Interview Transcript

Disclaimer: This interview is for informational purposes only and should not be relied upon as a basis for investment decisions. In Practise is an independent publisher and all opinions expressed by guests are solely their own opinions and do not reflect the opinion of In Practise.

My goal is to gain a deeper understanding of the long-term dynamics in this space. That's why I have some questions for you. Could we start with your background?

Certainly. As of last Friday, I concluded a multi-decade career at GEICO.

Congratulations. That's quite an achievement.

Thank you. Not many people my age have spent 38 years at one company. GEICO is the second largest insurer in America, recently surpassed by Progressive. Car insurance is our core business, and I've been in the claims business for all of those years. I've also overseen operational units and managed P&L for large markets. I can offer an opinion on almost any aspect of car insurance, whether it's sales, competition, claims, market dynamics, or different companies' strategies. I've studied almost all of it and have a deep understanding of the industry.

When it comes to claims and auto damage, I've been directly supervising and handling or overseeing people who handle these types of claims for 25 years. In auto damage, we currently have around 4,000 people. It's a significant part of our business. Approximately 60% of the premium in America goes towards paying claims related to car damage. This is an industry estimate, not specific to GEICO. Nothing I tell you will be specific to GEICO.

I'd like to begin by discussing the impact of technology on collisions and collision repair, focusing on the technology within the car.

When we talk about technology, most people think of ADAS, or Advanced Driver Assistance Systems, which influence vehicle movement. In the United States, the easiest way for me to explain this is through the Insurance Institute for Highway Safety, or IIHS. Are you familiar with it?

No, I'm not.

The IIHS is a non-profit organization primarily set up by insurers to study technology. This initiative dates back to the late 1980s when airbags were becoming prevalent. The IIHS conducts crash tests and you might have seen car manufacturers advertising their vehicles as safer, often mentioning that a particular model has received the IIHS top safety pick. They're constantly examining car technology to improve it and understand its impact on claims, safety, and insurance-related matters. Alongside the IIHS, there's a data compilation program. For instance, if the 10 biggest auto insurers in America account for 70% to 75% of the premium, all 10 are members of this program. If we consider the top 20, we're probably talking about over 90% of the companies in America. I'm going to share my screen to show you a few things.

What you see on the screen is data from the IIHS and the Highway Loss Data Institute. I'll start from the top and then we can go back to page nine. This data is due for an update at the end of this month. Let's start with the basic technology in cars and then we can delve into these other charts. The top figure, Figure 14, shows the percentage of new vehicles sold in America by model year that have a camera. The middle figure, Figure 15, shows the percentage of all registered vehicles in America with a camera. The most recent estimate is that there are 280 million registered vehicles in America, including cars, motorcycles, and trucks, with the majority being cars.

There are a lot of them. The average age of a vehicle in America is 12 years and it's increasing.

Yes, and getting older.

As cars age, they may not have rear cameras. In America, there's an entity that studies how long it takes for technology to be integrated into the fleet. This helps everyone understand what is likely to happen in the future. By looking at figures 14 and 15, you can use past history to make a projection. This method was used with airbags and third brake lights, as these weren't always standard on cars. Figure 16 is their prediction of when cameras will be in 80% of the fleet, and they're usually quite accurate.

They have this data for every type of technology. For example, the first car I bought in America that had a rear camera was a 2005 Lexus GS. But according to this data, as of 2022, only 40% of cars in America have a rear camera.

I can send you this information. It's publicly available and useful to know when you're interested in car technology.

Now, let's look at the same set of slides for curve adaptive headlights. These headlights turn with the steering wheel, improving visibility when you're steering. Does that make sense?

I have these charts for every technology. We'll discuss two things for you to understand. The first question is what percentage of the cars vehicles have this technology, in order to understand how the technology affects accident rates. The second thing we're going to talk about is when the car does have the technology, what does it do to the frequency of accidents and the severity of accidents?

That's exactly what I'm trying to understand.

I'll try to get there a bit faster. I'm just laying the groundwork so I can send you this information. I don't want to waste your time just showing slides. I want this to be a productive conversation. The IIHS releases this data every year, broken down by technology.

The next feature is the lane departure warning. This is activated when you're driving and you get too close to the edge of the road or the white line. Figure 21 shows the percentage of cars equipped with lane departure warning. Let's move on to blind spot monitoring. I've already explained what it looks like at the beginning. Now that you're familiar with this, I'll send you these 16 pages of information. Is my screen updating properly?

I'm currently on the instruction page, so I believe it is.

This middle chart, titled "Predicted Registered Vehicles Equipped with Collision Avoidance Systems", forecasts the technology in 2022 and 2027. It's a summary of all vehicles in America, registered and predicted to be registered. It breaks down the type of technology. I wanted you to see the detailed breakdown by component and type of technology. This is a summary of what it looks like. It shows how the American fleet is likely to adopt the technology, considering that vehicles are twelve or 13 years old on average. Older vehicles may not have blind spot monitoring or crash prevention, and may not even have a camera.

I'll send this to you. Now, let's move on to the next part of our conversation.

Can I download the latest version of this document from the website when it's released?

Yes, you should be able to. If you need any assistance, feel free to reach out to me. Coincidentally, I've been involved with the IIHS for a long time. I had lunch yesterday with the head of data analytics. If you want to speak with him, I can arrange that for you.

That sounds like a great contact.

I've invited him to speak at various industry events over the past 10 years, and I've taken industry representatives to their site where they conduct all the testing. It's a worthwhile field trip.

Can you see this page? It's a different exhibit. Does it say "Curve Adaptive Headlights, Page 12 of 24" at the top?


This is a separate presentation that they've produced. I'll send you both documents. You can read them, and we can discuss them further if you'd like. My perspective as an insurer might differ from theirs. Now, let's look at the "Curve Adaptive Headlights" technology. Figure 10 shows the change in PDL, which stands for Property Damage Liability.

In America, we offer collision insurance, which covers one's car in the event of a collision with another car or a tree. We also provide PDL insurance to someone like John Jones if I were to hit his car or property. These are two distinct coverages. Figure 10 combines these coverages to show the frequency of claims by type of manufacturer. For instance, BMW, GM, Mazda, and so on. It illustrates what happens to the frequency of claims when a car is equipped with curve adaptive headlights.

We previously discussed curve adaptive headlights. They have a very low adaptation into vehicles. However, here's what happens when you incorporate this technology into a car.

I have a basic question. Why would the impact be so different between manufacturers?

The difference is due to the varying technologies. Some manufacturers have superior technology. For example, curve adaptive headlights in GM vehicles may reach this far, but in a Mercedes Benz vehicle, they may only reach this far. All their technologies differ. There's no standard because the government hasn't imposed one for many of these technologies. Also, cars are built differently. The placement of components matters. When we discuss front and rear parking, everything is different. But as you can see in figure ten, the best performance for this technology results in a 10% drop in frequency.

There are different components here. This presentation focuses on frequency. Frequency refers to how often a claim occurs. Severity refers to the cost when it happens. So, you might have lower frequency, but higher severity. To an insurer, the premium may remain the same, be slightly higher, or slightly lower. You're still going to have 90% of the claims, and those 90% are going to cost more, especially on the BMW.

So, it's not necessarily intuitive, but when you add curve adaptive headlights to a BMW, it becomes more expensive to insure.

Correct, there will be fewer claims on the BMW.

So, it's safer, but more expensive. I have one more question regarding the top chart you pointed to. Why is there a difference in the frequency between PDL and collision? It seems you're much less likely to hit a tree than to damage your own car.

That could be because there are more collision claims in the daylight. It could also be that collision claims occur in a parking lot and it's a minor bump as opposed to driving down a highway. There could be any number of differences between the types of claims that might be made for collision versus property damage liability. I'm going to send you both of these items. We have the same types of charts as we did in the previous one, tracking along with lane departure warning.

We know how often these features will be integrated into cars. It's not going to happen overnight. We won't reach 80% for probably another 10 to 15 years of registered cars in America, referring back to the previous chart. This is just an example of what happens. You might wonder why lane departure warning doesn't have a significant impact on frequency, given the weighted average of these three types of vehicles

For instance, my son had a Jeep Cherokee. When I bought it for him eight years ago in 2016, it was a good car for a beginner driver, and I included the technology package. However, we both disliked the lane departure warning feature. When activated, the car's steering wheel automatically adjusts to keep you in the middle of the lane. If you get too close to either side, the car steers for you. This can lead to the driver battling with the car.

As an experienced driver, I didn't appreciate the car maneuvering against my control. The car assumes that I'm not paying attention or asleep, when I might just be too close to a line. No one drives perfectly in the center of the lane at all times.

This technology is a warning system. Not only does the vehicle try to steer against the driver's wishes, but the seat can start shaking. In many models in America, the seat vibrates if you get too close to the right or left line. This is a reason for me to deactivate this technology. I don't want my seat vibrating.

Is there any data on how often these features get deactivated in cars?

There have been studies, largely from American car rental companies. These businesses, unlike insurance companies, buy new cars which depreciate over time. Companies like Enterprise, National, and Alamo, all owned by Enterprise, typically keep cars for one and a half to two years due to depreciation. Studies have been conducted to determine what percentage of the time the technology in these rental cars remains activated by the end of the rental term. The results indicate a low percentage, although I can't provide a specific figure at the moment. I could find this information for you.

I'm curious if it's around 50%, 70%, or 80%.

I believe it's more than 50% that gets turned off. I'll make a note to find that information for you. I could discuss this with someone in the industry or perhaps IIHS may have the information.

I'm asking because I'm interested in the effectiveness of different technologies and whether any of the newer technologies, those just being released or only available for a year, will change the trends we've seen over the last five to seven years.

Over the next five to seven years? No, because it will take a long time for them to be adopted given there are 280 million registered cars in America. Even if all 15 million new vehicles sold each year in America were equipped with this technology, it wouldn't significantly impact the overall fleet because there are 240 million to 280 million vehicles that don't have it. As I mentioned in another presentation, it typically takes 30 to 40 years for newly introduced technology to reach 80% adoption.

Let me show you another slide. This one is about front crash prevention. This is a new technology. It was first introduced in the 2000 model year and by 2009, it became standard on 1% of new cars. As of 2022, it's standard on roughly 70% of new cars, but only 28% of the total fleet.

Let's focus on the bottom chart, but ignore the year 2000. Instead, let's look at 2010. To reach 80%, it will take 25 years, from 2010 to 2035.

I'll send this information to you and we can discuss it further. I want to ensure we make the best use of your time. The data is quite interesting. The topic of ADAS often comes up. It's not as frequent as it was six years ago, as the world is changing. We've discussed AI and the current trends. People used to talk about driverless cars, envisioning a world where cars drive themselves. However, in America, you can't currently buy a car that drives itself, one without a steering wheel.

If a car that drives itself isn't available by 2023 or 2024, it likely won't have a significant impact in America for 30 to 40 years. By that time, I'll be 88. As an insurance executive, I can tell you that while technology is being introduced into cars that will affect the frequency and cost of accidents, driverless cars won't have a significant impact on the American insurance market for 30 to 40 years.

I've been advising people to focus on other areas rather than driverless cars. In America, things are more complex due to the state-by-state judicial system and tailored negligence laws. The laws currently don't support the liability of a driver in a car without a steering wheel, and it will take decades for this to evolve.

Let's consider front crash prevention. If I'm a driver in America and my car has this feature activated, who is responsible in case of an accident? Is it me, the manufacturer, or both of us? What if I turn off the technology? There are countless issues that lead me to believe that driverless cars won't have a significant impact for a very long time.

Given what you see today, what's your outlook on claims inflation over the next decade?

I believe that the frequency of claims is likely to decrease. However, it's complex because the frequency of claims cannot be examined in isolation.

Each study I've presented assumes that things will remain constant, that what happened yesterday will happen tomorrow, and this will continue for 10 to 15 years. However, we know this isn't the case as there are no absolutes in the world. There are economic factors that affect frequency. For instance, if the economy declines or there's a recession, the frequency will drop. This happens because people change their habits during economic downturns, such as going out less frequently or reducing travel. There are countless factors that affect frequency. However, if we assume that everything remains constant, these studies suggest that the frequency of claims will decrease.
However, the severity of claims is likely to increase. But my prediction is not precise due to various economic factors. If the economy is booming and people are shopping, dining out, and vacationing more, the frequency might actually increase.

From the data we've gathered so far, it seems that the increase in severity is significantly higher than the decrease in frequency.

That's generally due to technology.

So, it's not about personal injury but about property damage and collision. It's more about the cost of repairing these cars because the components are more expensive.

Correct. But let me go back to the other presentation to show you something. We're back at the section on lane departure warning.

Look at figure 13. This shows the different technologies, PDL, and collision. Do you see it?

Here are the injury claims. We have the same set of charts for the same cars for injury claims. You mentioned personal injury, and yes, it's slightly different, but all the data is here. The frequency of a bodily injury claim on vehicles with lane departure warning might actually be a bit higher. I don't have a definitive answer as to why. But you can see a 6.2% increase in bodily injury liability and no statistically significant change in PIP. Do you understand the difference between BI and PIP in America?

No, I don't.

Collision and property damage refer to my car and the car I hit, respectively. Personal Injury Protection (PIP) and medical payments (med pay) cover injuries to people in my car.

BI stands for Bodily Injury, which refers to injuries to people I hit. These are just different coverages. However, they are measured differently and their effects vary.

That seems unusual.

However, it's data. It may not always make intuitive sense to me, but it's real. So, when you hear news or media pundits, or read something, the data doesn't always support what people say.

Yes, that's true. Some of this isn't intuitive either, right?

Indeed, it's all consistent though. We mentioned AI and ML as the current trends. We're going to see how the world adopts all of it, how it really works, and what the meaningful impact is. Let's just go back and use a different example related to cars. 10 years ago, I was on a panel discussing the shared economy, and Uber was becoming more mainstream. This was in 2010, and everyone was saying insurance companies would die out because nobody would need a car anymore due to Uber. But here we are 10 years later, and there are more vehicles registered in America to individuals. While Uber has its place, it hasn't really displaced anything. Except for taxicabs.

I'm going to send this data to you. We can get back together and discuss it further. It's real, meaningful data that tells a story. Given the history, the story usually doesn't lie. If we go back to the eighties and airbags, each of these technologies, the projections of this entity are time-tested and technology-tested. It's going to be pretty accurate. It's all driven by the size of the fleet and the number of new units that have the technology.

For instance, consider Lane Departure Warning. There's an Electronic Control Module, or ECM, in the car. If you were to compare the number of ECMs in a car in America 15 years ago to today, we're probably talking about a 600% to 1000% increase. None of these systems work just because there's a sensor. There's always technology behind the sensor doing something to communicate with the driver, and that's through one of these modules.

When it comes to the severity component of this, and other companies like CCC have looked at this. Sensors, a lot of this technology, parking sensors, rear bumper sensors, front bumper sensors, Lane Departure Warning sensors, cameras. 10 years ago, how often was a camera on an estimate? Today, how often is repair or replacement of a camera on an estimate? The beauty of this thing that I introduced you to at the beginning, is called IIHS and this other one, let me scroll to the top so you can see the full header, it's two things, IIHS and Highway Loss Data Institute.

Essentially, it's a unified entity with two names. We refer to it as the Highway Loss Data Institute. Every company in America that is a member sends monthly claims data to the Institute. This data includes information about whether an Audi, BMW, or GM was involved in an accident, the type of car, and the types of claims. They also receive the Vehicle Identification Number (VIN), which provides data on the technology in the car.

This is a detailed process used to compile data. It's not just people making assumptions about what's happening. It's all based on real data. The insurance industry sends this data to the Highway Loss Data Institute and both of these entities analyze it. This is how they know what technology is in the Audis, how often it's there, and its impact on the severity of an accident.

We send them what we call "estimatics", data from the estimates from CCC. This includes information like how often an airbag was in the estimate of repair of damage, how often sensors are there, and how much sensors cost. They receive all of this data and compile it from a very detailed perspective. I can confidently say that these are statistically significant numbers that are not going to change dramatically.

This is fantastic. It's very useful data. I think I should review it before asking you more questions. I suspect many of my questions are already answered in these reports.

Can I use the remaining time on our call to switch gears a bit and discuss a few other things? I'm thinking about demographic trends or societal trends that might be impacting frequency and severity. One term that comes to mind is "distracted driving", like people using their phones. Do you think this is significant and something I should keep in mind?

I believe this is already reflected in the data. If you were to look back at 2014 and 2015, you would notice a change in frequency. There could be numerous reasons for this change in frequency, and I don't have a definitive answer. It might be due to a combination of factors that I'm about to mention.

One factor that influences frequency is gas prices. When gas prices increase, people tend to drive less, and vice versa. We can discuss electric vehicles and other factors, but gas prices significantly influence driving frequency.

Another factor is the power of a vehicle. I don't have the exact data with me right now, but I can provide it later if you're interested. I have some additional data that I didn't present because I wasn't sure of its relevance.

The IIHS and HLDI also publish data on vehicle power, such as the percentage of cars with turbocharged or supercharged engines. Accidents often occur due to driver inattention and vehicle power. As more powerful cars become prevalent, there will be changes in frequency because these cars tend to drive faster, which increases braking time and collision force.

The severity of an accident is always going to be higher with faster vehicles. Insurance premiums are based on frequency and severity. If frequency decreases but severity increases, the premium might remain the same or even increase, depending on how these two factors combine.

From what I understand, engine size has significantly increased over the last decade or two. More trucks are being produced, and they are larger and equipped with bigger engines. Is there a trend or curve that shows this increase, similar to the adoption of seatbelts?

Let me check. I have an old study here. I don't have it in electronic form, but I can probably obtain it. The study is dated May of 2019. Let me find the section on vehicle power and glass. They have data on everything.

They also study electric vehicles. Can you see this?


They have data on front crash prevention cameras, calibration glass losses, panoramic roofs. I must have passed the section on engine size. They study the trend of engine size in vehicles over time. You're absolutely right. Engines are only getting bigger and more powerful.

Here it is. I knew it was here somewhere. I apologize, I would need to get this scanned in. There's a section here about engine power. Can you see this?

Yes, I see it. Supercharged engines.

They're examining data from 2016 and 2017, which is turbocharged. On my left, the low bar here indicates claim frequency. The middle bar represents claim severity.

It's increased by 20% or so?

It's increased by 4%. But look at this one. This is interesting and it's consistent with what I've observed. Look at the title. Can you see it? It reads, "Percent of vehicles with turbo and supercharged engines." Now look at this data, ranging from 1985 to 2016. Manufacturers are producing cars with more power.

So, for 2016, is it 22%?

Yes, it's around 22.5%, let's say 23%. This is likely referring to new vehicles. It doesn't specify, but I could get clarification. It must be new vehicles.

Because the change is too rapid to be the entire fleet.


I see.

When I say that the people at IIHS and HLDI have all the information regarding vehicles, I mean it. This presentation is from 2019. I must have printed out around 100 pages.

I believe I can get this in electronic format.

That would be helpful. May I ask about cameras in cars? In commercial vehicles, there's a camera facing the driver to monitor their actions. Is this likely to be implemented in consumer vehicles?

As far as I know, no manufacturer includes these cameras at the moment. I believe they're almost entirely absent.

It's not accurate to say 100% of them are aftermarket purchases. They are indeed in cars and used by Uber drivers. Over time, they will be adopted, but we're looking at decades before they have a meaningful impact. The most likely impact is not necessarily on the frequency of accidents, but on the evidence of who's responsible when accidents occur.

Apart from being an insurance executive, I'm also a lawyer. I've studied how claims are made and the legal consequences for drivers with different technologies. I've given a lot of thought to what happens in the legal system. The camera merely records what happens in front and behind a driver. It doesn't affect driver behavior. From my perspective, it's evidence that I didn't do anything wrong, or that I fell asleep, crossed a line, or whatever.

Eventually, it's likely that the camera will be incorporated into the black box. One thing I don't know is what percentage of cars are equipped with an EDR (Event Data Recorder), also known as the black box. Cars have been equipped with EDRs for at least 20 years. I imagine the camera would be sending information to the EDR, so anyone trying to reconstruct an accident would have that evidence available.

In cases where there's collision damage to cars, how is technology changing the way cars are repaired? I'm thinking about who repairs them, the type of parts they use, like OEM versus aftermarket versus salvage, and how the cost is shifting between labor, parts, and diagnostics?

Those are excellent questions. I have some background in this. In America, there's the Certified Auto Parts Association, which deals with non-OEM parts manufacturing, testing, and ensuring that any non-OEM part that is certified has the same safety specs. OEMs have done a lot to try and force the insurance industry to use their parts. Non-OEM parts are still available.

There are OEM parts, which are made by the manufacturer, and then there are non-OEM parts. Within non-OEM, there are aftermarket parts and refurbished parts.

Refurbished parts are taken from a wrecked car and brought back to life?

Exactly. Then there are recycled parts, which are taken off a car and resold without refurbishment. It's a used part. All of these, and there's data on this, I don't have it handy, but traditionally, the insurance industry would try to get 35% to 40% of its parts from non-OEM sources.

This all flows into the availability of these kinds of parts. The total losses that a carrier has, there are partial losses, claims that can be repaired, and total losses, cars that can't be repaired. Any insurance company is going to look at a claim and say, what is the value of the total loss, the salvage.

It's a massive business in both America and Europe. Insurers use scientific data to decide whether it's more economical to repair a car for $25,000 or declare it a total loss for $45,000, considering the remnants could be worth $30,000. So, the net payment might be $15,000 in one case and $20,000 in the other.

We're talking about a $25,000 repair versus a total loss. When cars are declared total losses, it's referred to as total loss frequency. The insurance industry measures all of this. There's data for everything, but it comes from different sources. You need to know where to find it. I could show you some of this data. I didn't think we'd get to it today, but total loss frequency is probably increasing for reasons I can explain. One reason is that parts for newer vehicles, like a Tesla, are in high demand and not always available.

Let's consider the batteries in Teslas. The average lifespan of a Tesla won't be twelve years, like most cars.

The batteries in an eight-year-old Tesla won't be as readily available as those in a combustion engine vehicle, like a Subaru or an Audi. The battery in a Tesla is similar to the battery in an iPhone. It changes in size, weight, and dimension over time. Eventually, Tesla will stop producing batteries for older models. So, the average age of a Tesla will be less than the average age of a car in America, which is twelve years.

I've heard that Tesla insists on using OEM parts for safety reasons.

Currently, Tesla is trying to enforce many regulations and is in disputes with several insurance companies. It's not as straightforward as Tesla initially thought. I've been dealing with them for 10 years on this issue. They demand higher labor rates and believe their cars should only use Tesla technology. Whether these mandates are justified is debatable. Let me find an article I recently shared with my colleague, that might be of interest.

Here's the article. Let me share it with you.

On March 21, it was reported that electric vehicle repairs are 28% more expensive than those of traditional combustion engines, also known as gas cars. This information is according to a survey conducted by Value Penguin in America. I can provide you with the article that I had previously shared with a colleague who was concerned about the high insurance rates for her electric vehicle. Electric vehicles are not only more expensive, but they also have more speed and advanced technology.

It would be helpful if you could send me those reports after our discussion today. Before we conclude, I have one last question. I'm interested in understanding the dynamics of non-OEM parts. They seem to be gradually gaining market share over time, infiltrating different parts of the car. Do you think this trend will continue? Given that the cost of parts like headlights could increase tenfold due to the addition of adaptive sensors, insurers might have to look for more affordable parts. Or do you believe that these parts will become harder to remove from vehicles, leading to a shift back to OEM parts?

I believe used parts will always be in demand. This question has been asked for the past 50 years. Even 38 years ago when I started in this industry, alternative parts were available and sought after because they were necessary. As technology continues to evolve and become more complex, the demand for alternative parts, whether they are aftermarket, recycled, or refurbished, will persist. For instance, Ford's EcoBoost engine, which is more of a hybrid than a pure electric vehicle, is a case in point.

Around 2018, Ford realized that it needed more engines for its 2014 models that had the EcoBoost engine. The company approached the insurance industry to inquire if more EcoBoost engines could be totaled to create an alternative supply in the market or if they could be sold back to Ford for refurbishment. This was due to the high cost of maintaining parts for many years. In America, there's a law that mandates car sellers to ensure parts are available for at least 10 years. I can provide more information on this if you'd like.

So, they don't necessarily have to manufacture the parts, but they have to ensure they're available, correct?

Yes, that's correct. For instance, if I buy a 2024 Tesla, the parts for my 2024 Tesla should be available until at least 2034. I might be slightly off with the years, but you could probably verify this with a quick Google search. Alternatively, I could find the information for you, or we could discuss this further at a later date.