A railway isn’t like a city street, where every additional intersection slows you down. Instead, it’s more like an expressway, where you can add a dozen interchanges but through traffic can pretty much ignore them. From the dawn of the railway age, there have been express and local trains. Modern high-speed rail is no different, since it’s an approach that massively improves the value you get from expensive infrastructure.

The Tokaido Shinkansen, which runs from Tokyo to Osaka, was the world’s first modern high-speed railway. The 515-km route, similar to Toronto-Montreal, has 15 intermediate stations. Yet a rider between the two megacities on either end wouldn’t even notice these stations existed. The trains on the line are sorted into three tiers: The Nozomi services are the fastest and provide nearly non-stop journeys between Tokyo, Nagoya, Kyoto, and Osaka. Hikari services are an intermediate class that add a few secondary stops. Kodama services are the high-speed locals making every stop (not an oxymoron!).

The secondary stations typically have four tracks, with two in the middle for the Nozomi to blast through at top speed and tracks on the side with platforms for the Kodama trains. The services are scheduled so that the fast trains can pass while the slow ones are stopped at the secondary stations.

The Kodama services are impressively well-used, especially at commuter times, because they still run at full high speeds while offering significantly lower fares, making it possible for people to commute to work in central Tokyo within an hour while living in a peaceful town over 150 km away.

Of course, Japan is one of the densest countries on earth, so one might think that’s the reason for the many intermediate stations. But just about every other country with high-speed rail has a similar model combining end-to-end express trips with services to intermediate destinations.

Consider the LGV Est, which is the high-speed line from Paris toward Strasbourg and Germany. Over 406 kilometres, it has three intermediate stations that rely mostly on park-and-ride traffic. The French high-speed lines keep speeds up and costs down by not building directly into the centre of every city (a very different approach from Germany, which I’ll write about later). Some of these outlying stations work better than others, but the more successful approach is the point-to-point service that France uses so effectively. TGVs branch off onto classic lines all over the network so that they can directly serve the main stations in the centres of secondary cities while allowing fast journeys along the high-speed line right to Paris. Each secondary city gets its own more-or-less dedicated train from Paris several times a day, running along the high-speed line until it intersects with a classic line, which it uses for the final run in to the station.

Subscribe for free, for infrastories from around the world.

Subscribe now

Nancy is a city of about 100,000 with a university, located about 350 km from Paris. If someone there wants to travel to the capital on a typical weekday, they have 9 dedicated non-stop or one-stop trains per day from the historic station in Nancy’s city centre. The trains use the old line through the Moselle valley for about 35 km to a junction with the high-speed line, and then run at up to 320 km/h the rest of the way to Paris. The opening of the high-speed line cut the journey from 3 hours to about 90 minutes.

The town of Ciudad Real is over a two-hour drive from Madrid, which is far too long for commuting. But when the Madrid to Seville AVE high-speed line opened in 1992, a train trip was only an hour, allowing Ciudad Real to effectively become a suburb of the capital overnight. The Spanish national railway eventually created a whole class of cheaper medium-distance high-speed services called Avant to carry commuters to places like Ciudad Real. In Britain, Southeastern’s domestic high-speed services carry Kent commuters on the same infrastructure built for the Eurostar. Commuter-oriented services to secondary cities, as long as they are priced appropriately, can massively improve the utilization of high-speed rail infrastructure. In effect, they’re almost free money that can make infrastructure intended for trips between major cities financially viable.

There is no trade-off between adding stations and running fast trains. The flagship express can run just as fast whether there are 15 intermediate stations or none. Everything else, like the Hikaris, the Kodamas, the Avants, and Southeastern High Speed, uses the vast additional capacity of the line.

This brings us to Alto, Canada’s project to finally build that Toronto-Quebec high-speed line that has been discussed for so long. As currently planned, it is a roughly 1,000-kilometre line from Toronto to Québec City, built for 300 km/h or more, with seven stations: Toronto, Peterborough, Ottawa, Montreal, Laval, Trois Rivières, and Québec.

The equivalent of the Nozomi services would run from Toronto to Montreal with only a single stop in Ottawa. Ideally, you could even build a bypass skirting the south end of Ottawa to avoid the slow, meandering track through the city. In that case, your Nozomis might be an hourly Toronto-Montreal non-stop, which would take around 2 hours, and the Hikari services would also stop at Ottawa and maybe one more station such as suburban Toronto or Montreal. Then, the Kodama equivalent would run from Toronto to Ottawa making intermediate stops, depending on the chosen alignment, at stations like suburban Toronto, Peterborough, Port Hope/Cobourg, Belleville/Prince Edward County, Kingston, and Smiths Falls. Now, don’t think of these as a pokey VIA service. They’d be running at 250km/h+ on a dedicated high-speed line, so they’d still be extremely fast—far faster than driving. The seating might be a little less luxurious than on typical high-speed trains, there’d be less room for luggage since most riders would just be going for the day, and there likely wouldn’t be extra amenities like a café car, but this would all help keep fares affordable enough for commuters.

Like the incredibly successful Ciudad Real or Southeastern High Speed services, these intermediate stops would likely be very well used, especially at rush hour. Nancy and Kingston have a lot in common. Given the housing crisis in Toronto, surely more than a few people would be quite happy to have a house in a beautiful historic town, less than an hour by comfortable train from Union Station, for hundreds of thousands of dollars below Toronto prices.

Fare policy is critical to making this work. We’ve seen it before with UP Express, the express train from downtown Toronto to Pearson Airport. When it opened in 2015, it had a high fare targeted at business travellers who would otherwise be taking a taxi. Unfortunately, revenues were tiny and reporters loved taking videos of the deserted trains. When the fares were slashed to match the GO fares, ridership quadrupled within a year. There was a gigantic commuter market that high fares had left completely untapped. Now, the main problem is overcrowding. The key with a new high-speed service is to plan for this from the outset, with high-capacity commuter-oriented trains and lower fares.

Different fare structures, like the monthly passes and frequent traveller discount cards available on the French TGV, can keep prices affordable for frequent riders. From Nancy to Paris, for example, you can get a monthly pass for about €436 that gives you enough free journeys for 3 round-trip commutes a week. Not bad for a 350+ km one-way journey. But French employers generally subsidize monthly commute passes at about 50%, so, all in, the commuter is generally paying about €9 out of pocket for each trip. That’s no more than a GO Train from Hamilton or LIRR from Hempstead. Even without the employer subsidization, that’s still quite reasonable given the much lower cost of living that far out.

The logical basic service pattern for Alto is one super-express train per hour to provide the fastest Toronto-Montreal journey times, one skip stop-style service per hour with a few intermediate stops but still a speedy end-to-end journey, and one Kodama-style train per hour making local stops. At peak times, you add more services as needed. Places like Peterborough or Belleville could likely fill a whole express train each on their own with commuters to Toronto at rush hour, if it’s priced right. The Kingston area has over 150,000 people. You wouldn’t need it to become a bedroom suburb of Toronto to get 600 people a day on a roughly one-hour train. In some cases, you could even have trains leave the high-speed line to run a short distance on a branch to the centre of the intermediate cities, allowing the high-speed corridor to be built relatively inexpensively on the outskirts.

So, the next time you hear that high-speed railways can’t serve intermediate destinations, remember that almost all of them do. The fast trains just don’t stop there.

Myth 1: The Yonge Subway was paid for out of the fare box

Torontonians voted on New Year’s Day 1946 to begin construction of Canada’s first subway system. The TTC had emerged from the war with a $25 million surplus and the expectation was that they could cover the new subway’s cost with the surplus and federal postwar reconstruction money. Alas, the federal money never materialized and a planned Queen Street tunnel was dropped. The Yonge line ended up costing $58.5 million, which much of it borrowed against future farebox revenues. As it turned out, ridership continued to decline and the farebox didn’t come close to covering the debt. On Dec 6, 1963, Metro Council voted to assume $51.8 million in TTC subway debt and to raise Metro’s share of future subway construction from 55% to 70%. It was Metro’s taxpayers that ended up covering a big chunk of the Yonge subway’s cost, not TTC farepayers.

Myth 2: The TTC was profitable before they got rid of the zone fare

In 1973, the TTC finally got rid of the double fare that residents outside the old City of Toronto had to pay when travelling downtown. We’ve often heard that decision blamed for the end of the TTC’s ability to support itself financially, but that’s not true. The TTC was already running an annual deficit in the mid-1950s and Metro began covering deficits with ad hoc subsidies. This was a perpetual political sore point as suburbanites pointed out that their taxes were bailing out the TTC while they were getting hit with much higher fares than downtowners. In August 1962, Metro Council voted 13–10, after eight hours of debate, to request formal provincial permission to subsidize transit. The TTC had lost 15 million riders in 1961 alone and the councillors embraced the concept of subsidy in order to fund the expansion of suburban transit service, as well as to minimize a planned fare hike.

The myth might come from the timing of the double fare’s elimination, but it gets the causality backwards. A few weeks after Bill Davis announced the cancellation of the Spadina Expressway in 1971, he announced a generous provincial subsidy program for transit. Queen’s Park would thenceforth cover as much as 50% of agency deficits. New TTC board member and North Yorker Paul Godfrey saw this as the opportunity to as last get rid of the double fare. Eliminating the double fare had not led to the subsidy. On the contrary, the subsidy had allowed for the elimination of the double fare. As it turned out, eliminating the zone fare ended up costing far less than expected because ridership jumped 12.3% as a result.

Myth 3: The streetcars were doing great until recently

In the 1950s, the TTC’s streetcars moved incredible numbers of people. The Bloor streetcar alone moved almost as many people as a subway line, with streetcars coupled together to carry the heavy loads. The TTC’s slogan was “always a car in sight,” since they were so frequent you could always see the next one coming. Back then, several routes carried in excess of 100,000 people every day. As of 2024, even the busiest King car moves only 61,000 a day and the Queen car moves 42,000. Dundas and Carlton are even lower. And anyone who has been around a while knows that those corridors have seen massive development and remain home to businesses that draw people from all over the city. There are many reasons for the decline, including reduced service frequency, route management challenges, ever-longer travel times, and vehicle congestion. While subway and suburban bus ridership boomed for decades, streetcar service and ridership declined. Even amid strong overall TTC growth in the 80s, vehicle-miles of service sank from 9.3 million in 1981 down to 8.2 million in 1990. Ultimately, when service gets slower and less frequent, fewer people will ride.

I’m currently working on a project called FastTrackTO that is pushing for various interventions and changes that would improve the streetcar network and likely increase ridership - alongside Reece Martin of The Transit Brief.

Subscribe for free, for infrastories from around the world.

Subscribe now

Myth 4: Toronto’s transit has always been stronger than American cities’

As people were buying cars and suburban houses after the end of the Second World War, Toronto transit ridership declined much like in American cities. That is, until they suddenly turned it around. From 1952 to 1962, Toronto’s transit ridership fell 37.6%, which was quite comparable to Chicago (38.8%), Washington (40.3%), and Philadelphia (41.0%). The opening of the subway in 1954 did little to arrest the decline and the TTC hit its post-Depression nadir in 1961, when its financial survival was an open question. But after the major bus expansion to the suburbs in 1963, the TTC dramatically turned around its fortunes and boomed right up until the late 80s, while American systems typically continued their steady decline for at least another decade.

Myth 5: Expanding transit to the suburbs was a financial disaster

The TTC has operated far more transit service in its vast postwar suburban service area for more than half a century. And over much of that time, it had the lowest subsidy of any major transit system in North America. The TTC’s 1962 Annual Report warned that “bus routes into new suburban districts are heavy money losers, sometimes for many years.” But when they finally got pushed by suburban Metro councillors to radically expand suburban bus service, to everyone’s astonishment the routes turned a $3.9 million profit in the first year and drove suddenly and unexpectedly skyrocketing ridership. The TTC’s consultants had projected 8.25% citywide ridership growth from 1962 to 72, but the actual figure was 28.7%, pretty much all of which occurred in the suburbs. The expansion saved, rather than destroyed the TTC’s independence for many more years. Far from a drain, suburban service is what put the TTC on a much more stable footing.

Myth 6: Nobody rides the Sheppard Subway

We’ve often heard the Sheppard Subway described as a deserted subway to nowhere, but by normal subway standards it’s respectably busy. Toronto has the busiest subway per kilometre in North America, which means we expect trains to be packed for much of the day. That sets up rather unreasonably high standards. Sheppard in 2023-24, despite pandemic related declines, was moving about 39,000 people every day. Before the pandemic in 2019, it was about 48,000. That’s not bad for a suburban stub that’s only six kilometres long. Miami’s Metrorail moves about 50,000, and it’s nearly 40 kilometres long. It’s busier than many US light rail systems that are dozens of kilometres long. Toronto’s subway is successful because most of its ridership comes from connecting buses, so the catchment area of stations is far larger than just the distance people could walk. Considering that the Sheppard subway was truncated to Don Mills, so it only connects with one notably busy north-south bus route, its ridership is pretty high. If it continued east to connect with the busy Victoria Park, Warden, and Kennedy buses, or west to connect with the other leg of Line 1 it could be expected to be much busier.

Myth 7: Transit City was killed

We’ve often heard that Transit City was killed, but it’s been the basis for much of Toronto’s planning since it was unveiled in 2007, and we’ve built many of the most significant sections. The original Transit City plan included:

• Eglinton Crosstown LRT

  • By far the largest project of the whole Transit City plan, the line was built as planned and is now open.

  • Its west extension to Renforth is now under construction, although it will be underground while the Transit City plan likely intended for that section to be above ground. If anything this is an upgrade.

• Finch West LRT

  • Built as planned and now in service

• Don Mills LRT

  • The Don Mills LRT from Pape Station to Don Mills and Sheppard faced the same constrained road alignment on its southern section. More than half of the line is now under construction as the Ontario Line, which will cover the Pape Station to Don Mills and Eglinton section and is another upgrade.

  • Long term planning suggests potential northern extensions of the Ontario Line that would effectively complete the Don Mills line.

• Sheppard East LRT

  • This project was planned to run from Don Mills station on Line 4 to somewhere around Twin Rivers and Sheppard.

  • It is now being studied, likely as a subway to a junction point with Line 2 near Scarborough Town Centre.

• Scarborough-Malvern LRT

  • This line was planned to run from Kennedy Station east on Eglinton, Kingston Road, and Morningside to UTSC and Sheppard.

  • This line is currently under study and considered a priority of the City of Toronto.

• Waterfront West LRT.

  • This was a short line from Exhibition Loop to connect to the Queensway streetcar ROW, so that the streetcars from Long Branch could run onto the 509 corridor.

  • This project is still planned by the city as part of work in the area around Exhibition place with a new section of track connecting to King Street along Dufferin.

• Jane LRT

  • This line was to run from Bloor to Steeles West station.

  • It was likely unbuildable as the fully surface LRT that Transit City had called for, since the southern section of Jane Street is too narrow to fit a full ROW.

Notably, the original Transit City did not include rebuilding the former Scarborough RT to run low-floor LRVs, which was a concept that later became associated with the plan.

Out of all the seven Transit City lines, two have been built as planned, including the largest single project. One is being mostly built as part of the Ontario Line subway. Two are currently high priorities in the planning process, with another likely to happen as Liberty Village and the Ex continue to develop. The last one was likely unbuildable as planned. That’s a pretty successful record of follow-through for any transit plan.

Myth 8: The Danforth urgently needs intensification to support subway ridership

The point here is not that intensification is not valuable. It’s that Line 2 along the Danforth is already an extremely busy line. At peak periods, riders at many of the stations along the Danforth struggle to squeeze onto trains that are already packed with riders coming from further east in Scarborough. (Capacity should increase as new trains and CBTC signalling are introduced into the 2030s.) The Danforth’s stations are also already pretty busy, especially when they are fed with the connecting buses that bring most subway riders to stations. But even Chester station, with no bus service at all, is busier than every BART station except a handful in downtown San Francisco. This is because a bustling commercial area and a good transit mode share can create high transit ridership even without very high residential density. And sometimes height isn’t everything. Even a large condo with 300 units might generate a couple hundred trips a day, which could be far less than a much smaller medical office building, for example, that might have dozens of medical practitioners, each of whom might see an average of a dozen patients per day coming from all around. Dense commercial development usually generates the most ridership of all.

If there is a corridor, however, that cries out for intensification, it’s the western leg of Line 1. Of all the four radial routes into central Toronto, the Spadina line is the only one that has always had ample excess capacity. There’s no better place in the city, from a transportation perspective, to put high-density development.

North American freight railways do a brilliant job of the main thing that matters for freight, which is low cost per tonne. Freight shippers generally don’t care if it takes a little longer, whether it arrives a few hours earlier or later in the day, or how comfortable the ride is. They want it to get there cheaply. The way to do that is to pack as much cargo into each train as possible. North American freight trains make European freight trains look like toys. A typical European freight train is about 1,500 tonnes and 750 metres long. North American freight trains can be well over 20,000 tonnes and four kilometres. On an intermodal train, it works out to nearly ten times as many containers per train as in Europe. Technology allows the crew in front to control locomotives distributed throughout the train so that the couplers at the front aren’t pulling the train’s full weight. North American freight railways have undertaken massive reconstruction of their network over the past decades to increase clearances so that trains on most main lines can handle two containers stacked on top of one another. All of this means two people in a locomotive cab can move an incredible amount of cargo. Since labour is often the largest cost or close to it, that’s an immense efficiency difference.

Trains usually leave once a day between any two points and, in many cases, far less frequently. Though there’s much talk about precision scheduled railroading in North American freight, that — ironically — doesn’t mean a precise schedule to the minute like a passenger would expect. Freight tracks don’t need high speeds, and their signalling systems don’t need to run trains very often. Tunnels are often built without complex ventilation, because it’s fine to just wait an hour or two for the diesel exhaust to clear before the next train passes.

The North American freight network is also genuinely integrated across the continent. Every day, trains run right through from Mexico to Canada without needing to swap locomotives or follow significantly different rules. That’s very different from Europe, where cross-border freight has to contend with different signalling, different electrification, and often different regulations altogether. It doesn’t mean there’s no cross-border freight traffic, but it’s significantly more complex and that makes it less competitive. Oil bound for southern Germany from the Middle East takes the most direct route through the Suez Canal to Trieste and then travels the rest of the way by pipeline. But most Germany-bound container traffic goes all the way across the Mediterranean, around Iberia, through the English Channel, and up to the North Sea ports at Antwerp, Hamburg, and Rotterdam, where it can finally be transferred to rail or barge. Freight trains from Italy to Germany would have to cross the Alps, which is a legitimate but hardly insurmountable barrier. Huge amounts of freight cross the Rockies every day.

Subscribe for free, for infrastories from around the world.

Subscribe now

Even the massive new tunnels being planned and built under the Alps to ostensibly solve the problem are not always working as part of an integrated freight network. Switzerland’s immense Loetschberg Base Tunnel, for example, for years carried huge numbers of trucks that get on trains at the Italian border, ride trains under the mountains, and then get back on the road once they reach Germany. Though it’s gradually getting better with more or the traffic on traditional intermodal trains, that’s a tell-tale sign of a poorly integrated freight network.

Rather than length and weight, successful passenger service needs frequency, speed, and reliability. Greater weight per axle improves freight efficiency but also significantly increases wear on tracks, which hurts speed and comfort. Inconsistent freight schedules don’t mix well with passenger trains running every few minutes. And high freight-train weights have long prompted regulators to impose safety standards that also require passenger equipment in North America to be far heavier and therefore less efficient than in Europe and Asia, though this is finally starting to change. We design passenger trains like a 1950s Chevy, where sheer mass is the main safety feature, while European rolling stock uses far more advanced technology like the crumple zones that are used on a much lighter modern Toyota.

North American signalling, designed around freight needs, can afford to keep trains widely spaced because almost no freight line runs more than a few trains per hour. In Europe, or on a North American subway that’s not part of the national rail network, a double-track line can routinely carry 24 trains every hour in each direction. In North America, the same two rails 1,435 mm apart would be considered aggressively used if they handled four per hour, aside from a handful of exceptions like the Hudson River tunnels where freight traffic is excluded.

Freight standards can also massively increase cost in an urban environment relative to what might be possible using European passenger rail standards. In general, a conventional North American railway is built for 1% grades, while modern electric passenger trains can easily handle 3%. That means that raising a track seven metres to clear a road or another rail line needs about 1,400 metres of construction, rather than potentially only about 467 metres. That’s a lot of extra money and disruption, and in a city it often means closing off street crossings entirely.

High locomotive weights compound the problem. In North America, even lightweight passenger cars are generally pulled by a locomotive that is as heavy on each axle as one designed for hauling freight over the Rockies. They typically are about 30,000 kg per axle, while a comparable European locomotive is about 20,000 kg. This means bridges and other structures have to be built far heavier than they’d otherwise need to be.

The solution here isn’t kicking freight off major parts of the rail network or somehow trying to make it copy the far less successful models in Europe. But we also can’t keep running our passenger railways, especially in dense urban and intercity corridors, like they’re freight trains with chairs.

Meanwhile, in Europe, even some new high-speed railways are built to accommodate freight, requiring freight trains to move at far higher speeds than is most efficient while also forcing the engineering of the line to be more expensive than it would be otherwise. They are making basically no serious efforts to permit double-stacked containers or longer trains. Even a dedicated new freight-only railway has far lower weight limits than North America and won’t allow double stacks. In most cases, freight cars are still manually coupled using a chain, rather than the semi-automated Janney couplers that North American railways have been using for the past century-and-a-half. In other words, they’re trying to run passenger trains with cargo instead of people.

The problem may fundamentally be that we force two fundamentally incompatible users together on the same network and wonder why one always comes out the loser. In a future post, I’ll write about what a better way might look like.

^[1] Statistics Canada. Table 23-10-0057-01 Railway industry summary statistics on freight and passenger transportation
Passenger-kilometres per capita equals total passenger-kilometres divided by population. One passenger-kilometre represents one passenger transported one kilometre

The US military is famously the land of the $435 hammer and $640 toilet seat. Those infamous purchases were uncovered back in the 80s, in the “Pentagon Wars” era of relentless agitation for reform from activists within the Defense department, the philosopher-colonel “Genghis John” Boyd most famous among them. But despite their best efforts, there is still a lot going wrong.

There are a few ways that things most often seem to go awry. An old one, dating back to the 60s, is the multirole trap. Robert McNamara famously tried to get one aircraft to work simultaneously as a bomber, a fighter, and a carrier-based interceptor. The F-111 ended up an okay bomber at an extremely high cost, but was entirely unusable for the other two jobs. “There isn’t enough thrust in all Christendom to make a Navy fighter out of that airplane,” said one disgusted admiral. The Navy eventually ended up with the F-14 made famous in Top Gun, while the USAF astonishingly didn’t have a conventional air-superiority fighter at all for the first years of the Vietnam War until they adopted the Navy’s F-4s.

It seems like having one piece of equipment perform many jobs would save money, and sometimes it works, but often you end up with something so complex that it costs a fortune. If you need scissors, a can opener, and a corkscrew for your home, you’ll get much better ones, and probably for less money, if you buy them separately rather than buying a Swiss Army Knife.

Then there’s the issue of designing around a piece of equipment that hasn’t been invented yet. It can be a great way to advance the general state of technology, but it’s usually not cheap, and sometimes it doesn’t work at all.

Most recently, we’ve seen the problem of buying an off-the-shelf design and then “Americanizing” it in ways that seem completely reasonable in isolation, but that combine to make the thing nonsensical or even unbuildable.

As the Cold War ended, the backbone of the US surface fleet, responsible for escorting all the carriers and most other tasks, was the Ticonderoga-class cruisers dating to the early 80s and based on a 70s hull design, along with the Arleigh Burke destroyers, which were first launched in 1989. They’re still the backbone today, though the Ticos are now being retired basically without replacement.

Over decades, the Navy has had program after program to replace them. And every one of them has failed. Finally, in despair as they saw the huge numbers of destroyers being churned out of Chinese yards, they decided to simply start building more Burkes. At least they knew how to do that.

First, the CG(X). In 2001, the Navy launched a program to replace the Ticonderogas with a next-generation cruiser. The CG(X) grew, unsurprisingly, into a 20,000-plus-ton nuclear-powered behemoth carrying an enormous radar and designed around ballistic missile defence. Nineteen were planned and zero were built. The whole project was cancelled in 2010 due to unaffordability, and the Navy fell back on more Burkes.

And then there was DD(X). The Navy actually managed to build a few of those. The Zumwalt class was supposed to be a super-destroyer with a radar-evading tumblehome hull and a giant gun ostensibly to replace the Iowa class battleships in the shore bombardment role. While 32 were planned, three were built, at almost $8 billion apiece. The DD(X) had a pair of 155mm Advanced Gun Systems firing a custom guided round. The cost of that ammunition rose to somewhere between $800,000 and $1,000,000 per shell, so the Navy simply stopped buying the ammunition in 2016 and the guns sat on the ships for years with nothing to fire. They are now being physically torn out to make room for hypersonic missiles that can’t be reloaded at sea, which the Navy hopes will finally give the ships some kind of purpose.

One of the most curious fiascos was the Littoral Combat Ship. The Navy, not unreasonably, decided in the early 2000s that the wars of the future would be fought in shallow coastal waters against small boats and diesel submarines. They needed a fast, cheap ship and decided that they would swap out modules based on the mission they wanted the ship to do (the multirole dream again!), from anti-submarine to minesweeping to surface warfare.

It didn’t work out as planned. They developed two separate classes, one of which was especially unsuccessful. The modules took forever to get right, and the anti-submarine one was cancelled outright. A 2013 Pentagon report judged both variants as not survivable in actual combat, which is not a great trait for a warship meant to operate close to hostile shores. The promised price of about $220 million a hull became $500–600 million. The Navy was so thrilled with them that they started retiring them not long after they were completed, because the upgrades to make them useful cost more than the ships were worth. One ship didn’t even make it past the five-year mark. They commissioned the last Littoral Combat Ship in 2026, having already retired seven of them.

Subscribe for free, for infrastories from around the world.

Subscribe now

The Ford-class carrier was built on the proven Nimitz hull, but it still managed to go billions over budget and enter service years late. The Navy took a proven design and then stuffed it full of unproven technology. They replaced the proven steam catapults with electromagnetic ones and brought in weapons elevators that took years after delivery to get working, along with replacing other assorted technologies. They shovelled enough money at the problem that they finally managed to get the ship mostly functioning.

Now, the Navy has started a new DDG(X) program to finally replace the Ticos and older Burkes. It has been in development for years, but the Trump Administration decided it wasn’t exciting enough and is trying to replace it with the Trump-class “battleships,” which will clearly go well. The press release features a sci-fi railgun, an electromagnetic cannon that can fire at Mach 6. The Navy already spent 15 years trying unsuccessfully to develop a railgun before finally giving up in 2021.

The response to this litany of failures was meant to be the Constellation-class frigates. The Littoral Combat Ships clearly weren’t working, and a big destroyer with nearly a hundred missiles was clearly overkill for many of the basic patrol missions the Navy is regularly asked to undertake. Recognizing that developing a ship class from scratch was doomed to failure, they decided to play it safe. The US has many allies with navies that built perfectly good frigates, designed for NATO compatibility. They could just buy an existing design and skip all the development steps.

After running a competition, they settled on the Franco-Italian FREMM design. The plan, as stated to Congress, was for the new ship to keep about 85% of the parent design. Then the Navy got to work. They decided it needed an American radar, which meant a new superstructure. It needed an American combat system, American missiles, American machinery. The survivability standards for French and Italian sailors were apparently not good enough. There also needed to be some margin for future growth. Each change seemed reasonable on its face, but by the time they were done, the commonality with the original design was down to about 15%. The ships had also gained so much weight that they might even struggle to keep up with the carriers they were intended to escort. Still, the yard started building the first one. By April 2025, it was about 10% complete. Seven months later, when the program was cancelled, they had made it up to 12%. At that rate, it could be ready for launch by 2050.

They had taken a ship that worked and was in service with allied fleets, and they redesigned it into something basically unbuildable. The Navy’s conclusion, unsurprisingly, was that they should never again try to use someone else’s design. Obviously the endless changes were all completely unavoidable and the perfectly good design could not possibly have just been built as planned.

So, now they’ve decided to replace the Constellation with a new frigate program. This time it is going to be based on the Coast Guard’s Legend-class cutter, so it’s at least a red-white-and-blue American design. But we’ll see how it actually goes once they start “refining” the design.

It’s not just an American issue, of course. Canada also went shopping for a new class of surface vessels and settled on Britain’s capable Type 26, a relatively compact anti-submarine warship. Then, Canada bolted on a very large American air-defence radar and the Aegis combat system to turn it into a sophisticated air defence destroyer. While the radar is extremely impressive, the ship doesn’t actually carry many missiles for the radar to control. The River Class has room for only 24 vertical launch cells able to accommodate longer-range surface-to-air missiles, while Burkes and other large air defence destroyers have two or even four times that number. Fifteen of them are expected to cost nearly C$80 billion.

One might wonder if this could be the old multirole problem again. For a smaller navy, it might make sense to have one ship class do everything. But it might also have ended up cheaper to buy a small, proven class of dedicated air defence destroyers with the advanced systems of the River Class and a hull designed to handle them, alongside a proven class of frigates like the original Type 26, optimized for Canada’s traditional antisubmarine warfare role. It matches the old Canadian mixed fleet of a small number of Iroquois-class destroyers upgraded for air defence, and a larger number of Halifax-class patrol frigates well-equipped for anti-submarine work.

With the DDG(X) stuck on paper and the Trump class little more than a fantasy, the future of the US Navy for at least the next decade will almost certainly be the Miami-Vice-era Arleigh Burkes.

If you haven’t had enough of the joy of defence procurement, I strongly recommend the movie Pentagon Wars. It’s appropriately a comedy.

There were only four decades between the flimsy propeller planes of the 1930s and the Concorde in the 1970s. In that time, airports massively changed form, evolving from, in many cases, simple grassy fields to the massive cities unto themselves with kilometres-long runways that we are familiar with today. Major metropolises have generally had several generations of main airports. In 1940, London had Croydon, Paris had Le Bourget, and New York had LaGuardia. The latter survived, but all were inadequate to meet the demands of postwar growth, so the next generation were built, at Heathrow, Orly, and Idlewild (later JFK), respectively. By the 1960s, aviation was in the jet age and big cities had huge jets flying overhead every few minutes.

Those jets were loud. If you watch a video of a 707 or DC-8 taking off in the 60s, you still won’t fully understand how loud they roared, spewing heavy black smoke. Pretty much the only time you can encounter plane noise like that today is if you happen to find yourself under a B-52, an old Soviet aircraft, or the handful of classic 60s-era jets still flying. The large jets of today really are remarkably quieter than earlier generations. The noise of the 707s, however, paled in comparison to what everyone thought was coming next.

From the small, unpressurized DC-3s and flying boats of the 30s, to the high-flying pressurized airliners of the 1940s, to the jets of the 1950s, aviation technology was surging ahead. By the 60s, military fighters were already routinely flying supersonic, and everyone assumed that passenger planes soon would be too. Britain and France cooperated on Concorde. The US government lavishly supported the SST program, prompting Seattle to name its new basketball team the SuperSonics, until funding was cut in 1971, throwing the city into a huge recession. The Soviet Union actually beat both to first flight with the Tu-144.

These new planes were astonishingly loud, and not just when they were flying at top speed blasting everyone with sonic booms. Right up to its retirement, Concorde’s takeoff was infamous for setting off car alarms all around the airport. The battle against aviation was one of the first big battles of the modern environmental movement, right up there with opposition to dams and urban expressways.

In an era of supersonic planes, the old model of houses just metres off the end of runways, like at Heathrow or JFK or LAX, was no longer tenable. Instead, modern “jetports”, as they were often dubbed, would be built well outside the city, with huge land areas bought up to serve as permanent buffers. Out of this came a whole generation of new mega-airports. Some of them were built, but many never made it off the drawing board. The most interesting, perhaps, were the ones where they started but never finished. This article is meant to be a bit of a high-level overview and, at some point, I plan to write about several of these in a bit more detail.

Subscribe for free, for infrastories from around the world.

Subscribe now

First, the successes. The biggest airport developed in Europe in that era is Charles de Gaulle, which has served as Paris’ main international airport ever since. Dallas-Fort Worth International’s immense size has allowed it to become one of the top US hubs. Washington’s Dulles is a successful international airport and just about the only airport in the Northeast with ample room for expansion, but it never quite reached its potential because National Airport continued to skim off all the premium domestic traffic. Tokyo built Narita, though heated disputes with landowners have left it heavily compromised.

There were many more airports that stalled at the concept stage. British officials were already concerned about the limitations of Heathrow in the early 70s. Their solution was a new airport in the estuary of the Thames built on landfill at Maplin Sands. It’s a scheme that is periodically revived in various nearby locations, most recently as “Boris Island.” The Port Authority of New York and New Jersey looked at the Great Swamp in New Jersey to build a jetport for the New York area. It sparked passionate environmental opposition, eventually resulting in the designation of the Great Swamp as a protected area for wildlife.

The main limitation of these jetports was their sheer distance to their primary cities. Where strong rail links were built, as they eventually were at Narita, they tended to be reasonably successful. But when the ground access infrastructure never really arrived, the airports generally never really ‘took off.’

This leads us to the third category of airports, which are the ones that failed to become major airports even though land was acquired and, in some cases, construction even began. Perhaps the most infamous is Mirabel Airport in Montreal. The airport did end up operating as the city’s main international airport for several decades, despite the failure to build the promised rail link. Unfortunately for Montreal, however, domestic flights remained at the much accessible Dorval airport, which spelled the end of the city as Canada’s main hub. Connecting traffic shifted to Toronto Pearson, where passengers could transfer from overseas to domestic destinations at a single airport. Eventually Montreal’s international flights shifted back to Dorval and Montreal ended up back where it started, with Mirabel as an industrial airport and manufacturing centre for the Airbus A220 (formerly Bombardier C-Series). Toronto had its own jetport plan, with land acquired at Pickering to the northeast of the city, but the airport was never built and most of the land was eventually turned into park and an agricultural preserve, with the last of it now being considered for more parkland.

Los Angeles World Airports acquired an immense site in the desert north of Los Angeles, near Palmdale, for a new airport, though they never ended up needing to replace LAX and instead the region ended up served by several small reliever airports at Burbank, Ontario, and Santa Ana.

Maybe the strangest story is the Everglades Jetport. As tourism to South Florida was booming in the 60s, planners proposed an immense airport, tailor-made for supersonic aircraft, in the middle of the Everglades. Tens of thousands of acres were acquired for the airport, which was meant to serve both Miami and the cities of the Gulf Coast with strong highway and rail connections. But all that was built was a single runway before the project was abandoned. Ultimately, the mega airport proved unnecessary to meet even the massively growing demand in south Florida. As most of the air traffic was destined to Florida itself, a single massive hub was not as needed. Additional airports like Fort Lauderdale and later Southwest Florida took much of the load from Miami. The intended mega-hub became a sleepy airport mostly used for training, and in a bizarre twist, the Florida government recently took advantage of its isolated location to temporarily turn it into an immigration detention centre nicknamed Alligator Alcatraz.

In the end, the supersonic era never really came. Planes also got much quieter, and airport planners and air traffic controllers got better at squeezing more, and bigger, planes into the older generation of airports. Still, the lost jetports are a fascinating what-might-have-been in aviation history.

That decision quietly rearranged the economic geography of the United States. Until then, infrastructure was based largely on existing demand, with notable exceptions. But infrastructure built to serve existing demand tends to reinforce existing patterns, and it can lock entire regions into a low equilibrium where demand stays low because the infrastructure is poor, and the infrastructure stays poor because demand is low. This, in turn, can lock places into relative poverty and underdevelopment. Breaking that trap requires building ahead of demand, to a standard the current traffic does not justify. The Interstate program is a strong example of this.

To see what makes the Interstates unusual, we must look at the previous generation of dominant transportation infrastructure. The American railway boom was certainly shaped by politics, with land grants, municipal subsidies, and various other government programs. But fundamentally, railways were private businesses that built lines where the traffic existed and improved their lines in proportion to the tonnage and passengers using them. The result was a very uneven network: The Northeast and Midwest, dense with factories, mines, and mills, were criss-crossed with railways of an exceptionally high standard, while other parts of the country, like the South, made do with something far more limited.

The difference was as much about quality as quantity. The Pennsylvania Railroad, which exaggerated only mildly when it dubbed itself the Standard Railroad of the World, built a magnificent, mostly four-track corridor between New York and Pittsburgh on alignments engineered with shallow grades, generous curves, heavy rail, extensive grade separation, and even partial electrification because the sheer volume of passenger, steel, coal, and merchandise traffic justified engineering on that scale. A factory in Pittsburgh or Philadelphia sat on some of the finest freight infrastructure ever constructed, and those cities became major industrial centres in large part because of it. The corridors of the South, by contrast, were predominantly single-track lines, more lightly-built and steeply-graded, serving a region whose traffic largely didn’t warrant much more. For decades, the Southern railways even ran on a different track gauge from the rest of the country, meaning that freight had to be transloaded at the borders of the region, until the southern lines converted thousands of miles to standard gauge in a single coordinated effort over two days in 1886. Even after the gauges matched, the quality gap endured.

The South’s railways were limited because the region had few large rail-dependent industries, and the region had few large rail-dependent industries in part because its railways were slower, less reliable, and more expensive to use, with less competition. A manufacturer comparing Charlotte with Cleveland was weighing winding single-track against multi-track corridors owned by several major competing railways. Infrastructure was hardly the only force at work, since politics, governance, education, and capital markets all played their roles, and these things are deeply interrelated. But the absence of industrialization and the broad rural poverty that resulted were defining characteristics of the prewar South that Franklin Roosevelt and the New Deal sought to overcome. It wasn’t just transportation: Much of the rural South also lacked electricity because the market was considered too limited to bother serving. This led to the creation of the federal Tennessee Valley Authority, along with the Rural Electrification Administration, to get power lines running to every house and farm.

It is worth understanding that the migration out of the South was not only the Great Migration of Black Southerners fleeing the extreme legal discrimination of the Jim Crow segregation system. White Southerners not facing discrimination also poured into the industrial cities of the North in comparable numbers, filling the factories of Detroit and Chicago, as historians like Thomas Sugrue and James Gregory have shown. The South was stuck in an earlier economic era where finding opportunity usually required moving elsewhere.

A Network to Serve the Whole Country

The Interstate program deliberately broke this pattern. President Dwight Eisenhower secured passage of the Federal-Aid Highway Act in 1956, authorizing a 41,000-mile network of expressways with the fig leaf that these roads were being built largely for defence purposes. The network was planned as a complete national system, connecting every significant urban area and reaching every state in the continental U.S., with routes laid out to form a coherent grid. Ninety percent of the cost would be covered by the federal government, meaning that these expressways could be afforded by every state, whether rich or poor. And every mile was built to uniform national design standards, developed through AASHTO and enforced as a condition of the federal funding share. The Interstates required full limited access, meaning no connecting driveways, a minimum of two travel lanes in each direction, twelve-foot lanes, engineered shoulders, grade separation throughout, and design speeds of 50 to 70 miles per hour. A rural Interstate in Mississippi was built to the same geometric standard as one in New Jersey. Other places like the UK or Canada often contented themselves with “good enough” infrastructure, building short bypasses and upgrades where existing roads ran through towns or were unusually congested. But on the Interstate system, you could drive from Maine to Florida or New York to California without hitting a single traffic light.

It was a radical choice, emerging from the universal service provision ideals of the New Deal and fundamentally different from anything that had been built before. No private railroad would have double-tracked across sparsely settled Alabama, because the existing single tracks were far from overloaded. And the federal government had never sought to guarantee, let alone fund, very high-quality railways in every part of the country. The Interstate program built the equivalent of well-graded double track everywhere, financed through a national trust fund, on the basis that every region of the country should have equal access to high-quality transportation infrastructure.

The Rise of the Sunbelt

It took a generation, but the upshot was the inversion of a century of American economic geography. By the 1980s, the fastest-growing cities in the country were overwhelmingly in the Sunbelt, and above all in the South, the same region that had been hemorrhaging population within living memory. This change was not only driven by transportation infrastructure. Air conditioning made life in warm places much more pleasant, Texas oil produced a series of economic booms, firms chased cheaper labour and lighter regulation, and the dismantling of legally-mandated segregation removed a barrier that had deterred both investment and migration. But it likely never would have been possible if Pennsylvanians and New Yorkers were driving on fast turnpikes while Southerners trundled through every town on winding country highways. For the first time in American history, the South was at least as well connected as the North.

By the 1980s, the factory in Pittsburgh and the factory in Charlotte were now both served by Interstate highways built to identical standards. If anything, the advantage had reversed, because the Southern Interstates were newer and generally less congested, with fewer urban bottlenecks. The infrastructure deficit that had made a Southern location a non-starter for a freight-intensive business had been eliminated. Location decisions that had been foreclosed for a century were suddenly open, and companies took the opening. The South industrialized along the Interstates. Auto plants are now nearly as common in Tennessee or the Carolinas as they are in Michigan or Ohio. These days, you are much more likely to meet an ex-Detroiter in Charlotte than an ex-Georgian in Michigan.

The classic American freight geography ran east and west, from the industrial Midwest to the Northeastern seaboard, along the corridors of the great Pennsylvania and the New York Central Railroads. Today, two of the busiest trucking corridors in the country are Interstate 81 and Interstate 75, running north and south along the Appalachians and down through the manufacturing belt of Kentucky, Tennessee, and Georgia. On stretches of I-81, trucks make up as much as 40% of all traffic, and the corridor carries more freight than the parallel and far more famous I-95. The region along those corridors had been an industrial backwater for generations, and the railroads typically served them with poor (if spectacular) infrastructure like the Saluda Grade. No reasonable transportation planner in 1956 would have expected the Shenandoah Valley to become one of the major freight corridors of the country.

Induced Demand Has Many Forms

Infrastructure investment that follows demand is an approach that sustains the existing distribution of economic activity, and compounded over decades it hardens that distribution into something that looks like inevitability. The railway network served demand effectively, and in so doing helped freeze the South out of industrial America for the better part of a century. The interstate network ignored demand in favour of a uniform national standard, and suddenly vast new patterns of demand emerged far from where anyone would have expected them. We can always forecast a certain amount of uplift from a given investment, but no model can capture the kind of transformative region-wide effects that the interstate network produced. As a result, any rational evaluation of the interstate network would have had people expressing bafflement over the crazy idea of superhighways to rural Mississippi or southern Texas or a dusty town in the Arizona desert like Phoenix. The planners of 1956 could not have included the Sunbelt boom in their traffic models, because the boom was a transformation without precedent that the network itself would help create.

None of this is an argument for building bridges to nowhere, and the interstate program has had tremendous damaging effects—most importantly what its urban segments did to the neighbourhoods in their path, but also its role in damaging the more efficient and environmentally sustainable rail network. Of course, those urban segments were not initially part of the original ideal of smooth intercity transportation (and the highway networks of France or Germany are much closer to the ideal of expressways for intercity travel), but serving the highest-demand corridors in cities with infrastructure 90% paid for by the federal government was irresistible to the states. Transit and intercity railways, by contrast, got minimal federal funding until the 1970s.

Despite its flaws, the interstate system makes clear that a complete network, built to a uniform high standard everywhere, can significantly level the economic playing field and massively reshape patterns of demand. The next time someone argues that a corridor lacks the demand to justify the investment, it is worth asking whether the causation runs the other way, and thinking about whether the demand is absent because the corridor lacks decent infrastructure. The long stagnation and recent boom of the South suggest this should be kept in mind.

By the early 1970s the private railroads of the Northeast had collapsed into bankruptcy, the tracks between Washington and Boston were crumbling, resulting in trains that ran late more often than not. The Northeast Corridor, especially the segment between New York and Washington, had been a showpiece built with lavish infrastructure by the Pennsylvania Railroad, the self-proclaimed Standard Railroad of the World whose vast scale made it one of the biggest private corporations on earth. But by the 1970s, the state-of-the-art infrastructure of the 1910s to 1930s was crumbling and antiquated. Passenger rail simply was not profitable anymore and the whole enterprise of American passenger rail had been handed to a brand-new government corporation, known as Amtrak, that almost nobody expected to survive, let alone thrive. Then the 1973 oil crisis hit and suddenly electrified trains that needed no imported petroleum looked less like museum relics and more like strategic assets. A long-debated modernization that had been drifting acquired sudden urgency. The corridor carried nearly a thousand trains a day despite its decay, most of them commuters who had no other reasonable way to get to their downtown jobs, and the Northeast’s large congressional delegation made sure something got underway despite scepticism in the White House.

The corridor was in terrible shape. The electric Metroliners that ran from New York to Washington were meant to be a modern showpiece, yet a reporter noted that passengers “find it difficult to walk in the aisles or even read a book or drink a cup of coffee.” The diesel Turbotrain from New York to Boston fared worse still, described by Christopher Lydon in the New York Times as “a 170-mile-an-hour train, straining most of the way under a 75-mile-an-hour track limit,” a machine that “matches airline decor with railroad grime” while running on jet-engine power with an engineer in a steam-engine smock. By 1976 the Metroliners kept their schedule only forty percent of the time, which was hardly mysterious given that more than 180 miles of the corridor sat under slow orders. Something clearly had to be done.

Management by Committee

The first sign of trouble was structural, and it emerged before a single tie had been replaced. Rather than vest responsibility in a single competent authority, the project was carved up among three organizations that did not get along and were not especially well-suited to their assignments. The Federal Railroad Administration, an executive-branch body, held the purse strings and the oversight power; an engineering firm, DeLeuw Cather/Parsons (DCP), was hired to do the actual planning and design; and Amtrak, all of four years old and with no experience whatsoever in major construction, was named the supervisor of construction. The subtext was that the FRA was the executive branch’s way of keeping control of Amtrak, which was seen as a “congressional plaything”. The House Committee on Government Operations dubbed the arrangement the “three-headed monster”.

The FRA, tasked with overseeing a major railroad infrastructure project, employed only eight people with any railroad experience out of seventy-five, and had to borrow highway engineers from the Federal Highway Administration to make up the difference. Amtrak had ballooned from a few hundred employees to more than fifteen hundred in a single fiscal year, and most of the new arrivals had never worked on a railroad either, while many of those it inherited from the bankrupt private lines had spent their careers overseeing the contraction of the network rather than its expansion. They were, as one report put it, “lacking in the skills needed for undertaking large construction projects.” Operating a railway is not the same as building a new one, and nobody had done a major intercity passenger rail project in the United States for nearly four decades, meaning everyone that had worked on one was long since retired.

Subscribe for free, for infrastories from around the world.

Subscribe now

One Amtrak official offered the assessment that “there is no such thing as job continuity in the NEC.” Over a single four-month stretch, the railroad managed to burn through four different officials responsible for coordinating its bridge program with the other two heads of the monster. DCP, the group that actually knew how to run a large construction project, was given no authority to direct anyone. When its engineers watched Amtrak’s crews doing work incorrectly, all they could do was file a report with the FRA, which might eventually order Amtrak to fix it. The competent party could observe the mistakes but not prevent them, which is a very serious flaw for effective project delivery.

The master plan that was supposed to guide the entire project still had not been finalized or approved two full years into construction. The absence of a plan produced exactly the cascade of waste you would predict. Crews were told, in the words of the GAO, to “surface and align track without being given final alignments and elevation specifications,” which meant they laid track without a plan and then had to tear much of it out and do it again once someone finally decided what the plan was. Forty percent of a sampled batch of designs turned out to be useless because the requirements kept changing. In one representative episode, the engineers spent nearly a year drawing up plans to replace a bridge, at which point Amtrak decided to rehabilitate the bridge instead. They bought five million dollars of hopper cars and materials they never used, at a time when five million dollars was real money, because they didn’t have a clear plan for what they would need. They leased equipment on the assumption the project would finish in 1981 even though Amtrak already knew, at the moment of leasing, that it would not finish before 1983 at the earliest and therefore purchase would have been the more economical choice. A March 1979 report found that ninety-eight percent of the project’s components had fallen behind their original schedules.

While no one at the top could produce a plan for the actual railroad, the FRA fixated on minutiae in its oversight role, forcing Amtrak to extensively justify, in a billion-dollar project, the “purchase of some padlocks, supplies, and $2,500 in equipment rentals.” The requests often went unanswered or were answered late, consuming vast staff time, and after all that scrutiny not a single one of these minor expenses was ever actually rejected. The FRA, with its lack of rail expertise, did not know how to oversee the planning of a railroad, but it did know how to fight over padlocks.

Amtrak paid above-market rates on every one of the ten equipment leases the FRA bothered to examine, leasing pickup trucks for more than $6,400 apiece when they could have been bought outright for around $5,500. Two million dollars of concrete ties failed their tests. Most of the $800,000 price of a track geometry car was handed over even though the car arrived without all the equipment it was supposed to contain. A three-million-dollar contract with Arthur Andersen to build management systems for the project was cancelled after much of the money had been spent but before the systems could actually be installed.

The relationship between Amtrak and DCP frequently descended into time-consuming and petty feuding. DCP sent a proposed, supposedly jointly developed, 3-year track program to Amtrak for comment in December 1977, only to meet with no response. Amtrak instead sent a different schedule to the FRA that had not been coordinated with DCP. Work on interlockings, which was intended to be over 20 percent of the planned 1978 track program, wasn’t started until August of that year because Amtrak and DCP could not agree on work plans. Unsurprisingly, only two of five planned interlockings were completed that year.

Chaotic Management Doesn’t Come Cheap

The chaos was costly and unsurprisingly resulted in the traditional remedy of descoping the project. Management alone consumed $127 million, roughly a third of the entire budget and about three times what the GAO thought the job required, though the management was clearly not particularly effective. And because the money spent on confusion was money not available for infrastructure, the escalating costs forced round after round of cuts to the elements the project had actually been created to accomplish.

Many of the most important goals of the project ended up dropped. Elimination of $41 million in work to improve the track roadbed to make the rails more stable increased future maintenance costs, made it difficult to run at high speeds, and reduced passenger comfort. The modernization of the ancient New York–Washington electrification system, so antiquated that it used (and still uses) 25 hz equipment incompatible with the North American power grid and 1930s catenary that limits speed, was abandoned. The electrification to Boston was also dropped, not to be completed for several more decades. The Baltimore B&P tunnel, dug during the Grant administration and identified by the project’s own engineers as the single largest bottleneck on the corridor, were so waterlogged that their seepage had dropped the local water table by ten to twenty feet. In fact, it had been called obsolete in the 1930s. Its replacement, however, was scrapped at the time and is only now under construction.

The project ended up a bit like renovating the living room while the foundation crumbles. It focused on the headline travel-time improvements that had been used to sell the program in the first place, while the fundamental state-of-good-repair work was deferred into the future.

FRA conceded, even in the official history, that the corridor’s decay had grown so severe that its two goals had become “paradoxically contradictory.” Trip times could be reduced only by neglecting essential state-of-good-repair work, and the state-of-good-repair work could be afforded only by abandoning the trip-time improvements. A high-speed railway built on an unstable foundation and fed by an antique electrical system is fundamentally limited.

Meanwhile, in France

While Amtrak was struggling to figure out how to modernize its railway, France was building the TGV. SNCF was many times larger than Amtrak and had a cohort of experienced engineers who had been doing major double tracking and electrification projects for decades. France had a government that committed to finishing what it started, and a plan finalized before the digging began. The French line opened in 1981, close to schedule, and ran from Paris to Lyon at consistent high speeds far greater than Amtrak could achieve. The New York Times described the project as “much more ambitious” than its American counterpart, noting that it is “close to schedule,” an even starker contrast. “When New York-Washington trains are averaging 84 miles an hour (including five stops),” it lamented, “the French plan to be making the new 262-mile Paris-Lyon run in two hours flat, an average of 131 miles an hour, with top speeds of 160.”^[1] When American reporters asked French engineers to comment on the NECIP, the engineers were frankly baffled that anyone would attempt to run heavy freight that pounds the rails, slow commuter traffic, and high-speed expresses all on the same track. And yet, despite everything, an Amtrak vice president stood up upon the project’s completion and pronounced the upgraded corridor “the state-of-the-art railroad in the world,” a claim that would have startled anyone who had recently ridden the actual state of the art at 270 kilometers per hour through the French countryside.

Worse, France was able to build an entirely new corridor about the length of New York to Washington for the equivalent of about $1.7 billion, in 1981 dollars. The half-delivered modernization of the NEC ended up costing $2.19 billion.

The United States became burdened by arbitrary regulations just as France was abandoning them. Even on the relatively well-built NEC, regulators were reluctant to grant permission for operations at or above 110 mph despite years of operations at that speed by the Pennsylvania Railroad and many operations at higher speeds around the world. Until the end of the Second World War, while American trains were routinely operating at over 160km/h, France had maintained a blanket limit of 120km/h. It is worth noting that these regulations were developed and enforced by the same FRA that had so few staff members with significant railway experience. I’ll be writing more about these regulatory changes in a future post.

Subscribe now

A few decades later, another Northeast Corridor modernization project sought to bring in some of the French technology pioneered on the TGV. However, American safety regulations forced the Acela trainsets to be built at roughly twice the weight of comparable European trains, akin to relying on the brute strength of a 1950s automobile rather than the crumple zones every modern car and foreign train had adopted. The French engineers who worked on the project dubbed the resulting behemoth le cochon.

A Flawed Project with a Legitimate Legacy

The NECIP did, in the end, have some real accomplishments. Most importantly, it arrested decades of decay under the control of the struggling private railroads. It pushed Metroliner punctuality from 53% to 89%. It also lifted the New York-Washington time to two hours and thirty-six minutes (down from three hours), though the 27-minute New York-Boston improvement to 3 hours, 57 minutes was far worse than the goal and neither proved sustainable over the long term. Still, the gap between its ambitions and results, and between its results and those achieved by the French at the same time, is striking. It is a gap explained in large part by the chaotic management of the project.

We often tell ourselves that infrastructure fails for grand structural reasons, that it is about political will or funding levels or national character, and sometimes that’s true. But the NECIP failed for the same the reason a home renovation fails when you start knocking down walls before you have drawings. A successful project needs to be well conceived from the outset, with coherent designs flowing from that conception. This is especially important when you are building with limited experience. Every hour skipped on conception and design, or every choice to postpone making a crucial design decision, returns as ten hours of tearing out track that was laid to a specification that did not yet exist. All too often, the decision to start building before finalizing design comes from an inability or unwillingness to make critical decisions, resulting in a culture of endless deferral amid pressure to show progress. The three-headed monster did not fail because of a lack of money or political will. It failed because they didn’t know the project they wanted to build before they started building.

There is one American city that strikes me as especially well-suited to this approach, and it isn’t one that is usually seen as an ideal environment for transit.

Los Angeles is built on a grid much like Toronto’s. Its suburbs have a population density similar to Toronto’s, which is considerably higher than that of the suburbs of the eastern cities we tend to view as friendlier to transit. A grid of buses is an ideal transit system for a decentralized city like L.A., and the city has already built much of the rapid transit backbone the model calls for. But what Los Angeles has never had, and what turns out to be the most essential element, is the unglamourous bus frequency on which Toronto’s success actually rests.

A number of years ago, I sat on a panel at Chapman University in Orange County, alongside Darrell Johnson, then-director of the Orange County Transportation Authority; Laurie Berman, director of Caltrans (the California Department of Transportation); and Michael Dukakis, the former governor of Massachusetts. I used my time to make the case for the Toronto model and its relevance to Orange County, and I’ll admit the argument met with a fair measure of skepticism from Johnson. The core of my case was that Orange County has far more in common with the successful suburbs of Toronto than its reputation would suggest.

Consider the densities, in people per square mile, of several Toronto suburbs set beside Los Angeles and Santa Ana, one of Orange County’s principal cities:

Santa Ana is denser than any of the Toronto suburbs on this list, Los Angeles’ density sits between Etobicoke’s and North York’s, and Orange County as a whole is a good deal denser than Peel Region, the suburban county that contains major transit success story Brampton. The built form is also more similar than one might expect. While Toronto might have a few more apartment towers, most people still live in low-rise housing on meandering cul-de-sacs, nestled within a grid of wide arterial roads lined with strip malls, office parks, and big-box stores.

Subscribe for free, for infrastories from around the world.

Subscribe now

If density is the primary determinant of transit success, there is no reason why Toronto should have about six times the transit ridership per capita of Los Angeles.

The real determinant of transit success is consistent high frequency, and this is where the gap is very stark. When I spoke at Chapman, Orange County—with a population of over 3 million people—had only two bus routes that cleared the rather modest bar of running every 30 minutes until midnight, seven days a week. In Toronto, virtually everyone in the city lives within walking distance of a bus scheduled at least every ten minutes until late in the evening, and that gap in frequency—far more than anything to do with the built environment—is what separates a city where transit is a genuine option from one where it functions as a last resort.

Los Angeles faces the classic transit chicken-and-egg problem. It has never had transit attractive enough to persuade people to choose it (okay, not in recent memory), and so people have not chosen it, which has in turn supplied the justification for never making it attractive. The experience of Toronto and Brampton is that this loop can be broken, because even committed suburban drivers prove perfectly willing to ride the bus when it comes often enough. What’s striking is how little else it takes. The service needn’t be especially fast or reliable, though improvements on both fronts would obviously be welcome, so long as a rider can count on a bus turning up within a few minutes of arriving at the stop.

This is why Los Angeles strikes me as the ideal place to run the experiment, even though doing so would demand the same kind of rapid, concentrated improvement in service that Brampton pulled off, scaled up to a metropolis many times its size. The task would be to establish a grid of routes with guaranteed ten-minute frequencies along all the major arterial roads, and while that is a genuine undertaking, Los Angeles and California have already shown a willingness to make major investments in transit. Set against the billions in capital funding routinely lavished on rail projects, the operating money needed to blanket the city in frequent service is comparatively modest. California is a wealthy state and can certainly afford it, along with the modest capital expenditure on more buses and garages. Better still, Toronto and Brampton show that good service tends to attract so many new riders that the additional fare revenue holds subsidy levels flat or even drives them down, so that the frequent-service grid becomes far less of a fiscal burden than its up-front cost would imply.

Orange County has, in its own modest way, already demonstrated the principle. Its OC Bus 360 program added service on a handful of selected routes. Sure enough, service improvements brought ridership gains. In 2017, even as the system as a whole shed 2.9% of its ridership, the improved routes gained 10.4%.

If anything, the greatest risk in all of this is a pleasant one. Were Los Angeles to actually boost ridership the way Toronto and Brampton have, its comparatively low-capacity rail network would very quickly find itself swamped by the riders the buses delivered to its stations. That, however, is the kind of problem a transit system should hope to have.

Canada has always been a major player in nuclear energy. Its important role in the Manhattan Project led to the creation of Atomic Energy of Canada Limited, a crown corporation owned by the federal government and charged with developing nuclear energy for civilian purposes. Out of this work came the CANDU reactor, a distinctive design that had some significant advantages over other designs like the Pressurized Water Reactors (PWRs) and Boiling Water Reactors (BWRs) more commonly used in other countries. Most notably, CANDU reactors can be refuelled without being shut down, and they can use natural uranium rather than enriched fuel. But the ability to refuel while running comes from the fuel being housed in pressure tubes. Those pressure tubes are under enormous stress and, over time, can develop cracks.

CANDU reactors have been used around the world, in Argentina, South Korea, India, and Romania. But by far the biggest operator is Ontario, which has built twenty full-sized reactors. Back in 1983, Ontario Hydro discovered the pressure tube cracking issue the hard way when one of them suddenly ruptured, forcing an emergency shutdown. No real harm was done and no radiation was released, but it was a wakeup call that prompted much closer attention to the potential issue. The discovery of significant pressure tube problems as well as a power surplus and other issues in the late 90s led to the sudden idling of eight of Ontario’s twenty reactors.

A few years later, power demand was growing again and it was apparent that more generation capacity would be needed. Bruce, then the largest nuclear plant in the world with eight reactors (four of which had been idled) was leased to Bruce Power, a private company. A decision was taken to begin refurbishing some of the reactors so that they could be restarted and their lives extended for several more decades.

There are genuine advantages to this kind of midlife reconstruction since you emerge with a much more modern plant, but it is an extraordinarily complex process. You are essentially rebuilding the reactor from scratch, in a radioactive environment. And nuclear projects, generally speaking, do not go smoothly. Virtually every nuclear project around the world for decades has come in over budget and behind schedule. Ontario’s first refurbishments were no exception, with significant delays, major cost overruns, and serious questions about whether it made more sense to just build natural gas plants instead. Natural gas plants have high and volatile long-term fuel costs, but they are cheap to build. They feature what are essentially oversized jet engines, making them manufactured products rather than bespoke megaprojects like nuclear power plants.

Subscribe for free, for infrastories from around the world.

Subscribe now

But this is where the story gets interesting. The Canadian nuclear industry did not look at those early failures and say, “These projects are all unique, there’s nothing to learn, just budget more time and money next time.” They did something much harder and far more valuable: they acknowledged that the projects had gone wrong, not as an unavoidable fate, but as something that could be understood and corrected. And they asked how it could be done better.

One of the most consequential things they did was build a full-scale physical model of the reactor. Construction crews, engineers, and technicians practiced on it over and over until the procedures were essentially muscle memory. When it came time to work on the actual reactor, which is of course radioactive (meaning you want to minimize time spent inside it), they could carry out the work precisely and efficiently. They applied this same mindset across many dimensions of the work with systematic analysis of what had gone wrong before and deliberate redesign around those lessons.

The results speak for themselves. At Darlington, Ontario Power Generation and its partners were able to completely rebuild four reactors, in a radioactive environment, for $12.8 billion dollars. They also managed to do it four months ahead of schedule and $150 million under budget. In the nuclear world, this is absolutely unheard of. France’s new EPR reactor at Flamanville cost over C$21 billion and was about twelve years late and four times the original budget, while the two new AP1000 units at Georgia’s Plant Vogtle cost close to C$50 billion and were about seven years late and 2.5 times over budget. Even China’s new European Pressurized Reactors (EPRs) and AP1000s were about four years late and far over budget.

Obviously rebuilding an existing reactor is in some ways less complex than building a new plant from the ground up. On the other hand, it is far more complex since everything is radioactive! Canada is now arguably the only place on earth that can deliver major nuclear projects on time and on budget. If Canadian nuclear workers can rebuild a reactor—replacing pressure tubes, steam generators, and countless other components—in a radioactive environment, it seems reasonable to ask whether those same people could build a new reactor in a non-radioactive greenfield environment. We won’t know until we try, but if new CANDU reactors could be built anywhere near the cost-per-unit implied by the Darlington refurbishments, recognizing the additional work required for the rest of the plant, it would represent a dramatically more affordable source of electricity than the international nuclear alternatives and, arguably, many other sources of baseload power.

Because of the sheer number of reactors it has been refurbishing, Canada has done more major nuclear construction work than almost any jurisdiction on earth over the past fifteen years. It has built real expertise, institutional knowledge, and a supply chain, and it has proven that it can do huge amounts of construction work that can meet the incredibly exacting standards of nuclear regulators. Shoddy construction work was responsible for many of the delays and overruns on the nuclear projects in France, Georgia, and elsewhere. People who have rebuilt sixteen reactors in a radioactive environment might well be able to build new reactors just about as well, especially if they keep the changes from existing designs to a minimum. After all, design changes are what can produce years of added regulatory processes and ensuing delays. We should recognize what we have built, and think seriously about what we could do with it.

There is a broader lesson from this, too. Infrastructure costs have been skyrocketing around the world in recent years, especially in countries like Canada, the United States, and the UK. But this is not inevitable, and Canada’s nuclear sector is a rare example of a big cost surge that was arrested and even turned around. That happened because of a willingness to recognize that projects went badly and a determination to figure out why and how this could be prevented in the future. We all know of many projects that have faced serious delays and cost overruns; the question is whether their builders followed that same path of figuring out why and how to make sure it doesn’t happen on the next projects.

We tend to view commercial aviation through a purely human lens, as a system built to move us from Point A to Point B. But from an infrastructure and economic standpoint, your passenger plane is actually a dual-purpose vehicle where what’s under your feet can be almost as valuable as you. To understand how a backpacker can afford to visit Machu Picchu, or how a low-cost carrier can fly you across the Atlantic for a few hundred dollars, you also have to think about the bottom half of the pressurized metal tube that flies through the sky.

An airliner is, fundamentally, a long aluminum or composite cylinder. Because of structural engineering and aerodynamics, a circle is the most efficient shape to pressurize. But humans prefer flat floors to walk on.

When aircraft designers place the cabin floor, they place it roughly at the widest point of the circle. This design choice creates a spacious upper deck for passengers, but it leaves an enormous, curved cavity right underneath them. That’s the belly.

Some of this lower space is reserved for passenger luggage. On certain ultra-long-haul routes, a portion is partitioned off for auxiliary fuel tanks or crew rest quarters. But on a typical widebody aircraft, a substantial amount of that real estate remains completely empty. To an airline, that space is a financial opportunity.

While familiar shippers like FedEx and UPS, alongside behind-the-scenes players like Atlas Air and Cargolux, operate massive fleets of pure freighters, a major percentage of the world’s air freight doesn’t travel on cargo planes at all. It moves quietly in the belly holds of the regular commercial flights people take every day.

Subscribe for free, for infrastories from around the world.

Subscribe now

What actually goes into the belly? Air freight is expensive, so it is reserved for items where speed is worth a massive premium. Some of it is ultra-high-value goods like pharmaceuticals, electronics, or even famous paintings. Every autumn, there is intricate logistical choreography designed to move billions of dollars’ worth of new iPhones from factories in Asia to store shelves globally, precisely in time for launch day (a major environmental initiative for Apple turns out to be better supply chain management so more iPhones travel on ships instead of on planes).

One of the most important types of air cargo, however, are the fresh foods that would rot into goo if they had to survive a long container ship ride across the ocean. When you eat fresh berries in the dead of a northern winter, or enjoy sashimi thousands of kilometers from the nearest ocean, you are often tasting something that spent some time under airline passengers’ feet.

This reality introduces a fascinating economic dynamic. While passenger ticket sales generate the lion’s share of an airline’s revenue, belly cargo provides highly lucrative, high-margin supplemental income. On a route where demand is mostly price-sensitive tourists, the predictable revenue from cargo can be what tips a flight into profitability. Not all flights, however, have nearly as much cargo demand as others.

Looking at a map, one might expect that Peru would struggle as a major tourist destination despite its great attractions, since it’s quite far from big tourist markets like Europe, North America, and northeast Asia. This would ordinarily make airline tickets much more expensive than ones to closer destinations like Mexico, the Caribbean, or North Africa.

But Peru has a huge advantage. It is a massive global exporter of fresh agricultural products. This creates a symbiotic relationship between tourists making a trip to Machu Picchu and grocery store shoppers looking to buy some fresh asparagus even in the dead of winter.

By filling both the upper deck with humans and the lower deck with vegetables, airlines maximize their utilization of the aircraft. The cargo revenue and passenger revenue effectively cross-subsidize each other, bringing down both ticket prices for tourists and shipping costs for agricultural exporters. It also makes non-stop routes feasible that might not otherwise work economically, which is valuable for making a place a top-tier tourist destination. The asparagus in the belly means that a flight from North America to Lima might not be much more expensive than a much shorter flight to the Caribbean, where the belly often flies mostly empty.

Iceland has famously transformed itself into a massive hub for low-cost transatlantic travel. Millions of people have taken advantage of the Icelandic stopover, where passengers break up a flight between Europe and North America with a quick trip to the Blue Lagoon.

But the math behind those shockingly cheap Icelandair tickets relies heavily on what is happening beneath the passenger cabin. The waters surrounding Iceland are some of the richest fishing grounds in the world (so much so that they sparked a Cod War between Iceland and the UK).

The bellies of those transatlantic flights are consistently packed tight with fresh Icelandic fish bound for restaurants and supermarkets in both North America and Europe. Your cheap vacation to Reykjavik is, quite literally, being carried on the backs of fresh cod and salmon.

The next time you walk down the jet bridge, take a moment to look down at the tarmac. Beneath the floor of your cabin, wrapped in the dark belly of the plane, a completely different journey is taking place. Your ticket might well be getting subsidized by someone’s iPhone or sushi.

This is the kind of mega project that seems completely fanciful today in North America. It was likely pretty fanciful at the time this report was written, but on the other hand it is exactly the kind of project that many cities around the world that are able to build big infrastructure more affordably would take as not merely possible but obvious. If you go to Spain or France, even fairly small towns have expressways buried invisibly underground and ambitious projects to reorganize rail stations to improve access and free up land can be found in many big European and Asian cities. Stuttgart 21 is one contemporary example.

The rail corridor is a gash across the the city’s core, playing at least as big a role in severing downtown from the waterfront as the much-maligned Gardiner Expressway. The expression “the wrong side of the tracks” has real roots, as big rail corridors, like expressways, have long created major physical and psychological barriers between neighbourhoods. For decades, while the area north of the tracks was packed with office towers, the area just on the other side was mostly empty parking lots. That was true at the time of this study. Even today, though the former rail yards on the south side have finally been mostly redeveloped with offices and condos, walking from Front Street to the other side of the tracks is a pretty dark, dingy experience.

This problem was long recognized. As freight operations largely shifted to the suburbs, a large land area in the heart of the city that had previously been used for rail yards was freed up. In the late 1960s and early 1970s, CN and CP Rail jointly proposed the Metro Centre scheme, which would have demolished Union Station altogether, relocated the rail corridor adjacent to the Gardiner Expressway, and freed the entire railway lands for a vast development of office towers and residential buildings. These kinds of megaprojects typically deliver on their most prosaic elements while the visionary ambitions are quietly shelved. Curiously, Metro Centre did the opposite. The office blocks and condos never made it off the drawing board, but the eye-catching showpiece in the renderings, the world’s tallest tower, amazingly actually got built.

Preservationists successfully resisted the demolition of Union Station and its magnificent Great Hall. A revised scheme put forward in 1974 kept Union Station, shifted the office towers nearby, and proposed an ambitious public space instead. A park-like plaza was to be built over the tracks and platforms, accessible by escalators from the surrounding streets. Again, nothing much came of the plans. The study team at Peat Marwick, however, acknowledged that decking alone wouldn’t resolve the fundamental problem of connectivity between downtown and the waterfront, and suggested that burying the corridor entirely deserved further investigation.

A dozen years later, the Morrison Hershfield report was commissioned by a real estate company with ties to one of the railways. It concluded that it was “technically and operationally feasible to depress the rail corridor and platform tracks and allow for ‘at grade’ north-south streets, and major area building/covering of the rail corridor from as far west as Bathurst Street to as far east as the Don River.” York, Bay, and Yonge Streets (potentially along with additional streets further west) would cross the tracks invisibly and imperceptibly.

The study even claimed that traffic on the former two could be maintained during construction by routing vehicles through the teamways, the existing passageways on the either side of York and Bay that are now used as pedestrian routes and additional platform access points. Full closures to traffic could be as short as a single weekend. Yonge, lacking an equivalent laneway, would present a harder case, but the report suggested closures even there could be limited to very short periods. As they crossed the tracks, the streets would then connect to a new Esplanade-Bremner link, with York, Yonge, and Bay ramping gradually down to meet the existing street level at Lakeshore.

Subscribe for free, for infrastories from around the world.

Subscribe now

The report offered a cost estimate that seems unimaginable today: between 528 and 633 million dollars. This furthermore was for the higher cost option that maintained 1% grades suitable for freight trains throughout. It even included 25% to 50% contingencies. A solution that accepted 2% grades, which are perfectly reasonable for passenger trains, would have had even lower costs. The report was careful to describe these as order-of-magnitude figures.

Obviously, this study was extremely preliminary and perhaps flawed. It’s hard to take the cost estimates or the purported ease of construction at face value. But it does illustrate the incredibly ambitious kinds of infrastructure projects that were being contemplated and even built in that era. The Morrison Hershfield report was written at the tail end of a generation that had built enormously complex urban infrastructure through dense city centres across North America, not all of it good. Thousands of kilometres of expressways had been built including through the heart of major cities, like Montreal’s underground Ville-Marie Expressway. Ambitious transit projects like Washington’s Metro, San Francisco’s BART, and Montreal’s Métro were completed or underway. Even urban rail networks had been reorganized to reduce their aesthetic impact and provide land for construction. Philadelphia’s Penn Centre was built on a buried rail corridor, as the office towers along New York’s Park Avenue had been built on the yards of Grand Central Terminal a generation earlier. The sense that large, technically demanding projects were achievable was grounded in recent experience, which gave the report a confidence that looks surreal to us today.

Boston’s Big Dig was supposed to be the capstone of this tradition, a project that would correct the expressway-building sins of earlier decades by burying the Central Artery and replacing it with a linear park. Its eventual cost, rising from initial estimates around $2.5 billion to over $14 billion in actual expenditure, effectively discredited a certain kind of North American infrastructure imagination for a generation or more. The original vision for the Big Dig had included a north-south rail link that would have been transformative for Boston’s regional connectivity, but that component was stripped out as costs spiralled. They were left with a highway project that, despite its genuine urban benefits, became the emblem of everything that could go wrong with big infrastructure building. It became a cautionary tale that has tamed ambition across North America.

In other parts of the world, however, that ambition remains strong. As mentioned above, Stuttgart 21, the project to relocate Stuttgart’s through station underground and free the corridor for redevelopment, has been controversial, over budget, and interminably delayed, but it is being built. Vienna has successfully undertaken a comparable rationalization of its rail infrastructure. Across France and Spain, even small towns have buried sections of urban expressway as a matter of routine civic improvement, and Chongqing is in the midst of something even more ambitious with its rail network. Burying tracks, or highways for that matter, through a city centre is, in much of the world, a relatively prosaic project.

What changed in North America is partly about costs, as construction costs here have diverged sharply from European and Asian levels over the past several decades for reasons that are themselves worth understanding. It’s also because the mistakes of the expressway era and the delivery challenges of projects like the Big Dig and more recently the Eglinton Crosstown have made proposing boldly ambitious (or just regular) urban infrastructure almost unserious. The Rail Deck Park proposal, which would create a park over the already-below-grade section of the corridor between Spadina and Bathurst, has been discussed for years and remains unrealized. It is far more modest than what the Morrison Hershfield study contemplated, and even that has proven difficult to advance, with each plan seemingly further reducing scope.

It’s entirely plausible that if Toronto had tried to build this, it would have encountered the same cost surge and problems as the Big Dig. It was certainly far beyond the risk level that private companies like a real estate developer could contemplate. But it’s a bit sad to see the decline of ambition. Such a scheme would never even be contemplated in Toronto or most North American cities today. Perhaps it will take a North American city to try a project as ambitious as the Big Dig, but executed successfully this time, to make people willing to think boldly once again.

Freight transport is something that is too frequently brushed aside in favor of transit and aviation, and inland waterway freight transport is even less discussed.

This is a mistake: Ignoring a form of transportation that greatly impacts our lives while being sustainable, efficient, having massive economies-of-scale, the ability to use existing infrastructure, and which plays key role in national industrial policy. Inland waterway transport (IWT) is a silent giant: A vital component of the global logistics network moving vast quantities of the world’s industrial raw materials, food, and manufactured products along a network of rivers, lakes and canals to the hinterlands of otherwise landlocked areas. Most global trade is done via ocean shipping; IWT is usually a key feeder system for this ocean trade between the economic giants of the EU, the U.S., China and everything in between.

Unlike passenger transport, actors in freight transport must strategically balance price and frequency to maximize profit/value, while speed and frequency becomes a market differentiator. This is vastly different from passenger transport where speed and frequency are generally of the upmost importance to the consumer and comfort becomes a market differentiator. When choosing modes for travel, passengers may find the span of travel times offered between various competing bus, plane or train modes measured in hours. In contrast, a freight shipping market between two places can have travel times of various competing modes that vary from hours (airfreight) to weeks (ocean freight). Here, unit cost is the major factor. As such freight logistics in general still heavily relies on massification and economies-of-scale, which is how modes that to a passenger seem large, slow, and infrequent remain economically competitive across broad sectors of the freight transport economy.

Unlike passenger transport, speed or absolute travel time plays a lesser role in route and mode choice for general freight movement. Cost per unit-distance moved, of reliability of arrival times, and consistency in travel times are much more important metrics for shippers and industrial customers than the as-much speed and frequency of service preference that passenger transport offers. The exception is the rapidly growing market of Less-than-Container-Load (LCL) traffic from the rise in E-commerce: LCL cargo traffic mirrors dynamics found in passenger transport, favouring frequent, fast, and responsive transport similar to what passengers expect in a transportation system.

IWT has huge potential to be a low-emission, cost-effective alternative to road and rail transport. Despite most IWT systems relying on fossil fuels for motive power, the scale of goods transported in a typical IWT vehicle (a ship) has such high efficiency the fuel used and pollution emitted per unit of goods moved competes with the operational cost and environmental benefits advantages offered by electric freight railways. This is similar to the cost and environmental scale dynamics of an American diesel freight train when compared to trucks (even electric ones) but with more monumental proportions. For example, a single multi-kilometre diesel-powered American freight train may take hundreds of diesel-powered trucks off the road, but a single large diesel-powered American 40 barge tow down the Mississippi IWT system can take around 3 to 4 multi-kilometre American diesel-powered unit freight trains off the rail system. Europe’s freight transportation system operates on a similar paradigm but at a smaller scale, with lower axle weight, short sub 1-kilometre freight trains, and tows with barges numbering in the single digits.

IWT is sometimes the only mode of transport capable of shipping a large volume of low value goods at an attractive cost. The cost of a product is the labor, production and transport costs with profit on top. The cost-effective scale at which IWT operates is the reason national industries such as soybean growers in the US Midwest remain globally-competitive to price-sensitive consumers in China. Using massive diesel towboats pushing barges lashed together down the famous Mississippi River IWT System (which includes the Ohio, Missouri, and Illinois rivers), farmers receive significantly lower unit transport costs for getting their goods to market. This IWT barge traffic heads down to New Orleans and transfers into large international ocean bulkers for global export. Using the Mississippi River System infrastructure, US producers can accept higher labour and production costs while maintaining a competitive overall unit price for their goods to global consumers using the low transport costs IWT. IWT is especially tuned to export sectors of agricultural products and manufactured goods; barges/ships filled with full containers, cars and agricultural feedstock from inland industries just need float downstream with minimal engine power, riding the current of the waterway to ocean export hubs. When those same barges/ships return upstream running their engines hard fighting the water currents, they are carrying less freight: empty containers and half-full bulkers of fertilizers.

Apart from agricultural products (corn or soybeans), the US Mississippi River System also moves enormous volumes of bulk goods (coal or gravel) and petrochemicals. Other North American IWT systems include the deep-draft but less used Great Lakes/St. Lawrence “Seaway” shared by Canada and the United States. Unlike the barge tow-based Mississippi River IWT system, the Seaway uses a ship-based fleet. Uniquely, the Seaway has a fleet of massive “Laker” lake freighters that would not look out of place in trans-oceanic operations. These ships exceed “Seaway-max” sizes of the locks in Welland, Ontario that bypass the Niagara Falls, becoming captive to the Great Lakes west of Lake Erie. This captive fleet carries bulk goods such as iron ore, coal, and grain. Strangely, the Great Lakes/St. Lawrence Seaway does not carry much container freight with the Port of Montreal being the only significant container port on the Seaway.

Across the pond, Europe has a dense, highly-integrated IWT network with the EU’s “TEN-T” policy designating core corridors for improvements. Europe’s busiest freight waterway, the Rhine, connects the Port of Rotterdam (Europe’s busiest port) to the industrial heartlands of Germany’s Ruhr region, Switzerland, and France. It carries about two-thirds of EU inland waterway freight, moving chemicals, ores, coal, containers, and agricultural products. Of all EU IWT networks, the waterway system centred on the Rhine and the feeder network in the Low Countries are arguably one of the factors of success for Germany’s continuing dominance in European manufacturing.

Of course, there are other key IWT systems around the world like the Danube, an east-west artery linking 10 European countries from Germany to the Black Sea. In South America, the Amazon IWT System allows vessels to reach Manaus, Brazil 1,600 km inland and is commonly the only way for many communities in the Amazon to access the rest of Brazil; or the Paraná-Paraguay waterway, which is the economic lifeline to landlocked Paraguay and southern provinces of Brazil to export to global agricultural markets. Lastly, the IWT along the Mekong and Chao Phraya Rivers of Mainland Southeast Asia play a large role in industrial development and even geopolitics.

Which brings me, to the Goliath — China. Like High Speed Rail (HSR), skylines, metro systems, and Ultra High Voltage (UHV) transmission lines, China has the largest and busiest inland freight waterway specimens in the world by a country mile, exceeding aforementioned examples in the US and EU by a few multiples in scale and volume. As shown below, the Chinese IWT network consists of three parts: the Grand Canal (yellow), Yangtze River (red), and the West Pearl River (green) systems.

The Yangtze IWT system is the largest and serves as the primary artery for China’s manufacturing and resource flows between industrial megacities in the interior with the global shipping hub of Shanghai, the world’s busiest port. Her branch routes allow IWT traffic to reach Chinese cities off of the Yangtze such as Changsha, Hefei, and Nanchang. 13,000-ton ships with up to 6 meter draft can reach Wuhan, almost 1,000km upstream from Shanghai. A 13,000-ton IWT container ship like the Hanhai 5 pictured below can carry over 1,100 TEUs of containers, about the capacity of typical large North American double-stack container train.

The result? Cities like Wuhu, Jiujiang, and Changsha, hundreds of kilometres away from any ocean, each handle almost 1 million TEUs of containers per year via inland shipping. Chongqing is over 1,000km away from any ocean and handles over 1 million TEUs of containers per year, while Wuhan’s River Port alone handles over 2 million TEUs per year, the same volume of containers handled by the Port of Charleston, despite Wuhan being over 800km inland. In 2025, 26.2 million TEUs of containers were transported along the Yangtze IWT system.

Annual freight volumes on the Yangtze reached 4.2 billion tonnes in 2025. This is the traffic volume of several entire Mississippi (~450 million tonnes) or Rhine/Low Country (284.5 million tonnes in 2024) waterway systems and a larger freight volume than all other inland waterway systems in the world outside of China, combined. I remember when I was in Nanjing looking at the Yangtze River seeing ships pass by every few seconds. I encourage readers to go on Google Earth and follow the Yangtze River up from Shanghai.

This growth of IWT on the Yangtze has put a huge strain on infrastructure and formed traffic bottlenecks all over the river. To address this capacity gap, large upgrade projects are underway, including work on a container portage railway and a second, much larger, lock system at the Three Gorges Dam. In addition, new markets are being explored such as development of fleet and riverside infrastructure for IWT of finished Electric Vehicles. In addition to expanding capacity, China is leading the way in the emerging trend in electrification of the IWT fleet. China started production of the world’s largest electric battery powered container ships and bulk carriers for use on the Yangtze IWT. With swappable batteries in standard shipping containers sufficient to get from Shanghai to Wuhan (a distance of 1,000km) on one charge.

Apart from the Yangtze River, the southern network of the Grand Canal, which parts date back to the Ming dynasty, play a key local freight IWT network for Shanghai, Jiangsu, Zhejiang and Anhui Provinces. IWT traffic on the southern Grand Canal network connecting Nanjing, Hangzhou, Shanghai and everything in between was so high that bypass and relief canals had to be constructed to combat ship and barge congestion around several urban areas.

Lastly, there is the Pearl River IWT system — a latecomer with development only happening in the past 5-10 years. The Pearl River IWT system allows inland waterway traffic from Foshan, Nanning, Liuzhou, Zhaoqing, Jiangmen, Qingyuan, and Shaoguan to access the large international ocean ports of Zhuhai, Guangzhou, Shenzhen, and Hong Kong, which all have IWT barge transshipment berths. You commonly see river barges (called “pigs” or 豬 by the local Hong Kong seafarers) in HK heading into China. HK even has a dedicated “River Trade Terminal” in Tuen Mun for Pearl River IWT traffic.

Today, massive investments have been made to increase shipping capacity of the Pearl River IWT system. Including the completion of a large dam in Guiping, Guangxi to increase allowable vessel sizes. Whenever I travel through the Pearl River Delta, I see fleets of ships ply the various IWT channels in the delta. For example this one I saw when I was relaxing in a riverside park near Jiangmen: two dry bulk carriers, a container ship and a bulk chemical IWT ship.

The governments of US, EU, and China have well-developed networks of all three freight transportation modes. How they can best balance investment priorities and develop clear policies for integration between these freight transport modes remains a big question. The EU struggles with scaling operations of their IWT and railway freight systems, with each both operating at much smaller scales than China and the US respectively. China has traditionally followed a similar freight development policy as the EU where heavy freight is mostly handled by inland waterway shipping with railways playing a lesser role, conflicting with extensive passenger operations. This doesn’t stop China Railways from being the busiest freight railway system in the world. In contrast, railways play broad significant role in freight logistics in North America with the popularization of “land barge” style freight train operations, at the expense of proper passenger rail transport. US IWT development is also complicated by the Jones Act. On the Chinese side the opposite is happening. China is still developing and improving the rail and waterway network in a bid to shift freight traffic away from trucks. However, there is clearly some confusion on China’s freight railway development policy vis a vis IWT. This leads to some underperformance and (dare I say) lack of ambition on freight railway development despite world topping tonnage conveyed by China Railways every year. In contrast, big national IWT upgrades underway across China, especially in the South with the Pinglu Canal nearing completion will greatly help economic development of Guangxi Province and facilitate trade with SE Asia. I’m personally very excited about the Jinghan Canal project that bypasses Wuhan, which suffers from too much heavy freight barge congestion.

However, on the Chinese side, the question must be asked: where can a dedicated electric traction freight railway network fit in national industrial development plans. It is genuinely not clear what China’s optimal funding allocations and priorities should be between expanding rail freight and IWT infrastructure scale, capacity, coverage and efficiency. In China, these two modes increasingly overlap in market positioning, particularly in various scales of container, specialty chemical, automotive and finished manufactured goods market. This does not exist as strongly in the EU (in terms of scale) or the US (in terms of overlapping roles between modes).

Inland waterway transport is an important and often invisible means of transporting vast quantities of freight. It’s worth thinking about next time you spot a barge floating down a river.

This post was written by JR Urbane Network, who covers urban rail development in Asia, with a particular focus on China.

JRUrbaneNetwork

Trains are like dragons: massive, majestic, mystical and capable of quickly taking you on an exotic adventure. Longform posts on trains and urbanism in East Asia and Beyond.

We often hear about disastrous megaprojects with high costs, delays, and failures to meet demand projections. I’d argue that part of the reason for this is that we only remember the unsuccessful ones or the recent ones that were delivered poorly. When you think of a megaproject, it might be something like Montreal Mirabel Airport, Athen’s Olympic facilities, or Boston’s Big Dig. We remember these because they’re standing there underused or they were the cause of recent controversy; a successful megaproject, or even one that happened just long enough ago that its construction travails are forgotten, becomes largely-ignored but essential infrastructure. Our attitudes toward megaprojects are largely coloured by survivor bias: once a megaproject is successful, it becomes invisible, but everybody notices that white elephant airport or transit line few people use.

For every Mirabel — Montreal’s underused jetport that no longer sees any passengers — there are dozens of Hartsfield-Jackson Atlantas or Munich Airports whose immense infrastructure is extremely useful. These projects were no less “mega,” but their great success means that complaints about them are more about overcrowding than project delivery or high cost. Paris’ RER, which I wrote about recently, was deeply controversial for its high costs. The expenses put immense strain on the region’s budgets, forcing more efficient approaches, a dedicated tax, and some compromises one might wish could have been avoided, like the shared RER B and D tunnels.

But today, those budgetary challenges are long since forgotten and nobody could imagine the Paris region functioning without the RER. Even the negative memories of Boston’s Big Dig are starting to fade and people are generally pleased that they have a long park through the centre of the city rather than the Green Monster.

The same is true for countless megaprojects across the world. Projects don’t get much more “mega” than the Delaware River Aqueduct. Begun in the 1930s, for decades it was the longest tunnel in the world. The cost was a major burden on the City of New York’s budget. Yet, does anyone even think about its existence when they turn on a shower in New York City? That megaproject provides the water that makes today’s New York City possible. No matter how much it cost, its benefits are incalculable.

New York’s Sixth Avenue Subway was ruinously expensive and complex to build in the 1930s. But does anybody worry about that today? The Beauharnois hydroelectric complex outside Montreal was a scandal that brought down a government. A century later, it remains a hugely important generating station that provides steady emissions-free power for a fraction of a cent per kilowatt-hour. The original American and Canadian transcontinental railways also resulted in huge costs and government-shaking scandal, but both countries could scarcely exist without them. Dallas-Fort Worth airport, which took forever to build and cost a fortune, is now one of the busiest airports in the world and a critical economic anchor for the whole state. The Jubilee Line Extension in London was delayed, far over budget, and nearly cancelled; a couple decades after its opening, it has become the essential link to many of the city’s fastest growing neighbourhoods. No doubt in Ancient Rome, there was someone in a toga complaining about the cost / benefit of a road that we still use two thousand years later.

This is another fundamental point about major infrastructure: it often functions far longer than we are accustomed to planning for and enables things we couldn't imagine. Hydro infrastructure in Quebec now powers AI inference and electric cars, things which were not even a glimmer in some planner’s eye when the concrete was being poured. Financial calculations of cost and benefit generally only look up to about 30 years out. Likewise, demand projections rarely even try to calculate more than a few decades. It’s easy to imagine the builders of the original London Underground worrying about meeting demand projections in year one and two, and how they were going to pay the interest on their bonds. It would not have even occurred to them to think that those tunnels would be carrying hundreds of thousands of passengers per day a century and a half hence.

The point of this argument is that short-term debates about infrastructure often miss that it usually has a life much, much longer than what we are comfortable planning for. That’s not to say that cost doesn’t matter; if it prevents us from building a project in the first place, or requires us to make serious compromises, the longevity of big infrastructure means those choices can haunt us for decades or more. And there are some projects that are genuinely ill-conceived, like Mirabel Airport where splitting domestic and international flights between two airports killed the connecting traffic that is essential for any major airport. But deciding whether to build infrastructure that will last a century or more based on the construction period or the first few years of operation is usually a big mistake. We should build intelligently and as cost-effectively as possible, certainly, but we shouldn’t let bad experiences discourage us from building the big infrastructure that is the basis of our modern civilization.

In the mid-1960s, Paris was on the verge of spending billions of francs to duplicate rail infrastructure that already existed. It was a classic phone call problem. The RATP, the local transit authority that ran the Métro, and the SNCF, the national railway running regional and intercity trains, had their own separate corridors, technologies, and operating practices that made sharing infrastructure seemingly impossible. But it wasn’t, as we will see.

Paris retained its status as a city of global importance in the postwar period due in part to its extraordinarily extensive urban infrastructure. Following the development of the world’s densest metro network within the historic city walls, national and regional governments recognized the need to serve a rapidly suburbanizing population. At the dawn of the Fifth Republic in the late 1950s, trains from historic and newly developed suburban towns ran to century-old terminal stations arrayed around the perimeter of Central Paris. Commuters who couldn’t walk from those stations to their destination were forced into cumbersome transfers onto the congested Metro.

In the 1960s, the French national government embarked on a program of comprehensive regional planning that sought to resolve these problems through large-scale infrastructure investment. A centrepiece of these plans was the development of a new rapid transit system that would be both faster and more capacious than the existing Métro, permitting it to extend beyond the historic city walls to serve the rapidly expanding metropolitan region. Although the plans were beset by cost overruns and jurisdictional squabbles at first, the national government persisted in developing the network. Drawing on inspiration from sources as distant as Japan, French planners devised a creative strategy that overcame technical, jurisdictional, and bureaucratic obstacles to connect disparate suburban lines of the national railway into a high-frequency, high-capacity urban network that doubled as an express rapid transit system within the city. The development of the RER transformed the metropolitan region and enabled the development of transit-oriented suburban communities that house and employ millions, while retaining a close connection with the historic centre of the capital city.

Though Paris enjoyed an exceptionally dense urban metro network, most of it was the legacy of construction before the Second World War. Public transportation expansion had stalled by the 1950s, with only insignificant extensions to the urban Métro reaching completion. The national government during the postwar Fourth Republic was wracked by repeated changes in government, which hindered long-term planning. No progress was made on the development of a system that would unify the broader, rapidly suburbanizing urban region.

The Fifth Republic, proclaimed by Charles de Gaulle in 1958, had aspirations to demonstrate and celebrate France’s grandeur through planning and great infrastructure projects. The modernization of Paris was to be, as it had been for previous aspirational French leaders, a centrepiece of that effort. The population of Greater Paris was growing rapidly in the postwar period, adding between 100,000 and 150,000 new residents per year, and much of this growth was occurring outside the historic city walls. Despite the ideas of planners like Le Corbusier, who sought to rebuild the historic city in the modernist style, the city built by the Baron Haussmann in the 19th century was left largely intact while most new development was relegated to the suburbs. A new public transportation system was needed to tie these rapidly growing communities to the heart of the capital.

De Gaulle’s government radically restructured Paris’ governance with the creation of a unified District de la région parisienne (District of the Paris Region), an autonomously financed body that would be responsible for implementing the national government’s projects in the metropolitan area of the capital. At its head, de Gaulle appointed Paul Delouvrier, a veteran administrator who was given substantial powers to push through projects. He would play a critical role in the development of the RER program.

The grand plans of de Gaulle and Delouvrier for Paris were intended to rapidly overcome the infrastructure deficit that had developed during the Fourth Republic. The Gaullist plan, the Schéma Directeur d’Aménagement et d’Urbanisme de la Région de Paris, was far more ambitious, seeking to accommodate 7 to 8 million more residents in the Paris region. The most optimistic projections, based on extrapolating rapid postwar population growth, anticipated a metropolitan population of 20 million by the year 2000. The new plan included major infrastructure investments, such as a new airport to the north of the city in the town of Roissy that would become Charles de Gaulle International, and extensive suburban growth concentrated in two axes to the north and south of the city. It also finally elaborated on plans for a comprehensive regional rail network incorporating the new east-west line (now RER A) then under construction. It would add two additional north-south lines to create an ‘H’-shaped network centred on the historic city centre, amounting to more than 250 km of new railway lines throughout the urban area. Delouvrier was enthusiastic about the potential of the plans to encourage development in the historically less prosperous eastern half of the metropolitan region.

This plan resembled contemporary projects in the United States, such as the Bay Area Rapid Transit in the San Francisco Bay Area and the Metro in Washington, D.C., in that it relied almost exclusively on new construction and was not designed to interoperate with the existing regional rail network. The new lines would parallel or completely rebuild hundreds of kilometres of existing rail corridors then operated by the SNCF, expending vast sums on duplicating infrastructure that already existed.

Construction of the first stations on the east-west line began as underground stations were gradually added to the suburban lines in the east and west of the city, with the ultimate intention of connecting the two corridors. Several lines were ceded by the SNCF, which was then uninterested in urban transportation, to the RATP, which sought to rebuild and repurpose the suburban lines as part of a segregated regional rapid transit system. These first stations spawned one of the most enduring controversies about the initial plans. The architecture of the stations was intended to mirror that of the Paris Métro stations designed by the celebrated engineer Fulgence Bienvenue, featuring a single arch ceiling rising over the tracks and platforms and no central supports.

While this engineering task was relatively simple in small and shallow Métro stations, constructing the vastly larger and deeper stations of the RER—250 metres long, 45 metres wide, and 30 metres high—posed a severe challenge. Unsurprisingly, these stations faced immense cost overruns, ultimately costing 1.3 billion francs. The high cost of constructing the new lines placed the prospect of further construction in doubt.

Around this time, a group of French engineers and administrators went on a junket to Japan. Most junkets might have a limited impact at best, but this one was worth billions. The fundamental problem in Paris was not engineering. These were phone call problems: disputes over jurisdiction between the RATP and the SNCF. Instead of running trains seamlessly from the SNCF network onto the RATP-administered RER lines in the city centre, the RATP sought to either take over the SNCF lines to the exclusion of all other rail traffic, or to duplicate them, largely through expensive underground construction. Tokyo showed them another way. Though no fundamental technical obstacles existed that would prevent shared operation, the two organizations were hesitant to cooperate on a plan to mitigate their operational differences, which included different electrical systems, different operating rules, and even which side of the rail corridor was dedicated to which direction of travel. On the SNCF, trains drove on the left, while the RATP trains drove on the right. But some of the specific barriers were even more minute. SNCF drivers preferred to stand while driving; RATP drivers sat. The brake handle was on the right on RATP trains, and the left on SNCF trains.

Subscribe for free, for infrastories from around the world.

Subscribe now

Christian Gerondeau, acting as adviser to the prime minister, formed a steering committee that included directors of all of the organizations participating in the RER project. The committee was established to study the project in detail and to find ways to manage and overcome the challenges that it had faced. The committee bravely chose to look halfway across the world to find a model for emulation, pointing to the metro system of Tokyo that effectively transported millions of passengers per day across a vast urban area that dwarfed even the French capital. They were stunned to see the Chuo line, which moved 100,000 passengers per hour into a single terminal station consisting of one 10-m wide platform and two tracks. Most significantly for Paris, many of Tokyo’s private railway lines that served the city’s far-flung suburbs did not deposit their passengers at a single urban terminal. Trains would run on private railway lines from deep out in the suburbs until they reached central Tokyo. Then, they’d dive into the regular municipal metro, mixing with normal metro trains. At the other end, they’d often emerge from the metro and run on another private railway deep out into the suburbs once again. This required the co-operation of the municipal subway corporation and two private railway companies on either side of the metropolis. This example of cooperation between private companies convinced the Parisian planners that integration between the RATP and SNCF, which were, after all, both bodies of the French State, was possible.

A subsequent tour of the Tokyo Metro system by the French planners also demonstrated the value of cross-platform transfers, which greatly facilitated the handling of vast numbers of riders transferring between lines, as it eliminated the need for them to crowd onto stairs or escalators. Before their visit to Tokyo, French engineers had been anxious about the potential safety implications of such a practice, but viewing the crowds safely handled in the Japanese capital convinced them of its feasibility. This model would be employed at Châtelet-Les Halles station, the critical junction in the centre of Paris between the north-south and east-west lines that would be at the heart of the RER network.

New technical approaches had to be devised to permit the interoperation of SNCF and RATP trains. New MI 79 trains were designed specially to operate on the different electrification and signalling systems of the two railways. They even had separate controls in the cab designed around the preferences of each operator. At Gare du Nord, the SNCF operator would get off and the RATP operator would get on and the train would switch between the different systems, all in a matter of seconds and basically without being noticed by riders. All trains were made a part of the RATP fare system, which facilitated connections with the urban métro. An agreement between the SNCF and RATP allocated revenues proportionally. Once the agencies decided that they wanted to cooperate, the technical obstacles were all surmountable.

Once cooperation between the RATP and SNCF was possible, a vast array of new opportunities became available to the Parisian planners. Before, new lines required extensive construction of new lines and the rebuilding of existing lines. After, RER could be rapidly deployed on existing lines. A few short segments of tunnels in the centre of Paris could now connect nearly the entire network of SNCF suburban lines. Instead of tunnelling through the suburbs, the legacy of 19th-century railway construction provided already-built corridors that spanned the metropolitan region.

The RER worked because it had strong and clear political backing, a clear objective, technically informed leadership that could make decisions and force disparate agencies to collaborate, and ultimately the willingness to use creative means and new technologies to bridge gaps.

In the mid-1960s, Paris was considering a plan that would spend billions of francs to duplicate regional rail networks that already existed. When the people who were planning the RER went to Tokyo, they saw completely separate rail operators sharing infrastructure and realized that they too could change the way they operated to be able to work together. They didn’t conclude that Japan was somehow too different to emulate; their reaction was that if Japan can do it, France can do it too, if not better. Out of that, we got the most successful regional rail system in Europe.

New York has three major airports, and none of them works particularly well.

LaGuardia handles the premium domestic traffic from cities within 1,500 miles plus Denver that are allowed to fly there (other than on Saturdays). Its perch in northern Queens is a short cab ride from Midtown. JFK handles the vast number of airlines from around the world that want to fly to New York, plus smaller hub operations for Delta, American, and JetBlue. Newark has become New York’s only true major hub (for United), but it’s hobbled by its reputation as the inconvenient airport and is strong in the New Jersey market and not really anywhere else.

It’s an arrangement that is far from ideal. A passenger on Delta arriving at LaGuardia and connecting to a transatlantic flight has to take a cab to JFK. The airspace over the city is among the most congested in the world, exacerbated by coordinating the approaches to three major airports in a small area. And the city’s international gateway is accessible by train only through a sequence of transfers that takes the better part of an hour from Midtown and is extremely difficult from many other parts of the region.

This is the product of parochial politics dating back to a stunt at a Brooklyn airfield in 1934.

Newark had geography on its side from the beginning. It’s closer to Manhattan than JFK, well-positioned for New Jersey’s substantial population, and naturally connected to the highway and rail corridors that bind the Northeastern megalopolis. When commercial aviation took off in the 1920s and 30s, Newark was an obvious choice. It became the region’s first major airport for good reason.

Fiorello LaGuardia wasn’t happy about it. The mayor of New York was a man of volcanic civic pride, and the idea that the New York metropolitan area’s main airport was in New Jersey offended him on principle. He landed on a flight to Newark, claimed that his ticket said New York, and demanded to be taken to the city. The plane eventually landed at Floyd Bennett Field in Brooklyn, a publicity stunt that launched the political campaign for a “real” New York City airport.

Subscribe for free, for infrastories from around the world.

Subscribe now

LaGuardia relentlessly pushed for a new airport within city limits, and the project he finally got built came to bear his name. The airport’s lovely Marine Air Terminal, recently vacated by Spirit Airlines, was built in the streamline moderne style for Pan American’s flying boats. It remains one of the most beautiful airport buildings in the US. The airport itself, however, was always constrained. It was built on reclaimed land with limited room to expand, surrounded on three sides by water and on the fourth by dense residential neighbourhoods.

The 1930s generation of airports like LaGuardia was designed for propeller aircraft carrying dozens of passengers. By the late 1940s, with jets on the horizon and air travel booming, every major city’s airport was undersized. London outgrew Croydon, Paris outgrew Le Bourget, and even a divided Berlin outgrew Tempelhof. New York had LaGuardia, which was already at its limits, and Newark, which was sitting largely underused with land available.

Instead of recognizing Newark’s longstanding advantages, the Port Authority decided to build a whole new airport. The site they selected was on Jamaica Bay in the far southeast corner of Queens. It was mostly marshland but included the Idlewild Beach Golf Club. The area was remote and the terrain was difficult, requiring enormous landfill and drainage work to make it buildable. The site was the biggest in the world at the time, though it had a surprisingly antiquated layout at AA and other airlines’ insistence. Runways pointed in all directions to enable takeoffs and landings into the wind, something that was essential for small propeller aircraft but much less important for big jets. Its main access infrastructure was to be the Van Wyck Expressway and the Belt Parkway, which couldn’t accommodate trucks. Robert Moses, who controlled most of the region’s highway planning, had no interest in a rail connection. The airport opened as Idlewild in 1948, and was renamed John F. Kennedy International after the assassination.

Compared to Newark, access to Idlewild was pretty limited. Newark was bracketed by the New Jersey Turnpike, Routes 1 and 9, Interstate 78, and the Pulaski Skyway corridor. The major highway infrastructure of the Northeast runs directly through Newark, so EWR got exceptional highway access essentially for free, as a byproduct of being located at a natural transportation nexus. JFK got one constantly congested expressway and one parkway because it was built in a swamp substantially for its address in the city limits.

The rail comparison is even starker. Newark sits directly on the Northeast Corridor, the busiest intercity rail line in the Americas, connecting Boston, New York, Philadelphia, and Washington. The connection that exists today between EWR and Penn Station via NJ Transit is useful but awkward, a product of underfunded improvisation rather than serious planning. Done properly with a rail station adjacent to the terminal rather than a rattly old people mover, you could get to Midtown or downtown via PATH in around 25 minutes. With through-running at Penn Station, you could offer Jamaica in under 45 minutes, and feeder trains from Philadelphia, Trenton, New Haven, and beyond. JFK’s AirTrain is a genuine improvement over what existed before, but it still terminates at Jamaica, itself a congested transfer point, and requires another train to reach Midtown. It ends up taking the better part of an hour. Compare this with London’s Heathrow, which is served by the Piccadilly Line, Elizabeth Line, and Heathrow Express services, or Frankfurt, which has both regional S-Bahn and ICE high-speed trains right at the terminal.

It is worst of all for freight, since the region’s logistics centre is in New Jersey and much of the airfreight arriving in New York City is destined for other parts of the country. Since the Belt Parkway is inaccessible to trucks, that means thousands of trucks are clogging the Van Wyck and Cross Bronx to take cargo flying from overseas to the rest of the US. Most of them drive right past Newark.

The decision to build Idlewild was worse for every mode. Road access was worse, rail access was incomparably worse, freight access was worse, and even terrain was arguably worse. The only thing Idlewild had that Newark didn’t was a Queens address.

To understand the effect on air travel, it helps to have a short digression on how hubs work. Nobody particularly enjoys connecting flights. But you are never going to be able to fly nonstop from everywhere to everywhere. Hubs solve this by aggregating passengers: a traveller from Louisville and a traveller from Madison, both wanting to go to Frankfurt, can share a plane from Newark, making all three flights more viable than they would be relying on Newark-bound traffic alone. The hub city benefits from more nonstops, more frequencies, and, at least when the hub is competitive rather than a monopoly fortress, lower prices driven by the efficiency of consolidated operations.

Hub airports need capacity and consolidation. Capacity means enough runways to handle the volume, sometimes compounded by “banks” that mean large numbers of flights arrive around the same time and then large numbers depart an hour or two later, so that connection times are minimized. Consolidation means that long-haul and regional feeder flights aren’t split across multiple terminals or even multiple airports. New York fails on both counts.

Atlanta’s Hartsfield and Chicago’s O’Hare each operate roughly as many flights per year as Newark and JFK combined, and they generally have fewer delays, better on-time performance, and lower ground delay costs. The reason, in part, is runway geometry. Both airports have three independent parallel runways, which means three aircraft can land or take off simultaneously on adjacent runways regardless of weather. True independent parallels require sufficient runway separation (generally 1,310m), which requires land.

In the 1950s, Newark had ample land. The marshes and industrial land around the airport that eventually became the Port Newark container port were available for airport expansion. The container port could have been accommodated to the north, south, in Bayonne, and/or in Staten Island. Instead, because there were no big plans for expansion, that land became a port, and EWR remained constrained to a runway configuration that limits its capacity in poor visibility. Even today, there is enough space for two independent parallels, though the Port Authority recently decided to build a new Terminal A right where the new runway would need to go.

Three independent parallels at a properly-developed EWR, allowing for true simultaneous independent operations in any weather, could have handled the combined traffic of JFK and EWR. Add to that the upgauging effect that comes with consolidation, since when you eliminate duplicated routes between the same city pairs served by airlines at different airports, you replace two small planes with one larger one. The combined operation is more efficient, and the per-seat costs fall.

What we have instead is a three-airport system in uneasy equilibrium, each facility optimized for a different market segment by a combination of regulation and legacy.

LaGuardia serves the regional premium market it was always suited for, but it can’t expand and its ground access degrades every year as traffic grows. The new LaGuardia central terminal, finally completed after years of construction, is a genuine improvement on the dismal original, but has done nothing to help the airfield limitations that constrain throughput and result in constant delays.

JFK became the flagship international airport and the gateway to the world. TWA built Eero Saarinen’s swooping terminal, a genuine icon that now serves as a hotel lobby. Pan Am built a circular terminal that also captured mid-century ambition. The “terminal city” concept, giving each airline its own building, seemed visionary in the jet age but became a serious operational handicap as airlines merged and operations ended up split over multiple terminals. JFK has spent decades and billions trying to rationalize a terminal layout designed for the pre-deregulation airline industry.

And Newark settled into the role of the other airport, used by New Jerseyites, United loyalists, and people who want to fly nonstop to somewhere they can’t get to from the other two airports without full hub operations. Continental built it into a real hub after acquiring People Express, which had spotted EWR’s underutilized potential in the early 1980s and built a budget operation there before being absorbed. United, after its merger with Continental, inherited that operation and runs it as a significant domestic and international hub, albeit much smaller than Chicago, Atlanta, or even Charlotte. It works, but is still constrained and delay-prone.

The original sin was that New York’s great international airport had to be in New York.

LaGuardia was built to satisfy that requirement, but its scale eventually limited it to short-haul flying. When the jet age demanded a major international facility, the Port Authority couldn’t bring itself to invest seriously in an airport that was already well-located and well-connected to the regional transportation system. It built a new one, on a swamp, because the new one would have a New York address.

Had EWR been developed instead, including the runways, the rail connection, and the integrated freight access that its location made possible, the New York metropolitan area would have a different shape today. The hub inefficiencies that fragment connecting traffic across three airports would be reduced, thousands of trucks carrying air freight to the rest of the country wouldn’t have to rumble through Queens and the Bronx spewing emissions, and the airspace congestion that makes New York one of the biggest causes of aviation delay in the country would be less severe. It would mean many more Americans would benefit from single-connection flights to the many overseas destinations served only from New York. Even better, more passengers from Washington and Philadelphia would take a train to the airport rather than a plane.

What New York needed was simply to accept, once, that a thing could be located in New Jersey and still serve New York. In a recurring theme of New York planning, that was not acceptable. Instead of one great airport, the region got three compromised ones, stitched together with crowded cab rides. The lesson of all this is that letting arbitrary jurisdictional lines define your planning choices is never the right way to go.

It’s hard not to feel nostalgia when looking at pictures of passengers flying Pan Am across the Atlantic, sitting in plush seats while eating roast beef dinners served on white tablecloths by flight attendants in glamourous outfits. The implicit comparison is with some bleak flight on Ryanair or some other ultra low-cost airline, featuring the joys of wrestling your overstuffed cabin bag into a metal sizer, fighting over an armrest the width of a twig, paying for water, and staring at ads for lottery tickets.

But the truth is, we paid dearly for the privileges of the bygone days of luxury. In 1967, an economy class flight from New York to London cost US$399. That’s equivalent to US$4,000 today, by consumer price inflation alone. If I go on Expedia today, despite the latest fuel price crisis, I can book a round-trip flight from New York to London for under US$600. (Coincidentally, business class is just over US$4,000.)

What you are actually mourning, when you stare at a vintage photograph, is not a lost era of accessible luxury. Instead, you’re looking at an era that offered luxury at luxury prices, just as today does. The only difference is that today’s business class is considerably better, with lie-flat beds, noise-cancelling headphones, champagne, and the best food that can possibly be cooked in a microwave at 35,000 feet. The golden-age nostalgia is for something that was never available to most people.

Before airlines were deregulated in the US in 1978, and somewhat later in other jurisdictions, fares and routes were tightly regulated, which largely restricted price competition. That meant the only thing airlines could compete on was on-board service, frequency, and other amenities. Once airlines came to be allowed to charge whatever the market would bear, it became clear that the vast majority of economy class passengers cared first and foremost about price.

Air travel is a hyper-competitive industry, which is a large part of why airlines so often go bankrupt (including, over the past couple decades, all three big US airlines; major European airlines like Alitalia, Swiss, and SAS; Varig in Brazil; and Air Canada; among many others). Unlike many industries, airlines tend to face some form of competition on most of their routes. As a result, their profit margins tend to be pretty thin at the best of times, and every time there’s a downturn they tend to slip deep into the red. It’s surprisingly easy to start an airline with a few leased planes (something I’ll write about later), and there’s always someone who dreams of becoming the next Michael O’Leary.

Subscribe for free, for infrastories from around the world.

Subscribe now

Passengers these days are booking online and easily comparing prices, and airlines know that most of them will pick the cheapest option. That’s why airline workers’ wages are constantly being squeezed and many airlines have figured out ways to cram more seats onto existing planes: Most economy class passengers simply aren’t willing to pay an extra hundred dollars or more for a slightly wider seat, let alone for more spacious aisles.

When you compare travel options today with the supposedly even more glamorous age of the great ocean liners, it becomes clear how far we have come.

By the standards of 1912, RMS Titanic’s steerage accommodation was considered quite good, with small shared cabins rather than large dormitories. But those cabins were still buried deep in the ship, and the Atlantic crossing took several days rather than several hours. (Also, passengers housed so far below decks were at a pretty big disadvantage when things went awry: only about a quarter of Titanic’s third-class passengers survived its sinking). The price for a one-way third-class ticket on the Titanic was about £7. That doesn’t sound like a lot, but it’s equivalent to about £700 (US$938) today (post-World War One inflation was wild). By contrast, the cheapest first-class fare was £30 one way, which works out to around £3,000 or US$4,000 today. (The most expensive suites would be well over $100,000 in current dollars.)

Today, you can fly across the Atlantic in roughly seven hours for the price of a few really nice family dinners out. It might not be very comfortable, but it’s a lot more comfortable than spending a week below decks on an ocean liner rolling in a North Atlantic storm. Even more dramatically, a return trip from London to Spain, a journey that two centuries ago was the domain of the upper class, can be done today on a discount airline for little more than the cost of a family meal at McDonald’s. Modern air travel compresses both time and cost in ways that would have seemed like science fiction to the people who saved for years to make a one-way ship crossing.

The deal on offer today is that you have, essentially, two choices. The first is to fly cheap. You will get a lightly padded seat (still more comfortable than the benches we often endure every day on the subway), and you will get from A to B. You may not get a free meal, the ability to bring a whole wardrobe of luggage, much leg or shoulder room, or any particular coddling. You will, however, get to go somewhere that previous generations would have taken weeks to reach, and you will do it for much less than they would have paid.

The second is to fly well. Pay what an economy class airline passenger would’ve paid in the 1960s and you will receive, in return, a genuinely comfortable experience. You will sleep on a flat bed, drink champagne, and eat lobster thermidor. That’s considerably nicer than coach in the 60s (when even first class seats didn’t go flat).

The genius of the modern airline industry, which goes almost entirely unacknowledged, is that it has served both the ‘Fly Cheap’ and ‘Fly Well’ markets simultaneously and effectively. Budget carriers have made the world accessible to people for whom it was previously out of reach. Premium cabins have delivered genuine luxury to those willing to pay for it. And it’s worth noting that, now that Concorde is gone, everyone arrives in exactly the same amount of time. The millionaire in the flat-bed suite and the backpacker in seat 47E touch down together.

This isn’t to say that this comes without an environmental cost. The growth of air travel has resulted in a meaningful increase in carbon emissions. Despite experiments with biofuels, most flights still run on kerosene. Nevertheless, the same things that bring down costs by squeezing more people onto planes also reduce emissions per person.

When the 777 was launched, it was designed to seat nine across quite comfortably (3-3-3). Airlines later figured out that by getting slightly narrower seats with skinnier armrests and narrower aisles, they could squeeze an extra seat in every row. This took the capacity of a 777-300ER on Air Canada, for example, from 349 seats, including 42 business and 307 economy class, to 400 seats, including 40 business, 24 premium economy (similar to old-style business class), and 336 economy. Some with a reduced business class cabin wedge in an incredible 450 passengers—more people than certain configurations of the A380. It’s not very pleasant on a trans-Pacific flight if you have broad shoulders, but it is part of the reason you can book a Vancouver-Tokyo flight right now for under C$1500, despite a global energy crisis. It also means that air travel uses far less fuel per passenger, which is not unimportant for a major source of emissions.

We want the prices of the budget era and the service of the luxury era, as if these were ever available simultaneously. They never were. The golden age of air travel that lives in our collective imagination did not exist. We got premium economy seats with fancy meals at business class prices, and normal people didn’t go on routine overseas vacations or make regular visits to their cousins across oceans. They watched movies or wrote letters.

Airlines like Ryanair have made it possible for ordinary Europeans to take weekend trips to places as far flung as Estonia or Bulgaria. At the same time, Eastern European young people who’ve moved to big Western European cities can affordably travel home regularly to see family and friends. It would take 24 hours or more to drive or take the train from Tallinn to Frankfurt. This has done more for European unification than countless EU programs.

Yes, airports are overcrowded and people are herded around like cattle. But this is because so many people can afford to fly. If you really want the classic luxury experience, you can pay the equivalent of an economy class fare in the 60s to sit in a private lounge and maybe even get chauffeured to your plane in a limo. The golden age of air travel, if it has ever existed, is not in some bygone age of glamour. It’s today, when a trip across the ocean is affordable for a middle-class family and the recently arrived immigrant can afford to fly halfway around the world to visit his ailing mother.

What’s most striking is that throughout the 1950s, including after its first subway opened in 1954, Toronto was following pretty much the same trajectory of decline as those American cities. But then something happened to change the story entirely.

On Finch Avenue East in the postwar suburbs of North York and Scarborough, a bus (both express and local) is scheduled to come better than every three minutes during rush hour, every four minutes through the middle of the day, and about every six minutes even at 11 at night. Of course it’s not always right on schedule, but even when buses bunch, there’s almost always one every few minutes. Even after midnight, it’s every half hour. It runs twenty-four hours a day, seven days a week, on a road lined with the backyard fences of bungalows and the parking lots of strip malls selling discount furniture and bubble tea. On an average weekday, about 38,700 people ride it, which is considerably more than any bus route in New York City.

Subscribe for free, for infrastories from around the world.

Subscribe now

According to conventional planning wisdom, this bus should not exist. A 2012 Ontario Ministry of Transportation report defines a minimum density of 50 jobs and residents per hectare as the threshold required to justify basic bus service every 20 to 30 minutes. The density along much of Finch East is well below that. If that standard had been applied in 1963 when the Finch bus was created, it never would have been more than a limited service for people with no alternative. And yet here it is, carrying more riders daily than entire light rail systems in American cities that spent billions building them.

Finch East isn’t unique. Wilson, Steeles West, Don Mills, Jane, Lawrence East, and many others all run through neighbourhoods that look, to any honest observer, pretty much like the suburbs of any other North American city, with strip malls, cul-de-sacs, and garages. They’re predominantly middle-class communities with high car ownership, and while they have some high-rise apartments, those came after the transit success. The transit that shouldn’t attract many riders has attracted an awful lot of them.

Toronto’s historic transit success has a mythology, and like most mythologies, it’s mostly wrong.

The standard version gives credit to Jane Jacobs, who arrived in 1968 and helped stop the Spadina Expressway in 1971. Toronto chose people over cars and transit over highways. The rest followed naturally. It’s a good story and it no doubt helped, but the problem is that Toronto’s transit revival began a decade before Jacobs arrived. The year the TTC broke away from the continental pattern of collapse was 1963, when Jacobs was still in Greenwich Village and Spadina was beginning construction.

The key event of 1963 wasn’t a highway cancellation. It was an expansion of the TTC’s suburban bus service as a political sop to suburban elected officials who really wanted the elimination of the double fare outside the old City of Toronto, a move that the TTC leadership long resisted. The 1963 service expansion created a grid of routes along the old concession roads of Scarborough, North York, and Etobicoke, putting a bus within walking distance of nearly every new house on a cul-de-sac in the fast-growing new suburbs.

Nobody expected it to work. The TTC’s own staff were certain it was a financial disaster in the making. Everyone knew that bus routes into undeveloped suburban territory were an inevitable money-loser. But within six months, the new services were astonishingly covering their costs. Within a few years, they were among the most financially productive in the system. Consultants had projected 8% ridership growth between 1962 and 1972; the actual figure was 28.7%.

The people who moved to the suburbs for a house with a big yard turned out, in enormous numbers, to be willing to walk out to the main road and catch the bus, as long as the bus came often enough that they could count on it.

That’s the Toronto model—it’s not about multibillion dollar megaprojects. Everyone knows Toronto’s subway map is much less impressive than Washington’s or San Francisco’s, but Toronto has far more riders. The Toronto model is simply a grid of frequent buses running along the old concession roads, built not from visionary planning but from political compromise and institutional necessity, succeeding far beyond anyone’s expectations. The lesson isn’t complicated: If transit is frequent and convenient enough—even if it’s slow and sometimes unreliable—even many people who can afford a car will choose to use it. It’s not a complex lesson, but it’s one that much of North America needs to learn.

An engineering problem is a problem that requires engineering to solve. A river or a mountain is in the way, or the soil is unstable. These are real constraints, and the job of the engineer is to find a way through, over, under, or around them. That is what engineers are for, and they are very good at it.

A phone call problem is a problem that looks like an engineering problem but isn’t. Another agency controls the right-of-way, a utility needs to be moved, a private company doesn’t want to cooperate, a municipality is being difficult, or a landowner is reluctant to sell. These are human problems. They’re problems of incentive, authority, and will. They have human solutions. But infrastructure projects, which are generally run as engineering projects, have a persistent tendency to treat them as engineering problems instead. The results are predictable and expensive.

Metro North, serving the New York Metropolitan Area, is a mid-sized commuter railway with around a quarter-million riders per day on its three main lines, about a fifth the ridership of the Paris RER A alone. It operates its trains out of Grand Central Terminal, which has no less than 44 platform tracks—the most of any station on the planet. By comparison, RER A’s stations generally have 2 or 4 platform tracks each. Back in the 80s, the MTA built a new two-level tunnel under the East River, with the upper level for the subway and the lower level intended to bring Long Island Rail Road trains to the Grand Central area, which is a much bigger office district (more on the history of that district in a future article) than around the main LIRR terminal at Penn Station. Though the subway opened in 1989, this being New York, it took until the early 2000s to get serious work underway to use the lower level for the LIRR.

Subscribe for free, for infrastories from around the world.

Subscribe now

There was, however, a problem. Metro North made very clear that they couldn’t possibly fit the LIRR’s additional trains when they only had 44 tracks to work with. Now, one might reasonably note that since London’s Liverpool Street Station moves significantly more trains per day on fewer than half as many platform tracks, it might just be possible for Metro North to squeeze a few of its fellow MTA railroad’s trains in. But New York may operate with different physics than the British capital.

Unfortunately, this problem was treated as an engineering problem, not a phone call problem. Metro North had given their requirements, and they now needed to design within these constraints. So the engineering solution became a gigantic cavern deep under the existing station with 8 more tracks completely disconnected from the others. It ended up costing about $12 billion and took 15 years to complete, even though much of the tunnel had already been sitting there since the 80s.

This is hardly a unique situation. The Caltrain HSR Compatibility Blog has long discussed similar problems at San Jose Diridon station, where unwillingness to share infrastructure between Caltrain and new high-speed service will force a billion-dollar elevated structure. The same problem is occurring at the other end of the line. There is serious discussion of tunnelling into Los Angeles Union Station rather than using its existing capacity and approaches. Union Station is not full and its approaches are not unusable or particularly congested by global standards. The tunnel is being driven by the difficulty of reaching agreement with the agencies and operators that control access to the existing infrastructure. This is an engineering solution to a coordination problem at a very high price.

We’ve also recently heard talk that Toronto’s Union Station has been deemed full (it’s not) and planned Alto high-speed services might have to go somewhere else.

These are all phone call problems. It may be a genuinely difficult phone call: the negotiations may be complex, the competing interests real, and the other parties not obviously wrong to want what they want, but the answer to a difficult negotiation is a better negotiator, or a more senior one, or a bigger cheque, or all three. It is not a multi-billion-dollar structure built to avoid having the conversation.

The pattern is not accidental. It reflects something structural about how infrastructure projects are organized and evaluated. When a project hits an engineering obstacle, there is a clear professional framework for responding. Engineers assess the constraint, develop options, model the costs, and recommend a solution.

When a project hits a human obstacle, the framework breaks down. Negotiation where very high-level action is needed to move the needle is often not in the toolkit. Escalating to senior political leadership requires admitting that the project has a political problem, which project managers are reluctant to do. Paying a private company to move is often perceived as rewarding obstruction. The human problem therefore gets quietly reclassified as a technical constraint, and the engineers are enlisted to find a way around it.

This reclassification is almost never made explicit. Nobody writes a memo saying, “We have decided to spend four billion dollars to avoid a difficult conversation.” It happens gradually, through a series of individually defensible decisions, each of which adds cost and complexity, until the project that emerges looks almost nothing like the one that was originally proposed.

Treating a problem as a phone call problem rather than an engineering problem does not mean that the problem is easy. It means it requires a different kind of effort. It requires political seniority. Many coordination failures in infrastructure projects persist because no single person with enough authority over all the parties involved has chosen to make them their problem. When that person shows up, such as a premier, a minister, or a mayor with genuine clout, problems that festered for years can sometimes be resolved in days or hours. The history of successful large infrastructure projects is very often a history of someone deciding that the human obstacles were their personal responsibility to clear.

One of the legendary cases was the new Munich airport. The 70s and 80s had seen fierce battles over new airport construction around the world. Tokyo’s Narita airport still has farm houses in the middle of the runways (and is missing a crosswind runway, which makes for amazing crosswind landing videos) because they couldn’t manage to buy the land. Bavaria’s long-time minister-president, Franz-Josef Strauss, took a different tactic. As the story goes, he personally knocked on the door of the farmhouse of every affected farmer with a bottle of schnapps, explained the importance of the airport for Bavaria, and hammered out a deal. The new airport opened in 1992 with, by the standards of the era, remarkably few obstacles.

Solving phone call problems requires deploying money differently. Buying out a difficult private party, compensating a utility generously, or paying an existing operator enough to make cooperation worth their while is almost always cheaper than the engineering alternative. The reluctance to do this is partly cultural, since it sometimes feels like rewarding bad behaviour, and partly bureaucratic, because procurement rules make it easier to spend money on construction than on negotiation or compensation.

And it requires a willingness to escalate. The phone call only works if the person on the other end believes that the caller has the authority to make the financial or other commitments necessary. A mid-level person negotiating can’t make compelling firm offers, and a mid-level person on the receiving end will never get fired for saying no and preserving the status quo. Sometimes getting a deal done also requires someone with the authority and the will to make non-cooperation painful. That requires political cover that project managers rarely have and rarely ask for.

There is a version of infrastructure planning that treats every human obstacle as a fixed constraint, no different from a granite formation or a flood plain. In that version, the engineer’s job is to route around whatever is in the way, at whatever cost that routing requires. This version is very expensive.

There is another version that begins by asking, for every obstacle encountered: is this a rock, or is this a person? If it is a person, what would it take to move them? And is that cheaper than building around them? The answer, almost always, is yes.

The infrastructure projects that have been built cheaply (there are many, though they rarely get the same headlines) tend to share a common feature. Someone in authority decided early that coordination failures were not requirements to be accommodated through engineering; they were problems to be solved. So they picked up the phone.

Sometimes that is true, but more often than most people realize, it is not.

The technologies that have most transformed modern life, like the microchip, the internet, GPS, and the touchscreen, were not conjured by the invisible hand. They were paid for, in their most expensive and uncertain early stages, by the United States government providing a guaranteed market.

This is the first in a series of articles about how that mechanism worked, and what it produced.

Subscribe for free, for infrastories from around the world.

Subscribe now

Before the transistor, electronics were bulky and expensive. Computers filled entire rooms, their processing done by vacuum tubes that glowed hot, burned out constantly, and consumed enormous amounts of power. Radios were large wooden cabinets that became the centrepieces of living rooms.

The transistor had the potential to radically change the shape of electronics. Bell Labs scientists invented it in 1947, demonstrating for the first time that a small solid-state device could do what a vacuum tube did, but in a much smaller, more durable, and less energy-consuming package. But a transistor was still a discrete component. A radio or a computer required hundreds or thousands of them, each one wired to the others by hand. Miniaturization had limits as long as the components themselves had to be assembled one by one.

The integrated circuit dissolved that constraint. In 1958, Jack Kilby at Texas Instruments demonstrated that multiple electronic components could be fabricated together on a single piece of germanium. A few months later Robert Noyce at Fairchild Semiconductor independently arrived at a better solution. He printed the entire circuit onto a flat silicon chip using a photographic process. Noyce’s planar method made mass production conceivable. (Chip War: The Fight for the World's Most Critical Technology by Chris Miller is an excellent source on this history, which I have of course simplified here).

Noyce was one of eight engineers who had left William Shockley’s semiconductor laboratory in 1957 to found Fairchild. Shockley, who had won the Nobel Prize for the transistor, was by then a difficult and paranoid employer. Shockley called them the Traitorous Eight, a name by which they eventually became famous as some of the founders of Silicon Valley (itself named after Noyce’s invention). Nearly every major semiconductor company that followed traces its lineage back to those eight defectors.

But in 1959, the integrated circuit was an invention in search of a market. The chips were expensive and finicky to manufacture. Nobody had yet built the specialized equipment needed to produce them reliably at scale, because nobody had a reason to invest in that equipment. And nobody had a reason to invest in that equipment because there were no customers. And there were no customers because the chips were expensive. It was a classic chicken-and-egg problem, which all too often kills promising technologies before they mature.

What breaks that kind of deadlock is a buyer willing to pay whatever the technology costs, in whatever quantities the manufacturer can produce, for long enough that the manufacturer can afford to get good at making it. In 1959, there was exactly one such buyer in the world.

On October 4, 1957, the Soviet Union launched Sputnik. It wasn’t anything more than a metal sphere about the size of a beach ball, with four radio antennas, orbiting the Earth and doing nothing more consequential than emitting a beep. But the rocket that put it there was not modest at all: The same rocket that had lofted Sputnik into orbit could lob a nuclear warhead onto an American city. And unlike a bomber, which took hours to cross the Arctic and could be intercepted, a ballistic missile arrived in thirty minutes and could not be stopped.

The strategic implications were immediate and frightening. American cities, which had spent the postwar years believing themselves protected by distance and air defenses, were suddenly vulnerable in a way they had never been. And missiles were so fast that they might destroy American bombers before they could even take off. Fears of a “missile gap” with the Soviets far ahead in ballistic missile technology and production, later shown to be exaggerated, became one of the key issues that got John F. Kennedy elected president.

The American response was the Minuteman. To give a little bit of background, the original ballistic missiles going back to the V2 were liquid fuelled. Since liquid fuel was highly corrosive, the missiles could only be fuelled a short time before launch (which could also create a scary “use it or lose it” situation once they’re fuelled). That process took longer than the flying time of a missile from the Soviet Union to the USA, meaning that they could potentially be destroyed sitting on the ground while the fuel was still getting pumped in. The solution was solid fuel, which is basically like a slow-moving explosive. That’s what changed with the Minuteman. Its solid fuel meant that, if warning of an incoming Soviet attack was received, the retaliation could be on its way within minutes. But then the American leadership decided they also wanted to be able to destroy the Soviet missiles in their hardened silos before they could be launched, rather than just aiming at cities. That meant much higher accuracy than older missiles, which people were satisfied with hitting within a few kilometres (not a problem if you’re targeting a city with a megaton-class nuclear weapon). Accuracy within a few hundred metres was incredibly technically demanding, as you’re calculating and minutely adjusting the trajectory of a missile travelling halfway around the world at Mach 20. The upgraded Minuteman II required a guidance system of extraordinary precision, but small and light enough to fit inside a missile. It required computing power that 1950s electronics could not provide.

The Air Force turned to the integrated circuit. Beginning in the early 1960s, it bought microchips in quantities and at prices that had no relationship to normal commercial logic. It paid a premium because it needed the chips, and it needed them before anyone had figured out how to make them cheaply. In doing so, it funded something the market would never have funded on its own. That guaranteed market and essentially unlimited budget enabled the development of the capital equipment, the manufacturing processes, and the engineering knowledge needed to produce integrated circuits reliably and at scale. By the time the Minuteman II entered service in 1965, American manufacturers had solved most of the hard problems of semiconductor fabrication.

The Apollo program compounded the effect. NASA’s guidance computers required the most advanced integrated circuits available, and NASA bought them in enormous quantities at whatever price was asked. At the peak of the Apollo program, the federal government was purchasing roughly half of all integrated circuits produced in the United States. No one would call it a subsidy per se, but a guaranteed market that would pay whatever the cost is even better.

In 1965, Gordon Moore, then at Fairchild, published a paper observing that the number of transistors on an integrated circuit had been doubling roughly every year (later amended to two), and predicting that this trend would continue. Moore’s Law, as it came to be called, came to be seen almost as a law of nature. But, in truth, it was a consequence of heavy investment, and the investment was a consequence of demand, and the demand came from the Pentagon and NASA.

The timing is not coincidental. Moore published his observation in 1965, the same year the Minuteman II entered service. The doubling he described had been happening since the early 1960s, precisely the years when military procurement of microchips was at its most intense. The engineers at Fairchild and Texas Instruments were not improving their processes because improvement was intrinsically rewarding. They were improving them because they had customers with deep pockets and exacting specifications who needed better chips every year.

Once the manufacturing problems were solved and production volumes were high, costs predictably plummeted. In 1964, a chip containing about sixty-four components cost about $32. But by 1971, the price of a chip containing over a thousand components was about $1. By the mid-1970s, they were selling for cents. The guaranteed market had done its work. It had paid for the transition from laboratory curiosity to mass-produced commodity, and then stepped back.

Soon, entrepreneurs figured out all kinds of new ways to use these suddenly cheap chips, from the pocket calculator eventually to the personal computer, the mobile phone, the chip in your credit card, and the sensor in your dishwasher. None of that was planned by the Air Force, nor was it anticipated by NASA; it was simply what happened when a technology that had been expensive became cheap, and the engineers and entrepreneurs of the commercial world could finally afford to use it.

The market gets credit for all of it, and the role of the government is long forgotten. If we wonder why certain promising technologies seem to take off suddenly while others languish in obscurity for decades, often a guaranteed market is the invisible hand making the takeoff possible.

You can see it this chart below. American production sat around five million barrels per day (bpd) in 2006, then climbed steeply through the shale revolution of the 2010s to over thirteen million bpd today. It is one of the most dramatic industrial reversals in modern economic history.

Hidden inside that transformation is a quieter story. While the US was becoming energy self-sufficient, one foreign supplier was not being displaced. Instead, it was expanding. Canada now accounts for roughly sixty percent of all US crude oil imports, nine times the volume of the next largest supplier. It’s been an eight-decade-long story.

My great-grandfather was a part time oil prospector in Alberta back in the 1920s. He was convinced that Alberta would become a major oil producing region. Unfortunately, he never lived to see February 13, 1947 when, after 133 consecutive dry holes, a drilling crew working for Imperial Oil struck a gusher thirty kilometres southwest of Edmonton. Ontario had been one of the world’s first oil-producing regions, with the wells at Petrolia coming in 1866, just a few years after the famed first strike at Titusville, Pennsylvania. But they never managed to find much more. This find in Alberta was different. Leduc No. 1 and subsequent discoveries were big enough to firmly establish Alberta as having internationally significant oil reserves. It was set on a path from insolvency and a federal bailout in the 1930s to becoming Canada’s wealthiest province.

Imperial Oil was a subsidiary of Standard Oil of New Jersey (which eventually became ExxonMobil after a series of mergers), and it took the lead on figuring out how to get the oil to market. The Western Canadian market was tiny and Alberta was far inland, so infrastructure to get its oil to larger markets was needed. Oil pipeline construction had boomed during the Second World War, in part because oil tankers were prime targets for U-boats, but this would be the most ambitious pipeline ever.

The Interprovincial Pipeline was incorporated and moving oil 1,850 kilometres to Superior, Wisconsin by 1950. American-owned Imperial Oil didn’t see the need for an all-Canadian route. A 1953 extension (Line 5, now the scene of major political controversy) ran east from Superior through Michigan’s Upper Peninsula, crossed at the bottom of the Straits of Mackinac, and terminated at Sarnia, Ontario, connecting Alberta crude to Canada’s petrochemical heartland (near Petrolia) along with a big market at Detroit. By 1968, a separate loop extended south through Wisconsin and Illinois to serve the Chicago refining complex. This pipeline network, now owned by Enbridge, is one of the main North American energy corridors.

Getting the pipeline built wasn’t the only problem, however. In the 1950s and early 1960s, the Persian Gulf was producing so much oil that they couldn’t find enough places to sell it. It was generally much cheaper to load a tanker in the Gulf and sail it to the East Coast or Montreal than to pipe oil across the continent.

Subscribe for free for infrastories from around the world.

Subscribe now

The solution was political. In 1961, the Diefenbaker government drew the Borden Line at the Ottawa River and created two Canadian oil markets. West of the line, meaning Ontario and everything further west, refineries would be forced to buy Alberta crude at a controlled price, sometimes as much as a third above world price. East of the line, Quebec and the Atlantic provinces continued importing from overseas. Throughout the 60s, Ontario consumers paid a substantial premium to keep Canadian oil production alive, much as the US had protected a substantial part of its domestic market with quotas for domestic and Canadian production. Then 1973 arrived and the OPEC embargo sent world prices through the roof. Suddenly Alberta’s oil was comparatively cheap, and it was their turn to feel aggrieved at the controlled price.

Northern Alberta had long been known to sit atop deposits of bitumen mixed with sand so vast they constituted, in reserve terms, the third-largest oil accumulation on earth. The problem was that extracting anything useful from this substance was extremely difficult, expensive, and environmentally destructive. Through the 1960s and 1970s, federal and provincial governments poured investment into developing the extraction technology, and Syncrude, the flagship project, came online in 1978.

The oil sands remained a marginal proposition until the 2000s, when world oil prices rose sharply enough to make them genuinely economic. What followed was one of the largest industrial booms in Canadian history. The flood of production brought back the old 1950s logistics problem of how to get the oil from Alberta to major markets. Building to the Pacific, which would allow Canada to finally sell oil on global markets, meant crossing the Rockies, which is no mean engineering feat and also meant environmental and political challenges. Eastern Canada’s market was already largely saturated, as overall North American oil demand has been growing slowly, if at all. The path of least resistance ran south into a market that seemed to have infinite appetite for imported crude and was keen, in the wake of 9/11, to wean itself off its dependence on the Middle East for energy.

First, more and more supply was directed to Midwest refineries, pushing aside oil that had previously been imported from overseas via pipeline from the Gulf Coast. Initially, multi-billion-dollar upgraders were built in Alberta to turn the bitumen, which isn’t really a liquid, into normal crude oil suitable for refineries. But later, it was determined that the bitumen could simply be diluted with liquid petroleum and shipped efficiently by pipeline directly to Midwestern refineries, which were then upgraded at great cost to handle heavy oil.

But eventually the Midwestern market was also saturated with Canadian oil that was cheaper and closer at hand than oil from overseas. The great opportunity lay further south in the biggest market of all, the immense refining complex in Texas and adjacent states that produces a huge proportion of the oil products used across the US. (The decline of East Coast and West Coast refining in favour of the Gulf is itself an interesting story I’ll write about at some point.) The Gulf Coast refining complex had been built, across decades of investment, specifically to process heavy crude. In addition to local production, it imported vast quantities from Mexico’s giant Cantarell field, which peaked in 2004 at over two million barrels per day (one of the largest oil fields in the world) before declining sharply, and from Venezuela, which also sits on the same kinds of ultra-heavy oil reserves as Alberta. But after the major battle between the Venezuelan government and the previously independently managed national oil company PDVSA in 2002, many of its skilled workers and engineers left, infrastructure decayed badly, and production began a precipitous decline from over three million bpd in 2006 to barely one million in 2025. As those suppliers faltered, a void opened in the Gulf Coast refining system just as Canada was producing more oil than its existing pipelines could handle.

Pipelines that had historically run northward, carrying imported crude from Gulf Coast tanker terminals up into the Midwest, were reversed (notably the Capline system) to carry Canadian crude south. The Enbridge mainline system, descendant of the original Interprovincial Pipeline, was expanded repeatedly. Canadian oil not only displaced Gulf imports in the Midwest but eventually flowed all the way to the Gulf refineries themselves, which turned out to have most of the equipment needed for processing Alberta bitumen.

An oil sands plant costs billions and takes years to build, but once it’s up and running, it has low marginal production costs and can operate for decades. Canadian producers could reliably undercut overseas suppliers because they had a low marginal cost of production (once the hugely expensive infrastructure was built, it costs comparatively little per barrel to operate), low shipping costs, and no geopolitical risk premium to boot. Canada was, for American refiners, effectively a domestic supplier at a discount price.

This arrangement turned out to suit the United States extremely well. With a guaranteed and cheap heavy crude feedstock flowing in from the north, American shale producers since the 2010s have been able to pump their light sweet crude and sell it abroad at world prices. The US imports the heavy crude its complex refineries need from Canada at discount prices, given their monopsony power, while exporting the light crude it produces domestically at higher world prices. Even with recent discussions of Venezuelan oil coming back onto the market, it’s not likely that it will be able to displace Alberta oil. The need to spend billions to rebuild Venezuelan infrastructure will mean that it will struggle to compete on price with Alberta’s longstanding plants and decades-old infrastructure.

It’s a great deal for the United States, but a bit more of a mixed bag for Canada. The way Canada’s oil infrastructure is built, nearly the only feasible international market is the United States. Imperial Oil built the pipeline in such a way that even Canadian oil markets are served via the United States. Leaving aside the environmental challenges of oil production, this means that Canadian producers pretty much have to accept the price that US buyers are interested in paying, which has often been far less than the world price. That’s why the opening of the expanded TransMountain pipeline to the Pacific, which, for the first time, allows for large scale exports of Canadian oil to world markets, has led to such a big jump in the overall price for Alberta oil. But even now, it can only carry a fraction of Alberta’s production. As in so many things, infrastructure dictates economics and, for the foreseeable future (likely as long as we’re still consuming oil), the Canadian and American markets are locked together.