mills could accommodate either Eurostar’s planned fleet growth or one new operator, not both. In October 2025 the regulator approved Virgin’s application and declined the other three; by early 2026 Eurostar and Virgin had submitted a draft access agreement, though Eurostar had earlier weighed a legal challenge to the decision. New services are not expected before the end of the decade.
5 John W. Patty and Elizabeth Maggie Penn, “Analyzing Big Data: Social Choice and Measurement,” PS: Political Science & Politics 48(1): 95–101, 2015. The general point — that a measurement encodes choices, and that those choices can be contested in the same way an election can — runs throughout the work Maggie and I are doing on classification and evaluation.
6 Having been refused Temple Mills, the FS Italiane group — Trenitalia’s parent — has shifted to a France-based plan built on a new depot near Paris, with a target of cross-Channel services later this decade. The bottleneck, denied, is simply being rebuilt on the other side of the water: conservation of impossibility, observed in real time.