A short post on short trains
The lesser urbanism hack
Epistemic status: Main part is well-supported but may have some minor errors. The parts about potential future lines are inherently speculative.
Small Train is Good Train
A while ago, I wrote about how elevated trains are the greatest urbanism cheat code, increasing the amount of track miles you can build per dollar (or per year) by a factor of 2-4. And while I don’t have anything else on that order of magnitude, I do have one more easy 20-50% gain1: Run shorter trains.
The basic idea is simple: The single biggest cost of any metro system is the stations, whose cost scales with size. Therefore, if we run a system for smaller trains, we can build smaller stations for these trains, saving a huge amount on station costs. This costs us in reduced total capacity, but this can easily be made up for by increasing train frequency.
This also has some non-obvious advantages: Smaller systems let us have smaller, more incremental transit projects, gaining project experience and institutional knowledge for subsequent projects. Especially in places like the US, which don’t have recent experience in building infrastructure, it’s better to start small where possible.
It’s also just better to split up your capacity as much as reasonably possible. Having to wait for a train that comes every three minutes is a lot better than waiting for an equally-crowded train that comes every six minutes. Wait times are the second-most underrated component in making metro systems effectively faster (right after street to platform time).
Some details
Cost and Timeline improvements
Stations usually dominate metro construction costs (23 – 62 % of hard cost on the Transitcosts European projects sample, mostly depending on underground vs elevated) and can be made roughly half as long (50-55 meter platforms vs ~110 meters) for low-capacity short trains. Station construction costs scale roughly with station size, so cutting station size in half implies a total cost reduction of roughly 20% on underground lines2 and 15% on elevated lines.
Timeline improvements are even sharper. Since the stations are the single biggest (and thus longest to build) component of the system, anything that makes them easier to build makes the whole system open faster. You can’t quite build a station twice as fast if it’s twice as small, but you can typically hit 30-40% timeline reduction (3-4 years in absolute terms) on a medium-sized project (e.g. a 10-15km subway line) where stations are on the critical path.
Capacity
Capacity limits are why we didn’t always do this: Train operators are expensive3, which limits the number of trains you can run since it makes it more efficient to run fewer longer trains. It’s the invention of fully driverless modern trains that makes this obsolete, making running more trains basically free4. But now that we can run fully automated trains, the only reason to run longer trains is if we’re hitting hard capacity limits after maximizing frequency.
This isn’t impossible - it happens in Hong Kong, and can easily happen in other dense places like Manhattan - but it’s uncommon. 3-car trains running at 30-40 trains per hour (a normal peak frequency for automated or even some human-driven metro lines) reach a capacity of about 18,000 passengers per hour per direction, well above the expected demand of any American line that doesn’t go through Manhattan. Outside of the US, most Israeli or European lines would also be served by this level of capacity5. Experience with lines of this nature shows they’re fairly robust to passenger surges.
Finally, it’s worth noting that small automated trains are the easiest way to keep good weekend service. Systems with large trains typically have borderline unusable frequencies on weekends (sadly true for most American systems); switching to small frequent trains lets you run usable frequencies even when you’re running with low total capacity.
Examples, and a tangent on New Jersey
This isn’t a theoretical idea. It’s the concept behind North America’s greatest transit project of the last century, the Vancouver Skytrain
Other great examples include London’s DLR and the Copenhagen Metro. All three were successfully built at relatively low cost (which at least for the first two is extremely unusual, given that both Canada and the UK are almost as bad as the US at infrastructure).
Where else could we replace existing plans with smaller more frequent trains? Basically every currently planned project in America (and most of those in Israel). Asking GPT for examples got me this table
But this doesn’t include my favorite possibility, which is Jersey City.
New Jersey has severely underrated potential to be the next great American city: It’s right across the Hudson from New York, so it could gradually bleed over people in jobs without having to build a new center in the middle of nowhere. While it has its political dysfunctions, New York levels of political dysfunction at both the city and state level are extreme6. There’s a severe housing shortage pushing people out, which Jersey City and its surrounding towns could take advantage of.
Jersey has a lot of wins. It’s visibly cleaner, nicer and with less homelessness. It builds a lot more housing (Jersey City built around 13 housing units per 1000 residents in 2020-2023, compared to 4/1000 for NYC). It has lower crime and (slightly) lower taxes. And it’s right across the river. While NYC proper has been losing population, Jersey City has been gaining.
The biggest barrier to scaling up Jersey is transportation. Commuting to Manhattan is difficult (Manhattan doesn’t have the road capacity for mass driving, the PATH is kind of terrible, and ferries and buses are too low-frequency to be convenient). We could solve this by moving more jobs into Jersey, but then we’ve just moved the commute burden further west.
For Jersey City to become the next New York, it needs the thing that makes New York unique in North America: An actual large-scale metro system. We can’t build one from scratch anymore, but building a small number of small-station lines lets us do it one step at a time, then scale up as the city grows. And if Jersey can build a transportation system that’s both more convenient and nicer than New York’s, it can form an alternative, less dysfunctional hub for the northeast. If we’re really lucky, it could even push NYC to try to get its act together.
Failure modes
The final warning here is the failure mode: Don’t build “small trains” just to build a small train (because you don’t have need for more capacity). When building small trains, do it by actually keeping everything small and simple. Otherwise you end up with the Macau LRT - a system with giant tracks, giant stations, and everything needlessly huge and expensive except the trains themselves. You’ve gained almost nothing in cost or simplicity in exchange for the lower capacity.
On the other hand, the actual trains it uses are beautiful. I don’t have the style knowledge to explain why, but their (japanese-designed) trains are just prettier than the chinese-designed trains of the Hong Kong MTR7.
Or about a billion dollars in savings on a 10-mile subway line.
about $80 million dollars per route-mile
It tends to surprise people that operator salaries are a significant issue, but operators are the single biggest line-item of most transit systems. In New York labor is the MTA’s single biggest operating cost (roughly 58 % of the MTA’s entire expense base), and roughly three‑quarters of that is vehicle operators.
Once you eliminate drivers, the main costs are things like station maintenance and central operations (unaffected by number of trains), power (scales roughly linearly with total train-mass - so running twice as many half length trains costs about the same as fewer longer trains), and maintenance (goes up slightly for more short trains vs a few long ones - there’s more things to break down - but not by much). Even adding more total train cars is relatively cheap - without drivers, central operations account for roughly half the operating costs, which means that even doubling the number of trains only raises the operating costs by at most 50% - and without drivers, you’re paying 25% less than you were when you had drivers while running four times as many trains.
Downtown Tel Aviv lines might have issues - especially with Israel’s growing population - but this is probably addressable by building more lines, which would be a good idea since the area is fairly noncentralized.
And, given the current state of the mayoral race, likely to get much worse in the near future.
There’s a more subtle complaint about my whole approach, which goes “by hyperfocusing on efficiency like this, aren’t we sacrificing the ability to build beautiful monumental stations like Grand Central or the Moscow metro?”
I’m actually not entirely unsympathetic to this, but I think it’s mostly wrong. Grand beauty comes from using excess capacity well, not from raising requirements. Although the Moscow Metro is pretty impressive.
Short automated trains at 90 second frequencies are what I dream about.
Small trains and small stations are good until they aren't, and then you're screwed.
Singapore built the orbital Circle MRT line exactly as you suggest: small trains (3 carriages, vs 6-8 on other line), small stations to fit these trains, frequent fully automated operation.
However, the line turned out to be much more popular than planners anticipated, with the result that at rush hour, it's common to have to queue just to enter the platform where you need to wait for multiple trains to arrive before you can squeeze in.
And the tricky bit is that there is essentially nothing you can do now to fix it. There's a hard physical limit (around 90 secs) on how fast the cycle of a train arriving, people getting on and off, and departing can get, and retrofitting all 30+ deep underground stations to be larger would be an insanely expensive and disruptive exercise.