Time of Day Matters for Auto Insurance Risk

Take the time to get it right with fairer, more accurate premiums

For over two decades, insurance companies have been using driving data to deliver fairer and more accurate premiums in the form of monitored insurance products, commonly called usage-based insurance (UBI).  Market penetration of UBI policies continues to grow.  Several major insurers have publicly stated that more than 40% of new auto business is opting for these types of policies.

In the Q4 2022 Investor Call, Progressive’s CEO, Tricia Griffith, stated “Today, UBI is our most predictive rating variable, and it provides unparalleled rate accuracy to our customers… Part of our road map is to integrate telematics deeper into our model.”  Driving data directly measures auto insurance risk, so it should be more accurate than traditional factors that are proxies for how, how much, when and where a vehicle is operated.  Insurers, like Progressive, that continue to integrate driving data will have a major competitive advantage, subjecting their competitors to adverse selection. 

If insurers want to compete in the future, they will need to develop driving scores that cultivate the full signal from all four components:  how, how much, when and where a vehicle is operated. Most driving scores are still relatively immature; typically, the where and when components are the least developed.  This article highlights the importance of correctly quantifying the risk associated with “when you drive”.

Time of Day Risk

Many of today’s driving scores only surcharge late-night driving and frequently do not vary based on the day of week. TNEDICCA (www.tnedicca.com), a crash data and analytics provider, captures data from police reports and knows precisely when and where over 30 million US crashes occurred.  TNEDICCA uses that data to determine the relative risk of trips based on when and where the trips occur. 

TNEDICCA’s Ohio data was used to explore the “when you drive” component of risk. The following chart shows there is a significant difference in crash risk depending on the time of day and day of week. The lines represent the crash rate relative to the average crash rate, and the bars depict the percentage of traffic volume by hour and day of week. Note, Monday through Thursday were combined due to the similarity of the patterns.

As expected, the crash risk is highest in the late-night hours.  The data suggests that the riskiest hours (Sunday from 2AM-4AM) are about 10 times riskier than the safest times.  The data also confirms that the risk from 2AM-4AM is very different on the weekends than during the week.  In fact, the 3-4AM weekday crash rate is lower than the crash rate during the evening rush hour.  This data supports regulators’ concerns that shift workers may be unfairly impacted by programs that surcharge late-night driving without differentiating by day of week (or using some other mechanism such as repeat trips).

Time of Day Impact on Insurance Risk

The most meaningful rating characteristics identify significant variation in risk for a significant percentage of exposure.  While the prior chart confirms late-night weekend driving is significantly riskier than the rest of the week, it represents less than 1% of the driving.  To understand the practical importance of the time variable, we need to answer the question “is there material variation in crash rates when most of the driving occurs?”

The data highlights that over half of the driving volume occurs during hours when the crash rate is either 10% above or below the average crash rate.  For example, 3PM to 6PM on weekdays represents almost 20% of the driving volume and has a crash rate that is 15-20% higher than average.  Additionally, 4AM to 7AM on weekdays represents almost 7% of the driving volume and has a crash rate that is 40-50% lower than average.  More generally, the average absolute deviation across all hours of the week is more than 15%.   Moving beyond a simple late-night surcharge is clearly a significant segmentation opportunity.

Animal-Involved Crash Risk

Animal-related crashes can be a significant percentage of the claims for Comprehensive coverage.  The following chart shows the Ohio crash rate for animal-related crashes only.  The animal crash rate is much more consistent across the days of the week and is lowest during the high-volume daylight hours

Insurance companies generally price auto policies based on an “annual” risk basis even though they typically write semi-annual policies.  This practice avoids see-sawing of premiums and isn’t consequential assuming most insureds renew their policy.  That said, some carriers have implemented telematics-based policies that can significantly vary month-by-month (e.g., pay-per-mile products). 

The preceding charts aggregated the data over 12 months.  Even if insurers chose to ignore seasonality when determining premium relativities, quantifying the seasonal impact helps product managers better interpret quarterly financial results.  The next chart displays the relative crash rate by month for animal and non-animal crashes. 

Seasonality Risk

Insurance companies generally price auto policies based on an “annual” risk basis even though they typically write semi-annual policies.  This practice avoids see-sawing of premiums and isn’t consequential assuming most insureds renew their policy.  That said, some carriers have implemented telematics-based policies that can significantly vary month-by-month (e.g., pay-per-mile products). 

The preceding charts aggregated the data over 12 months.  Even if insurers chose to ignore seasonality when determining premium relativities, quantifying the seasonal impact helps product managers better interpret quarterly financial results.  The next chart displays the relative crash rate by month for animal and non-animal crashes. 

There is a clear seasonal impact for animal-involved crashes in Ohio. The crash rate is over 250% higher in the fourth quarter than the third quarter.  This is likely due to increased deer activity in the Fall due to mating season.

The magnitude of the animal-involved crash rate relativities disguises the seasonality of the non-animal or “normal” crash rate, but it does exist.  July has the lowest “normal” crash rate and the winter months’ crash rate is about 25% higher than the summer months’ crash rate.   

To better understand the cause of these differences, we compared the weekday hourly crash rate for each of the four seasons: Winter (Dec-Feb), Spring (Mar-May), Summer (Jun-Jul), and Fall (Sep-Nov).   We used the weekday crash rates hypothesizing the summer school holidays were a major contributor to the low summer crash rates and, consequently, any differences would be most noticeable during the week.

The most significant percentage spreads in crash rates occurred during the 5-8AM and 6-9PM time periods.  Interestingly, those three-hour periods in the morning and evening correspond to sunrise and sunset, respectively. 

Monthly data from the Astronomical Applications Department was used to estimate the percentage of the morning (5-8AM) and evening (6-9PM) periods that occurred before sunrise or after sunset for each of the four seasons (see Table).  The percentages appear to be directly related to the seasonal crash rates during those morning and evening hours.  Clearly, improved visibility due to daylight has a major impact on crash rates.   Even if an insurer isn’t going to do seasonal pricing, an insurer using time of day for telematics should consider differences in sunrise/sunset times throughout the year.

Final Thoughts

Why haven’t all driving scores gone further?  Perhaps insurers lack the necessary data volume, don’t think this variance matters, or have intentionally chosen to keep the time-of-day factor simple for non-actuarial reasons.   Whatever the reason, insurers not using “when you drive” or doing so simplistically are leaving potential segmentation on the table. 

Insurers should remember that the world is changing, and people’s individual driving patterns can be very diverse.  Companies who effectively use actual driving data will have a huge competitive advantage versus those who do not.  The provided information could be implemented in several ways resulting in a step-change improvement over what is done today. 

This article hopefully convinced you to take the time to get the time right!