Hosted April 15, 2020: Part one of a multi-part series
How did driving behavior shift due to the COVID-19 lockdown? In this webinar, Arity experts share an in-depth driving behavior analysis on how mobility behaviors and risk have changed throughout the pandemic.
Speakers:
Arity’s webinar delves into the impact of COVID-19 on transportation, featuring insights from Louisa Harbage, Grady Irey, Rob Nendorf, and Megan Klein. The discussion covers significant reductions in driving behavior, with daily active connections and miles driven dropping over 50% since early March. The analysis highlights variations by state and population density, showing more substantial declines in urban areas. Changes in driving behavior, such as reduced aggressive acceleration and increased speeding on highways, are also noted. Megan Klein discusses potential impacts on the auto insurance industry, including repair costs, medical expenses, claim development, and fraud. The webinar concludes with a Q&A session addressing data specifics and future insights.
Louisa Harbage:
Welcome everyone. I’m conscious we still have quite a few people logging in, but we also have quite a bit of content to share with you, so we’re going to go ahead and jump right in. I’d like to thank you all for joining Arity for our first ever webinar, looking at the impact of COVID-19 on transportation. I’m Louisa Harbage, the director of market intelligence and strategy, and we at Arity are really looking forward to spending the next hour with you sharing some insights into how things have changed in light of the current pandemic. We have a few logistics before we get started. You’ll notice it’s very quiet on the line. We do have all of your phone lines muted. There are quite a few people on the phone, so we’re trying to manage the amount of background noise that you hear. So, if you have questions, we encourage you to type them into the Q and A, or the chat panel. Either one works and either one will go just to the panelists on this call, so they won’t be broadly distributed.
We will have a little bit of time for Q and A at the end and we’ll try to answer some questions there, but if you type them in, the other nice thing is that we can take any of them offline so we can potentially answer them after the meeting as well. We will also be sharing a copy of the slides with all of our attendees, so those will come through via an email directly to all of you who have registered tomorrow.
So, if you haven’t gotten something before then, that’s when to expect it. Before we get started, we do have a quick legal disclaimer. So, as I’m sure you are all familiar with, when we conduct forums like this, we need to be very mindful of adhering to both the letter and the spirit of all antitrust laws.
So, we’re asking that you all please read this statement that we’ve prepared to that effect. This will guide all of our discussions here, so it’s assumed that everyone has read and will be following it. In summary, essentially what it says is that we’ve designed this presentation in such a way to ensure that no competitively sensitive information will be disclosed and that there’s no opportunity for any anti-competitive understandings. That means we’ll stick pretty closely to our outlined agenda and not discuss any individual company or industry trends, or rates, or other competitive information, and that you should not either. All right, so, with that, I’m delighted to introduce Grady Irey. Grady is the oversight for our data science and analytics group at Arity, and he and his team are responsible for business analytics, data quality, radiant services, and basically all the data science supporting the Arity platform and the product teams. Grady…
Grady Irey:
Thank you, Louisa. Good afternoon, everybody. Arity is a mobility data and analytics company that provides data-driven solutions to companies invested in transportation to enable them to make mobility services smarter, safer, and more economical. Insurance companies, automobile OEMs, shared mobility companies, and governments turn to Arity to better understand driving behavior, manage risk, operate more safely, and ultimately increase their bottom line. The Arity platform is built on more than 280 billion miles of historical, anonymized driving data; more than 23 million active telematics connections; and 10 years’ experience analyzing driving data from cars and from mobile devices. You might ask, “Why is Arity sharing insights into the impact of COVID-19 on driving behavior?” Well, a lot has changed for all of us in a short amount of time, both personally and within our industry, and with our telematics capabilities and partnerships, we have unique insight into the impact COVID-19 is having on personal mobility.
We want to share the insight we have to help you assess and navigate the change. For those of you who engage with Arity on a regular basis, I expect we’ll be having extended discussions about these insights and more over the next few days. For those of you who are new to Arity, we hope you’ll reach out to us to do the same. Let me introduce you to our speakers today. First, Rob Nendorf is the director of data science at Arity. Rob leads the data scientists, data engineers, and analysts across the company who turn our driving data into meaningful insights. He has led data science as well as analytics deployment initiatives since 2013. Rob received his PhD in mathematics from Northwestern University in 2011. Megan Klein is the director of actuarial and rating services at Arity. Her team is responsible for the actuarial support of Arity’s telematics models enabling insurance companies to execute on their goals around telematics. Now, here is the outline for today’s discussion. First, Rob will guide us through an overview of mobility trends. Then, Megan will address how the pandemic may impact the insurance industry. We’ll reference some resources you may find interesting as you continue to monitor COVID-19-related events. And finally, we’ll answer as many questions as we can in the time we have left. Rob, and Megan, and I will all be available for questions at the end and we’ll be joined by Arity’s Director of Insurance Solutions, Katie DeGraaf. Now, let me turn it over to Rob.
Rob Nendorf:
Great, thanks Grady, and thanks everyone for joining us remotely. So, I’m going to talk through the trends that we’re seeing in our data across the country and kind of share what impacts we see of the disease. So, just to talk a bit about the dataset that’s behind this analysis itself. All of the insights and data in this presentation are based on representative samplings from Arity’s multisource dataset. So, that includes anonymized and aggregated driving behavior data from multiple insurance and non-insurance sources. It’s definitely not solely reflective of any particular industry or source. And our data is collected from both mobile apps with our software installed and onboard device methods that are inside vehicles. So, this is just a quick visual heat map that just shows the geographic coverage kind of cut off at the continental United States. And it’s really just there to say we have full coverage of the U.S. This is just based on a month’s worth of trips, for instance. We have over 23 million connections through both our insurance and non-insurance partners. So, we have frankly more data than we need to do this analysis at a countrywide level and we have the data we need to dig in and do this analysis at a very granular level, so, things like state, even county, or even census tracked. Okay, so, let’s get into some of the countrywide numbers and trends that we’re seeing. So, these next few slides will be some views of our core driving metrics on a daily basis. I’m going to go slow to go fast on this first one and just make sure everyone understands what’s in the chart here and then we’ll go through some of the other metrics. So, this is daily active connections. So, this could be a smartphone or an onboard device, and active means it sent us at least one trip for that day. So, that’s what daily active means. The blue curve is the actual numbers of active connections we’re seeing on a daily basis over time. So, starting early February and going through April 12th, so, fairly recently. We normalize the numbers, so that whatever the values of that metric, whatever the value of that metric was on February 2nd is set to a hundred. So, this really keeps the view of the relative differences, but isn’t showing the absolute values.
The dotted gray line is what we call the counterfactual. So, this just means what is our view of what would’ve happened with this metric if the pandemic didn’t occur. And we create this counterfactual using time series forecasting methods. So, looking at the 2020 data and forecasting forward and also using 2019 data to inform that forecasting. And so, you can see the counterfactual proceeds along mildly increasing over time as we get through March and April, whereas the 2020 actuals start to drop extremely quickly around March 8th to March 10th and then start to stabilize and stay down, but start to level off around late March, so around March 27th say. And so, we see over a 30% drop in daily active connections. So, that’s just the connections that are returning at least one trip for that day. So, the orange curve below shows that difference between the blue curve and the dotted line below there so you can see kind of directly what the magnitude is. So, let’s look at a few more of these. So, this is trips per connection. So, of those connections that are recording trips, how many trips are they recording on a daily basis? And again, you can see the counterfactual go along and continue to increase mildly, whereas the 2020 actuals drop significantly around exactly the same time around March 8th to March 10th, they start really dropping, and then around the 27th they stabilize, but are still going down as we head into April. So, fewer trips obviously leads to fewer miles and those trips may be shorter themselves. So, miles is what we really focus on. And so miles driven per active connection also drops significantly over this time period. Again, it drops right around the same time and starts to stabilize around the same time in late March. We’re seeing even a slight uptick in miles driven per connection, but generally it’s been stable since around the 27th of March. So, we’re seeing about 30% down on a daily basis per active connection. So, this is of the connections that returned at least a trip during the day, how many miles did they travel that day? And so, when you put miles driven per connection and daily active connections together, you get the total miles driven for the population that we’re measuring and this is what that looks like. And so, very similar, all these graphs have very similar shapes. The
signal’s very clear, but this one is actually down more so total miles driven is down over 50% over the time period that we’re looking relative to the counterfactual. And so, again, it starts to stabilize on the 27th, but it’s stayed down and it’s been down over 50% each day since the end of March. So, while the overall population metrics are interesting, it’s important to note that not everyone has exactly the same experience as the overall population average. So, most people have reduced their travel, but some people are driving about the same as what they used to or even more, so 9% of our connections are actually driving more and taking more trips. Fifteen percent have more miles. So, this is kind of the power of having more longitudinal data as opposed to just getting traffic aggregations or what not. We can see how our users who are kept anonymized in all of this analysis are behaving relative to the population.
So, let’s start to break things out by state. So, those are all countrywide numbers up to this point. So, here’s a map with all 50 states and the District of Columbia showing the percent change in daily mileage as of April 12th. So, this is the daily difference from where we think the counterfactual would be—so where we’d be without the pandemic. And there’s some things that you’d expect here. States like New York and California where the prevalence of the disease is higher and have more urban areas are down more than a state like, say, Idaho, which is down less. So, let’s take a closer look at that. So, on the right hand side here, the length of the bars represents that current deviance from the counterfactual and so you can see the actual numbers for each state. And then, the bars are colored by the prevalence of the disease, which we chose to use the number of cases of COVID per 100,000 people, so we adjust for population. And so, right now, that ranges from one to New York’s at about a thousand, actually it might’ve just tracked over a thousand within the last day or so. And so, you can see some correlation of darker colors, so more prevalence of the disease and the longer bars. So, places where the disease is more prevalent, you’re getting more extreme reduction in driving behaviors. So, I’ll just take a second. I know people want to find their favorite state in this plot and look exactly where they fall. So, I’ll give everyone a second just to look at this.
So, just to call out some of the numbers on the high and low end. So, New York is down nearly 70%. D.C., similarly. Vermont, and Michigan, New Jersey trail that just a bit. Then, on the other end, we have states like Idaho, Utah, and other states that are on the lower end and the lower end, we’re seeing kind of 45 to 50% down as of the most recent snapshot in this data, which is April 12th.
So, this is a different view of the state breakout numbers. Here, we’re looking at how those numbers tracked over time, similar to what we showed in the countrywide metrics. So, there’s a whole kind of mess of spaghetti of curves going along over time. Most of them we’ve just grayed out so that you can see a few of the highest and lowest impact states. So, again, this is total miles driven over time. This is the daily deviance from the counterfactual and this is just showing the change, not the two actual versus counterfactual. And so, one thing to take away from this plot is that New York started dropping a bit earlier than most states, but most states dropped about the same time in the grand scheme of things. So, right starting that March 8th, certainly by the 13th or 14th, most of the states were starting to move into that dropping phase and they all dropped about at the same rate and have all started to kind of stabilize. But, there are differences. Some states were earlier, and some states were later, and some states have dropped to lower overall total miles driven than others. So, a lot of the states went through the same kind of general pattern, but there are material differences.
So, we also wanted to overlay some of the timings of the stay-at-home orders and see how they might’ve played a role. So, on the right hand side here, you can see the first stay-at-home order was not for the entire state of California. It started in San Francisco. But, as these orders rolled out over time, I think that the first obvious thing to notice is that we were already down 40% countrywide before the first such order was put in place. So, it certainly is not the case that these orders were the cause of the change in driving behavior that we’re seeing. And then, what that blue curve represents is any trip that was in a state where one of these orders were in place, we switched them from the gray curve to the blue curve so people move from gray to blue over time. So, what you can see is as these orders go in place, if you are in one of these states with an order, you do show up a little bit lower than the states that didn’t have an order in place at that day, and it might contribute to overall staying lower in terms of driving behavior, but it’s not a super obvious signal. I think this one will be really interesting to keep track of as potentially states start to remove these orders and see the changes in behavior by state. We’ll be keeping a close eye on that.
So, as we process the data, we attach a lot of metadata and attributes to the data so that we can analyze it later. Here’s an example of that. So, to every trip we look at the trip start location and say what census tract is that location in, what county is that location in, and what state is that location in. So, obviously you’ve been seeing the results of state level analysis in the last couple slides. So, this is one that’s based on census tract. So, by looking at the census tract to the trip start, we can then attach the population density for that trip start, and then we can group trips based on the population density. So, that’s what you’re seeing on the Y-axis. So, it goes from a density of less than 100 people per square mile and then in groups 100 to 500 and so on, all the way up to the largest density is if you’re over 3000 people per square mile.
And so, then, we can look at the distribution for users of changes in miles driven and that’s what you’re seeing mapped out here. And so, obviously there’s a strong correlation between density and the change in miles driven. So, up at the top, at low density areas, that’s where you’re seeing the drop is more 40%, give or take some, and then by the time you get to the highest density areas, you’re up above 60%, give or take some. So, there’s a lot of overlap, but there’s obviously strong signal that if you’re in more urban, higher density areas, you’re going to likely be driving less than if you’re in a rural or lower density area.
So, let’s turn our attention to driving behavior. So, beyond capturing these core driving metrics that are really about how much you’re driving, we also have a very deep understanding of the risk of the drivers the road. So, what we’re looking at here is based on Arity’s Drivesight 2.0 score, and that score is based on your historical trips. So, it’s a pretty stable view into your driving behavior. It doesn’t change on a daily basis, or anything like that. And so, what we did was we looked at the week starting March 2nd, so that was the week, kind of like the last week of normal driving behavior before the COVID impacts really started to change things. And then, we compare it to the last week we have in the data, so that’s the week that starts April 6th. And what we did was we looked at those two populations which are different, right? Because we know that fewer people are on the road, now—those daily active connections went way down. And so, we asked ourselves is that population—does it somehow have a different mix of risk driving risks than the overall population from before the pandemic? And it’s hard to see that there’s actually two kind of density curves here that show you the distributions. So, the distributions are almost right on top of each other. And so, the answer is no. There hasn’t been a significant change in the mix of driving risks for those who are still on the road. So, some people might’ve thought, Hey, maybe folks who are less risky drivers, maybe their ability to work from home or generally stay home and avoid driving was greater than riskier drivers. And so you might expect that mix of driving risks to change over time. And so, what we found is that we’re not seeing that as of today. And so, you can see the mean score is moving a bit, but it hasn’t moved in a significant way.
So, digging in more to behaviors that do change and can change on a day-to-day basis, these are some more granular driving behaviors. So, at the top left we have basically the rate at which people are driving at extremely high speeds. So, this is a little bit different than just saying someone’s speeding relative to the speed limit. This is someone who’s going really fast regardless of what the speed limit is. And so, what you can see is the rates between 2020 and 2019 on a daily basis tracked fairly well together. And then, now, as of late March, we’re starting to see some differences. And so, in 2020 the rate at which people are driving at extremely high speeds is down a bit. You’ll see in another slide where speed relative to speed limit, there’s a bit of a different story there. So, this is extremely high speeds.
And so, below that you have aggressive accelerations. And so, the rate at which people are having these aggressive accelerations is definitely down in 2020 and it drops right around the same time that a lot of these core driving metrics dropped, right? It drops around March 8th and then kind of bottoms out around the 22nd. It’s actually gone up a bit since then, but basically we’re seeing less aggressive acceleration on the road and that seems pretty clearly due to COVID-19 just given when it drops in 2020. On the other hand, aggressive breaking behaviors have tracked pretty closely with 2019, so we haven’t seen a significant difference there. And then, phone use in 2020 is actually a bit above 2019, but there are issues in our data for this event in terms of how we capture this and how it interacts with features of the OS, which consistently change over time. So, this is definitely one where more analysis is needed and I would probably put that caveat for all four of these graphs. Like, we’re going to continue to dig into this, validate, and I think we’re starting to have some preliminary results. We wanted to share those out, but definitely more to come on these.
OK, so I promised a different view of speeding and here it is. So, we have the ability to look at the road segments that all of our trips are driving on and use that geospatial context to layer on a lot of things that we can use in the analysis. And so, one of those is the posted speed limit for that road that you’re driving on. And so, this is a look at the distribution of speed relative to that posted speed limit. Again, kind of before and after the pandemic was in full effect. And these look pretty close to each other, but, you have to remember this is on a very large population of connections. And so, that shift to the right is actually significant and we’re seeing a shift of about between two and three miles per hour on average between these two different times periods. And that, we think, is significant at a population level. We do see some differences for different road types. So, in the top-right plot, we see the same picture but restricted from all roads—which is what’s on the left—to roads that are high-volume, high-speed roads, and there you see a significant effect, whereas the speed relative to posted limit for low volume, low speed roads is different. Of course, you’re not going to see as much of a change because the speeds are lower in general, but we’re seeing a stronger effect on the highways.
So, this is the first set of charts where we’re actually showing you the unsmoothed daily driving data. So, anyone who’s looked at telematics data would have realized in the first chart that I’ve shown that the numbers were much too smooth and didn’t have any day of week variation to be showing the raw daily counts. And so, here we’re looking at those raw daily counts and in 2019 you can see these big changes between weekdays and the weekends where you get these big drops in aggressive acceleration. And so, that’s a pretty typical feature of looking at driving data. And what you see is starting around early March, maybe mid-March for aggressive acceleration in 2020, you lose that day-to-day variation, those big spikes go away. And, in fact, every day sort of looks a bit like the weekend. And so, this makes sense, the roads are not nearly as congested as they used to be.
We don’t have nearly as many people that have to commute to work. And so, you don’t get that weekly commuting driving behavior that we’re used to seeing. And so, this is why we put up there, the weekends and the weekdays have blurred together for a lot of people. I know that resonates with me. And so, you see similar patterns for driving at high speeds and even just miles per trip where you don’t see the big peaks from a day-to-day, weekly standpoint that we’re used to seeing in the data. And so, it’s not just that these high-level trends have changed significantly and overall driving is down significantly. It’s also that some more granular behaviors have also changed and that’s what we’re seeing here. So, at this point I’m going to hand it over to Megan. We have definitely seen a significant amount of impact from COVID-19 in our data. I hope it was helpful to share these insights. We’ll continue to layer on more into these analysis and continue to provide updated views of what we’re seeing. So, Megan, take it away.
Megan (Klein) Jones:
All right, thanks Rob. So, now that you all understand a bit more about how the mobility trends have been changing, we did want to take some time to share our thoughts on how the pandemic might impact the auto insurance industry. Now, I’ll touch a bit on not only the potential impact due to the mileage and behavior changes we at Arity have been observing, but we’ll also point out some other considerations that companies might want to keep in mind and evaluate as you look to understand the ultimate potential impact that is specific to your book of business.
All right, so, first, let’s go ahead and start by looking at the response from insurance regulators in the U.S. so far. Now, most states are encouraging, or even requiring that companies provide relief for insurance payments to their customers. This has come in the form of relaxing due dates and extending grace periods as well as requesting insurers waive fees for late payments. We’ve also seen that some states are even prohibiting cancellations of policies during this time and other states are strongly encouraging that companies work through all available alternative options as best as they can before moving forward with canceling the policy. Additionally, there are some states that are encouraging insurance companies to extend the protection for personal lines drivers to cover the increased commercial driving activity such as delivery of food, medicine, and other essential items that small businesses are using to adapt to the current circumstances. Now, it’s important that we note that every state’s approach has been different. Some states are mandating these changes while others are just recommending them. So, we encourage that you check in with your respective departments of insurance to ensure that you fully understand the regulations as it relates to the COVID response.
All right, so, as I move beyond talking about the coverage and payment adjustments that regulators are requesting, it’s also important that insurers understand that the shifts in mobility patterns Rob just shared can lead to insurance companies experiencing lower rates of accident. Now, if we look at historical data, you can see there is a strong relationship between miles driven and collision frequency. So, here on the slide is a graph that was initially presented by the Insurance Information Institute back in March of 2018. And, on this graph, they depicted mileage data from the Federal Highway Administration and compared that data to average collision frequency data that was derived from Fast Track information to help illustrate that there was a relationship between miles driven and collision frequency. Now, to orient you a bit, on this graph, the orange line represents the Federal Highway Administration’s estimated miles driven for the year that ended on each quarter, and the blue line reflects that rolling four-quarter-average collision frequency. And, what you can see here is that when miles driven was decreasing back in 2008 and 2009, around the time of the recession, so did collision frequency.
Then, we can see that both of these trend lines remained pretty flat after the recession. And then starting in 2014, both mileage and collision frequencies began to increase again.
Now, while this graph only shows results through 2017, it does help us see that there is a relationship that exists between miles driven and collision frequency. However, I would say it’s important to note that this is only the relationship between miles driven and frequency for collision coverage. And, we all know that collision coverage is a personal lines coverage that has much more of a likelihood to be impacted by a reduction in miles. There is and there will be variability in how these reduced mileage trends can impact any one specific insurance company’s results, how they impact severity, and even how they impact the frequency associated with other coverages. So, with that in mind, we wanted to talk through some of the other considerations beyond just mileage that companies might want to take into account as they try to understand what the ultimate impact of this pandemic can look like for their book of business.
So, I’m going to preface with this list is not comprehensive, but I did want to remind folks that there are many additional aspects that affect the business of insurance, that insurance companies could take into account as you make your own independent decisions. So, first, let’s talk about the potential impact on repair costs. So, I think some of us have seen, but the supply chain may be disrupted. Imported goods could be delayed, small repair shops could be closed, and these and other impacts can have an overall impact on the costs that insurance companies will, or could, ultimately see to repair vehicles.
Next, beyond just vehicle repair, there could also be an impact on medical costs. Now, medical costs could be impacted due to the stress that facilities might be experiencing during this pandemic. And, on top of that, situations may be different now for any of the victims that are in accidents and need medical assistance because of the other influences and the impact on the facilities themselves. So, these can all lead to changes that may be observed in average bodily injury claim amounts. So, next, let’s talk about the impact, or a potential impact, on claim development. And, here, there are a couple of considerations that you may want to keep in mind. First, it’s possible that insured behavior has or will change in a way that affects their claim reporting patterns. Whether sped up or delayed because of the circumstances that we’re in. Additionally, claim settlements might be delayed given there are many courts that are closed during this time. Now, as an interesting aside, a friend of ours, Robin Harbage—for those of you that might know him—he was on jury duty, the third week of March, in Cleveland, and noted that there wasn’t a single civil case that was tried the entire week. So, he actually didn’t even need to go in for jury duty. So, it’s not unreasonable to expect some changes here as we see those courts being slowed down or even closed during this time.
All right, moving on. There might be coverage or policy changes that insurance company might want to keep in mind. Related to that, many restaurants and small businesses are adapting to include delivery to manage through the disruption of the pandemic. And, often for these small businesses, these are personal autos that are put into commercial delivery service without an associated or explicit policy change. So, as mentioned before, with regulators extending coverage for personal lines, there might be a new mix of covered losses that could come through. Additionally, consumers themselves might change their coverage or policy information. For example, some consumer groups are advocating that the customers, or the insureds, evaluate their driver classification and change their classification from driving for work to driving for pleasure, and then evaluating that, and changing back again, once the insured reverts, or returns to driving to and from work.
So, any of these types of changes, whether it be changes in coverage or change in the types of coverage and for how much could impact the overall losses and premiums that can be expected for any one insurance company. All right, so, as it relates to coverage level differences or any variances by coverage, I hinted at it earlier, but remember that not all coverages will be impacted by a reduction in mileage or, frankly, any of these other considerations on the slide right now, in the same way. We know—and if anyone lives in Chicago—it was snowing earlier. So, there are still going to be weather events, and hail, and also events like theft, and, really, mileage driven will not directly impact those types of claims in the way that they might impact other types of claims. In fact, if we use the last recession as a reference, it is possible that insurance companies might actually see some increases in certain types of claim activity, such as vehicle theft or theft claims, in general. And, related to that, that brings us to fraud as well. As a society, we’ve never experienced this type of or such an increase in unemployment rates. And, the reported unemployment completely misses the impact on gig workers. So, while insurance companies will be able to or could look at prior trends to inform how changes might look, especially as it relates to fraudulent claim behavior, it really will be difficult for our past trends or the industry’s past trends to accurately reflect what the industry could experience now.
All right, so moving on to premiums received, there is additional liability that is associated with the non-cancelable policies and the deferment of payments that regulators are requesting companies to offer. So, companies might want to take into account this potential increased liability in unrecoverable or uncollected premium that will come or could come from that. And then, customer retention as with normal considerations for a business, understanding the levels and the types of customers that remain in an insurance company’s book will be important to fully understand how average losses and premiums might shift. And then, as we talk about investment returns, let’s not forget that an insurance company’s performance is dependent on more than just underwriting returns, more than just how much premium received relative to the losses and expenses it’s paid out. But, variability in the investment returns on those reserves that are held for unearned premium or losses will also impact the company’s overall returns.
So, even beyond, I know these were additional considerations beyond mileage to keep in mind, but even as we talk about mileage trends and what Arity is seeing, there are still many unknowns in when those trends will begin to revert, how quickly the trends will revert, and even to what levels those mileage levels will come back to. And frankly, it’s entirely possible that driving trends might not return to the same prior levels that Arity was showing earlier. Given companies are learning more about how to enable their employees to work from home and individual consumers are learning more about different types of auto products such as more usage-based insurance products. So, it’s even too early to tell how mileage will ultimately impact a company’s results. So, I’ll wrap my section with really just saying there’s still so much unknown about how this might ultimately impact each aspect of an insurer’s business, but it will be good for companies to consider taking into account these other considerations to best understand how their own books might be impacted. And as Rob mentioned, Arity will continue to monitor trends as they develop. So, with that I’m going to pass it back over to Louisa for our wrap up.
Louisa Harbage:
Excellent, thanks Megan. So, you’ve heard from Rob about how much less people are driving and the ways their driving behavior both has and hasn’t changed. You’ve heard from Megan about how it’s reduction in driving may impact frequency as well as other factors insurers may want to consider as they contemplate the impact of the current pandemic. Before we open things up to a few questions—and we do have some time for that here—I just wanted to share a few additional data sources that are out there tracking the impact of COVID-19, that you may find of interest. Google Community Mobility Reports has been specifically looking at point-of-interest category changes—so retail, recreation, or groceries, in order to monitor economic impact. The Kaiser Family Foundation has been looking at cities and the states issuing stay-at-home orders and what the specifics of those look like. And the COVID Tracking Project is obviously specifically looking at COVID-related testing data. So, all of these potentially helpful resources that you might want to take a look at. If you’re panicking—I know some of these are very long URLs, so I would just remind you that we will send you a copy of the slides. This will be included there, so no need to feel like you need to jot those down right now.
All right, and with that we will open it up to questions and I know I’ve seen quite a few of them coming through. It looks like we’ve got several around some of the trip data here. So, one asking you showed broad geographic coverage of trip data. Can you share more about the mix on the countrywide heat map slide? It looks like you have as much data in the upper peninsula as in Chicago. You guys are much better at geography than I am apparently. How accurate is that? So, Rob, could you maybe just talk to us a little bit more about some of that data?
Rob Nendorf:
Yeah, sure, I’d be happy to. So, I’ll get into the mix by geography. I think I also saw some other questions in the Q and A, just about the basic volume and different attributes of the data. So, the data is multi-source as we mentioned, and it’s a large set, and we do have all customer segments covered. So, family single standard, non-standard. It is mostly personal driving and personal vehicles. The over 23 million connections is really monthly active connections. I think someone asked about the daily active connections is generally two thirds of that number, at least pre COVID because we saw that that daily active connections number is going down significantly because of the pandemic. So, in terms of the geographic coverage, as I mentioned before, we have strong coverage throughout the country and even outside the country, the scale that we have is important, right, as you dig into the geographic breakouts like county or census track that I mentioned earlier.
So, while we keep the data anonymized and we do attach census track, county, and state for the trip start, and that’s been really important for us. For instance, early on in February into March, as the epidemic broke out in and near Seattle, we are actually tracking King County, Washington—the county that covers that location—and keeping an eye on that as it progressed. And so, we have the trip volume to look even at small counties, not just King County, Washington, and get solid trip volume on a daily and weekly basis to be able to see these trends. So, even when we break them out at a very granular level, you actually see charts that look a lot like the countrywide chart because we have the credibility and the data and so we’re able to see those impacts on a daily basis.
Louisa Harbage:
Excellent, thanks Rob. I also see a few questions around the driving score. What’s included? What are we monitoring there? Megan, do you want to talk to that?
Megan (Klein) Jones:
Sure, I’ll cover a little bit about that. So, as Rob mentioned earlier, Arity develops scores that help us identify and segment driving risk based on different behaviors, driving behaviors, that correlate with insured losses.
And these proprietary Drivesight scores use variables like the ones that Rob covered earlier, including features that reflect things like speed, change in speed, the time of day, day of week that an individual user is driving as well as distracted driving for those sources of data where we’re collecting information from a mobile app that someone has downloaded on their phone. And we do operate as a rating service organization, so when we work directly with companies, we can dive more deeply into exactly how that score works and how it could be leveraged by an insurance company to help inform their book of business.
Louisa Harbage:
Thanks, Megan. I have another question here and I think it’s probably a really good one as we start to look forward—will Arity have the same level of insight as we continue on, or even as things start to return to normal? How will you share and leverage that information? Katie, we haven’t heard from you yet and, as our Director of Insurance Solutions, might be a good one for you to tackle.
Katie DeGraaf:
Thanks, Louisa. Hey everybody, it’s Katie. Deb, this is a good question. So, we do have this information ongoing through our partnerships and we’ll certainly be monitoring it ongoing during the pandemic as Rob and Megan have already talked about. Now, we want to decide the scope and frequency, kind of what we share based on your feedback and the changes in the environment. I think Louisa has already mentioned you will be receiving a copy of this deck. That was another question we got a few times. So, that will be coming, you should get it tomorrow. You’ll also get a survey and you’ll have some contact information for us. And so, we hope that you use that to give us some feedback. Let us know what you think and what would matter to you, and from that we’ll figure out how to respond appropriately.
Louisa Harbage:
Great, thanks Katie. I know we’re creeping closer to the top of the hour and, if any of you are like me, you’re back-to-back meetings. So let’s take maybe one more question here before we wrap up. I’m seeing something about changes to trip durations. So I know we’ve talked about frequency and miles driven, but a little bit about durations. Rob, do you want to chat to that?
Rob Nendorf:
Yeah, so, trip duration, as you might expect is fairly correlated with miles driven and you see the same pattern that you see in terms of miles driven per active connections. So, yeah, to summarize the countrywide metrics, there are fewer people on the road. The people that are on the road are driving less, they’re taking shorter trips for less mileage. And so when you combine those two things together, you get significant drops in miles driven. And then, when we break this out at a state level, you see a lot of variation by state. All states have dropped essentially an unprecedented amount relative to historical trends, but there are significant differences at the state level. And then when you look at even the county level, again within state, you see a lot of variation and that’s kind of what that population density graph was picking up. So, I live in the state of Illinois and Chicago is a very different place than Southern Illinois. And those two locations are experiencing this pandemic very differently so far, and that shows up in the driving behavior. So, yeah, there are a number of things that we sort of could have shared, but we didn’t want to inundate everyone, so we kind of picked the key metrics that told the story and kind of kept the presentation to that level. But we’ll continue to layer on more analysis over time. And, we’d love to have your feedback on what you’d find interesting that isn’t in here and what really resonated and you want us to continue to show.
Louisa Harbage:
Great, thank you so much. All right, so I know we still have quite a few questions in the queue and we are running a little bit short on time. As I mentioned before, we will be documenting all of those and doing what we can to follow up and continue to provide additional insights and potentially address some of those offline. I do want to thank you all for joining us today. We will, as I mentioned, be automatically sending out copies of the slides to all of you tomorrow. In that email, you’ve heard Katie and Rob both plug it, but we will also be including a link to a quick survey which will give you an opportunity to tell us what other topics, or insights, might be of interest to you in the future. For those of you who engage on Arity on a regular basis, I know Grady mentioned at the beginning, but we are expecting that we’ll be having extended discussions about these insights and more over the next few days. For those of you who are new to us, we hope you’ll reach out and continue so the conversation, so that we can do some of the same for you. And with that, thank you all very much. Have a wonderful day and please stay safe.
Read our accompanying report, in which Arity experts share an in-depth analysis on how mobility behaviors and risk have changed throughout the pandemic.
