> The suit also cites Allstate as gathering direct car use data from Toyota, Lexus, Mazda, Chrysler, Dodge, Fiat, Jeep, Maserati, and Ram vehicles.
Seems like the bigger part of the story is at the bottom. You can uninstall GasBuddy from your phone but finding and buying a new car that doesn't track you is a bigger hassle.
>Mozilla’s latest edition of Privacy Not Included reveals how 25 major car brands collect and share deeply personal data, including sexual activity, facial expressions, and genetic and health information
How do they triangulate that from car telemetry?...blood pressure/heart rate signatures from face? Suspension rocking in a park at night? Maybe they pull it from your phone.
I don't trust the manufacturers, but Kia seems to have said they have the right to collect sex info, but haven't. [0]
> “To clarify, Kia does not and has never collected “sex life or sexual orientation” information from vehicles or consumers in the context of providing the Kia Connect Services.”
Where you drive and how long you stay there at night, correlated with your partner's connected car.
It's not a camera in the bedroom but you can pretty easily extract relationship graphs from geolocation tracking and proximity. US intelligence agencies have been doing it in the middle east for ages...
“For every GM vehicle, before any connected vehicle services are activated and before any data is ever collected, the vehicle owner must accept the OnStar Terms & Conditions and Privacy Statement. These detail our data practices and are available online for consumers to review before they even walk into a dealership.”
It's all spelled out in an 4pt light grey font that you clicked through without reading. So it's your fault we know you're pregnant before you do.
Even in the flood of terrible news about privacy and other things, this exposé stands out as especially disturbing. I was considering getting a new electric car to replace my combustion, but now I'm going to stretch it for as long as I can instead.
Great article. I remember seeing that a year or two ago but had long since forgotten about it.
> Several car brands also note that it is a driver’s responsibility to tell passengers about the vehicle's privacy policies.
"Hey Bill, before we go to lunch, gloss over my Nissan's EULA."
I think I might start buying older cars and just swapping out the engine instead of putting up with the telemetry. If the car has a SIM card, I don't want it.
Why would society not mandate cameras to make sure the operators of thousands of kilograms of metal at high speeds are paying attention to the road instead of their phone? And to be able to punish them if they are not.
Assuming pedestrian and children’s safety is a priority.
I guess being permanently paralyzed because someone rams into you because they were checking their texts or TikTok is not harm, but a video recording of someone not paying attention to the road is harm.
Isn’t one of the most common ruminations of modern society that children cannot roam freely due to excess risk of being hurt or killed by a distracted driver?
>IIHS says pedestrian crash deaths have risen 80% since hitting their low in 2009. The statistics show that 2021 was almost as deadly to people on foot as last year. Nearly 7,400 walkers — more than 20 people a day — lost their lives in 2021 after being struck by a vehicle.
Airline pilots get recorded, why shouldn’t drivers?
> Airline pilots get recorded, why shouldn’t drivers?
AFAIK, general aviation pilots are not recorded. Black boxes are only a thing in commercial aviation, so a more appropriate analogy would be the recording of bus or semi-truck drivers.
There is far less moral hazard as a general aviation pilot because death/grave injury is far more likely in an airplane collision, whereas a personal vehicle driver is relatively safe, especially in vehicles most dangerous to others
Regardless, what is and is not required of all pilots is beside the point. The point is society implements a safety/accountability measure to prevent x rate of injuries/deaths…but society does not implement the same safety/accountability measure to prevent y rate of injuries/deaths where y is far greater than x.
The discrepancy is because it is politically unpopular to hold the people causing the larger rate of injury accountable (voters who drive personal cars), whereas it is politically popular to hold the people causing a smaller rate of injury accountable (commercial pilots).
This seems to run completely contrary to Mazda's own Connected Services Privacy Policy, which states that by default (ie if you aren't paying the subscription for connected services, or never enrolled in it), they don't share info outside of Mazda, other than aggregate data for research purposes. Even the section discussing if you are an active subscriber indicates that your written consent is needed before it's shared with unrelated third parties.
Third-party service providers: We may share your Personal Information with third-party service providers that we contract with to provide you with the Mazda Connected Services, as detailed in this Connectivity Privacy Statement.
The Artity (allstate) SDK offers some utility like "fuel efficiency tips", but it is a cover to collect data.
Folks your emergency button we all have in our cars (onstar) that supposedly calls 911 in case of an accident (they don’t unless you pay the subscription) they just sell all the gps to data brokers.
For older cars the telematics module might not even function. The phase out of legacy cellular networks will clean out some of this monitoring. Disabling the telematics module isn’t even difficult and it’s easily accessible in my vehicle.
It is intensely frustrating that you need to be an aero embedded remote sensing SWE to defeat all this bullshit though. It used to be that having a basic knowledge of javascript was sufficient.
This is Allstate, the house, trying to maintain their edge over their customers, the players. If they get additional info that tells them you are an unsafe driver, then their odds of you not getting into an accident diminish and their goes their premiums, the bets.
Yah yah insurance reduces risks instead of adding it like gambling, but the analogies are still there.
I hate insurances and their hidden shenanigany-like algorithms but at the end it is a fair game I when you look at the big picture.
I never owned a car but I feel deeply concerned as I use the road by foot, bike and rentals and am often scared by the drivers usage of the road. You know : texting, updating gps, driving full speed in turns without visibility, taking over as Schumacher in the last lap… I’m mean yes it’s safe for the other ones in bigger cars and themselves until no deer encounter.
The speed limit is a LIMIT not a requirement. And don’t start me with "you’re dangerous driving so slow" lol.
>> The speed limit is a LIMIT not a requirement. And don’t start me with "you’re dangerous driving so slow" lol.
It certainly is. Speed differential is a huge cause of serious accidents. That can be driving too fast or too slow. Many roads have minimum speeds, and as a commuting cyclist I can tell you the biggest threat is not being able to move with the general flow of traffic.
I'm also not sure how you came up with driving slower and smooth breaking being correlated, especially if you do limited driving.
I don’t known roads with minimum speed limit here in France but the highway (80km.h). I agree it’s a good idea to respect the min and max limit when they exist. I also 100% agree that its way safer to move at the speed of the flow. I’m also a commuter and don’t engage in flows I can’t follow. However I see too much people on low traffic condition driving at full speed where they couldn’t stop in case of… a deer/cyclist/dizzy child. Things happen and blaming random hazards won’t make the road safer. Slowing down, will.
> slower and smooth breaking being correlated
The energy of a moving object is proportional to the square of the speed, but your braving force is constant. If you want to decelerate from Xkm.h to 0km.h before the deer at 100m ahead, you’ll have a smoother stop if you drive slower. And that don’t even take reaction time into account.
> especially if you do limited driving.
Limited on a daily basis but it’s been 20 years I learned to drive and do it regularly for occasions like week ends, road trips, helping parents, going to buy heavy appliances etc… also riding a bicycle share many thinks with driving a car, a truc or a motorcycle.
This is how all the tech bro mass management of people systems work. Do away with any customer service, throw away randos impacted by any situations outside the norm because your system is large enough to keep the needed critical mass even without those people. Scale over nuanced edge cases to minimize cost/friction/complexity.
The future world will become more and more brittle/brutal for us randos because we are 100% disposable in the future tech bro libertarian dystopia. If any situation puts us outside their standard use cases and automated support systems ability to resolve issues we are purged from the system. And when all of society is manage by those types of systems you are hit.
Plot twist: at some point in our lives we're all in the "randos" group. So these policies are default inhumane to some portion of the populace all the time.
I guess you could argue that driving near deer at all is still more risk of accidents and damage to the car than a driver who never goes out of their suburb?
When your kid runs in the road. Do you want the to avoid an accident, or do you want to optimize for them just turning enough so they don't lose their insurance.
When traffic ahead of you makes an emergency stop, do you want the person behind you at a safe distance to error on stopping earlier or do you want to add an incentive to just stop right before your bumper?
There is a reason emergency braking is a popular feature. Humans tend to no apply brakes early or hard enough, even when no distracted.
Otherwise safe drivers are punished also because it isn't a reliable metric, so the actuaries know that they can't tell crash avoidance from risky behavior.
If you know about training fleets and semis, you can see these new drivers taking risks that they should not be taking.
Give it a decade and a lot lost or shattered lives and I would guess these will either abandoned or not supported by the reinsurance industry. Possibly just being another source of income to the companies to sell to data brokers.
>90% of drivers consider themselves above average,
They're probably right because the average includes a lot of people who chronically create carnage like drunks, teenagers and that old woman everyone has in their extended family who swears she's a good because she managed to not be found responsible for the dozen accidents she's been in.
Now, if 90% of people said they were better than median that would be concerning.
> While reducing speed when you see deer is important, you won't see the one you hit. In my case it jumped over a road barrier from below, it would have been impossible to see.
"Won't see it?"
You can't argue that all else equal, driving during the day is as likely to hit a pronghorn as driving at night. Not calling you a liar, but I'm skeptical.
Never made that claim, but dawn and dusk are the most risky.
Mine happened on a lunch break, can't see through rock either.
But you only see a fraction of the wildlife that is there. If your sole stratagy is to see them you may be surprised how many deer hit the sides of cars.
Complacency and selective attention are very real human problems.
The classic basketball game example of selective attention if you don't buy it.
Then it seems reasonable to charge people who live in those suburbs more.
I think they key sticking points are 1) Why isnt aggregate claim data good enough for this purpose? 2) Do we prefer erroneous risk estimation from bad personal profiles over erroneous risk estimation from group averages.
If you actually have a tracking system like this (a lot of insurance companies will offer it), you'll find the bar for unsafe driving is ridiculously low.
The truth is insurance agencies always want to find a reason to increase your premiums, because the law says you can't just increase the premiums for no reason. Most premiums are lower than what it "needs" to be, so if they find an excuse, they will jack up the premiums.
> The truth is insurance agencies always want to find a reason to increase your premiums, because the law says you can't just increase the premiums for no reason.
How does this work? Are you saying all the insurance companies share data and collude to set your rate? Or they just hope you don't price shop every year or two and jump to an offering that is under the legal limit (whatever that is)? My state has over 50 auto insurance companies, so I don't see why they would have to resort to tricks like that to increase your rate in an unfair way.
Yes, models have to be shared with regulators and the regulators have to be convinced that the premiums insurers wish to charge are actuarially justified (supported by the data).
We all do. However, the false premise here is that this data is able to reliably classify drivers into "safe" and "unsafe". If you brake hard, you might save a pedestrian's life. If you don't brake at all, you will kill them. This data lacks context for driver actions to reliably perform that classification. Also, it is also not subject to any kind of appeal where you can bring such other context and facts to bear on the determination if it determines you are "unsafe".
By that logic should we stop having red light and speed cameras? After all, it's not hard to come up with plausible sounding reasons why you might be forced to speed or run a red light.
It is not “by that logic,” since the comment specifically referenced “this data”, meaning inertial and positional telemetry collected by insurers from apps like Gas Buddy, as the article references. Introducing external speed cameras is a strawman.
They do not have a monopoly. There is plenty of room for a handful of companies charging too much to fail. there is also plenty of room for companies charging too little to fail.
You buy insurance if there is an uncertain outcome, e.g. a 1% probability of having a $100k claim. In which case you can expect to be charged around 1% of $100k (plus admin costs and whatnot).
Perfect information means you know the exact probability of the event happening. It doesn't mean that you know with certainty whether an event will happen or not. That would require a crystal ball or a time machine.
If I'm going to have a $100k claim I'm not going to drive. There is nothing in life worth doing despite it causing a $100k claim. It can wait (or I'll pay for a taxi). However I need to do things and there is a small chance I will do something wrong and thus incure that $100k claim.
That sounds... fine? If the prediction algorithm really is perfect, and it predicts my next trip is going to result in a fender bender, I'm certainly going to take uber or public transit. Even if it ends up costing an extra $50 or 2 hours of my time, that's still better than paying for hundreds/thousands of dollars worth of repairs.
Hanford has low rates and payouts. But their inspections were very demanding. They invented the Hanford Loop that keeps feed water in the boiler in case of a break, etc, etc. Being insured by them meant you'd be safe!
If you know the future exactly, then sure. If your knowledge of the future is just in expected values, then it can still make sense. Imagine the radio station doing a "we'll give you a million bucks if X happens" promotion, when X is truly random. The radio station doesn't have a million bucks to front that randomness, but the insurance company can do it.
Correct. As a State Farm policy holder, I get an annual letter reminding me that I am also an owner and have governance rights (basically a proxy vote). More similar to Vanguard than a credit union.
No, because the amount of money you have in a bank account accumulates linearly, so you can only pay up to what you have put in. With insurance, you can get a payout more than what you have contributed up to that point, which is necessary for covering catastrophic damages.
But in this hypothetical the insurance company has perfect information, so they won’t sell you that policy that has to be paid out for more than you’ve contributed.
It’s just a thought experiment, but the more information they have on us, the more relevant it becomes.
They can’t predict the future, they can’t predict exactly when you’ll get in an accident.
Perfect information means they know your risk level to the best possible accuracy, which would really only apply to populations.
Perfect information means they insure 1000 people and predict they’ll have one bad accident per year. After ten years they covered for ten accidents. All ten could have occurred in the first year and they would still be correct.
Sure. I can’t speak for Scoundreller, but I don’t think it was meant to be particularly interesting, just pointing out to wbl that if insurance was to become perfectly fair, it will also have become pointless.
There’s a pretty big leap between perfectly fair and having perfect knowledge of the future. You can know that a fair coin gives exactly 50% chance to flip heads or tails without knowing what the outcome of the flip is going to be.
The original context was unsafe drivers, with the gist of the response being that when you have everyone paying for exactly the costs they themselves incur, insurance has become meaningless.
It’s hyperbole of sorts, but it highlights that until such a time, raising the cost of insurance doesn’t just punish the people who actually cause the damage.
Except insurance also covers for costs that are not your fault or not anyone’s fault. An insurance premium could be divided into two components: one based on individual risk and the other component based on no-fault risk that applies to pretty much everyone equally. How are you going to get bad drivers to pay for hail damage? How are you going to get bad drivers to pay for a tree falling on your car? How are you going to get bad drivers to pay for an accident caused by a random tire blowout?
The personal risk component can be accounted by “perfect” information and that component can get bigger or smaller depending on your definition of perfect, but there’s another component which can’t.
What if you have perfect information and conclude that your chance of being in a terrible accident, requiring tens of millions of dollars of medical care and property repair, is 1% for every billion miles of driving, and you’re the unlucky person who happened to be behind the wheel. Will you have enough money in your piggy bank to cover it?
Maybe you could argue you shouldn’t have to cover medical expenses if we had a single payer system—the money to mitigate medical risk from driving still has to come from somewhere.
Maybe you could argue that damage to property should come from those property owners’ insurance.
What if you don’t get into a terrible accident, you just get into a boring accident where you total your car and don’t hurt anyone. You know the odds of this are low, chances are it won’t happen in your life, but it will probably happen to someone you know. What if it happens to you, when you’re very young and have a new car? You haven’t had a chance to put away any money in your piggy bank yet. You need to replace your car now. How does a piggy bank help you?
'Unsafe' is a subjective label often applied haphazardly. I find people quick to label others as 'unsafe' tend to behave in unsafe ways of their own on the road, such as parking themselves in the left lane at or below the speed limit and refusing to move over.
Oh yeah, I loved paying more than the purchase price of the car per year as a young driver. I have since crashed 0 cars, I am yet to kill anyone. In fact, the car mechanically totaled itself before I got to finish my years worth of insurance. On which I never got a refund on.
The pay-as-you-drive model might paint a picture of a more equal playing field, where everyone pays what they are due, but I think it is mostly a sham to squeeze more money out of people.
At least GasBuddy still works well if you only share while using app and you can generally do without precise location.
In terms of its benefits, at least in Canada, it saves me a bunch of dough on gas. Highly recommended.
Though I like the French approach where gas prices are all on a government website without needing to depend on crowd-sourcing: https://www.prix-carburants.gouv.fr/
I’m more worried about a navigation app spying on me.
At least in my area, it’s helped me realize which city I’m usually in typically has the best prices and realized that prices tend to go up during the day and drop late in the evening.
There’s about a 13% difference between the cheapest and most expensive at the moment.
It's not that extreme by me I guess, but it's at least 10 minutes minimum. And if you have a car with a small tank and decent mpg, its just not really worth the trip to costco or the time just to save $2 (20-30c a gallon).
My gap is widening and is a three-digit dollar savings at this point, even with the paid subscription for their savings gas card. I’m a rural commuter in a hybrid running ~500 miles a week in the field and it definitely stretches the gas dollars.
Not the person you asked, but I normally buy a fair amount of gas and I use GasBuddy. Actually, I pay for GasBuddy by the month at $9.99 for discounts on gas and roadside assistance. I've done so for years.
For various personal reasons I haven't been driving a ton lately, so I'll pick a more typical month.
My direct discounts were $17.88 in that month on gasoline that was priced at $269.49, at a cost of $9.99, for a total savings of $7.89.
Plus whatever extra I might have paid if I didn't use GasBuddy at all to help with planning to buy gas where the price at the pump is cheapest (which anyone can do without a subscription or installing an app). The potential savings gained by being very selective of where I buy gas, using information provided by GasBuddy, is impossible for me to tabulate.
If I had to guess, then I would guess that being selective saves me an average of around 10%. So, about $27 in my example month.
(I do not participate in any of GasBuddy's drive-tracking programs, and it only knows my location when the app is open. Drive tracking is a thing they offer people to opt into (for "free") but I could see the writing on the wall with that, vis-a-vis this Allstate incident.)
This may not be true of all such apps, but I open mine and it subdivides trips by every stop (including stop signs and lights). Then when you don't do it for 3-4 days, you have 50 trips you need to click through.
My time was worth more than the savings they offered. In fact, it was insulting.
I should not have to spend 3 hours per month categorizing my trips for their data mining expedition for a $30 discount (especially not one that evaporated at the next renewal, because I wasn't safe enough). That app was rigged.
I tend to agree. But how do you avoid that experience becoming terrible? Competition between insurance companies does keep the experience good and probably keeps the price down.
simple: you build the cost into the price of gasoline and electric charging. one risk pool and removing advertising probably also keeps the price down.
It makes sense to share your data with an insurance provider if you are a well-behaved driver. It is well-behaved drivers that subside the insurance cost of reckless drivers.
There are two assumptions hidden in that which may not be true.
First, that the way the insurance company uses the data is accurate. For example, if you work nights, noticing that you drive at night might cause them to raise your rates, because more of the people who drive at night are drunk. You're not drunk, you're just commuting, but now you're getting punished for giving them data.
Second, that you know what "well-behaved" is, because people who know they're being tracked will try to conform to what they think will give them a better score. So now you're focusing your attention on strictly adhering to some mental model of what you think the tracking device wants to see in order to lower your rates instead of leaving that attention to focus on actual hazards. Then you end up having to brake rapidly because you notice an actual hazard too late, which makes the tracking device penalize you more than all of your machinations were helping, or even causes you to get into a collision.
> It makes sense to share your data with an insurance provider if you are a well-behaved driver.
I would like to see evidence of this.
My experience: discount for sharing is low, maybe $100/year.
My expectation: if I actually get in an accident (1) the data I shared will be used against me; (2) my insurance provider will put me in a high risk pool or drop me.
It seems to me that the downside is really high and the upside is not.
Is that wrong?
P.S. If I actually had faith in all of the companies involved, I would certainly agree with the parent. However, I don't.
I mean, $100 a year is 15% of my bill - I might have to look into the data sharing stuff. I guess if you're already burning money on comprehensive and large deductibles that might not matter as much to you.
Rest of your analysis seems spot on. Important to note however that getting in accidents already increases your premiums - I don't really think the data sharing is gonna change that much.
They do pay higher premiums but that might not be enough to pay for a claim. The economics are often such that you price low risk drivers to subsidize higher risk drivers if realistically pricing the high risk drivers is not economical, i.e. would lead to losing business. It's a balance.
As another commenter pointed out, if insurance companies had a crystal ball and could perfectly predict each customer's probably of having an accident, and charged them an individualized rate based on future payouts, then we would not need insurance. You could just put those premiums in a savings account and pay for the accident(s) when they happen.
The whole point of insurance is it's a shared risk pool. Arguments like "Bad drivers make YOUR rates go up" just serve to pit customer vs. customer while insurance companies profit.
That would only be true if people could expect multiple insurance events over their life time.
But, if the expected time to an insurance payout for a good driver is longer than their life time, then good drivers will never have enough money in their account to cover an accident that occurs.
If you had a perfectly accurate crystal ball and the number of accidents you would have in your lifetime is zero then your account wouldn't need to have any money in it.
It's true that perfectly accurate crystal balls would render insurance irrelevant. It doesn't strike me as that interesting of a revelation, but it is true.
I guess it's a response to people saying "They need to charge bad drivers more that they do, in proportion to how badly they drive." If insurance could accurately do that, it wouldn't be insurance--it would just be an individualized pre-paid accident savings account.
More to the point, it erodes the value of the system as insurance.
Suppose the insurance thinks that Alice has a 90% probability of an accident and Carol has a 15% probability, so they want to charge Alice six times the premiums of Carol. Then in practice Carol is the one who has the accident and not Alice, because it's not perfect.
But the pool Alice is in is much smaller than the other one, so if they were merged, the combined group would only be paying 10% more than Carol does, which would be serving the purpose of insurance -- spreading risk. Whereas if you separate them, Alice is screwed -- even though she isn't even going to have an accident -- because now she has to pay >$7000/year in insurance rather than ~$1300 when you combine these imperfect predictions with smaller risk pools.
as another commenter said, they lose money on the policies and paying out. they make money on investing the policies before they have to pay out the claims.
The data collection methods are not foolproof and you can get dinged for "hard braking" even if you aren't. You're basically opting into a temporary rate decrease until they see data coming in that they don't like.
Are these apps able to track location and accelerometer data from the background? Or does this only apply to people who open something like GasBuddy while driving the car?
For GasBuddy, specifically, it was used for an opt-in feature known as "Trips" whose purpose was to show you driving insights such as those mentioned in the article (hard braking, hard turning, hard acceleration). Not enabled by default and looks to have been removed at some point.
There are too many proactively dangerous drivers on the road. And if these invasive methods make it harder for them to continue to drive this way, I am OK with it.
me too, but in the recent political climate any type of enforcement has been out of fashion. I guess things will change soon, and probably too quickly and too far (as they always do).
Surveillance doesn't make it harder to commit crimes. It might help catching people after the fact a little bit easier, but people committing crimes aren't really considering future consequences.
This is your friendly reminder that the insurance industry does not generally make money on underwriting. In other words they are losing money on insuring you. Typically the costs of the insurance side of the business are greater than their revenue from premiums. This typically means that insurance actually is a good deal on net. They make their money on investing the float. They can only do that if they attract premium payers by having the most competitive pricing. They can only have competitive pricing by having accurate data and models.
Allstate, for example, is billions of dollars in the hole on paying out claims for the past decade. Most insurance companies are.
Edit: I'm talking about underwriting losses. Selling insurance and paying claims is a sort of loss leader for the insurance industry. Insurance companies are net losers on paying out claims, but they make enough back by investing the premiums before claims have to get paid, that it is still a net profitable business.
Essentially insurance companies are loaned money by their customers since they are selling a promise of future payment on claims. They invest this pool of money and keep the gains until it is time to pay out on claims. Many companies (like allstate) pay out more in claims than they collect in premiums. For 2022 and 2023 they paid out 5 billion more in claims than they took in in premiums. Those were pretty bad years, but over the long run, just about every insurance company loses money this way.
Additional Edit: this comment is particular to the big property and casualty insurance products (home and car insurance on the consumer side). Best Buy extended warranties, health insurance, and other products may have business models that rhyme, but that don't follow this exact fact pattern.
Generally agree about InsuranceCos, but I don't think your claim is true about profitability of core insurance operations.
Their combined ratios (insurance payouts + expenses / premiums collected) are closely tracked by investors and if they're creeping up to 100% without a systemic reason that affects all InsuranceCos, they get a lot of scrutiny.
Here's just one lens [0] for home insurance going back to 2004 (~20 years). It appears that the combined ratios were under 100% for ~8 years.
>Selling insurance and paying claims is a sort of loss leader for the insurance industry.
Now you say
> I didn't say that they ALWAYS post losses
And
> that insurance premiums are a net money loser for the industry.
This is circular. As for the article, it shows the context that the last couple years have been exceptional and core operations aren't sustainable.
Your reasoning doesn't explain why these companies are cutting coverage, skyrocketing premiums, and leaving markets. They wouldn't be retrenching if the status quo was absorbing losses.
> In 2023, insurers lost money on homeowners coverage in 18 states, more than a third of the country, according to a New York Times analysis of newly available financial data. That’s up from 12 states five years ago, and eight states in 2013. The result is that insurance companies are raising premiums by as much as 50% or more, cutting back on coverage or leaving entire states altogether. Nationally, over the last decade, insurers paid out more in claims than they received in premiums, according to the ratings firm Moody’s, and those losses are increasing. [0]
The original discussion about insurance companies accepting small underwriting losses to gain investment capital (float) refers to the traditional insurance business model operating under normal conditions. This is different from the current market disruptions we're seeing in specific high-risk regions.
Insurance companies are indeed leaving certain markets and raising premiums dramatically, but this is happening because:
1. Climate Change Risk Unpredictability
* Traditional insurance models rely on being able to predict risk with reasonable accuracy
* Climate change is making weather-related disasters more frequent and severe
* Historical data becomes less reliable for predicting future losses
* This uncertainty makes it impossible to price policies appropriately
2. Regulatory Constraints
* State regulators often limit how much insurers can charge for coverage
* Companies can't price premiums high enough to cover increasing risks
* They're forced to choose between unsustainable losses or market exit
* Political pressure often prevents charging actuarially sound rates
3. Concentration of Risk
* Some areas face multiple overlapping risks (fire, flood, hurricane)
* Large-scale disasters can trigger many claims simultaneously
* This violates the insurance principle of risk diversification
* Even investment returns can't offset such concentrated losses
4. Scale of Potential Losses
* Traditional model accepts small predictable underwriting losses
* Current climate risks create potential for catastrophic losses
* Example: California wildfires can destroy entire communities at once
* No amount of investment income can offset such massive losses
5. Market Structure Issues
* Some markets require insurers to take all risks (can't be selective)
* Cross-subsidization between markets becoming unsustainable
* Limited ability to diversify within regulated markets
While the traditional insurance model can handle planned small underwriting losses offset by investment gains, that model breaks down when facing large-scale unpredictable risks that can't be properly priced or diversified. This explains why insurers are withdrawing from certain markets while still operating profitably in others.
> refers to the traditional insurance business model operating under normal conditions. This is different from the current market disruptions we're seeing in specific high-risk regions.
I've yet to see a capital markets discussion with an InsuranceCo where someone says, "Oh! Your combined ratio under normal conditions is at/over 100%. Good job management team!" Please provide some sources supporting combined ratios >= 100% as OK/sustainable, or this will continue to be circular.
> Insurance companies are indeed leaving certain markets and raising premiums dramatically, but this is happening because...
Dude, your #1-5 just listed reasons why the core insurance operations have to be profitable, or they cease operations. That's making my point. You made no reference to their investing activities overcoming losses as you implied. If they can't price risk effectively in their core business "under normal conditions," then they will get competed out of the market.
> I've yet to see a capital markets discussion with an InsuranceCo where someone says, "Oh! Your combined ratio under normal conditions is at/over 100%. Good job management team!" Please provide some sources supporting combined ratios >= 100% as OK/sustainable, or this will continue to be circular.
Dude, your source shows that over a multi-decade time-span the industry loses underwriting money more often than it makes money (12/20 years), and if you tally it out in the long run, they are net negative (101.5 dollars paid per 100 in premium income). https://www.spglobal.com/marketintelligence/en/news-insights...
Of course they aren't going to congratulate themselves on losing money on underwriting in those words. The way you will see it phrased is congratulating themselves on increased premium income from sales, while elsewhere in the report acknowledging underwriting losses. Maintaining the same rate of underwriting loss while increasing sales will also increase the total loss, but it is not at all rare to see an investor report congratulating management on increased sales in a circumstance where a larger loss was made.
> Dude, your #1-5 just listed reasons why the core insurance operations have to be profitable, or they cease operations. That's making my point. You made no reference to their investing activities overcoming losses as you implied. If they can't price risk effectively in their core business "under normal conditions," then they will get competed out of the market.
You asked me to address those specific circumstances, which I noted are largely separated from the aggregate of all circumstances and markets that drive the insurance industry as a whole. You are missing my big point; the insurance industry do not get rich by charging you premiums higher than your claims. Any profits they make are slim, and normally eaten up by losses in the long term (again, see your own source). The insurance operations don't have to be profitable, and they aren't according to both your source and me.
They have to be predictable so that you can price your premiums in a way that the underwriting losses aren't greater than your investment income. Those reasons I listed are what make those policies unpredictable, un-priceable effectively, and therefore not an acceptable way to lose money.
I didn't address the fact that they make their losses back in investments because it was the whole point of my original post, and all of the sourcing that I have done; it is broadly made in the quarterly reports I linked to, as well as the letters to investors from Warren Buffet.
Per your own source and decades of investor documents, the industry posts underwriting losses constantly, and makes their profits by investing the float. It isn't a big secret.
no worries, not trying to be pedantic, just wading into new territory with a lot of questions. Sometimes financial equations take unintuitive forms for niche use cases.
They make a profit on investing the float that offsets their underwriting loss. I never claimed that insurance companies aren't profitable. I said that their insurance is a money loser.
But... they don't have money to invest without the insurance premiums coming in. My family finances look a lot worse if you exclude interest and investment proceeds at random from the analysis, too.
$10-20B profit annually is not something I'd describe as "in the hole". The "hole" is overflowing with money.
If you keep selling a product for $95 that costs you $100 to deliver you will be "in the hole" on that product line. For Allstate, that product line is insurance.
If you invest that $95, get a $15 dollar return, and then deliver the $100 product, you have made a net profit even though you sold the product for below cost. You are still in the hole for the insurance product no matter how you slice it.
Yes, I get that the insurance product is what allows them to invest the float. But any accountant can tell you that they are in the hole on that product.
They are still selling the product at a loss, even if their time arbitrage lets them make money elsewhere. Its kind of magic in that its a win/win scenario in what looks like a zero sum game. The point is that people think that insurance companies make money off of them paying a premium higher than what they will recover in claims, and saying that insurance is always a bad deal for the consumer. My point is that isn't true. Insurance companies pay out more in underwriting costs than they collect in premiums. Insurance is very frequently and insanely good deal in aggregate.
is it really a float? Im not an accountant, but when I think of float, I think of the balance in an account that regularly clears.
Is it analogous to interest in household checking account where income = bills, or is it analogous to the same household with a large retirement account that sweeps the checking and pays bills.
Maybe another way of asking this is whats the relative size of the interest bearing investments to the annual premium collection.
Can you show me where this is listed in the allstate financial results?
The float in the insurance industry is simply the term of art for the pool of money from premiums paid that is available for investment. On that particular statement, I would look at page 8. The assets are, more or less, the float which comes pretty directly from the liabilities of unearned premiums and estimated future claim payments in the liabilities.
It is hardly a coincidence that their unearned premiums + claim payment reserve happens to be roughly the amount they have invested ($65b and $66b respectively).
Okay, so if I'm reading this correctly, annual premium income is around $30 billion, annual insurance payouts is also around $30 billion ( cost of administration). Investment assets under management around 66 billion.
This is all right, I still don't see how the insurance business unit add value unless there are profitable years on average.
I'll have to check out the letters, maybe they will explain this.
Edit: Having read that section of the letter, my understanding is the existence of insurance companies relies on the belief that in the long run, the insurance side will not be a loss leader, at least for insurance types without a significant time lag between customer acquisition and insurance payout. For example, operating a "loss leader" could be profitable for something like life insurance where there average claim for a new customer is in the future.
It seems like insurance types where there is no intrinsic time delay between customer acquisition and payout (e.g. auto, home) can not operate this way. That is to say, you cant build up a profitable interest bearing float if average policy payouts are >100% in month 1 of the policy.
Insurance companies' investments come almost entirely from premiums that customers pay. These premiums show up as liabilities on the balance sheet because they represent future claims the company expects to pay out. Think of it this way: customers are paying now for a service (insurance coverage) they might need later.
Let's use an example: If an insurance company has $66 billion in premiums collected, they know they'll likely need to pay out about $66 billion in claims over the coming years. Instead of letting this money sit idle in a bank account, they invest it.
This is where the two sides of an insurance company come into play:
- The underwriting side (selling insurance policies) collects premiums
- The investment side uses these premiums as capital to make investments
It's similar to a loan system, but with a twist. When customers pay premiums, they're essentially "lending" money to the insurance company. This "loan" only gets "repaid" when the customer files a claim. Meanwhile, the insurance company invests the premium money. While they eventually have to pay out claims (repay the "loan"), they get to keep all the investment profits they made.
This explains why insurance companies often continue selling insurance even if they lose a small amount on the underwriting side - it's like getting a very cheap loan. Historical data shows insurance companies lose about 1.5% per year on underwriting. That's their effective borrowing cost, which is much cheaper than other forms of borrowing.
Why don't they just raise prices to make both underwriting and investments profitable? Because insurance is highly price-competitive. Customers will quickly switch companies for a better rate. If Company A tries to make a 2.5% profit on underwriting, while Company B is willing to lose 2.5%, Company B's prices will be 5% lower - and they'll attract more customers, giving them more premium money to invest.
please see my edit about the relevance of time delay. Im curious what you will say. Using the loan analogy, I understand how a firm can make money on loans with a repayment delay. This doesnt make sense to me if the payout is >100% and the average time delay converges to 0.
No, that's not accurate. The time delay doesn't converge to 0 just because payouts match incoming premiums. Here's why:
Think of it like a water tank:
- The tank contains 66B gallons (total liabilities)
- 30B gallons flow in annually (new premiums)
- 30B gallons flow out annually (claim payments)
- The tank stays at 66B gallons (stable liability pool)
Even though the annual inflow equals the outflow (30B), it would still take about 2.2 years to drain the entire tank (66B/30B = 2.2) if you stopped adding new water. This is the average time delay.
The matching of annual inflows and outflows just means the system is in steady state; it doesn't affect the average duration of how long money stays in the system. That duration is determined by:
- Total liability pool ($66B) divided by
- Annual payout rate ($30B)
Another way to think about it:
- Each premium dollar collected today is promised against future claims
- Those future claims are spread out over the next several years
- Even as old claims are paid, new premiums create new future obligations
- The ratio of total obligations to annual payments (66/30) determines the average delay
So while the annual cash flows may match, the time delay is a structural feature of how insurance obligations are spread out over time. The matching of annual inflows and outflows maintains the system's stability but doesn't eliminate the time delay inherent in the insurance model.
It is intrinsic to the nature of insurance that there is a time delay. Even if the insured were to suffer a loss on the same day that they paid their premium, there will still be a delay. Even the most efficient, benevolent insurance operation cannot process a claim, value a loss, and settle the claim within a day.
I get how you can have a profitable steady state in with equal inflow and outflows given a surplus tank.
I get how even an immediate claim takes time to settle.
Would you agree that this puts an upper limit on the loss they can run? If more aggregate auto claims are submitted than premiums paid in a day, you can only make interest on they delay duration. If payout delay is say 3 months, annual interest is 4%, you break even at a 1% loss on premiums, right?
If you are consistently drawing down your float to pay claims instead of adding to it, you are better off leaving the auto insurance industry and simply operating an equity investment firm.
> Would you agree that this puts an upper limit on the loss they can run? If more aggregate auto claims are submitted than premiums paid in a day, you can only make interest on they delay duration. If payout delay is say 3 months, annual interest is 4%, you break even at a 1% loss on premiums, right?
Of course there is an upper limit on losses they can run. The upper limit is base on their investment returns and the time frame. The time frame is longer than you think. Based on a quick run of the numbers it is measured in years.
> If you are consistently drawing down your float to pay claims instead of adding to it, you are better off leaving the auto insurance industry and simply operating an equity investment firm.
They are continually replenishing the float at the same time. The whole point is that they are operating an equity investment firm, except the investment capital comes from a revolving pool of insurance premiums instead of some other pool of money.
There's something I'm still not getting. Car insurance should not have a significant lag because the probability of claim is not related to the date of policy. I'm just as likely to get in an accident on day 1 of my policy as day 1,000.
If daily claims meet or exceed the daily premium, the only time available to earn interest is between premium receipt and claim payment.
Imagine founding an insurance company and on day one you receive $100 in premiums and $100 in claims.
That's for the corporation as a whole and counts the net of investment profits and insurance losses. If you open their earnings release, you can see combined ratios over 100%, indicating that the total cost of operating the insurance is negative, but this is offset by investment profits (as the original commenter mentioned - they make money off the float).
The insurer losing money insuring your does not mean it's a good deal on net.
I pay $100, they pay out $90 on my behalf in that time period, total cost to them is $110 because overhead. That it cost them $110 and I only paid $100 doesn't mean I'm not losing $10.
Of course they actually make $111 because they invested my money but that's beside the point.
Seems like the bigger part of the story is at the bottom. You can uninstall GasBuddy from your phone but finding and buying a new car that doesn't track you is a bigger hassle.