Coronavirus and traffic crash stats

How will the novel corona virus lockdown affect traffic safety?

CORRRECTIONS : The graphics and data here are based on the state’s dataset as of mid-February 2021 — it turns out there were a significant number of fatalities that were not yet reported, so many so that the fatal count for 2020 is HIGHER (not lower); see here for final numbers. If I have time I will re-run the numbers and re-create the graphs.

This streetsblog article had a random assortment of related factoids; from a grab-bag of places. One theory has it that with the significantly reduced vehicle traffic, the remaining drivers are driving faster; thus offsetting what should be significant safety improvements.

Another additional theory that seems to bear out in the AZ data is while overall vehicle travel is down a lot, the riskiest behaviors (e.g. DUI, unlit bicyclist crossing street in the dark, etc) perhaps haven’t changed much; this would support why fatalities are down just a small amount, while all other severity categories (overall, incapacitating injury, injury, etc) are down much more. This is true for not only motorists, but bicyclists and pedestrians, too.

Traffic safety rates, particularly for motorists, are usually considered in terms of VMT, Vehicle Miles Traveled by motorists. This number is quite accurately measured, but isn’t yet available (at least in the places I’ve looked). It’s quite likely that this most-watched indicator: overall fatalities divided by VMT is going to go UP, because it’s expected that fatalities haven’t fallen as much as VMT [see the NHTSA reference, below]

For Arizona bicyclists, it’s more of a mixed bag; the number of bicyclist fatalities last year was nearly exactly the same as in previous years — it’s actually up a small amount — but since the number is quite small, about 30 / year, it jumps around a lot from year-to-year, and changes are not necessarily significant. And while its reasonable to assume that all fatal traffic crashes get reported, it’s well known that lower severity crashes go underreported in normal times. Due to the pandemic it’s quite possible there is more underreporting, but at the same time the more severe the injury the more likely it will be reported. The number of incapacitating injuries is down by a lot. (30 to 40% compared to previous 5 years).

Here are some month-by-month crash counts; first all crashes, then bicyclist injury (injury, incapacitating injury, or fatality); as compared to the average for the same month over previous 5 years…

I would argue anecdotally that there are significantly more bicyclists during the year 2020, than past years. I feel like I’ve seen multiple bicyclists every trip I take, when in other years it seems like I unusual to see any.  Because of the pandemic, many people  have way more free time on their hands to ride a bike or walk around, and at the same time far fewer other (car) trips. As such the bicyclist injury and fatality rate per trip is probably safer.

 

Notes

See this spreadsheet for detailed counts which underlie the graphics above.

All data is from ADOT’s ACIS (Arizona Crash Information System); data for 2020 was retrieved 2/17/2021 and is not yet final (In other words, there will be a tiny percentage of stragglers that are yet to be reported into the database). (as noted above, see CORRECTION, the stragglers on fatals were significant this year)

As such, the incident counts reflect ONLY traffic crashes reported to or by police. The pandemic has probably reduced the likelihood that people involved in a crash will report it to police; particularly for low-severity crashes (though that has always been the case; i.e. there are no fatal undercounts, but lots of no-injury undercounts)

As mentioned above, this is raw counts; there are no rates.

Counts are incident counts, not persons count

References

NHTSA: Early Estimate of Motor Vehicle Traffic Fatalities for the First Half (Jan–Jun) of 2020 DOT HS 813 004

The order “Proactive Measures To Protect Against COVID-19” was issued March 19, 2020 Arizona’s “Stay Home, Stay Healthy, Stay Connected” orders specifically allowed unlimited outdoor exercise. This was issued March 30 when traffic was already greatly reduced

There’s an excellent collection of data driven new on the GIcycle  — “Geographic Information support for bicyclists”, and often deals with the important topic of cyclist/ped counts.

Unacast offers mobility data down to the county-level. Though i wonder if traffic here (in the Phx metro) is now as of late May close(r) to what it would normally be in a late May. Winter visitors would be gone by now, and schools would be done by now as normal. Would be better to compare year-over-year if they have that data?

The queries (for my internal reference):

# By month: note these are INCIDENT, not person, counts; 2020 data as of Feb 2021 but not yet final.
SELECT IncidentMonth, IncidentYear, count(*) "All" FROM incident GROUP BY 1, 2 ORDER BY IncidentMonth;
SELECT IncidentMonth, IncidentYear, count(*) "Bike all incl no injury" FROM incident 
 WHERE sF_Bicycle AND InjurySeverity IN (1,2,3,4,5) GROUP BY 1, 2 ORDER BY IncidentMonth; 
SELECT IncidentMonth, IncidentYear, count(*) "Bike, injury" FROM incident 
 WHERE sF_Bicycle AND InjurySeverity IN (3,4,5) GROUP BY 1, 2 ORDER BY IncidentMonth;
SELECT IncidentMonth, IncidentYear, count(*) "Bike, incap injury" FROM incident 
 WHERE sF_Bicycle AND InjurySeverity IN (4) GROUP BY 1, 2 ORDER BY IncidentMonth; 

# by year: note these are INCIDENT, not person, counts; 2020 data as of Feb 2021 but not yet final.
SELECT IncidentYear, count(*) "All, incap severity" FROM incident 
  WHERE InjurySeverity IN (4) GROUP BY 1 ;
SELECT IncidentYear, count(*) "All, fatals" FROM incident 
  WHERE InjurySeverity IN (5) GROUP BY 1 ;
SELECT IncidentYear, count(*) "Motorist, fatals" FROM incident 
  WHERE NOT (sF_Pedestrian OR sF_Bicycle) AND InjurySeverity IN (5) GROUP BY 1 ;
SELECT IncidentYear, count(*) "Motorist, incap severity" FROM incident 
  WHERE NOT (sF_Pedestrian OR sF_Bicycle) AND InjurySeverity IN (4) GROUP BY 1 ;
   SELECT IncidentYear, count(*) "Bike, incap severity" FROM incident 
  WHERE sF_Bicycle AND InjurySeverity IN (4) GROUP BY 1 ;
SELECT IncidentYear, count(*) "Bike, fatals" FROM incident 
  WHERE sF_Bicycle AND InjurySeverity IN (5) GROUP BY 1 ;
SELECT IncidentYear, count(*) "Ped, incap severity" FROM incident 
  WHERE sF_Pedestrian AND InjurySeverity IN (4) GROUP BY 1 ;
SELECT IncidentYear, count(*) "Ped, fatals" FROM incident 
  WHERE sF_Pedestrian AND InjurySeverity IN (5) GROUP BY 1 ;

SELECT IncidentYear "March - Nov", count(*) "Bike, fatals" FROM incident 
  WHERE sF_Bicycle AND InjurySeverity IN (5) AND (IncidentMonth > 2 AND IncidentMonth <12) GROUP BY 1 ;

#yearly; but restricted to specific time-of-year, chosen for covid traffic reductions 
SELECT IncidentYear "Mar 15 - Nov", count(*) "Bike, injury" FROM incident WHERE sF_Bicycle AND InjurySeverity IN (3,4,5) AND ( 
(IncidentDate > "2009-03-14" AND IncidentDate < "2009-12-01") OR 
(IncidentDate > "2010-03-14" AND IncidentDate < "2010-12-01") OR 
(IncidentDate > "2011-03-14" AND IncidentDate < "2011-12-01") OR 
(IncidentDate > "2012-03-14" AND IncidentDate < "2012-12-01") OR 
(IncidentDate > "2013-03-14" AND IncidentDate < "2013-12-01") OR 
(IncidentDate > "2014-03-14" AND IncidentDate < "2014-12-01") OR 
(IncidentDate > "2015-03-14" AND IncidentDate < "2015-12-01") OR 
(IncidentDate > "2016-03-14" AND IncidentDate < "2016-12-01") OR 
(IncidentDate > "2017-03-14" AND IncidentDate < "2017-12-01") OR 
(IncidentDate > "2018-03-14" AND IncidentDate < "2018-12-01") OR 
(IncidentDate > "2019-03-14" AND IncidentDate < "2019-12-01") OR 
(IncidentDate > "2020-03-14" AND IncidentDate < "2020-12-01") ) GROUP BY 1;
   
# Just for reference (not in attached spreadsheet):
#seasonal fluctuation
select  IncidentYear, IncidentMonth, count(*) from incident 
where sF_Bicycle AND InjurySeverity IN (3,4,5) GROUP BY 1,2;

# stuff about ReceivedDate
select IncidentDate, ADOTReceivedDate from incident 
where IncidentYear=2019 AND sF_Bicycle AND InjurySeverity IN (3,4,5) AND IncidentMonth = 2 ;

# looks like stuff flows into database same day as received...
select ADOTReceivedDate, count(*) from incident 
where IncidentYear=2020 AND sF_Bicycle AND InjurySeverity IN (3,4,5)
GROUP BY 1 ORDER BY ADOTReceivedDate;

.