“To Predict with Confidence, Plan for Freedom”


The Journal of Public Transportation has a special issue out consisting of thinkpieces by a range of figures in the business.  I’m honored to be there alongside industry leaders like Susan Shaheen of UC Berkeley, Graham Currie of Australia’s Monash University, Kari Watkins of Georgia Tech and Brian Taylor of UCLA, as well as our favorite operations and scheduling consultant, Dan Boyle.

My contribution is called “To Predict with Confidence, Plan for Freedom.”  It basically outlines the argument of my next book, so this would be a great time to hear some critiques of it.  Here’s the opening:

What will urban transportation be like in 10-20 years? How will automated vehicles interact with social and cultural trends to define the city of tomorrow? Will the vehicles of the future be owned or shared? How will pricing evolve to motivate behavior? What will happen to public mass transit? What other innovations can we expect that will transform the landscape? This paper, which is merely the outline of a larger argument, suggests three interconnected answers.

  • We can’t possibly know. History has always been unpredictable, punctuated with shocks, but if the pace of change is accelerating, then unpredictability may be increasing too.
  • We can reach many strong conclusions without knowing. A surprising number of facts about transportation, including some fairly counterintuitive insights that would be transformative if widely understood, can be described and justified solidly with little or no empirical ground, because they are matters of geometry and physics or of nearly axiomatic principles of biology.
  • Prediction may not be what matters anyway. If we abandoned hope of predicting the future, we could still describe a compelling outcome of transportation investment, one that motivates many people who will never care about a ridership prediction or economic impact analysis. We could also predict it in the sense that we can predict the continued value of pi. That idea is freedom, as transportation expands or reduces it.

So if that catches your interest, read the whole thing, and share your comments below!

41 Responses to “To Predict with Confidence, Plan for Freedom”

  1. albaby April 9, 2018 at 8:06 am #

    An interesting piece, Jarrett. I think you’re spot on in emphasizing the sort of freedom (as opposed to nominal *access*) represented by isochrones. A travel mode that goes to lots of places slowly, but few places quickly, will be less valuable to users, other things being equal.

    If you are looking for constructive criticism, I will offer the same comment I’ve made from time to time: the claim that “[c]ities, by definition, are places where space is scarce” is neither definitional nor necessarily true.

    It is true of a certain type of city, epitomized by cities that had significant development before widespread penetration of the automobile: cities with pronounced urban cores that have very high residential and job densities. But that is not the only type of city. There are other large cities in the U.S. where existing transit use is incredibly low, and where well north of 80% of commuters travel to work by car. These are cities with dispersed, low-density development patterns for both jobs and homes. They are cities where nearly all originations and destinations, apart from a very small proportion of peak-hour trips, are in locations where space is not “scarce” as I think you’re using the word (a sense of geographic and spatial limitation, not economic scarcity). Cities like Houston, Dallas/Fort Worth, Phoenix, San Diego, Las Vegas, San Antonio, Oklahoma City, Charlotte.

    So when you write that freedom is largely predictable if you look at “the design of a transportation network in relation to urban structure,” I think that’s true only to the degree you make a clear-eyed and accurate assessment of the urban structure.

    • Jarrett Walker April 9, 2018 at 8:19 am #

      It’s a big problem that we don’t have a word for the “dense city.” But if your city has traffic congestion, then it’s a place where space is scarce relative to demand, and these issues begin to apply.

      • albaby April 9, 2018 at 9:20 am #

        Perhaps if you are planning to flesh this argument out in a much longer work (as you suggested), it might be helpful to talk about the “dense city” rather than cities broadly – or clarify that you’re going to use the term city only to refer to those types of cities. Someplace like Jacksonville may not be anyone’s idea of The Big City….but it’s still a city of a million people.

        In such cities, there may be *some* traffic congestion – but not necessarily enough to make any other transportation mode apart from private passenger automobile feasible or necessary. The non-monetary cost of queuing for a small number of trips at limited times may be small, and may be the most efficient response to congestion, since most of the city has few space constraints. There may be economic reasons why they don’t build a roadway network with more capacity – the sheer impossibility of space constraints may not be a significant factor, or a factor at all.

        If you’re writing for a U.S. audience, I think it might help if you just clarify which cities you’re talking about, rather than make this a definitional attribute of “City.” There’s about 32 million people who live in just the nine low-density, low-transit-using metros I listed above.

        • Jarrett Walker April 9, 2018 at 10:01 am #

          Thanks. This is helpful!

        • Dorian April 9, 2018 at 10:17 am #

          One thing I find especially interesting is that even in what you call low density areas, transit is still more efficient than private cars as a means of transportation. Especially during the peak, even low density areas depend on quality public transportation. So if “freedom” is your goal, you should pay for more (efficient) suburban transit. Not because it is the most efficient expense a transit agency can come up with, but because it is vastly better than relying solely on private vehicles. I am thinking specifically of the south bay area, like san jose. Services that are separate from traffic like caltrain do really well during the peak, because they are not stuck in traffic. School buses are also popular in some areas.

          There are really two things a transit agency should pay attention to when measuring a transit route. I think these are 1. cost, and 2. how appealing the service is in isolation. Cost doesn’t really translate to service one to one (speed, one wayness and labor differences for example). So obviously level playing field would be needed to justify these suburban services, unless massive budget increases are considered.

          • albaby April 9, 2018 at 11:55 am #

            Dorian writes:

            “Not because it is the most efficient expense a transit agency can come up with, but because it is vastly better than relying solely on private vehicles.”

            That’s open for debate in low-density areas, where the costs of providing transit are higher and the efficiencies lower. But more importantly, I think that Jarrett’s piece (and perhaps upcoming longer work) is an attempt to isolate the question of cost/resource allocation efficiencies from more fundamental transportation issues. In other words, that *even if* private passenger cars were to become more “efficient” in terms of cost, emissions, or time, that the spatial geometries of cities would impose constraints that prevent their replacing transit.

            I look forward to seeing that larger work. My own inclination is that this is unquestionably true in the intermediate-term (and shorter) for the half-dozen cities that make up 70% of U.S. transit use (NYC, DC, SF, Boston, Chicago, Philadelphia); that this is possibly true for the next handful of cities that are trying to adopt their urban form to that model (Seattle, Portland, Austin, Los Angeles, Denver); and that it is not true for the rest of U.S. urban areas that are low-density, have a relatively small portion of jobs in the urban core, and therefore have relatively little transit use today. So I would love to see Jarrett’s take on those different types of urban form.

          • Dorian April 9, 2018 at 2:49 pm #

            I definitely support and understand the value of all day, high frequency. And of good network design. I was in Athens recently and it strikes me as a city that is very much dependent on transit, because it is dense. (And their bus system is very complex, as opposed to connective)

            So, good on Jarrett for the piece, it was an intriguing read and I cannot wait to read the book (I’ve already read human transit).

            As for the “That’s open for debate in low-density areas”, not really. A bus takes maybe two times the space of a private vehicle, and during the commute I’ll say an average of two people per car (which really I think the median is one), then you’d only need a bus with four people to be better off than driving in terms of space. Cost is variable but suburban expresses and peak service tends to be a “high ridership” investment, at least that’s what I read in the transit choices report for the VTA. I wonder what doing peak service well would look like…

        • Jack April 9, 2018 at 11:44 am #

          Density is a spectrum. However, cities are always more dense (and therefore have less space per person) than non-cities, which seems to be Jared’s point. Your debate seems to be on degree. Maybe a clarification would be a breakdown of density levels and at what points different constraints enter the equation.

          Of course, the problem with density is how and where you measure it, which Jared has also talked about.

          • Jack April 9, 2018 at 11:45 am #

            I’m sorry.. Jarrett, not Jared. It’s been a long day.

          • Sean Gillis April 11, 2018 at 5:01 am #

            I’m an urban planner, and would be very interested to see a discussion of density levels and at what point different constraints enter the equation. I have looked for this kind of empirical information, but it doesn’t seem to be readily available. The minute you move into this kind of work, you start to make a huge number of assumptions: likely trip choices people will make; number of vehicles people own; cost of parking; quality of transit service; distribution of jobs and services; etc. These can vary widely from city to city. The amount of cars people own and how often they drive is also pretty tightly correlated to income.

            The legacy cities are clearly where this problem of space is toughest. In Canada, that means Montreal and Toronto, our only two industrial-era metropolises. It would also include the cores of smaller cities like Halifax, Ottawa, Quebec City and Hamilton.

            In my hometown of Halifax (urban and core suburban population just over 300,000) congestion is a problem at choke points entering the city, and at points throughout the inner city, which built out between 1750 and 1950. The core of the city is a mix of old walking city and streetcar suburbs. Residential densities (at a neighbourhood level) in the core are between 4,000 ppl/ square km and 12,000 ppl/ square km. Job density in the central business district are probably above 25,000 jobs/ square km. This is definitely high enough density to produce consistent delays* in rush hour, and tough conditions for finding parking.

            The other places with consistent congestion* problems are the suburban big-box shopping areas, especially evenings and Saturdays. So low density can produce congestion as well.

            * With the proviso that congestion and delay is relative to the expections/ norms of whatever city we’re talking about. We have bad small city traffic!

        • Mike April 9, 2018 at 11:50 am #

          The VA part is in reply to RossB’s argument below.

      • Mike April 9, 2018 at 11:36 am #

        Cities are still places with less space per person than non-cities. Even the recent low-density car-oriented cities are denser than rural areas. The most useful definition of urban vs suburban is walkable vs non-walkable, but not everybody accepts this and some readers may consider it an insult, because it leads to paradoxes like most of San Jose being a million-person “suburb”, and Bellevue, Washington’s central area becoming a “city”. And then what are Everett and Tacoma, which were historically called cities but now appear to be suburbs? It may be best to just use “urban” and “suburban” and leave “city” and “suburb” for whatever the area’s popular term is. Christopher Leinberger has two useful terms: “walkable urban” and “driveable sub-urban”. That clearly states the difference and avoids confusion, although it can still sound like an insult to suburbanites (so it may be less acceptable for Walker’s purpose than for Leinberger’s).

        The US is an outlier among industrialized countries in our low level of both urban, suburban, and rural transit. So the answer is, we should double all of it, and make sure not to leave out intermodal connections like airports and ferry terminals, or recreational opportunities like trailheads, or all-day rural and exurban service including weekends. That doesn’t mean a bus to every isolated house, but reasonable service between town centers on arterials. And last-mile fiexible-route service to every house can be filled in after these larger gaps are addressed.

        It is like the VA in some aspects. Neither the VA, nor public transit, nor Johnson’s “war on poverty” (as the need evolved in the 1970s and 80s, and racial gaps were eventually recognized) ever got enough resources to solve their problem and deal with the increasing demand that came with the natural population increase and the rising inequality caused by the same people who want to shut down those programs. In contrast to other countries that put the public first and just solve the original problem, at least to a large extent. So the people who artificially restrict resources below the adequate level turn around and say the program isn’t working and government isn’t effective so let’s shut it down or privatize it. Again it comes down to the need to put the public first, and to view transit as “access to places” as Jarrett says.

        • albaby April 9, 2018 at 12:55 pm #

          Mike wrote, “The US is an outlier among industrialized countries in our low level of both urban, suburban, and rural transit. So the answer is, we should double all of it….”

          Not necessarily. After all, it’s very expensive to “double it” – and in the suburban and rural areas, you might not get very much bang for your buck. So if only 2% of trips are by transit, and you “double” the service (and expense) and only get 3% of trips now by transit, you’ve spent a lot of money for little benefit. The argument autonomous vehicle supporters make is that you can probably provide better transportation for those residents for a lower cost if you just give them access to cars (through an AV program) than if you try to do fixed route transit.

          I think that Jarrett’s point in this article is that if you look solely at the spatial geometry of the urban environment, you can isolate out questions of what particular mode is more costly or more popular. Based on *spatial* limitations, it doesn’t matter whether someone comes up with a new private passenger car that’s super cheap – physics limits the extent to which private cars can be deployed in cities. *That* claim is very dependent on the density of the city you’re talking about, because spatial limitations are very different in Jacksonville from Manhattan.


          • Sean Gillis April 11, 2018 at 5:10 am #

            If your only goal is to grow ridership, than yes you’re not getting much bang for your buck by doubling your service and going from 2% to 3% mode split. But is your goal ridership? Or is it providing a better travel option.

            I’ll have to re-read the section from the Human Transit blog and the book, but I’m not sure we have a really great way to talk about the effectiveness of coverage services. There’s a small town (pop. 8500) near where I live that just started a bus route. It costs $200,000 a year, and generally serves 60 people a day. Is that worth it? I have no idea, but it sounds like the service is helping a lot of people.

            What’s the value of helping a small number of people a lot? How do we ration our resources? What choice might they make in Omelas?

          • albaby April 11, 2018 at 6:37 am #

            Sean, in this context it’s not a question of *whether* to help a small number of people a lot – but *how* to help them. In areas like you describe, it might be cheaper to provide transportation services to those ~60 daily riders through private passenger cars than a bus route *if* autonomous vehicles become cheap enough. And it might be a better travel option for them than buses. In a small town, you’re (usually) not looking at the type of space constraints that Jarrett has argued compel mass transit.

          • Sean Gillis April 16, 2018 at 10:08 am #

            Autonomous vehicles could change the price point/ quality of service hugely. Will be interesting to see if the timelines for autonomous vehicles live up to the hype.

    • Rick R April 17, 2018 at 5:46 am #

      Space may not be “scarce” in car dependant cities, but the very need to provide parking in the core or dispersed clusters of a city increase the scarcity of land. This pushes up the cost of land and investments for all types of land use in these areas.

      The fact that many car parks in these cities are ground level only helps conceal how much land is being consumed by transport through their lack of bulk. It is common for over 50% of land in a free-standing supermarket to be taken up by car parking, with customers who are pedestrians or have used public transport for their journey required to walk across this carpark to gain access.

  2. RossB April 9, 2018 at 9:30 am #

    Looks good to me. The basic argument is that space is limited, and no change in the way vehicles are operated will change that.

    I agree, but I think there is another argument that is just as interesting, and to my knowledge, lacks much study. That is that automated vehicles could represent a breakthrough that could provide much of North America with high quality transit at affordable prices.

    It is a complicated argument, and I would love to see some mathematical modeling for it. But the basic idea is that an automated transit system (consisting of vans and buses) would provide better transit than that found in just about any city, at a fraction of the cost of automated taxis.

    Proponents of automated taxi-cab systems have claimed that futuristic systems could provide the same level of quality as most transit systems at a lower cost. That may be true. But if so, then it stands to reason that a futuristic transit system would provide service that is better than the automated taxi-cabs, at the same cost.

    As operating costs shrink to zero, vehicle cost (and maintenance) becomes the biggest issue. The more vehicles, the more expensive it is to operate the system. While the miles that each vehicle travels is important (and complicated) the total number of vehicles is probably the biggest cost. In each case, peak usage is key. It really doesn’t matter if you can get by with a dozen vehicles at 3:00 AM — what matters is how many you use during rush hour. There are really three possibilities.

    1) Pure taxi-cab. In this case, you don’t share a ride (unless you are going to the exact same place). This is very much like a traditional cab. During rush hour, this is the least efficient system. If there are 10,000 people trying to get somewhere at a particular moment during rush hour, you will need 10,000 cars.

    2) Dynamic, ride sharing taxi-cabs and buses. On the surface, this sounds very promising. A “smart” system, that knows where everyone is headed, would be able to send the appropriate vehicle to your door. You would, in turn, pick up other riders along the way. If lots of people are headed your way, you get picked up in a bus, not a car. This is the only way that you could compete (in terms of vehicle cost) with:

    3) Fixed route, fixed time transit system on a grid. Buses would run on the most popular corridors, vans on less popular ones. There are efficiencies gained by requiring riders to walk a few blocks or make a transfer.

    That is where the mathematical modeling comes in. How often is the dynamic system better than the fixed route system from a user or cost effective standpoint? The dynamic system benefits from one sort of efficiency: the vehicles are empty only when they are on the way to pick someone up. At 3:00 in the morning, a traditional bus might travel the streets for its entire run without anyone on it. The dynamic system also provides a better ride — for some. But as efficiency is gained, ride quality deteriorates. If you are the first one being picked up, but the bus zig-zags all over town picking up up riders, then you would be better off walking a couple blocks and riding the main line. It also sounds great to be picked up “on demand”, but there is always some sort of lag time between when you request a ride, and when you can be picked up. As any New Yorker can tell you, you can often get on a train faster than you can get into a cab. As the system becomes more efficient, fixed route, fixed time service is actually better for the rider, as well as more cost effective.

    My guess is if someone actually does the math, they will find that the fixed route system becomes a much better system for the vast majority of riders than a dynamic one, assuming you spent the same amount of money. Since a dynamic system has to have enough vehicles to provide for peak usage, you are talking about a lot of vehicles. Enough to provide most cities with five minute headways on every corridor, in a grid. Even in smaller cities, like Spokane, this would provide high quality service at a relatively low cost.

    It seems quite possible that even within the same area, a system would transition from fixed route to on-demand in the wee hours of the evening. If ridership is really low along a corridor, that corridor itself could be on-demand. If it is really low overall, then the entire region could be on-demand. But my guess is that it really takes very few rides for a fixed route system to be more cost effective. Sending empty buses (or in this case, cars) down the street seems very wasteful, but an on-demand system has empty vehicles as well (on the way to a pickup).

    The biggest obstacle to that sort of transit system is not technological, but political. Replacing a public entity with a private one is a popular idea, regardless of the evidence. The V. A., for example, is considered a model for health care and is a leader in several areas (use of nurse practitioners, electronic records, etc.) but privatization efforts continue, without any evidence that care would be better. Inadequate financing is the root of the problem (way too many veterans, and way too little money) but it is much easier to blame “the government”.

    The same thing could happen to transit. Private companies and anti-government ideologues have every incentive to avoid something that could provide high quality cost effective mobility for most people. They wouldn’t have to show evidence that it actually works better.

    • Mike April 9, 2018 at 11:48 am #

      The most efficient system would be driverless buses and trains for most of it, and driverless cars for the smallest gaps, and for the inevitable “holes” in the transit network (such as where going from an arbitrary A or B requires a long detour to C, simply because you can’t have stations everywhere and lines in every direction). The danger is people seeing driverless cars as “making transit obsolete”, leading to underinvestment in transit right when it’s starting to catch up to its backlog.

      I’m still skeptical that driverless vehicles can perform on anything more than limited routes. So this future of driverless taxis may be several decades away rather than one or two. But if Jarrett is making a long-term argument, then maybe it doesn’t matter whether it takes several decades or never.

    • Sean Gillis April 11, 2018 at 5:14 am #

      “The biggest obstacle to that sort of transit system is not technological, but political. Replacing a public entity with a private one is a popular idea, regardless of the evidence.”

      Privatizing is popular now, in the English speaking countries, inthe post Regan-Thatcher era. Hard to predict if that will continue.

      • Jarrett Walker April 12, 2018 at 9:02 am #

        Actually, the pendulum is swinging back away from the radical privatization of the Thatcher era, and toward re-assertion of gov’t control of planning, at least in Australia, NZ, and Ireland, and to a lesser extent the UK. And it’s a good thing too. I clear up this distinction here: https://humantransit.org/2010/09/on-privatization-scares.html

  3. Georgist Economist April 9, 2018 at 3:38 pm #

    “can be described and justified solidly with little or no empirical ground”—err, what? Your insights are supported by empirical observation on the one hand. On the other, if they were refuted by observation, they would have to be discarded as flawed (even if the actual flaw isn’t found). Page 2: “ours is an age when empiricism reigns.” It is the foundation of science that experimental observation has the final say. Easier said than done in the fields that deal with human beings, naturally.

    “These principles are undoubtable axioms of the world at human scale, the world we are talking about in urban planning.”
    Unfortunately, much of the urban sprawl—known for being unwalkable, automobile-dependent—is decidedly not at a human scale but at automobile scale. In fact, there is more than one way in which some parts of cities can be not-human scale: urbankchoze.blogspot.hu/2014/11/malls-and-urbanism.html picture 6, “commercial strip” classification explained here: urbankchoze.blogspot.hu/2014/04/transport-concept-door-to-door-or.html

    “Because borts have invented such a vehicle, their cities are now too large to be reached solely by hopping, drifting, or slithering.”
    Indeed, the bortcar’s speed increases the size of the city and the available space per bort. In fully general terms, if the bortcar (or the bortmover, for that matter) can move (on average) N times as fast as an unassisted bort, then there is (approximately) N^2 times as much land available, from which travel time is the same via bortcar (e.g. within Marchetti’s constant). Consequently, it is not a geometric contradiction if the bortcar takes N^2 times as much space as an unassisted bort, because overall space per bort grows in exactly the same proportion. This is the same conclusion that can be drawn by stepping through supply and demand: the greater speed of the bortcar creates a larger supply of land, which drives down its price, leading to more land being used less efficiently. This lower efficiency use can very well cover the share dedicated to bortcar infrastructure.

    Imagine a world where the “flying car” is a viable (safe, affordable) method of transportation. At a speed of 2-300 mph, even with last-mile(s) driving, the area of a practical isochrone divided by several million people gives less than 10 people/acre average density. This is low enough that the space required for runways is not an issue. Of course, concentrated office or retail might run into a problem (though if a few miles of driving on the ground as a car are included, even this limit is dramatically loosened).

    Now, it is entirely possible that bort society has some other reason for incompatibility with the bortcar.
    – It is possible that “zoning” restricts non-residential uses to such a small fraction of the available land that traffic converges there to a disproportionate degree, and space is scarce due to this self-same zoning.
    – It is plausible that the number of borts per vehicle weight is more favorable, the larger the vehicle, and/or that the bortmover can exploit mechanical returns to scale that the bortcar cannot, leading to lower cost per bort-distance.
    – It is possible that bort cities (perhaps individual buildings, but mostly the street network) are significantly slower in responding to changes in transportation technology than those technologies change. Thus for a very long time (possibly much more than the average bort lifespan) there is a transient effect, where bort cities built around a slower bortmoving vehicle’s speed and size continue to exist without significant differences, even though they are surrounded by a skirt of city built around a later, faster and more space-extensive vehicle, and thus there is temporarily (for several bortlifespans temporarily) a traffic of such nature that the old city was never supposed to encounter. Eventually, equilibrium could be reached, whereby the old bort city would be somehow rebuilt in the likeness of the surrounding new city. However, this can be largely theoretical, primarily because the “target” equilibrium itself moves as yet faster and larger vehicles are invented. Note that this paragraph doesn’t say anything about such adaptation being desirable.

    “high frequency grid”—not an actual objection, but I regularly wonder whether other patterns, e.g. triangular/hexagonal, would be better than a rectilinear one.

    “Prediction and freedom are opposites: to the extent we can predict your behavior, you are not free.”—This is philosophical, but I disagree. Just because anyone can correctly predict that (reductio ad absurdum) I won’t jump out the window today doesn’t mean that I’m incapable of doing so or that in any sense I am not free of doing so. It’s only that I have no intention whatsoever to do so.

    • Jarrett Walker April 10, 2018 at 8:32 am #

      I can’t believe that you don’t accept the notion of axiomatic knowledge — the kind of knowledge that geometry describes or that follow from our definitions of concepts. Do you really need to do experiments to know that an elephant can’t fit in a wineglass?

    • Jarrett Walker April 10, 2018 at 9:19 am #

      Your window-jumping case is not the interesting case. It’s not hard to predict behavior that’s necessary to survival, as this is almost an axiom of biology.

      Freedom of speech and association are the interesting cases, and these involve highly unpredictable things like cultural and religious expression, the arts, political action, and simply the expression of personal tastes.

      • Georgist Economist April 11, 2018 at 10:37 am #

        Induction and axiomatic knowledge are different things. Generalizing laws from experiments works fine and obviates the need to do the experiment fitting the elephant into the wineglass. However, induction always leaves falsification open, so if it turns out that our previous induction was erroneous, we can throw it out. But once you declare what could be a theorem to be an axiom (and perhaps have an “unlucky” set of already-declared axioms) you could very well derive insane results, or a contradiction.

        There is a great multitude of examples where a bad assumption lead to ridiculous conclusions. I like the story of a village where, facing a crowded commons, the locals assumed the number of cows was an externally given variable, and tried to extend the commons as a solution. When it didn’t work, they concluded they must not have extended it sufficiently, and tried again. Furthermore, when a politician suggested introducing a fee for grazing cows on the commons, the villagers booed him. They already paid several sorts of taxes on cows, they said. You will recognise that this is analogous to cities trying to widen roads. Even better, this is just as much about predictions: engineers predicted steady growth in VMT, politicians tried to build just as much road capacity that would forestall congestion, and by induced demand, this resulted in approximately as much VMT increase as the engineers predicted.

        An example I think you will like is this quote:
        “In a void, no one could say why a thing once set in motion should stop anywhere; for why should it stop here rather than here? So that a thing will either be at rest or must be moved ad infinitum, unless something more powerful gets in its way.” — Aristotle, Physics, Book IV, section 8
        (from en.wikipedia.org/wiki/Horror_vacui_(physics) )
        If you squint a little, this is Newton’s first law! “In an inertial frame of reference, an object either remains at rest or continues to move at a constant velocity, unless acted upon by a force.” Except Aristotle took an incorrect equation of motion as an axiom, promptly ran into a division-by-zero error, and declared that vacuum cannot exist. Had he done a few experiments, he could have founded classical phyics in the classical era, beating the likes of Galileo and Newton by two millennia or so.

        I think by freedom and its incompatibility with prediction, you meant the philosophical debate of “free will versus determinism”.

        • Jarrett Walker April 12, 2018 at 9:13 am #


          I agree that we may be deluded about what constitutes an axiom, and much of my work is about clearing up obvious cases where people confuse their narcissism with immutable facts about the universe, as in your “tragedy of the commons” example.

          But by axiom I really mean “a thing that it is not possible to doubt.” Biology cannot begin without a concept of what counts as an organism, which clearly involves seeking resources and producing more energy from these resources than you expend in seeking them. Likewise, our daily life cannot proceed without our certainty that we experience the human world in three dimensions of space, that objects do not fit into containers smaller than themselves, etc etc. That’s the kind of knowledge I’m trying to build on.

          I am also open to the possibility that some future insight may cause things that I see as permanent to turn out to be contingent or even wrong at some higher level. Indeed, physics has already done that. But for the purposes of talking about cities and transportation, which is my subject, we do live in Newton’s physics and Euclid’s geometry. This is what I mean about “human scale,” as opposed to the realms of very big and very small where other rules apply.


          • Georgist Economist April 12, 2018 at 11:57 am #

            Certainly. I thought by “axiom” you meant its theoretical definition; apparently, I misunderstood what you were trying to say.

  4. albaby April 10, 2018 at 10:23 am #

    Doesn’t that rather depend on the size of the wineglass? 🙂

    I think that Georgist Economist is presenting the same issue that you and I talked about up top – that your assumptions about what a “city” looks like are not axiomatic. Thus, the type of transportation that can fit into a “city” (the relative size of the elephants and wineglasses) isn’t axiomatic, either.

    It is axiomatic that borts can’t simultaneously exclusively rely on bortcars and have cities that look like NYC or London. But they could easily rely on bortcars and have cities that look like Jacksonville. It’s not axiomatic that they would choose the former type of city, and not the latter – that depends heavily on the attributes of bortcars relative to transit.

    • Jarrett Walker April 12, 2018 at 9:15 am #

      You are right that I need to clarify my definition of a city. Or more exactly, when I say that a city is people living close together, so that there’s not much space per person, that IS my definition. Any patch of ground with some people can legally call itself a City, but that’s just a brand name, not a meaning. Anyone who knows the difference between a banana republic and Banana Republic ought to be able to track that.

      • albaby April 13, 2018 at 5:49 am #

        I think that would help – especially since I don’t think your “people living close together” definition is nearly as self-evident as your comment above suggests. There are plenty of metros in the U.S. where people are living close together, but have virtually no mass transit usage and still function. Jacksonville, Phoenix, San Antonio, Charlotte, Tampa, Raleigh, Indianapolis, Orlando, Memphis, even Dallas…. These are all metros with *commuter* transit mode shares at 3% or below, and thus are not compelled to have transit as a mandatory solution to spacial limitations – but are certainly “cities” as that term would commonly be understood.

        That’s the open issue, I think, with your axiomatic argument. It would not be difficult to design a community for our Borts that had no transit whatsoever, and relied entirely on Bortcars, that still had all the Borts living fairly close together and would scan as a “city” to our eyes. That “city” would look more like Oklahoma City than Seattle, of course – but it would still be a city, not merely a patch of ground where rural residents have decided to re-brand.

      • Georgist Economist April 14, 2018 at 10:11 am #

        As albaby says, one of the easiest ways to define “city” is in terms of economics and transportation isochrones. A blot-shaped area of land where, for every point P in the area, the isochrone drawn at P (during the day) covers a huge number of residences, jobs and retail.

        It is the large number of people within reach that enables either large or “picky” institutions and employers, because enough people with the desired uncommon trait(s) live within the area. From high-tech companies (what fraction of people are competent programmers?) and universities to by-now-obsolete “classical” factories employing thousands of people, transportation is vital. The same can be said about malls and other, large-scale retail establishments, as well as things like zoos and museums. Also, from the point of view of residents, a wide variety of jobs are available.

        Now, there is an alternative to cities, which I call the phalanstery. Locate the factory in the middle of nowhere, build housing around it, and a handful of shops to serve the residents. Centrally organise visits to e.g. zoos, and on the selected day lease a few buses to take the interested people there.

        There are a couple of drawbacks to this system, though. One is that if somebody’s employment or social status changes (e.g. entering a higher level of education), they need to move. Another is the paucity of choices, also known as personal freedom. If you wanted to go to a different museum, or on a different day, you are out of luck.

        In abstract terms, in the city a large number of people can repeatedly use the transportation to self-select by employment, by retail choice, by entertainment choice, etc. Whereas in the phalanstery, out of the same original pool of people some are selected by employment—and that’s it. The other choices are left on the floor, as it is neither possible to select according to all choices (due to combinatorial explosion), nor is there a transportation network that the people can use to self-select.

        • Georgist Economist April 14, 2018 at 12:46 pm #

          I only left out the main part. 😀

          Nothing in the above definition/description implies high population density *in terms of space*. Cities require high population density *in terms of travel time*. The average speed of movement (and headway, if applicable) does give a minimum spatial population density; however, this lower limit is inversely proportionate to the square of the average speed.

          This constraint (and the general concept of cities being defined by transportation) generates several varieties of Platonic ideas for the forms cities can take.
          ⚫ Medieval city: limited in size by the speed of walking. Extremely high density. Some
          principal streets are 6 meters (20 feet) or wider, to allow two horse-drawn vehicles to
          pass each other, but most streets are as little as 2-3 meters (10′). If only people (and
          hand-carts) are the only thing that ever goes there, you don’t need more.
          ⚫ “Japanese” transit city: a low number of heavy rail lines (metro and suburban service
          on mainlines) define the city. Frequencies are high, capacities are astonishing.
          Adjacent to each station of these lines, extremely-high-density buildings stand
          (multi-story indoor malls and residential or office towers), but density and building height
          decrease quickly as distance increases. The map of built density, or the city’s skyline,
          clearly resemble isochrones. Because many important parts in the city (trip destinations,
          in particular) have very high density, cars often cannot penetrate the city. Residential
          streets are often narrow, 3-4 meters (10-13′) are not uncommon.
          ⚫ “European” transit city: a comparatively high number of light rail lines and bus lines.
          These lines usually form a somewhat regular high-frequency grid. Built density is much
          more even than in the case of the “Japanese” city. If heavy rail is present, it tends to
          serve as a shortcut in the tram/bus lines, i.e. passengers tend to connect to/from a
          surface line, rather than directly the origin/destination of the trip.
          Obviously, these are ideals, and heavy rail lines in particular will always tend to show
          traits of the “Japanese” city. Where multiple heavy rail lines meet in a small space, such
          as under Midtown Manhattan, you get Midtown Manhattan.
          Unfortunately, perhaps, because most of the city is at medium density, the appearance of
          cars is possible.
          ⚫ “American” pre-car city: as far as I can tell, an exception. My understanding is that many
          American cities were built to plans from the ground up, rather than “evolving” over time.
          Consequently, many of their characteristics reflect the aesthetic tastes of their designers,
          rather than the transportation-economic reality of their day. In particular, most American
          cities (and centrally-planned features of European ones) have VERY wide avenues even in
          settlements barely qualifying as cities. Residential streets in American towns are often over
          10 meters (30′) between the outside edges of the sidewalk, and then there are front lawns.
          My understanding is that many parts laid out BEFORE cars today accommodate a traffic
          lane in each direction, a parking lane in each, some green buffer, sidewalk. Again, this is
          about non-main streets where the dominant for of mobility was walking, for which 3 m is
          known to be sufficient.
          ⚫ American automotive city: when the oversized streets of the previous were filled by cars,
          it started to evolve into the more modern urban sprawl. Residential streets patterns are no
          longer connected grids, but discontinuous, to exclude through traffic. Unfortunately, often
          pedestrian shortcuts are left out. Trip destinations are surrounded by swathes of parking.
          ⚫ “Mature” American automotive city: in addition to the above, a grid of grade-separated
          highways provide high-speed connections across the network of medium-speed streets or
          avenues. The built form of the city becomes very low-density, a.k.a. urban sprawl.

          These “ideal” cities only reflect the economics of transportation, and assume that anything could be built anywhere. The reason anything is *not* built is because it is simply less profitable to build that, than to build something else. In practice, especially in America, there are *very* strict zoning codes. Not only are offices, retail and residential areas completely separate, even different building types are zoned separately. This causes land to be used inefficiently. Unfortunately from this point of view, even if zoning were completely abolished, the street network would stay unchanged. This is a problem because, for one, the sizes of building plots determine the way in which they can be used. For another, if the street itself is oversized, that cannot be rectified in a simple way.

          • Philip George Hayward April 30, 2018 at 10:15 pm #

            Georgist: your classification of city types through history and evolution, is helpful knowledge. I wish I had more time to comment on this thread. I just want to add some comments to your very helpful one, seeing you are obviously so interested in the essential details.

            Shlomo Angel et al’s “Making Room for a Planet Full of Cities” is full of useful information. So is the UN Habitat Program report “Streets as Public Space and Drivers of Urban Prosperity”. There is a lot of overlap between cities of the “different” classifications, in the developed world. Every first-world city has had a sustained era of suburban sprawl, falling population in their old core, and falling overall “urban area” density as they expand spatially. Most cities have not had a reversal of this trend. Some special cases have, but this has more to do with the kind of local economy they have been lucky to evolve as. Some local economies, “global finance” based ones, for example, are what they are because they are unique. If it were possible for them to be ubiquitous, there wouldn’t be any of them.

            In so far as “old” cities are still denser than newer ones, it is because their density fell from a very high start point. Without this high start point, it is impossible to replicate the old cities densities in new cities by the currently fashionable urban planning approaches.

            The “intensity of street network by city” data in the UN Habitat Program report contains many surprises. Pre automobile “planning” if it was done, tended to allocate a lot of space to streets, much more space in fact than automobile based development does. This is of course because the actual development in between the streets is of such different density. But it does make eventual evolution of a post-automobile locale into something higher density, very problematic, as it is likely to implode under its own congestion diseconomies and the lack of space to allocate between multiple urban “public purposes”.

            It also makes “economic development” in a chaotic, under-planned third world city very problematic, something the UN authors were wishing to emphasize. They suggest an observable correlation between a high proportion of surface area dedicated to streets, and sustaining a relatively high density, successful urban economy. Manhattan has exceptionally high proportion of its surface area dedicated to street space. Phoenix does not. The outer suburbs in New York also do not. Suburbs everywhere, as a general rule, do not have an intense street network.

            Highways per se are not the essential cause of “sprawl”; as long as a rural road network exists and people are free to “develop”, the cheapness of rural land will always attract escapees from the urban “economic rent” trap. Then government attempting to cope with the resulting infrastructure needs, would be best advised to get maximum lane-miles for the dollar with low-cost surface arterials. Highways are a waste of money because most travel on them is at surface-arterial speeds or less anyway.

            You would enjoy my online essay “The History of Urban Land Rent and Cyclical Volatility”, being a Georgist (although it might challenge some of your current assumptions). I also recommend my “Rocket Scientists Haven’t Forgotten about Gravity, so Why Have Urban Planners Forgotten About Land Rent”? This discussion thread and the headline article, is typical of this problem.

            Of major importance, is what Jarrett Walker touches on in his point: “Reduce the amount of travel through mixed-use urban planning. A more egalitarian way to the same end is to design the city to minimize the travel borts need to do to do whatever they do. We call this mixed-use planning.” But it would be more correct to say that planning prevents mixed uses; mixed uses do not need to be “planned”, they need to be “allowed”. This is another massive natural evolutionary trend in cities to whatever extent it has been allowed; employment disperses, and whether the employment is following the people or the people are following the employment, is a question that appears to have complex answers: either one can be true under different conditions.

            It should be obvious that co-location efficiencies are the biggest game-changer. A hypothetical extreme, would be everyone living on 4 acres, but working at home. But planners are thinking in the exact opposite terms to “beneficial dispersion” because they want to concentrate jobs and households on mass transit routes.Then they come up against another reality that they don’t understand, that “land rent” increases at the locations they are trying to push everyone into, which “prices out” people.

            Ironically dispersion of everything (employment and amenities) and a flat urban land rent curve, and housing of all types at all locations being more affordable, will enable far more people to achieve a walking or cycling based lifestyle if they want it; it is a matter of them co-locating with their favorite transport destinations. It is unnecessary to “concentrate” anything other than in small splattered nodes, to help enable walking and cycling – concentration is purely for transit, it is transit that will not work at all under dispersion, whereas walking and cycling will do just fine (provided the design of the streets is OK). I have lived in “suburbs” all my life and have always been a walker, to the local amenities which have always been adequate for my daily needs. I have also been a keen recreational and commuting cyclist in my younger days, and I do not understand the antipathy of modern “pro cycling” planners towards suburban street networks which I regard as far more pleasant to cycle on than anything in a high density location (the utopia of separate rights of way for cyclists being non existent and impossible to retrospectively enable in a dense, built-out city). I am a strong believer in John Forester’s approach to cycling issues.

            Lastly, I want to explain more about the excellent Japanese “articulated density” urban form that you so insightfully identify. This is because integrated enterprises that developed property as well as built and operated transit, were the norm from the 19th century in Japan, and have continued to be so. There are multiple competing networks of transit in every city, which not only compete for ridership per se, they are also competing, via the property they own surrounding their own stops, for tenants and “trip attractors”. Their network of routes and property is like a “network” master planned community. Like suburban master planned communities under US style urban sprawl, they are competing on value-for-money, only they are competing for tenants-cum-riders, in existing built areas, not competing for residents only that will then be driving everywhere.

            The Japanese system is absolutely brilliant and it is a travesty in this era where urban transport, planning and housing is such a hot issue, that no-one understands the difference between the Japanese system and everywhere else where private site rents escalate in response to transit investments and transit-oriented upzoning, sabotaging the intentions of the investments and the planning. As a Georgist, you will get this – and land taxes are actually not a “solution”; they might suppress land rents to a predictable extent, but the patterns of land rent and the locational incentives involved would still exist. “Targeted” land taxes would have other unintended consequences. But land taxes in conjunction with road pricing and proper pricing of infrastructure are the right approach to the problems that urban planning claims to be trying to solve. Anthony Downs uses a simple analogy: you have a picture on the living-room wall that looks like it is in the wrong place: do you move the wall (urban planning) or the picture? (direct fiscal incentives to behaviour).

  5. Bjorn April 10, 2018 at 11:37 am #

    Are there any published transportation planning documents that talk in terms of distributions of probability? i.e. “This new street is projected to have between 15,000 and 25,000 vehicles daily (90% confidence) in 2040.”

    • Sailor Boy April 11, 2018 at 5:04 pm #

      All prediction is bulls**t. We get the traffic we build for.

    • Jarrett Walker April 12, 2018 at 3:43 pm #

      The closest thing to this that I’ve seen is scenario planning, where different scenarios with different assumptions are used.

  6. Ben G April 14, 2018 at 3:45 am #

    My only beef is the whole “bort” abstraction. Why not just say “people” — I don’t think inventing a fantasy planet and inhabitants, while then explaining their existence in completely human terms (more or less), brings any benefit whatsoever… only potential confusion.

    Looking forward to the book!

    • Jarrett Walker April 26, 2018 at 9:42 am #

      As soon as I talk about humans, readers start thinking about psychology, and my point is that geometry and some axioms of biology give you all you need. Welcome further suggestions. I know the bort approach is polarizing.

  7. Eric Goodman April 16, 2018 at 10:11 pm #

    Congestion is itself a definition of urban or city in the way Jarrett uses it. If we think of all kinds of congestion, not just the cars on a road variety.

    Cities are crowded restaurants, lines at attractions or the hottest nightclub, busloads of kids and teachers on museum fieldtrips, shuffling to squeeze in the office tower elevator.They are hard to get tickets, no vacancy, penthouse apartment with a view on top of 200 without, convention center hotels, stadiums, stock exchanges. They are ports and parliaments, great stations, markets full of goods where everyone and everything connects. Cities are crowded sidewalks, and crowded parks and crowded beaches, and crowded halls and crowded streets. To anyone seeking relief of congestion, of people in the way, the only solution is to be somewhere more rural on the spectrum from urban to rural, even if that is only a small park with the illusion of pastoral environs. I grant that there may be congestion around a constraint or concentration in a rural area, and that exception is a proof for the definition.

    The more rural an area, the more there is relief from congestion, not because it has been solved, but because they are definitionally opposites. So, trying to solve congestion leads us to try and make the city rural. Answering that question makes us build suburbs. Congestion is a land use question. How intensely do we want to use the land. Do we want a 60k seat stadium or 80k seat? Apartment complexes or high rise towers? acres of parking or wide sidewalks and interesting plazas? Shopping mall or mixed use? Isolated or integrated campus? A city or a town or a sprawlville zombie?

    The transportation question is not how to solve congestion or get one person around or over or under it. The nugget is how to move masses of people through congested places and spaces, in a way that enhances and embraces, rather than destroys or degrades the more urban mix and breadth of activity there and the experience of it all.

    Cities are places where the value and desirability of land for a wide variety of other uses makes it more costly and less beneficial to dedicate it to transport, even as high levels of activity generate high demand for movement warranting more supply. That is the essential tension that demands a more efficient use of right of way and vehicle capacity, at whatever size and speed and stop distance ratios happen to fit your town form, culture and choices.

    Every place has to determine how they allocate land between dispersed and congested forms. For a century now that conversation has been deterministic. The math around how to move has been weaponized to advance a car centered form leaving no refuge for the person scale city. If we use our models, not to predict a foregone outcome, but to test different inputs, might we find actions more in line with our values and our envisioned future? Perhaps actions that maximize freedom?

    Technological improvements don’t need to be oversold as magical elixirs. We should be excited about some of the possibilities for better resource and time allocation with improved location awareness and automation. Models can be more enlightening when used to test scenarios rather than producing predictions that demand fealty. They could be used simply as tools in the design kit of a community as they create a desired future. Humans seem to be better at design than prediction, no?

  8. Mark Bahner April 21, 2018 at 9:06 am #

    I don’t see the future of transportation as being particularly hard to predict. In the U.S. and most other developed countries (at least…I’ve given less thought to developing countries, and know less about them) in less than 30 years, we will have predominantly:

    1) Computer-driven vehicles,
    2) Transportation as a service (TAAS), and
    3) Electric vehicles.

    I see the synergies between those three aspects (e.g. computer-driven promotes TAAS, and TAAS promotes electric vehicles) as being very strong, so locking in the first aspect (computer-driven vehicles) strongly promotes the other two.

  9. Peter R November 13, 2019 at 6:45 pm #

    Looks good, but there’s one branch of thought that I think could be better: as written, the article assumes the large size and armor of bortcars as a given.

    That said, I think there’s a strong physics and biology based case for why in order to be safe, vehicles that move much faster than walking must occupy much more space than their occupants. To put it in the bortworld context:

    1. Borts in motion might cross paths and end up trying to occupy the same space at the same time. This is a collision.
    2. Through evolutionary adaptation, borts most probably have developed tolerance for collisions near hopping, drifting, or slithering speed, but not collisions that are significantly faster. This is because organisms adapt to immediate environmental pressures but not to non-pressures.
    3. Bort-sized vehicles might exist that enable borts to travel faster than they can hop, drift, or slither. But vehicles that don’t protect borts from collisions can’t enable a major increase in travel speed without having those collisions become a major safety issue.
    4. Providing the necessary armor for protecting borts from faster speed collisions requires space much larger than the bort. The primary function of armor for collision protection from similar-mass vehicles is to reduce collision accelerations to tolerable magnitude, and reducing acceleration means more space will be needed to effect the necessary velocity change. Human cars have crumple zones (which use non-occupied space for deceleration), airbags (which provide deceleration through cabin space), and seat belts (which keep occupants away from the sides of the car), and it is likely that bortcars would feature similar inventions.
    5. Similarly, fast-moving bortcars need to be operated with significant buffers between them. The space is needed so that bortcars can slow down in the event of an imminent collision, either reducing its severity or preventing it altogether. The buffer space required increases with speed.
    6. In addition to occupying more space, fast-moving bortcars don’t mix well with borts hopping, drifting, slithering, or using bortcycles because the latter would then be vulnerable to fast collisions. This problem is easily solved for cases in 3D or beyond, but it’s not possible to have multiple networks in 2D in the same region of space with paths that don’t cross. In human cities, going into 3D is possible but expensive and space-consuming.

    Now, by assuming that a bortcar must be fast and safe, we know that it must be large, that it must be operated with plenty of buffer space, and that it must be kept away from borts and bortcycles.

    And now I’m curious about your next book. I’m looking forward to learning more!