The BusGraphs Project

When trying to solve a complicated problem, I find it useful to ignore obvious constraints at first. It’s obvious that a transit vehicle cannot hold an infinite number of people. Yet when I proposed a redesign of King County Metro’s transit network in Seattle with frequent, 24 hour, seven day per week service, that is exactly what I did. I find that being confronted with a large number of constraints early on encourages an overly conservative end goal.

Beyond the objective of uniformly frequent service, only one constraint governed my redesign. I did not want the new network to increase the number of person-hours required to operate the system’s transit vehicles. This is the predominant expense of transit agencies. Keeping this net-neutral separates a pragmatic proposal from a fantasy. All other constraints could be sorted out later. Thinking in this way produced a network that doesn’t resemble typical transit service, though a rider could easily see the benefits.

Due to a computational error, the data and conclusions in this post may not be accurate.

I’m continuing to explore access-based route productivity measurements by examining four routes that, like route 1, serve Seattle’s Queen Anne neighborhood. Routes 2, 3, 4, and 13 make their way up its eponymous hill. These are some of the quirkiest routes in King County Metro’s system. The former three are putatively long routes connecting downtown Seattle with mostly-residential neighborhoods on either end, but run a variety of shortened trips. Meanwhile, route 13 is linked with some, but not all, abbreviated route 2 trips. It might be more pragmatic to think of these as six shorter routes with a common trunk downtown, but that’s not how Metro treats them. It evaluates the productivity of each route in its totality, so I will do the same. How do these routes compare when it comes to access-based route productivity measurements?

Due to a computational error, the data and conclusions in this post may not be accurate.

I received a package from Amazon the other day, and I noticed something hidden under the shipping label. My first full-time job was writing software for its warehouse operations, and I had a passionate love-hate relationship with it. I had to take a closer look.

PackTypes! I remember those! Without delving too much into the minutiae of Amazon’s warehouse operations, they’re broad categories of packaging. Back then, I saw their use in enforcing appropriate packaging as clunky and inflexible. In my view, the solution was to relegate the PackType concept to the dustbin of history, in favor of granular packaging instructions that would impose restrictions on a shipment by shipment basis. I wrote design documents for this. I navigated the web of systems that referenced PackType and proposed migration plans for each one. I advocated for it as an organizational priority whenever I had the ear of management.

I listen to the Effectively Wild baseball podcast religiously. Generally, though, I despise sports commentary. I associate it with red-faced former athletes ranting about heart, hustle, and generalities concerning who’s a winner and who’s a loser. They recycle a tiresome collection of arguments, invariably returning to the trope that anyone who didn’t play the sport professionally can offer no insight. Effectively Wild is different.

Due to a computational error, the data and conclusions in this post may not be accurate.

I initially wrote this op-ed for the The Stranger. That publication passed on it, citing the busy election endorsement season. I then submitted it to The Urbanist, where it was rejected as a “too extreme of a version of the argument.” While the publisher’s feedback did highlight the risk of overcrowding under the restructure, I felt that it largely disregarded the benefits of transit service that makes journeys throughout the city uniformly convenient, amidst concerns of low ridership and farebox recovery on certain trips. I thought that transit service like this would be compelling to Seattle’s progressive and urbanist press. I plan on reflecting on that in a subsequent post, but, for now, I want to make this piece publicly available. I still think that it is an idea worthy of study.

In the 12 years that I have lived in Seattle, I can’t remember a time when it has felt more bleak to be a transit rider. I find the car-free life to generally be a happy one. In exchange for a little patience, planning, and pliability, I pay a fraction of the typical American’s largest expense after housing. I haven’t worried about parking spaces, oil changes, or gas prices in seven years. The downside is that this leaves me, and those for whom this lifestyle is not a choice, at the mercy of decision makers who can lose sight of what’s important for running useful transit service. That is happening now for Seattle’s transit riders. I find this uniquely frustrating. In my estimation, there is a way for King County Metro, with its present resources, to run frequent service at nearly every current transit stop in Seattle, 24 hours per day, seven days per week, but its leadership seems disinterested in going down this path.

I recently received feedback critical of the Seattle transit network restructure that I previously proposed. It raised concerns about overcrowding during peak travel times—those associated with the commutes for a traditional 9 to 5 office job. My restructure provides uniform frequency on each route throughout the day. Presently, King County Metro and Sound Transit run extra trips beyond the normal frequencies on some routes during these times. In theory, riders distribute themselves among the additional trips, increasing the capacity of the system at the heaviest-traveled times.

I understand this point, but still think it should be questioned. Even with additional trips, a lot of people could still want to travel at the same time, resulting in a crowded bus, followed by a nearly-empty one. Even in that scenario, though, it’s nice that people who can’t or choose not to board the packed bus don’t have to wait as long until the next one comes. I also feel that Seattle’s local transit agencies don’t approach the crowding problem with mechanisms beyond frequency. Different vehicle layouts and passenger arrangement staff can better utilize limited space. Nevertheless, I must acknowledge that there is a physical limit to how many people can fit in a single vehicle. In my proposal, I tried to account for this by giving the Link 1-Line 8-minute frequency and select other routes, including Rapid Rides, 10-minute frequency, on account of their popularity.

The contention that the average commute duration has remained constant over time, and is independent of cultural differences and the built environment, underlies Marchetti’s constant. Technological changes might determine how far one can go, but the time allotted remains the same. That allotment is thought to be an hour total, or 30 minutes each way. I came across Marchetti’s constant early in my exploration of transit planning, and my analyses typically use a time budget of 30 minutes. That being said, I never made an explicit decision that I would use that time budget because it’s Marchetti’s constant. That 30 minutes seemed like a reasonable amount is probably influenced by having encountered it, though.

In spite of that choice, I simply don’t like the idea of assessing the quality of transit using a parameter justified by Marchetti’s constant. That a preferred amount of commuting time is built into human nature seems preposterous. While it’s interesting that this average value shows a persistence over circumstance and time, is there any merit to the average value itself? It feels like the sort of average that folds a diversity of individual preferences into a single, flawed description of a human stripped of agency. There are plenty of people who will tolerate longer commutes, and plenty who would balk at devoting an hour per day to it, and plenty whose tolerance varies depending on the current priorities in their lives. There are trips that people need and want to take that are not part of commutes. How long should be considered tolerable for those? Contemplating what people collectively “prefer”, “are like”, or “will do” seems to inform decision making in the transit planning field, but individuals, not populations, make the choice to board a transit vehicle.

I grew up in Massachusetts. When I go back and spend a day in “Boston”, it might actually involve being in Brookline, Cambridge, Medford, Somerville, Chelsea, Everett, and Revere. That is not my relationship with Seattle and the municipalities around it. Sometimes I don’t leave its boundaries for months at a time. Much of what I need and want to do is located within the city limits. That’s not true all the time, but it’s true often enough that it heavily influences the way that I think about measuring transit access in Seattle. I presuppose that a person in Seattle mostly cares about getting around Seattle. The access analyses of the city that I produce measure the ability to do that.

Making trips within Seattle is, of course, not everyone’s priority all of the time. Looking at access at the city level is also a somewhat awkward way to analyze transit in this area. Seattle does not run or plan the transit service located within it. It provides some funding and participates in the planning process, but most transit in the city is under the purview of King County. Sound Transit, operating the Link Light Rail, exists at the level of a multi-county consortium. City-level access analysis feels like the most instructive tool for evaluating everyday transit quality, but it’s unlikely to be the most relevant tool for the agencies making the decisions. Recognizing that, I produced a thirty-minute transit and walking access analysis for King County. I feared this would be impractically slow and expensive—a fear that was neither entirely born out nor totally meritless. There are caveats to calculating and considering access for the county that are not exposed at the city level.

Due to a computational error, the data and conclusions in this post may not be accurate.

On September 2nd, King County Metro enacted its plan to suspend or reduce the frequency of 20 routes. Faced with a shortage of operators and mechanics, it could not reliably operate its existing service. Meanwhile, I proposed a reimagined transit network for Seattle, subject to the same service hour constraints. In contrast to Metro’s plan, it achieved greater access than Metro’s pre-reduction network. With schedules for the reduced service now available and running, it’s possible to analyze the exact access implications of the service reductions.

Due to a computational error, the data and conclusions in this post may not be accurate.

In the last post, I presented a restructured transit network for Seattle. Overall, the new network, versus that of an existing weekday, improves access by 19.7%, while reducing the amount of service by 13.9%. In theory, this would allow King County Metro to operate transit service at least every 15 minutes, on nearly all the current transit corridors in Seattle, 24 hours per day, seven days per week, while still accounting for its current labor shortage. To accomplish this, the restructure eliminates service that appears duplicative. Most notably, rather than having many bus lines go through a downtown already served by the Link 1-Line, routes are altered to serve Link stations that would allow the trip to continue there. It would be an adjustment for transit riders in Seattle, but one that would improve the ability to use transit to reach destinations throughout the city.

The 19.7% access increase throughout Seattle is an average, and those can hide information. It would be hard to accept a network restructure that concentrated all the improvement in one area, at the expense of many others. A positive change in average could indicate that is happening, instead of a more desirable across the board increase. The access map is useful to see that spatial distribution, but it’s hard to look at two different maps and make comparisons. I wanted to look at the restructure in a different way. In previous posts, I have constructed a variety of measurements based on access. This post applies them to the restructured transit network.