Understanding Collision Summaries

Cities and the state (ADOT) issue various reports with regard to traffic safety, e.g. the City of Phoenix issues the Bicycle Collision Summary, the city of Mesa calls theirs the Bicycle Crash Analysis. Phoenix also likewise publishes a Traffic Collision Summary. Here are links to a selected list of such reports:

ADOT : currently offer a yearly .pdf, 1997 – 2008: Motor Vehicle Crash Facts no bicycle-specific report is available.
Phoenix Streets Dept : currently offering the 2007 .pdf: Traffic Collision Summary, Bicycle Collision Summary, and Pedestrian Collision Summary. There are a number of prior years available, but for some strange reason they do not seem to archive them. Here is the year 2000 bike that I have saved.
Mesa Transportation Dept , Bicycle Program : I can’t find either a general traffic, or a ped report. But the Bicycle Program does have Bicycle Crash Analysis 2008 .pdf.
I am looking in particular for Tempe but can’t find it
Tuscon: can’t find just the city, but PAG 2001-2008PAG Bicycle Crash Analysis 2001-2006 ; PAG Bicycle Crash Analysis 2001-2005

The National Center for Statistics and Analysis of the National Highway Transportation Agency — the NCSA of the NHTSA? — has the crash forms, and instructions that are the basis for reporting here. In particular, the Arizona Crash Form Manual is enlightening because it shows what data gets collected.


No usage data

There is no bicycle usage data — so nothing can be said about the rate of bicycle collisions or injuries, except for per capita rates. The Phoenix report addresses this directly “unlike automobile crash data we have no direct measure of how many bicycles were on the road… We have no understanding, therefore, of the bicycle crash rate”

No distinction of injury severity

There is no distinction in type of injury; the only outcomes are injury, no injury, and fatality.(i need to check: does aliss have any place for injury severity?)

No crash typing

There is no “crash typing” a la Cross and Fisher. Only the aggregate number of relative position; e.g. 32.6% of cyclists were on the sidewalk just prior to the crash, or 47.8% of motorists were turning right just before the crash. But they are not correlated, you can only hypothesize that this amounts to a lot of right hooks.

Can’t draw conclusions, or propose countermeasures

Since there is no rate, injury severity, or typing — nothing useful can be drawn as to the risk posed by the various categories presented. For example: does the data show that it is riskier to ride on the sidewalk or in the street? The answer is resounding; you can’t tell from this data.

Quality of data — too much “other”

This goes without saying, but the quality of the report depends entirely on the quality of the data. The main data is gleaned from ALISS. (The Phoenix Bicycle Collision Summary also claims that Phoenix police reports are also used?). Under the “violations” (which is more accurately called the Violations/Behaivor, since several of them are not violations) section, the bicycle data purports to show that bicyclists are far more often at fault than motorists. A look at the data shows a disturbingly high number of “Other” violations/behaviors for bicyclists as opposed to motorists. For example of the 480 Phoenix collisions in 2005, there were only 15 Other for motorists but 83 Other for bicyclists. Numbers in the Mesa report were very similar, 10 vs. 60 out of 276 collisions. This makes me suspicious that a large number of these PARs (Police Accident Reports) are not be thoroughly completed and/or liability is being assigned without regard to applicable statutes (Title 28), but rather some ill-conceived notion of the way bicyclists “ought” to ride. There is a section of the form where the investigating officer lists out the actual citations issued — I don’t know if this information is captured in ALISS.

Other thoughts

Sometimes there are nuggets of wisdom, e.g. this from the Phoenix Traffic Collision Summary: “Sharing the roadway is one of the most common events in society today yet it can be the most dangerous thing most Americans do every day. It also remains one of the least appreciated dangers in life”.

From A Compendium of NHTSA’s Pedestrian and Bicyclist Traffic Safety Research Projects
NHSTA DOT HS 810 793 , my emphasis added: “Prior to the landmark crash type identification studies of Snyder and Knoblauch (1971) and Cross and Fisher (1977) , pedestrian and bicycle crash taxonomies were largely based on the demographics of the event: the victim’s personal information (age, gender, etc.), crash location, time of day, day of week, weather conditions, road geometry, injuries incurred, and other data typically contained on a police accident report. While informative and useful, this type of classification system is not very effective in revealing why crashes occur and how to prevent them”.

Other Links: there is a collection of crash-statistic links here.

Long ADOT-sponsored report: Remedies for Driver Error Final Report 567. The abstract begins with this peculiar statement: “Driver error is estimated to cause about half of all traffic accidents in Arizona and the United States”. There is no source for this, they apparently they are relying on the “No Improper Driving” category derived from collision reports (ALISS i would guess?). This just shows that this data is incomplete — police reports simply don’t assign fault or error in many cases… as the report notes: “Given the fact that ‘No improper driving’ and ‘Not stated’ were reported in 52 percent of cases, statistics on driver error may be underestimated”. Nevertheless, almost without exception (an earthquake perhaps?) an error was obviously made by a driver (or to a much lesser extent, by a bicyclist or pedestrian). Equipment failure is rare.

“Across the United States, driver error is estimated to cause between 45 percent and 75
percent of all crashes and is a contributing factor in the majority of crashes (FHWA, 2002)”. The reference points to Driver Error Report FHWA-RD-02-06 with a dead link so I need to look that up (but there’s another similar title Identification and Evaluation of Driver Errors, FHWA-RD-02-003). Elsewhere the 45 and 75 percent figures are referenced to Hankey, J. M., Wierwille, W. W., Cannell, W. J., Kieliszewski, C. A., Medina, A., Dingus, T. A., and Cooper, L. M. (1999). Identification and evaluation of driver errors: Task C report; Driver error taxonomy development. (Contract No. DTFH61-97-C-00051). Washington, DC: Federal Highway Administration.

For Further Reading

There are many (easily, thousands) of crash studies floating around out there. Here is a particularly valuable and thoughtful one, the Orlando Area Bicyclist Crash Study: A Role-Based Approach to Crash Countermeasures. It in detail breaks down 885 crashes over a two year period (2003-4)  categorizing according to Pedestrian and Bicycle Crash Analysis Tool (PBCAT). The 17 fatalities are detailed seperately. There is an intersting reference to “group ride” behavior and its potential impact on crash rates: approximately 0% (zero percent). Of the 885 crashes, only8 involved multi-cyclists and motorists. And in 6 of the 8 incidents the motorist was found at fault.

On the too-much-other problem; apparently this is also unsurprisingly a factor in pedestrian crash analysis: “In addition, there
were many cases coded as “Other/Unknown” for the pedestrian action in the FARS…” DOTHS809585

Pretty interesting subject: “This thesis explores the way in which the criminal law deals with drivers who kill…”, it explores a bunch of cases in Britian.

Britian’s TRL has issued a large study. Here is a news story about it: Risky cycling rarely to blame for bike accidents, study finds. This confirms what I am seeing on a smaller basis — most cyclist serious injury and death is clearly the fault of a motor vehicle driver. e.g. survey of 2009 AZ fatalities; pre-preliminary results.