Sunday, October 15, 2017

Understanding Safety: Crash Rates vs Counts on State Street

One of the claims here is that making distracted driving our primary safety focus is a kind of error.

It attacks a secondary problem or phenomenon or factor rather than the primary one. It's not an error in the sense of something false, but is an error in the sense of something misguided or inefficient.

The new cellphone law
Behind that focus on distraction is an analytical preference for rate statistics rather than counting statistics, it seems to me.

When we focus mainly on distracted driving, we say we will keep the rate and amount of driving more or less the same, and we are going to try to squeeze out incremental improvements in our rates of safer driving. It's really a kind of qualitative improvement: We're going to try to drive better.

In and of itself that is not an error of course. We should always try better! Distracted driving is a problem, and we should want to make inroads on that.

But that may not be the best approach to have as the primary focus.

A different approach would be to say that humans are fallible creatures and that since driving a vehicle weighing thousands of pounds at lethal speeds is inherently dangerous, it is more prudent and more effective to drive less distance, drive less often, and drive more slowly. Even if our rate of crashes remains the same, because we are driving fewer miles and making fewer trips and driving more slowly there will be fewer crashes and fewer deaths. This is mostly a quantitative argument.

A table of modeled crash rates on State - the dead are erased
(Tier 2 Evaluation of the Street Design Alternatives)

And a similar table for Statewide targets
(at SKATS last month)
Because at the end of the day, at the end of the month, at the end of the year, it is the number of dead people that matter. It's the count that matters. It's mourning 17 dead instead of 24. It's binary: dead, not-dead.

Reducing the death rate from 0.24/1M miles traveled to 0.17/1M miles traveled (or whatever is the denominator - I don't think the memo defines it for "crash rate," and these metrics are presumably standard guild knowledge) may or may not affect that count. It also moves toward abstraction. The relevant metric should instead squarely focus on the awful costs to the dead, their families, and their friends. It's not just some efficiency rating that allows us to abstract away from the dead, as if death and injury were merely some minor, incidental cost on our roads to be waved away.

Road Diet reduces crashes by at least 20%
In the draft State Street Corridor Plan we see this tension exactly:
In summary, Alterative [sic] 1 – Improved Four-Lane and the No Build scenarios yield very similar crash predictions for all intersections as the volumes and intersection configurations are mostly the same. Although predicted crash rates are slightly higher under Alternative 2 – Road Diet, it provides the most consistent predicted safety improvement of all of the alternatives in terms of total number of crashes for all intersections. The roadway changes result in the predicted total number of crashes reduced by at least 20 percent at six intersections along State Street. This stems from the underlying approach of Alternative 2 – Road Diet, which reduces the number of lanes on State Street and adds a two-way left turn median that becomes a left-turn lane in both the eastbound and westbound directions. Alternative 3 – Hybrid is predicted to provide a slight improvement in the total number of predicted crashes and to see crash rates very similar to those of Alternative 1 – Improved Four-Lane.
As modeled the Road Diet concept yielded "the most consistent predicted safety improvement of all of the alternatives in terms of total number of crashes for all intersections. The roadway changes result in the predicted total number of crashes reduced by at least 20 percent at six intersections..." The underlying technical memo adds that "the intersection of Court Street and 13th Street...sees a reduction in total crashes of 35 percent."

But all the tables (see above) are about rates, not counts.

This is an image from a road diet
(chapter divider in State Street Plan booklet)
One of the ways that the team justifies dropping the full road diet is to focus on crash rates instead of raw totals. That helps to mystify the actual trade-off: Less cut-through traffic on side streets for more death and injury on State Street.

But if we want fewer people to die or be hurt on State Street, we should employ the full road diet. This is a straight-forward conclusion from the analysis.

And then if cut-through traffic remains a problem on the neighborhood side streets, we should employ other actions to address that.

More generally, we should think more about what a focus on counts or rates might mean. I don't pay very close attention to the national and international conversation about Vision Zero. I don't know what that methodological debate might look like. (Do you?) It's probably true that rate statistics are more stable than raw counts, and that the year-to-year variation in counts easily leads to over- or under-estimating success. Truly, the multi-year trends are more important than year-to-year fluctuation, which is subject to more randomness.

But as we consider the large issues:
  • Climate disruption and greenhouse gas emissions
  • Road safety and public health
  • Urban aesthetics and livability
  • Municipal budgets and the expense of construction and maintenance 
They all converge on finding value in driving less and making fewer car trips. A safety approach that is working in tandem with a climate disruption strategy would focus on reducing driving rather than on distracted driving, for example. Distracted driving campaigns allow us more easily to disconnect driving from other important debates and policy changes. It makes progress on a coordinated approach more difficult.

By counting miles traveled, counting trips, and counting crashes, in these raw totals we have a better analytical framework for talking about transportation than we do when we focus on rates.

This may seem counter-intuitive, as most of the time when we want to assess safety, we want a probability, some kind of rate statistic. But right now, at this moment in thinking about road safety and other debates, a focus on counting statistics will actually serve us better.

Of course someone will object: By focusing on counts, the fastest way to achieve zero deaths is to have zero driving. Well, yes. This isn't exactly reductio ad absurdum, but it does express that our goal might be asymptotic. We should want to make driving the mobility choice of last resort, or a late resort anyway, rather than the default mobility choice of first resort. We should want to drive less and less and to substitute other mobility for drive-alone trips. Sometimes a car really is the best transportation tool. But we gravely overuse it right now. Maybe when we aren't overusing it, then we will have a different debate with different analytical tools.

Vision Zero also asserts that increasing minor fender-benders could be a good trade-off for reducing catastrophic crashes and deaths; rates of crashes might increase while count of deaths would decrease. That's another way that focusing on rates is less useful than focusing on counts.

No comments: