Disclaimer: I do not work with any kind of telematics or smartphone tracking, nor does anybody in my company to my knowledge.
I would consider tracking and rating on risky driving more socially acceptable than something like credit score, which I find mildly objectionable as a clear proxy for socioeconomic status and to a lesser degree race.
Risky driving definitely has correlations with socioeconomic status and race. People who drive in inner cities will display more risky acceleration/braking as well as driving in a riskier geographic area. They’re more likely to drive at risky times like 2 AM. So, I don’t find this kind of data completely innocuous, but compared to something like credit score which (I know I’m oversimplifying) is kind of a “surcharge for being poor”, it really doesn’t strike me as socially unacceptable.
If rating on risky driving is unacceptable, I’d argue that rating based on accident history or number of late payments is barely if at all more socially acceptable.
I am not asking about the merits of risky driving as a rating variable, but rather how the data was collected. Does sneakily collected data differ in acceptance from one where the driver signed up for a telematics device?
I think “sneakily collected data” is a broader issue than just insurance ratemaking and underwriting.
However, for years we’ve been using databases that aren’t necessarily dreamt of by ordinary consumers, although admittedly most of those databases are populated by the insurers themselves. I’m pretty sure that the average person driving down the street hasn’t really thought about the fact that lapses in coverage, or having selected only minimum limits will probably be known to future insurers regardless of whether it’s disclosed on their next application.
Legislation and regulation can certainly prescribe boundaries on what we can or can’t use. However, in their absence…I’d be hard pressed to objectively define which kinds of databases can or can’t be used in such a way as to allow for future developments.
In the absence of finding agreement on where to draw the line, some actuaries/insurers are going to want to make use of that information. Competitive pressure will push other actuaries/insurers along.
Back when I was working with credit scores for personal lines (over 20 years ago), I hypothesized that I could build scoring models in such a way as to exhibit no worse bias than rating/underwriting elements that are commonly tolerated.
The problem was lack of sensitive data. At the time, we could defend ourselves from rabid lawyers by pointing out that we did not have the data to discriminate against protected classes. Acquiring such data for the kind of project I wanted to tackle would doom us in certain courtrooms, in certain cases, even if doing so was an effort to minimize inappropriate bias.
I’m glad that we’re talking more about how to recognize and reduce bias these days. However, I’m still a little unclear on how the concern about rabid lawyers is being addressed.
Many comments pointed out that it is difficult to tell if the smartphone owner is driving or a passenger. Perhaps sophisticated pattern recognition will be able to identify them by driving habits. Could be problematic if they get it wrong. For instance, I’m driving with my SO and say “Honey, can you use my phone to text so-and-so and let them know our ETA” and it is recorded as me using my phone while driving.
“Wrong, sir! Wrong! Under section 37B of the contract signed by him, it states quite clearly that all offers shall become null and void if - and you can read it for yourself in this photostatic copy - “I, the undersigned, shall forfeit all rights, privileges, and licenses herein and herein contained,” et cetera, et cetera…“Fax mentis incendium gloria cultum,” et cetera, et cetera…“Memo bis punitor delicatum!” It’s all there, black and white, clear as crystal! You stole fizzy lifting drinks. You bumped into the ceiling which now has to be washed and sterilized, so you get nothing! You lose! Good day sir!”
I have long advocated that, similar to mortgages, the government should mandate that insurers collect certain data such as income and race.
No insurer will ever collect especially race on their own, for reasons you say and because of public perception. Without this data, we will never have serious progress on social equity in rating. We have little half-measures like inefficient ways to prove a company isn’t redlining, but it’s very hamstrung.
True. Even without knowing who was driving, telematics will be able to tell that a particular car did a lot of hard braking, swerving etc., so that car will be more likely to have an accident. It’s also happens often (but not always) that when that particular owner gets a replacement car, the same driver/s will be driving that car.
I would put down money that more frequently being a passenger in erratically driven vehicles shows a positive correlation with likelihood of you being in an accident while driving. Less than you driving erratically, sure. It’s certainly useful data when you’re a passenger in your own vehicle.
With less certainty, I assume the same correlation for giving your phone to a passenger while you are driving.
I never thought about my smartphone tracker when I was passenger in my partner’s vehicle, I just assumed it to be part of data collection. If it was designed to only track my driving, I’d just get my insurance and have my partner drive us everywhere in my car that I wasn’t going individually. After I take ~5 extremely careful, short trips anyway.
I know this is an old post, but if the telematics is coming from my phone how does it know whether the car is mine, BF’s, my mother’s, my boss’s, my Great Aunt Sally’s…
I believe I was referring to telematics coming from the vehicle.
It still may be possible from the phone but would require frequent matching with the car’s position. Even then it might not work if you’re in a car directly behind your car (say if you’re doing a trip with two vehicles). Not sure what would happen if you’re parked next to an amusement park and ride a rollercoaster (lots of acceleration).
My insurer wanted me to put this app on my phone… sheesh, I hadn’t even considered what roller coasters would register as. Glad I didn’t do it. Mini Me has entered the roller coaster years.
My app correctly identifies when I’m on a train vs a passenger vs driving. Or did, I’m past the initial data collection phase and now just get the savings from not being a psycho driver.
I just recently renewed my policy, and installed their app on my phone. I have the option to reclassify a trip as a trip where I’m a passenger, which seems to invite adverse selection.
But what do I know, I don’t work in P&C, I work in health. And I know the reason why some health plans offer gym memberships is not because they think gym memberships make people more healthy, rather that people who buy plans that offer gym memberships tend to be more healthy in the first place.
I think a similar thing is going on here. People willing to voluntarily install an app that tracks their driving habits are more likely to be safer drivers.
State Farm uses (or recently used) a tracker that’s just supposed to sit in your car and track you, then communicates to your phone.
It’s entirely possible to take 5 very careful, short trips, then remove the tracker from the vehicle and have no further data collection at maximum discount.
It makes me wonder what is better during the collection period. Is a lack of trips during data collection period good or bad? It’s less data, but, on the other hand, not driving is safer than driving.
y insurance data know that even problematic/noisy data can have predictive power.