For Some, Uber’s Role in Congesting Manhattan Still Hidden in Plain Sight

Graph: I Quant NY
Graph: I Quant NY

Maybe it’s the sleekness. Or the digital disruptiveness. Or something slipped in the water bottle on your seat. Whatever it is, some data mavens are contorting into pretzels to deny the obvious — that Uber is contributing to the slowdown in Manhattan traffic.

The latest Don’t Blame Uber entry was a New Yorker post over the weekend by Ben Wellington, a Pratt statistics prof and mainstay of the “Open Data Movement” via his I Quant NY website. “Uber Isn’t Causing New York City’s Traffic Slowdown,” screamed the headline, though curiously, Wellington didn’t quite write that.

Here’s the gist of his post:

At the start of 2013, cabs were getting faster by about 0.0015 miles per hour per day. By mid-2014, they were getting slower by about 0.0013 miles per hour per day — or about one mile per hour every two years. In other words, every day, cabs were getting slower less quickly than they had the previous day, even as Uber was expanding its fleet. This is the opposite of what we would expect if for-hire vehicles were the main force behind falling traffic speeds.

Let’s unpack that.

First, mining the Taxi and Limousine Commission’s humongous database of all yellow taxicab trips from 2011 through last month — the period in which yellows have been GPS-capable — Wellington computed daily and monthly average speeds for the entire fleet (excluding ultra-long trips that might skew the averages). As the graph shows, average taxi speeds climbed from 2011 before peaking in mid-2013 and heading south. This dovetails with City Hall’s insistence that traffic in the Manhattan core has been worsening, although the drop in Wellington’s graph is less severe than the city’s figures.

So far, so good. Wellington then zeroed in on the rates of change in speeds and produced the quoted passage above. Since use of Uber really took off only around two years ago, it makes more sense to focus on the most recent 24 months. Eyeballing Wellington’s graph, average cab speeds fell an estimated 0.48 miles an hour from mid-2013 to mid-2014 and another 0.32 mph from mid-2014 to mid-2015. In effect, the decline in speeds shrank by a third. Because Uber’s presence on NYC streets has been accelerating, he reasoned, the slowdown should have been getting more pronounced, not less.

Since it didn’t, something other than Uber must have caused the slowdown.

Q.E.D.? No. For a host of reasons:

  • No one is blaming Uber for all or even most of the slowdown in Manhattan traffic. In a recent post I charged it with around 40 percent. The factors making up the remaining 60 percent could have been more potent in 2013 and early 2014 and less so later.
  • Cabbies may have been slow to react to the drop in hails as Uber gained market share. If yellows didn’t start trimming their shifts in earnest until late 2014, the steeper decrease in cruising after then could have tempered the drop in speeds that would have been expected from the growth in Uber.
  • Perhaps revulsion over yellow drivers’ maiming of Sian Green in August 2013 and killing of Cooper Stock in January 2014 led cabbies to operate more cautiously — by yielding in crosswalks, say. This could have contributed to slower speeds for a time, until the effects of NYPD enforcement or social opprobrium fell away.
  • Regression to the mean: the mid-2013 peak in speeds might have been a statistical aberration that made the ensuing decline appear more pronounced.

“Traffic is an incredibly complicated thing, and the people who argue about it are prone to overstating their cases.” That’s how Wellington concluded his New Yorker post, though this precept may have eluded the headline writer.

There’s actually much to like in that post, including Wellington’s enterprising and skillful translation of data for an estimated three-quarters of a billion trips, and his exculpation of the citywide 25 mph speed limit (that’s what the vertical dotted line in the graph denotes) from major responsibility for the slower speeds. But what seems to suffuse the piece, as well as earlier ones in FiveThirtyEight and the New York Times, is a reluctance to apprehend two basic truths about Manhattan traffic.

First, removing hurdles to using automobiles leads to increased use. Second, additional use of autos in a congested environment like the Manhattan Central Business District contributes to a general slowing down of all autos. (In a post last month I reported that an average mile driven by a single automobile in the CBD slows down all other vehicles there by a combined 10 minutes, though the figure varies widely by time of day.)

These realities may lack Uber’s glamour. But they’re no less real — on both sides of the windshield.

  • chfong@yahoo.com

    Let the automobiles/taxies/uber slow down to standstill, We should be focusing on improving the mass/public transit. Add more subway and commuter tracks, more bridges and tunnels for trains between NJ, NYC, CT and upstate NY areas.

  • com63

    Just make a few bus lanes along the way.

  • Alexander Vucelic

    I believe Bike lanes have reduced congestition in The CBD. Roughly data seems to Support this

  • BBnet3000

    Unfortunately, as we may be about to find out with the bike lane proposal on Amsterdam Ave, the city is strangely beholden to maintaining level of service for automobiles. Congestion in the core will keep us from improving the streets in the core.

    This is also why our fabled Protected Bike Lanes don’t continue into Midtown.

  • c2check

    Not only that, but the congestion makes New York a terrible place to be a pedestrian, bicyclist, or transit rider in general.

    There’s so much potential to make Manhattan so much nicer. But someone has to do something to restrict the number of cars that can access Downtown (and Midtown, and perhaps elsewhere, *ahemMoveNYplan*), or increase the space for bicyclists and pedestrians (by removing general traffic lanes *ahemBikeLanes+extendedSidewalks*).

    Lately we haven’t seen much of either.

  • Joe R.

    Actually, eliminate private automobiles from Manhattan altogether. Repurpose taxis as primarily a mode for the disabled (i.e. have some disability requirement for taxi riders). That gets rid of all nonessential motor vehicles. My guess is traffic volumes fall by 90%. Manhattan would be a walking and biking paradise. You could finally get rid of all those traffic lights which are bane of every cyclist and pedestrian’s existence. The idea someone getting around on their own power should have to wait for motor vehicles in a big city is preposterous.

  • Larry Littlefield

    From your prior post…

    “My calculation that the presence of Uber vehicles in the CBD is reducing travel speeds by 4.3 percent, is predicated on 50 percent additionality, with half of Uber CBD pick-ups constituting additional traffic while the other half displace yellow cabs and traditional black car service.”

    But what would those additional trips been absent Uber? Non-existent? Bus.

    Part of what is going on is the boom in the city’s economy, with fixed street space and (unfortunately given the financial repercussions of past policy) fixed at best transit service. That’s a problem for everything — housing, schools, etc. unless some kind of efficiency balances the rising numbers.

    I don’t have a personal interest here. I don’t have a smartphone and have never used Uber.

  • c2check

    I’d love to see elegant trams down the avenues and major cross streets, too. There were plans to redesign 34th and 42nd to that effect, but they were scrapped. Maybe it’s time to try again.

    We could totally do something like this here, and it could be amazing
    (Starssbourg)

  • Joe R.

    Amazing all the nice stuff you see elsewhere. Meanwhile, NYC streets look like a combination of a yellow cab parade and a used car lot with all the curbside parking.

  • IQuantNY

    Thanks for writing about my post! Your paraphrase of my work seems a bit off though. You paraphrased my logic as: “Since it didn’t, something other than Uber must have caused the slowdown.” when in fact the post points out that there is flawed logic on both sides. Of course I agree with you that adding cars adds congestion- that is not up for debate. My gut is that 40% is too high based on what I did, but we don’t know for sure.

    Honestly, the saddest part of all of this is that our governments still refuse to show their work when making data driven claims… so it’s hard to really understand what they have to say. I hope one day that will change.

  • Komanoff

    I think your 150-200K figure is high, by around 2x. Ditto your 20%. If you have any basis for them, please provide.

  • Alexander Vucelic

    about 30,000 Citibikes trips are taken in CBD each Day.

    Traffic Counts by TA this Summer Indicate 20-30% of bikes in CBD are Citibikes. These Counts do Not Account for Hudson River Parkway (8,000 -10,000 bikes a Day).

    Privately owned bikes are Not Used as Much as Ciribikes (8x a Day) , bit take easily 3-4 trips a day Running errands etc.

    TA has some excellant traffic Counts from this summer Which definitely concludss bike traffic is at least 10% of traffic on Avenues w/o protected bikes lanes and as Much as 25%-35% of Traffic with protected bike lanes.

    agreed 200,000 is on The high side, bit 100,000 is certainly on The lowside for daily Bike trips in The CBD.

    20% of Trips taken by bikes would have otherwise been taken using motor Vehicle ? Just a reasonable estimate, Cycling in The CBD is more of a substitue for Cabs/VFH than non-cyclists Might realize. 1/5 feels about right, Since its definitely a Minority of Trips but not as Few as 1/10 and certainly not as high as 1/3.

  • Komanoff

    Yep. And A-Rod hit his 50th career grand slam last night. And Obama carried all 50 states in 2012 …

    Seriously: Citibikes racked up an avg of 20K daily trips in first six months 2015 (20,767 weekdays, 18,565 weekends). Current six months will be much higher, but even full-year 2014 avg’d only 24K. And less than 100% of Citibike trips are CBD.

    Your “otherwise” percentages are conjectural. Early NYCDOT survey did find 21% of Citibike users saying auto or cab would have been part of the trip they took instead on Citibike, but that shrinks to 12% when normalized for multiple answers. And corresponding %’s are likely (much) lower for non-Citibike bike trips.

    If you’re going to throw numbers around, how about putting in a little time to base them on actual data?

  • Alexander Vucelic

    citibike data
    http://datawrapper.dwcdn.net/rNb8Y/29/

    CBD Traffic Counts by TA this Summer are Hard data – not conjecture.

  • Komanoff

    Your link isn’t to CBD traffic counts. It’s to a July-August continuation of the Citibike datasets I drew on in my second graf above, and is in line w/ my (obvious) prediction that 2H 2015 usage will be much higher than 1H. I hope you understand the pitfalls of cherrypicking w/ summer data, when it comes to bike counts?

    Anyway, what I said was conjectural was your “otherwise” percentages.

    Alex, I’m happy to continue this, but offline (kea “at” igc.org). You clearly have a lot to contribute, and it behooves you to take a more systematic approach to using numbers.

  • Alexander Vucelic

    Thank you

  • J_12

    I think there is an open question of whether Uber adds a significant amount of new car trips, or whether it mostly shifts trips away from yellow cab/radio dispatch/private vehicles, which would have happened anyway.

    In order to claim that Uber is adding in a significant way to traffic congestion, I think there is a burden of proof to show that it actually adds enough new car trips to make a difference. This is difficult to measure since we can’t directly observe the decisions made by individuals regarding how to travel.

    Looking at average cab travel speeds is fine, but doesn’t seem that meaningful in isolation. How did overall travel times for all vehicle behave over that period, and what is the expected relationship between aggregate vehicle travel times and cab travel times? Is there an interaction with travel times for subways or other transit modes? Is there an economic effect that would lead to people switching modes over this period?

    I think this is a hard question, and oversimplifying to blame Uber, or any other proximate cause, seems more motivated by political agenda than by data analysis.

  • Komanoff

    Re your first and second grafs: The evidence you’re saying doesn’t exist was distilled nicely in a NYT “Upshot” piece last month, http://www.nytimes.com/2015/07/28/upshot/blame-uber-for-congestion-in-manhattan-not-so-fast.html, which I summarized in a Sblog post last month that I link to in my first bulleted graf.

    Re your third graf: The 3/4 billion taxi trips that Ben Wellington drew on for his New Yorker analysis (and summarized in his graph at the top of this post) are a “proxy” for general traffic speeds. More precisely, *changes* in taxi speeds over time will tend to track *changes* in overall traffic speeds fairly closely.

    Your charge of political motivation is gratuitous and baseless. I’m data-driven in everything I do. The fact that in my Sblog post last month I ascribed some CBD congestion causation to bike lanes should be proof of that.

  • J_12

    I admire a lot of the work you do, but in this case I disagree with the validity of the methods you are using. Lets just leave it at that as I don’t think this is the right place for a technical debate.

  • Pedicab Pete

    Would the absolute explosion of construction sites within the last one or two years extending into pinched lanes of traffic – beyond just lanes of parking – with red and white barriers or construction fences that all road users must navigate carefully around (including cyclists) be slowing traffic speeds down? I have never had to negotiate so much construction material and vehicles in all my years of cycling in NYC. It’s hellish.

  • chekpeds

    Guys ! You cannot be serious!
    The issue is not if Uber contributes to congestion ( obviously they do) but whether their contribution is meaningful.
    Because the economy is now in full swing again, we should compare all the numbers to pre – 2007 since the intervening years of depression are a statistical anomaly.
    Second, I would check the volume of construction going on in manhattan. Just in the west side 10 th avenue is down to one lane, 9 th avenue is down to 2 lanes because of water main constructions, 25 th street has two constructions going on, 40 th street has a huge hotel being built .
    I presume it is true in many districts where financing has resumed and there was pent up demand. Each costruction takes up at least one lane and often more . And generates large numbers of truck and contractors trips
    Verizon is still rolling out FIOS and MTA is building the second avenue subway.

    The reality is that we are a costruction zone which I believe is the main culprit by far.

  • chekpeds

    Exactly right . See my post above

  • HeyYou Chen

    I agree with constructions as one of the major contributing factors to congestion. But I would blame many on us individually that contribute to the problem collectively. I actually LOVE Uber. Below are my observation what cause the congestion (not including construction, which IMHO is the biggest contributor):
    1) traffic light pattern. We used to be able to traffic light surf on the Ave. I remember many times in the 90s up to late 00s where I can go practically from Houston/1st Ave all the way to 59th st without stopping for a red light (saturday night). This past Saturday, I had to stop for red lights at 10thst, 19th, 22nd, 23rd, 28th, 33rd, 49, 55th, 57th, 59th. I blame this on stupid pedestrians who don’t pay attention to roads and just look at their phones. This caused all the brain dead politicians enact laws to slow down the traffic in order to save some selfish idiots’ lives.
    2) Selfish drivers who block the box. This is the major cause of grid lock problem. Yet many a-hOles love doing this.
    3) Selfish individuals and drivers who double park. I’ve noticed so many times a car could just pull into a hydrant to temporarily offload cargo or discharge/pick up passengers. Nope, these morons have to block one lane that cause congestion and create more noise pollution.
    4) Bike lanes. That took one driving lane away. You don’t think that contribute to the congestion? Although I like bike lanes.

    5) Yellow cabs themselves that circle around to pick up a ride. Most Uber drivers often park on the side waiting for their next pick up. Who do you think cause more traffic problem here?

ALSO ON STREETSBLOG

Uber and Manhattan Gridlock Are Rising Together

|
How responsible is Uber for the 9 percent drop in Manhattan travel speeds that New York City transportation officials reported last month? The answer appears to be: quite a lot.  If — and it’s a big if — the surge in use of Uber and other app-based car services is not offset by a decline in use of […]

Heat Maps Show Where Traffic Takes the Most Lives in NYC

|
As city government started work on the Vision Zero Action Plan, statistics professor Ben Wellington saw an opportunity to use data on crashes and fatalities to show the magnitude of the challenge. Wellington teaches a statistics course to Pratt Institute city planning students using open data from New York City government. He also uses city […]

Uber’s Own Numbers Show It’s Making Traffic Worse

|
Uber blasted out an Excel spreadsheet to reporters this morning, accompanied by a story and editorial in the Daily News, with data providing a snapshot of how many Uber vehicles are on Manhattan streets south of 59th Street, New York’s central business district. While Uber claims the data shows its vehicles aren’t responsible for congestion in the city core, […]