COVID-19 Impact on Auto Crashes, An Update
My “Impact of COVID-19 on OH Auto Claims” post highlighted the significant reductions in OH crashes based on data provided by TNEDICCA. My friends at TNEDICCA continue to amass and analyze accident data daily and are allowing me to provide an update incorporating Texas data (my home state). This update provides deeper insights based on additional data elements and a first look at the impact of a “phased re-opening”.
Changes in Statewide Crash Counts
The following chart compares recent daily Texas crash counts to the same period in 2019 with the added context of cumulative confirmed COVID cases:
TNEDICCA calculated the percentage changes in crashes in half-month increments. The following chart shows the changes annotated with key COVID-19 related dates to better understand the effect on accident rates during different stages of the crisis:
The direct relationship between key COVID-19 dates and the percentage change in crashes is readily apparent. As we continue to open our economy and people drive more, it will be interesting to track this data to see if we return to 2019 crash levels or whether there is a “new normal”.
As I wrote in a prior post, insurance companies rely on historical data to project future costs when setting prices for traditional auto insurance. Given this an anomalous situation, the historical data is less relevant creating a major challenge for auto insurers trying to respond with credits, refunds, etc. Furthermore, any analysis of future rate needs will be difficult without more information on the “new normal”. These are bigger issues for the medium and small insurance companies who cannot quickly generate enough “new” data to uncover clear signals.
Changes in crash counts by time and geography
TNEDICCA explored the data at a granular level to quantify the reductions by geography and time of day. First, TNEDICCA plotted the change in crashes versus the population for each county and found the reduction was more significant in more populated counties:
Crashes in the counties with 1 million or more residents dropped 2.1 times more than crashes in counties with less than 5,000 residents (-54.6% versus -25.9%). The number of confirmed COVID cases per 10,000 residents is approximately 2 times higher in the more populous counties. Since the large counties were more at risk and implemented stay at home restrictions early, it is not surprising that the miles driven (and thereby crashes) would be more significantly reduced in the larger counties.
When actuaries determine territorial rates for traditional auto insurance products, they often start by aggregating loss experience in small building blocks (e.g., zip code or census tract). Even large insurers struggle with having enough data in each building block to draw credible conclusions without using multiple years of data and employing smoothing techniques. Actuaries will need to consider whether data from this period should be excluded from future territorial analysis.
TNEDICCA also examined the change in crashes by time of day/day of week which I have summarized into five categories:
COVID-19 social distancing restrictions intuitively led to larger reductions in miles driven during rush hour (with work at home) and late at night on weekends (with bars being closed). The crash data confirms the number of crashes changed more significantly during the higher-risk rush hour and late night/weekend times (-61% combined) versus all other times (-47% combined).
Time of day tends to be a common factor used in telematics-based insurance programs. Insurers with such programs should consider the potential impact this may have on their policyholders’ telematics risk score. My post “Is UBI the COVID-19 Economic Cure for Auto Insurers” explains how different types of telematics-based policies will react to changes in driving behavior.
Potential Impact on Severity of Crashes
TNEDICCA has done two things to help us understand this based on data from other states. (Texas did not have the data necessary to study severity.)
First, they compared the percentage of crashes with injuries during this period with the same period last year and found no discernable difference. In other words, injury-related crashes have decreased proportionately with overall crashes.
Second, they examined the reported speed of the crash. They found that the average speed of crashes during this crisis is 27.1 mph as compared to 21.8 last year. This, of course, suggests that accidents will result in more damage and more severe injuries.
TNEDICCA does not have data on fraud or repair costs. That said, it is reasonable to assume that severity may be increased due to supply chain shortages that increase costs or fraud as businesses or individuals inflate insurance claims because of hardship caused by the economic crisis.
Much of the focus has been on the drastic reduction in crash frequency due to fewer miles driven. While the frequency has clearly been much lower, this data suggests the severity may be higher than normal. Since frequency is a binary result (i.e., yes or no claim) and severity can be many different values (i.e., the final claim cost), it will be more difficult for actuaries to accurately quantify the severity impact.
Auto insurers are busy analyzing their own data to quantify COVID-19 impacts for purposes of refunding and setting future rates. Since these are unprecedented times, they cannot rely on historical data to uncover clear signals.
TNEDICCA possesses up-to-date crash data provides key insights:
Crashes have been significantly reduced considering the COVID-19 restrictions, but are already rebounding with the opening of the economy
The reductions in crashes have been larger in more populated counties
The reductions in crashes have been larger during high-risk driving times
The severity of crashes will likely be higher than normal during this time.
TNEDICCA has similar data in other states. If you are an insurer struggling with a lack of internal data or just want to benchmark your results, TNEDICCA can help you. Feel free to contact me or Yiem (TNEDICCA’s CEO) for more information.