My colleagues at the University of Minnesota just released Access Across America: Transit 2017. The time series here is a big deal, it is now possible to look at change at accessibility systematically from a national perspective, and compare cities. From the page:
Most U.S. metros increase access to jobs by transit
The 2017 edition of Access Across America: Transitreports that 36 of the 49 largest metros showed increases in job accessibility by transit. Though rankings of the top 10 metro areas for job accessibility by transit only changed slightly from the previous year, new data comparing changes within each of the 49 largest U.S. metros over one year helped researchers identify the places with the greatest increases in access to jobs by transit. Kansas City improved more than 17 percent. San Francisco, which ranks 2nd for job accessibility by transit, improved nearly 9 percent. In all, 42 of the 49 largest metros showed increases in job accessibility by transit.
“This new data makes it possible to see the change from year to year in how well a metro area is facilitating access to jobs by transit,” said Andrew Owen, director of the Observatory. “Transit is an essential transportation service for many Americans, and we directly compare the accessibility performance of America’s largest metropolitan areas.”
Key factors affecting the rankings for any metro area include the number of jobs available and where they are located, the availability of transit service, and population size, density, and location. Better coordination of transit service with the location of jobs and housing will improve job accessibility by transit.
Maps of cities with the greatest increases in job accessibility by transit
Why does Australia have higher transit use than the US?
This question has two major explanations: Driving is harder and using transit is easier. On the road side, as my colleague Wes Marshall says: “Policy-related differences include stronger and more extensive enforcement programs [in Australia], restrictive licensing programs, and higher driving costs.”
In places like central Sydney, narrower lanes and expensive parking also make driving a burden. The Australian motorway system is less developed than the US interstate highway system, though the government is funding major new urban motorways in Australia (e.g. WestConnex in Sydney).
Transit benefits because higher population and employment density (especially around transit stations ) within cities compared to most US cities (as well as a more urban population overall) reduces access time to and from transit and enables higher frequency service to serve the demand. The train, bus, and tram systems in Australian cities are relatively high frequency and fairly reliable, with all-day service. While the systems are imperfect (as any daily commuter will tell you) they are orders of magnitude better than most of the US.
Transit service is a positive feedback system (The Mohring Effect, named for Transport Economist Herb Mohring who first identified it). More demand calls for more service, the additional service is in the form of additional buses and trains running at different times than the original service, reducing schedule delay, making transit more convenient, calling for more service. This works two ways, so transit cutbacks increase headways (decrease frequency) making transit less convenient, lowering demand, resulting in more cutbacks.
From the 1920s when tram (streetcar) use peaked (notably excepting the spike during World War II) through the 1960s there was a process of Bustitution — substitution of buses for trams. Many cities around the world (notably excepting places like Melbourne, Toronto, San Francisco, and especially selected cities in Europe) instead of paying the costs of recapitalising their tram systems, opted to convert tram lines to buses that had much lower capital costs.
In the US, there is a grand conspiracy theory, about how this came about. While most of the conspiracy theory is over-blown, there was some evil doing, as is the wont of people infected with greed (better known as people). In Minneapolis the people who converted the streetcar to buses went to jail, not for the conversion but for crimes like bribing state legislators and giving kickbacks. In Brisbane, the Paddington tram depot caught (were set) on fire as bus conversion was being debated, answering the question.
In general, the reality is much more market-rational. Electric trams were first deployed in the late 1880s, so by 1950 the service was over 60 years old. Trams needed a major capital infusion to keep operating. That capital infusion was not forthcoming from fares; in the US trams had clearly been in decline for the better part of thirty years. It was a hard call for cities not to replace their trams with buses. The private sector, which financed trams initially, were unwilling to finance it again, leaving it to local governments to come up with money for the trams (or not, as it turned out).
So most cities became tramless. Those cities were losing transit riders before the conversion and lost more after the conversion. It’s a vicious cycle.
The new Light Rail mode (See Appendix) in North America kicked off with Edmonton (1978), San Diego (1981), and Portland (1986). In retrospect, many people regret the process of bustitution, and cities that later reinstalled LRT systems would with perfect foresight likely have kept their tram lines going and recapitalised them. Note that the actual coverage of these new system is much smaller than the historical trams, most tram lines were removed in most North American cites, as in Sydney.
Wikipedia reports the farebox recovery is lower in Australia than US cities, which implies a higher public subsidy. (I am not convinced there aren’t methodological differences in accounting here, but it is worth noting).
Why is Australia’s transit use rising when the US is falling?
The second question is more difficult. One response is that fuel prices remain higher in Australia. Another is that there has been more investment in transit, including more frequent service and continuous improvements to stations and vehicles. Third, Australian cities have recently rolled out smart cards (Opal in NSW) like the Oyster Card in London, and along with it pricing reforms to reduce the fare penalty for transfers, which has significantly boosted use of transit.
Australia does some other things differently from the US. Among them is increased use of contracting out to private firms to provide service. (This is not universal yet, but is growing.) This is also done in the UK and most of Europe, but not very much in the US. This has effects on costs and perception and unionization. The contractors are for-profit businesses aligned with the idea of higher ridership, so support for transit in Australia is bipartisan, while in the US, transit is considered a Democrat issue in most places, and Republicans are often actively hostile as it is not their constituents (or only support transit to their suburban districts with high cost, low value commuter rail systems like Northstar in Minnesota).
While transit in the US is perpetually in “crisis” (to listen to its supporters), in Australia (and Canada and Europe) it is a normal part of society that is widely used, and doesn’t have the same stigma associated with it.
What should the public sector do to increase ridership?
I asked on Twitter “Would restoring Sydney Trams to a network resembling that at their maximum extent (291 km), similar in scope to Melbourne’s Tram network today, be a good use of public resources?”
The response was
50% Yes, Benefits >> Costs
27% No, Benefits << Costs
23% Maybe, Benefits ~= Costs
Looking at Sydney the densities are much higher here than in most North American cites, aside from New York, San Francisco, Chicago. I previously examined the existing and planned trams in Sydney.
Because they are widely used, they have a strong constituency for their betterment, and government is responsive in expanding the system.
Convincing existing some-time riders to ride more is far easier than going from 0 to 1 as Peter Thiel might say.
I think early ridership gains come from going deep rather than going wide. A large fraction of the US still lives in areas designed around transit (basically pre-1920 America), including city cores and streetcar and commuter rail suburbs. Residents sometimes use transit now. These places are much easier to serve because the land use in conducive to transit, the densities are high enough and the networks are oriented for transit access and service.
Good, relatively cost-effective service like Minnesota’s arterial BRT (bus rapid transit ) (MetroTransit’s A Line, eg) have shown large ridership and user satisfaction gains with low investment. The system is made more efficient with things like payment before boarding, and all-door boarding, reducing time at stops and increasing driver and bus productivity.
The aim should be to serve users better, not help non-users by reducing congestion, which may be a happy byproduct, any more than building roads aims to reduce transit crowding.
Wesley E. Marshall (2018) “Understanding international road safety disparities: Why is Australia so much safer than the United States?” Accident Analysis and Prevention 111. 251–265
Mohring, H.(1972). “Optimization and Scale Economies in Urban Bus Transportation,” American Economic Review, 591-604.
Appendix: Streetcars and Trams vs. LRT
The difference between Light Rail and older streetcars or trams is primarily, but not entirely, one of branding. Anyone who says there is a clear formal difference that people abide hasn’t gotten out much. Different cities use the same words to mean different things. Still there are differences in degree:
Streetcars or trams often share right-of-way in the street, while Light Rail often has a mostly exclusive right-of-way with at-grade crossings, but either system can be operated either way.
Light Rail vehicles tend to be wider with higher capacity and longer with higher capacity, its longer vehicle is a heavier vehicle: Light Rail is not light, it’s only light with respect to commuter trains; it’s not light with respect to buses, cars, or people. Light sounds airy and like it should be less expensive, but it’s a only a little less expensive.
Transit vehicles and services form a continuum, you can operate streetcars in exclusive tunnels as in Boston. Both LRT and streetcars differ from commuter trains but it’s a continuum in regard to that as well.
We can assess full costs to consider internal costs, subsidies, and externalities [link].
This post looks at the idea of road rent. At the margins, what is the value of road space, and how might that cost look on a per vehicle-km traveled basis?
Land has value. Land used as roads has value both as a road and potentially for other uses. What if the value for other uses was higher than that for use as a road?
In Greater Sydney land values range from to $AU210,000 per m² in Barangaroo on Darling Harbour to under $AU1000 per m² in Western Sydney [link]. In Minneapolis, we estimated a few years ago that average assessed land value as $144 per m² for roads and $30 per m² for highways. [Junge, Jason and David Levinson (2013) Property Tax on Privatized Roads. Research in Transportation Business and Management. Volume 7. pp. 35-42.] It seems that assessed value is about 2/3 of market value in Minneapolis.
In some places it is much higher, in some places much lower, the examples used herein are simply an illustration.
The idea is that there is land holder (such as a government land agency) that has to decide whether to allocate land to road uses or for other purposes.
Consider a typical suburban residential neighborhood with `free’ parking in front of houses. The land is valued at $1,000 per m². Each house requires one parking space out front, and parking is permitted 24 h per day. Conservatively, a car takes 10 m² when parked (the road is the access lane, we consider that separately). It would cost $10,000 for the land owner to purchase the land equivalent of the parked car. The annual rent on that would be $400 (at 4% interest).
In this example $400 is how much the car owner should pay annually to their municipality for a permit to park their car to cover the cost of land (not the cost of infrastructure, or any other costs of roads and mobility, just the cost of land). This is a bit more than $1/day (more precisely $1.095/day). In more expensive neighborhoods, this would be higher, in less expensive neighborhoods, lower.
For Minneapolis, I have previously estimated about 220,000 on-street spaces. At $400/space per year, this would raise $88,000,000 per year, a not inconsiderable share of the city’s $1.3B annual budget. Instead it is mostly given away free.
Consider the implications if property taxes were reduced by up to $88M in total, and parking permits sold at $400/year (payable monthly with the water and trash bill). People would realize the cost of on-street parking, and there would be less of it, and less vehicle ownership at the margins, and fewer trips by car. Space freed up could be re-allocated.
Alternatively, $400 per year is the value of public subsidy from publicly-owned land to private car owners who get `free’ on-street parking. In short from the car-less to the car owners.
Alternative Uses of Road Space
The economic idea of opportunity cost is important here. Opportunity cost is value of the next best alternative. The next best alternative to road space might be renting it out. So for instance an urban US freeway that destroyed blocks of extant development when it was built has an opportunity cost associated with the value of that real estate.
So the question arises as to what other uses could be made of the road; for if there were no other uses, you might as well store cars for free. Here are several other uses that could be considered to replacing a parking lane:
Park or parklet,
Paid parking, via meters,
Shared car parking (rented to the car sharing company),
Shared bike parking (rented to the station-based or dockless bike sharing company),
Taxi or ride-hailing stand,
Shared scooter parking (rented to dockless scooter sharing company),
Food truck or ice cream vendor,
Road for moving motor vehicles (a parking lane could be another moving lane),
Sold off for development.
The last item deserves some discussion. Consider that our road with two parking lanes (one on each side) is maybe 10 or 12m wide (~32 to 40ft). This is wider than some houses are long. The city could in principle retain the sidewalks and sell off the roadbed for townhouses or single family homes. Given the houses are already serviced by alleys, and so long as not all roads were sold off, some roads could be. An illustration of this is the Milwaukee Avenue in Seward in Minneapolis, as shown in the figure. You will see there is no paved street in front of the houses. This could be tightened up further or realigned should there be demand.
This is not appropriate for every street. However, (1) there are places this can be done, where roads are in excess and housing scarce, and (2) this illustrates that land currently used as asphalt to store and move cars has value, and that houses have value even in the absence of streets for cars in the front.
There are always excuses — utilities may need to be relocated, fire trucks would need to go slower down narrower sidewalks. But these excuses can be overcome, there are numerous examples of narrow paths that function as roads.
Note: 1 are = 100 m² and 1 hectare (ha) = 10,000 m²
Typically each car is in use 1 – 1.5 hours per day, and parked for the remainder. In the previous section, we considered parking, the `remainder,’ in this section we look at the time in motion.
When in use, the car is occupying not simply its area (the 10m² = 2m x 5m), but also is preventing the use of other space around it. On a freeway with a capacity of 1800 vehicles per hour traveling at a freeflow speed of 100 km/h, (i.e. just before the speed and flow drop due to congestion sets in) there is a critical density of 18 vehicles/km.
18 vehicles per km is 55.5 meters per vehicle. Lane width is 3.65 meters, so the area occupied is 202 m². Let’s round to 200 m². Each moment a car is in use, it is using 200 m², on which it should pay rent. So for an hour a day, this is 720,000 m² s or 72 ha s. (The meter-squared by second (or hectare second) is a new unit of measurement (a time-volume) that needs a catchier name).
It is the density that is the relevant number here, since vehicles are occupying space that we are charging rent for in this thought experiment. Though they are moving, and so the space they are occupying moves with them, there is always some space being occupied for the duration of their travel. Each of those vehicles per hour is occupying a moving window of space.
Roads are a Time Share
When roads are less congested, cars are consuming more space per vehicle. So uncongested urban are much more expensive per traveler than congested rural roads. When traffic breaks down, they are consuming less space, but presumably are occupying that space for more time, since they are going slower. Induced demand [link] and travel time budgets [link] negate that to a significant extent.
In this example, the hourly rent on 200 m² is what we are interested in. Though cars move, over the course of 1 hour of travel in these conditions, they are claiming that much space. The specific space they are claiming moves with the vehicles, but this all balances out as other cars claim the space they vacated.
Empty roads still have to be paid for, and paid for by actual road users. Even when a road is not being used, it is available to be used. Travelers have the option of traveling. Pavements cannot be easily be rolled up and allocated to other purposes on the fly, particularly purposes like buildings. (Roads can occasionally be closed for special events, but this is rare during business hours.)
Road section 3 (downtown): l= 1 km, w=3.65, v=40 km/h, q=1600, k =40 veh/km , AADT=16,000 vehicles/day/lane, p= $10000/m².
where: l = length (km), w= lane width (m), v=velocity (km/h), q=flow (veh/h), k=density (veh/km), AADT = Average Annual Daily Traffic, p= land value ($/m²), i=interest rate = 0.04, r= land rent ($/year/m²), d = days/year
Consider each road section to be a homogenous pipeline. (With heterogenous traffic, this is obviously far more complicated, and we would make use of the q, k, and v variables to compute an area-time.)
The annual rent (R) for each road section is the R=p*i*l*w
Road 1: R=$1,000/m² y * 0.04 * 5,000 m * 3.65 m = $730,000/y
Road 2: R=$5,000/m² y * 0.04 * 10,000 m * 3.65 m = $7,300,000/y
Road 3: R=$10,000/m² y * 0.04 * 1,000 m * 3.65 m = $1,460,000/y
This annual rent is paid by the road agency to the land owner for the use of land as a road. The road agency then wants to recover this cost from its customers, the travelers.
The question of how to allocate always has some subjectiveness to it. Another way of thinking about it is based on elasticity of demand. Peak hour work trips are perhaps the least elastic (least sensitive to price), and so from an economic efficiency perspective should bear the greater cost.
In this example, we take a simpler tack.
The allocation is R/AADT to get cost per year per daily tripmaker, and divide by 365 to get cost per trip, and by section length to get cost per km. In this example:
The total is thus $529.25/year or $1.45/trip to cover land rent. `Your mileage may vary,’ as the saying goes.
The implications of this are several.
At an additional $1.45/trip, travel by car (and congestion) will diminish.
Road rent is essentially additive with annualized infrastructure costs, which generally does not consider the cost of land (rather, land is often implicitly considered `free’ or a sunk cost).
If travel by car diminishes sufficiently, road space can be clawed back and redeployed for other public purposes.
Narrower lanes impose less road rent. But not necessarily proportionately so, as the throughput on narrower lanes (with human drivers) may be lower as drivers are less keen to be immediately adjacent to nearby high-speed vehicles.
Slower moving vehicles take up less space, but take that space for longer.
While pedestrians and bicyclists use space as well, they use much less space. (See discussion of flux.) Sidewalks (footpaths) are often considered part of the adjacent private property, and are thus already paid for with property tax.
Land used for roads instead of development is not on the books for property taxes.
The revenue raised can be used for many transport purposes or redistributed back to taxpayers through some other means.
We expect the additional road rent reduces the effective land rent that landowners can charge. If people have to pay more for travel, they will pay less for real estate.
Rural areas have much lower, perhaps negligible, road rent. Though the number of users drops significantly (so there are fewer travelers who must pay the burden of road rent), the cost of land drops even more significantly.
Were there no (fewer) roads, land would also have very little (less) value, since it would be difficult to access and egress.
If roads were fully built on, views would be terrible and the existing buildings would diminish in value. But none of that is to say we have the correct amount of roads now. Clearly urban roads are undercharged in a real estate sense.
Using data from the Federal Chamber of Automotive Industries that was once freely available online, and is now behind a paywall, I have produced graphs illustrating the Australian vehicle market. The data show among the passenger cars: medium, small, light, and micro are all gaining in proportion of passenger cars, rising from half the passenger car market to 83% since 2000.
`Are Australian Vehicles Getting Bigger?’
The answer here is ‘Yes.’
As will be no surprise to Australians, or North Americans (See Canada data), the share of Sports Utility Vehicles has exploded since the beginning of the Millennium from about 13% to 39%, and now more SUVs are sold each year than passenger cars.
Now 50% of 70% is 35% (small cars share of all vehicles in 2000) while 80% of 38% is 30% (small cars share of all vehicles in 2017), so the share of small and medium cars of all vehicles is falling. But the total market of vehicles sold in Australia is still increasing from 787,000 in 2000 to 1,189,116 in 2017, and 30% of cars sold in 2017 is more than 35% of cars sold in 2000, so there are still more in terms of total number of small and medium cars sold in 2017 in total than 2000, even if it is a declining share of the market.
The Australian government also conducts a Motor Vehicle Census and just as the number of new cars sold each year rises with population growth, the total number of vehicles is also rising. This differs from the US, which has more or less peaked in cars per capita, and perhaps cars. I graphed this data for NSW for selected years (this data, is also, inconveniently, not in one place)
The reason for more SUVs vs. large cars are speculative. That is, why do people now prefer SUVs and not station wagons or big cars? It’s not as if people actually do a lot of off-road driving.
One is the idea of the extreme trip. Sometimes (say once a year or even once a month) a very large car would be useful. So instead of renting the specific vehicle when they want it, SUV-owners buy the vehicle they would use 1% of the their trips (or 0.05% of their time – since cars are only used 5% of the day anyway, and at rest the remainder, sleeping more than even cats), but which is too large 99.95% of the time.
One answer is the car Arms Race. In a taller car, the driver can see farther ahead (drivers are less likely to have their view obscured), which lets tall vehicle drivers anticipate better. It makes drivers feel safer, which they are for themselves, even when they are not for others.
More people are killed because of SUVs and light trucks, in the US, Michelle White estimated in 2004 “For each 1 million light trucks that replace cars, between 34 and 93 additional car occupants, pedestrians, bicyclists, or motorcyclists are killed per year, and the value of the lives lost is between $242 and $652 million per year.” Presumably the same logic holds in Australia.
Increasing the mass of vehicles on the road doesn’t do society any favours from an energy consumption, or air pollution perspective either. And of course, larger vehicles use more space, consuming more land in parking lots (which are now often restriped to accommodate more massive vehicles) and roads, where the width of lane consumed by larger cars rises, providing less manoeuvrability for other cars.
With the rise of autonomous vehicles, and especially vehicle sharing, the right sized vehicle will be summonable by app, so when travelers need the specific type of car for a large trip with many people, they can get it. The rest of the time, drivers will be able to use a car fit for purpose, one that holds one person for a one-person trip, and two people for two-person trips, and so on. This opens up the potential for skinny cars, enclosed electric cycles, and many other appropriate vehicles, which take up less road space, making it even easier to improve the environment for other road users, including walkers and bicyclists.
Passenger Motor Vehicles
Passenger vehicles are classified dependent on size, specification and average retail pricing. Selected vehicle types will be assessed on footprint defined as length (mm) x width (mm), rounded, as follows:
Sports Utility Vehicles
Vehicles classified as Sport Utility Vehicles (SUV) meet the FCAI criteria for classifying SUV vehicles based on a 2/4 door wagon body style and elevated ride height. Vehicles typically will feature some form of 4WD or AWD, however, where a 2WD variant of a model is available it will be included in the appropriate segment to that model.
Vehicles designed principally for commercial but may include designs intended for non-commercial applications.
Vehicles designed for exclusive heavy commercial application.
Hatch, sedan or wagon with a footprint < 6,300
Hatch, sedan or wagon with a footprint range 6,301 – 7,500
Hatch, sedan or wagon with a footprint range 7,501 – 8,300
Hatch, sedan or wagon with a footprint range 8,301 – 9,000
Hatch, sedan or wagon with a footprint range 9,001 – 9,500
Hatch, sedan or wagon with a footprint range 9,501 >
Wagon for passenger usage, seating capacity > 5 people
Car, coupe, convertible or roadster
3,501 – 8,000kg GVM
=> 8,001kg GVM & GCM < 39,001
8,001kg GVM & GCM > 39,000
Light Truck Sizes:
Light bus < 20 Seats
8+ seats, but less than 20 seats
Light Bus > 20 Seats
Vans/CC <= 2.5t
Blind/Window vans and Cab Chassis <= 2.5t GVM
Vans/CC > 2.5–3.5t
Blind/Window vans and Cab Chassis between 205 and 3.5 tonnes GVM
Pick-up / Chassis 4×2
Two driven wheels, normal control (bonnet), utility, cab chassis, one and a half cab and crew cab
Pick-up / Chassis 4×4
Four driven wheels, normal control (bonnet), utility, cab chassis, one and a half cab and crew cab
Abstract: This study evaluates routes followed by residents of the Minneapolis–St. Paul metropolitan area, as measured by the Global Positioning System (GPS) component of the 2010/11 Twin Cities Travel Behavior Inventory (TBI). It finds that most commuters used paths longer than the shortest path. This is in part a function of trip distance (+, longer distance trips deviate more), trip circuity (−, more circuitous trips deviate less), number of turns (+, trips with more turns per kilometer deviate more), age of driver (−, older drivers deviate less), employment status (+, part-time workers deviate more), flexibility in work hours (+, more flexibility deviate more), and household income (−, higher-income travelers deviate less). Some reasons for these findings are conjectured.
Author keywords: Global positioning system (GPS); Shortest path; Route choice; Wardrop’s principles; Travel behavior.
An edited version of this appeared on The Conversation June 11, 2018. The original is below.
Traffic signals are a source of great inequality in the urban realm, giving priority to motor vehicles over pedestrians. Cities and states say they want to encourage walking and biking for many reasons: it is space efficient, it has less environmental impact, it is healthier, it is safer for other travelers, and, since, it reduces the numbers of cars on the road, even motorists should be in favour of other people walking. To help achieve that, road management agencies should take the lead in reprioritising traffic signals by redistributing intersection delay from pedestrians to cars.
While planners tend to focus on the long-term decisions, like infrastructure and land development, it is the shortest of short-term decisions, how many seconds of green light each movement gets at an intersection, that shapes daily perception of the feasibility of walking or driving to a destination at a given time, and thus the choice of route, destination, and mode of travel. Traffic signal timing involves math, so it has been historically delegated to the engineers, but it also involves values and priorities, and so is the proper subject of public policy.
Since the early twentieth century dawn of what Peter Norton calls ‘Motordom’ in his book ‘Fighting Traffic‘, street space has steadily been regulated and enclosed, limiting the rights and privileges of pedestrians while promoting those of drivers as a class, in the name of safety and efficiency. But we should ask safety and efficiency for whom? Prior to traffic signals, pedestrians could and did cross the street whenever and wherever they wanted, before the term ‘jaywalking’ was invented and street crossing was regulated. The introduction of signals prioritised the movement of motor vehicles at the expense of pedestrians, whose effective walking speed through the city necessarily slowed. The consequences of making it easier to drive and harder to walk on people’s choice of mode is pretty straight-forward, and consistent with the rise of the automobile in the 20th century.
Pedestrians take longer to cross streets than cars because they move slower. As a result, the ‘don’t walk’ signal flashes before the light turns red for cars. But at many intersections it is worse than that. In Sydney, the traffic signal policy is set at many intersections to give less green time to pedestrians on a phase (from the time the light turns green to when it turns red, or from ‘walk’ to ‘don’t walk’) than to automobiles, to give autos a protected left turn without having to yield pedestrians. This guarantees the average pedestrian arriving randomly at the intersection waits longer than a random car.
The cycle length (time from the start of the green light to the start of the next green) tends to be longer at busier intersections (and busier times of day) as a longer cycle length reduces the number of phases per hour, and thus reduces the amount of lost time associated each phase, when the intersection is not being effectively used by any approach. Lost time can never be reclaimed, so one understands why engineers might want longer cycle lengths if the objective were moving cars.
However long cycle lengths particularly disadvantage pedestrians, who stand out in the open exposed to the elements and the tailpipe emissions of cars, motorcycles, trucks, and buses. Even more significantly, people systematically misperceive travel delay, so waiting at a traffic light feels even longer than it actually is.
First introduced in 1922 in New York City, traffic signal coordination aims to ensure vehicles arrive at the traffic signal when it is green, so they don’t have to stop. By correctly timing traffic signals in sequence, platoons of vehicles move together through a ‘green wave’. So let’s say the wave is set for a speed of 40 km/h. Then as long as a car accelerates from the first signal to 40 km/h, and maintains that speed, it should then hit the following lights on their green phase as well.
While this is relatively easy to maintain on a single road, it is more difficult on a network, especially a complex, asymmetric network. It also works against the idea of actuation, as interruptions to the pattern (extending or contracting phases) change the window in which cars can successfully hit a green light at a given speed. Of course, just because cars can make a green wave at a speed of 40 km/h doesn’t mean pedestrians will make a green wave unless they travel at exactly a divisor of 40 km/h (e.g. at exactly 5 km/h between intersections). This means that pedestrians will more likely wait at red lights at intersections timed for cars.
Actuation / Beg Buttons
While some signals are ‘fixed time’ which eases coordination at the expense of adapting to conditions, modern signals are ‘actuated’, that is, they respond by adjusting the phasing, and perhaps the cycle time, in response to the presence of vehicles. For vehicles, there is either a camera which detects their presence, or more commonly, a sensor in the road, often a magnetic loop. In either case, this is automatic for the car, and can detect cars upstream of the signal. This allows the signal to stay green longer for a phase if it detects a vehicle approaching, or turn red sooner when there are no vehicles. In contrast, for pedestrians, they are required to push a button to get a walk signal. If they arrive a second too late, they have to wait the entire cycle to get a walk signal. If there are many pedestrians, they don’t get a longer walk signal. Pushing the ‘beg button’ (so nicknamed as the pedestrian must request the signal) twice does not make it come faster or stay green longer. Ten, or a hundred, pedestrians do not make the ‘walk’ light come faster either. The beg button is often positioned out of the way, requiring the pedestrian to walk longer than would otherwise be required. A few seconds here, a few seconds there, add up.
There is a reason that traffic engineers don’t automatically allocate pedestrian phases. Suppose the car only warrants a six second phase but a pedestrian requires 18 seconds to cross the street at a 1 meter/second walking speed. Giving an automatic pedestrian phase will delay cars, even if the pedestrian is not there. And there is no sin worse than delaying a car. But it also guarantees a pedestrian who arrives just after the window to push the actuator passes will wait a full cycle.
The role of signal policy
It turns out that one of the world’s most widely deployed traffic signal control systems, the Sydney Coordinated Adaptive Traffic System (SCATS), was developed here in Australia. Just as Australia led in traffic control to more smoothly move cars, it should lead in pedestrian-oriented traffic control. There are a number of steps that those concerned about pedestrians should insist on. To start:
Pedestrian time must be considered (and prioritised) in the traffic signal timing algorithms so that their weight is equal to or higher than the weight of a passenger car.
Pedestrians should get the maximum feasible amount of green time on a phase, rather than the minimum, so that pedestrians arriving on the phase have a chance to take advantage of it, and slower moving pedestrians are not intimidated by cars.
Pedestrians should get a ‘leading interval’ so they can step into the street on a ‘walk’ signal before cars start to move on a green light, increasing their visibility to drivers.
Pedestrian phases should be automatic, even if no actuator is pushed. Instead, the actuator should make the pedestrian phase come sooner.
Many more intersections should have an all-pedestrian phase (what is referred to as a ‘Barnes Dance’) in addition to existing phases so pedestrians can make diagonal intersection crossings without having to wait twice.
There are numerous other steps as well that can improve the life of the pedestrian, and thus increase their number. Certainly we can demand more patience from drivers as well. The advent of the autonomous vehicles over the next few decades is unlikely, by itself, to eliminate the need for traffic control in cities. There will be places where the number of cars and people are such that they cannot efficiently organize themselves, and where other traffic controls, like stop signs or roundabouts, cannot be effectively implemented. But autonomous vehicles should help get more throughput out of intersections, losing less time than human drivers, and behaving far more safely.
According to Uber, emergency braking maneuvers are not enabled while the vehicle is under computer control, to reduce the potential for erratic vehicle behavior. The vehicle operator is relied on to intervene and take action. The system is not designed to alert the operator.