Shoddy Study Claims Light Rail Increased Congestion in Paris (Hoisted from Comments)

Jarrett points us to a just-published paper in World Transit Research that contends that Paris’s new T3 light rail line caused traffic congestion on the adjacent freeway, the Boulevard Périphérique, to increase, thereby causing a net increase in environmental damage and a negative social rate of return. Reading it at its original source requires academic access; here is a mirror on this blog, and thanks to ant6n for sending it. The study does not produce much evidence that an increase in traffic congestion indeed happened. As Angus Grieve-Smith explains in the comments on Human Transit:

It’s important to note that the authors did not measure traffic on the Périph. They just observed that average speeds on the highway declined from 45.9 km/h to 43.5 km/h, and that “many witnesses of the public hearing on the extension of the tramway to Porte de la Chapelle testified their fears to see an analogous shift increasing the congestion on Eastern Périphérique.” In other words, bullshit.

The fact is that a large portion of the traffic on the Périph is going from one side of the city to the other. If some of the drivers on the Maréchaux transfered to the Périph, increasing congestion there, some of the drivers on the Périph would take commuter trains across town instead. Some of the drivers would find it more convenient to take the metro instead of the tramway, or to drive an alternate route that doesn´t involve the Périph, possibly one of the parallel boulevards closer to the center of the city.

The study spends very little time arguing that an increase in traffic happened. It almost takes it for granted. The evidence it provides is that the average speed on the entire Périphérique went down 5%, from 45.9 to 43.5 km/h, whereas the average speed on the southern segment, which parallels the T3 line, went down 10%, from 37.9 to 33.9 km/h.

Instead of arguing that the reduction in speed represents extra traffic coming from the lanes removed to make room for the T3, the study assumes that 100% of the reduction in traffic on the Maréchaux, the boulevard on which the T3 runs, was transferred to the Périphérique. This is unlikely: the phenomenon of reduced demand is attested in the literature – see references here. Traffic shifts to less congested times of day, and sometimes disappears entirely as drivers choose not to take the trip. For one example, when the West Side Highway collapsed, about half its traffic disappeared; this percentage is high, presumably because Manhattan has good transit options, just like Paris.

It’s in fact worse than Angus says. Although the paper provides traffic counts on the Maréchaux, it provides no such counts for the Périphérique, although such counts should be very easy to find. Its computation of the traffic increase on the Périphérique comes entirely from prior assumptions about the traffic that disappeared from the Maréchaux. Another, more minor sleight of hand is the choice of years. For the Maréchaux, the paper argues for comparing present traffic to traffic in 2003, just before the tram’s construction began; for the Périphérique, the numbers provided use 2000 as a baseline.

Most of the paper’s effort is spent not on trying to prove that traffic increased, but on computing the social costs and benefits under questionable assumptions. Doing that is difficult to say the least without knowing more about the nature of traffic on the Périphérique, and the study makes even more questionable assumptions there. To be fair, the biggest smoking guns do not concern the social cost that according to the study is by far the highest, slower traffic speeds; those follow from the assumptions. Instead, they serve to showcase a careless and even biased thought process.

First, the difference in carbon emissions between free-flowing traffic at 38 km/h and 34 km/h is small; what causes fuel consumption to rise in traffic jams is not lower average speed but rather stop-and-go traffic. Thus, even a first-order estimate of extra fuel consumption is impossible given the study’s numbers and assumptions. Fortunately for the study, the carbon cost it uses is so low (€25/ton) and the overall effect posited not large enough that the overall magnitude posited is negligible.

Second, in its computation of economic costs, the study makes the following observation about the project’s cost:

Available information on the monetary costs associated with the project is scarce. One has only the ex ante costs envisioned in the official preliminary Public Inquiry: 341.8M€ for the initial investment and 43.9M€ for the exploitation of the tramway. Experience suggests that ex post costs are likely to be appreciably higher (Flyvbjerg et al. 2002).

For the record, it took me all of three minutes to search on Railway Gazette and Google and find ex post costs amounting to €311.5 million. Worse, the paper says it chooses to use the original cost estimate for lack of other numbers, but then multiplies the original budget by 1.3, the standard factor for public projects in France. As far as I can tell, the reason for multiplying budgets by 1.3 is to cushion against small budget overruns, which could turn slightly beneficial projects into net liabilities; it’s a more honest way of including a contingency budget. In other words, the paper claims that costs probably ran over but its cost estimate for net benefit purposes assumes they didn’t, while in reality they didn’t run over while the paper assumes they did.


  1. EngineerScotty

    Missed in all of this is the big double standard:

    * The time motorists spend in traffic is a matter of significant public concern.
    * The time transit users spend on their journeys, whether in-vehicle or waiting, is not.

    Switching from a mixed-traffic vehicle (in this case, bus) to a vehicle in an exclusive right-of-way (in this case, a train) generally represents a significant improvement in service quality to the transit user. However, that doesn’t appear to matter. Traffic congestion is a sign that roadways need to be expanded; but degraded transit service tends not to prompt the same response from officials.

  2. MobilMan

    The T3 is successful in terms of ridership but questions remain about travel speed which is 16.5-17.5 kph – less than the promised 20kph and not much higher than the 14.5kph of the bus. A (more expensive?) rebuild of the “Petite Ceinture” line for light rail would have avoided the speed and capacity issues.

    • EngineerScotty

      Even 20kp/h is pretty poor for rail with a separate ROW. 17kp/h is about as slow as the Portland MAX through downtown–and there, the train is running through a dense downtown grid on a surface alignment with stops every 500m, intersections every 75m, and extremely limited signal priority. The T3 runs in the median of the Maréchaux and appears to have far fewer crossings–is this simply a case of too many stops?

      • Alon Levy

        Local transit in Paris, and by imitation the rest of France, is really slow. The Métro (except Line 14) has average interstations of about 500 meters and an average speed of about 20 km/h; that’s why RATP chose to build the RER instead of to keep expanding the Métro out.

      • Miles Bader

        Surely the stop-spacing of an existing system doesn’t need to constrain the stop-spacing of extensions to that system (especially if the extensions are in areas with a different character, where wider spacing might seem natural)…

        • Alon Levy

          The narrow stop spacing extends throughout the legacy system; it doesn’t matter much if the trains have wide stop spacing and can average 45 km/h in the suburbs if they slow down to an average of 20 km/h once they cross city limits. The RER A’s central section, which closely parallels Line 1, functions as an express bypass. Between Nation and Etoile, the RER A makes 4 stops; Line 1 makes 14, all of which opened in 1900. RATP made the right choice to first construct the RER A to connect to La Défense and only later extend Line 1 there, after the RER A had become saturated.

          • michaelrjames

            I admit to being a late developer so I couldn’t resist making a comment here almost exactly ten years, give or take a few weeks, after this article first appeared (which I just stumbled across). And ten internet years is like a century in normal years … it makes me the Jeanne Calmentof PO! Showing lamentable lack of longevity none of the other commenters here have persisted to today. (Other than the author who is a mere callow youth:-)

            it doesn’t matter much if the trains have wide stop spacing and can average 45 km/h in the suburbs if they slow down to an average of 20 km/h once they cross city limits

            Technically you are correct but of course it does make a difference, especially for those travelling from anywhere intramuros to the Petite Couronne. Actually it also makes a difference to those using the lines only intramuros in that it affects service throughout the system (or at least affects the resources needed to maintain the equivalent service prior to the suburban extension). There is a limit to how much Metro lines can be extended due to these factors. I’ve mentioned it here previously (on much later articles) that for my first year in Paris I had to use a bus to get to my workplace at the Cancer research hospital complex at Villejuif about 5km south of Paris. It was tedious and exacerbated by my dislike of buses. However the extension of M7 from Maison-Blanche* to Villejuif-Louis-Aragon was completed in that first year and it transformed my commute and also my willingness to use the city more after work etc (opposite to the way a multi-seat bus transit saps your will). The 4 ‘new’ stations to Villejuif are spaced about 1 to 1.8km compared to an average of 600m for the whole line, and it is the only part where the trains actually get up to their maximum speed; and I don’t know if it is just an illusion created by the unusual speed but it seems one is racing downhill.

            A final observation is that while M14, built in the 90s, has much longer interstation spacing of av. 1.13km–and thus traverses Paris much faster–and is why this line is being extended significantly at both its ends; in fact to connect to orbital M16 at Villejuif-IGR (Institut Gustave Roussy, ie. the Cancer centre which is just a bit too distant from M7, about 1.5km). This shows that the thing that stops extension of older Metro lines deep into the suburbs is the density of stations within Paris, not the extension itself. Given the sheer population and destination density in intramuros Paris it wouldn’t necessarily desirable to reduce the interstation distances of the old lines, and of course it was only feasible for M14 because most of its route (intramuros) was already well served by historic Metro.
            *apparently the extension to Kremlin-Bicetre already existed but the southbound Villejuif line branches from the south-east Ivry line at Maison-Blanche which will also be the branch point for M14.

  3. Dist

    Scotty, what you are describing is the T3.

    – (very) dense urban area
    – dense street grid
    – tram stop every 500m

    Plus you have to add the fact that the T3 have to intersect with heavily trafficked arteries that give access from the suburbs (and the Périphérique) to the inner Paris.

    17kph is slow but already better than the previous buses that were averaging 15kph if I remember correctly. I think that it’s actually the line engineers (or operators if you prefer) through one of their union that refuses to raise the speed on the line. They find it too dangerous. And I also don’t know if it will be possible to make the tram run faster.

  4. Pingback: Cost Overruns: How I Learned to Stop Worrying and Hate Bent Flyvbjerg | Pedestrian Observations

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.