Study finds a slimmer York Boulevard has cut down on collisions

The removal of traffic lanes on the western section of York Boulevard through a “road diet”  has helped improve traffic safety, including a sharp decline in hit-and-runs, according  a review of traffic collision figures for the road that passes through Eagle Rock and Highland Park. However, the number of collisions involving pedestrians continued to rise – albeit slightly – as did those involving cyclists, said an analysis published by the L.A. Department of Transportation’s Bike Blog.

While many motorists have complained about increased traffic congestion on York, the analysis found that collisions and injuries per mile declined 23% and 27% respectively after a 1.3 mile-section of York between Eagle Rock Boulevard and Avenue 54 was reduced to one traffic lane in each direction in 2006.  The results of the study compare five-years worth of collision figures before  the  traffic lanes were removed in 2006 and the five years after the road diet took place.

The study also compared the slimmed down, western section of York to an .09-mile  stretch between between Avenue 54 and North Figueroa Street that kept all its four traffic lanes until October 2012, when a road diet reconfigured west bound lanes. Collisions and injuries dropped in both sections of York but the declines were more pronounced on the west end of the street that had traffic lanes removed.

The LADOT Bike Blog hailed the results as proof that road diets improve traffic safety:

Road diets aren’t without their critics, but these projects ultimately make us all safer and everyone is a beneficiary of a street made safer in their neighborhood. Not only does it mean fewer collisions are happening when streets are reconfigured to be made safer, but it also means the city resources that would otherwise go towards responding to traffic collisions can now aid other emergencies and neighborhood needs.

Click here for Bike Blog’s analysis of the collision figures.


  1. Next up: The Hyperion and Glendale Blvd freeways.

    • I don’t happen to commute that direction, so I don’t know if Glendale Blvd through Atwater and into Glendale gets very congested. Certainly by the time it gets down to Fountain, any congestion is fairly limited in time – I live two blocks from there and I make a left turn from southbound Bates to eastbound Fountain to Myra as part of my commute, and traffic is light enough that I never have to wait more than 20 or 30 seconds for a gap big enough to do the left turn, even on a bike.

      That stretch of Hyperion and Glendale has lots of local restaurants and bars that I would love to visit more often, but it’s just an incredibly unfriendly street, so I normally just go over to Sunset Blvd, which is much more friendly, despite probably having just as many cars moving just as fast. If they can improve the amount of space on Hyperion and Glendale for pedestrians and cyclists, the local businesses would probably see a lot more local patronage (without needing to add parking), and they should be able to accommodate all the existing car traffic, though I’d have to hear more from traffic engineers about that.

      Glendale Blvd in Atwater particularly probably doesn’t need to have so much space devoted to cars. It was originally designed before the 2 existed, and when some of its space was devoted to the Pacific Electric. With cars having a much better option on the freeway now, I suspect there’s even less demand than there was back then, though I’d need to see more historical traffic studies to be sure. If that space is turned over to pedestrians, Atwater could become a great pedestrian space, though of course we all might prefer to see that old red car line revived.

  2. This just goes to show that pedestrians and cyclists aren’t the root of the problem; bad motor vehicle drivers are.

  3. That’s because we avoid driving down York blvd period. Good bye York blvd. Hello Yosemite DR. and figueroa.

  4. As a driver and occasional bike rider, I don’t mind the road diet on York. It slows things down a bit, but that’s a good thing, since it’s a busy commercial strip with lots of pedestrians.

    A road diet on Colorado in Eagle Rock, dropping it from 3 car lanes to 2, would make that stretch more walkable and less of a freeway. One danger of reducing speeds on main arterial roads is that it pushes traffic onto side streets — and I fear Yosemite could bear the brunt of this. And on Colorado, a road diet won’t solve the worst parts for bikers — by the 134 onramp near Wiota, and then, in Pasadena, the horrible stretch on Colorado near the San Rafael onramp.

    • What will most likely happen in the installation of stop signs on Yosemite and you’re right on for that broken stretch on Colorado to San Rafael. A rainstorm THREE years ago damaged and the city is sitting on it’s aXX, THREE years. I heard that it costs 350K to fix and because of the delay the damage has spread.

  5. I find it interesting that the Bike Blog presents the positive numbers as percentages but the negative numbers as rates. The decrease in collisions and injuries are presented as significant at -23% and -27% respectively, but the increase in pedestrian accidents is called modest at 29%. It then goes on to downplay the 100% increase in bike related accidents by claiming that bike traffic increased by 150%. I decided to check this and found that the bike count for that intersection actually lists a 140% increase not 150%. Something stuck me as odd about the graphs on the bike count. There were no error bars. This is usually a bad sign. So I did some napkin math. The error in the calculation for York/Ave 50 is plus or minus 57%. This is undoubtedly because the sample size is rediculously small (3 samples in 2009 and 3 samples in 2011). As of right now there is not enough data to draw any meaningful conclusions about the relationship between increased use and increased collisions. On the bright side, with a margin of error that large the BikeBlog wasn’t exactly lying when it claimed 150% instead of 140%.

    • It would also be useful if they had made note of the change in motor vehicle counts during the time period. My guess is that as mike and Lisa suggested, drivers might be avoiding the street. It would be really useful to know how much of the decrease in collisions is due to decrease in number of cars and how much is due to improved safety for the existing cars. It looks like collisions involving bicycles haven’t changed in any significant way per bicycle, but the number of bicycles has increased hugely, which is still a win for local parking and the environment, even though it’s not much of a win for safety.

    • Error bars? Really? The data was pulled from discreet reports filed by police officers. What, does he need an error bar because maybe the author of the report misread a report or something?

      This isn’t a chemical test to see if chemical X causes cancer in albino rats – this is a stacking up of paper reports of things that actually happened and were documented.

      Do you want another study for crashes that weren’t reported? Hey! Now you are going to need error bars because you will be guessing as to the unreported crashes.

      Please, take your back of the napkin retorts back to the burger shack,

      • The concept of error bars is actually relevant. The motivation is not that there might be an error in a report (that’s what your reply is thinking). It’s that we only have a small number of reports to estimate the true rate.

        Like if you try to estimate the chance that a thumbtack will land point-up by throwing it 6 times. You definitely observe it correctly each time (i.e., you read the police report correctly), but because of the limited number of tests, your estimate of the odds will have an error. The size of the error can be estimated in terms of the number of trials. Small number of trials gives a big expected error.

        (I did not check @skr’s error bars, but they should go like 1/sqrt(n) where n is the number of trials. For 6 trials, 57% would be higher than I’d guess, but there are several ways to do the computation.)

        That was more than you wanted to know, but @skr has a point.

        • Jesus Christ, WTF. We are not looking for one of 10,000 possible causes of a disease here.

          If anything, the reported numbers of crashes are drastically understated because police do not get called nor write a report, nor submit every report to the SWITRS database.

          However, OF THE REPORTED CRASHES, there is a clear trend over a4 year period both before and after the lanes went in. There is no need to provide error bars because we are not measuring anything – we are comparing reported crashes from one group of years with reported crashes from another group of years.

          There is no prediction being made, there are no estimates being used.

          The sample size you are looking for with respect to the direct correlation between road diets (and vehicle speeds on urban streets) and reduce numbers of reported injuries and crashes can be found in the vast and growing literature on road diets wherein precisely the same drop in crashes and injuries can be found.

          • since the graphs to which I was referring were the bike count totals, yes there are estimates being used. The large margin of error is becaue the sample size is incredibly small. The data is junk. But thanks for getting all irate over something that I didn’t even say.

          • Any time you compare small samples of random data, the concept of error bars, or a like concept, is highly relevant. Since you chose to respond to my reasonable reply with an emotional rant, I won’t say more. Just point out that you clearly know less about this than you think you do, and it makes you look dogmatic and ignorant. (I’m a supporter of bike lanes, and I showed up at both the public comment sessions to support them.)

        • Additionally, more reports means more people injured or killed. That sounds like a shitty experimental method. How long do you want to wait for results? 8 years is too short a window? 8 years of traffic crash reports, and their trends as compared to various other sections of a roadway, sounds like a pretty solid sampling window within which to make these sorts of findings. That is, the industry standards are pretty low and actually just looking at the reported number of crashes on a street is a novel means of judging the value of planning decisions done in the right of way in Los Angeles (no sarcasm, that is really the truth).

      • um no, the police reports have absolutely nothing todo with the graphs I was talking about. They were generated by a different group conducting a bike count.

        • Darn it, you are right. The LA bike counts are a pretty shoddy affair as these things go. With respect to those bike count figures, we definitely need some error bars when it comes to extrapolating into a city0wide measure. The LACBC used a procedure developed by a federal body but has only been able to do 1/2 of what the procedure calls for in terms of intersection measurement. That said, it is an all-volunteer effort done by an agency that receives little to money to do said bike counts. You would think, with all the money spent on things like ATSAC, that our systems of traffic monitoring would be some of the best in the world and account for a full modal split measurement – and you would be wrong in that assumption. We toss our data and ignore traffic counts in LA (and most of America) preferring instead to rely on highway studies done in the 1960’s and 1970’s and computer models developed by SCAG over any real data culled from scientifically rigorous study of our own streets.

          Apologies for being such a dumbass on this one.

          • no problem.

            But one little quibble. They only monitored more like 1/3 of the necessary intersections and only had ‘complete’ data for about 17% of the necessary intersections. And what they considered complete was having a dataset of 3 data out of a possible 730 for a given year for the two monitored years.

  6. Also, there is an error on the bike count. It lists a 2009 count for that intersection as 70 and a 2011 count as 168. This is how the 140% increase is arrived at. However, if you add up the reported counts you get 87 and 168 respectively which results in only a 93% increase. This correlates more stongly with the 100% increase in bike collisions and suggests that the bike lane only gives the appearance of safety while doing nothing to decrease collisions. Of course such conclusions are spurious with such incomplete data.

    • skr,

      Thanks for pointing out the bike count error and inconsistency in the post!

      All changes were reported in rate, which is often more telling than percentages (especially when dealing with small numbers), but we’ll update the post so that they are all also reported in percentages as well.

      A note on the bike increase percentage: The LACBC .pdf initially reports 150% though subsequently reports a 140% increase. As you pointed out, the actual increase based on the individual reported counts is 93% so we’ve corrected the post to use this figure.

  7. Human scaled streets are safer for humans, not really all that surprising if you think about it… LA’s just a few decades behind the times when it comes to transportation and urban planning.

Post a Comment

Please keep your comments civil and on topic and refrain from personal attacks. The moderator reserves the right to edit or delete any comments. The Eastsider's Terms of Use and Privacy Policy apply to comments submitted by readers. Required fields are marked *