Tuesday, September 23, 2014

Transit and congestion part 2: how congestion can cripple transit

In the previous article, I said that transit isn't really a solution to congestion, but a way to increase the capacity of congested roads and avoid congestion from choking the life out of urban areas. Basically, I spoke of the effect of transit on congestion, but it is very important not to underestimate the effect of congestion on transit viability.

Indeed, speed is vital for transit. Not only relative speed to other modes of travel as speed is probably the greatest factor to determine what mode of travel people will opt for, but also absolute speed, and for a simple reason: the cost of transit is directly impacted by the speed at which they travel. That is because a major part of the cost of transit is the driver, and he is paid by the hour. Furthermore, the slower each vehicle goes, the more vehicles you need to buy to be able offer the same capacity.

I know it is evident for most people who read this, but let me explain in detail through an example.

Let's take a given bus route:
  • It is 10-kilometer long (a bit over 6 miles)
  • It is a round-trip
  • To satisfy demand, you need to have a frequency of 4 buses per hour (capacity for a transit line = frequency of the vehicles times the capacity of each vehicle)
  • You can expect an average speed of 15 km/h (around 10 mph)
  • You have 400 riders who will take the bus
  • We will neglect the issue of layover for clarity, buses that end the route can start again just as fast as they arrive
Let's simulate this for 3 hours of the peak period.

Each line represents a bus going from the start of the route (0) to the end (10)
 In this case, at the most, you have three buses running on the route, so you need to buy 3 buses, each costing 400 000$, for an initial capital cost of 1 200 000$. During the three hours, buses are running in total for 7,25 hours. If each bus costs you 150$ per hour of running time to run, then that is about 1 100$.

So, to sum up, to provide this level of service and this capacity you need:
  • 3 buses, so initial capital cost of 1 200 000$
  • 7,25 vehicle-hours of running time, so 1 100$ operating cost
  • To pay back operating costs, each ticket must cost 2,75$
Now imagine that instead of an average speed of 15 km/h, you have an average speed of 12 km/h (7,5 mph), just 20% slower. What happens? Well, instead of taking 40 minutes to make the entire route, buses now need 50 minutes to do so.
The result? You need 4 buses instead of 3, as at the worst time, you will have 4 buses running the route at the same time, you will also require 8,85 vehicle-hours of running time for buses. In short:
  • The capital cost is now 33% higher, to 1 600 000$
  • The operating cost is 22% higher at 1 350$
  • To pay back operating costs, each ticket must cost 3,38$ if all 400 riders stick to the bus...
  • ...but if the slow speed convinces say 50 of them to stop taking the bus, the cost per remaining rider is 3,86$, 40% higher than previously
So just being 20% slower has just increased costs by around 25% and may increase ticket prices by even more than that if the slow speed gets people back into cars (and buses have to self-fund). The problem is evident: transit needs to be fast to be cost-effective, if transit is slow, then it is going to be unattractive and very expensive.

Overall, in this case, the cost of the transit service in relation to speed can be represented the following way:
Cost per passenger of transit service depending on whether we suppose ridership to be static (always 400 during peak hours) or dynamic (4% gain or loss per 1 km/h difference in average speed)
In practice, the effects of faster service are probably not as pronounced because of layover time between routes for bus drivers and the like, but the dynamic is still there.

Case study: streetcars

This is likely what happened to old streetcars. When they were introduced at first, roads were largely empty of vehicular traffic, with only pedestrians and a few horse-drawn vehicles around, generally keeping to the sides of the street or to the sidewalk. As traffic speed was about 5 km/h (3 mph), ie walking speed, there was no need for signalization at the vast majority of intersections. No traffic lights and no stop sign, some cops directing traffic at a few major intersections. So streetcars in that context almost only had to stop to allow passengers to get on or get off. Even if they couldn't quite go fast, their average speed was still pretty decent. Since labor was cheap, they even had another employee besides the driver, the conductor, whose job was to collect the fare from riders, so that passengers could pay the fare after boarding and thus avoid losing time dwelling at stops.

Some sources mention an average speed of 10 to 12 miles per hour (16 to 20 km/h) for streetcars in the US in the early 20th century (Source and source). Though it doesn't look that fast, take into account that the average speed of buses in New York City nowadays is 7,5 mph (12 km/h). In Montréal too, the average speed of buses in Montréal during the peak hours (when most people take transit) is about 12-13 km/h, around 7-8 mph. Despite nearly 100 years of technological advancement, transit is slower than back then.

Though data is scarce on the early 20th century (at least as far as I can find on the internet), most sources mention growing congestion as more and more cars came on the road. Traffic lights and stop signs started appearing in the 1910s and 1920s to deal with the higher vehicular traffic. These signalizations, though they ordered traffic, also had the effect (and still do) of causing delays on arterial roads. Even a well-programmed traffic light at a relatively little used intersection will frequently incur around 10-20 seconds of delay on average to cars passing through. So if you have one traffic light every 500 meters (around a third of a mile) on a 10-kilometer trip (6-mile), you will likely lose 3 to 6 minutes at traffic lights. So if you used to be able to do that length at an average speed of 20 km/h (12 mph), now you could have seen your speed actually fall down to 16-18 km/h (10-11 mph).

I imagine visually this is how traffic should have evolved:

Streetcars with mainly horse-drawn carriages and plenty of pedestrians (not shown), traffic is generally fluid due to few vehicles and low speeds of vehicles, allowing them to easily negotiate intersections

Cars become much more common, now there is friction at intersections and police directing traffic

Finally, we start adding signals to intersections

In old sources, often streetcars were blamed for congestion, due to the windshield perspective of those who used to have access to the media and the attention of the powerful. This is for a simple reason: most streetcars actually had tracks in the center of the street, in the left lane, the fast lane. How did passengers board? Well, at first, passengers waited in the street, forcing cars to slow down not to hit them. As traffic became more ordered, here is how it worked, and still works in Toronto:

Is it any wonder then that car drivers seethed at the sight of streetcars that blocked two lanes of cars when allowing passengers to board or alight? And why car drivers were so receptive to arguments for replacing streetcars by buses?

So, anyway, some posts back, I made a graph about the average speed of transit lines depending on how many stops they had to make and their maximum speed. I will re-use this to illustrate the effect of congestion on streetcar lines.

First, the original situation where streetcars largely had to stop only for passengers and faced little to no congestion to slow them down.
Initial situation, where streetcars only had to stop for passengers and faced no congestion
Now, after cars started congesting city streets and imposed traffic lights and stop signs, streetcars stop more often and go slower
Now combine this fact with the earlier fact about how slower transit is more expensive to run for the same capacity. This means that an operator in this situation has a choice of either reducing the capacity of transit lines by 30 to 40% or increasing the number of vehicles by 50% to maintain the same capacity. No matter what, it results in an increase of costs per passenger of at least 50%.

More importantly, this thought experiment reveals that streetcars, far from causing congestion, were actually a VICTIM of the congestion caused by cars. Car congestion increased transit operating costs and lowered their average speed, making them more expensive and less attractive at the same time.

Streetcar companies, that used to be profitable, started requiring influx of public funds to stay afloat, many were bought by cities. Even when the streetcars were replaced by buses, it didn't help matters, and transit companies, now public, often require up to 70% of funds to come from public money and not fares. A result of three factors:
  • Less ridership
  • More expensive operations as the lines are now slower
  • Transit as welfare policy in cities now built around motorized transport

Lesson for future surface transit

The lesson we can take from these facts and what happened to streetcars should be clear: in highly motorized cities, transit requires reserved right-of-way to protect itself from congestion. In some small cities, maybe they don't need it yet, but if they plan on growing, they should do it as quickly as they can, because when congestion becomes common, taking any space away from cars will be more and more politically contentious.

This is a lesson I think developing cities especially need to take to heart. As people start being able to afford motorbikes or cars, they risk overwhelming streets and causing monstrous congestion. When this happens, bus services running in mixed traffic will suffer most, because the speed reduction will result in transit companies either requiring massive subsidies to keep running, or going bankrupt, thus pushing people into private motorized transport. Even worse, those who will not be able to afford motorized transport will have to walk in congested roads where they are likely to be hit, wounded or killed. Congestion may also choke cities and lead to good jobs and wealthy residents fleeing to sprawl, copying the North American mode of development that left desperate poor people in urban areas.

Considering this, I think developing cities need to implement transit-only lanes and BRT quickly, because it will only get harder to implement as more and more people get motorized. BRT exists here to essentially protect buses from the congestion that cars and motorbikes create and safeguard transit riders' access to an acceptably fast and safe form of transport. If buses are left in mixed traffic, transit use will simply evaporate over time.

For developed cities, modern tramways on their own ROW or even better, grade-separated LRTs and subways, should be preferable, because their lower labor requirement means they will cost less in the long-term and be more attractive to people who already have cars. But even bus lanes to go around congestion will do a lot of good in order to provide for more affordable transit that is more attractive to people.

Eventually, developing countries should also invest in rail, I think, for as they become richer, labor will become more expensive and the buses that used to be cheap will become more and more expensive. Rail can avoid that issue, and also provide for much higher capacity for less space.


  1. In highly motorized cities, transit requires right-of-way to protect itself from congestion
    Absolutely. Even bus bulbs, which save the bus from having to wait for cars to pass before it can pull out into the travel lane, can make a real difference. One interesting thing about this, though, is that a lot of the congestion comes at certain very specific bottlenecks, and bypassing that can give most of the benefit of a fully dedicated right of way. And it won't necessarily impact the traffic throughput that much either, if the space being taken is just used to store cars queueing to get through that bottleneck.

    Also, somewhat relatedly: the modern trend of putting streetcars into the curb lanes of roads is thus the height of stupidity, as that's where there's the most friction with cars parking (or just stopping to drop someone off), cyclists, etc. One of the few advantages of the traditional streetcar model is that at least they run in the "fast" center lane, as opposed to a bus that has to deal with curbside friction. Curb lane streetcars are the worst of both worlds.

  2. I wouldn't say curb lane streetcars are necessarily the worst of both worlds. Yes there's friction with parking cars, but cyclists tend to avoid tracks, and with good reason. Also, unless there's a dedicated right-of-way in the middle of the street, then a left lane streetcar has conflicts with left turning cars. It's also not practical to restrict cars from turning left at every intersection, so there's going to be a conflict point somewhere.

    Anyway, great post, and I'm glad to hear some of my own thoughts about the historical streetcar pattern being vindicated. I think this is one instance where the excessively wide "hypertrophic" streets of the 19th century North American cities was a benefit, at least at first, because in the busy downtown city cores they could put in multiple tracks if necessary and there'd still be room for horses and carriages and people to mill about in the street and even park without getting in the way. Of course, those same wide streets also invited automobile usage, especially outside downtowns where they were still excessively wide but also mostly empty. At least there they usually installed some island platforms for streetcar stops, but those became obstacles for cars and put waiting riders in a rather precarious position.

    It's often forgotten that a slow running speed can still add up to a decent average speed when there's few stops, like the aforementioned historical streetcar systems. The opposite is also true, where a high speed limit on a suburban arterial stroad still results in slow average speeds because of all the stop lights and traffic backups. I don't know what the average/schedule speed was, but Cincinnati's cable car lines (there were 3 independent companies) operated with cable speeds of 8 mph on the inner leg and 10 mph on the outer (the cable power houses were in the middle, about 2 miles from downtown). That sounds pretty lousy, but again, if they only had to stop to pick up and let people off, then their average could be pretty close to the running speed. I'd also add that the historical streetcar systems, due to the lower labor costs, also maintained incredibly tight frequencies compared to today. Again in Cincinnati, off-peak outer lines still had headways of around 20 minutes, and that quickly became 10 minutes as you get closer in. So not only does this make the service immensely more convenient, it also speeds up running time because each individual streetcar doesn't have to stop as often to pick up or let off passengers. Poorly used bus lines of today can have that advantage too, but since they're usually slave to an infrequent schedule, they have to wait at certain points to prevent getting too far ahead, thus negating the benefit.