More about missing Hit and Runs

I find any hit-and-run abhorrent. A hit-and-run driver involved in any traffic crash with any injury is a criminal felon; and if the injury is serious or results in death, it’s a serious felony (class 3 or 2). It’s worth pointing out that hit-and-run is a crime for the driver regardless of who caused the crash (i.e. which party had the right of way is irrelevant, except for sentencing).

The background is I noticed here and there over the years what seemed to me to be an odd listing in the traffic database of a fatal bicyclist, and an unknown vehicle with unknown everything else (direction, lane, lighting, etc etc) except for bicyclist demographic and it usually had a location, but was NOT listed as a hit-and-run. I inquired about a few of them and the story was always something like the coroner(I think?) submitted something somehow to traffic records. I brushed it off as something that happens very rarely; some sort of freak occurrence. However. It turns out there are dozens of such cases every year., unsurprisingly most of them are pedestrian.

I believe this is an incorrect coding.  The whole point is they don’t have a crash report that was submitted by law enforcement. If they are, indeed, traffic crashes then they are by definition hit-and-run. In the Arizona traffic database an incident is a binary flag; and in the FARS (national) database it is also, as of 2020, a binary. There is no in between, there is no maybe.

Arizona has a strangely large number of such fatal incidents. Although it’s a fairly small statistical percentage of traffic fatalities, it’s significant under-reporting of fatal hit-and-runs.

Here’s a bunch of extracted data in a google sheet:

'injuriesByPersonType' worksheet  'HR'

In some years, the percentage of peds killed in hit-and-runs are significantly under-reported; e.g. 2018 and 2023 stand out — the official percentage is supposed to be 23 to 25%, but including the ‘missing’ hit-and-runs it’s more like 33 to 34%

Is Arizona really special?

In the years where FARS allowed an “unknown” result for HIT_RUN (in addition to yes/no. Before 2020), year-after-year Arizona accounted for something over half of the unknowns hit-and-runs in the entire county(?!). E.g. 2019 of the 26 nationally, 16 of them were in Arizona; 2018 it was 36 out of 64. To put it another way, in most years, most states reported ZERO unknowns, and a handful of states reported just a few. (Florida had several in some years; but then again FL is a large and populous state)

One other anomaly jumped out, in that in 2016 and 2017, Connecticut (state 9) had a huge number (64 and 56 respectively. Plus 13 in 2018) whereas in all the other years (2010-2015, 2019) it reported ZERO unknowns. Is anybody keeping an eye on this data?

Definitions

MMUCC 6th Edition (2024) V39. Hit and Run / Element Definition:
Refers to cases where the vehicle or the driver of the motor vehicle in transport is a contact vehicle in the crash and departs the scene without stopping to render aid or report the crash.
Attribute Values: (Select 1) No or Yes

Arizona Crash Form Manual box 2f ‐ Hit & Run ‐ A crash may be considered hit and run if any driver involved in the event fled the scene, even if the driver later was apprehended or reported the crash at a later time. Enter the Unit # of the vehicle that fled the scene in the blank. If more than one unit left the scene list all units

In ACIS (Arizona Crash Information System, a database based on the info on the crash form), Hit and Run is a “flag” (either yes or no)

FARS (as of January 2025, this document is no longer available from the federal gov’t; fortunately, it is archived)
V6 Hit-and-Run, SAS name HIT_RUN
Definition: This data element identifies whether this vehicle was a contact vehicle in the crash that did not stop to render aid (this can include drivers who flee the scene on foot). Hit-and-run is coded when a motor vehicle in-transport, or its driver, departs from the scene…

Beginning in 2009, this was on a per-vehicle basis; and the allowable values were not just 0 (no) or 1 (yes) but also unknown (an 8 or a 9); what’s weird is as of year 2020 data and going forward, they removed the unknown., so we’re left with just 0 and 1. I have no idea why, but it seems odd/wrong. The FARS manual for 2020 simply says “removed attribute” without explanation.

Queries

Hunt for common/re-used asdm FileNumbers might offer a clue that the incident might not have come from a law enforcement filed crash report; besides 0 (or might have been NULL), the only number that recurs more than a few times are the 4 or 2 digit year (typically at least dozens of times a year). I also wonder why, but haven’t looked into it, as to why suddenly around 2015 and going forward there are thousands of ‘0’ file numbers??

SELECT FileNumber, count(1) FROM asdm.2023_incident GROUP BY 1 ORDER BY 2 DESC LIMIT 10;

As expected, Pedestrians suffer more hit and runs, this query gives an overview of Hit and Run involved fatalities and injuries. For example in 2018 there were 60 Nonmotorist HR fatalities but only 15 motorist HR fatalities:

SELECT SUM(TotalFatalities), SUM(TotalInjuries), SUM(TotalNonMotoristsFatalities), SUM(TotalNonMotoristsInjuries), SUM(sF_Pedestrian), SUM(sF_Bicycle), SUM(sF_Motorcycle) FROM 2023_incident WHERE HitAndRunFlag=1;

Compare “unknown” Hit and Runs from the FARS vehicle table (a value of 9. In some years, 8 was allowed but does not appear anywhere in the data back to 2010), nationally vs. just Arizona. Arizona often “leads the nation” in unknown number of hit-and-runs. Why?
Another aberration in the nationwide FARS data is state #9 (Connecticut), some years particularly 2016 and 2017 suddenly had a huge number of unknown hit-and-runs; went from 0 in 2015 to 64 in 2016, and back to 0 in 2019.  There’s obviously something wrong there.

SELECT HIT_RUN, count(*) FROM fars.2018_vehicle GROUP BY 1; 
SELECT HIT_RUN, count(*) FROM fars.2018_vehicle WHERE STATE=4 GROUP BY 1;
SELECT HIT_RUN, STATE, count(*) FROM fars.2018_vehicle WHERE HIT_RUN > 1 GROUP BY 1,2;

Fatal counts, and broken out by hit and runs, and person type, according to the database. (these match the number and severity spreadsheet):

SELECT count(*) "Ped HR"
FROM 2018_incident AS i JOIN 2018_person AS p ON i.IncidentID = p.IncidentID
WHERE p.eInjuryStatus = 'FATAL' AND p.ePersonType = 'PEDESTRIAN' AND HitAndRunFlag;
SELECT count(*) "Ped not HR"
FROM 2018_incident AS i JOIN 2018_person AS p ON i.IncidentID = p.IncidentID
WHERE p.eInjuryStatus = 'FATAL' AND p.ePersonType = 'PEDESTRIAN' AND NOT HitAndRunFlag;

SELECT count(*) "Bike HR"
FROM 2018_incident AS i JOIN 2018_person AS p ON i.IncidentID = p.IncidentID
WHERE p.eInjuryStatus = 'FATAL' AND p.ePersonType = 'PEDALCYCLIST' AND HitAndRunFlag;
SELECT count(*) "Bike not HR"
FROM 2018_incident AS i JOIN 2018_person AS p ON i.IncidentID = p.IncidentID
WHERE p.eInjuryStatus = 'FATAL' AND p.ePersonType = 'PEDALCYCLIST' AND NOT HitAndRunFlag;

SELECT count(*) "Motorist HR"
FROM 2018_incident AS i JOIN 2018_person AS p ON i.IncidentID = p.IncidentID
WHERE p.eInjuryStatus = 'FATAL' AND (p.ePersonType != 'PEDALCYCLIST' AND p.ePersonType != 'PEDESTRIAN') AND HitAndRunFlag;
SELECT count(*) "Motorist not HR"
FROM 2018_incident AS i JOIN 2018_person AS p ON i.IncidentID = p.IncidentID
WHERE p.eInjuryStatus = 'FATAL' AND (p.ePersonType != 'PEDALCYCLIST' AND p.ePersonType != 'PEDESTRIAN') AND NOT HitAndRunFlag;

These are probably Pedestrian (or Bicyclist) fatalities that have been entered into the database without a crash report (e.g. a report from a hospital or medical examiner or whatever); but are not listed as HR. It looks at only two-unit incidents (i.e. one of the units is an MV, the other is a ped/bike)

SELECT *
SELECT i.IncidentID, i.FileNumber, i.IncidentDateTime, LOVCounty.name County,LOVCity.name City, OffNcic.name Oncic,ExtNcic.name Encic, i.eCollisionManner, i.eLightCondition, i.TotalUnits, i.TotalNonMotorists, i.HitAndRunFlag, i.OnRoad, i.CrossingFeature, i.Latitude, i.Longitude, "maplink", i.eTrafficWayType, i.eIntersectionType, i.eJunctionRelation, i.eWeather, p_bike.UnitNumber b_UnitNumber, p_bike.eNonMotoristLocation, p_bike.eSex b_Sex, p_bike.Age b_Age, p_bike.eViolation1 b_Violation1, p_bike.eSafetyDevice b_SafetyDevice, p_bike.ePersonType b_PersonType, u_bike.eLane b_Lane, u_bike.eTravelDirection b_Direction, u_bike.eUnitAction b_Action, p_car.UnitNumber c_UnitNumber, p_car.eSex c_Sex, p_car.Age c_Age, p_car.eViolation1 c_Violation1, u_car.eLane c_Lane, u_car.eTravelDirection c_Direction, u_car.eUnitAction c_Action 

SELECT count(1)
FROM (((( 2022_incident AS i JOIN 2022_person AS p_bike ON i.IncidentID = p_bike.IncidentID) 
JOIN 2022_unit AS u_bike ON p_bike.UnitID = u_bike.UnitID) 
JOIN 2022_person AS p_car ON i.IncidentID = p_car.IncidentID) 
JOIN 2022_unit AS u_car ON p_car.UnitID = u_car.UnitID ) 
LEFT OUTER JOIN LOVCounty ON i.CountyId = LOVCounty.id 
LEFT OUTER JOIN LOVCity ON i.CityId = LOVCity.id 
LEFT OUTER JOIN LOVNcic AS OffNcic ON i.OfficerNcic = OffNcic.id 
LEFT OUTER JOIN LOVNcic AS ExtNcic ON i.ExtendedNcic = ExtNcic.id 
WHERE p_bike.eInjuryStatus = 'FATAL' AND p_bike.ePersonType = 'PEDESTRIAN' AND p_car.ePersonType='DRIVER' 
AND NOT i.HitAndRunFlag AND p_car.Sex = "U"
AND TotalUnits=2
ORDER BY i.IncidentDateTime;

 

This identifies the quantity of LIKELY suspicious incidents, those that may have gotten into the database WITHOUT a crash report because there are too many unknowns. Before 2018 the counts were just a few, suddenly in 2018 and beyond there were over a hundred/yr! Most of those involve just MVs but the ones with a Ped are significant

SELECT count(*)
FROM 2023_incident
WHERE TrafficWayType=99 AND IntersectionType=255 AND JunctionRelation=99 AND Weather=99;
SELECT sF_Bicycle, sF_Pedestrian, sF_Motorcycle, TotalFatalities, HitAndRunFlag, count(*)
FROM 2023_incident
WHERE TrafficWayType=99 AND IntersectionType=255 AND JunctionRelation=99 AND Weather=99
GROUP BY 1,2,3,4,5;

SELECT count(*) "Ped LIKELY HR"
FROM 2023_incident AS i JOIN 2023_person AS p ON i.IncidentID = p.IncidentID
WHERE  i.eInjurySeverity = 'FATAL' AND sF_Pedestrian
 AND p.ePersonType= 'DRIVER' AND NOT i.HitAndRunFlag AND p.Sex = "U" ;

SELECT count(*) "Bike LIKELY HR"
FROM 2023_incident AS i JOIN 2023_person AS p ON i.IncidentID = p.IncidentID
WHERE  i.eInjurySeverity = 'FATAL' AND sF_Bicycle
 AND p.ePersonType= 'DRIVER' AND NOT i.HitAndRunFlag AND p.Sex = "U" ;

SELECT count(*) "Motorist POSSIBLE HR"
FROM 2023_incident AS i JOIN 2023_person AS p ON i.IncidentID = p.IncidentID
WHERE  i.eInjurySeverity = 'FATAL' AND (NOT sF_Bicycle AND NOT sF_Pedestrian)
 AND p.ePersonType= 'DRIVER' AND NOT i.HitAndRunFlag AND p.Sex = "U" ;


.

References

Hit-and-Run Crashes: Prevalence, Contributing Factors and Countermeasures:  Hit-and-run crashes and fatalities rates have been increasing at an alarming rate. AAA, 2018. This is now old, and probably (?) would show that after 2018-2022 data shows an even more alarming rate of increase.

Fatality Facts 2022, Pedestrians, IIHS. Shows national HR fatal peds were about 20% from 2013 to 2019, but then shot up to ~ 25% by 2022. Arizona’s published 2022 rate of 27% is actually more like 33% if “missing hit-and-runs” are included.

Leave a Reply

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