Credit: WSDOT

Good news: traffic deaths in Seattle were down significantly in 2018, according to SDOT according to the latest available data from SDOT. But we’re still far from Vision Zero.

Please note that collision data for 2018 is incomplete, as reporting is still in progress. The data in this story was released on December 19, 2018 and downloaded on December 25, 2018. The City’s internal data collection is ongoing. 

Ten people were killed in collisions this last year, down from 19 the previous year, the 2010s annual average of 19, and a decade-high 27 in 2016.

In 2018, more people were killed traveling by car than any other mode. That’s the norm in Seattle. In the 2010s, nearly 60% of people who died in traffic in Seattle were killed while driving or riding in a car.

Over the course of the 2010s, the average day saw 38 collisions across town. So it’s no surprise that collisions happen everywhere in Seattle.

However, some intersections and blocks are more dangerous than others. Highway 99 and Rainier Avenue South are particular standouts. Downtown, as a whole, is by far the most dangerous place to get around.

Downtown, where people in Seattle walk the most, has seen the most collisions by cars and cycles with pedestrians. Rainier, Aurora, 45th Avenue North, Lake City Way Northeast, California Avenue Southwest, and Delridge Way Southwest are also notably dangerous places to walk.

Rainier is dangerous for its entire length. Highway 99 is especially bad on the Aurora Bridge and the Viaduct. Fortunately, the Viaduct is coming down, and rechannellization plans are in the works for Rainier.

However, the Aurora Bridge is just as dangerous as it was in 2015, when the horrific Ride the Ducks collision killed five people and injured 78 others.

The high-volume highway has no median, and the city and the state were fighting over who should pay for installing one. As of June, no plan to fix the bridge was even in development.

Of course, the Viaduct was identified as a similar threat to life and limb about 20 years ago, and it’s only now being demolished. So expect an improved bridge in 2040 or so.

Notes on the data

Each location is aggregated for the entire time period of the data set: if separate collisions happened in the same place in 2010 and 2014, they have been combined into the same point. Each location is geolocated using coordinates supplied by SDOT. The dataset was last updated on December 19, 2018.

38 Replies to “Seattle’s most dangerous roads”

    1. I mean, Bellevue is a different city. So, yes, ‘Seattle”, the city, stops at the bridges.

      1. Yes, but many articles on this blog reference the region. And readers certainly come from across the metropolitan area. The link seems like it might be useful or interesting to some.

  1. Essentially this data is saying “busy streets have more crashes” which doesn’t really tell us that much. I’m not sure Downtown is the most dangerous area in terms of crashes relative to the number of people present – there are far more people Downtown than anywhere else.

    This analysis can’t identify a street that isn’t very busy but is nonetheless dangerously designed, has aggressive drivers, etc. They blend in with all the other streets.

    1. I think you could generate that sort of analysis by taking the number of accidents/deaths per intersection and dividing them by the total number of drivers to get accidents as a percentage of the total traffic, and then just look for anywhere that stands out as particularly dangerous.

    2. I don’t think that’s necessarily true. Broadway, Queen Anne Ave, there are plenty of other places that are “busy” and not listed here. The issue with some of the streets that are is that they are in lower-income areas and have far fewer cross-walks. They have not been properly planned. Rainier can go on for a long time without a cross-walk, let alone a stop light. Pedestrians have no choice but to cross in dangerous locations. Drivers are not expecting it.

  2. For clarification roadway fatalities statistics include when a person is killed at the scene of the collision. A recent resident in Ballard was in a car collision but survived one month in critical condition at a hospital and the cause of death was a brain aneurysm, she is not listed as a traffic death. There was also a motorcyclist who was critically injured on Market street in Ballard in August and later died from his injuries who is not listed as a traffic death.

    I’m assuming there are many many many more which makes traffic fatality statistics dubious at best. Maybe the only redeeming factor is death statics are likely consistently unreliable every year so trend analysis could show changes in road safety or police/hospital reporting, but I would say it’s a horrible indication of absolute values for roadway fatalities.

    1. Good point. The FHWA is no recommending combining ‘Fatal’ crashes with ‘Serious Injury Crashes’ for analysis; their point is that the differences between Fatal and Serious Injury crashes are indistinguishable in many cases, and a person may have survived due to random chance (or age, or luck, or whatever, or as you point out, just lived long enough to not be listed as a fatality). It looks like these data only have the three-level distinction of crashes: Fatality, Injury, and PDO (property-damage-only). For a much more reasonable analysis with far more data on dangerous road locations, one should include Serious Injury Crashes into these numbers.

  3. “Downtown, where people in Seattle walk the most, has seen the most collisions by cars and cycles with pedestrians.”

    We have to get these dangerous cycles (whatever those are) off the road! Based on this sentence it appears they are just as dangerous as cars!

    Both sides. AMIRITE

  4. 45 mph is too high a speed limit for Aurora. A lot of people are traveling as high as 55 mph It probably should be lowered to 35 mph. And particularly on the bridge it needs camera enforcement.

    by the way – citations should be greatly increased, and fines for 1st or 2nd time offenders greatly reduced. Purpose is to increase safety and reduce injuries – not raise money.

    1. Or, just reduce it to two wider lanes per direction, and add a center barrier and shoulders.

      It won’t actually reduce capacity because, with trucks and buses forced to straddle two lanes the 3 lane capacity was fiction to begin with.

      1. The other option is 3 lanes one direction and 2 lanes in the other, with a movable barrier that matches the additional lane to peak traffic.

      2. @asdf2 — I wouldn’t mind the change, but switching from 3 to 2 lanes each way would alter the capacity. There are only so many buses that go across Aurora. You will never see a long line of buses, one after the other, crossing. I think there are four bus routes (5, 26, 28 and the E) that cross the Aurora bridge. The E is most frequent, peaking at about every four minutes. So if every bus has similar peak rate, that is one bus a minute. That just doesn’t take up much space, even if each bus takes up two lanes.

        You can kind of see what I’m talking about in this picture:×855.jpg. Now imagine that there are twice as many cars. They can still fit, and are using up all three lanes behind (and in front of) the bus. Now imagine there are two lanes (and twice as many cars). Something has to give. They won’t all fit. I’m not saying I’m against the change, but capacity would be reduced.

      3. I’m sure a 5-lane configuration would really save enough road space to include all of a center barrier, wider lanes, and shoulders. A 4-lane configuration would. Shoulders would make a huge difference in comfort for people on the sidewalk.

        Also, assuming the curb lane on the approaches to the bridge are bus-only in both directions, the effect on bus travel times would be negligible, and, even for cars, the merge down from 3 lanes to 2 wouldn’t back up much they’d only have to stop to let someone in when there’s a bus coming.

      4. I think the best option is 3 lanes southbound and 2 northbound. The 39th on ramp would work better with the 3rd lane. Bus lanes could be extended up to the bridge on all sides, so things wouldn’t get worse at least.

    2. For every arterial inside Seattle city limits, including and especially Aurora, I wonder if we couldn’t use traffic signals not to “enforce” safe speed through threat of punishment but rather re-INforce the instinctive smooth driving of which I think every licensed motorist should be capable.

      Have real-time signage fix drivers’ awareness not on the legal speed – which goes without saying- but on the speed that will least likely to face force you to stop. My sense is that this awareness will tend to put drivers in same mind-frame as their driverless digital counterparts. And tune tempers accordingly.

      Like every good habit, correct driving speed starts with an idea. In this case, two ideas. That within any city limit, vehicle speed is only legal and design priority for public transit in its own fully-reserved lanes, and answering its own signals. And for everybody else and their machines, safety and comfort come first.

      Mark Dublin

    3. Speed limits are the one part of Vision Zero that remain. 15th Ave NE and NE 65th Street have both been lowered to a ridiculous 25 mph. Slowing down arterials makes the buses go slower too, and makes already-long trips across the city even longer.

      I don’t know what Aurora’s speed limit is on the expressway but between 80th and 130th at least it’s 30 mph, and I think in Shoreline it’s 35. You only get to 45 in Snohomish County.

  5. Is there any analysis of time of day? Many Seattle streets are poorly lit because the street lights are up in the trees, and the new LED lights don’t seem to sync well with the old light poles. It’s hard for drivers to see pedestrians when lighting is poor.

    1. Yes this. I am constantly shocked by how many lights are off pre-dawn on major arterial roads. Or how many people are stupid enough to run across a 5 lane road in the dark, wearing dark clothes.

      1. Swear that some people think it’s their Constitutional right to save battery by leaving their headlights off until at least midnight.

        Can anybody think of a single reason why every single car’s electrical system should not put headlights on as soon as the engine starts?

        Mark Dublin

    1. Yes. And the collision data that goes online typically lags by several weeks to several months. All of 2018’s data is not yet online. Here’s an example of all collisions every year:

      dilinger@e7470:~$ grep 2014 Collisions.csv |wc -l
      dilinger@e7470:~$ grep 2015 Collisions.csv |wc -l
      dilinger@e7470:~$ grep 2016 Collisions.csv |wc -l
      dilinger@e7470:~$ grep 2017 Collisions.csv |wc -l
      dilinger@e7470:~$ grep 2018 Collisions.csv |wc -l

      Either total collisions dropped massively, or 2018 collision data is incomplete.

      Let’s look by month:

      dilinger@e7470:~$ grep “2018-01-” Collisions.csv |wc -l
      dilinger@e7470:~$ grep “2018-02-” Collisions.csv |wc -l
      dilinger@e7470:~$ grep “2018-03-” Collisions.csv |wc -l
      dilinger@e7470:~$ grep “2018-04-” Collisions.csv |wc -l
      dilinger@e7470:~$ grep “2018-05-” Collisions.csv |wc -l
      dilinger@e7470:~$ grep “2018-06-” Collisions.csv |wc -l
      dilinger@e7470:~$ grep “2018-07-” Collisions.csv |wc -l
      dilinger@e7470:~$ grep “2018-08-” Collisions.csv |wc -l
      dilinger@e7470:~$ grep “2018-09-” Collisions.csv |wc -l
      dilinger@e7470:~$ grep “2018-10-” Collisions.csv |wc -l
      dilinger@e7470:~$ grep “2018-11-” Collisions.csv |wc -l
      dilinger@e7470:~$ grep “2018-12-” Collisions.csv |wc -l

      So either November was a very safe month, and December even more so, or.. it’s not all there yet. Please take this article down.

      1. I’m not clear what data Peter used for this post, but if it’s *data* that data isn’t supposed to be used any more. The city switched to this:

        From that data, you can pretty easily see that there were at least 13 people killed last year.

        dilinger@e7470:~$ grep “2018-[0-9][0-9]-” Collisions.csv |grep -i fatal|wc -l
        dilinger@e7470:~$ echo -n “Number of bike deaths: “; grep “2018-[0-9][0-9]-” Collisions.csv |grep -i fatal |grep -i pedalcy|wc -l
        Number of bike deaths: 1
        dilinger@e7470:~$ echo -n “Number of pedestrian deaths: “; grep “2018-[0-9][0-9]-” Collisions.csv |grep -i fatal |grep -i pedestri|wc -l
        Number of pedestrian deaths: 5

        And again, that’s not even the full year.

      2. Thanks for flagging this. I should have been much more explicit that the 2018 data is incomplete—the lead overstates the case. That’s my mistake, and I’ve made a clarification. This is definitely something that I’ll continue to track, and work directly with SDOT to know definitively when 2018 reporting is over.

        Looks like you’ve found the dataset that I was using. Looks like it was updated yesterday (, which accounts for the discrepancies that you noticed. I was working off a .CSV file current to December 19 that I downloaded on the 25th, during a break in festivities.

        According to the 2018 Open Data Plan (, is still the repository for official city data. The GIS API is hosted on a different domain.

  6. Need some details about these accidents. From personal and professional driving experience over the years, I’d say driver training has to get a lot more rigorous, and licenses a lot harder to get. Car body styles- would like a word with a design engineer.

    Long time ago was trained to “Shoulder Check” before lane change. Appreciate stronger structure, but wonder how much lost window space now is just to make the car look deadly. Sorry, cameras, but the corner of my eye never once blinked out when most needed- noted downside of electronic vs. physical
    mechanical for anything.

    Comparison I also noticed some years back. Around Los Angeles, freeway signage lets you know at least five miles ahead which lane you need to be in. Tacoma’s (hopefully) world’s worst, but signs in Seattle also make your lane your best guess, with a half hour penalty for wrong choice. And perfect state of mind for a fatal last-instant attempt to correct.

    Remedy? Nationwide, once a year every motorist gets an hour with a State Police instructor (after small- to- no bookwork brushup). Who’s in turn got the choice of issuing a driver’s license or a Prepaid!!!!!! pass good on every ground transit system in the US, Canada, and Mexico.

    Finally (you win, John Niles!) a real cure for congestion. And a drop in insurance rates that’ll make whole tax base see next ten ST-s as their travel lives’ best bargain. But really would like to see break-down on categories of accident details. Thoughts from anybody else?

    Mark Dublin

    1. I would observe that the reports don’t discuss our freeways. Are freeways less or more dangerous than city streets in Seattle? If safer, does that mean that design is more of a factor than speed? How are the two related?

      It’s common in many places to put larger street name signs that are lit from behind hanging off of signal poles at major intersections in many cities. Seattle doesn’t typically do this.

    2. How much elecyricity do those signs use? Another thing the US does is put a traffic light above every friggin’ lane. In Vancouver you have two highways crossing with one light in each direction.

      1. LED is a huge power savings, broadly estimated as only 25 percent of power needed for prior technologies. Backlit street sign power would only be needed at night, and don’t require the brightness of street lights. I don’t see energy consumption as a major issue here.

  7. The FHWA is now recommending combining ‘Fatal’ crashes with ‘Serious Injury Crashes’ for analysis; their point is that the differences between Fatal and Serious Injury crashes are indistinguishable in many cases, and a person may have survived due to random chance. It looks like these data only have the three-level distinction of crashes: Fatality, Injury, and PDO (property-damage-only). For a much more reasonable analysis with far more data on dangerous road locations, one should include Serious Injury Crashes into these numbers.

  8. Am I right that almost every preventable fatal accident boils down to somebody driving too fast for location and conditions?


  9. It was a bit surprising to read “Uninsured motorist rates varied from 4.5% in Maine to 26.7% in Florida. Other struggling states include Mississippi 23.7%, Michigan 20.3%, New Mexico 20.8%, Tennessee 20.0%, and Washington 17.4%”

    Having been hit by an uninsured motorist I think Washington is playing with fire.

  10. Why is this post up at all? The analysis here is entirely flawed and paints a false picture. The data is totally wrong as others have posted. Please remove this article and issue a correction.

  11. Ooops.

    All snark aside, I await the OP’s eventual (corrected) reporting on these metrics. With that said, I think it’s silly to put any stock into the numbers asserted for cyclist/pedestrian accidents as many of these incidents, even those involving injuries, simply go unreported.

  12. “Downtown, as a whole, is by far the most dangerous place to get around.”

    NO! Bad data.

    Without normalizing for pedestrian population, we have no idea if it’s more dangerous to get around downtown compared to anywhere else. If there were 500x more people downtown wearing hats than a place with 1/500th the population would you say you’re 500x more likely to wear a hat if you’re downtown? Of course not. The same is true about getting hit. It’s possible that’s true, but your data has not shown that one way or the other.

    Explanation of this sort of bad data:

    This very blog.


Comments are closed.