Over the last few years, transit agencies have created incentives to take transit to New Years’ celebrations. Trains run later, and in 2017 Metro was free for the first time. While it’s worthwhile to provide attractive, safe alternatives for people, I’m sad to say that these improvements haven’t shown up in the accident data, courtesy of SDOT:

Year 2013 2014 2015 2016 2017
Total Collisions 31 17 26 25 26

The collision count is from 5pm in December 31st of the listed year to 5am on January 1st.

Of course, there’s missing context here. The city’s Department of Special Events didn’t respond to my request for crowd estimates, which is the denominator of the numbers above. Metro said 160,000 people boarded a bus on NYE 2017, compared to 170,000 in 2016 and 300,000 in 2015. But that difference is almost certainly due to 2017 being a Sunday, while 2016 was a Saturday and 2015 a Thursday. The 160,000 figure is “about 17 percent more than average Sunday-service ridership in fall,” according to spokesman Scott Gutierrez.

The nice thing about no-fare days during major events is that we get a lot of the benefits of free transit without the downside (significant lost revenue). No one is going to skip buying a pass because of the holiday, but riders don’t have to deal with a flood of new (and drunk) users struggling through fare payment. If there were a way to make buses like a newspaper paywall, where the first few trips of the month are free, that would be a revenue and operations sweet spot. But quite aside from the safety impacts, it’s good to take advantage of these sorts of opportunities when they occur in the calendar.

9 Replies to “NYE Service Improvements Not in the Data”

  1. Even ignoring the issue of the denominator, I don’t think there’s enough collisions for a trend to be statistically significant. With every single collision being around 5% of the total, any “trend” is going to have a lot of random noise.

  2. I am so confused by this post. As I gather, the untitled chart depicts “collision count” on each new year’s eve, as reported by SDOT. What does that even mean? Does that mean each count is the number of accidents by buses on new year’s eve? I find it hard to believe that there were 26 incidents of buses crashing in a 12-hour period this past NYE. Or maybe this is the total number of traffic accidents in Seattle that eve, which seems to contradict the context which is entirely about transit safety. Are buses striking pedestrians counted as a collision?

    1. I assume a collision is any collision with a car. The logic connecting this to buses is that by extending the hours of transit and making it (Metro, at least) free through New Years, we’d see a decrease in car accidents, since fewer people would be driving home drunk. The data does not support this, though it does seem to be rather noisy.

      1. Injuries and fatalities would be a more important measure. People can get hurt or killed without getting into a car. But, like collisions, the number within this time frame is already so small, that it is of little statistical use.

        Expectations of fewer people driving also ignores the induction of additional drivers onto the road as other drivers shift to free transit. Only reducing road capacity (but not turning it into additional parking) will solve that problem. But I will settle for getting parked cars out of the transit lanes 24/7.

  3. Ditto clarification of “collisions.” But do think that the more drunk people transit carries, the fewer cars they’ll be driving. Which should substantially pay back the cost of the lost fares.

    Will also introduce people to transit whose driving would not be visibly impaired by the amount of alcohol in their blood, but will get punished severely if they get pulled over for a bad tail-light. Call transit “#Designated Driver!” Or something that sounds like at app. And also make it an if it isn’t already.


  4. Wouldn’t we need vehicle-miles-driven or some other information to normalize the data? There’s lots more people in Seattle now than even a year ago, so that the rate of change is flat in the collision numbers could actually be a good thing.

    1. Population seems like a more relevant denominator, to reflect growth, than vehicle miles driven.

      Reducing vehicle miles driven during that 24-hour period would be its own measure of success. Using it as the denominator would make increasing vehicle miles as important as reducing collisions in the effort to reduce collisions per vehicle mile.

      The ideal measure would be person-injuries per person-trip, but compiling that would be a logistical nightmare.

      1. True, though my pre-coffee thinking included transit in vehicle-miles-driven (i.e. 20 people on a bus driven a mile would be the same as 20 SOVs driven a mile). There’s probably a more precise term for it (per-capita distance traveled?).

  5. We also need to review DUI stops from the police data, but again, one night is not a large enough sample.

Comments are closed.