Hacker News Reader

Comment by rcar | Hacker News Reader

This has a lot more to do with the risk pooling of insurance and competition among carriers than it does to do with some desire to mindlessly violate privacy.

To give a simplified example, say a good driver costs an average of $100 per year and a bad driver costs $10,000. If without trackers, you can get 100 customers where 10% end up ultimately being bad, the cost over the year ends up being $100 * 90 + $10,000 * 10 = $109,000 or $1,090 per customer. If trackers let you get down to just 8% bad drivers, your expected costs are $100 * 92 + $10,000 * 8 = $89,200 or $892 each.

That lowered cost means more competitive rates, which in turn drives positive selection (i.e., good drivers will be more likely to sign up for your service in the first place). Those bad drivers you passed on also aren't just going away - they're going to competitors and driving up their costs.

View on HN

I think it's clear that they want to use this to minimize expenditures. The problem arises with how they use the data. The trackers don't intrinsically let them get the bad driver rate down, but lets them tune the premiums that those drivers have to pay.

Let's use your data; they currently pay ~$1,090 per customer. So for the sake of argument, let's say that they charge $1,100 and keep that $10 in expectation. Now they install GPS trackers, and they can identify two classes of behavior; class A drivers (50% of drivers), who have a 4% accident rate, and class B drivers (50% of drivers), who have a 16% accident rate. The combined rate is still 10%.

Class A drivers cost 48 * 1e2 + 2 * 1e4 = $24,800 = $496/driver

Class B drivers cost 42 * 1e2 + 8 * 1e4 = $84,200 = $1,684/driver.

So this is great -- good drivers pay less, risk profiles are narrowed, etc. Maybe the bad drivers can be given feedback that will tell them what things they can do to minimize their accident rate based on the GPS data, and get some feedback on that -- "drive slower", "stop at red lights", "stop running into parked cars" -- things that both improve the metrics for the GPS and result in safer drivers.

But maybe the metric that the adjusters found are "frequently drives in historically Irish neighborhoods", or "drives to fast food restaurants more often". Now things get diciers. And what if the things that the GPS is measuring are not proxies for how bad a driver they are, but how willing they are to settle with cash without involving insurance. Or the correlation is a false one -- it could be that most of the Class B drivers are excellent drivers, as good as Class A, but they have to pay Class B rates because they got lumped in with a secondary association not related to quality of driving.

This is a problem with any attempt to bring more metrics to bear in computing insurance rates, but is especially exacerbated by the invasive nature of GPS-based metrics, which can be used as proxies for all sorts of socioeconomic signals that allow open discrimination.

The tracker I have doesn't have any kind of location capabilities, and only tracks certain kinds of events like acceleration rates and speed. So basically as long as you keep it under 85mph and don't slam on the brakes or the accelerator frequently, you get the full discount.

Do you mind saying what kind of device you have? In some cases the only thing preventing other information being used is privacy policies, and the data may be used for gathering metrics to plan future devices.

If the data collection is verbose enough, dead reckoning from speed/acceleration information can be used to unmask some non-driving information. It can't be generally unmasked completely, but even small correlations, like "they start slow in the morning but go faster after xx seconds" -> "they live near a freeway" -> <regional information> -> "they are slightly more likely to be rich" -> "they under-report accidents" can be inferred, leading down the same path. This is the sort of thing that I was indicating above -- a lot of "good" drivers can be lumped in with the "bad" ones to make Class B, if you can establish an attributes that appears to be causal but whose accuracy is low.

It sounds like the one that Progressive used. They would monitor the time of day you drove, your speed, and acceleration/braking. So if you drove fast, late at night, and braked hard constantly, you would not qualify for a discount but at least they didn't raise your rates. And the best (worst?) part was that you only had to have the thing in your car for 1-3 months, I forget which, but when you were done you sent it back and were free to go back to your original driving style.

We could raise gas taxes instead of letting insurance companies capture the profit from punishing inefficient drivers (speeding, not looking ahead and braking instead of easing off accelerator).

> Maybe the bad drivers can be given feedback that will tell them what things they can do to minimize their accident rate based on the GPS data

The issue here and my concerns with GPS trackers in general is if they are seen as providing valuable insight to how a person drives. This simply is not the case because so much of driving is context dependent. A few weeks ago, my brakes gave out briefly and I swerved sharply into the shoulder of the on-ramp I was on and finally came to a stop in the median between the on-ramp and the road I was trying to merge on to. Purely from GPS data that looks like erratic and dangerous driving when in reality I saved the insurance company money by managing to avoid hitting anything with my car.

Whoa, what caused your brakes to give out briefly?

Brake fade probably. In stop and go traffic in a hot climate brakes can get slower to respond. The brake pedal feels mushy is the best way to describe it. I've never experienced complete failure in South Florida traffic but I do know they I have to press harder each subsequent time.

AKA: When it’s time to flush your brake fluid.

Also: when you have an automatic and can’t depend on your engine for some braking power to take a load off the brakes.

If you had left more following distance you wouldn't have had to stop as quickly or as frequently. Either that or you need to maintain the car's braking system better. The point is, the brake fading is almost certainly a result of poor driving or maintenance habits, either of which should result in paying higher insurance premiums.

1. I'm already paying a higher premium just for driving in South Florida.

2. It's called bumper to bumper traffic for a reason. Following distance is <1 cars length. Anymore than that and some bozo is going to slide into your lane.

3. You know nothing of me nor my driving habits. I simply pointed out that brakes can fail or appear to fail for reasons other than mechanical failure.

> 1. I'm already paying a higher premium just for driving in South Florida.

Yes, but that's not because of your driving habits, is it?

> 2. It's called bumper to bumper traffic for a reason. Following distance is <1 cars length. Anymore than that and some bozo is going to slide into your lane.

Heaven forbid your ego take a little blow to make everybody safer.

> 3. You know nothing of me nor my driving habits. I simply pointed out that brakes can fail or appear to fail for reasons other than mechanical failure.

I know exactly what you've told me: You drive in a manner likely to cause your brakes to overheat and fade, a problem that many, many other people driving in the same conditions as you are able to avoid. Therefore, you should pay more than them for insurance, all else being equal.

I'm not entirely sure but the mechanic I trust said it was probably a bad master cylinder. That or just needed to change the brake fluid. It was definitely freaky because I like to think I do a decent job of keeping up on maintenance.

Especially when your GPS position jumps around in a dense urban environment! Let's hope the interns programming the ETL are actually looking and thinking and caring about the location data

When I had one on my car for a while (my wife signed us up for it with our new insurance, had to leave it on for a couple months) I had to struggle with driving safely and trying to keep that stupid thing happy (often by guesswork). If no-one modified their behavior with one on it wouldn't matter, but I'm sure everyone does so you're gonna risk ending up on the "naughty driver" list if you don't go out of your way to perform for the damn thing. Net effect, I was definitely at least a little less safe with it on. Hated it.

> so you're gonna risk ending up on the "naughty driver" list if you don't go out of your way to perform for the damn thing.

Worse than this, who say's they'll use the data honestly or competently?

Maybe they will draw false inferences from your innocuous driving data and still label you a 'bad' driver.

Then, if you try and argue with them, they'll make an appeal to the data, and attempt to shut down all debate.

This is a nice follow-on. The one additional thing I'd point out is that when a Class B driver gets the quote from this carrier of $1,684 but gets a quote of $1,100 from another carrier, they're likely to head for the cheaper option, providing further benefit to the carrier using the trackers.

Also, I haven't worked in property/casualty insurance, but my understanding of these trackers is that they don't include GPS capabilities, almost certainly for those sorts of reasons you mention. The risk of bad headlines incentivizes carriers to do what they can to ensure that the false-positive Class B drivers are not correlated with anything potentially sensitive.

> my understanding of these trackers is that they don't include GPS capabilities

A friend has one of these in her car and it literally has an LED labelled GPS. Which makes no sense because one of the arguments I heard they need GPS to get mileage and speed, which you can already get from OBD.

Not if you install bigger tires ;)

While insurance is a different industry, the fact that tech companies continue to make decisions that play badly in the media spotlight suggests that 'bad headlines' isn't a huge deterrent in today's world.

Poor people (should be) the least likely to care about headlines over cost.

Isn't this the purpose of insurance adjustments? When an event occurs that would cost the insurance company money, they change their prior of what you should be paying, no need to try and "pre-crime" with tracking chips.

I would rather we not normalize technology that really has no reason to exist but to allow insurance companies to optimize the last meaningless % of their margins at the cost of a slow erosion of an expectation of privacy. (companies who already have a history of doing questionable things to deny claims and avoid liability, enoughso that I have personal stories on this matter)

This is more like advertising your insurance company in women’s magazines to pre filter your customers without offering different prices based on gender.

As the insurance company couldn’t determine your driving ability themselves, and therefore offers the same price regardless of ability, the black box allows customers to pre filter themselves before applying, hopefully reducing the average accident rate.

As insurance becomes perfect, you’re just paying your actual damages plus a premium to the insurance company.

There’s decreasing value to pooling when the risks are perfectly quantified at the customer level.

Reply on news.ycombinator.com