Hacker News Reader

Comment by whack | Hacker News Reader

Discriminating against insurance customers on the basis of things they cannot control, like their age or gender, is not good.

Discriminating against customers on the basis of driving habits that they have control over, like how fast they drive, or how hard they swerve, is completely fine.

If driver A drives safely at the speed limit, and driver B constantly pulls stunts at speeds well over the speed limit, B is much more likely to get into an accident than A. Why should A have his premiums raised just to subsidize B's voluntary driving habits?

By making B pay for the costs of his unsafe driving habits, and rewarding A for his safe driving habits, this feels like a great marketplace improvement.

View on HN

How about a real world scenario:

- driver C who drives fast yet safe (over the speed limit like most drivers) but has to deal with:

- driver D who drives the speed limit in the fast lane, changes lanes without signaling, is essentially clueless, and forces C drivers to navigate around them.

Stunts at any speed is bad, sure, but speed can't automatically be assumed to be stunts. That's the flaw.

If the behavior is not predictively relate increased claims activity, then it wouldn't be penalized. They won't just say "driving 5 mph over the speed limit on the highways raises rates" for no reason. The reality is probably though, at some speed increase above the posted limits there is a relationship with claims activity. If that limit is 4 mph or 10 mph, I think it's fair to charge more.

And can car computers tell if blinkers are on? At some point, changing without signaling is likely to be trackable if it isn't already. Ultimately, clueless driving will leave trackable indicators. This is driving, it's behavior based, and the outcomes are observable.

At any rate, I hope this makes usage-based insurance more common. I currently use MetroMile as my car insurer, which charges per mile of car usage. I'm a city-driver, my miles are insanely low. I'm glad to be benefiting from smarter, behavior-based insurance even if it is in its infancy.

> If the behavior is not predictively relate increased claims activity, then it wouldn't be penalized

yeah, sure. i totally believe that. ever actually deal with an insurance company? they pretty much do whatever the hell they want unless there's enough people willing to sue in a class action suit. this goes for consumer, business, financial, whatever.

when this becomes mainstream i will go with alternative insurance providers that do not track my driving habits even if it comes at a huge premium.

> they pretty much do whatever the hell they want unless there's enough people willing to sue in a class action suit. this goes for consumer, business, financial, whatever.

Okay fine, please explain to me how you think they would be worse with behavior-based rates than they are today. If they're just looking to bump rates for everyone on a flimsy pretext, there is nothing stopping them from pushing the price envelope as high as they can right now. Well, except the competitive environment they operate in, of course, which isn't going away even with behavior-based rates. They already do whatever they want, so this lets them do it more? It seems unlikely to me, and my early experience is incredibly positive.

Behavior-based rates can at least remove some of the mystery behind rates rather than add to it. Behavior is personally trackable, shareable, and demonstrable. Ultimately, it's a way to increase price discrimination, something I'm a fan of.

> If that limit is 4 mph or 10 mph, I think it's fair to charge more.

But there's already telltales for that: if a driver consistently drives more than 10 mph over the speed limit, they'll eventually get caught and ticketed.

That's where age & experience come into play. If a driver has 20 years of experience with no tickets nor accidents, does any of the other stuff really matter? It's not just observable, it's proof - instead of assumption.

Yes, because there are many people with 20 years of experience with no tickets nor accidents, and they are not all equally likely to file a claim next year. It's not proof or assumption. It's actuarial science.

Those telltales are reactive, not predictive. If someone is consistently driving over 10 mph then they should be charged higher rates even if they themselves have not had an accident because they are engaging in quantifiably riskier (to themselves, others, and the claims department) behavior. I'm sure there's people out there that drive like maniacs 90%+ of the time and haven't had an accident, but that doesn't mean they aren't significantly more likely to do so in the future.

Knowing people engage in behavior objectively quantifiable as more risky is also strong evidence of increased level of risk for that driver, not an assumption that the given individual will have a given outcome.

200 people a year die from nuts and peanuts in the US, therefore eating nuts and bringing them to social situations is "quantifiably [risky] (to themselves, others, and the claims department) behavior."

There is a subtlety here in correlation and causation.

>I'm sure there's people out there that drive like maniacs 90%+ of the time and haven't had an accident, but that doesn't mean they aren't significantly more likely to do so in the future.

It's not 'they', it's 'their cohort', and the apparent risk can be very sensitive to the level of detail and factors accounted for - such as nut allergies.

As long as we know the ruler IS bent - because we need some way to charge people appropriately for insurance.

> It's not 'they', it's 'their cohort'

I would argue it is both: the "drive like a maniac" cohort is likely to see a higher incident rate across the population, and that a given individual exhibiting "drive like a maniac" behavior is more likely to have an incident themselves vs. a given individual that does not "drive like a maniac."

This is driving in the open world, not art you create in your basement. Drivers are not fully in control of their environment, and riskier driving behaviors are risky even for fantastic drivers because they cannot account for how others will react to their "mania", to keep the vocabulary consistent.

Mania is quite subjective and increasing the difficulty of this conversation :) Yes, someone I'd call a 100% maniac very likely would have poor individual outcomes as well.

You could chart the relationship between levels of mania and some of the unaccounted factors, like driving skill or restraint. Slippery slope ends up in a trap of having to account for everything for the sake of fairness.

> If the behavior is not predictively relate increased claims activity, then it wouldn't be penalized. They won't just say "driving 5 mph over the speed limit on the highways raises rates" for no reason.

You're assuming the insurance company has perfect information.

That's the problem with all of this. The insurance company isn't in the car. They don't know why you're doing what you're doing.

And the problem is you get what you measure. Changing lanes without signalling will make an accident more likely 85% of the time but make it less likely 15% of the time. If you're penalized 100% of the time then all the people who were doing the right thing in every case will now, in order to keep their insurance rates low, start doing the wrong thing 15% of the time.

Perhaps the insurance companies' models would (with enough data) learn to use 85% as the optimal value, and penalise deviation from that, rather than deviation from 100%

So then people will look at their statistics and start making unnecessary lane changes and so on for the sole purpose of fixing their numbers.

Unless you don't actually tell them what you expect them to do. Then you end up creating a whole new class of crazy folklore about what to do to lower your insurance rates that people follow religiously even when it's obviously irrational.

If you want a computer to decide how to drive then get a self-driving car.

If that limit is 4 mph or 10 mph, I think it's fair to charge more.

The problem is, speed limits in the US are often underposted to raise revenue and/or give the police an excuse to stop almost any driver. Certain studies have shown that drivers doing 4-to-10 MPH over the speed limit are among the safest. The slower ones are often either not paying attention, actively obstructing traffic, or impaired.

I'd be more likely to accept behavior-based insurance rates if they're based on sound traffic engineering principles rather than politics.

Yep, it would penalize those who have high risk markers but are nonetheless actually safer. Seems to be the same problem as charging more to an individual 16 year old who is actually a safe driver.

They charge the 16 year old more today. Better data collection would lower his rates.

They already do in the UK, lots of young drivers can only afford to drive if they have a "black box" fitted to their car by the insurer to track their driving behaviour.

I recall that one of the benchmarks used for one of the car-tracking-fob-based insurance policies was how "hard" they were braking and how often, and possibly high acceleration. Repeated hard braking would imply tailgating and distracted driving, and repeated high acceleration would imply generally reckless driving.

I have an Android phone and use the Google app "My Tracks" to record walks I go on. Sometimes the rate that it tells me I'm walking jumps wildly high, even though I walk at a very constant rate. Sometimes it's 2x what it really is.

I would hate to be a good driver and be penalized for bad software/hardware. Given the track record of the auto companies (on which the insurance companies will rely), I think it's a given the software will be really shitty.

Mine used to reliably have me teleport into the United States then back into Canada during my morning commute. They've since fixed that bug :)

Edit: it wasn't actually teleporting, but it had me driving across fields, mountains, rivers, and the boarder all in about 1 or 2 minutes.

The answer to your question is - because speed is not the only thing you can judge "safeness" of driving by.

If the driver A drives at a speed limit, but is drunk, or 90 years old, or playing with their phone all the time, or just got their licence two days earlier, they might be actually much more likely to get into an accident than someone who is going faster than the speed limit but is in perfect mental and physical condition.

I'm not trying to excuse speeding, I'm just saying there's a hundred other factors that are just as important when judging how good a driver someone is. And I feel like insurance companies won't care - they will use the sensor data to say "hey, you accelerate really fast from the traffic lights, that must mean you are a really shitty driver and you have to pay more than other drivers!"

You can absolutely discriminate based on age and gender for insurance purposes, though. Men are charged higher rates than women. Teenagers are charged higher rates than those in their 20's or 30's. A newly licensed 16 year old male will cost more than a newly licensed 26 year old male. My insurance went down when I got married (which is not outside of my control, but is still a protected class).

If there is valid statistical evidence that a certain group of people (male teenagers with divorced parents, single women in their 40's earning more than $60k a year, divorced white men in Massachusetts, whatever) is likely to behave in riskier behavior for whatever reason, you can bet an insurance company somewhere has the data and is charging accordingly.

The parent was saying it was not good, not that it was illegal, as your comment seems to be addressing.

You can, but you shouldn't. I hate to use the term slippery slope, but it's crazy to me that they still do this. You could easily apply to same logic to race and prison statistics and how officers treat who they're dealing with, who you hire, etc.

Much better to do it on an person by person basis.

It is much better, that's why insurers try their best to do so. The more data is available, the more feasible that is.

> If there is valid statistical evidence

Is that true? Can they charge me a different rate because of my race or religion?

It's a grey area legally, and is usually prevented by ethics committees and industry bodies. Practically, if you collect the information you run the risk of a large anti-discrimination suit down the track. The usual proxy for race/religion is first name, last name and zip-code - name clustering gives very strong risk differentiation (I crunch numbers for insurance companies).

Depends if the jurisdiction you buy insurance in outlaws it.

This is a good overview:

http://nakedlaw.avvo.com/consumer-protection/why-is-it-legal...

What is currently legal to do and what is "right" to do are two different things. GP was talking about what "should" be. Of course, a lot of that is opinion that is up for debate.

I personally agree. I think charging males more (for example) was only done because we had no way of getting more individualized data. It is all very crude and probably will be an outdated concept in 20 years.

*It is illegal to discriminate based on sex for auto insurance purposes in many countries as well as Montana. Insurance companies in most US states take your credit score into consideration (if you don't pay your credit card bill you pay more for car insurance) but this practice is illegal in a handful of states. Massachusetts is one of them, if I recall correctly.

I get the logic, and I'm personally fine with this approach[1], but there's a tradeoff between privacy and non-freeloading (don't know the technical term). Some types of data collection are so "icky" that people prefer to tolerate free loaders, and this feeling is common enough not to cause the adverse selection problem that forces everyone to accept the loss of privacy or pay impossible rates.

Consider the extreme case: let's say technology allows an external observer to directly read your thoughts as they happen.[2] Hey, people that think worse things are riskier, so what's the problem if they get higher rates? If you don't want higher rates, don't think about cutting through traffic and running red lights!

I just hope no one breaches that database and learns what you think of Suzy...

[1] as a safe driver long resentful of the statistics of my demographic

[2] Because of the "control/choice" issue, assume there's a difference between deliberate and unchosen thoughts and the device only takes the former.

But again, thinking about doing something and actually doing it makes all the difference. You do not have 100% control over what your think. Sometimes you think about something and then decide whether to do it it not by predicting the possible outcomes. I swear i have thought about murdering several people, but of course i will never follow through. And I'm sure everybody does that. It's part of our critical thinking. We think of something and then weigh the pros and cons in our mind and then decide what to do.

See footnote [2], and the problem is that (as with driving styles) there may turn out to be a correlation between how often you think about murder and how often you do it, which can be used to calibrate insurance-risk premiums.

Sure, not every thought about murder leads to murder, and not every red light you run leads to an accident. But people that run more red lights have more accidents, and people who think about bad things more (let's posit) will do more bad things.

"Oh, you only think about murder 3 times a week? Discount!"

Still ready to let your thoughts about Suzy get archived on some hard drive?

All of that does not matter one bit to an insurer. If what you think (or what someone manages to measure of it) can be used to predict damages better than without the data, that allows the insurer to offer those with predicted lower risk a lower premium and thus gain them as customers. It doesn't matter that you possibly know that you would never follow through--if people who think about murdering other people on average are more likely to actually do so and they cannot distinguish between those who do follow through and you, it is rational for them to put you in the same category.

It's not freeloading. The WHOLE POINT of insurance is amortizing risk across a wider range to individuals. If you can correctly predict risk exactly for every single individual then the whole model is pointless.

The "whole point" of insurance does not require vastly different risk classes to be lumped together and pay the same premiums; just the opposite. Someone who is known to carry a higher risk but pays as if they had a lower risk, can reasonably be characterized as freeloading, and that is the sense in which I meant the term -- not the mere sense of "they made a claim".

(Imagine e.g. someone in Florida being rated for hurricane insurance on the mistaken assumption they lived in Kansas, or a cab driver rated as if the car was just used for a daily commute.)

That insight in no way contradicts the principle that a completely predictable event is not insurable, or that insurance allows distribution of the costs of an event across a class.

By that definition though, anyone who comes out "ahead" on their premium is a freeloader. Perhaps a less charged term is needed? Freeloading sounds like this person has done something morally wrong, when in fact that have just used insurance as it is meant to be used.

I've thought about this since my post, and I think a better term for the concept I intended, would be "risk assessment accuracy". That is, you can increase RAA, but at the cost of a loss of some privacy -- that's the dynamic I was describing.

But I still think it's fair to call someone a freeloader if they deliberately get mixed in with a low risk class while knowing they're higher risk.

Edit: And you're right, that in the case of perfect RAA, the concept of insurance breaks down because it means you know people either will or won't experience the insured event.

Generally i think companies leverage data to ratchet rates upwards, not offer unsolicited discounts. Why invest in infrastructure that will reduce profit? That's why when I drive through an ez-pass-only onramp without an ez-pass, they know exactly where my car entered and exited, but they still charge me the max-fare for not having a ticket.

I agree that insurers will not offer unprovoked discounts. But the accessibility of this data will allow competitors to more accurately estimate accident likelihoods at the individual level, price the insured risks, and undercut the incumbents for a subset of the population.

Unless, of course, the federal government mandates that insurers cover all drivers at the same rates, regardless of pre-existing reckless driving habits.

There is massive competition in the auto insurance industry. I was about to say that this would imply that profits wouldn't go up because folks would save 10% by switching to GEICO.

However, I realized a problem.

Currently, all insurers are looking at the same demographic data. If this scheme is put into practice, then insurers will have a proprietary information on an individuals' driving habits, letting them offer discounts to good drivers. This leads to a situation where to switch to a different insurer will always cost more for a good driver, so few drivers will switch. This means that trying to switch insurers will become a signal of bad driving, so insurers will charge more to new customers. This leads to fewer and fewer people switching insurance companies, leading to decreased competition.

Unless you have access to your own data. Perhaps the insurance companies could be forced to share it upon user request, for example. Then the new company knows everything the old one did.

Sure, but that would require a change in legislation - one that doesn't seem very likely in America.

This is exactly the goal - these systems are used to ratchet up prices on bad drivers (read them as bad customers), pushing them to other companies.

One solution would be to have laws like Europe has that require companies to provide to each customer all data it's collected about that customer on request Then you could "take your data with you" and show you're a save driver.

Interestingly enough, Geico was started by someone who successfully predicted that government employees would be less risky (the GE stands for Government Employees):

https://en.wikipedia.org/wiki/GEICO#History

In practice, it's the opposite. Switching may be a sign that you are a bad driver but it's also a sign that you are price sensitive which is usually the more important signal. Never switch and you are flagged as price insensitive and therefore a fine candidate for being over-charged.

> Generally i think companies leverage data to ratchet rates upwards, not offer unsolicited discounts. Why invest in infrastructure that will reduce profit?

Because it's a competitive market, and if they don't offer competitive rates, other insurers will take all their market share.

I'm really not seeing how this isn't obvious. If there's a clearly more efficient way of doing insurance rates, companies can either adopt it, or die at the hands of the ones who do.

>Why invest in infrastructure that will reduce profit?

Well ratcheting premiums upward is not the only way to increase profit. Same premiums with reduced number of claims at adds to the profit too.

The sophistication of these devices aren't that high - what constitutes an "unsafe driving habit"? Speed? How hard you turn? Most do some sort of banding around your speed and braking force, which is super crude and doesn't reflect actual driving habits.

They can pick up a lot of signals, and they have claims data: it seems like they should be able to build a pretty good model.

Even just "number of miles driven" is a strong signal.

Progressive also logged the time of day you drive. Between 12-3am (IIRC) was marked the highest risk time to drive.

Still probably less crude than your age and gender, no?

Those methods have a lot more refinement, but eventually, most certainly.

I suppose the more important argument is not about "what's the quality of the model?" vs. "what goal is the insurance company trying to accomplish with better risk pricing?"

Might actually be pretty interesting to see how the random statistics collected correlate with claims. It might indeed turn out fast, aggressive drivers never file claims, so ... that's what should be rewarded.

Sort of the inverse of the self driving car problem - they follow all of the rules, all the time, and are more accident prone, even though they're not at fault for the accident.

http://time.com/4098303/self-driving-cars-accident/

On many major roads in the US, refusing to exceed the speed limit is unsafe behavior.

With respect to speed, safety comes from matching your speed to that of the surrounding traffic, even if the surrounding traffic is on average exceeding the posted limit. Road-raging vigilante behaviors by drivers intent on enforcing their (incorrect) ideas about safety (i.e., deliberately driving slowly or occupying a passing lane to try to force passing traffic to slow) is actually far more dangerous and in many states is illegal and ticketable.

I'm not saying I don't believe you, but I want to see the law that states you can be given a ticket for driving 65 in a 65 just because traffic in the left lane is going 80.

Many states (including Kansas, where I used to live) have enforceable minimums as well as maximums on major highways. All states can pull over and ticket vehicles for, in the judgment of an officer, traveling at a speed slow enough to be unsafe (witness the stories a little while back about self-driving cars getting into trouble for obstructing traffic by being too slow). In recent years there's been a big push to begin enforcing this on highways, largely because of the rise of road-raging self-appointed "enforcers" who try to force other traffic to obey their personal interpretations of traffic laws.

And in all 50 states it is now some type of infraction to be anywhere but the rightmost lane when travelling slower than other traffic, even if by "slower" you mean "exactly the posted limit". This year-old article has some notes on the left-lane stuff, including the fact that yes, you could be ticketed for doing 65 in a 65 if the passing traffic is doing 80 and your slower driving is interfering with that:

http://www.vox.com/2014/6/16/5804590/why-you-shouldnt-drive-...

And also points out what was already well-known to everyone except vigilante traffic rangers: that doing only the speed limit, when everyone else is faster, is demonstrably unsafe.

Yeah, but those minimums are usually 40 or 45mph. You can absolutely be ticketed in many states for going at the speed limit in the left hand lane if you're blocking faster traffic although I expect that's fairly rare. I'd have to see someone driving at the speed limit in the right hand lane be ticketed for going slower than the flow of traffic to believe it. That would be a definite man bites dog event.

Utah law states that being tailgated in the left lane without traffic ahead of you is evidence of driving too slowly (on a freeway).

I should have been more clear, I meant that you're in the right lane and traffic in the left lane is going much faster.

Several states have laws stating that slower traffic must keep to the right, but it's not necessarily dependent upon speed limit.

I suppose it is conceivable that going 65 in the right lane when all the other right lane and merging traffic is expecting to go 70 or 75 could still be unsafe. If there's a big gap ahead of you and a long line of cars behind, it's probably a good idea to speed up regardless of lane (assuming weather's good and your car can handle it).

What's really frustrating and unsafe though is when drivers try to merge onto the freeway going 35 or 45 and traffic is going 70 in a 70 or 75 zone.

LMGTFY: http://www.cnbc.com/2015/06/11/pay-up-slowpokes-or-get-out-o...

Here is a specific law in Utah, for example: http://le.utah.gov/xcode/Title41/Chapter6A/41-6a-S704.html?v...

As I stated elsewhere (and prior to you posting this comment), I meant if you are in the right lane going 65.

In California:

CVC 21654:

(a) Notwithstanding the prima facie speed limits, any vehicle proceeding upon a highway at a speed less than the normal speed of traffic moving in the same direction at such time shall be driven in the right-hand lane for traffic or as close as practicable to the right-hand edge or curb, except when overtaking and passing another vehicle proceeding in the same direction or when preparing for a left turn at an intersection or into a private road or driveway.

(b) If a vehicle is being driven at a speed less than the normal speed of traffic moving in the same direction at such time, and is not being driven in the right-hand lane for traffic or as close as practicable to the right-hand edge or curb, it shall constitute prima facie evidence that the driver is operating the vehicle in violation of subdivision (a) of this section.

(AKA "Slower traffic, keep right" law)

Indiana passed a law like that less than a year ago

I like the idea, but in practice, we'd all likely see our rates changing inscrutably as the insurance companies change their models.

That's fine. What I object to is the surveillance.

The trouble with this logic is so: why not make both A and B pay extra?

It brings in more profit than making B alone pay extra. All that it requires is some half-assed justification, which can easily be fabricated.

One danger is that driver B will respond by driving without taking out insurance if it becomes too expensive. Which would leave people in a bad situation when he finally runs someone over.

Reply on news.ycombinator.com