Should Transit Agencies Panic?

I’m in Atlantic Citylab today, responding to the widespread notion that transit agencies face some kind of existential crisis.  So far, the two most quoted bits are:

“A general message of ‘technology changes everything’ has become one of the most powerful arguments for letting fixed transit wither, even though this means worse traffic and higher transportation costs for cost-sensitive people.”

“Technology is a tool, not a goal. The job of local government—including transit agencies—is to serve the goals and aspirations of citizens. That, not fear of technological change, should be the foundation of their decisions.”

Hope you’ll read the whole thing.

Over the weekend, I also attempted a longer thinkpiece on why transit agencies can be frustrating to deal with, and how transit advocates can work with them more constructively.

 

 

 

83 Responses to Should Transit Agencies Panic?

  1. albaby January 22, 2018 at 7:30 am #

    Interesting piece, as always. However, as has been frequently noted here in the comments, your argument against public transit being displaced by technology still overlooks two major points that *should* have most transit agencies panicking:

    1) Although you again caveat the spatial geometry argument as applying to “dense cities where space is scarce, and where huge numbers of people are moving in generally the same direction,” most U.S. cities aren’t like that. Austin, which you mentioned in the article, has a 4% transit mode share. New Orleans (also mentioned) has a better, but still low, 8% transit mode share. Neither has enough *current* transit users for a shift away from transit to bump up against spatial limitations.

    2) Arguing for a shift away from transit to private passenger vehicles does *not* require arguing that “everyone who can’t afford those fares should be “left behind.” You can still subsidize low-income folks’ transportation – you would just do it by subsidizing their Lyft/Uber rides, rather than subsidizing transit. This doesn’t make sense today (except for maybe para-transit), because even with the artificial Lyft/Uber fares there’s still too big a cost differential. But in a world with affordable autonomous vehicles (if such were to come to pass), that could be the more efficient outcome for many transit agencies; after all, if you subsidize individual riders rather than a *system*, you can also avoid the expense of subsidizing people who *could* afford their own transportation.

    So while the NYC MTA has no reason to worry, transit agencies in most other U.S cities should be deeply concerned that their electeds might switch from subsidizing buses to subsidizing Lyft/Uber passes.

    • Novacek January 22, 2018 at 9:20 am #

      >>Neither has enough *current* transit users for a shift away from transit to bump up against spatial limitations.

      Of course they do.

      Current Austin resident here.

      The statement that Austin only “has a 4% transit mode share” is so broad as to be functionally meaningless. The denominator of that statement includes residents over 15 miles from downtown (as the crow flies, 20 miles by road).

      While other, close in urban census tracts (like West Campus) may have a transit mode share nearly an order of magnitude higher.

      Get rid of transit in those areas of Austin, and see how the spatial limitations of (for instance) Guadalupe St are hit even more than they are now.

      • albaby January 22, 2018 at 10:37 am #

        “While other, close in urban census tracts (like West Campus) may have a transit mode share nearly an order of magnitude higher.”

        They *might* – but do they? I kind of doubt it. There’s a million people who live in the City of Austin proper, and the boundaries of the city don’t go much past ten miles (with low population density out one the fringes). Meanwhile, average weekday boardings for CapMetro run about 100K (of which some 20% are the university shuttle) – so figure something less than 50K users, considering that’s boardings (not trips) and some users come from outside the city proper? That’s going to be less than 5% of the population *within the City*, which means you’re not likely to get 40% mode share in too many places.

        And even granting that there are likely isolated pockets of density (both residential and job/destination density) in even the most auto-centric cities, is that going to provide enough ridership to keep a transit system functional if autonomous vehicles take out the rest of the network? Given how low the usage is from a metro/regional level, Just taking a look at Austin’s system map:

        https://www.capmetro.org/uploadedfiles/Capmetroorg/Schedules_and_Maps/System_Map.pdf

        …there’s a lot of the system that’s outside of the dense, urban core. If you were to cut back the system from all of the area that has sufficient space to handle a modest uptick in traffic, is there enough left to be viable? And if not, I think you’re more likely to see the physical built environment change to adapt to the new transportation technology rather than the other way around.

        • Novacek January 22, 2018 at 11:28 am #

          >>of which some 20% are the university shuttle

          Exactly. So 20k trips by 50k students right there. Add in UT student boardings onto the other routes and students (as a population overall) are probably running _at least_ 10% mode share.

          So in the dense West Campus area, being at 20%, 30% or higher seems very possible (I looked, but I couldn’t find updated numbers that geographically specific).

          >>And even granting that there are likely isolated pockets of density (both residential and job/destination density) in even the most auto-centric cities, is that going to provide enough ridership to keep a transit system functional if autonomous vehicles take out the rest of the network?

          That wasn’t the claim of yours I was addressing (yet). You were claiming that nowhere in Austin had spatial constraints that were helped by transit. That’s obviously false.

          >>And if not, I think you’re more likely to see the physical built environment change to adapt to the new transportation technology rather than the other way around

          How? By leveling half of downtown, half of campus, half of west campus, to free up space for new lanes for all these new SOVs?

          • albaby January 22, 2018 at 12:19 pm #

            I don’t claim that nowhere in Austin has spatial constraints that are *helped* by transit – just that the city has such a low use of transit already that it is not *impossible* for all (or nearly all) of it to be cannibalized by autonomous vehicles (if we ever find ourselves in a world where that tech is affordable). Washington’s experience with the marked drop in Metro ridership during their safety/maintenance problem’s is instructive – although ridership dropped *precipitously* for a while, and traffic got somewhat worse, the overall transportation network did *not* stop functioning.

            You don’t need to level “half” of anything to free up space – again, the point is that the current transit system *isn’t taking that many cars off the road*. Rather, you would expect to see more physical changes that optimize the type of pick-up and drop-off nature of AV’s – bulbouts and bays, reducing the number of parking spaces in peak areas, and the like. So UT might lose some of their surface lots and build some more residence halls, for example.

          • Mike January 23, 2018 at 1:22 pm #

            “Washington’s experience with the marked drop in Metro ridership during their safety/maintenance problem’s is instructive”

            Are you surprised that people stop taking the subway when it becomes unreliable, overcrowded, and liable to kill people? Look at the opposite situation: when a city like DC makes a 10-year investment in its subway, fixes outstanding maintenance issues, and expands capacity or frequency: ridership goes up for some reason. The same applies to bus networks, at least for cities down to the size of Portland. (Maybe not Tacoma, due to small size and low density throughout; the jury is still out how much potential ridership Pierce Transit could get if it had more service.)

        • Novacek January 22, 2018 at 11:45 am #

          >>is that going to provide enough ridership to keep a transit system functional if autonomous vehicles take out the rest of the network?

          Okay, so let’s consider this claim now.

          Why would a transit agency (planned and run by professionals) shrink the coverage of the network down to non-viability?

          It’s not like small ridership losses on the fringes of the network would lead to a ridership/funding death spiral. That fringe rider is already being subsidized at a higher rate than a core system rider, with a fractional FRR.

          >> If you were to cut back the system from all of the area that has sufficient space to handle a modest uptick in traffic

          I’m guessing you’re vastly over-estimating where this could occur. Most of the arterials in Austin are at the point of functional failure as far as traffic and LOS are concerned.

          • albaby January 22, 2018 at 12:28 pm #

            The transit agency might not have a choice in the matter (again, this all presupposes a world where autonomous vehicle tech is affordable enough to be economically viable).

            Some riders will just leave on their own – that’s already happening even with more expensive Lyft/Uber without AV’s. And if an AV fleet were able to provide transportation services *more cheaply* than the transit agency over most of its routes, the politicians will face enormous pressure to abandon the fixed route model in favor of using AV’s itself: they’ll be cheaper to run, *and* provide quicker service to the users (who now get the benefit of point-to-point transportation). More likely, though, the government body will find it more convenient and efficient to just subsidize the *user* to use market-provided transportation rather than run an AV fleet itself, as we do with food stamps and (some) housing for low-income folks.

          • Novacek January 22, 2018 at 1:05 pm #

            >>Washington’s experience with the marked drop in Metro ridership during their safety/maintenance problem’s is instructive – although ridership dropped *precipitously* for a while, and traffic got somewhat worse, the overall transportation network did *not* stop functioning.

            So a 10% drop in ridership* didn’t totally destroy the transportation system, great.

            So you think that’s evidence that completely dropping transit wouldn’t increase traffic significantly?

            *That’s for WMATA, several of the other area systems had smaller drops or even ridership increases.

            >>the point is that the current transit system *isn’t taking that many cars off the road*.

            It’s taking a significant fraction from the chokepoint arterials in the system.

            Again, let’d look at Guadalupe in Austin. Traffic counts are ~30k vehicles/day. Ridership on just two of the bus routes that travel that corridor (1/801) is 12k /day. Move all of that ridership into SOVs (or 1.2 occupant vehicles on average) is a huge increase in traffic.

          • Novacek January 22, 2018 at 1:14 pm #

            >>More likely, though, the government body will find it more convenient and efficient to just subsidize the *user* to use market-provided transportation rather than run an AV fleet itself, as we do with food stamps and (some) housing for low-income folks.

            >>you can also avoid the expense of subsidizing people who *could* afford their own transportation.

            You’re showing a massive misunderstanding of the basic purpose of transit here.

            Is transit social welfare? Sometimes.

            It’s also economic development.

            Just handing out uber passes to low income riders would, for many cities, destroy their economic viability. Their transportation systems would grind to a halt.

          • albaby January 22, 2018 at 1:19 pm #

            The question isn’t whether it will increase traffic – it’s whether it is possible for the city to still function. Jarrett’s spatial geometry argument isn’t predicated on the idea that AV’s can’t displace transit because they will make traffic worse – we’ve already seen in New York (and other areas) that it’s perfectly possible for people to switch to ride-hailing even though it makes traffic worse. His argument is that it is literally not possible.

            So yes – traffic may get worse on Guadalupe. Perhaps not 10K cars per day worse, since some of the folks that are traversing Guadalupe by bus today may not actually need to go down that arterial (but are routed there because it’s works best for the agency).

            BTW, that figure seems impossibly high. Average weekday ridership on the MetroRapid portion of the system (of which the 801 is only part, and of which only part runs on Guadalupe) is only about 10.5K daily boardings. It can’t possibly carry 12K riders per day. Can you provide a link to your numbers?

            https://www.capmetro.org/uploadedFiles/New2016/About_Capital_Metro/Data_and_Statistics/2017-10-Monthly-Ridership-Web-Report.pdf

          • Novacek January 22, 2018 at 2:51 pm #

            Like I said, that’s the ridership of two of the routes that travel on Guadalupe, not the subset of the boardings that happen to occur on Guadalupe itself. About 5300 for the 1 and 7400 for the 801.

            https://docs.google.com/spreadsheets/d/1AyqgGqDCAipHn1WyVnzbeZfuSA9g-6hj5Vk6RwF5h_Y

            If you want boardings on Guadalupe stops specifically (but of all routes, not just those two) there’s also data (but harder to read).

            https://docs.google.com/spreadsheets/d/1f5zu7SFquRfz5v9net3Q5GG5UhdSZwmsLv3I8epYkVw

            Or slightly older, from 2014.

            https://www.bizjournals.com/austin/news/2014/05/20/stop-by-stop-mapping-transit-ridership-in-austin.html

            I count ~7000 boardings/exits (each of which would be a separate SOV trip on Guadalupe, one coming and one going) on just the top 4 Guadalupe stops.

    • Jeff Wegerson January 22, 2018 at 9:40 am #

      Your arguments favoring ride-hail subsidies over transit system subsidies fall within Jarrett’s ridership/coverage dichotomy. Intelligent agencies will, and already do, operate within that dynamic.All that changes for such agencies is the technology used within the subsidized coverage budget. If the same dollars can cover more area with ride-hailing subsidies, of course, they go there. And vice versa.

      The technologies of the future are already accounted for now, no matter the densities of ridership nor the technologies used to serve them.

      • albaby January 22, 2018 at 12:41 pm #

        But the changes wrought by the technology might be enormous, well beyond what’s accounted for now. We don’t know, of course – autonomy may always be too expensive to be practical. But if autonomy ever gets to the point where it is cheap, it could not only swallow up the “coverage” part of the ridership/coverage dichotomy, but eat into the “ridership” portion – particularly for buses. After all, there are some advantages to a personal AV compared to riding a bus (privacy, door-to-door delivery, perhaps a quicker travel time) that transit agencies may have trouble competing against even on their “ridership” routes.

        Again, it’s all up in the air right now – but I think that it would be a mistake to assume that there is no possible existential threat to fixed-route transit systems. A cheap AV could decimate transit in a lot of urban areas.

        • Quite Pedestrian January 26, 2018 at 3:59 am #

          Dense areas will have high-frequency autonomous larger vehicles that can support set stops and schedules, also known as fixed-route buses. Sprawling areas will have on-demand autonomous smaller vehicles, also known as ride-hail. Both are forms of Transit. And if TNCs want to operate the latter, then public transit properties can more easily focus on the former, where productivity is much more cost-effective. That’s not a threat. That’s a relief.

      • Mike January 23, 2018 at 1:39 pm #

        *if* more share dramatically shifts and it doesn’t create massively more congestion in some areas, then we can look at scaling down the fixed-transit footprint or modifying the agency’s mission. It’s the same issue as building new P&Rs: suburbanites demand them now but it’s possible that in a few decades cars will become less popular (due to climate, convenience, lifestyle, fuel prices, dwindling paychecks, etc) and the P&Rs will become mostly empty. At that point we can convert them into TOD housing or something. And some forward-thinking agencies are building lots and garages with convertability in mind. The same principle applies to robotaxis and robocars. We can plan for possible conversion someday, but don’t reduce transit now or avoid investment in trunk lines (rail, BRT, etc) because you think they won’t be needed. That’s depending on an uncertain future, *unproven technology*. They *might* not be needed in the outer suburbs and exurbs, so if you’re concerned about that then start with low-cost BRT. But *don’t* skimp on frequency or transit lanes. But in cities and inner-ring suburbs, go full speed ahead with rail and BRT plans and don’t downscale, because you’ll probably need them. Most American cities have less transit infrastructure than the need (i.e., less than Canadian or European cities), so these investments are merely catching up, and will be needed even in a robotaxi future. Even if robocars *can* transport everyone without causing undue congestion, there are still advantages to having fixed-route transit in medium-density areas, especially if it’s 5-15 minute frequent.

    • JHBW January 23, 2018 at 10:18 pm #

      What I took away from the piece, and what I believe, is that the cheaper cost of operating rideshare AVs would merely recoup the expense of the subsidies being spent currently on keeping Uber prices as they are. This is in accordance with what I think the company is doing, structuring its business model prematurely to prepare for their ultimate planned goal of full automation.They won’t be losing money anymore but to the rider fares would remain unchanged. In that case, why would the potential for low-income riders flipping to rideshare be any greater than today?

      • Mark Bahner January 24, 2018 at 7:36 am #

        “What I took away from the piece, and what I believe, is that the cheaper cost of operating rideshare AVs would merely recoup the expense of the subsidies being spent currently on keeping Uber prices as they are.”

        When I read the word “subsidies,” I think of tax dollars spent on something to keep its apparent cost to users low. Are there any tax dollars being given to Uber?

        • el_slapper January 24, 2018 at 7:56 am #

          Well, their insane capitalization is what works as subsidies. Kind of private subsidies.

      • albaby January 24, 2018 at 8:33 am #

        JHBW, no one yet knows whether AV’s will even be cheap enough to be commercially viable, let alone the impact on fare pricing.

        But if AV’s become inexpensive, then it is certainly possible for AV fares to fall below current ride-hailing fares. Those fares are not currently high enough to cover costs of operation (hence the references to “subsidies” in Uber/Lyft discussions, as capital investors are covering those operating losses through their equity). However, that’s in no small part because it’s expensive to have a personal driver. If you can remove the driver for a low enough cost, then ride-hailing can be cheaper than it is today while breaking even.

        Of course, that “low enough cost” is still very speculative. Again, we don’t know how AV tech will price out. But that’s the scenario many transit agencies should be concerned about.

    • RossB January 27, 2018 at 6:56 pm #

      >> You can still subsidize low-income folks’ transportation – you would just do it by subsidizing their Lyft/Uber rides, rather than subsidizing transit. T

      How would that be cheaper? How would driverless taxicab service be cheaper than driverless bus service?

      That is the part of the argument that just doesn’t add up. Right now, just about every transit agency runs fixed route service, as well as on demand van pool service. They run the former because it is more efficient with moderate demand. As long as the buses are reasonably full, it is the cheapest way to move the most people. They run the latter when demand is very low (no sense running empty buses all day).

      Driverless vehicles don’t change the dynamic. On demand service is still more expensive. Consider a bus that carries fifty people on its half hour trip. It is by no means a high efficiency bus. People are getting on and off at every stop, so if you only pick up fifty people, it is really nothing special. Yet it is still way more efficient than if you ran fifty cabs to pick up those people. Fifty cabs means a lot more cars, a lot more maintenance, and a lot more gasoline for all those trips back and forth. You are better off running the bus.

      Now consider how each system scales. In both cases (cabs and transit) you’ve eliminated the most expensive part of the equation. But that’s it for taxi-cabs. You have more cabs on the street, but that doesn’t mean that service is any better. If all the cabs are downtown at 5:00 PM, or at the arena after the big ball game, or the airport at noon, you are out of luck in your neighborhood. The fact that you have thousands and thousands of extra cabs roaming the streets doesn’t mean it is easier to catch one (as any New Yorker will attest). Meanwhile, what if those taxi-cabs clog the streets? What if they are ubiquitous — what if they are so common, and so ready to pick up a rider that they pile up, and traffic moves at a standstill (again, New Yorkers are familiar with this phenomenon)? Is that actually any better? Of course not.

      Now consider what happens if you have a lot more buses on the street. You have much better headways (five minutes should do). You have a grid, instead of cherry picking the most popular trips. That means anywhere to anywhere service with minimal delay. As you ramp up, the improvement in taxi-cab service is minimal (a little less waiting at best) while the improvement in transit is dramatic (a lot less waiting along with a faster, more useful set of routes).

      Now take it a step further, and consider express bus service. This makes sense, as long as other people are also making the same trip. Otherwise, you will simply have to wait while the bus picks people up along the way. If you want express, stop to stop service, and you are the only one who wants to make that trip, then by definition you are asking for luxury service, and you better pay your own way (and in my opinion, a little on top of that).

      Meanwhile, what about traffic? Has congestion just magically gone away? Of course not. In most cities, it is still an issue, which means that once again, you want to consider things like throughput. This means that transit lanes make the most sense, which means that in many cases, even a very expensive, point to point cab ride will be slower than the fast bus. Yes, the bus stops to pick up and drop off people along the way, but it least it doesn’t wait for every car.

      It all comes down to scalability. Transit scales, cabs don’t.

      You can’t ignore the hardware. It cracks me up when writers start comparing automated transit systems to computer systems, and ignore this fundamental difference. The software revolution would not have happened without the hardware improvements. The PC revolution came about because CPUs became cheap. The Internet would still be used by a small portion of the population if not for relatively cheap wiring built on top of an even cheaper phone network. Streaming videos, search engines and online encyclopedias would not exist if not for for dirt cheap hard drives. The hardware improvements went hand in hand with the software improvements.

      But no such improvement exists when it comes to transportation. This reality still exists, which is why bus service will always make sense: https://humantransit.org/2012/09/the-photo-that-explains-almost-everything.html.

      • albaby January 29, 2018 at 6:50 am #

        Sure, a bus that carries fifty people for its entire half-hour trip might be somewhat more expensive to replace with individual cars (though not much – buses aren’t as fuel efficient, and buses are much more expensive, than single-occupancy vehicles). But there are a *lot* of cities with a lot of times that buses *aren’t* carrying fifty people. They’re carrying a lot fewer. The *typical* cost to run the system across all the times when buses are both full and more empty may not be any lower than AV’s.

        Similarly, while in “most cities” congestion is an issue, in “most cities” transit makes up a tiny fraction of even commuter mode share (let alone total mode share). So while an exodus from transit to AV ride-hailing would likely make congestion worse, it could only be a *bit* worse. After all, DC’s transit breakdowns showed that even a non-trivial loss of ridership in a transit-heavy system can sometimes be accommodated without much of a problem. And if AV’s are able to improve traffic movements among vehicles, the net might be relatively tiny.

        So while transit agencies in major transit-using metros (like NYC’s MTA) have nothing to worry about, transit agencies in other metros with relatively modest usage (like Austin or New Orleans, mention in the original article) might face a serious problem if AV’s come along at an affordable price point.

        • RossB January 29, 2018 at 5:44 pm #

          >> The *typical* cost to run the system across all the times when buses are both full and more empty may not be any lower than AV’s.

          I still don’t buy that. If you want to move a city full of people, it is much cheaper if people are willing to share a ride, and are willing to make a transfer. Even when ridership is relatively low, you are better off building a network, as opposed to having a city full of vehicles, each making point to point trips. Point to point transportation is amazingly inefficient.

          As Walker has pointed out, we are quick to admire the benefits of automated taxi-cabs, but slow to understand the benefits of automated transit service. Places that have low ridership also have poor service and Austin is no exception. They have only six bus routes with headways 15 minutes or better. I get it — I understand why the agency doesn’t bother running buses more often (it is expensive) but ridership suffers because of it. It is much faster to drive, or take a cab.

          But as you scale up, everything changes. It really wouldn’t take that many vehicles to provide a typical city with five minute headways in a grid. Let me try some math here.

          I’ll use Austin, which is a relatively big city from a physical standpoint. It is about 300 square miles. With buses running every quarter mile in a grid, that works out to 2,400 bus miles (8 linear miles per square mile). All these buses are basically going back and forth, so you need to double that. To get decent headways, you can run the buses about a mile apart (meaning two buses heading the same direction on the same street are a mile apart). If the buses are going at 12 MPH, that works out to a bus every five minutes. Since traffic is apparently nonexistent in Austin, that should be easy.

          OK, so that works out 4,800 buses all together. This is service that would be the envy of every city in North America (if not the world). This is very frequent service (five minutes, all day, all night) along with good stop spacing and short walks to the nearest bus line.

          That is a lot of buses — 4,800. Can you beat that with a point to point system? I seriously doubt it. A point to point system has to provide one car per ride for peak ridership. You also need extra cabs, to minimize the wait time. If the entire fleet is out servicing rides, then it will take a while before one frees up in my neighborhood.

          In a city of a million people, with roughly 1.5 million cars, 5,000 taxi-cabs just aren’t going to cut it. 5,000 rides at peak time is probably less than what the current system provides (ridership is close to 100,000 a day), let alone what a better system would provide. The current transit system is only a tiny fraction of even commuter mode share — is that your goal? If so, how will you possibly keep the numbers down. If not, then you need orders of magnitude more cabs.

          Once you eliminate the cost of driving, the most expensive part of the system becomes the vehicles. The more vehicles, the more expensive. A transit system like Austin’s might consist of mostly vans (since it is a fairly low density city) and in many corridors, it might consist of station wagons. But unless you are looking at tumble-weeds rolling down the street all day, the cheapest high quality system will require sharing the ride, and making a transfer. In other words, a transit system. Providing something that avoids transfers, and avoids sharing would be lovely, but it is a luxury system, whether it is provided by private agencies or public ones.

          • albaby January 30, 2018 at 7:43 am #

            It’s not a “luxury” system – it’s merely one that’s more efficient for users. Point to point transportation is quicker and more convenient than one that relies on transfers and mode-switching, in addition to providing benefits like shelter from the elements and more convenient cargo transportation. You’d certainly need more than 5K AV’s to make it work in a city like Austin – but since buses cost 15x what private passenger cars cost (and are much less fuel-efficient) you’re not going to have much additional expense to get a fairly sizable fleet. If you’re running vans and station wagons, you have a lower cost to the transit vehicles – but you also need far fewer AV’s to provide the same coverage, and perhaps not even more than the number of vans needed for fixed-route sharing.

            Your service might “be the envy of North America” only if someone else is paying the operating costs, because you’re going to be moving a fair amount of empty metal around through most of your service. Austin averages only about 25 passengers per revenue hour, a bit less than the national average of just over 30 – but either way, buses just aren’t running at 140% of seating capacity 24/7 over your whole network. Which means that if AV tech is not too expensive, an AV fleet could price close to or cheaper than some transit system costs…which would basically erode those systems.

            In cities that are so dense that physical constrains are highly relevant (like in frequently-referenced Manhattan, but other cities with high current transit mode share), transit will of course persist – there is literally no alternative. In other cities that are today almost entirely auto-centric (Houston, Oklahoma City, Phoenix, Orlando, etc.), you might see transit almost completely absorbed into point-to-point AV systems. So while some transit agencies need not panic, many others should be at least somewhat concerned about what an inexpensive AV technology could do to their networks.

    • RossB January 30, 2018 at 8:29 am #

      One of the things that people haven’t talked about much is how private companies operate. If you own a small bakery and run out of muffins every day, you are doing something right. You could argue that you are losing customers (since many muffin fans go away disappointed) but if this has been going on for some time, there is nothing to worry about. That reminds me of the Yogi Berra line “No one goes there anymore, it’s too crowded.”

      The point is, that is a fine business model. There are alternatives at that point, but no one would question your decision to just stay put. One option would be to ramp up. But that involves risk. There may also be constraints (do you have enough space to bake all the extra muffins, or handle all of the extra customers, etc.).

      Another option would be to simply charge more.

      A better example is if you run a theater. During weekdays, the theater is half empty, but during the weekends, it is full. Do you expand, so that everyone will continue to love going to your theater on Friday night? That would involve a lot of risk, since it isn’t full on weekdays. Maybe instead you should raise your prices on the weekends, and lower them a bit on weekdays.

      That is exactly what the ride sharing companies have done, and is exactly what they will do when they are automated. Buying a bunch of new cars for rush hour isn’t a great business decision if they sit idle the rest of the day. Capital costs become your biggest expense and while operational costs go down with fewer miles, it is cheaper to drive a small number of cars a lot instead of a bunch of cars a little. Just as the theater has no interest in expanding, neither will the taxi-cab company.

      That simply won’t work if your goal is to replace the transit organization. You can’t tell a city that their transit system won’t be available during rush hour. That means that you are either going to subsidize the taxi-cab company by buying them a lot more cars, or you are going to run automated buses (or vans). The cheapest way to provide this extra service would be a grid, not point to point service. Thus you would still need the transit system during rush hour.

      During times of low demand, you could contract with the automated taxi-cab service, because they have some idle cars. Except you do as well. You’ve bought the vehicles, they ran frequently during rush hour in a grid — why not run them all day? At worse you have a case where the vehicles are too big. You have a bus where a van, or even just a passenger car would do. But running buses in a very efficient manner (in a grid) is still cheaper than running cars in a point to point system, unless ridership is very low. You would to do some fancy mathematical modelling to figure out where the breakeven point is, but it is unlikely that you will find moving to a point to point system (with your vehicles or the taxi-cab companies) make sense, unless it is very late at night (where transit is currently rare).

  2. Mark Bahner January 23, 2018 at 9:15 pm #

    “Once we have wide uptake of full automation—sometime between 2020 and 2100 (or never), depending on who you ask—labor cost goes away and driverless taxi fares theoretically get cheaper. But labor cost of fixed transit also goes away. Labor is the dominant element of transit’s operating cost, so driverless buses and trains could be vastly more abundant.”

    Let’s leave aside trains. If driverless buses exist, why would governments operating them be better than private companies operating them?

    If transportation is provided as a service, then the people who provide that service will know exactly where everyone they serve is coming from, where they’re going to, and when they want to arrive at their destinations. I don’t see the sense in a government, which won’t know those things, owning buses and providing bus service in that situation.

    • el_slapper January 24, 2018 at 8:01 am #

      Because the privatization of buses in the UK has shown the private sector is even worse than state-led companies for this specific task.

      Other example, in France, buses were open to run freely on intercity lines in 2012. Against the SNCF, the national state-owned railway company. Today, the market leader, with the Lion’s market share, is Ouibus. Which belongs to SNCF.

      The assumption that private sector is more efficient than state-led agencies for running buses is not backed up by facts.(OTOH, I would not seat in a state-built car, with a lot of history to back up my fears.)

      • Mark Bahner January 24, 2018 at 11:13 am #

        “The assumption that private sector is more efficient than state-led agencies for running buses is not backed up by facts.”

        When fully autonomous vehicles are common (probably by the early 2020s), do you agree that they will almost certainly be predominantly in the form of transportation-as-a-service? (The reason being that with transportation-as-a-service, vehicles will typically be on the road 8+ hours per day, versus less than 2 hours per day with a privately owned vehicle. So the costs per mile will be dramatically reduced.)

        So if fully autonomous vehicles are available in the form of door-to-door transportation-as-a-service, what is the incentive for either the transportation-as-a-service provider or the customer to switch over to a bus owned by the city? For example, if a person lives even as close as 200 yards from a bus stop, and the person’s work is even as close as 200 yards from a bus stop, why wouldn’t the person choose door-to-door service, especially on any day it was raining, snowing (or simply had snow on the ground), cold, or hot?

        And if the service provider picked the person up at the door, what is the incentive for either the service provider or the customer to be switched to a city bus?

        • Mike January 24, 2018 at 11:46 am #

          *IF* autonomous vehicles can someday distinguish a paper bag on the road from a stone, and drive in snow/ice, and recognize a thousand hard-to-list situations instantly, then they might have a significant mode-share if they can get the cost low enough that working-class people can buy them or hire them every day. But that contradicts the corporate profit incentive, which says to target high-paying people and avoid the higher expense of serving everybody. Or we could make streets autonomous-car only, as some are suggestion for a lane for certain Interstates. But doing that in a city on grid streets would be massive displacement of those without autonomous cars. On the other hand, a dedicated BRT lane is by definition exclusive, so a robobus could serve it if it can do the simpler tasks of staying in the BRT lane and making sure intersections are clear, with a supervisor monitoring the intersections in case she has to intervene to tell the car to stop or go.

        • el_slapper January 25, 2018 at 5:28 am #

          Simple question of available place. Here in Montpellier, right under my office window, I’ve got a 2*2 route plus a tramway lane. Considering the red lights, and other traffick difficulties, each lane carries one car every 6/7 seconds. Counting 1.2 people per car, thats around 800 people per hour per lane.

          Tramway lanes are slightly broader, but each tramway here can carry 200 people(and often do). So we just need 5 per hour for the tramway lane to be as efficient in terms of space use than a car lane. in some parts of the city, there are up to 18 per hour. 3600 people per lane per hour, instead of 800 per car.

          Buses are not as impressive, but they still increase the capacity of the road network. Much more than autonomous cars can do. OTOH, autonomous buses would be even better. Still, count 20 buses per hour(every 3 minutes, still doable), 50 people per bus, and you’re already more productive per available space than the individual car. And you’ve still got some place remaining for a few individual cars.

          Public transit is a multiplier of transport capacity per unit of space.

          Of course, if you live in an endless suburb, local density is too low for any kind of public transit being useful to you. But as soon as you go to the dense centertown, public transit is the only way to transport all people needed for the economic machine to work. I take my car when I’m going to the outskirts of the town. I take the tram for going to the centertown. Cheaper, quicker, easier. It’s a dense centertown, over there.

          • Mark Bahner January 25, 2018 at 6:41 pm #

            “Simple question of available place. Here in Montpellier, right under my office window, I’ve got a 2*2 route plus a tramway lane.”

            It’s interesting that you should mention Montpellier. I see the population of Montpellier is listed at 275,000, with a population density of 4800/km2. I live in Durham, NC (USA), with a population of 228,000, but a density of 810/km2. In other words, Montpellier has a population density that six times the population density of Durham.

            For perspective, we’re just up the road from Raleigh, which has a population of 469,000, but a density of only 1100/km2.

            In other words, Montpellier is much more densely populated than or Durham or Raleigh.

            “Of course, if you live in an endless suburb, local density is too low for any kind of public transit being useful to you.”

            Here is a list of the incorporated and unincorporated places in the U.S. with the highest population density:

            https://en.wikipedia.org/wiki/List_of_United_States_cities_by_population_density

            Only about 80 incorporated or unincorporated places, with a total combined population of only 13 million–that’s out of the U.S. population of over 300 million–have a population density higher than Montpellier. So virtually the whole U.S. is an “endless suburb,” by the standards of Montpellier. 😉

          • Sailor Boy January 29, 2018 at 12:40 pm #

            Average density is an utterly useless metric in this conversation.

            Firstly, it is calculated differently everywhere: Some places don’t count large parks, some places have different cutoffs for what is part of the city, some include temporary visitors.

            Secondly, I would imagine that even in Durham, the city has congested corridors, probably in the higher density areas (downtown).

            You’ve also ignored cost in your mode selection. If I can catch a double decker bus with 90 other people for $1, or a taxi with 3 other people for $4.50 (assuming that the bus costs 5 times more to run), I’m pretty likely to catch a bus. I might catch a taxi to the bus stop, but I’ll still catch the bus.

          • Mike Robinson February 2, 2018 at 9:00 am #

            According to the stats, London has an average population density of only 1,510 per km2… that’s the trouble with averages.

    • Jarrett Walker January 24, 2018 at 11:47 am #

      Pls distinguish between privatization of service planning (a disaster) and privatization of operations under contract to government, which is fine and done in many places.

      • Mark Bahner January 24, 2018 at 2:23 pm #

        “Pls distinguish between privatization of service planning (a disaster) and privatization of operations under contract to government, which is fine and done in many places.”

        What would be the contract? The company would get money from the city? For what? People would pay money to the city (in taxes) and then the city would contract the company to provide transportation? For free, or some reduced price (relative to people simply paying the company the full price at the time the service is rendered)?

        • Sailor Boy January 24, 2018 at 3:05 pm #

          In Auckland we do this.

          The city plans the network and then contracts private companies to supply and drive the buses. The city receives all fare income. The total network subsidy is about 50%, some of the busier routes are very profitable, but are used to cross subsidise their feeder routes.

          Privatisation of route planning has been a disaster everywhere. If company A runs a bus on the hour and at the half hour, another private company B is best to run their services at 1 minute to and 29 minutes past the hour. This is no better for riders, who gain the most use when additional services run at quarter past and quarter to the hour.

        • Jason January 25, 2018 at 3:22 am #

          In Taipei, Taiwan how it works is that the city plans the network and contracts the bus driving out to private operators, who collect fares as well. As part of the contract, city provides some agreed upon subsidies to the operators for money-losing routes. Despite high ridership (some lines are wildly profitable) the system is still losing money (i.e. subsidies are still necessary) because bus fares have stayed the same for over 20 years and are around $0.50 USD per ride.

        • el_slapper January 25, 2018 at 5:18 am #

          That’s how it works in Israel as well, IIRC. The state says there must be public transit on a defined line, and asks private companies how much it would cost to cover the route. Does not work that bad – as long as the route is designed to serve the transport network, and not the interest of the private company.

    • RossB January 28, 2018 at 9:13 am #

      >> Let’s leave aside trains. If driverless buses exist, why would governments operating them be better than private companies operating them?

      Because government owned organizations operate better than private company in a limited environment. Seattle City Light is a publicly owned utility and it is as good, if not better than any private utility. The chief benefit of private companies is that they compete. But it would be crazy to have dozens of different electric utilities all serving the customers in Seattle. Dozens of different cab companies can serve Seattle, but you run into a couple problems. To begin with, there is no reason why any of them have to try to meet the needs of society. They are free to choose the most valuable riders (the ones willing to spend the most money) and ignore others. Congestion is also not an issue. If the cab companies can make the most money by serving downtown, then downtown becomes congested. There is no incentive for any of the cab companies to avoid it (the tragedy of the commons). These problems can be dealt with via regulation, but there is no reason to believe that a highly regulated private company operates any better than a public organization. Seattle City Light has a better record than Comcast Cable, in my book.

      >> If transportation is provided as a service, then the people who provide that service will know exactly where everyone they serve is coming from, where they’re going to, and when they want to arrive at their destinations. I don’t see the sense in a government, which won’t know those things, owning buses and providing bus service in that situation.

      First of all, why couldn’t the government know those things, too?

      But that misses the point. If you have even a decent amount of ridership (streets that aren’t empty) a fixed route, fixed time system is the fastest, most efficient way to move the most people. Consider several factors:

      1) One person in a car is inefficient from a road space perspective (https://humantransit.org/2012/09/the-photo-that-explains-almost-everything.html).

      2) Even without drivers, one person in a car is inefficient from a cost perspective. A bus carrying fifty people is a lot cheaper to operate than fifty cars each carrying someone. You have fewer vehicles to purchase, along with fewer miles driven, which means less maintenance and lower fuel costs.

      3) A fleet of vehicles, each carrying one person, must meet the peak demand by purchasing extra cars. In other words, if a city expects 1,000 riders every afternoon at 5:00, then they have to purchase 1,000 cars, or expect people to wait a long time. With a transit agency, you meet peak demand with larger vehicles. So a route that would be fine with a van at noon requires an articulated bus at 5:00.

      In short, ride sharing is a lot more efficient. So efficient that it behooves an area to prioritize transit passage. This is common throughout this (and other countries). Agencies have special lanes that can only be used by buses, because buses can move more riders through that corridor.

      Of course, you could have on demand van pooling. However, for reasons that are again not obvious, an on demand van pool system gravitates towards a fixed route, fixed time system. As you start scaling up — as demand starts increasing — you are better off running buses frequently on a grid.

      I’ve used this explanation before (I should probably just write it down somewhere and reference it) but consider an airport van service. This is exactly the service you describe and it exists now. They know where I am going and when I want to leave. For simplicity’s sake, consider the trip from the airport. Now imagine you are the person in charge of the routes, or the person writing the software for it.

      The first question is when do the vans leave? If you leave as soon as someone arrives, then you are nothing more than a taxi-cab service, and that is expensive. If you wait until the van is full, then it could be a long wait until that van leaves. Immediately it becomes obvious that if you really want to balance customer service with a cost effective option, then you leave at a set time, even if your vans vary in terms of ridership.

      Then there is the question of routing. It doesn’t make sense to serve people north of the airport and those south with the same van. It makes sense to send a van to one part of town. If there are enough people, it makes sense to send a van to one corridor (e. g. a highway).

      Now the path of the van is beginning to look like a bus route. But unlike a bus route, the van has front door service. If someone lives a couple blocks away from the highway, the van pulls off the main road (which has bus only lanes) then drops off the passenger and gets back on the highway. This is great for that rider, but terrible for everyone else. Those at the end of the line have a very long wait. If you are trying to save money and provide the greatest good for the greatest number, you should simply ask that rider to walk the last couple of blocks.

      Now imagine that someone lives close to the highway, but a mile away, not a couple blocks. They are the only one in the area. You could run a van from the airport to their house, but again, that is like a taxi-cab (expensive). If the van deviates to serve that rider, every other rider suffers. Ideally, you would take advantage of a crossing route, which would provide service for other people making similar connections (although not from the airport).

      Now you’ve basically created a transit network. As ridership grows, it gets better and better. It scales, while a taxi-cab or instant carpool system does not. As a transit network gets more riders, you can increase headways. Increased headways greatly increases the value of the system, which in turn leads to more riders (a virtuous cycle). With better headways, the transfer penalty is minimized, which means that bus service can operate in a grid, which provides the most efficient service for anywhere to anywhere trips.

      On demand service has a role to play, but it is primarily for luxury service or for very low demand areas or times.

      • asdf2 February 11, 2018 at 4:00 pm #

        If you’ve ever carpooled places around with friends, one thing that quickly becomes apparent is that the extra time to pick up and drop passengers adds fast. At times, just picking up a mere 3-4 people (with a mile’ish deviation for each person) was enough to make the carpool slower than the regular old bus. Granted, when it’s a ride among friends, you’re probably too busy socializing to notice the extra time.

        But, if it’s a shared-ride car service, like UberPool, where the passengers are strangers, all of sudden, the extra time to loop around the block to drive to someone’s front door becomes pure overhead from your perspective, justified only by the dollars and cents you save over simply ordering a private ride. It doesn’t take very many times of paying 10x the bus fare for a trip that’s no faster than just riding the bus(*) before you start to feel cheated by the whole thing, and just go back to riding the bus.

        (*) If you’re lucky, there will be no other passengers, and you end up with what’s essentially a private ride for a shared-ride price, subsidized by the ride-sharing company’s investors. Obviously, a system that depends on this happening more than a tiny percentage of the time to keep customers happy is not going to scale.

  3. Mark Bahner January 24, 2018 at 3:29 pm #

    “The city plans the network and then contracts private companies to supply and drive the buses. The city receives all fare income.”

    With autonomous vehicles providing transportation-as-a-service, there is no “network.” People will simply tell their smartphone that they need to go from X location to Y location, and need to be at their destination at Z time.

    So a city the size of Auckland might have 100,000+ requests for transportation from all different locations to travel to other different locations at any particular time. How does it make sense to have buses of a fixed size traveling fixed routes at fixed times in that situation?

    Also, in that situation, how does it make sense for the city to get the fare income for each of those 100,000+ rides, rather than the person taking the ride simply paying the company that provided that particular ride?

    • Jason January 25, 2018 at 3:29 am #

      > How does it make sense to have buses of a fixed size traveling fixed routes at fixed times in that situation?

      Because it’s more efficient (read: cheaper), and because road space constraints preclude door to door travel for everyone (you will need a lot more vehicles than can be fit into the roads). Autonomous vehicles do not solve the fundamental problem that we can’t fit as many cars as we wish on the roads today.

      • RossB January 28, 2018 at 10:25 am #

        >> So a city the size of Auckland might have 100,000+ requests for transportation from all different locations to travel to other different locations at any particular time. How does it make sense to have buses of a fixed size traveling fixed routes at fixed times in that situation?

        Because it the fastest and cheapest way to move them. Consider two systems, yours and mine. Mine works like this:

        Buses run on a grid with five minute headways. Stop spacing on most routes are as Walker described (https://humantransit.org/2010/11/san-francisco-a-rational-stop-spacing-plan.html). But there are also a handful of express buses that also operate on five minute headways along the more popular routes. Since buses take up a lot less space — since throughput on a bus is much better — the city has bus lanes on most of the crowded corridors.

        Now compare the two systems:

        Performance — For most trips, my system wins. The savings made by avoiding congestion make up for the extra stops. The extra stops aren’t really an issue for short trips, and as I’ve described, the more common long trips have express service. It is only the rare trip (a long distance trip where express service isn’t warranted) where your system wins.

        Cost — My system wins easily. You have a lot less vehicles and a lot less miles driven. You might have a van driving a lonely road every five minutes, only rarely picking up someone, but running big vehicles on the major corridors (in their own lanes) more than makes up for that. With your system, you would still need to have a car in that lonely neighborhood, just in case someone wanted a ride. You would probably need several, actually, unless you are willing to just ask people to wait a long time if they decide they want to go somewhere.

        Scaling — Let’s assume both systems become very popular. I don’t see how that makes your system better. You have more cars on the road, but you still have to deal with peak demand by buying more vehicles, which means more cost and more congestion. With my system, you simply increase the size of the vehicles (vans to buses, buses to trains) or run them every two minutes instead of five. You also run more express routes.

        Convenience — Your system promises door to door service, mine does not. On the other hand, mine is available if I forget to charge my phone. I can wander to the nearest major street, look for a bus stop, and know that a bus will take me down the road.

        Your system is an edge system. It makes sense when ridership is low, for those with mobility challenges, and for those willing to pay a premium price for door to door service. In the first two cases, it is reasonable to see that as part of the transit agency. It also makes sense as a complementary system. The city itself might have five minute service, but suburbs have on demand service. Someone would take a bus (or two) then pick up a cab at the outskirts of town (as one picks up a cab in a small town by the train station).

        • Mark Bahner January 29, 2018 at 8:22 am #

          “Because it the fastest and cheapest way to move them. Consider two systems, yours and mine.”

          That’s a bit difficult, since mine won’t be fully built for another ~20-25 years.

          “Cost — My system wins easily. You have a lot less vehicles and a lot less miles driven. You might have a van driving a lonely road every five minutes, only rarely picking up someone, but running big vehicles on the major corridors (in their own lanes) more than makes up for that. With your system, you would still need to have a car in that lonely neighborhood, just in case someone wanted a ride. You would probably need several, actually, unless you are willing to just ask people to wait a long time if they decide they want to go somewhere.”

          No, every study *I* know of concludes that autonomous vehicles operating in transportation-as-a-service mode will be far, far less expensive per mile than the total transportation system (i.e., including all forms of ground transportation) is for current system is for countries like the U.S.A., New Zealand, Australia, Canada, Germany, Italy, etc.:

          https://en.wikipedia.org/wiki/List_of_countries_by_vehicles_per_capita

          The reasons are almost too numerous to list, but they include:

          1) Car ownership will dramatically decrease
          2) Annual miles driven per vehicle will dramatically increase
          3) Electric vehicles will be favored
          4) Vehicles usage will be much more closely tied to vehicle design (e.g. single-seat vehicles will go from extremely uncommon to very common…probably the most common vehicle on the roads, in terms of numbers)
          5) Ride sharing will be greatly promoted over current levels
          6) Vehicles will remain warm in the winter and cool in the summer, because they will be on the road constantly
          7) Accidents will be dramatically reduced, and insurance costs likewise reduced
          8) Travel to brick-and-mortar retail will be replaced by e-commerce with goods delivered by autonomous delivery vehicles
          9) Road capacities will increase dramatically, meaning fewer roads will be needed per vehicle mile traveled
          10) Parking lots will nearly disappear

          …etc., etc. I’m probably missing 10 other good reasons why autonomous vehicles providing transportation-as-a-service will provide far, far lower costs per passenger mile traveled than the current system of combined private vehicle ownership and mass transit.

          There’s a revolution coming in transportation, and it will probably start within 5 years and be over within 25 years (when 90%+ of the passenger miles traveled will be by autonomous vehicles providing transportation as a service).

    • Sailor Boy January 29, 2018 at 12:45 pm #

      “So a city the size of Auckland might have 100,000+ requests for transportation from all different locations to travel to other different locations at any particular time. How does it make sense to have buses of a fixed size traveling fixed routes at fixed times in that situation?”

      Because the people who travel down a route aren’t fixed, but the number of people who do are.

      We get 100,000 people into out city centre at peak everyday. This wouldn’t be possible without using buses and trains. Letting people off all of the buses wouldn’t be possible without them running on set routes at set times so that kerb space can be allocated.

      It is geometrically impossible for ride hailing to serve this transport demand.

      • RossB January 29, 2018 at 7:28 pm #

        >> No, every study *I* know of concludes that autonomous vehicles operating in transportation-as-a-service mode will be far, far less expensive per mile than the total transportation system (i.e., including all forms of ground transportation) is for current system is for countries like the U.S.A., New Zealand, Australia, Canada, Germany, Italy, etc.:

        I’m not describing the current system! I am describing an *automated* transit system. It is cray to think that somehow we will have automated cars running around everywhere, yet buses will still operate like they do now.

        Holy cow — five minute headways on *every single corridor*, all day and night? No one has that. Such a system would be the envy of every city in North America (if not the world). Building that system would be expensive, but not if you simply ran it with vans (or cars). But operating it *now* would be extremely expensive. Even if you had a grant to buy all the vehicles, they would sit idle, because labor remains the most expensive cost of operations (by far).

        With automation, the biggest expense becomes buying the cars. Sure, gas is expensive, but let’s just look at cars for a second. In an ideal system, you have a grid, with parallel lines about a quarter mile apart. Even that is probably overkill, but it does allow better stop spacing (https://humantransit.org/2010/11/san-francisco-a-rational-stop-spacing-plan.html). Each bus goes back and forth, to the end of the service area. A grid with quarter mile lines means four vertical lines and four horizontal lines per square mile. Since each line is a mile long, that works out to 8 linear miles per square mile. Double that for the round trip (you want to serve both directions), and you have 16 miles of bus service per square mile.

        To get good headways, you spread the buses out. The faster the bus, the fewer you need. At an average speed of 12 MPH, one bus every mile would provide 5 minute headways.

        So with a hundred square miles to cover, you would have 1,600 buses. That is a lot of buses. But don’t think of them buses. There is no reason why they would all be buses. Some would be vans ,some would be cars. It really is about peak usage per corridor. If it needs to be a bus, it is a bus. If a van will do, it will do.

        The point is, for a 100 square mile city, you need 1,600 vehicles. You would more to add reliability (make sure those buses come very five minutes). But 2,000 vehicles would do it. That is peak — that is the most you would need to provide a fantasy system with very small headways, excellent stop spacing and short walks for every single person in this city.

        It is absurd to think that a point to point system could provide that kind of service, in all but the most desolate of places. I’m only talking about 2,000 vehicles serving 100 square miles. There are a couple cities that happen to be very close to 100 square miles (according to Wikipedia). Both are in Florida. Tallahassee and North Port.

        Neither are especially big. Tallahassee is the 126th biggest city in the U. S. North Port is much smaller, and only has 57,000 people. But can you even provide for the people of North Port, with only 2,000 vehicles? During rush hour, are you saying there are (or would be) only a couple thousand cars on the road? Sorry, but that just seems absurd.

        With an automated system, the biggest concern is peak usage. The savings that result from having cars sit around during the dead time aren’t really much of a savings. I need to buy a lot of cars to serve peak time (when everyone is riding in one my vehicles). In fact, I need to buy more than that, so that the riders can get a ride when the want it. If all the vehicles are in use, and I don’t care (e. g. current ride share systems) then it doesn’t matter. So what if it takes an extra half hour for your cab (or they jack up the price). But if I’m running a system for the public good, then I want to serve the entire area, which means I need extra vehicles sitting idle at least reasonably close to every potential customer.

        As far as gas usage goes, it gets complicated. Buses could be running without customers. But that is also the case with an automated taxi-cab system. You still have to get the cab out to the customer, and in many cases, it needs to be in the neighborhood, so that it can respond in a reasonable fashion. It really isn’t that different — except that on many corridors, you would pick up way more people. That means on average, I’m guessing a transit system like the one I described would use less fuel as well.

        In many cases (certainly in North Port) an automated, point to point system would be an improvement. Since North Port probably doesn’t have much in the way of transit infrastructure (transit lanes) a cab would be faster. But transit would still be very good (much better than it any system in North America, let alone better than it is now). More to the point, it would be cheaper to operate. It would be silly to have the public pay extra so that people have something ideal, when we could deliver for them something much better than what they have now.

        • Mark Bahner January 29, 2018 at 9:44 pm #

          Hi Ross,

          If we’re looking to talk about a city that’s about 100 square miles, I nominate Durham, NC, at 108 square miles, population 260,000.

          The one thing that I know should absolutely *not* be done would be to build a Durham-Orange Light Rail System. Anyonet who works for any government at any level who wants that should be fired:

          https://www.indyweek.com/news/archives/2017/04/07/the-durham-orange-light-rail-now-costs-more-than-3-billion-and-durham-will-have-to-pay-more-of-it

          I would put a smiley face there, but I don’t think there’s anything funny about it.

          • RossB January 30, 2018 at 9:12 am #

            I agree. For a city like that, a project like that makes no sense. Population density is very low throughout the region. In Durham itself there are only two census blocks above 10,000 people per square mile. In Chapel Hill, there are four. The biggest is still less than 20,000 ppsm.

            In contrast, the River Line is one of the worst light rail lines in terms of ridership per mile. It has less than 10,000 riders a day. It didn’t cost that much to build, as it leveraged existing rail lines. But the low ridership exists despite the fact that Trenton is much more densely populated. That line connects to Camden, which is also more densely populated than Durham or Chapel Hill, and sits across the river from Philadelphia! It is absurd to think that the Durham-Orange line will be worth the cost compared to other lines that really don’t have many riders, even though they are much, much more densely populated. 3 billion? Insane.

          • Mark Bahner January 30, 2018 at 1:14 pm #

            Hi Ross,

            “I agree. For a city like that, a project like that makes no sense. Population density is very low throughout the region. In Durham itself there are only two census blocks above 10,000 people per square mile. In Chapel Hill, there are four.”

            Yes, and of the 648 towns and cities in North Carolina, Durham is in the top 5% in population density. Further, the light rail would essentially connect the University of North Carolina Hospital in Chapel Hill to Duke University Hospital in Durham. But *nobody* makes that trip! Nobody goes from one hospital to the other. Or if they do, they wouldn’t ever stand for the 45+ minutes it was going to take to go from UNC Hospital to Duke Hospital. (Oh…and then they later added a connection to North Carolina Central University, NCCU, a historically black university.)

            But despite all this insanity, I went to the Durham City Council meeting in which the go-ahead was given for engineering studies for the project, and the measure passed unanimously and without any debate. Why? Because everyone thinks light rail is cool. How could any decent person be against light rail?

            Plus, much more importantly, the expectation was that the federal and state government were going to pay the huge majority of the cost.

          • Mark Bahner January 30, 2018 at 1:17 pm #

            I forgot to include a map for the proposed Durham-Orange Light Rail Project (after the addition of the spur to North Carolina Central University (NCCU).

            http://ourtransitfuture.com/wp-content/uploads/2017/04/0509_map_DOLRT-withNCCU-blue_160900-1.jpg

  4. Mark Bahner January 25, 2018 at 8:50 pm #

    “Autonomous vehicles do not solve the fundamental problem that we can’t fit as many cars as we wish on the roads today.”

    We want to get from our origin to our destination as quickly as possible. I think autonomous vehicles are going to boost the speed by which we get from our origin to our destination *tremendously*. Particularly for the places on earth where humans move most slowly.

    For example, page 22 of this report has speeds (in miles per hour) for taxis in the Manhattan central business district (CBD) and midtown core from 2010 to 2013:

    http://www.nyc.gov/html/dot/downloads/pdf/mobility-report-2016-print.pdf

    Central Business District: 9.4 mph in 2010, down to 8.9 mph in 2013

    Midtown Core: 6.5 mph in 2010, down to 6.1 mph in 2013.

    I expect when the system is fully automated, within 25 years, speeds will easily be 4 times higher in the central business district (35-40+ mph), and 5 times higher in the midtown core (30 mph).

    P.S. By way of comparison, the average subway speed in NYC appears to be about 17 mph.

    • Jason January 26, 2018 at 11:22 pm #

      If you have enough cars on the road, autonomous or not, you are going to have congestion. You simply can’t carry NYC subway passengers on the road with cars. You either restrict number of users (through e.g. price increases), or you add more and more cars until congestion causes cars to slow to a crawl such that people at the margin choose alternatives (travel at a different time, go to different places, take the subway).

      • Mark Bahner January 27, 2018 at 6:11 pm #

        “If you have enough cars on the road, autonomous or not, you are going to have congestion.”

        I think most people are *vastly* underestimating the potential for autonomous vehicles to relieve congestion, especially if they are operated in terms of transportation-as-a-service. The reasons why autonomous vehicles operating as transportation-as-a-service can dramatically reduce congestion include:

        1) It will be possible to know every person’s origination point, destination, and desired arrival time. Therefore, it will be possible to look at where “flash” buses or minivans can start and end to deliver people. A person might go from a single-seat, single passenger vehicle at the origin, to a minivan or small bus on a heavily traveled road, and back to a single seat vehicle to get to the destination.

        2) Notice how I mentioned “single seat” vehicles. What percentage of the cars on the road today in the U.S. (or any country) are single-seat vehicles? The answer is far less than 1 percent. But when autonomous vehicles are operating in the mode of transportation-as-a service, probably more than 50 percent of the vehicles on the road will be single-seat vehicles. Why? Because why have room for two or more people if only one person is in the car?

        3) A single-seat vehicle operated autonomously means that two, or potentially even three, vehicles can be operated abreast in a single lane.

        4) Vehicles will be able to go much faster, and be operated much more closely spaced. For example, there’s the old rule of thumb of one car length per 10 mph. With autonomous vehicles operating in communication with one another, even one foot per 10 mph is probably more than necessary.

        I’ll discuss the example of the capacity of the George Washington Bridge with human drivers versus with autonomous vehicles operating in transportation-as-a-service mode, after I get some time to think about some actual numbers. (I think you’ll be amazed by the potential numbers for autonomous vehicles operating in transportation-as-a-service mode.)

        • Mark Bahner January 27, 2018 at 7:23 pm #

          An analysis of the George Washington Bridge:

          Current peak traffic is about 20,000 vehicles per hour (11,000 headed west to New Jersey and 9,000 headed east to Manhattan at 5 PM). Speed limit is 45 mph.

          Let’s look as the potential capacity of the bridge operating with autonomous vehicles (AVs) operating as transportation-as-a-service. Assume:

          1) Speed of approximately 60 mph.

          2) Vehicles separated by approximately one car length.

          3) Every vehicle is a minivan carrying an average of 5 passengers.

          4) Each vehicle is 17 feet long, and 13 feet separate each vehicle.

          5) Assume that 8 lanes are dedicated to westbound traffic and 6 lanes are dedicated to eastbound traffic. (With autonomous vehicles operating as transportation-as-a-service, it’s possible to switch a lane from one direction to another in just a few minutes…with some additional pavement added around the bridge.)

          6) Speed of 60 mph is 88 ft/s…let’s call the speed 90 ft/s for easy computation.

          7) 90 feet would be 3 minivans (each 17 feet long with a 13 foot gap, for a total per minivan of 30 feet). So that means each second you could get 15 people (3 minivans) per lane. If we have a total of 14 lanes (8 westbound and 6 eastbound) we could transfer 15 people per lane per second x 14 lanes = 210 people per second. That’s an astounding 756,000 people per hour!

          8) To look at this amazing number in a different perspective, Manhattan on a typical day has a population of about 4 million. Let’s assume that the real limit to passengers on the GW bridge with AVs operating in transportation-as-a-service mode was “only” 400,000 people per hour. That would still mean Manhattan could be essentially evacuated by the GW Bridge in about 10 hours!

          • Jason January 28, 2018 at 12:11 am #

            Three cars per second is fantasy. Wireless networks are not reliable enough for this. At 60mph, one mistake and you have a huge chain accident in making. Plus, for this to happen we will have to wait until the day 100% of cars are autonomous. This will not happen for decades. For the foreseeable futures most streets will be limited to a car every 2 seconds.

            It’s generally not disputed that a 14 lane highway can have subway-like capacities today. The problem is that we can’t build 14 lane highways everywhere we currently have subways.

          • Mark Bahner January 29, 2018 at 8:45 am #

            “Three cars per second is fantasy.”

            No it isn’t. I don’t think you’ve read sufficiently on the matter.

            Here is a Daimler webpage talking about platooning (vehicles following each other closely) of their trucks:

            http://media.daimler.com/marsMediaSite/en/instance/ko/Daimler-Trucks-is-connecting-its-trucks-with-the-internet.xhtml?oid=9920445

            Connected vehicles in a platoon require a distance of only 15 instead of 50 metres between them. This considerably smaller distance produces a significant reduction in aerodynamic drag – comparable to slipstream riding in cycling competitions. In this way a platoon of three trucks can achieve a fuel saving of around seven percent, reducing CO2 emissions in the same measure. 

            So that’s discussing semi-trucks (which have a much longer stopping distance) separated by only 15 meters (49 feet) at highway speeds.

            And here’s an IEEE (Institute of Electrical and Electronics Engineers) discussion of platooning for cars:

            https://spectrum.ieee.org/automaton/robotics/industrial-robots/sartre-autonomous-car-platoons

            “The researchers and their industrial partners expect the technology to be ready for production in a few years. The biggest remaining hurdles are no longer technological, but legal and social: Not sure I’d feel safe taking my hands off the wheel and foot off the brakes when going 130km/h and driving only 5 meters behind a truck.”

            So that’s talking about a car following a truck to within 5 meters (16 feet) at 130 km/h (81 mph).

            My proposed situation was minivans following one another to within 13 feet at 60 mph on the GW Bridge, which is extremely straight and could be extensively instrumented, in terms of sensing vehicle distances and sending wi-fi signals to all vehicles. (Also, I chose 13 feet mainly for ease of calculation. :-))

            My proposed situation is entirely compatible with Daimler webpage and the IEEE webpage.

            “It’s generally not disputed that a 14 lane highway can have subway-like capacities today. The problem is that we can’t build 14 lane highways everywhere we currently have subways.”

            I think both you and asdf2 may have the misimpression that I’m proposing to eliminate the NYC subway system. I’m not. I’m simply pointing out that autonomous vehicles operating in transportation-as-a-service (TAAS) mode have the possibility to increase road capacities to truly mind-boggling levels.

          • Jarrett January 29, 2018 at 9:09 am #

            Mark.

            You need to explain the space needs of assembling and disassembling these platoons. Elon Musk’s tunnel, like all road tunnels, works fine with everything working at high speed. It’s the space taken by accel and decel, and assembly/disassembly of platoons, that eats up too much of the city.

            The other issue is that I’m thinking about the long, long period when we have a mixture of human and automated drivers on the road. Solving that issue politically will take decades longer than it will to solve it technically.

          • Mark Bahner January 29, 2018 at 11:12 am #

            “You need to explain the space needs of assembling and disassembling these platoons.”

            Jarrett, these specific platoon situations across the GW Bridge that I described–either minivans carrying an average of 5 passengers, or buses carrying an average of 50 passengers–will almost certainly never occur…even by 2040, when the overwhelming majority of passenger miles traveling in the country will be by autonomous vehicles providing transportation as a service.

            And the reason why they will *not* occur, is because they simply won’t be needed. As I noted, the peak hourly traffic on the GW Bridge is about 20,000 vehicles per hour, and even at the ridiculously large average occupancy of 2.5 people per vehicle, that’s only 50,000 people per hour. The situations I was talking about involved transferring over 500,000 people per hour…more than 10 times the present peak hourly load.

            Geek note: Per page 24 of this document, the average car occupancy on the GW Bridge is actually 1.74 per vehicle, and about 80% of the morning crossings are in cars carrying only one person. (!)

            http://www.nyc.gov/html/dot/downloads/pdf/manhattan-river-crossings-2013.pdf

            So what will actually happen is that there will be some mix of single-occupant vehicles, minivans, small buses, large buses, and trucks. Let’s assume that the tolls remain as at present:

            “E-ZPass users are charged $10.50 for cars and $9.50 for motorcycles during off-peak hours, and $12.50 for cars and $11.50 for motorcycles during peak hours. Trucks are charged cash tolls of $20.00 per axle, with discounted peak, off-peak, and overnight E-ZPass tolls.”

            If tolls remain as at present, then with autonomous vehicles operating with transportation-as-a-service (TAAS) the number of cars/minivans carrying 3, 4, 5, 6+ passengers should *dramatically* increase. After all, if passengers ride alone, they will be charged the full $10.50 or $12.50 toll, but if they’re riding in a minivan with 6 passengers, the cost of the toll will only be one-sixth as much.

            How will they be assembled? If a transportation network company knows where tens of thousands of people are coming from and going to at any instant, it should be very easy to match up several people to a single vehicle. People would just go to a curb, and the vehicle would be there in a minute or two. The persons already in the vehicle would have to wait even less. Then, when the minivan carrying say 4 people gets over the GW Bridge, it might take one or more people to their individual destinations, or it might let other people off at the curb of a street, where they’d either take a *single-seat* single occupancy vehicle, that would only take up half a lane, to the their destination. (Or they might get deposited at a subway station.)

            The key is that the system is tremendously flexible. A person’s smart phone might show 4-5 options, with varying prices and arrival times. Or after a period of time, the smart phone would know the type of option its owner would typically choose, and just show that option unless directed otherwise.

            The system would also be flexible to an extent that would be insane with human drivers…with lanes on the GW changing from eastbound to westbound with only a few minutes delay…and with roads inside Manhattan going from 2 lanes each direction to 3 lanes one direction or the other, or even switch to one-way and back, just on a few minutes notice. Not to mention the cars passing each other at 90 degree angles at intersections without stopping. 🙂

          • Sailor Boy January 29, 2018 at 12:46 pm #

            You’re joking right? You genuinely think you can get that many vehicles off of the freeway?

    • RossB January 28, 2018 at 10:36 am #

      @Mark — Every leave a stadium? One of the things you notice is that even though everyone is packed quite tightly, moving with great efficiency, it still takes a long time to empty the place. There simply isn’t enough space for people to fit in the aisles.

      That is the problem. Even bumper to bumper cars take up more space than a bus. I noticed even in your examples you’ve moved away from “anywhere to anywhere” cars (i. e. taxi-cabs) to mini-vans. Another word for those is a bus. For lightly traveled corridors, bus service can be handled by mini-vans. But if you want to move the most people across a bridge, then a fleet of big buses is going to be better than a bigger fleet of cars, regardless of the space allowed per vehicle.

      • Mark Bahner January 28, 2018 at 8:54 pm #

        “@Mark — Ever leave a stadium?”

        Yes, I have, and I was going to discuss precisely that situation as an example of one of the ways that autonomous vehicles providing transportation-as-a-service are going to revolutionize transportation.

        “One of the things you notice is that even though everyone is packed quite tightly, moving with great efficiency, it still takes a long time to empty the place. There simply isn’t enough space for people to fit in the aisles.”

        Yes, to give one example…I lived in Farmington Hills, MI in the middle 1980s. I went to a basketball game between the Pistons and the Atlanta Hawks. The game was played in the Pontiac Silverdome, because that’s where the Pistons played home games that year. Anyway, my house in Farmington Hills was about 30 minutes’ drive from the Silverdome. However, after the game, which was over at like 10 PM, as I recall I spent *almost an hour* in the parking lot at the Silverdome! As I recall, the total time to get home was almost 2 hours, where the time to get to the parking lot at the start of the game was about 30 minutes.

        Here’s how the situation might have worked if autonomous vehicles provided transportation-as-a-service. I would come out of my gated-community apartment in Farmington Hills, and get into a single-seat car. I’d be driven a few miles to a place in Farmington Hills where a minivan or small bus would be waiting, to carry people from Farmington Hills to the Silverdome. We’d get out and go to the game. After the game, we’d come out of the stadium and walk *directly* to one of several hundred very large buses sitting in the parking lot. I’m talking about double-decker buses that can handle 80+ people. Those very large buses would drive to parking lots in places like shopping malls or businesses around the stadium. I’d then get into a minivan to go to a central point in Farmington Hills, and then a single-seat vehicle to get to my home. (Unless there were other people from my apartment complex, which was *huge*, who also went to the game.)

        So to get to the game, I’d make one transfer from a single-seat car to a minivan (which would deposit us right at the gate, not in the parking lot). Then coming home, I’d make two transfers, because the double-decker buses would be almost the only vehicles in the parking lot. (There’d of course be some separate buses or vehicles for people with mobility problems.)

        “But if you want to move the most people across a bridge, then a fleet of big buses is going to be better than a bigger fleet of cars, regardless of the space allowed per vehicle.”

        Yes, I agree completely. But in the situation I described, currently the peak number of vehicles on the GW Bridge in NYC is about 20,000 per hour. Let’s probably be extremely conservative and say that their are on average 2.5 people per vehicle. (With the 0.5 person obviously being in a lot of pain.) So that’s 50,000 people per hour is the current peak traffic on the GW Bridge. I described a plausible situation for autonomous vehicles operating in transportation-as-a-service mode in which more than 500,000 people per hour could cross the GW Bridge. That’s more than 10 times the current maximum capacity! So there probably would never be a need for all buses to cross the GW Bridge. But let’s buses with 60 people on them crossed the bridge at 60 mph (90 feet/second) and with one bus every 90 feet. That is, the bus is about 60 feet long, and the buses have a 30 foot distance between them. And let’s say there are 50 people in each bus. Now we’re talking about 50 people per second per lane. There are 14 lanes, so the GW Bridge capacity would be an astounding 700 people per second, or 2.5 million people per hour (!!!).

        But the GW Bridge will never need anywhere near that capacity, even in a situation where all of Manhattan needed to be evacuated quickly.

        P.S. I’m not positive the GW Bridge is designed to handle the loads from a situation where there are fully loaded buses every 90 feet. But since a capacity of 2.5 million people per hour will almost certainly never occur, it doesn’t really matter.

        • Sailor Boy January 29, 2018 at 12:48 pm #

          So, in other words, you’d use a taxi to get to a fixed route bus station then catch a bus.

          • Mark Bahner January 29, 2018 at 7:35 pm #

            “So, in other words, you’d use a taxi to get to a fixed route bus station then catch a bus.”

            You need to read more carefully. Nowhere did I mention any “fixed route”…let alone a “fixed route bus station.”

            My whole point was that it doesn’t make any sense to have fixed-route and fixed-time vehicles, if the system involves autonomous vehicles operating as transportation-as-a-service.

        • RossB January 30, 2018 at 7:03 am #

          Just to be clear, I was talking about pedestrians leaving the stadium. The same thing happens when people leave a movie theater. If there is no one in the arena, you can walk right out. But if is full, you have to wait. That is because even if people are walking in lock step, marching out, they can’t all fit in the aisles.

          The same is true on the freeway. During rush hour, there is simply too much demand for space on freeway, even if everyone is driving bumper to bumper. In fact, the image that every references doesn’t even show safe driving distance (http://urbanist.typepad.com/.a/6a00d83454714d69e2017d3c37d8ac970c-popup). Yet the cars still take up more room.

          Anyway, Sailor Boy has summarized the solution quite well. You are talking about taking a taxi then a bus. Works for me, and works for about every other city. Except as transit gets much, much cheaper, a lot of people will simply take transit the whole way.

          • Mark Bahner January 30, 2018 at 7:55 am #

            “The same is true on the freeway. During rush hour, there is simply too much demand for space on freeway, even if everyone is driving bumper to bumper.”

            No, with autonomous vehicles operating in transportation-as-a-service (TASS) mode, there is no real-world limit to freeway traffic, in terms of the number of persons per hour that can be transported. As I pointed out in the GW Bridge example, in the real world, the present-day maximum hourly demand on the GW Bridge is about 20,000 vehicles per hour, with about 1.75 people per vehicle, for a total of 35,000 people per hour. But with autonomous vehicles operating in TAAS mode, it would be *easy* to get *10 times* as many people per hour across the GW Bridge as the current peak hourly demand. That would be done not simply by increasing the speed and decreasing the distance between vehicles, but by boosting the number of people per vehicle well above the current value of 1.75/vehicle.

            “Anyway, Sailor Boy has summarized the solution quite well. You are talking about taking a taxi then a bus.”

            No, I’m not. I don’t understand why y’all can’t follow what I’m writing.

            “Works for me, and works for about every other city. Except as transit gets much, much cheaper, a lot of people will simply take transit the whole way.”

            If you mean “transit” including more than 50 percent of the vehicles being single-occupant vehicles, then I agree. But that transit won’t be provided by city governments. Why not? Because leaders of city governments behave like idiots. The Durham-Orange Light Rail Project is a perfect example. And why do the leaders of city governments behave like idiots? Because they don’t care, because it’s ***not their money.***

            https://www.indyweek.com/news/archives/2018/01/29/what-will-trumps-infrastructure-plan-mean-for-the-durham-orange-light-rail-project

  5. R. W. Rynerson January 26, 2018 at 12:57 am #

    Notes on privatization:

    Over half of Denver RTD’s rubber tire service is contracted out under a state legislative mandate. The original law had the contractors buying their own buses and retaining the cash revenue that they collected. It was asserted that this would provide better service than the public agency at a lower cost. Very quickly it turned out to be unattractive to run a business with the inherent risks of those requirements and the switch was made to leasing transit district owned buses and turning over cash revenue to the district.

    Which brings up some questions for the hypothetical AV service provider. When a bridge collapses and cycle times skyrocket, will they be able to impose demand pricing on the customers? If so, added vehicles might be rushed from another city. If a limit is imposed on demand pricing, the service provider will expect a subsidy. When a drunk urinates in private in an AV, he may feel better, but what about the next customer? Can the private AV provider blacklist the drunk? (In the evil, big-city bus, if the drunk gets past the operator there still is SOME social restraint and other seat options.) Will the AV provider have limited service areas or may I proceed to North Platte, where if it breaks down, I can just abandon it? If service areas are limited, is someone responsible for cleaning and securing a transfer station?

    These are all real-world hazards that can be dealt with, but my surmise is that as in Colorado the AV providers will try to enlist government to take the risk.

  6. Mark Bahner January 26, 2018 at 3:17 pm #

    “When a drunk urinates in private in an AV, he may feel better, but what about the next customer? Can the private AV provider blacklist the drunk?”

    Right now, everyone owns their own car. I don’t see any reason why drunks would be *prevented* from owning their own AVs. Or paying other AV fleet operators who might be more willing to forgive previous disgusting behavior. So I don’t see any good reason for fleet operators to be prohibited from serving customers who have previously behaved badly.

    “Will the AV provider have limited service areas or may I proceed to North Platte,…”

    If one provider doesn’t serve the area of your destination, I don’t see why there wouldn’t be some other provider that does serve your destination. Or again, you could own your own AV, or borrow a friend’s AV. (I would be much more inclined to lend a friend my car if it was an AV…when presumably any accidents would be the AV’s fault, not the person sitting in the car.)

    “…where if it breaks down, I can just abandon it?”

    I don’t know of any current rental car company that does not come and deliver a replacement if the car breaks down. It seems like it would be a terrible business decision to provide a means to get to the destination if there is a contract to go to a particular destination.

    “These are all real-world hazards that can be dealt with, but my surmise is that as in Colorado the AV providers will try to enlist government to take the risk.”

    Hopefully, governments wouldn’t take those risks. I don’t see the reason for governments to assume risks in the situations you described. As I noted, all of those situations seem to have close parallels in today’s world where the government isn’t stepping in (e.g., rental car companies dealing with customers damaging cars, or dealing with cars that break down, AAA dealing with cars that break down, etc.).

    P.S. One thing we don’t have much of today is situational pricing. There’s a little bit of that in the form of congestion pricing for some roads in some situations (e.g variable tolls around Washington DC or congestion pricing for roads in downtown London, England). We should have a lot more of that, because it would allow people to see the real costs of transportation.

    • R. W, Rynerson January 26, 2018 at 9:42 pm #

      I agree that the car rental model could work for those with credit accounts, although I have experienced territorial restrictions on rentals and had credit cards declined due to just having made an overseas purchase. Government’s entrée will be concern for those who need transportation, but have no credit. I should clarify that I’m not necessarily talking about direct subsidies. It might be in the form of regulated monopolies that would then have the resources to overcome major disruptions or social issues.

      Note also the customers boarding buses who still pay cash, even when buying a day pass or investing in a prepaid fare card may offer a saving.

  7. asdf2 January 26, 2018 at 7:43 pm #

    Even if the technology exists to operate autonomous buses, it is far from given that mainstream public transit agencies will actually do so. Many of them have powerful labor unions to contend with who will simply not accept bus drivers’ jobs getting outsourced to machines. These unions are often impossible for the agency to fight because the official that heads the agency (or has the authority to hire/fire the person in charge) needs the union’s political support in order to get re-elected.

    Look no further than the New York Subway and its two-man crew on every subway train as an example. The conductor is paid to basically just sit there, the same way that a bus driver would get paid to sit there if and when the bus is ever made capable of running itself.

    This is where the private sectors has the advantage. Simply by not being run by officials in the pockets of the unions, private companies can and will make the driverless vehicles actually driverless, once the technology allows it.

    • Mark Bahner January 26, 2018 at 8:25 pm #

      “Even if the technology exists to operate autonomous buses, it is far from given that mainstream public transit agencies will actually do so.”

      When transportation-as-a-service comes–I don’t think there’s any question about “if”–the cost per mile traveled will go down significantly. Plus, worries such as finding parking and car theft or damage will disappear. Finally, people will no longer have to drive, so even that current advantage of mass transit will disappear. So if transit systems persist in any practices that raise prices, there will simply be no competitive advantage for mass transit.

      • asdf2 January 27, 2018 at 10:04 pm #

        The need for transit isn’t going to go away for the simple reason of space constraints.

        Imagine what would happen if all the subway riders in Manhattan suddenly decided to switch to autonomous cars? The answer would be total gridlock. And the only way to allow people to bypass the gridlock, which scales up to the population of NYC is the subway. So, AV technology or not, the subway and the need for the subway, is not going anywhere.

        • Mark Bahner January 29, 2018 at 11:33 am #

          As I wrote previously, I’m not suggesting eliminating the NYC subway system. But we all need to recognize that Manhattan is unique in the U.S., and close to unique in the world, in terms of population density.

          Manhattan has a daytime population density (or “occupancy density”…since many people come in to work, but live outside Manhattan) of about 70,000 people per square kilometer. As I’ve noted previously, my home town of Durham NC has a population density of about 800 people per square kilometer, and the NC capital of Raleigh has a density of about 1,100 people per square mile. And both Durham and Raleigh are in the top 5 percent for population density for all towns/cities in NC. So the fact that subways are reasonable in Manhattan really doesn’t say much about transit in most other cities in the U.S. or around the world.

          • Sailor Boy January 29, 2018 at 12:52 pm #

            That’s not unique. Downtown Auckland has daytime occupancy density of over 50,000. I’d imagine that Vancouver, Toronto, LA, Chicago, would all be similar.

            I think you are grossly overestimating the spatial efficiency of autonomous cars, and grossly underestimating the spatial efficiency of buses.

      • RossB January 30, 2018 at 7:20 am #

        When automated taxi-cab service arrives, so will automated transit service. When that happens, an automated transit network is cheaper to operate and build than a point to point system. The cheapest way to get people around even a small city would be to have them share rides and transfer vehicles. A point to point system would be more expensive in all but the smallest of loads. Ridership only has to average more than one person in a vehicle to be cheaper with a transit system. A point to point system would be faster (assuming no preferential treatment for transit vehicles) but a transit system scales, and thus becomes much better as you ramp up service. In other words, an automated taxi-cab service becomes much cheaper, but an automated transit system becomes much cheaper and much faster. As automation takes hold and drops the cost of both systems, they begin to narrow in terms of performance (a very good transit network is not that much slower than a good taxi-cab system). With Taxi-cab service remaining more expensive, and transit service improving dramatically, automated taxi-cab service would still be considered a luxury in all but the most remote places (where people might be inclined to buy their own vehicle).

        • asdf2 February 11, 2018 at 4:09 pm #

          “When automated taxi-cab service arrives, so will automated transit service.”

          No, it won’t. The bus driver’s union will see to it that it doesn’t.

          They’ll justify it to the public in terms of safety and people with special needs. They’ll say we need to especially cautious about safety on a vehicle carrying large numbers of people, above and beyond what would be demanded from a taxicab. They’ll say that some people (e.g. old people who don’t like technology) won’t feel comfortable riding on a vehicle driven by a robot, yet they can’t drive, and if the transit system doesn’t have human bus drivers, those people will be trapped in their house. They’ll argue that an official presence on each bus is necessary to deter harassment and bad behavior. They’ll argue that human drivers are needed to collect fares and answer questions from people who don’t know where they’re going.

          At the end of the day, all of the above is just excuses. But, watch the union make them anyway. And watch the politicians in the union’s pocket accede to their demands.

  8. Mark Bahner January 29, 2018 at 7:22 pm #

    “That’s not unique. Downtown Auckland has daytime occupancy density of over 50,000.”

    ????!!!

    I wrote that Manhattan has a daytime occupancy of about 67,000 people per square kilometer (that was from 3.94 million people in 59 square kilometers). You say *Downtown Auckland* has a daytime density of “over 50,000”?! Per square kilometer?!

    I don’t see how Downtown Auckland can even be close to Manhattan. The tallest building in Auckland is 574 feet (per word on the Internet). That doesn’t even make the *top 100* of tall buildings in NYC…the overwhelming majority of which are in Manhattan:

    https://en.wikipedia.org/wiki/List_of_tallest_buildings_in_New_York_City

    I think one of us is very wrong…or we’re talking different units…or you’re limiting “Downtown Auckland” to a very small area!

  9. Sailor Boy January 30, 2018 at 1:22 pm #

    Downtown Auckland is ~4 square kilometres inside the motorway noose. It’s daytime occupation is ~200-250,000. That’s over 50,000 per square kilometre. This is the definition of downtown used by the Ministry of Statistics , the Ministry of Finance, and the Ministry of Business Innovation, and Employment in New Zealand.

    Manhattan is 59 km2 in a metro area of approximately 20m people, Downtown Auckland is 4km2 in a metro area of approximately 1.5m people. In each case, we have taken roughly 3km2 per million people in the metro area.

    You do not need tall building to achieve high densities. The 11th arondisemnt in Paris has almost no buildings over 10 floors and has almost 50,000 residents per km2 https://en.wikipedia.org/wiki/11th_arrondissement_of_Paris

    You think that dense downtowns are unique because you live in an area (Triangle from New York to Chicago, to Atlanta) where massive subsidies were paid to demolish buildings for freeways and parking lots in downtown. That is almost unique in the world. Dense downtowns beyond car scale are ubiquitous worldwide.

    • Mark Bahner January 30, 2018 at 4:13 pm #

      “Manhattan is 59 km2 in a metro area of approximately 20m people, Downtown Auckland is 4km2 in a metro area of approximately 1.5m people.”

      Yes, so Downtown Auckland–no offense intended–is teeny tiny compared to Manhattan. It’s about 1/15th of the area of Manhattan! And Manhattan is just one part of a city that has many times the number of people in all of New Zealand. As I mentioned before, the tallest building in Auckland doesn’t even crack the top 100 tall buildings in Manhattan.

      That’s my point. Cities with subways are extraordinary on earth. There are only about 200 cities on earth with subways. And New York’s subway system is extraordinary even among that extraordinary set of cities that have subway systems. It is #1 on earth in the number of stations. It’s one of the oldest on earth. It’s in the top 10 for ridership. It’s in the top 5 for length.

      It simply does not make sense to talk about New York City and its subway system and cities like Auckland and its transportation system as though they’re remotely similar. It *does* make sense to talk about Auckland and…maybe Indianapolis? Fort Worth? Dallas? Nashville? Charlotte?…as though their transportation systems are similar.

      “You think that dense downtowns are unique because you live in an area (Triangle from New York to Chicago, to Atlanta) where massive subsidies were paid to demolish buildings for freeways and parking lots in downtown.”

      Have you lived in the U.S.? Or visited for extended periods of time? If so, what parts of the U.S.?

      • Sailor Boy January 30, 2018 at 7:34 pm #

        “Yes, so Downtown Auckland–no offense intended–is teeny tiny compared to Manhattan. It’s about 1/15th of the area of Manhattan! And Manhattan is just one part of a city that has many times the number of people in all of New Zealand.”

        Thank you for agreeing with me. Auckland is far smaller than Manhattan, yet has exactly the same geometric problems with density.

        “It simply does not make sense to talk about New York City and its subway system and cities like Auckland and its transportation system as though they’re remotely similar.”

        Except that I just proved that they are similar. Both have sprawling metro regions with incredibly dense centres visited by roughly 1 in 6 people from the metro area every day.

        I think it is incredibly disingenuous that you have spent the last week arguing that the density of New York is unique and as soon as I demonstrate that it isn’t, you’ve shifted the goalposts.

        New York wouldn’t function without fixed route transit, Auckland wouldn’t either. Our city centre has doubled in employment, student numbers, and residents over the last 15 years without a single new trip and we are now upgrading our commuter rail to enable a four line metro (subway) pattern and we are about to start funding two light rail lines.

        In New Zealand and Australia there are 6 cities with over 1 million people. Once Auckland finishes it’s tunnel, all of them will operate above ground trains the same way New York operates it’s subway: high frequency all day service. Wellington and Canberra are smaller than 1m people. Wellington already has one of the highest rates of transit ridership in the region and Canberra has had to build a light rail line to cope with congestion. These cities all have dense city centres that can only function with transit.

        I have never lived in the US, I’ve also never visited. I’ve read all about your subsidies for freeways and suburban homes. This is the result https://acme.com/same_scale/#39.76876,-86.15825,-36.84927,174.76455,15,H,H
        https://acme.com/same_scale/#33.75419,-84.38666,-33.86762,151.20698,16,H,H

        Cities like yours are the global exception. Most developed world cities over 1,000,00 people have rapid transit, with very dense city centres.

      • Jason February 2, 2018 at 10:57 pm #

        4km^2 may be small compared to Manhattan but it is absolutely large enough to cause total chaos if everyone going there travel using cars, even autonomous ones.

        Sendai, Japan has even less people in a similar area yet it needs subways and trains to function. Rail transit is not rare in Japan. Practically every city worth mentioning has it. I’m not sure what’s your point. You need to realize that the American development pattern is really quite different from a large fraction of the world. It has comparably way too many freeways and parking lots.

  10. RossB January 30, 2018 at 4:20 pm #

    There has been a lot of talk about automated taxi-cab service, but I feel like there hasn’t been enough focus on what an automated transit system would actually look like. I’ve outlined such a system in some of the comments, and they are based on the ideas of transit experts, as expressed by this author. A grid system, with frequent stop spacing, and appropriate stop distances makes sense. There are discussions about stop spacing and how far away the grid should be (https://humantransit.org/2010/11/san-francisco-a-rational-stop-spacing-plan.html) as well as how often is “freedom”, especially for a grid system (https://humantransit.org/2011/12/how-frequent-is-freedom.html).

    But what about the vehicles? For a lot of streets, I think vans, instead of bigger buses, make more sense. Ridership may go up as you increase headways, but often not enough to justify the use of bigger vehicles. I can see several reasons for agencies buying a bunch of vans:

    1) Ridership is simply too low on many of the streets. This is where an automated system will become very popular, as you can see new service, and very good service at that. For example, here is a neighborhood of Seattle called Sunset Hill (which is part of Ballard): https://goo.gl/maps/cBgS5reN6K62. It is low density, but not super low (the houses are packed relatively tightly together). It has about 8,000 people per square mile (https://arcg.is/1DyGz5). The county serves it with buses, but only during rush hour. Since it doesn’t have weekday service, it creates a pretty big hole for transit riders (the nearest bus line is a half mile away). This seems like a great place for a fleet of vans.

    2) Vans can go places buses can’t. There are a couple examples I can think of in Seattle where this would make a big difference. In that same neck of the woods, there is an east-west street called 65th. It lacks bus service, as shown here (https://seattletransitmap.com/app/) (On the map you can find 65th by first locating Green Lake north of downtown — it pretty much bisects it). There is bus service to the east of Green Lake, but not on the west side. You can see how a bus cutting across 65th would be quite popular, as it would connect various bus routes, and serve a relatively densely populated area. However, intersections like this are a problem (https://goo.gl/maps/jikcXJyLjmR2). Just as big of a problem is the steep grade to the east of that intersection. Another area that is very steep is in West Manor Place, in Magnolia, here: https://goo.gl/maps/RfsrriLzVhx. You can see all the apartments (density) but those folks have to either walk up a steep hill or down quite a ways to catch a bus. A van could do that just fine, and it isn’t so densely populated that capacity would be a problem.

    3) Cost. This is where I would love to see some numbers. The biggest capital cost is buying the vehicles. Once they are automated, the operational costs involve maintenance and fuel. In both cases, I would assume that with full vehicles, the cost per passenger is cheapest with bigger vehicles. But what if they aren’t full? I honestly don’t know the cost per passenger numbers or the mile per gallon numbers on the range of vehicles that are out there. I wonder how full a bus has to be before it is actually a better value than a van? Right now, it doesn’t have to be very full (because driver costs are a major part of transit costs) but as that goes away, I wonder how vans stack up. I wouldn’t be surprised if a bus has to be very full for it to be the better value.

    4) Labor Transition. A lot of cities would love to fire all their bus drivers and run them automatically. But most would run into major opposition. It makes a lot more sense to simply allow people to retire, and slowly buy out workers. That means adding service, rather than replacing it. I would assume that in most cities, the biggest problem is not capacity, but service. In other words, given the choice of buying 100 buses so that the most popular runs aren’t crowded, or 200 vans so that buses come every 15 minutes instead of every half hour, most people would prefer the latter. It would likely generate greater farebox recovery (which would make it more cost effective).

    Thoughts?

  11. Wanderer February 5, 2018 at 4:54 pm #

    The calculation for Manhattan is misleading, because most of the jobs in Manhattan are concentrated in much smaller areas–in Midtown and Lower Manhattan/Wall Street. The job densities in those areas must be much higher.

    Downtown San Francisco has over 300,000 jobs in a few square kilometers. Even Downtown Oakland has something like 70,000 jobs in maybe 3 square kilometers. You can’t serve either of these areas solely with cars, autonomous or otherwise. Even many roads in much more spread out Silicon Valley already come to a screeching halt at rush hour.

    The capital cost of autonomous vehicles is almost never considered in these discussions. Uber has made a completely fantastical claim that it will be able to charge people $1 per mile, but some transit people panic around that number.

    Unless you have autonomous vehicles driving right off the freeway into garages (might work in Houston?) their speed can’t be all that high. That is if you want people to be able to walk and bike on city streets.

    • Jason February 7, 2018 at 10:55 am #

      I can concur. Look at the Google campus in silicon valley. a square km of 2 story buildings in the middle of parking craters, people can go work at any time of their choosing. Cars lined up to exit the freeway for 15-30min everyday.