[ Check out Paul Schimek’s visualization of FARS bicycling data ]

Cindie Holub’s death on March 1,  (Cindy’s death was written up on bicyclelaw.com, also see 2010 fatalities), from injuries sustained in a Feb 24 collision with a garbage truck caused me to look up the rule for categorization purposes. “To be included in this census of crashes, a crash had to involve a motor vehicle traveling on a trafficway customarily open to the public, and must result in the death of a person (occupant of a vehicle or a nonmotorist) within 30 days of the crash.” from DOT HS 811 137.

The US DOT runs a very elaborate, publicly available, query-able database for every traffic fatality in the US called FARS — Fatality Analysis and Reporting System.

There is a web interface that leads the user through building a query, e.g. below is a tabular report for all State=4 (Arizona), and Person Type=6 or 7 (bicyclist or other cyclist) in 2008, 20 cases were found for 19 collisions (this is a slight discrepancy of the reported 19 figure in the official NHTSA state-by-state report).

some handy info can be gleaned, e.g. the gps coordinates for most incidents are listed, you can use google to map it, e.g. 32.81931389, -111.68361944 is the intersection of I-8 and I-10.

obs Case# Date Hit&Run? Latitude Longitude Road Age
1 18 01122008 2 32.22082778 -110.78518056 BROADWAY BLVD 14
2 18 01122008 2 32.22082778 -110.78518056 BROADWAY BLVD 14
3 58 02122008 0 33.55228611 -112.22566389 NORTHERN AVE 46
4 79 02292008 0 32.16764444 -110.95620000 PARK AVE 43
5 255 04302008 0 33.58914444 -112.14539167 CHOLLA ST 5
6 284 05202008 0 32.31297500 -111.03741389 CAMINO DE LA TIERRA 43
7 463 07092008 2 33.53823889 -112.05129722 GLENDALE AVE 40
8 505 08132008 0 32.81931389 -111.68361944 I-10 50
9 506 08142008 0 32.25069722 -110.83816389 TANQUE VERDE RD 54
10 553 08292008 0 35TH AVE 28
11 566 09072008 2 32.16296111 -111.00652500 MISSION RD 50
12 573 09122008 0 32.26481389 -110.95234167 FORT LOWELL RD 15
13 585 09232008 0 33.50880556 -112.23706389 CAMELBACK RD 69
14 604 09272008 0 32.26353889 -110.96099722 1ST AVE 46
15 616 10042008 0 SR-87 40
16 652 10092008 0 US-89 38
17 666 09172008 2 33.49768056 -112.09130278 15TH AVE 56
18 673 09202008 0 CASA GRANDE HWY 60
19 699 10262008 2 33.66721667 -112.24961667 BEARDSLEY RD 65
20 825 12052008 0 32ND ST 42

One of these 20 (case number 18) was a seriously injured cyclist — thus the total for 2008 is 19 fatalites despite there being 20 rows of information.

Here is a really cool visualization map map.itoworld.com of all 2001-2009 fatals. (they also have UK)

The Technical Details

The gory details of how the data is coded is in Electronic 2009 FARS Coding and Validation Manual, HS 811 353; along with the 1975-2008 FARS Analytic Reference Guide HS 811 352. These are both huge documents loaded with nitty-gritty details and both seem to cover similar material.

I’ve never had much luck with the web query front end; and for big jobs I download the actual datasets which are zipped together as a bunch of .dbf files. These files can be read by many applications; I used a (very old, from Office 97) Microsoft Excel and it worked fine. Incidentally I could not open the large table with OpenOffice ver 3.2 — it apparently limits to 64K rows of data and the one set is too big.

Statistical Hole?

“”To be included in the file set, a crash had to involve a motor vehicle traveling on a traffic way customarily open to the public” — FARS Analytic Reference Guide.

Where, then, do bicyclist traffic fatalities that involve either parked motor vehicles (see e.g. this one on 12/20/2008 ), or no motor vehicle at all (reported 4/27/2010 loss of control, near Sedona) get recorded? Apparently, and its not completely consistent, the answer is FARS doesn’t usually record non-roadway / non motor-vehicle involved fatalities (one counter-example is in 2009, there was one cyclist killed in a driveway, and that was in FARS). Ohiobikelawyer.com details a couple of motorist fatalities in 2010 where bicyclists were involved; one incidentally, and the other was listed as the cause of the fatality (April 30, and July 11 respectively). It will be interesting to see how/if these get coded in FARS 2010 when it becomes available.

Jane C. Stutts and William W. Hunter 1999 study Injuries to Pedestrians and Bicyclists: An Analysis Based on Hospital Emergency Department Data, (the FHWA contact is Carol Tan Esse) says that:

More bicyclists are injured in bicycle-only events than in collisions with motor vehicles

And in their data, out of 1,066 incidents, only 280 (26%) were bicycle-MV road (so-called quadrant A), Table 7. And perhaps noteworthy is that bicycle-MV non-road was quite insignificant (23 incidents).But of course many many of those were not serious, in Table 14 they split it three ways; Treat&Release, Admitted, Fatal — i.e. from least to most serious. Most of those Admitted (68/120) and nearly all Fatals (5/6) were bicycle-MV road (as i would have expected), versus the other three possibilities. Here is the way they couched the fatals in their intro/literature seach:

The Rodgers (1993; 1995) and Baker et al. (1993) studies also made comparisons between bicycle-related deaths reported by the National Center for Health Statistics (NCHS) based on death certificate data and the FARS data based on State police crash reports. Generally, these comparisons revealed that FARS contains 8-10 percent fewer bicycle-related deaths than does the NCHS database. This was attributed to the fact that approximately 10-14 percent of bicycle-related fatalities do not involve a motor vehicle and/or do not occur on public roadways. These findings are supported by earlier analyses carried out by the Centers for Disease Control and Prevention, showing that 90 percent of bicyclist fatalities involve motor vehicles and 86 percent occur on public roadways (Sacks et al., 1991)

So I think the answer to the question: How many bicyclists are killed NOT as a result of MV collision? Is like 10% based on these studies they cited above (Rodgers and Baker, and Sacks).Or in terms of sweeping statements; I think it is fair to say something like:

Nearly all bicyclists are killed as a result of a MV collision.

Unreported and Underreported Crashes

While nearly all fatalities get reported, less severe of course tends to be significantly not reported in some fashion. There is also a separate category of excluded crashes for various reasons, i have a list of such non-traffic crashes.

Interesting material here from National Telephone Survey of
Reported and Unreported Motor Vehicle Crashes 812183.pdf

NHTSA publishes crash statistics based on police accident reports, but many crashes are never reported by the police. In 1981 NHTSA sponsored a telephone survey to estimate the incidence of unreported crashes. That survey estimated 47 percent of crashes go unreported. In 2008 NHTSA paid for an updated survey, reported here. The present survey, completed in 2010, collected data on 2,299 crashes, 697 of which were unreported to police. When the data were properly weighted, the participant responses indicated that approximately 30 percent of crashes go unreported. In both surveys the crashes were mostly property-damage-only crashes, although some unreported injury crashes were found. However, the data in this report is only one aspect of the unreported crash problem. See Chapter 5 of The Economic and Societal Impact of Motor Vehicle Crashes, 2010 (DOT HS 812 013), for a more thorough explanation of this problem.


10 thoughts on “FARS”

  1. List of known ‘missing’ traffic Fatalities from FARS?
    Milton Olin Jr / Dec 8, 2013
    Inthis comment, BJToepper, correctly noted this well-known Dec 8, 2013 CA bicyclist fatality is missing from FARS; given what I was told by ADOT about it being too late (in early 2017) to fix/correct some 2015 data, I would surmise this one will be forever missing… The driver was a L.A. County Sheriff’s Deputy (LASD); victim Milton Olin Jr; Calabasas on 22400 block of Mulholland Highway. Indeed, Toepper is correct, various searches turn up nothing in FARS
    select * from incident where MONTH=12 AND DAY=08 AND YEAR=2013 AND sF_Bicycle;
    select * from incident where TWAY_ID LIKE “Mulholl%” AND sF_Bicycle;

    Incidentally, NHTSA data comes from state-level police reports, which for cyclists mostly only contain crashes that rise to high injury levels, and even then miss deaths. Here in L.A., for instance, the widely-reported death of Milt Olin still has not appeared in SWITRS (which feeds into NHTSA), now nearly four years after the fact. One wonders what else the official record has missed that newspapers did not.

    Here’s one I thought might be missing but isn’t; news reports March 1,2015; there were two that day in FARS:
    select * from incident where MONTH=03 AND DAY=01 AND YEAR=2015 AND sF_Bicycle
    This is it: 15 / 150016;
    The initial news reports has some comment about sunrise and being hard to see — the crash was 6:25A. Sunrise that day was like 6:43A so, no, it wasn’t a rising sun issue. The cop/driver that killed was arrested and apparently fired; later news stories say he didn’t stop at the scene and was initially charged on suspicion of evidence tampering. The last mention of the trial at this time (early 2018) seems to be this May 2017 story “Buddemeyer, of Pahoa, pleaded not guilty in October to charges of first-degree negligent homicide, tampering with physical evidence and making a false report to law enforcement in connection with the March 1, 2015, crash that killed Jeffrey Surnow, a 69-year-old visitor from Michigan…Police say Buddemeyer was operating a subsidized patrol car eastbound and was on-duty when his vehicle struck and killed Surnow, of West Bloomfield Hills, Michigan… An officer involved in circumstances matching Buddemeyer’s has since been terminated from the force, according to a discipline report that did not name him”
    There’s obviously alledged wrong-doing that has NEVER been in the press. Other news reports said “First responders with the Hawaii Fire Department initially reported the incident as a hit-and-run”. Well, was it? (it’s not according to FARS) Anyway, the civil case Surnow_et_al_v_Buddemeyer_et_al has activity as recently as Nov 2017. This is going to cost the taxpayers a bundle.

    “Nevram” gave me these samples as potentially missing in this f.b. comment:
    2014-02-19 at approximately 6:40 p.m., San Bernardino / Kendall Drive [update: this one is in fars, case 2014 / 60713
    Sat. 2014-12-27 Riverside, CA / Central Avenue (update: indeed, cannot find a bicyclist fatality this date; need to check some other searches like motorized bike? other dates? other streets?)
    monday 2014-12-29 ~ 6AM San Bernadino, CA / East Lynwood Drive
    (update: this one is in fars, case 2014 / 62316 )

    N Hollywood (LA) Vanowen Street near Longridge Avenue, 2014-11-04 fatal (electric) bike dooring. Not in FARS. This is obstensibly correct since did not appear to involve an MV in transport. I DID verify it also was not listed as a motorcycle/moped/motorized bike. In the event an elec bike is considered an MV, it should have been included. It’s not clear to me how those distinctions get made and whether they vary from state-to-state. See here for a diatribe about motorized bike reporting, it’s conceptually possible ebikes are supposed to be lumped with mopeds and motorized bikes, which get coded as motor vehicles and not bicycles.

  2. data project here
    by @nrippner
    but ended with 2018 data… there are scripts here that were used to assemble the data that may be helpful…

    Some possibly relevent chatter here:
    https://opendata.stackexchange.com/questions/2028/fatality-analysis-reporting-system “I recommend using ETL tool (Extract Transform and Load) such as Talend Open Studio (TOS) or Penthao (both are open source solution)”

Leave a Reply

Your email address will not be published. Required fields are marked *