The website h2020 has a very interesting examination of “traffic evaporation”-what happens to all that traffic when you close urban expressways in cities?  The Paris Region Planning and Development Agency (IAU Île-de-France) has been examining the impact of these closures in large cities around the world, and has come up with some startling conclusions.
Despite the initial fears, the removal of fast lanes does not worsen traffic conditions beyond the initial adjustments,” explains Paul Lecroart, urban planner at IAU and a specialist on this issue. “In all the cities studied, the evaporation of traffic is an important element to observe”.  Here is what is interesting-when a fast lane is removed, overall traffic decreases by 14 per cent after several months. Why? “The reasons that lead to traffic evaporation can be several, one of them is the so-called “induced traffic“: when you create a fast lane, you automatically create traffic.” 
Another study in 1992 by the Ministry of Transport in France estimated that the creation of a French motorway increased car volume by 40 per cent. Take the motorway away, and in the long term, traffic decreases. “The reduction of traffic is mainly due to behavioural change: people start adapting to the new spatial configuration. The behavioural changes that brings ‘traffic evaporation’ are: change of itinerary and of schedules, the frequency of travel, the mode of transport (shifting from cars to two-wheeled vehicles, bicycle, etc.), but also car-pooling, new family organization, moving or working remotely.”

The French term of  “traffic evaporation” is related to the Braess paradox which states” that adding extra capacity to a network may reduce overall performance and increase travel times. As in a game structure, if drivers have the possibility to choose their own route autonomously they will behave selfishly. This means that each driver will aim at improving its respective travel time by arriving first: all drivers will take the new “fast” road and will thus cause congestion.”

Take away the “fast lane” and you reduce that congestion, as seen in Paris’ work reclaiming the right bank of the Seine for pedestrians and cyclists which you can view here.  A YouTube explanation of the Braess  paradox is linked below.


  1. The phenomenon of disappearing traffic was recognized just as its counterpart — induced traffic — became a measureable element. Jeff Kenworthy, the Australian transportation planner, has written extensively on this and other related issues.
    From ‘The Ten Myths of Automobile Dependence’:
    The obvious response to the failure of freeways to
    cope with traffic congestion is to suggest that still
    further roads are urgently needed. The new roads are
    then justified again on technical grounds in terms of
    time, fuel and other perceived savings to the
    community from eliminating the congestion. This sets
    in motion a vicious circle or self-fulfilling prophecy of
    congestion, road building, sprawl, congestion and more
    road building. Automobile dependence is inevitable in
    such traffic engineering.
    Awareness of this phenomenon, now called induced
    or generated traffic, is increasingly common in the
    literature. […]
    There has developed an alternative to this kind of
    road planning treadmill which is comprehensive land
    use/transportation planning that develops
    alternative transportation systems and different land
    use patterns aimed at minimising unnecessary
    movement. The comprehensive plan is a much more
    community-based project that invites a city to
    envision its future and then seeks to find the
    appropriate infrastructure. This process requires a
    much more creative role from planners and engineers
    who need to provide the land use and transportation
    mix most able to meet the complex needs of the
    community. […] There is also clear evidence that if road
    capacity is removed then a high proportion of traffic
    just disappears; this ‘traffic evaporation’ or ‘traffic
    de-generation’ also gives another tool to cities
    struggling with how to manage their future

    1. Firstly, this Braess’s example assumes a scenario where drivers are essentially driving down the same path and follow each other like lemmings through a new road link. This will rarely happen in real life unless they are in a parade. Anyone can travel along a network longer than one should because of misinformation or perception.
      Secondly, it is obvious that adding another road segment (as in the case of the simple network example always used) will create an intersection, and intersections increase travel time. And roads are not always added to reduce travel times, but to provide access to places people want to go. So travel times will obviously and intentionally go up. This follows the “law of options” in that with additional options, the decision time goes up. That’s why in John Nash’s “blonde” example, less options can be more efficient.
      So yes, adding additional capacity can increase travel times, but it’s not like adding additional capacity will *permanently* increase travel times as many people misinterpret Braess’s example into real life. Braess’s paradox could happen, but it would not last long as traffic flows change like the waves of the sea, and does not stay still like a morning lake for very long. You may get instances of Brass’s but very briefly.
      Braess’s paradox is less likely to happen when traffic volumes are sufficiently low or high. But it also assume somewhat of a constant flow of traffic and homogeneous driver behaviour (human psychology being the main factor–even IAU Île-de-France admits to this). We know traffic flow varies quite often within even 15 minute time periods, and the type of drivers and their behaviours, based on their experience, driving style and preferences (including aggressiveness)–and the fact that now we have devices and in-vehicle navigation systems (of which some use real-time info)–all reduce the likelihood of a Braess’s “Nash equilibrium” moment (and this all assumes everyone acts selfishly i.e. tragedy of the commons, and everyone has uniform information/perception about traffic conditions which is almost impossible until we go completely autonomous and connected, which itself will have an interesting effect to this theory).
      Also, Braess’s analysis is based on simple networks and single value of flows, so it does not include non-linear traffic flows (which is unrealistic) nor does it include the complexities of traffic weaving and people’s propensity to travel along the same routes (not like water which will flow in a pipe system generally at equal pressure regulated by the physics of fluid dynamics). An example are some roads in Central LA neighbourhoods which are avoided for perceptual reasons.
      We have seen local cases where Oak St bridge was advertised to be under repair with a reduction of 1/2 the lanes and people become like lemmings and avoid that bridge to only clog up the alternate crossings (e.g. Knight St Bridge). Once they find (via news reports) Oak St Bridge, even at 1/2 capacity, was flowing freely for the few that risked crossing it, did they then revert in hordes back to that bridge, allowing the alternate crossings to breath the next day. After a few days, everything settles down into equilibrium and you get relatively equal (if not, perceived) travel times. So examples of Braess paradox may be due more to communications than simply infrastructure changes. Places where people “observed” Braess also happened to coincide on Earth days and massive marketing/communications campaigns, as well as the provision of alternate options like increased transit.
      And as the Brasses’ paradox applies to general networks, it is not just for road applications, but has also been studied in communications/internet networks, and can also be applied to transit and cycling networks. So then to accept road capacity increases will increase travel times, that would mean one accepts adding transit capacity will increase ridership times, etc. So does that mean the new Evergreen Line increased transit times through Port Moody? Even if it did, we know the policy purpose for the line is more than travel time but also about mode shift and sustainable development.
      And does Braess’s Paradox explain how traffic speeds went up when we had the 2001 bus strike? It’s simply the “addition” of lanes due to no buses stopping on parking lanes (that are restricted during peak directions) and no interactions of buses pulling back into traffic flows. People still got to work, just found other means (again, these numerous factors are beyond the scope of Braess and demonstrate the simplistic view of reality it considers)
      You have to also look at the fact that the past transport system was one with very little real-time information and with that selfish (person-optimal) travellers will cause inefficiencies and periodically, phenomena such as Braess. But with the introduction of real-time information (i.e. google maps) to a growing percentage of travellers, the system is becoming more “system-optimal” and eventually with full blown ITS (i.e. autonomous and connected travelling across most modes) it will become about as system-optimal as it can get, which will reduce most of the “inefficient” human psychology from the system.
      For those who want to still bash road building, a more common situation where increased capacity can cause delays is not necessarily network additions as exampled by Braess, but merely lane-capacity additions when you increase lanes say on a highway. Going from say 2 lanes to 4 lanes increases the lane options and subsequent jockeying for lane position (assumes people compete for less congested lanes). This is called “weaving” in traffic engineering and this causes “friction” or “turbulence” in flows especially near interchanges where people want to exit or enter the highway. Adding HOV lanes (or what Nash would call “blondes”) down the centre of the roadway also can cause more congestion as they are usually required to travel on their exclusive lanes until hundreds of metres before their exit points, and are therefore required to force their way across all lanes causing all kinds of slowdowns, safety issues, and ironically, reduced capacity. Adding lanes can be a practice of diminishing returns as long as humans are behind the wheel. That is until we have driverless vehicles en-mass when we will see an increase in the ROI for our roads (which we will need to build less of as the theory goes).
      Sorry for the lengthy reply but I hope you don’t think you can understand this concept and think it represents reality by watching a 3 min. Youtube video…

    1. #1: Cato Institute. Not what you’d call agenda-less; although the same can be said of many PT readers, to be fair. #2: the article you linked doen’t debunk induced demand. The author admits as such. It instead disputes one researcher’s flippant description of his findings relevant to induced demand – and presumes by extension to debunk the whole notion. It’s no different than “disbunking” global warming because one researcher in 2006 claimed the Earth’s average temperature would rise by 0.9 degrees by 2018 when in fact it had only raised 0.7. More bias disguised as rigor from the world’s biggest cowards: Libertarians.

    2. An excerpt from Todd Litman’s ‘Generated Traffic and Induced Travel, Implications for Transport Planning’, January 2017:
      >pi>Ignoring generated traffic results in self-fulfilling predict and provide planning: Planners extrapolate traffic growth rates to predict that congestion will reach gridlock unless capacity expands. Adding capacity generates traffic, which leads to renewed congestion with higher traffic volumes, and more automobile oriented transport and land use patterns. This cycle continues until road capacity expansion costs become unacceptable.
      The amount of traffic generated depends on specific conditions. Expanding highly congested roads with considerable latent demand tends to generate significant amounts of traffic, providing only temporary congestion reductions.
      Generated traffic does not mean that roadway expansion provides no benefits and should never be implemented. However, ignoring generated traffic results in inaccurate forecasts of impacts and benefits. Road projects considered cost effective by conventional analysis may actually provide little long-term benefit to motorists and make society overall worse off due to generated traffic. Other strategies may be better overall. Another implication is that highway capacity expansion projects should incorporate strategies to avoid increasing external costs, such as more stringent vehicle emission regulations to avoid increasing pollution and land use regulations to limit sprawl.
      There are 10 ½ pages of references.

      1. Adding road capacity does not generate new traffic as if it’s some reactive law. Adding new land uses (options for people to originate from and go to–and again giving people more options creates Braess problems) and more people can generate new traffic. Adding capacity can “induce” new car trips if prior to that there was “latent” or pent up demand to travel by car. But this is not always the case because at the end it comes down to a perceived utility or cost if you will. Why did the new Port Mann experience a decrease in traffic when its physical capacity was effectively doubled? Because tolls decreased the effective capacity (which is influenced by perceived costs). So it is not that simple and this “rule of thumb” in which increasing capacity creates a knee-jerk increase in net auto usage. There is also a temporal component most people who argue about this forget. Also to study this, you need to consider more of a screenline scope vs. counting cars on that one roadway that was widened (so this adds a spatial component).
        Travelling on surface networks is a means to an end and people do not travel simply because there are more roads. If more roads reduces the cost or utility to travel and this is lower than the latent demand to travel (if there is any), then you will induce more travel. And we know latent demand doesn’t occur 24/7 because if it did, people would rush out and drive in the early morning to take advantage of all that road capacity available just for them!
        But the academic question is if this latent demand exists and if so how much is there? And this is for the demand of a service that is not really desired (commuting) compared to say the latent demand for Canucks playoff tickets (ok bad example given their current standings). For most trips, if people have to travel, they usually have to get somewhere at a given time. Sure they can go earlier and avoid the rush, but that is also a cost in terms of perceived value of time. At some “optimal utility” point they go. In this case, this trips was not induced because it was planned. Less congestion through more road capacity could have affected their timing of their trip. If anything, from a time-perspective it could have been induced from a future hour, but the trip was predetermined to be made regardless. So you have temporal, spatial, modal, and economic factors that complicate our understanding of latent demand, which is critical to understand and suppose if any *new* net demand was induced.

  2. Yes as explained by Clark Lim, the “Traffic evaporation” theory and the Braess Paradox are 2 different things.
    the Braess paradox is explained at constant traffic (or demand)
    In the Parisian example (and as in the Seoul’s Cheonggye example , the closure of an expressway leads to “traffic evaporation”, (overall traffic decrease or demand)
    that is the premise that people do same travel choice (going from A to B) disregarding of the offer is wrong…. so as observed time after time, building more road induce more traffic unless a new barrier such as toll encourage people to do different choice:
    Tolls doesn’t decrease the effective capacity of a road, but impact heavily the perceived costs of the trip, what in turn impact heavily the demand for such trip.

Leave a Reply

Your email address will not be published. Required fields are marked *