The roads are not the same

Line graph showing trends in aggressive acceleration in 2019 vs 2020
The pandemic has caused driver behavior change that extends beyond reduced driving mileage.

It seems like there are so many changes happening right now that it’s hard to keep up. As we try to balance a return to “normal” alongside the continued risks posed by COVID-19, America is entering a new phase of its response to the COVID-19 pandemic.

Some cities have reopened restaurants and parks and other businesses, while others are still enforcing broad social distancing policies. On top of the response to the pandemic, people are coming together to demand solutions for the racial and economic inequality that Black Americans face. These historic events interact with each other in complex ways.

In prior posts, we discussed the drop we saw in core driving metrics, such as miles driven, and how driving changes at the local level are important to watch. Since then, we have seen driver behavior change again, with an upward trend in miles driven and many areas are now returning to pre-COVID-19 levels of driving. In fact, in states like Wyoming, we are actually seeing more traffic than we’d expect due to increased tourism as national parks reopen.

Total miles driven
Daily deviance by state over time

Line graph showing total miles driven sharply declining after March 10, 2020

We know that driving risk is measured not only by how long people are on the road but also by the choices people make during trips. Driving safely includes limiting your speed, predictable and smooth acceleration, and deceleration, avoiding high-risk hours, and, of course, setting down the phone.

At Arity, we are passionate about using data from phone and vehicle sensors to understand driving and accurately identify risky behaviors. Our driver data helps make roads safer and insurance pricing more transparent and accurate. In the charts below, our data shows driver behavior change during the pandemic. Here are some highlights:

  • The pandemic hasn’t changed how frequently we use our phones while driving, which is way too often.  
  • Aggressive driving behaviors such as sudden acceleration and braking are down as roads have become less congested.  
  • People are speeding more on both side streets and highways, including driving over 100 mph more frequently than they did pre-COVID-19. 
Aggressive acceleration
Indexed (100=Feb 02)

Line graph showing trends in aggressive acceleration in 2019 vs 2020

Aggressive braking
Indexed (100=Feb 02)

Line graph showing trends in aggressive braking 2019 vs 2020

Phone use
Indexed (100=Feb 02)

Line graph showing trends in phone use 2019 vs 2020

Driving over 100mph
Indexed (100=Feb 02)

Line graph showing the difference in driving over 100mph in 2019 vs 2020

As with mileage, we see a variance in behaviors between states. For example, in Texas and New Jersey, we see different responses to the pandemic likely due to different road conditions and driving cultures.  

Aggressive acceleration, Texas
Indexed (100=Feb 02)

Line graph showing the difference in aggressive acceleration in Texas in 2019 vs 2020

Aggressive acceleration, New Jersey
Indexed (100=Feb 02)

Line graph showing the difference in aggressive acceleration in New Jersey in 2019 vs 2020

Driving over 100mph, Texas
Indexed (100=Feb 02)

Line graph showing the difference in driving over 100mph in Texas in 2019 vs 2020

Driving over 100mph, New Jersey
Indexed (100=Feb 02)

Line graph showing the difference in driving over 100mph in New Jersey in 2019 vs 2020

The Arity Drivesight score combines these different driving behaviors into a single measure of driving risk that is highly predictive of insurance losses. We score driving trips each day to view driving risk and driver behavior change. From our driving scoring during the height of the pandemic response, we found that driving risk per mile driven increased a small, but noticeable, amount.

Driving risk during COVID
Rolling weekly average

Line graph showing driving risk levels during the beginning of the Covid-19 pandemic

Going beyond predicting driving risk, Arity also measures some outcomes directly. The Arity crash detection service uses machine learning algorithms and sensors on mobile phones to detect potential crashes. Arity customers use crash detection to alert friends, family and, sometimes, emergency services of potential accidents. For this analysis, we used the data from the crash detection service to see how the frequency and severity of crashes has been affected amid driver behavior change.

Unfortunatelythe percent of highspeed crashes is increasing. This is important because we know crashes at higher speeds are more dangerous than those at low speeds and typically result in more injuries and damageThis is one possible cause for the 14% increase in the fatality rate per mile recently reported by the National Safety Council.  

Speed at time of collision
Confirmed claims before and after COVID

Bar chart showing speed at time of collision before March 13, 2020 and after March 13, 2020

As people return to pre-COVID-19 driving levels, we must watch whether their driving behaviors also return to previous levels. This will be a crucial insight in order to stay on top of the actual outcomes that insurers will have to manage.

Stay safe on the roads!  

Headshot of Rob Nendorf
Rob Nendorf
Rob Nendorf is the Director of Data Science at Chicago-based subsidiary, Arity. He has led data science and software development projects within the Allstate Enterprise since 2013. Rob received his Ph.D. in Mathematics from Northwestern University in 2011.