Oakland deserves better

I’ve been living in SF for several years now. Muni and BART get me everywhere I need to go, and it’s been reliable enough that I don’t even have to consider other modes of transportation. No matter where I go or what time or day of the week it is, I’ve got a good idea of several ways to get there, and I’m generally not worried about making connections or waiting for a long time (except, perhaps, between 1 and 5 AM). This provides me with a certain peace of mind and feeling of freedom—the city is fully available to me at the low price of $94 per month, and I never feel stranded anywhere.

I suppose this has spoiled me to an extent. I’ve been spending a good deal of time in the East Bay recently, and have been taking AC Transit around. Most of the time, it works well for me, but my use case is probably very different from most actual residents of Oakland. I usually take AC Transit buses to get from BART stations to some location and back. BART stations are usually served by many bus lines, and the sheer number of options can paper over the problems with frequency and ease of making transfers. On the occasions I have taken AC Transit buses between areas of Oakland not served by BART, I’ve had a vastly different experience.

What makes this all the more frustrating is that Oakland is a city with a lot of great neighborhoods and interesting places to go. It deserves a bus network that lives up to this.

Oakland can support high all-purpose ridership

One of the main determinants of the quality of the bus network is density. When you have more people in an area, you can have more riders. These areas are also more likely to have commercial districts or other amenities that attract people from outside of the neighborhood. Oakland nominally has a density of 7,514 people per square mile, but this is too reductive. Here’s a map of population density by census block group from the 2016 American Community Survey:


The reason Oakland has such low density when taken as a whole is that it includes a lot of more or less suburban space. The hills drag down this aggregate population density quite a bit. On the other hand, East Oakland has many areas with over 20,000 people per square mile, which puts it on par with lots of neighborhoods in SF which have much better transit. There’s even quite a bit of space in the dense area taken up by freeways and interchanges. If the I-980 spur is taken down, Downtown Oakland will likely densify as well.

If we look at Oakland without all the hilly bits and industrial area around the port, it has around 15,128 people per square mile.

Over 300,000 people in about 20 square miles.

What’s surprising about this is that even though East Oakland contributes a lot to this density figure, the people who live there don’t seem to take a lot of buses, and rely heavily on cars. This should be our first sign that something is up and AC Transit isn’t providing a great service.

This slideshow requires JavaScript.

In most of the census blocks in East Oakland, people driving alone make up the majority of commuters, where bus riders end up in the single digits. Neighborhoods like Temescal have a much lower share of drivers, owing to some combination of better bus service and easy access to BART. Residents of downtown almost certainly take BART to work if they commute into SF, or walk if they stay in Oakland. BART is an important part of the transportation situation in downtown and northern Oakland, but the odd alignment and station spacing in East Oakland wastes a lot of its catchment area there. This means good bus service is all the more important unless BART is planning to fill in the gaps. (No serious proposal for East Oakland infill stations is in the works at the moment.)

So what’s the problem with Oakland bus service, then? There are plenty of high-ridership, frequent lines in the system. AC transit trunk lines can run at pretty high frequency. Line 1 runs at 10 minute frequency or better for most of the day. The 72, 72M and 72R put together achieve headways as low as 6-7 minutes if you’re at a stop served all three. But high frequencies among commute-focused routes during commute hours don’t make a good bus network by themselves. We have to consider lots of other things if we want people to use the bus network to get around in their daily lives—ability to make transfers, ease of understanding the network, frequency at non-peak hours and on weekends, and so on.

Understanding AC Transit’s decision framework

A lot of transit agencies divide their customers into two camps, the “choice” and “captive” groups of riders. Agencies that use this distinction are unlikely to improve service in areas where car ownership is comparatively low, because those areas are said to have a higher number of “captive” riders which will take transit no matter how poor the service is. “Choice” riders, by contrast, have other means of getting around and will only take transit when it’s more convenient or better for them. Plenty of articles and reports have been written about why this split is wrong-headed, such as this one. The short version is that this mode of thinking completely denies that incremental improvements in speed or frequency can have significant effects on ridership, and this leads agencies to send their money towards other “features” such as Wi-Fi which few riders actually care about.

I don’t think this is the framework AC Transit is operating under, because they tend to serve areas with lower car ownership better. Despite being denser and having more riders that might be considered “choice”, East Oakland does not see the same level of service as downtown and the north. San Pablo, Telegraph, Broadway, and College are all served well by AC Transit. Not all of this can be attributed to pressure from the university, either, though it surely supplies many regular riders.


I think the issue is more that AC Transit sees itself as a provider of transportation for commuting, and its responsibility largely ends there. Because the dense area of Oakland is fairly long and skinny, this means that a commute network can still provide service to a lot of the city. At the same time, it can’t serve most people for all-purpose trips which may not end downtown. This is apparent when we see the lack of frequent crosstown lines.

There are essentially two buses that function as frequent crosstown lines: the 57, which runs from 5am to midnight every day at 15 minute frequency, and the 80/81, which combined give a reasonably frequent crosstown line on Ashby, as long as it’s not a weekend or nighttime. The next major crosstown line is the 51B on University Ave in Berkeley. It’s over a mile from downtown to 40th St and around a mile and a half to Ashby from there, so a lot of people in between end up underserved. Frequent crosstown lines in East Oakland are basically nonexistent. The 20/21 combine to provide decent frequency on Fruitvale Ave until the 21 stops running at 10pm, and that’s about all there is. This poses a problem because BART stays close to the water, as does the frequent 1 for a good portion of its route. A lot of people between the water and MacArthur Blvd are left with few options outside of the weird buses that have complicated routes, mostly exist to take you to BART, and run at low frequencies outside of weekday commute hours.

This means that the bus network doesn’t function that well as a whole—as a way of getting people between arbitrary places. If you’re at San Pablo and Stanford, getting downtown is pretty easy. If instead you’re trying to get to 51st and Telegraph, your journey will probably take longer than you would expect. This ability to move well in one dimension but not in the other is often a problem even in good transit systems; my friends in Manhattan often tell me that they’d rather travel dozens of streets than, say, four avenues. In San Francisco, it is still generally easiest to get downtown from somewhere, but there are many more frequent crosstown lines, which makes this less of an issue.

If Oakland had frequent crosstown lines at a reasonable spacing, you could take advantage of the power of fast transfers and the grid to get anywhere quickly and reliably.

Complicated lines, weird service hours, downright bad maps

Instead, we get bus routes like the 14.

This is a bus which cuts right through some of Oakland’s densest areas near Lake Merritt. Car ownership rates in these neighborhoods are comparatively low. The bus has one terminus downtown and a BART station at the other, so it’s well-anchored. But it meanders around in between so the actual route is difficult to remember. If you don’t live or work in its immediate service area, I wouldn’t expect you to know where it goes. The line makes a U shape and the “middle” is anchored by the Laurel commercial district on MacArthur, but it doesn’t spend enough time on MacArthur to really serve that corridor, or any other. At best, this is emblematic of the direct service example from the Human Transit post about transfers I linked above. This bus also runs at 30 minute frequencies after 8pm on weekdays and at all times on weekends, and there’s no service after 10pm.

The 14 has about 3,000 weekday boardings. This is lower than the number of boardings on Muni’s 18-46th Avenue, despite the latter being one of its system’s most far-flung crosstown routes. It primarily goes through areas of lower density and high car ownership, the overwhelming majority of which is purely residential area. A good deal of its potential catchment area eaten by the Pacific Ocean and Golden Gate Park. It has worse weekday midday frequency than AC Transit’s 14. Why is this?

The 18 is a well-designed route. It is simple and most of it occurs on or close to the street for which it is named, so it’s easy to find and figure out where you can go with it. It runs on the same hours as just about every other bus in the Muni system (5am to midnight, every day) despite having relatively low ridership, serving a lower-density area, and having its ridership mostly concentrated between the AM and PM peak hours[1]. It does not closely duplicate service found on more frequent lines. Most importantly, it connects riders to several very frequent east-west or downtown-oriented routes. This means it is a critical component of the entire network and benefits from good network design as well as good line design. So, the 18 is simple, reliable, and useful.

Frequent transfers available from the 18. All of these are listed on the Muni system map as 10 minute frequency or better.

As we mentioned before, the 14 is complicated. It can’t be easily boiled down to one major street. Its service hours are relatively narrow and it has poor frequency at times when people may still even be using it to commute. In the densest areas of the its route, it is 2 blocks from the 40 and 4 blocks from the 1, both of which are among the most frequent lines in the entire system, and which I would be more likely to take once I knew they had better service.

It’s harder to get information about AC Transit routes compared to Muni routes because of the way the maps and schedules are made. For example, the route map for the 14 can be found here. This map is not very helpful if you are trying to figure out how it can get you around Oakland as a whole, because the map displays very little context outside of the streets themselves. It shows connecting routes, but does not show where they go and instead simply identifies them by number. It also does not distinguish which of the connections are frequent and important routes. Some of the connections listed will never happen in practice, such as with the 800 (owl) routes, which run at disjoint hours with the 14. (Depending on where you are going, such a transfer could happen at around 7am on a Saturday or Sunday morning, but it is impossible at all other times.) I would also argue that it’s easier to remember when Muni routes are around as they have a short name which describes where it goes. AC Transit buses are often described by the route number and every street they use; on the signage they often just display the terminal. I find myself having to check the maps anyway to make sure they go where I’m trying to go.

Some maps will just make your head spin. Try making sense of this route map, which lays out the 40 (one of the most important lines in the system) in an incredibly confusing way.

The system map has similar problems. For Muni, the system map uses progressively darker and bolder lines for more frequent routes, so you can tell how much service an area receives just by looking at the map and don’t have to inspect in too much detail. Express services and other routes that are not usable all day, while few, are marked to appear less important. AC Transit’s system map uses all manner of colors and symbols to draw routes. It does a poor job of teaching general characteristics of the network at a glance. One symbol is used for “all-day service,” but this does not describe a consistent set of service hours, and the map legend tells you to look at the timetables anyway. In addition, Muni lists frequencies for each route on the maps at bus stops. AC Transit prefers to lay things out in terms of schedules. Certainly, at least some of their routes are frequent enough that they could be listed in terms of frequencies as well, but the choice of using timetables makes it difficult for people to decide whether or not using or connecting to a particular route will result in them waiting a while. Transbay buses other than the F, NL, and O only run during commute hours in the commute direction (assuming your job is in SF), so they are not useful for all-purpose transit and so I think they should be kept to a different map to avoid clutter.

All-purpose trips require uses at all hours

Suppose we could wave a wand and fix the network design, such that we have high-frequency crosstown routes at regular intervals. The network as a whole might have higher ridership than it does now, but maybe not by much. The new routes mainly see a lot of riders at commute hours and are underused at other times of the day. AC Transit might consider this experiment to be a loser as it spends a lot of money on what it considers to be overservice, and it would be hard to blame them for that. What’s the whole point of doing this, then?

Well-designed crosstown routes may still mostly be used to get to other bus lines, which will actually take riders somewhere interesting. This will provide them some level of service even after commute hours, and does make the network more useful for everyone. But they are most effective when they are useful in and of themselves—that is, when they are used for other purposes than just getting people to and from their houses. This is part of the reason why a crosstown route like the 24-Divisadero is able to support such high ridership throughout the day. Other than being useful for getting around, it goes directly through mixed use areas of Divisadero and Cortland.

This means some of our network design should be supplemented by zoning changes. We shouldn’t take this too lightly, though; zoning changes without strong renter protections could lead to displacement of the very communities we are trying to serve by improving the transit network.

Oakland has a general plan which guides its decisions around zoning and transportation, among other things. The last general plan was adopted in 1998, and its high-level strategy separated neighborhoods or corridors into two categories: “maintain and enhance” (meaning little change in style or intensity of land use) and “grow and change” (meaning upzoning, mixed-use development, or the like). A map can be found early in this section of the plan. Some major crosstown corridors did fall into the latter category but it’s not clear how much they have affected zoning decisions since. For example, MacArthur Blvd, West Grand, and 14th Ave are all labeled as “grow and change” areas. These are prime corridors for major crosstown bus routes and AC Transit has, to some extent, paid attention: MacArthur is close to the 57, and Grand is served by the NL.

Obviously, high density around the corridor can be a key component of high ridership, but more important is having a mixture of uses along the corridor itself. This opens up the potential base of users for the route from just those who live near it to everyone who’s served by the network, and if that allows us to turn the frequency up on those routes, then the people who live there end up with better service as well. So if MacArthur is the corridor that the city of Oakland wants to “grow and change” in the direction of more mixed use, then one appropriate change might move the 57 to MacArthur from 40th St. Another option could be to zone 40th St for mixed use instead. High frequencies will make sense on these corridors even without big changes to residential density; SF achieves some of its highest ridership numbers on routes which are surrounded by 3-story buildings.

Transfers and fare structure

A few years ago, AC Transit made some significant changes to their fare structure. They creating a day pass, but removed discounted transfers. This was done in the name of speed and preventing “transfer fraud.” While the day pass automatically activates if you board 3 buses in a day and are using a Clipper card, the people who are most likely to be affected by this change are also the least likely to have Clipper cards, so they might be paying double the normal fare if they are using cash. Even with the day pass, there ends up being a slight surcharge for transfers if you take a round trip. This is a problem—if we are trying to reconfigure the network so it uses transfers to make it easier to get around, people should not have to think about taking a transfer. Transfers should always be free.

AC Transit attempted to enforce a number of rules around transfers, namely that they could only be used for a certain amount of time (which is reasonable) and only in “one direction” (which is not reasonable, let alone clear). Fare evasion also captures an outsize amount of attention from transit agencies. If we accept that there will be some level of fare evasion and we actually want to speed up buses, there are better ways to do it. Muni has implemented some of these, such as all-door boarding on all buses at all stops. If Clipper users are allowed to board through the rear doors, then the bus does not have to wait for everyone to file in through the front, which significantly reduces dwell times. AC Transit’s fare structure makes this somewhat more difficult, since it also has a concept of local vs. Transbay fares on Transbay buses. I suppose you could board through the rear if you are paying the full Transbay fare, but if you are taking the bus locally, you could still board in the front to tell the driver. You would also have to pay 25 cents when using a (local bus to local bus) transfer, forcing you to go in the front and stop while you put change into the farebox. If transfers were free and used a proof-of-payment system, then people with paper transfers could still board in the rear. In some sense, it seems like their speed problem was self-inflicted.

Fare evasion is definitely not a big enough deal to justify punishing people who need to transfer, or who don’t have Clipper cards. This fare structure may also inform some of their network design decisions, such as having duplicated and branching service (like the 20/21, 80/81 or 72/72M) to avoid transfers, or why so many lines go well out of their way to terminate at BART or downtown. If AC Transit adopted policies similar to Muni, they would be able to fix several of these problems at once.

Improving the grid

One of the first areas I’d like to explore is how to take the area in inner East Oakland and give it better connections to the rest of the network. We’ve already gone over the 14 and why it’s not a very useful bus. The 62 has a similarly odd route and falls off in frequency after commute hours. The 96 acts sort of like a true grid line, but is never frequent enough to be useful at 30 minute midday headways during the week.

First, I’d like to separate the 40 and the 1 a bit. Their catchment areas overlap almost completely in this area, and since they are both trunk lines this seems to be a wasteful decision. (If we do need this extra capacity in the area, we should run more 1 buses, perhaps with a short-turn during peak hours.) I chose to stick to E 21st St because that seems to give a good middle point between the 1 on International and the 57 up on MacArthur. Also, most buses in this area go around the south side of Lake Merritt; I chose to route around the north side, and use 20th St (for the transit center/BART connection) and Grand downtown.

The 40-E 21st/Foothill.

We can also think about some crosstown lines. First, we should identify important crosstown corridors. In East Oakland, this is probably Lakeshore or 5th Ave, then 14th Ave, 23rd Ave, Fruitvale, 35th, and High St, to start out. North of downtown, we could have 27th, MacArthur/40th, something in the 51st-55th area, perhaps Alcatraz, and then Ashby Ave. To the west, we have Market/Sacramento, then Adeline, then Peralta St or Mandela Pkwy. We might be able to stitch some of these together into a nicer layout.

Generally, I’d like to do the following:

  1. Make a frequent grid, with the spacing between adjacent in the neighborhood of half a mile, or maybe closer together in denser areas.
  2. Avoid moving existing high-ridership corridors, especially ones that get lots of people downtown
  3. Lay out gridlines on mixed-use streets where possible
  4. Keep lines simple; that is, avoid unnecessary detours and try to make lines straight and stick to as few streets as possible
  5. Avoid streets that are too narrow or otherwise adverse for buses (i.e. lots of speed bumps)
  6. Simplify duplicated service, especially if it’s duplicated across more than 2 buses
  7. Simplify branching service (which usually dovetails with #6)

My hope is that this will produce a grid where people can travel all over Oakland with at most one transfer. For this to work really well, we need high frequencies on all of the grid lines. I think that we can get some extra frequency without making huge capital and operational investments through line simplification—buses that don’t have to travel out of their way will have shorter routes that the same number of buses can cover more frequently.

This map is taking some time to put together, as it involves shuffling around a lot of currently existing lines. Changes made in one place often require cascading changes in other places. While working on it, I also realized that I would like to be able to explain some of my decisions, rather than just finish up this post with a map with a lot of lines and no justification for any specific part of it.

I’ll finish this up with maps and rationale in a follow-up post.

[1]Old data on ridership broken down by time of day was gathered as part of the Transit Effectiveness Project, the precursor to Muni Forward. This data can be found here. As the fundamental situation of the 18’s service area has not changed much in the past decade I figured that these data were still largely relevant, but I would urge caution when using TEP data to talk about other routes.

How bad is the last mile problem in San Francisco?

The “last mile problem” is often invoked by Uber and Lyft to justify their attempts to cannibalize public transit ridership. From this article in the Examiner:

Both Lyft and Uber offer bus-like services for their vehicles, Lyft Shuttle and Uber Express POOL. The Uber Express POOL service is set to launch in six more cities nationwide Wednesday, though it has operated in San Francisco since November 2017.

Kate Toran, head of taxi services at SFMTA, who also helped craft the rules around private mass transit, said Lyft Shuttle is regulated by the California Public Utilities Commission, not the SFMTA, because it seats so few passengers in each vehicle.

While SFMTA officials have expressed concern about Chariot potentially competing with Muni, Uber staff argued Tuesday that the company’s new carpool service also may address routes public transit may not serve well.

In a Tuesday briefing with reporters, Ethan Stock, head of Product at Uber, said he viewed Uber’s Express POOL service as a solution to the “last mile problem,” where public transit doesn’t bring people to their destination doorstep.

“One of my favorite examples is my own commute,” he said. “I would love to take Caltrain up the Peninsula and have a seamless Caltrain ride to the Uber office (in San Francisco).”

Yet, he said, “the amount of time and frustration” he has trying to get from Caltrain to his offices at Uber on Market Street is a barrier to using public transit.

“I think this is an example of a gap in the public transit system I think Uber is in a position to fill,” he said. “Public transit works really well on core, high volume routes, but in the diversity of moving all over the city, it doesn’t.”

I’d like to answer both the specific and abstract arguments made by Mr. Stock. I can’t tell from his statements alone whether or not he takes Uber to his office from his house or from the Caltrain station at 4th & King. In the former case, it’s not feasible for everyone to have a one-seat ride from far away to their destination. If you want to know what happens when we try to give everyone a one-seat ride, you need look no further than awful 101 traffic. Either way, since he brings up the “last mile problem” I assume he’s proposing Uber as a solution for the latter.

The Uber office is at 11th & Market, so getting there from the Caltrain station is a direct ride on the 47. This is an incredibly frequent bus—one of the most frequent in the entire system. Ridership on this line is over 10,000 per weekday, and it shares a significant portion of its length with the 49, which has much higher ridership. By all means, this is the kind of “core, high volume” route that Mr. Stock thinks transit works well on. It picks him up where he gets off the train and delivers him to a point about a minute’s walk from his destination. More generally and as I talked about in my first post, the frequent grid covers San Francisco with these high-ridership routes, and this is especially true downtown. So why does Mr. Stock consider this frustrating and difficult, and what part of the last mile problem really remains unsolved?

Mr. Stock does not consider taking the bus, and the last mile problem is only bad in San Francisco if you do not take the bus. SFMTA has designed the bus network to have nice grid properties and help people within the city commute downtown, but it has also created a system with high connectivity at all of the rapid transit stops. Because rapid transit stops are so busy, it makes sense to serve them heavily with buses, so people can make many different connections more easily. Accounting for this when designing your network makes it useful for as many people as possible, even when those people are coming from far away. From Caltrain alone, we can connect directly to almost every downtown subway stop. Embarcadero is accessible by the N or T. Montgomery is along the 10 and 12. Powell can be reached by the 30 or 45. Van Ness is on the 47. Civic Center is a bit harder as the 19 is a few blocks away, but Muni runs the 83X during peak hours to bridge that gap. The 81X and 82X also provide additional commute service from Caltrain to other destinations like the Transbay Terminal. The Central Subway will make moving people to and from the Caltrain station even easier when it is completed next year.

This is still true, if not to the same absurd level, with just about every other rapid transit stop in the city. I talked with a friend who works out in the Bayview and he revealed that his company is part of the Lyft corporate program, which helps people take Lyft from nearby BART stations to the office. His office is also accessible from Glen Park BART via the 23 and 44 buses. The 44 is especially frequent during commute hours.

Further discussion centered on the fact that buses are slow. I won’t argue with this, but we can fix it, and getting people into cars is not the right way. One part of the solution could be expanding the rapid network, as I wrote in this post. (If the 44R proposed in that post existed, his office would be a 4-stop ride and short walk from Glen Park.) Another could be installation of more red lanes, which would speed up the buses we already have. For some people, taking their bikes along or using bike-sharing services might provide a reasonable solution. But we can’t ferry everyone to their jobs from busy BART or Caltrain stations via cars, even with carpooling. The more efficient your carpooling is, the slower your ride gets, and the more it approximates standard bus service. As we take more cars, everyone’s ride gets slower. And this speaks to Mr. Stock’s final assertion—it is not mass transit that performs badly when people in a city are trying to go from anywhere to everywhere, but car-based transportation such as that provided by Uber.

In places which have less useful local bus networks, the last mile problem is worse, and car service might be the best option for people once they arrive at a nearby mass transit stop. It certainly is a more sustainable practice than driving the whole way yourself. But in San Francisco, we can do better, and tens of thousands of people already use buses to get around after taking BART or Caltrain in. It might take some investment to make them more useful, but in the end everyone will have better and faster service. Uber and Lyft have again tried to position themselves as a supplement to public transit, but it looks like they are continuing to interpret “public transit” solely as rail and if they supplant bus service, we will all lose out.

Making connections with station placement

I recently read this article about some potential development near downtown San Jose, and this along with the news of development happening around Diridon Station reminded me of the BART extension that’s supposed to come to San Jose in the mid-2020s. The San Jose subway has already been watered down from the initial proposal, in that there were two stations in the downtown area that have been consolidated into one to reduce the cost. However, the location of this single downtown station has not yet been finalized.

VTA presents two options for this station, both of which can be found here. The station will sit underneath E Santa Clara, which makes sense as that’s the main thoroughfare. The west option has the station spanning Market through 3rd, and the east option 2nd through 6th. The east option seems clearly worse: the west option is much closer to the center of downtown, and will also provide a more direct connection to the northbound buses and light rail that run on 1st and 2nd Streets. The west option would bring much higher benefit to the whole network.

Map excerpt from here, with the west option in purple and the east option in yellow.

The west option also leaves more room for an infill station between the Downtown and Alum Rock stations, which we might want to reconsider if the areas east of downtown densify significantly. An infrastructure project which has so much potential impact on San Jose should leave future options open as much as it can. Why would VTA even consider the east option, if it doesn’t fit as well with the current network and land use?

It’s likely that they would make concessions like this to avoid political battles. The west option—being closer to the middle of downtown—might engender more short-term resistance from merchants, since construction can lose them foot traffic and customers. Though this is clearly worth it for them in the long-term, merchants may not believe they will survive to get the benefit. Remember that Winter Walk SF, which shows up every December on Stockton St, is a compromise made with Union Square businesses that were afraid of losing customers. Central Subway construction in the area is suspended while this is going on, which makes it take longer and cost more.

It would also sit directly underneath the light rail tracks on 1st and 2nd Streets, which means construction on the station will probably disrupt light rail service. The city can work around this by stopping trains at Convention Center Station if they’re coming from the south and Japantown Station from the north, and running a frequent bus bridge between the two. Customers won’t be happy, though! By contrast, the east option would not require digging up 1st or 2nd Streets and light rail service could continue during the construction period.

This issue is based on the current locations of switches, and with some extra track along E St John and E San Fernando they could run trains to St James and San Antonio Stations. This would mean only Santa Clara Station would go out of service, and might provide some extra flexibility for moving trains around in the future. Additionally, these stations are less than half a mile apart and both are a short walk from downtown, so it might require less capacity on the bus bridge. I’m not sure this is enough upside for VTA to invest in something that involved, but it’s worth considering.

Unfortunately, this means the easy option is the one that does not serve the needs of riders quite as well. This shouldn’t be surprising—it’s the reason we see a lot of BART in freeway medians or why rail systems tend to expand interminably into the suburbs when they would benefit more from core capacity improvements. It should serve as a reminder that if we want a transit system that works for everyone, we’re going to need to make some noise. The VTA website lists some upcoming meetings here, so if you live in the South Bay and are interested in this, you might want to check them out.

The sparseness of the rapid network

San Francisco has complicated geography. While the hills are nice to look at and offer some great views and parks, they’ve also forced Muni to do some suboptimal network design. Hills will break up the street grid and be very hard to pass over. We can see this quite clearly with Mt. Sutro and the Twin Peaks, which cut off the Castro/Noe Valley street grid from the Sunset one. This makes it difficult to design an east-west line that fills in the gap between the N and the L. The 7 tries to avoid this by coming in from Lincoln Way, and the 48 up from West Portal (though only during peak hours), but large areas are still left uncovered and the innermost bits of the Sunset aren’t well served by this sort of design. Hills also break up Noe Valley and Glen Park, making continuous north-south lines difficult.

In addition, I think that the hills and narrower streets that some of the highest-ridership lines traverse have prevented them from getting rapid buses added to them. I would expect the criteria for adding a rapid bus to be something like the following:

  • High ridership
  • Lots of transfers to other high-ridership lines
  • Long route
  • Dense stop spacing on the local

If we look at just the rapid network, it seems to be very commute-focused, with few crosscutting lines except for the 28-19th Avenue. Entire areas of the city also seem to be lacking in rapid service.

I excluded the 7R-Haight/Noriega Rapid here because it skips so few stops that it might as well be a local.

Just about every line here serves to get people downtown! By contrast, the area just west of downtown LA seems to have no problem implementing a rapid grid in addition to its locals.


It’s missing a few pieces—perhaps rapid lines on Beverly and La Brea—but the grid structure is clear here. Can we look at something similar for SF? First, we need to decide which lines are due for rapid upgrades. The Van Ness section of the 47 and 49 will essentially be rapid once Van Ness BRT is complete, so that’s a start. Having a 49R to complete the rest of that journey would be nice as well, and Muni does seem to be planning something along these lines. I have some ideas for other upgrades, ranked roughly from what I think are the best options to the worst:

  • 29-Sunset (19k riders/weekday, crosstown line, connections to most rapids and Balboa Park Station)
  • 44-O’Shaughnessy (17k riders/weekday, crosstown line, connections to most rapids and Glen Park Station)
  • 1-California (26k riders/weekday, 29k with express routes, adds service north of Geary)
  • Either the 22-Fillmore (16k riders/weekday) or 24-Divisadero (10k riders/weekday, but longer route and denser spacing)
  • 45-Union/Stockton (10k riders/weekday, goes through an area with little rapid coverage, though the route is shorter)


So what’s stopping us from upgrading these lines? One of the interesting things about rapid buses is that if they are deployed along frequent local bus lines, they won’t really save much time or go much faster if they can’t pass the local buses. This means that trolleybus lines that are potentially too steep for a full motorcoach are immediately off the table! The 1, 22, 24, and 45 won’t be able to have rapid service for this reason. This is a real bummer, since the 22 or 24 would have filled a nice gap in the middle and added a crosstown rapid line. Additionally, the stop spacing on the 24 is so dense in some areas that it stops are not even 600 feet apart—sometimes even under 300 feet! It would have been nice to have a fast ride at times other than the middle of the night, but it’s not looking too good right now.

Can we at least add rapid service to the 29 and 44? Passing on normal two-lane streets is harder, but still possible. From the Presidio terminus to about Holloway & 19th Ave, the 29 is mostly on large streets where passing is easy. The same is true while it’s on Ocean. Otherwise, the streets are narrow and might not allow passing, and passing on Mansell would probably be dangerous for bikers. The 44 spends a lot of time on large roads as well, and Silver is a better street to pass on than most with only one lane in each direction.

What does the rapid network look like with these additions?



Still not perfect, but it’s getting there! We have some semblance of a grid in the southern and western areas of the city now. Twin Peaks has a faster way to go crosstown. The northern areas are still underserved, but I’m not sure there’s a good solution to that short of a Union St subway. Perhaps something could be done using Broadway since it is flat, wide, and the area seems like it could use the service, but it wouldn’t line up with any existing line and would probably be the result of some rejiggering of the lines that are already in the area.

A downtown subway grid

Recently, the SFCTA, SFMTA and various other transportation organizations released a website called Subway Vision. It aims to collect ideas from everybody about where to install new subways in SF. The Subway Master Plan is supposed to ensure that we are always building a subway, and we currently have the Central Subway under construction, along with planning for the M subway and construction hopefully in the not-too-distant future. Future expansion plans are also being considered right now, as the Rail Capacity Strategy report released back in February gives an outline of what Muni thinks its next steps will be.

People spend a lot of time talking about Geary whenever new subways are discussed. And that’s great! Having high-capacity transit on Geary is extremely important given that its bus lines ferry the same ridership as Caltrain. BRT will definitely help, and we’ll see how this develops in the future. However, I think the need for new subways downtown is often overshadowed by this discussion. Our current rail network has one subway that goes from downtown outwards, through which a large percentage of its 150,000 daily riders go. This is a huge bottleneck! Problems that occur in the subway here are liable to delay many people and cause lots of gaps and bunches. If there is one thing we can learn from New York here, it is that these problems are way more tolerable when there is some redundancy.

Having extra subways parallel to the Market Street Subway downtown is hardly a new idea. SPUR advanced the idea in this report and has implied it in various materials concerning a second Transbay Tube. By pulling together some ideas from these and the Rail Capacity Strategy, we can come up with a nice grid-like structure for subways in downtown SF.

First, if we look at the Rail Capacity Strategy, we find that Muni eventually wants to put LRT on Geary. LRT will improve capacity even more than BRT, but the most interesting part about the plan is the fact that it might be a hybrid subway/surface line. Muni gives cost estimates for all of the involved projects in the report, and the lowest estimate for Geary LRT appears to be the cost of installing surface rail along its entire length at a cost of about $1.4 billion. By using the estimated cost of installing other subways and doing some math, we can figure out that the $3 billion upper bound makes sense if Muni installs a mile of subway and 5 miles of surface rail. One mile of subway takes us from Market to about Van Ness along Geary, which is a reasonable option.

But why should we stop at Market? Merging into the Market Street Subway is probably not feasible, since it’s not possible to fit a flyover junction when there is already another subway below. Even if it were possible, it would probably be ill-advised, as we’d get one extra stop at the expense of subjecting the new subway to all of the same problems we get by combining the other lines. But there are other options—the report also has lots of figures for job and population density in various sections of the city, and we can use those to extend the route. 2nd St in particular has a very high density of both, and a subway in the Geary area could connect through Mongtomery Station and continue into SoMa along 2nd St. This line could be anchored at AT&T Park, which is surrounded by a commercial district, and which often relies on the easily-overwhelmed Muni Metro Extension to get people in and out on game days. 2nd St is also slated for an upgrade to LRT in the distant future, according to the Rail Capacity Strategy.

It might seem like overservice, and while the redundancy with the Central Subway is nice, that’s not really what we’re after. The grid really comes together when we install a subway along Folsom. Folsom is considered one of twelve high-priority corridors for expansion by Muni. By pulling a line out of the Market Street Subway between Church and Van Ness and instead sending it up Folsom, we solve a number of problems:

  • Fewer lines are merging at Van Ness, which means we can mitigate a bottleneck there
  • The already-existing flyover junction would make construction of the other one less disruptive, since we wouldn’t have to reconfigure the current subway
  • In the event of backups in the Market Street Subway, all KLM trains could reroute into the Folsom Street Subway instead, mitigating delays (J and N trains could stop at Church & Duboce for transfer) and still allowing decent access to Montgomery and Powell areas via the 2nd/Geary and Central Subways

In this proposal I choose the L train to be pulled out in standard service because we can’t pull out the J or N, as the new trains will leave the MSS before the J or N merge in; the M is going to improve capacity in the MSS after its subway expansion and use of longer trains, which is important; and the K can only run short trains at the time being, and may still be used to serve the Muni Metro Extension.

The end result is a grid-like structure centered on 2nd, 4th, Market and Folsom Sts, with additional transfer points along the Embarcadero and on King. Here’s what it looks like:

Above-ground stops are circled with gray, and Van Ness BRT is shown as the green dashed line.

There’s also room for further expansion: Leavenworth, Civic Center and 7th St stations all line up quite nicely, for instance (and would be easily connected by the bus I laid out in earlier articles). If the Caltrain railyard is rebuilt underground and the 280 spur is at least partially torn down, this could open up enough land that extending the Muni Metro Extension out to 7th St and building more rail along 7th St makes sense, as well. Some ideas for extending high-capacity transit along Van Ness send it down 11th St and Potrero Ave, in which case they would also fit nicely into this grid; extensions down Mission could also use the transfer to the new station at Duboce Ave. Sadly, I think the Chinatown T stop is too far north to have a really easy transfer, but a line down Sacramento St might be useful and connect Van Ness BRT with Embarcadero and Main St Stations. The T would be within reasonable walking distance of such a line.

This expansion provides a lot of extra options for people to move around on corridors that Muni considers important, and getting them into a shape where they work well together. Having redundant infrastructure makes the system more resilient and also makes it easier to justify arbitrary planned service changes, which people will need to deal with if we ever eventually want 24-hour rail service.

But don’t just take my word for what subways are worthwhile to build. With Subway Vision, Muni is giving all of us an opportunity to influence the evolution of the city. We should use this opportunity to tell them what we think!

Completing the grid: The Tenderloin gap, part I

If you look at a transit map of San Francisco, you might find some surprising holes in the coverage. Perhaps we can learn something about network planning by trying to fill them in.

One of the more confusing examples of this is the lack of north-south lines running through the Tenderloin, Nob Hill, and Russian Hill. The Tenderloin is one of the densest areas in the city and has a lot of below-market-rate housing, so adding service would increase mobility for a lot of people. Nob Hill and Russian Hill are also very dense and poorly connected by transit, which means we have a big opportunity here.

Fisherman’s Wharf employers have trouble filling positions because the area is out of the way, and having more service there could bring in more employees and tourists. Because it is such a big destination and it just north of Russian Hill, it can serve as an anchor for one end of the line. Another important point on this line would be a connection to Muni Metro and BART.

Lastly, if we want a frequent grid that functions well, we have to make sure the grid lines are not spaced too far apart, or our reliably-covered areas will have huge gaps.

Let’s look at the current map of north-south lines to see what we’re up against.


You might look at this and say, “Coverage seems fine!” That’s true, but for such an important area, there is startlingly little frequent transit, and the lines aren’t very simple. If we evaluate each of these lines, the picture comes out much bleaker.

First, let’s look at the 19-Polk. North of the Tenderloin, this bus is basically a less reliable Van Ness bus. It runs less frequently, there is no owl coverage, and any speed problems plaguing Van Ness buses will be put to rest by Van Ness BRT in a few years anyway. I’d wager that any time the 19 saves by going faster is cancelled out by the extra wait. The portion in the Tenderloin has one-way splits, and they cross over each other a few times, making the route harder to remember. Perhaps it is useful for connecting some areas to Civic Center Station, but it seems to me more like the old 26-Valencia, which was an infrequent bus next to a bunch of really important and frequent lines running on Mission. The 26-Valencia was discontinued in 2009 for this reason.

Let’s update the map.


Next, let’s talk about the 27-Bryant. This bus has a number of problems. First and foremost, it does serve the Tenderloin, but its route is plagued with so many leapfrogging one-way splits that it’s impossible to remember where the stops are! It seems like planners were trying to follow an imaginary extension of 5th St northwest on the other side of Market. Dragging this diagonal line across a normal grid is a recipe for disaster and the incredibly confusing route proves this.

This line also has a poorly-anchored terminus at Van Ness between Washington and Jackson. I don’t know of anything there that will attract riders other that the connection to the 47 and 49 lines and possibly the Academy of Art. It seems like the decision to terminate the line there was an attempt to keep Tenderloin bus riders out of Russian Hill, especially when you compare the route of the 27 to its neighboring cable car routes (which we’ll get to in a minute).

In addition to problems with the alignment, the bus is not that frequent. I don’t think this is a suitable north-south line in the frequent grid. One more time:


The cable cars are interesting, because they are actually quite frequent. They are anchored at both ends by Union Square in the south and Fisherman’s Wharf/Aquatic Park and North Beach in the north. However, they are specifically routed around the Tenderloin, likely to avoid serving its residents. This shows they are aimed at wealthy Russian Hill/Nob Hill residents and tourists, and aren’t a useful part of the grid, either.

It is important to note that the Tenderloin used to have more cable car routes, and the cable cars used to be a more equitable mode of transit. This fell apart in the mid-1950s when the city government was trying to dismantle the Tenderloin. One of their weapons was removal of cable cars that went through the neighborhood, which city officials claimed were “dragging down” the city. The city said the cable cars were slowing down automobile traffic in Union Square, which it was desperately trying to expand into the Tenderloin. They also promised to replace the affected lines with buses as part of a “modernization” effort.[1]

It’s clear that this was a move to contain the Tenderloin and its people rather than improve the transit network. It wrecked transit access in the Tenderloin and we can still see the effects today. Here’s what we’re left with.


This is the north-south component of our frequent grid in this area. It looks a little bare, doesn’t it? This, too, has problems for anyone living in the center. The one-way splits on the 8-Bayshore and 30-Stockton are extremely wide (4 blocks apart!) and so the area which you can really call “covered” by those buses is greatly diminished. The 8-Bayshore is split all the way up to Columbus, making it especially useless in the middle area.

The distance between the Van Ness buses and the Stockton buses is also very far:


Keep in mind that all of the northbound Stockton buses use Kearny when they are south of Sutter. That means your nearest frequent bus might be half a mile away, and south of O’Farrell, it might be even worse! In the average case, it’s still 8/10 of a mile between routes. Noted transit planner Jarrett Walker often talks about frequent grids in network design and says that parallel lines should be generally 1/2 mile apart and at most 3/4 mile apart, so a gap this large means we are underserving the middle area.

I think we’ve laid out the case for having another frequent north-south line here. Next time I’ll discuss what we should think about when we design it.

  1. The Tenderloin: Sex, Crime, and Resistance in the Heart of San Francisco by Randy Shaw, pp. 96-98