Preliminary observations and important questions about traffic safety during COVID-19 based on provisional weekly police-reported injury crashes on state highways in California
How to study traffic safety in a rapidly changing environment?
The COVID-19 pandemic has had an abrupt impact on many aspects of our lives, including mobility. This has a direct impact on exposure (i.e., the level of activity on the roads), but can also trigger other responses that can affect road user behavior. Due to the rapid rate of change there is a need to monitor things at a higher frequency than what we are used to in the traffic safety community. As an example, if we look at the Centers for Disease Control and Prevention (CDC) Provisional Death Counts for Coronavirus Disease (COVID-19), two slightly different narratives emerge based on the unit of observation.
Provisional death counts for COVID-19 (CDC)
The first narrative is derived by looking at the total deaths from February, 2020 and shows a cumulative single digit percent increase in death counts relative to prior years. The second arises when we observe it at a weekly level. Here, we are able to more explicitly see the progression and impact of this pandemic, which reveals an increase of 30 percent or more over what was expected in some weeks. Moreover, since the weekly death toll is regularly reported by the media, it can have short-term implications on travel behavior due to different routines and higher levels of anxiety. Accordingly, the corresponding impact on safety is expected to be realized on a comparable time interval. Considering this, efforts to study the safety impacts of COVID-19 should try to evaluate it on a relatively short time scale, when possible.
What data are we able to get?
Following the above logic, UC Berkeley SafeTREC has been aggregating injury crash data and exposure data on a weekly level as part of the launch of an Injury Crashes During COVID-19 dashboard on our Transportation Injury Mapping System (TIMS). The crash data is coming from police-reported crashes that are added to the California Statewide Integrated Traffic Records System (SWITRS) by the California Highway Patrol (CHP). There is a delay related to submitting, processing, and tabulating crash data into SWITRS. The delay is too restrictive for crashes that occur on non-state roads so those are excluded here, but for crashes on the state highway the delay is more timely. However, due to this lag, the data shown in the dashboard may be missing some relevant crashes, particularly those occurring in the most recent 2-3 weeks. The exposure data is based on Vehicle Miles Traveled (VMT) as provided by Caltrans Performance Measurement System (PeMS) and is limited to loop detector reliability but does not suffer from a reporting lag. We also gathered crash and VMT data for the corresponding weeks last year.
What are we looking at?
The data is shown for three geographical areas. One for all the state highways in California and two covering the two main urban metros (including 6 Bay Area Counties and 5 Southern California Counties).
Overall statewide summary table (TIMS)
The main summary tables include a comparison between 2019 and 2020 of the Total crash counts, Fatal and Severe crash counts, and VMT. Additional tables include a breakdown by primary collision factors with an emphasis on Unsafe Speed, Improper Turning, and DUI crashes and a breakdown for pedestrian and bicycle crashes. All these data can be examined along two avenues: (i) crash frequency or rates; and (ii) crash distributions in terms of severity or types of crashes. Due to the reporting delay, the frequency or rates should be limited to identifying important patterns and comparisons to 2019 should be done with caution. However, findings regarding the crash distributions can be compared to 2019 with more confidence.
What are we learning so far?
An initial descriptive analysis of the data reveals two leading observations:
there is a noticeable reduction in injury crashes along state highways in California. While it is too soon to analyze this data and produce a point estimate, it is already apparent that the frequency of crashes has changed dramatically starting from the week of March 16, 2020 when California’s stay home order went into effect. This very clear change-point splits the first six weeks of our data demonstrating a provisional average of 1,122 injury crashes per week (2/3/20-3/9/20) with the following six weeks that show a provisional weekly average of 544 (3/16/20-4/20/20). While we expect relatively more injury crashes to be added to the latter period, it is unreasonable to think that this large gap will become small.
reduction in fatal & severe crashes is smaller than the reduction in lower-level injuries. We can quantify it by comparing the odds between the six weeks prior to the stay at home order and the six weeks after. For fatal and severe crashes the odds are 0.71 compared to 0.46 for other injury crashes. This results in an odds ratio of 1.55 and indicates that during the stay at home order the odds of a crash being fatal/severe crash is 1.55 more than it was before. While this result is less sensitive to reporting delays, it can still be biased if the reporting delay is different across levels of severity (e.g., it takes longer to complete a report for a severe crash). Noting that the proportion of fatal/severe fluctuates through the latter period makes it reasonable to expect that this result is not driven by reporting delays. The finding that injury crashes along the California state highways have become more severe can be caused by different factors. It can be because people's driving behavior has gotten worse and they are taking more risks (e.g., a lot more excessive speeding citations per CHP), but can also be driven by fewer congestion-related injury crashes which usually involve a minor injury.
What are some important questions moving forward?
While this preliminary descriptive analysis is not mature enough to understand the impact of COVID-19 on traffic safety, it helps us make some timely observations and also draw a path for more follow-up research.
As we continue to track safety data during this complicated and unprecedented period, we identify three important questions to consider moving forward:
as things open up and mobility and VMT bounce back, what safety data do we need to monitor and how do we need to respond to make sure things don't end up higher than what they would have been without a pandemic?
what is happening to safety on non-state roads?
how is California different from other states that have seen an increase in fatalities despite a comparable reduction in mobility and why?
We will keep updating the data on the dashboard to allow others to analyze it too, and will also be generating new research based on using this data with the intention to identify insights that can help save lives on California’s roads.
Today's blog post was from UC Berkeley SafeTREC Co-Director Offer Grembek. Learn more about Dr. Grembek here.