Wednesday, January 7, 2015

World's worst investment: why urban highways destroy wealth

Infrastructure investment is often seen as a good way to stimulate the economy and to provide facilities to allow the economy to grow. It's a pretty uncontroversial responsibility of government to provide for such infrastructure. However, even as a left-winger, I know that we have to be careful with government investment and make sure the money spent is spent effectively. Since government has a huge ability to spend and subsidize, it can effectively maintain unprofitable, wasteful ventures almost forever, taxing the rest of society to subsidize them, creating a drag on the economy. In the private sector, this is much less likely as actors do not have the financial capacity of government and so trying to support an unprofitable venture will inevitably bankrupt them, forcing the venture to halt.

Now, sometimes unprofitable ventures still have social benefits or moral imperatives that justify them. So I'm not condemning every government spending program here. But clear-minded reflection and analysis is necessary to make sure money is well-spent.

Which brings me to the matter of urban highways, which is the single most glaring example of an infrastructure spending that is ill-advised, wasteful and yields no positive outcome for society. They are largely the single most wasteful infrastructure spending project that we nonetheless keep doing over and over.

An example: highway vs subway, case of Montréal

The 1960s in Montréal were a time of big public projects, like in most of North America. In 1967, Montréal was slated to receive the International Exposition, and in order to welcome the world, projects that had been on ice for years finally got funding from governments now afraid to look like incompetent buffoons in front of an international audience. Consequently, two major transport projects were launched and done in the same lapse of time. The first was the Décarie Highway, a trench highway built along a previous boulevard, minimizing expropriations, and the second was the first 25 km of the Montréal Metro. 

This provides a good comparison of the costs and capacities of highways versus rail-based transit (though Montréal's metro is on tyres), as in both cases, the projects would be built underground, one in a trench, the other in tunnels. Too often, people claim that highways are cheaper than transit, or at least act like they are, but they do so comparing highways built on the surface in suburbs or fields with underground subways, which is an apple-to-oranges comparison.

Without further ado, let me present the Décarie Highway:

Décarie highway, 3 lanes per direction in a trench located in a densely inhabited section of Montréal

Décarie highway on a map of Montréal, connecting the A-40 to the north to A-20 to the south
Now, some numbers:
  • The highway is 6,4 km long
  • With 3 lanes per direction, it offers a passenger capacity of roughly 7 000 to 8 000 people per hour per direction
  • Its cost was 370 million 1967 Canadian dollars, or about 60 million dollars per km
  • In 2014 dollars, the cost would be 2,6 billion dollars, or 405 million dollars per km
  • If the highway had to be funded entirely through tolls, the toll would likely have had to have been around 35 cents per km, or around 60 cents per mile (in 2014 dollars).
 At the same time, the initial Metro network was built, opening also for the 1967 Expo:
The original three lines of the Montréal Metro, opened for the '67 Expo, versus the current system
The system was built entirely underground. Even if we wanted to take it on the surface, we likely couldn't due to the rubber tyres and the harsh winter (Sapporo does it but by covering the surface lines entirely).

Some numbers for the Metro:
  • The original network was 25 km long, including a tunnel under the St. Lawrence River.
  • Each train has up to 9 16-meter-long wagons (54 feet) with a capacity of 100 people per wagon, 150 in crush conditions, and a frequency of up to one train per 3 minutes. That is a capacity between 18 000 and 29 000 people per hour per direction.
  • This original network cost 214 million dollars in 1967 Canadian currency, or 8,5 million dollars per km
  • In 2014 terms, that is 1,5 billion dollars, or about 60 million dollars per km
Just for fun, let's mention the Expo Express, to my knowledge the first fully-automatic Heavy Rail Transit line in North America that ran on a line 5,7-km-long, built for 18 million dollars (125 millions in 2014 dollars), with a capacity of around 20 000 people per hour per direction. Unfortunately, after the Expo, the line went from nowhere to nowhere, and so the service was discontinued in 1972.

So, to sum up...

  • The highway was 7 times more expensive per km than the subway to build
  • The highway passenger capacity was 3 to 4 times less than the subway
  • So overall, the highway is 20 to 30 times more expensive to build per capacity-km
So what does that say about what is the better investment? Sure, some people can point out the highway is also used for freight, but with the passenger capacity of subways, this allows freight an easier time to navigate surface streets.

The reasons for the higher costs of highways are simple: they take more space, thus when land is expensive, they will be more expensive to build, likewise when you need to build elevated structures or to dig in the ground. Otherwise, laying tracks or building highway-grade pavement is roughly the same price. Subways also include the cost of wagons, of electrification and of maintenance centers too, which makes highways even more expensive because these costs have to be paid by its users and not accounted for when you build them.

OK, it was more expensive, but is it worth it? Did it create wealth?

Something being more expensive doesn't necessarily make it a bad investment. A pickup truck is more expensive than a compact car, but to some people, it may be a better investment if they need to haul a lot of stuff for their work. So, do urban highways help create exponentially more wealth than transit?

First of all, please note that I speak of URBAN highways, highways built in inhabited areas and that cut across them in order to facilitate commuting and movements inside the urbanized area, and not highways between cities that allow movements BETWEEN cities to be done faster. These are two completely different beasts.

Anyway, urban highways have a terrible legacy regarding wealth in the cities they have been built in. In almost every case, the areas adjacent to highways have become blighted by their presence. Highways achieve this because:
  • They cut off parts of the city from the rest of it, making it harder to get on the other side for residents, which makes living in the area less desirable.
  • They create a lot of nuisance through increased traffic and visual and aerial pollution, making the areas even less desirable.
  • They suck out traffic from existing healthy commercial arterials, which eliminates pass-by trips to the businesses on these arterials, weakening them and even leading to their closure.
  • As people require cars to use the highway, any business or service that is not in a car-oriented area will not benefit from them, so older urban areas with few parking spots cannot cater to highway users. Buildings have to be destroyed to make way for parking lots in order to attract highway users, but doing so eliminates a lot of business and services, reducing the area's vitality and the amount of wealth in them.
  • Since the highway is so fast, proximity matters less, and as greenfield developments are always cheaper than urban infill development or redevelopment, it makes housing and businesses in sprawl cheaper and more convenient in comparison to urban development, so development money flees to undeveloped areas, leading to under-investment in existing areas and their decline.
Elevated and underground highways can reduce the impact, but they are extremely expensive to build, as the Décarie Highway clearly demonstrates. 

An elevated highway in Nagoya, tall and narrow to minimize impacts on the area... but with a toll around 1$ per mile traveled
Urban highways do not create wealth, they rather destroy the wealth of the areas they're built in. However, for short-sighted suburban mayors and officials, they bring development to their suburbs as they favor greenfield development. Yet, this "development" needs to be called what it actually is, "displacement". Every new mall in the suburbs tends to be balanced by dying malls or commercial areas in the city.

On a macro level, there is no link between the amount of highway lanes per capita and the wealth of metro areas. San Francisco is very rich yet has next to no urban highways. Vancouver famously has no highway within the limits of the core city and is doing extremely well.

The richest metropolitan areas in the US are San José, San Francisco, Seattle, Boston, Washington, Houston, New York, Portland, Hartford and Salt Lake City. Areas with relatively few urban highways are overrepresented in that list, with Houston being the one big exception.

Meanwhile, transit is well-known for generating plenty of development around it and revitalizing areas it is built in. This is actually hurt by zoning in much of North America, but the few places that have embraced it like Arlington in Virginia or Vancouver seem to be positively booming.


So overall, urban highways, which are supremely expensive compared to alternatives for urban passenger trips, do not seem particularly efficient at creating wealth for society at large, they even seem to actually hurt wealth creation. Worse, highways help create a donut effect where the old neighborhoods at the center of a metropolitan area decline and become blighted as wealth flees to the periphery.

So as far as public infrastructure investments are concerned, urban highways are certainly one of the worst ones. They are expensive and yield very little benefit for society overall. However, they do benefit a certain group of people who naturally gravitate to suburbs, at the expense of city residents. Fairness dictates then that highway users have to shoulder the cost of them rather than highway costs being collectivized regardless of use. Tolling urban highways should therefore be the preferred way ahead for cities stuck with them.

An even better way forward for highways in North America would be to give highway construction and maintenance over to public State corporations rather than highways being directly built and maintained by DOTs and ministries of Transport. The State corporation could rely on the government's line of credit to borrow funds but would be asked to self-fund itself through tolls or other revenues. That way, highways would be built only where drivers would be willing to pay for their full cost. If they do not, then highways would not be built. If alternatives exist to do the job for less, they would stand an higher chance of seeing the light of day.


  1. "this allows freight an easier time to navigate surface streets" <- Static demand fallacy, beloved of transit boosters too.

    Dedicated space (below or above ground) for more efficient modes (transit/bike/ped) boosts aggregate freedom of movement (relative to car-dedicated/dominated space) but is independent of congestion levels throughout the day and year.

    1. Personally, I believe that induced demand is more about speed than capacity.I wrote a post about the subject, why packed street grids that have much higher capacity than highways do not have to be feared.

      The point is that the main benefit of roads and streets for car users is speed, not capacity. So speed induces demand, not capacity. But when demand for that speed benefit is greater than supply of space on the roads, you have congestion which reduces speed. Increasing capacity on highways does induce demand, but not in and of itself, but rather by allowing speed, limited by congestion, to return to free flow levels.

      A subway or other rapid transit that can be speed-competitive with cars, especially if it's also cheaper (including parking costs), can thus largely prevent the formation of congestion on streets. Induced demand will tend to gravitate towards transit rather than towards streets. Of course, this is valid only in the corridor of those rapid transit lines, however, these also happen to be, generally speaking, the corridors where congestion is most likely to emerge.

  2. I'm trying to plan an event for 2015 and want to get in contact with you. Could you email me at ? Thanks!

  3. I didn't realize the metro was so cheap... it explains why many of these rapid transit systems were built around that time and expanded little since. $60 million per km for a subway would be amazing today, the Spadina subway extension cost about $300 million per km, so 5x more even after adjusting for inflation.

    Has the cost of building urban freeways also increased comparably?

    Seattle's Alaskan Way Viaduct replacement tunnel seems to have also cost 3-4x more per km even after adjusting for inflation. Big Dig in Boston is considerably worse.

    Are these higher costs due to tunnelling vs cut and cover (or just cut in Decarie's case)?

    1. Montréal's subway is quite narrow so they get away with a single tunnel (most subways require 2 tunnels), and Montréal's ground is easy to tunnel through. The latest extension to Laval came in at roughly 150 million dollars per km and that was the early 2000s. Construction costs of public works seem to have increased far beyond inflation all over the world, maybe PPP would explain it better as labor has gotten more expensive but maybe techniques haven't kept up with labor reduction (or maybe union rules still impose the same number of workers for the same job).

      I do think urban freeways have seen the same cost increase. Even at-grade freeways in the suburbs in Montréal now cost 50-60 million dollars per km (case in point A-19 in the northern suburbs). In Detroit, a widening plan for a 6,7 mile stretch of the I-94 is expected to cost at least 2,7 billion dollars. That is 400 million dollars per mile for one lane per direction more. In Seattle, the Alaskan Viaduct tunnel, 2-mile long is expected to cost 4,25 billion dollars, that's 2,125 billion per mile, a bit over 1,4 billion dollar per km. Making subways look positively cheap in comparison.

      I don't know of any modern cut-and-cover freeways being built in cities in recent times, but I'd be curious too to know how much they would cost. The closest thing we get is freeway widenings, which are plenty expensive as well. We also get interchange reconstructions which are often disastrously expensive, the Turcot interchange will cost 3,7 billion dollars according to recent estimates, in Milwaukee, the Zoo interchange alone will cost 1,7 billion dollars to rebuild.

      As an engineer, I know that the idea that we are ever "finished paying" for infrastructure is BS. When we have reimbursed the original loan for the infrastructure, we have to borrow a new loan in order to rebuild the infrastructure.

  4. Have any analyses been done that compare the relative changes in value when an urban freeway is built depending on use type? I know that the auto-oriented commercial uses like drive-through fast food restaurants, gas stations, etc. are pretty low-value, but at the same time they occupy some of the best sites for car access at highway interchanges. Also, the residential neighborhood that's been decimated by a highway and traffic could become more valuable as a commercial corridor.

    On a more broad level, while these highways tend to destroy land value near them, and especially in cities, they also increase value in suburbs (the notion of displacement that was mentioned) once outside the noise-shed of the highway itself. It might look like a significant decrease in value over a small relative area of the city translates to a very large area of somewhat increased value out on the periphery. That seems to fit with the planning notions that were in place when these highways were being proposed in the first place.

    Might there even be some increased value in commercial central business districts due to good highway access? I suspect the answer is that it might be a sort of hidden form of displacement, where higher value more dense development is replaced with lower density but larger buildings that have more space dedicated to parking. It may be more expensive to build, and appear to be of a higher value, but it's still a loss on a per acre or per occupant kind of basis.

    1. Increasing value in the suburbs is a poor trade because the low density reduces the tax revenue per unit of infrastructure.
      In the city the higher density gives a lot of taxpayers supporting fairly short stretches of infrastructure, so that even fairly low levels of taxation pay for maintenance and repair with a surplus. In the suburbs the ow number of taxpayers per unit of infrastructure doesn't provide enough revenue to pay for maintenance, usually costing a dollar for every 50 cents of revenue, even with relatively high taxes.
      Checkout Strong Towns for an engineer's look at sustainable infrastructure design.

  5. The truth of your premise depends a lot on how generalizable your figures are. Harris County, Texas is wrapping up construction on a 61 km stretch of the Grand Parkway that cost $1.1 billion—about a third the inflation adjusted cost/km of the Montreal Metro, according to your figures above. Houston's Green and Purple light rail lines, on the other hand (underground rail is not practical in the Houston due to the water table) cost $80 million/km, or 25% more than the Montreal Metro.

    1. You're comparing apples to oranges here, exactly the type of comparison I warned against of comparing at-grade rural highways built in greenfields with underground urban infrastructure.

      First, the price figure presented here is only the inflation-adjusted price for the subway, infrastructure was much less expensive back then. The most recent subway extension cost 150 millions $ per km. No trench highway has been built in decades around here, so I don't have an estimate for the construction cost of one, but I think the ratio would be the same and it would probably be 1 billion per kilometer or so (certainly the Big Dig bears this out).

      I've checked the Green Parkway and it is being built on flatlands, with nothing near them. Land is cheap, work is easy to do and it will mostly be at-grade, being on viaducts for interchanges (about one per mile). Of course it's affordable. Building a heavy rail line in the same context would not be more expensive (these are usually called "commuter rail" when in rural areas), maybe even cheaper, and commuter rail can essentially be surface metros... In Japan, many commuter rail lines in Tokyo just go underground and serve as metros in Tokyo while passing through it.

      Here's another example: the French TGV (high-speed rail) costs 16 million euros per kilometer, around 24 million Canadian dollars (at current exchange rate). Why don't we just build TGVs all over cities if it's so affordable? Because their low price per kilometer is due to their location in rural areas, build them in cities and their price explodes. Same thing for highways, in fact, it's worse for highways because they need more space.

      As to Houston's LRT lines, again, these are in the middle of urban areas, where work is difficult and land is expensive.

      That's why I compared the two projects in Montréal, because both were done at roughly the same time and have the same installation (underground), making the comparison an apple-to-apple one.

  6. Roads have a qualitative advantage over mass transit that vanishes only in high density cities: mass transit cannot get most passengers from points A to points B, but only from points A' to points B', often at huge inconvenience to the user. The value of time spent stick on congested roads, waiting for trains, walking between transit stops and desired points, and so on, plus the longer waits during off-peak times, etcetera -- commute tone matters a great deal. There are other costs too. For example, in New York there's parking cost for people who want to drive (and the cost of having too have these parking garages). In many cities in the North Americas mass transit outside smallish downtown areas makes little sense to the people who would have to use it.

    But i agree that roads have significant qualitative costs that are often not evident until after they are built. The cross-Bronx expressway is a prime example of a road that caused a great deal of destruction of value.

    What i like a lot is ridesharing: it reduces the needed for parking and decreases the inconvenience to mass transit users of having to get from points A' to points B.

    1. There is a chicken-and-egg problem here. Mass transit is inconvenient to many users in North America because the network is not well-developed. This translates to low demand and usage of mass transit. The low demand translates to conclusion of mass transit being expensive in cost benefit studies. Demand for mass transit would increase significantly if the commute does not take three hours each way due to poor coverage.

      Cost burden is another problem of urban highway when it is owned/funded by municipalities. These highways often serve suburb commuters. The suburbs are often located outside the municipality owning the highway. Yet, the suburbs do not share the cost of building, expanding, and maintaining the highway. The Gardiner Expressway and Don Valley Parkway in Toronto are prime examples.