# In Transit, Inefficiency is Exclusion

That’s the real title of my piece in Citylab today.  Read it here.

### 18 Responses to In Transit, Inefficiency is Exclusion

1. Peter L June 12, 2018 at 10:56 am #

Is it just me or are there a whole lotta people that confuse the word “efficient” with the word “convenient”?

• el_slapper June 14, 2018 at 6:21 am #

I would rephrase it : they confuse “efficient” with “convenient for me”. But you’re right, indeed.

2. David June 12, 2018 at 12:57 pm #

I read the piece at Citylab and my first reaction was also that you’d be putting a lot fewer people in those tunnels. I was curious what the numbers on that were, so I did some back of the envelope calculations.

You can fit one train per signal block, obviously, and Wikipedia says that a block in NYC is 300m long. Choosing more or less at random, let’s use a train consisting of 10 R179B cars. That gives you 2580 passengers per block.

What about cars? Let’s say you have something the size of a Toyota Corolla. A Corolla is 4.6 meters long and holds 5 passengers. Double the length to allow for spacing between the cars and you get 32 cars per block, or 160 passengers. This is comically fewer people — 6% of the subway numbers. Even the the cars ran touching each other it’d only be 320 people.

But cars aren’t that dense. What about busses in the tunnels? If we use the new Proterra BE40 electric busses that NYC is testing we have a length of 12m. Double, again, to space them out for 300/24 or about 12 busses per block. Proterra’s web site says each bus holds 40 passengers, so you have 480 passengers per block – 18% of subway capacity.

Even if you could run the busses five times faster than the existing subway (seems unlikely and possibly terrifying) you’d still not quite get up to current throughput.

So, the idea of running cars in the subway isn’t just a bad idea, it’s absurdly ill thought out. It took me literally five minutes to run this calculation. I probably made some mistakes and incorrect assumptions, but still, it’s going to be close. It frustrating to see people promoting transit ideas without putting even the slightest bit of thought into them.

• Anonymous June 12, 2018 at 9:39 pm #

Unless you want the cars to crawl through you’ll need a lot more than a 1 car length following distance, unless of course you like the idea of car crashes.

At 60 km/h you’ll need more like 30 m following distance unless the cars are driven by computers.

Train following distance is even longer due to the requirement that a train be able to stop completely in the following distance of the previous one but it’ll still be much higher capacity and except for high speed lines dwell time comprises a lot of that.

• Peter L June 13, 2018 at 8:52 am #

The #RoboCar true believers think that cars will be “fleeted” so that the following distance is inconsequential. I’ve always maintained that if that actually worked, then there is no reason not to apply the same technology to trains (or even buses). Of course, stopping a car that’s “fleeted” is waved away as being a minor technical issue to deal with.

• Anonymous June 14, 2018 at 12:32 am #

If you were serious about increasing capacity with robocars and could keep meatbag driven cars off the roads then not only could you set the following distance to bumper to bumper (not completely since cars are going to have to merge and demerge at some point but once merged you’d probably want them up close to get the aerodynamic benefits) but you could make the cars a lot smaller due to not bothering with any crumple zones.

A robotaxi operator running in a road network without human drivers would be likely to have a fleet mostly consisting of single or two seat vehicles given the average occupancy of 1.2 which may go down if children and oldies no longer need to be driven by a human.

Something like the Renault Twizy would allow for much greater capacity since you fit two cars in one lane (could be a bit smaller if you can trust the computers enough not to crash).

That trains can derail means it’s unlikely that the stopping distance standard would be lowered, might be easier getting an exemption for monorails, most types of which require catastrophic damage to the track to derail but monorail switches are slower and dwell times even longer so only Personal or Group Rapid Transit with offline stations and steering done by the vehicle could actually make use of headways measured in seconds or less.

• David June 13, 2018 at 6:08 pm #

For an autonomous car a single car length is probably plenty. But the point is that it doesn’t matter.

• JJJJ June 13, 2018 at 10:29 am #

David, just a minor correction on the buses. Thats 40 seated, so around 60 including standees per bus

• David June 13, 2018 at 6:00 pm #

I thought 40 seemed low. So that’s about 720 passengers per block — 28% of subway capacity. Better.

• Georgist Economist June 20, 2018 at 3:36 pm #

I think it’s misleading to start from vehicle dimensions and spacing-in-distance. For theoretical calculations of throughput, the main factors are:
1) “length fill ratio” given by spacing-in-time, determined by safety considerations;
2) vehicle width;
3) type of interior layout (yes, really, read below).

I think the points are easier to understand in reverse order.
3) A metro train with many doors and purposefully wide aisles packs way more (standing) passengers onto each square meter of floorspace than a long-distance train full of seating. This is important not just for throughput, but also for things like “over-capacity shock”: if less space is left over after all seats are taken, additional passengers increase apparent overcrowding much faster. But the key thing is that once you fix the interior design, all you need to compare is the quantity of floorspace moving.
If you want to really grok this concept, I recommend this video: https://www.youtube.com/watch?v=BzB5xtGGsTc
In the limit case, you could possibly want to strip out seats altogether, whether the technology is bus, metro or anything else—if passengers take short enough trips that they don’t mind. Multifunction spaces (pram, wheelchair, etc.) are your friend.

2) I think it is pretty obvious that the width of the vehicle is independent of its speed, acceleration, etc. While digging larger tunnels is more expensive, building any tunnel at all is very expensive in the first place. Building tunnels for a one-larger-size technology (mainline train, metro, bus/LRT, car) can be a good deal, especially if they can connect to surface-running stretches.

1) Technology and safety requirements limit how much time must elapse between the *end* of one vehicle and the *front* of the next passing. (For everything railbound: the next vehicle must be able to come to a halt before something suddenly falling off the end of the previous. For road vehicles, this is not required, even though in tunnels they can’t swerve, either.) Note that simply making the vehicles longer can push the fraction of line length that is carrying passengers upwards. Using short cars is a bad idea, ceteris paribus. However, technological solutions (e.g. platooning) *can* dramatically increase this fraction, thus the techies pushing them *have a point*. Let’s make a list:
– better deceleration (braking) improves throughput;
– automation (no lag between detection and brake application, no sight range limits) dramatically improves throughput;
– avoiding braking requirements entirely because they don’t apply to rubber-tyre vehicles is a hack;
– longer vehicles improve throughput if they need to be widely spaced (trains), are irrelevant if not (platooning cars).

Putting the points together: if the rules are not fixed, the highest throughput can be achieved with automatically-driven articulated buses with decreased number of seats. If they are fixed, and railway equipment catches up to platooning, then trains are a better bet, partly because they are wider.

Of course, this is wildly theoretical. In practice, platform dwell is a crucial factor in determining achievable throughput. There are some nice tricks, such as having a platform on both sides of the train, and designating one as for alighting and another for boarding, so the passengers getting on and off don’t block each other’s way. And even then, it’s quite possible to move so many people to a station that they can’t exit the station fast enough, because walking on stairs and escalators have insufficient throughput. Or, speaking more broadly, because the next link in the trip is overwhelmed by the passengers trying to make a connection. Say, a train connecting to a bus is a PITA, because no matter how frequently the bus runs, all the passengers from the train will want to get onto the one bus that comes first—in my city, an ongoing metro renovation and, in the peak, ~1.5 artic buses/minute(!!) demonstrates this. The other way around is OK.

3. Rick R June 14, 2018 at 5:02 pm #

But which definition of Efficient should be used?

Least number of vehicles?
Least number of staff?
Least distance travelled by vehicles?
Least fuel/energy used?
Customer wait time at stops?

You can’t minimise all of these, even trying to prioritise two can lead to contradictory results. I comes down to those who make the decisions on what the transport system is trying to achieve coming up with clear priorities, explaining them and setting relevant KPI for operators to follow.

I often read outraged commentary on how those running some transport system are fools for not meeting the definition of efficiency the author holds closest to their hearts. It helps if the priorities of the system can easily be explained.

• asdf2 June 14, 2018 at 9:44 pm #

Taxpayer dollars spent per rider. Money is the ultimate constraint that everything boils down to.

• Jack June 18, 2018 at 5:02 am #

Well said, asdf2.

On Rick’s list, 1, 2, 3, 4, and 6 are all efforts to minimize something that cost money. Customer wait time is convenience, and least distance traveled might be convenience if it affects ride time. So, the two factors being weighed here are really efficiency, which is about the group’s total resource usage, versus convenience, which is about an individual’s resource usage (usually time). That gets back to what Jarrett was saying in this article: convenience and efficiency pull at each other, and any pull towards convenience equals a greater amount of the group’s resources being spent on fewer people, which inherently means those with the fewest resources of their own suffer the most.

4. Cererean June 15, 2018 at 8:02 am #

Cars are a bad idea, but what about electric scooters running in underground tunnels?

Okay, for a big city that won’t be workable, but for a small city it might be a practical system that’s sort of like PRT. Bike lanes, but underground and so sheltered from the weather, with a 12mph speed limit.

• Anonymous June 16, 2018 at 2:31 am #

Given the extra cost of underground construction over elevated it’d probably be more expensive than PRT.

I wouldn’t call cars a bad idea, they do work pretty well if density is low and even with high density there are still plenty of circumstances in which they work better than anything else, it’s just that they don’t scale to high density very well so anywhere with high density will have to have the car as a minority transport mode.

5. Georgist Economist June 20, 2018 at 4:05 pm #

“There’s also labor efficiency, but automation of high-capacity transit is always possible if we decide we care about that.”
Is labor efficiency really a problem? A single train driver carries as many passengers as ~10 bus drivers, as many as ~1000 car drivers.

“In cities, the only truly finite resource is space.”
As you correctly explain in several articles, space is created by door-to-door speeds. Lower headways, lower platform/stop dwell (proof-of-payment), longer stop spacing quite literally create more space.

As far as the petrolhead proposal can be considered to have any thought behind it, it does make a few points:
– due to obvious reasons of friction, rubber-on-asphalt vehicles can pull higher deceleration than steel-on-steel vehicles, thus can be packed with less time elapsed between them (though technological improvements like magnetic rail brakes can even the field somewhat);
– due to a quirk of regulation, road vehicles are allowed to be packed more closely than rail vehicles even relative to the stopping distance;
– due to same, road vehicle control technology (e.g everything necessary for platooning) are more developed, and cannot simply be copied over;
– road vehicles tend to be better motorized (power/mass, etc.) than rail vehicles, giving higher acceleration as well.

• Anonymous June 21, 2018 at 12:45 pm #

Acceleration of public transport vehicles is limited by the existed of standing passengers, if you required everyone to wear a seatbelt and banned standees then you could out accelerate sports cars (maybe not if you’re relying on steel on steel but LIMs certainly could if you applied enough power).

On train drivers, a peak hour train may carry a thousand people per staff member but what about off-peak?

• asdf2 June 23, 2018 at 5:24 pm #

“Is labor efficiency really a problem? A single train driver carries as many passengers as ~10 bus drivers, as many as ~1000 car drivers.”

During peak hours, generally, no, but during off-peak hours, absolutely, yes. The problem is that in order to get peak labor efficiency, you need to fill all (or nearly all) of the seats. If 6,000 people want to travel the line every hour, you can run the train every 10 minutes and get good-quality service. But, at a time of day when only 1,000 people want to travel per hour, you can either run a train every 10 minutes that’s 1/6th full, and cough up the money for the labor inefficiency. Or, you can run the train once per hour, which will be fuller, but cost the agency less. Of course, running the train less often means a service that’s less convenient for riders, which means a service that fewer people will want to ride, so even the once-per-hour train still won’t be full.

This is a problem with commuter rail systems all over the country – they spend all their money on capital to get the train set up, and leave nothing left to spend on labor for more than a skeletal service outside of rush hour – if the line runs at all.

It is the cost of labor that dictates how often the train can run during the hours when the train will not be filled to capacity, as well as hour many hours per day or days per week, the train is in service.

See the Vancouver Skytrain for an example of how frequent all-day, seven-day-a-week service can be when the trains are automated, and the cost of labor is not a concern.