I don't see nearly enough complexity in this analysis to justify the claim of having found any insurance fraud.

Firstly, there's no account for correlation between the features identified. The article mentions VINs which have several single-vehicle accidents, for example, but someone who has one single-vehicle accident is probably more likely to have another. Switching coverage is another of those potentially-correlated features; if you claim and it bumps your premium, aren't you likely to shop around as a result?

Secondly, there's no attempt to account for the law of large numbers. It's incredibly unlikely that someone has three single vehicle accidents in a year, but because the probability of that is nonzero, we know that with enough vehicles on the road then someone is going to do it.

The article covers itself by acknowledging this, of course, but if you title your blog post "We Found Insurance Fraud in Our Crash Data" then you should actually do that.

Totally agree.

The conclusion is a little too strong for my tastes. Changing coverage looks sketchy, but I’d bet people shop around after a crash and the ensuing rate hike.

Being involved in multiple accidents could suggest fraud but it could also just be bad driving, maybe someone very young or old.

It’d really help to know the base rates for many of these things: how often do people switch insurance, for example?

I'm kind of irritated that the article doesn't say anything about probabilities.

Given they have enough data, at some point it's perfectly reasonable to have cars with 5+ crashes per 12 months - just because of chance.

This is exactly why statistics was invented, damnit!

I think “in the exact same way with all the known indicators of insurance fraud” adds a couple orders of magnitude to the expected by-chance odds.

Well they do say they calculate propensity scores, but the whole language of the blog post is very hand-wavy. Warning over and over again that something is not causal proof, and then dramaticizing it as if it were proof is just unprofessional.

If you have a fraud model, just show the model and the data and the validation - everything else is marketing fluff.

It's an interesting article, but the title ('we found fraud!') doesn't quite match the article, which continually points out that there are suspicious patterns in the data but that they aren't ironclad evidence of fraud.

That said ... yeah those particular VINs do look seriously dodgy to me! If nothing else, then the drivers reporting 5 crashes in a 12 month window probably shouldn't be driving.

> Some years ago, a police officer casually told me, “You should get a dashcam, insurance fraud is common around here.” His offhand comment stuck with me, but life moved on.

This is very, very good advice. Dashcams are cheap, and easy to install. Treat yourself - spending $80 on a camera could save you thousands in lawsuits, insurance hikes, costs of a new car, etc.

To me it seems nearly as important as having a smoke detector in your house.

I don't remember them being easy to install; you need mounts and in the case of a rear dashcam, someone to wire them into your rear seat. Car dealerships should advertise adding webcams while you're in there -- that would really make it easy.

I got a dashcam that ran off USB, which I knew my car had a port for on the center console. There's ways to run the wire under the trim for the most part, but it's a long run.. and at the end there's a part that remains visible to some degree.

There are cams that can do a rear view as well from inside the cab, which likely provides enough evidence if you're rear ended.

I only opted for a forward facing dash cam.

In my state, you are 100% at fault for rear ending someone unless you can prove your innocence -- which a dashcam can do assuming the person in front does something shady (like lane change + intentionally slam breaks).

However, do note that dash cams are not going to magically make rear ending the person in front of you somehow that person's fault. Virtually no one seems to leave enough follow distance by default because doing so means someone merged into the space.. and a dash cam doesn't shift the blame for simple rear endings unless it can prove some kind of malice or inattentiveness on the other driver (but even then, inattentiveness of the other driver is not necessarily a legal defense for you not leaving enough room to react.. perhaps if they stopped faster than a car could be expected to break, e.g. hit a concrete wall....) -- of course, laws vary by state

In my car taking the headliner off involves removing 6 bolts (one in each pillar) and then popping out some retaining clips. I was surprised at how easy it was when I ran the wire to the rear camera. Headliners were notoriously difficult to deal with in the old days. There was even space to tuck the excess wire away to keep it neat and tidy looking. My camera just sticks to the glass using a sticker so attaching it was no problem either.

Cars should just give you access to the cameras that are already there. Why get a rear facing camera when your car is already legally required to have one?

I've wondered this too. Even if it's only the rear camera all cars should have a few seconds of rolling buffer that they save whenever an accident is detected. At the very least if the airbags blow it should trigger the save. You only need enough storage for maybe 10-15 seconds of footage.

I would also like the ability to turn on the rear camera when moving forward, for those times when the rear view mirror is blocked, especially on vehicles where the backup camera is integrated into the rear view mirror.

You tuck the wire along the headliner with an interior tool. It's about 10 minutes of work.

They are a big fat pain. I just added a front-facing one, done the right way (into the fusebox).

Didn't bother with the rear-facing one.

Driving defensively and slowly is much more skin to having a smoke detector as it actually aims at preventing harm rather than just figuring out who’s at fault.

The typical fraud "accident" is a vehicle with 4-5 people (including a child) who slam on their breaks or in heavy traffic on the freeway. If that doesn't work, they reverse into you, and then claim whiplash for all the people in the car with a doctor who has everything to gain. It's $50k minimum.

If someone backs into you, or slams on their breaks on the freeway without reason you can drive as defensively as you want and still get unlucky. People can cause "accidents" much more effectively than you can avoid one. Careful motorcyclists know that watching the behavior of all the people around them is critical to survival, but they still get into collisions.

If you do try to drive so "defensively" that you can never get into an accident another person tries to cause, then you end up with 7-10 car lengths of separation, and people will regularly cut you off increasing risks. Please don't be that guy stopped 70ft back from the traffic light and stopping on on-ramps.

Also fun to see dashcam video where someone jumps out in front of the car for the insurance claim, but the driver is too alert and stops before hitting them, only for the person to then throw themselves on the hood of the car while their friend pretends to freak out on the sidewalk and call the cops.

Getting falsely accused of hitting someone, being sued, and seeing a huge insurance rate hike doesn't count as harm?

I suppose the broken metaphor would be "I don't deep fry turkeys in my living room and I have smoke detectors."

But what you're saying ignores the fact that you're not the only turkey on the road.

This story doesn't make sense: it's not clear who hit who, whether you were scamming them by not giving insurance, or how a dash cam would help when no real damage was done.

The older couple hit him, with their motionless, parking-braked vehicle. Possibly by using dark matter/energy to cause space-time expansion that pushed his car into theirs.

I don't really understand what you're trying to say here.

Did you try to avoid giving them your insurance details? Why?

What would the dash cam have shown?

I would find it weird too if someone would stall.

And you always need to be able to stop your car independently of the other do. You know minimum braking distance?

Why would it be unlikely that the person committing fraud is under the influence? There's no reason to assume that.

Because they know that a cop is about to show up, is my guess.

I read that as assuming the fraudsters were professionals in a sense, i.e. doing it in an organized way. Wouldn't those kinds of criminals try to keep it looking clean?

It’s very unfortunate that the worst people in society ruin insurance for everyone else - so that when you need to use it you are investigated as a criminal. Too many systems are being broken by a tiny minority of people with no skin in the game and no social consciousness.

Almost everything wrong in society can be traced back to some small percentage of people ruining it for everyone else.

The article says

>After filtering out invalid VINs, we narrowed the dataset to roughly ~15 million crashes. (We also removed all drug and alcohol related crashes, since it’s unlikely someone committing insurance fraud would be under the influence.)

Which I immediately found extremely implausible. Seems like the author is very much NOT familiar with the type of people who commit insurance fraud.

In the next paragraph the article links to the FBI website on the fraud ring in Connecticut.

https://www.fbi.gov/news/stories/staged-accident-ring

Which says

>After an autumn evening of drinking and using drugs in 2013, a group of friends got into an Audi A6 and drove to the remote Wilderness Road in Norwich, Connecticut. The car slid off the road, hitting a tree.

>Everyone in the car survived, but this seemingly typical crash was no accident.

>Despite their impairment, the driver and passengers had purposely planned the crash to collect the insurance money.

So yeah, don't exclude drug/alcohol related crashes. Also the author should read the first paragraph of the stuff they link.

>Which I found very much not plausible.

Yeah. I'm not a fraud expert, but "impaired judgement" is a common effect of alcohol, and "desperation" is a common effect of drug addiction, so it seems weird to assume that people using drugs are committing fraud at a lower rate than the overall population rate.

As a side note, I grew up in Norwich so it's funny to see it mentioned in that report as "remote" because there isn't really anything remote in Norwich. Wilderness Rd, despite its name, does a ring around Mohegan Park, which is an urban-ish park with some trees and a rose garden. You can walk to the rose garden from the high school, which I sometimes did. Anyway.