We’ve been working with San Francisco Municipal Transportation Agency on the design of the next phase of their service restoration. Read all about it here!
11 Responses to San Francisco: Options for Service Restoration
You are here: Home › San Francisco: Options for Service Restoration
Bus routes with 5-minute frequency is often a waste of money. Compared to 10-minute routes, you double the cost but you only get 2.5 minutes less average waiting time. When you have a 5-minute bus route, it’s usually a good idea to split it into two 10-minute routes. There are different ways this can be done:
a) Have one route that make all stops and another route that only makes major stops. This has been done already by Muni with the “Rapid” routes. The shorter riding time sometimes outweighs the longer waiting time.
b) Create new routes for direct access to different parts of the city without the need to transfer. Such routes will often duplicate existing routes for most of the length. This is often frowned upon in transit planning circles, but it can in fact lead to less average waiting+transferring time if you have already have very high frequency.
In any case, it’s very important that any routes that parallels each other have equal frequency. It’s never a good idea to split a 5-minute route into a 7-minute and a 20-minute route, since most passengers will choose the 7-minute route and the 20-minute route will run with empty buses. This was the reason why the 2 Clement was unsuccessful.
Also, one problem with 5-minute routes is bunching. When you have two buses on the same route next to each other, only one of them is useful. The other one is a waste of money. When you have two buses on different routes next to each other, both of them are useful, since they target different passengers.
There is diminishing returns when you increase frequency, but little evidence that 10 minutes is a great cut-off point. If there is an international standard, it is 6 minutes. The main advantage of 5 minute frequency is that it provides good clock face scheduling. Often it is used because of branching. Two (lightly used) 10 minute branches combine for 5 minute service on the core. But the main reason they choose 10 minutes (instead of 12 or 8) is because of the clock. The trips aren’t expected to be spontaneous. People are expected to know the schedule, and plan accordingly. This isn’t “freedom”.
The New York Subway runs every 10 minutes because of this branching. This should provide 5 minute frequency within the core, but doesn’t, because of delays. It works great in Berlin, which has 5 minute frequency within the Ring (for the most part). The Paris Metro is mostly a 5 minute system. Toronto runs the two key lines every 5 minutes off peak. There is great value in going from 10 to 6, which leads to articles like this: https://pedestrianobservations.com/2020/08/21/new-york-as-a-six-minute-city/.
The difference between five and six minutes is minimal in this case. Buses tend to be less reliable, so you gain little in terms of scheduling. The main benefit in this case is in promoting the change. A “5-minute network” just sounds a lot better than a “6-minute network”. Hopefully Jarrett can comment on this (e. g. “Why not a 6-minute network?).
The other thing I would like Jarrett to chime in on is the literature for frequency and ridership. I found this: http://onlinepubs.trb.org/Onlinepubs/trr/1981/818/818-003.pdf, which confirms that ridership still increases when frequency increases beyond 10 minutes. There are plenty of examples of people writing about frequency (“up to six minutes”) but that may simply be a convention, with little evidence to support it.
I never said that 10 minutes was a cut-off point where extra frequency is meaningless. What I said was that 5-minute bus routes was a waste of money. You gain much more with two 10-minute routes than one 5-minute route.
You give Berlin, Paris, Toronto as examples of succesfull 5-minute routes, but all of them are metro routes. These don’t have the bunching issue that buses have.
Having a 5-minute network rather than a 6-minute network only for the promoting is just rediculous. 5-minute frequency costs 20% more than 6-minute frequency. I much prefer these money to be used for something that adds real value compared to something that just “sounds better”.
“You gain much more with two 10-minute routes than one 5-minute route.”
That’s ridiculous. Imagine a corridor, with a bus running every ten minutes. Now imagine another corridor, one block over, with a bus running every ten minutes. There is no way that is better than just picking a corridor and running the bus every five minutes. Yes, riders can save a tiny amount of walking (roughly one minute) — but they lose frequency.
As far as bunching goes, trains bunch too (in New York, Boston, San Fransisco). Bus bunching can be minimized with off-board payment (which San Fransisco has) and relatively short routes (ditto).
“Have one route that make all stops and another route that only makes major stops.”
My understanding is that the combination of a limited-stop express and a “local” is mostly a North American thing. It probably grew out of the poor stop spacing. The stops were too close together, but rather than do the tough work of spreading them out, they ran a second route on top, with fewer stops. This only makes sense if:
1) A significant number of riders are traveling between destinations on the limited stop express.
2) The main bus is very frequent.
If a bus is running every three minutes it is prone to bus bunching. This is what I mean by “very frequent”. Agencies only run buses that frequently when they are dealing with crowding. In that regard, this configuration actually saves the agency money. By running a few express buses, they don’t have to run as many “locals”.
But this type of demand is rare. It tends to happen only during rush hour. It is hard to see it making sense in the middle of the day, even though it is common in the United States. But just like with health care, the fact that it is more common here is not a good sign. It is pretty easy to see how it can fail to provide much value. Part of the problem is that you have two routes on the same street, that can’t be synchronized. The buses have identical headways, but at a particular bus stop, they may arrive at the same time. Then you have the fact that the express may not come close to where you are going.
For example, imagine this scenario: You show up at the bus stop, only to find that your bus left 2 minutes ago. The express will be here in 6 minutes, but that would require an additional 5 minute walk. You save a couple minutes on the bus, but you might as well wait 8 minutes for the other bus. Thus the express doesn’t get you anything. You would be better off with another regular bus equally spaced with the first one.
The basic concept you are describing is correct. Along a very busy corridor, it doesn’t make sense to run buses very frequently. But the definition of “very frequently” is much closer to 3 minutes than 10. In the case of this restructure, the “hybrid” proposal has the 5 running every six minutes, while the “five minute” proposal has it running every five. In both cases, having an express overlay running every ten minutes is quite sensible.
You misunderstood my post. 10 minutes is not “very frequently”, 5 minutes is.
The limited-stop express concept is probably over-used around North America. It doesn’t make sense where the time saved by skipping stops is so small that most riders pick the bus that comes first.
In your described scenario, you think that it would be better with another regular bus equally spaced with the first one. The problem is that, due to bunching, they probably won’t be equally spaced. Instead of one bus every 5 minutes you get two buses every 10 minutes.
It’s not a good idea to have a 10-minute express overlaying a 5-minute local. Like I said before, it’s very important that any routes that parallels each other have equal frequency. Otherwise, most passengers will choose the most frequent route and the less frequent route will run with empty vehicles. So in this case, it would be better with 7-minute frequency on both routes.
Transit planners tend to underestimate this phenomenon, so when they see that the frequent route have have more passengers per vehicle, they move vehicles from the less frequent route to the more frequent route, which creates a spiral until the less frequent route is removed completely. This is one of the origins behind the myth that “fewer frequent routes is always better than less frequent routes”.
We can quibble as to what constitutes “very frequent”, or the point at which adding frequency doesn’t improve service. For sake of argument, let’s say 5. Fine (it actually makes the math easy).
Now consider this scenario: A corridor has buses running every ten minutes. The city gets a grant, which it can use to improve frequency. They have two choices:
1) Run the buses every five minutes along that corridor.
2) Run a limited stop express every ten minutes along that corridor.
In the first scenario, there is less waiting. In the second scenario, many riders still have to wait ten minutes for a bus. It is only those who can use the express take advantage of it. The walking distance exceeds the time saved. But here is the kicker: You haven’t improved frequency. You can’t possibly synchronize the two routes, as the second bus is much faster than the other (otherwise there is no point). Thus for many of the stops, the buses arrive simultaneously. By your own writing, people prefer frequency over speed, which means this would be less popular than the first choice. There is clearly a trade-off, but with a normal corridor (lots of popular stops along the way) the first choice is best.
(See previous comment regarding bus bunching. )
Now assume that they implemented that first system. The buses are running every five minutes. Only now, the city has grown, and the buses are very crowded. So crowded, in fact, that they are going to run them more often, just to deal with the crowding. What then?
1) Run another 6 buses an hour. This means that instead of the buses running every 5 minutes, the buses run every 3 minutes, 20 seconds. The user experience isn’t better, but you have less crowding.
2) Run a limited stop express, every ten minutes. You reduce crowding, and give some riders a faster trip.
Clearly things have reversed, and the second proposal is better. Riders gain nothing by running the same old buses more often (it is like running bigger buses). But riders (some riders, anyway) gain by having an express.
It is only when you have reached the point where the rider experience doesn’t improve that an express overlay makes sense. Again, we can quibble as to where that occurs, but I don’t think anyone believe that running the bus every five minutes isn’t a huge improvement over running them every ten.
OK, that isn’t quite true. There are cases where the main corridor is full of big extremes in the bus stops (some have hundreds of riders a day, others have single digits). The low performing stops are often clustered together (a low performing section). In that case the express overlay is essentially the main bus route (the one that will generate the most riders) while the “local” is essentially a coverage route.
First, having lived in several situations with limited stop buses overlaying all stop locals, I heartily support them. The worst options is the rapid with no local which in an increasingly senior population is rider hostile.
Those are technical issues. The real challenge is reliability. In the half century living in the SF Bay Area, Muni has NEVER lived up to the announced schedules. Whether the oft reported three drivers having coffee at an outer terminal and then running a couple minutes apart after a gap of 15 or more, or simply Muni’s decades old deliberate understaffing. Whether directly funded by the city/county board of supervisors, or the current mixof parking ticket revenue, etc, the pre-covid missed runs ## are shameful. How many other systems have a premium pay clause in the driver contract based on “no leader”?
I think either the “frequent” or the “hybrid” as described on Muni’s web page would be good transit; I just don’t believe they can or will actually put 5 minute service on the street with any consistency. I have spent too many chilly evenings awaiting a bus which claimed 20 minute headways but was missing two drivers.
I would further note that very little is said in the currently announced plans about cleaning up the streetcar mess. With all of the bells and whistles computer assisted signalling in the Market Street Tunnel, Muni is un able to match the headways achievements of CTA in the State Street Subway in the 1950s using “obsolete” (but actually functional) relay based wayside block signals.
The net result is that (again, pre-covid) outbound rush hour service has usually been a disaster. One reason is the drat if there are any issues inbound on the surface, casrs arrive “out of slot”. Because drivers pick work assignments tied to a specific route, 3 Ls showing up to then go out remain Ls. Meanwhile the platforms become crowded with riders waiting for other routes. The obvious solution IMHO, is dynamic reassignment to preserve service on each route. This will require negotiation with the union, but it would greatly reduce rider frustration and speed actual throughput.
Plans voiced some months back proposed forcing transfers on several routes at the West Portal end of the tunnel–inconveniencing riders especially in bad weather or if (no surprise) the connecting train is delayed.
FWIW, before the pandemic the 5 (really 6) streetcar lins carried 1/5 +/- of the daily ridership. It should also be noted that both the N Judah and the T Third Street face so much automobile sabotage that each was given an express bus rush hour overlay.
You could reduce the amount of bus bunching by adopting traffic signal transit priority and by expanding bus stop zones so that two buses can load and unload simultaneously. Is Muni proposing to do that here? I didn’t see it anywhere on their website.
They promise TSP every so often. Supposedly it was installed on on the Embarcadero for the routes to Caltrain and 3rd st; but AFAIK, it was never turned on. Once again, the issue w/Muni is credibility.
As someone who lives in San Francisco, in my opinion frequency really is not as important as a transit service than being on time. WAIT TIME I believe is the most important. MUNI is notorious unreliable in wait times. Frequency is cities and not suburbs is a way to reduce wait time as much as overcrowding, but may or may not be very effective (buses bunching up), especially in San Francisco. Tough to solve in major cities, but if the goal is to reduce driving this needs to be solved. Of course overcrowding, good for transit budget, but huge turn off to riders, especially with pandemic. But crisis brings opportunity, and this moment is very critical for future of transit, so thanks Jarrett for helping make transit in San Francisco better.