Thoughts on a Jarrett Walker Post

“What if we planned public transit with the goal of freedom?” asks public transportation consultant Jarrett Walker in a blog post from this March. A little over five years ago, a presentation that he gave planted this question in my mind. Having a background in software, and unaware of anyone else who was thinking about this question¹, I started creating the tools that I would need to measure the freedom conferred by public transit systems. Within a year, I had sold my car and quit my job to focus full time on transit. I envisioned myself as establishing my own consultancy that would guide transit agencies through network redesigns using my software for generating measurements of transit network quality.

It never worked out that way, and despite exploring some other avenues², a path to success never seemed clear. Within another year and a half of starting, I was back to working in the software industry. Recently I’ve been thinking about that period of my life. I’ve wondered whether I should continue to have any professional aspirations in public transit planning. Am I adding anything at all? In coming across this post, I achieved no clarity in this quandary, but it did give some definition to the conflict that I was feeling.

This is a Great Post

I like this post and I wish I had written something like it. As a prospective consultant I constantly felt that I was failing to communicate. When I first started, I would find agencies that were considering restructures and I’d email their staff and try to interest them in my access-based analyses. The responses to my overtures, when they arrived at all, often seemed confused. One planner, looking at my measurement through the lens of the four-step model, characterized it as only performing path building. For it to be useful at all, it would need to incorporate existing travel demand and transit mode preferences. I was out of my depth and wasn’t able to formulate a strong response. Walker’s post anticipates this criticism and explains why an access-based analysis need not, and should not, emphasize these factors.

Structurally, the post reminds me of how I’d formulate my dialogue with transit agencies, but with some critical differences. When I’d describe my methodology, I’d start by showing a single isochrone, and then explain what insight would be gained by computing isochrones over an entire region. Walker’s post starts similarly, but from there, it goes in a totally a different direction. To me, the value of this isochrone was too obvious to describe; what was important was describing how it was generated. Walker makes the point of explicitly framing the isochrone boundary as the limiter of freedom.

The “Why Access Matters” section anticipates roadblocks that I never learned how to overcome. As I started attending pre-proposal meetings, I got to know some of the prime contractors who would bid on planning projects. One person at a planning firm revealed that he would be interested in employing access-based approaches more often, but felt that the local agencies had a set of criteria in mind. The post anticipates this. Not buying this high-minded talk that access is freedom? Make choices to improve access and that precious ridership will come.

Often, I’ll lament the planning choices that a transit agency will make and think that I need to do something. I’ll storm around the house composing a response in my head, only to realize I’m probably just unintentionally plagiarizing Jarrett Walker. There are competent, credible people working to bring access-based measurements to transit agencies. I had no aptitude for it. At the same time I’m not entirely comfortable stepping aside altogether. I have some reservations with the details of the post and I think they are worth exploring.

Is There Really a Wall, Though?

A key point in the post’s treatment of access is that people are at the center of a boundary that serves as a wall around their life. An isochrone, though, is a function of position, duration, and starting time. Variations in transit schedule radically alter the boundary on a minute-by-minute basis throughout the day. In short, there is not a single wall.

Many analyses work around this by creating a single boundary that is representative of a time range. Some unsophisticated tools will use the headways of transit routes to compute average wait times for catching vehicles on each route and base their isochrones off of that. More advanced tools (I’m thinking of conveyal’s in specific here), will compute an isochrone for every minute of a time period, and call something reachable if it is reached some configurable portion of those times. This is a much more accurate approach than averaged wait times, but it still throws away information. Consider a a map where a location is considered reachable if it can be reached in 50% of the cases. If changes to the network cause a location that is reachable 49% of the time to increase to 51%, that change will alter the map. But changes that make a location reachable 10% of the time rather than 40% will not have an observable impact. This throws away quite a bit of information about how often a destination is reachable throughout the day.

It’s unnecessary to accept this information loss, but it does require weakening the notion of the wall. A measurement can instead express access to a location not as a yes-or-no proposition, but as a reachability ratio: a ratio of how many times of day it is reachable to the total number of measurement points in the time range. Visually, the map will have to choose a fixed number of color bands that represent ratio ranges. However, a numeric aggregation of access over the whole map can use the exact reachability ratios for every point. This does mean that it’s no longer possible to ask something like “how many jobs are within the wall?”, but it is still possible to compare two locations. Adding up the reachability ratios associated with each job would generate a job access score. It’s a less intuitive measurement, but one that more accurately describes the change in access.

Is this loss of comprehensibility a big deal? Any measurement of access is going to make use of cutoff points for the sake of measurability and explainability. (No one who is willing to take a 45 minute trip is going to consider a destination 46 minutes away totally unreachable.³) From that standpoint, using a hard cutoff in the name of simplification may seem worthwhile. But when each cutoff degrades how well the measurement tracks with a person’s ability to access destinations, each cutoff must fulfill a purpose. Typically, service change analyses present the percent change in available opportunities that a redesign generates. It’s the percentage that’s important, not the raw counts or the raw amount of change. Thus it doesn’t matter if what is changing is a countable quantity or a sum of ratios. Since the latter better reflects the actual freedom that the network allows, I have never felt compelled to use counts. Yet Walker’s post, and many other access studies I’ve seen, both professional and academic, use a hard cutoff in order to treat access to opportunities as a count. That feels like a misstep to me.

What Counts?

Aside from my concerns about how to count, I have reservations about what to count. The post cites the improvement made as a result of a network redesign in Dublin: “the average Dubliner [could] reach 20% more jobs (and other useful destinations) in 30 minutes.” In this case, the other useful destination are student enrollments. While I don’t think there’s anything wrong about calling these important destinations, I’m deeply uncomfortable with most any determination of what locations are important, and consequently, which ones are not. I don’t want to do it myself. I don’t want anyone else deciding it either.

“Importance” feels entirely subjective to me. Is a job important? Not if you’re retired. Is a school important? Probably not unless you’re a student, a teacher, or a parent of a student. Is a methadone clinic important? Not if you aren’t dealing with opioid dependence. Is a polling place important? Not if you’re not allowed to vote. These locations are of deep importance to some, but of no importance to others. Maybe there are some destinations that are objectively unimportant and just those can be thrown out. The Magnolia Bridge in Seattle is just an ordinary bridge over some filled in tidelands that are mostly used as a parking lot. Yet it offers one of the most striking views of Mount Rainier in the city. Who has the authority to assign a level of importance to that?

My approach is to say, if the public may access a location, the ability to access that location via transit should be measured and included in any measurement of opportunity. By providing the public with a path to an area, public institutions have made an ongoing investment in it. They have invested public money into constructing it. If a sinkhole opens, they’ll repair it with public money. If a tree falls across it, they’ll pay a crew to clear it. When there is poor transit access to a location, it restricts access to a public resource to those who can live close to it or can operate, can afford, and decide to accept the negative societal and environmental impacts of operating a private automobile. We allow what is communally supported to become exclusive. If a measurement doesn’t penalize a public institution for doing that, by not deeming every publicly accessible location important, that measurement is not in service of the public.

Another way to look at this is to consider that for most people, even in cities, a private car represents the transportation choice or at least the aspiration. Imagine a car that would ask your destination before it allowed you to drive. If you responded with your cousin’s house or a liquor store, it could decide that the Ford Motor Company has determined that these are unimportant destinations and you’ll have to wait an hour before beginning your trip. People like their cars, in part, because they don’t do this. If the goal of redesigning transit networks is to attract more riders, making access to destinations convenient or inconvenient based on the judgement of an external party seems counterproductive. Public transit systems can’t provide an experience that is exactly like owning a private car, but that is not the purpose of measuring it in a way that treats the convenience of one as a benchmark. The goal is to ensure that any change made to the network narrows the access gap.

My proposed measurement is a simple one. For the area under review (typically a city or a transit agencies service area), I divide it into a grid of 80 meter by 80 meter sectors, discarding any sector that does not contain a public path or road. I then select a time limit, and for each minute of a full day, compute how many combinations of origin sector, destination sector, and starting time correspond to a trip that can be made in less than the time limit. I divide this number by the total number of combinations to create a ratio. Changes that increase access increase this ratio and ones that don’t will decrease it. It would be possible to assign weights to origins, destinations, or starting times. For example, a measurement that counts access to jobs would weight origins by population and destinations by a count of jobs. But by default, I use no weighting. This has been a common source of misgivings when I present this measurement. Most often, people want to at least weight origins by population. This weighting is useful in some contexts—it’s absolutely necessary for doing equity analyses. But using it for all analyses introduces a major problem. It devalues trips between non-residential destinations, such as errands that require multiple stops. I have yet to find a reason to weight destinations or starting times that does not hew to the belief that we can and should make statements about what trips are important. I simply don’t think that’s justified. I’m dismayed that many of the applications of access that I’ve come across, again both academic and professional, are comfortable doing it.

Is This Just the Beginning?

I don’t want to give the impression that I’m disparaging Jarrett Walker’s post. Very many transit agencies would be improved by basing their service change processes around measurements that constructed the most simplistic isochrones and considered a very limited set of “important” destinations. That’s the worst case scenario, and that’s not what being described in the post. In spite of the limitations that I perceive in the measuring process described in this post, having the leadership of more transit agencies on board with this it and committed to acting on it would make me a happier transit rider.

I understand that an access measurement that talks in terms of whole numbers of opportunities is easier to understand and explain than a measurement that frames access in terms of ratios of reachability over a grid of land. I don’t think simplicity alone is justification for use; I can’t get over the two issues that follow from the former. Throwing away information about how often one can access destinations throughout the day feels like a blunder. It attempts to fit reality into a simplified didactic model of a single wall. The idea of deciding which locations contain important opportunities is even more troubling to me. Regardless of the empathy of planners, regardless of the breadth of public feedback collected, regardless of the ability of individuals to suppress their biases, the very idea of an agency deciding what is an important opportunity, and thus what is not, strikes me as arrogant and counterproductive. I cringe when I see access to jobs used in studies. It’s a dreary measurement that shuts out so much about what is good about the human experience.

My hope is that framing access in terms of counting jobs and other opportunities is an incremental step on the path to something better. I wish that Jarrett Walker’s post gave me any indication that that was the case. Either way, though, I find myself in a frustrating place. I can reassure myself that there’s still plenty to do in the realm of software to make access measurements fast, inexpensive, flexible, and universally available. But if it’s just going to be used to count jobs, what has this all been for? It certainly doesn’t feel like it’s achieving the goal of measuring freedom, at least not yet.