Number and Severity of Arizona bike-MV crashes

[quick update on 2021 Crash Facts data — the pattern continues, the total number of reported cyclist crashes was only 1025. Notwithstanding a historically high fatality count of 43]

[at some point, ADOT began “scrubbing” PDO. As explained in the 12th Edition of the crash from instructions, linked here, effective 8/2019 AZ law bumped up the PDO reporting threshold from $1,000 to $2,000. As can be seen easily at the spreadsheet linked at Crash Facts 2021,  No-injury (and unknown injury) reports became exactly zero commencing with 2019).
In any event the PDO change would not explain all or even most of the dramatic decrease in reported non-fatal bicyclist crashes]

It’s been noted that the number of reported Bike-MV crashes reported by police in Arizona have seen a mysterious sharp decrease; even as the number of MV crashes in general has risen. In round numbers through the 2009-2016 (some of the tables below have been updated to add 2017 & 2018) period, Bike-MV crashes have fallen by about a third, while MV crashes overall have increased by 20%. This time period covers the economic downturn following the Financial Crisis of 2008 and subsequent recovery. That trend has been used to explain the MV crash rate; but it doesn’t seem to have anything to do with Bike-MV rates; which would logically rise with more traffic and more (MV) traffic crashes.

Fewer crashes would be a good thing of course, but it feel like there’s something else going here. Also, the number of fatalities, while statistically small,  remained relatively flat but above their longer-term trend of 24 (rolling 10 yr average 2007-2016; refer to internal spreadsheet \crashReports\asdm\AZstats.xls)

Number of Arizona bike-MV crashes 2009-2019

*  presumably depressed from pandemic, explored more here.

The large drop-off in Bike-MV figures came in a significant drop in 2014, followed by another significant drop in 2015. Later on I’m going to refer to the “before-time” as 2009-2013, and the “after-time” as 2015 & 16 (& 17). The was a yearly drop of 543 cases/yr in the after-time compared to the before-time.

The crash reporting system recognizes a 5 step injury severity scale: No Injury, Possible Injury, Non-incapacitating Injury, Incapacitating Injury, and Fatality. Except for Fatality, the categories are somewhat subjective, and it represents a snap-judgement of the officer filling out the crash report.

Unpacking the Bike-MV by injury severity reveals that there were significant drops in the number of  least-serious (No Injury, Possible, and Non-Incapacitating Injury) crashes, and the more severe category of Incapacitating Injury generally didn’t change much. (Fatals are difficult to intuit because they are very small in absolute numbers, and aren’t useful statistically). Here are the 5 severity categories by year (TODO: collapsed into one table)

Smoking Gun?

If we divide the time period into two epochs with 2014 as the dividing line, the before-time (2009-2013) compared to the after-time (2015 & 2016)  saw a drop (or missing?) of 543 crashes/yr on average.

There are 9 jurisdictions that contribute about 3/4ths of all reported crashes across the state, the other 1/4th comes from dozens of smaller jurisdictions (comfortingly, this hews closely to the Pareto principle of 80/20). Each of the Big 9 is broken out separately in the next section.

Something suspicious is happening in 7 of those 9 places, so in ‘Chandler’, ‘Flagstaff’, ‘Glendale’, ‘Gilbert’, ‘Mesa’, ‘Scottsdale’, ‘Tempe’ (the big 9 excluding Phoenix and Tucson);

— Concern & Disclaimer: is this “Data Mining”? & I almost failed statistics in college —

The number of lesser severity crashes in the before-time is several hundred less per year (392/yr using averages, nearly a 50% reduction!) than in the after-time. This suggests in at least several of these 7 jurisdictions some sort of policy change occurred around the year 2014 that discourages or otherwise doesn’t result in a crash report being filed for lesser severity crashes. There was a drop in serious+fatal but those numbers are both far smaller (question statistical significance) and the drop was fractionally small (that data is not presented below, but it’s readily available upon request).

That 392/yr drop in these 7 places constitutes the lion’s share of the statewide drop of 543/yr. [there’s some remarks below about per city population as a proxy for expected crash rates]

It’s possible, of course, that there could have been a dramatic decrease around 2014 in the number of such crashes in these 7 communities. Statistics can’t tell us whether a drop is a crash that didn’t happen, or merely one that wasn’t reported.

.

The Big 9

The nine jurisdictions, ‘Chandler’, ‘Flagstaff’, ‘Glendale’, ‘Gilbert’, ‘Mesa’, ‘Phoenix’, ‘Scottsdale’, ‘Tempe’, ‘Tucson’ account for around 3/4ths of all reported crashes.

Population-wise the Big 9 are 55% of the state’s population, while the Big 9 excluding Phoenix and Tucson is only 24%. Note that Peoria is high population, 154,065 just a bit fewer than Tempe, with relatively few bike crashes. (2010 census data; there’s a worksheet in the spreadsheet crashReports/asdm/cityTraffic.xls not necessarily online)

.

Below are results by each of the nine cities, along with overall counts per city, as well as counts by injury severity.

The non-Nine jurisdictions total counts (all severities) were very consistent; though they did fall by ~ 50/yr before-time vs. after-time.

The table below is just raw total counts for all the larger cities —

.

Below is a long dump (sorry for the poor formatting) of most larger jurisdictions, and counts of bike-MV crashes by year and severity

City of Phoenix

As a related corollary, I had pulled together figures for  City of Phoenix only; and made a table of motorist / Peds / Bicyclists killed and injured… (2019 data for SUM of total motorist injuries is incorrect for 2019?)

The number of reported bicyclist injuries have been relatively flat throughout the period. Pedestrian injuries are up, and Ped killed is up sharply.

Queries

select IncidentYear "Year", SUM(sF_Bicycle) "Bike-MV", SUM(sF_Motorist) "all MV" from incident GROUP BY 1;
select IncidentYear, count(*) "Bike-NoInj" from incident WHERE sF_Bicycle AND eInjurySeverity LIKE ('NO_INJURY') GROUP BY 1;
select IncidentYear, count(*) "Bike-PossInj" from incident WHERE sF_Bicycle AND eInjurySeverity LIKE ('POSSIBLE_INJURY') GROUP BY 1;
select IncidentYear, count(*) "Bike-Injury" from incident WHERE sF_Bicycle AND eInjurySeverity LIKE ('NON_INCAPACITATING_INJURY') GROUP BY 1;
select IncidentYear, count(*) "Bike-IncapInj" from incident WHERE sF_Bicycle AND eInjurySeverity LIKE ('INCAPACITATING_INJURY') GROUP BY 1;
select IncidentYear, count(*) "Bike-Fatal" from incident WHERE sF_Bicycle AND eInjurySeverity LIKE ('FATAL') GROUP BY 1;

SELECT IncidentYear, eInjurySeverity "injSeverity in selected jurisdictions", count(1)
FROM ( incident AS i )
LEFT OUTER JOIN LOVNcic ON i.OfficerNcic = LOVNcic.ID
WHERE  sF_Bicycle AND LOVNcic.name IN ('Chandler', 'Flagstaff', 'Glendale', 'Gilbert', 'Mesa', 'Scottsdale', 'Tempe') GROUP BY 2,1 ;

SELECT IncidentYear, count(1) "No + Possible + Non-Incap Injuries in selected jurisdictions"
FROM ( incident AS i )
LEFT OUTER JOIN LOVNcic ON i.OfficerNcic = LOVNcic.ID
WHERE  sF_Bicycle AND LOVNcic.name IN ('Chandler', 'Flagstaff', 'Glendale', 'Gilbert', 'Mesa', 'Scottsdale', 'Tempe')
AND eInjurySeverity IN ('NO_INJURY', 'POSSIBLE_INJURY', 'NON_INCAPACITATING_INJURY' )  GROUP BY 1 ;

SELECT IncidentYear, count(1) "Bike-MV in Big 9 Jurisdictions"
FROM ( incident AS i )
LEFT OUTER JOIN LOVNcic ON i.OfficerNcic = LOVNcic.ID
WHERE sF_Bicycle AND LOVNcic.name IN ('Chandler', 'Flagstaff', 'Glendale', 'Gilbert', 'Mesa', 'Phoenix', 'Scottsdale', 'Tempe', 'Tucson') GROUP BY 1;

SELECT IncidentYear, count(1) "Bike-MV All Other Jurisdictions"
FROM ( incident AS i )
LEFT OUTER JOIN LOVNcic ON i.OfficerNcic = LOVNcic.ID
WHERE sF_Bicycle AND LOVNcic.name NOT IN ('Chandler', 'Flagstaff', 'Glendale', 'Gilbert', 'Mesa', 'Phoenix', 'Scottsdale', 'Tempe', 'Tucson') GROUP BY 1 ;

SELECT IncidentYear, LOVNcic.name "Jurisdiction", count(1)
FROM ( incident AS i )
LEFT OUTER JOIN LOVNcic ON i.OfficerNcic = LOVNcic.ID
WHERE sF_Bicycle AND LOVNcic.name IN ('Chandler', 'Flagstaff', 'Glendale', 'Gilbert', 'Mesa', 'Phoenix', 'Scottsdale', 'Tempe', 'Tucson') GROUP BY 2,1 HAVING count(1) > 1 ;

SELECT IncidentYear, LOVNcic.name "Jurisdiction", eInjurySeverity, count(1)
FROM ( incident AS i )
LEFT OUTER JOIN LOVNcic ON i.OfficerNcic = LOVNcic.ID
WHERE sF_Bicycle AND LOVNcic.name IN ('Chandler', 'Flagstaff', 'Glendale', 'Gilbert', 'Mesa', 'Phoenix', 'Scottsdale', 'Tempe', 'Tucson') GROUP BY 3,2,1 ;

SELECT IncidentYear, eInjurySeverity "injSeverity other than the Big 9", count(1)
FROM ( incident AS i )
LEFT OUTER JOIN LOVNcic ON i.OfficerNcic = LOVNcic.ID
WHERE sF_Bicycle AND LOVNcic.name NOT IN ('Chandler', 'Flagstaff', 'Glendale', 'Gilbert', 'Mesa', 'Phoenix', 'Scottsdale', 'Tempe', 'Tucson') GROUP BY 2,1 ;

Here are queries for City of Phoenix vs. State; note that these are people counts, not crash counts (reminder; to build table, run as “all years”); (2nd reminder, as of 2019 the field TotalMotoristsInjuries somehow inexplicably became a count of every motorist, not just injured ones; so to correctly find that number, simply subract TotalNonMotoristsInjuries from TotalInjuries):

SELECT IncidentYear "Phx -----", SUM(TotalMotoristsFatalities), SUM(TotalNonMotoristsFatalities),
SUM(TotalInjuries), SUM(TotalMotoristsInjuries), SUM(TotalNonMotoristsInjuries)
FROM ( 2019_incident AS i )
LEFT OUTER JOIN LOVNcic ON i.OfficerNcic = LOVNcic.ID
WHERE LOVNcic.name ='Phoenix' ;

SELECT IncidentYear "Phx peds ", SUM(TotalMotoristsFatalities), SUM(TotalNonMotoristsFatalities),
SUM(TotalInjuries), SUM(TotalMotoristsInjuries), SUM(TotalNonMotoristsInjuries)
FROM ( 2019_incident AS i )
LEFT OUTER JOIN LOVNcic ON i.OfficerNcic = LOVNcic.ID
WHERE LOVNcic.name ='Phoenix' AND sF_Pedestrian ;

SELECT IncidentYear "Phx bike ", SUM(TotalMotoristsFatalities), SUM(TotalNonMotoristsFatalities),
SUM(TotalInjuries), SUM(TotalMotoristsInjuries), SUM(TotalNonMotoristsInjuries)
FROM ( 2019_incident AS i )
LEFT OUTER JOIN LOVNcic ON i.OfficerNcic = LOVNcic.ID
WHERE LOVNcic.name ='Phoenix' AND sF_Bicycle ;

SELECT IncidentYear "all AZ -----", SUM(TotalMotoristsFatalities), SUM(TotalNonMotoristsFatalities),
SUM(TotalInjuries), SUM(TotalMotoristsInjuries), SUM(TotalNonMotoristsInjuries)
FROM ( 2019_incident AS i );

SELECT IncidentYear "all AZ peds ", SUM(TotalMotoristsFatalities), SUM(TotalNonMotoristsFatalities),
SUM(TotalInjuries), SUM(TotalMotoristsInjuries), SUM(TotalNonMotoristsInjuries)
FROM ( 2019_incident AS i )
WHERE sF_Pedestrian ;

SELECT IncidentYear "all AZ bike ", SUM(TotalMotoristsFatalities), SUM(TotalNonMotoristsFatalities),
SUM(TotalInjuries), SUM(TotalMotoristsInjuries), SUM(TotalNonMotoristsInjuries)
FROM ( 2019_incident AS i )
WHERE sF_Bicycle ;

.

.