Comment by wjnc | Hacker News Reader

Comment by wjnc | Hacker News Reader

Actuarial pricing is pretty simple in it's basis. Estimate claim frequency and claim severity (loss) per model, age group, state, [100s more categories]. Use proper statistical distribution. Apply markup for costs. Apply commercial markup or discount when comparing to competition. Repeat yearly.

Since claim size is highly skewed, don't trust the new entrant to have enough data.

Often 5+ yrs of data are used. So a new model / development like Tesla will give you headaches. Usually some leap of faith and pressure from product development will get the early premiums coming in too low, since everyone wants a piece of the pie. I've seen that played out in many markets.

There are lots of interesting things going on in the data, so a proper PHYD-scheme should be able to beat old school pricing rather quickly. But then again, adoption of that is usually in sub-markets that are really different (f.e. young drivers in a direct market are really not what you want dominating your portfolio).

With regards to Tesla's reply. AAA will be modelling on an individual brand basis. They are just using the data as a outside source comparison. The Volvo comparison is not really important. Perhaps in the group F1-cars for regular people they wouldn't stand out, but that doesn't change the equation for AAA. Even with low N, if all the simple claims come in 1000s $ higher, you don't need to be able estimate the entire distribution to expect a loss leading proposition.

Another possible consideration, if the average cost of repair is higher (assuming for the sake of the argument that all Tesla's drivers involved in accidents were not at fault) the insurance companies will need (ahem, cough, cough) to raise the costs of all other contracts to compensate.

This of course is not limited to Tesla's, wherever there are more "luxury" cars driven it is more likely that insurance claims will be higher and the insurance companies are likely to spread the delta on all contracts.

Doesn't insurance segment by model? Or are you saying that 10% of a $10k tesla repair is $1k in overhead, and that gets amortized across the customer base?

No, the argument is that when a Tesla is not at fault then some other insurance company is paying more money to fix a Tesla than say a similar BMW.

Exactly.

Imagine that there exists only one insurance company (additionally non-profit) and that only 100 cars - all in the same, single segment - exist, that have (invariably) 10 collisions with another car per year.

If the average cost for repairs is US$ 10,000, for the system to work the total amount of rates must be 10x10,000=100,000 and thus each car owner will pay 100,000/100= 1,000 US$.

Now introduce a new car that replaces a large part, let's say 50% of the (only) segment.

When these new cars are involved in an accident, (and their driver is not at fault) the repair costs double to 20,000 US$, so if on average you will have 5 accidents costing 10,000 as before and 5 costing 20,000, the total costs will raise to 150,000.

The insurance company will either leave the 50 "normal" cars at 1,000 and thus raise the rates of the 50 "more expensive" cars to 2,000 OR spread the difference by raising everyone's rates to 1,500.

Why don't companies price new models to the market relatively high instead of low? That seems like they're leaving a lot of money on the table. And I can't see why an actuary would care about pressure from product. Do you mean from the car manufacturer, or some internal insurance product team?

Even if it's internal product pressure, I don't see why any insurance company would do anything but give the actuaries a massive amount of say. That's the core competency of insurance companies.

Internal pressure probably. Not all insurance products are profitable. Especially big insurers are willing to take a hit to get in "good" customers, for example rich Tesla drivers. Then cross sell other products and raise the price of the original insurance yearly.

Why would a rich tesla driver be a better customer? What if they make a bunch of claims? Insurance is about having predictable data. In this case it looks like Tesla drivers are making too many claims so they need to make premiums higher.

I can answer that. I am not rich, and when the company that insured my car raised it rates a lot I went looking elsewhere.

The new company got my car insurance, and when my truck insurance was due they got that too. Later that year I added my house insurance and since I was getting discounts I also added my cabin up at Parry Sound.

So the result of having a cheaper car insurance is they sold me four policy not just one.

More likely to purchase other profitable lines. Sell them an inexpensive auto insurance policy, then make more money on Homeowners, Liability, Umbrella, Boat, Jewelery, Life, etc.

This is the answer. Homeowner's is the big profit maker for those companies which sell both. Presumably a Tesla owner has an expensive home. Auto is just for breaking even.

While not rich and not a tesla driver. I'd speculate richer folks tend to not micromanage prices changes(raises) in insurance premiums. Someone who is less rich might reconsider insurance purely on a financial decision. And someone a little more wealthy will never look twice at it.

Most people don't get in accidents and most people need auto insurance for many decades. So those lazy rich people will likely not change.

This is also assuming most people don't make many claims in insurance. (as is the base concept of how insurance works?)

Yeah, I don't usually cross shop renewals. I just renew whatever I have, unless the premium increase is considerable (+15-20%).

I might be leaving a lot of money on the table, and I'm not even that wealthy, but it's just not worth it to me.

And I'd add that the few times I've made relatively minor claims on both my home and auto insurance (handled by the same company), it's been pretty much completely painless. It's just not worth it for me to go to a bargain basement insurer to save maybe a few hundred bucks a year.

I can only guess - after you cross 25 years of age, I have never heard of insurance going down. I know people who are 50-60 years old, with zero claims ever, and they never get a discount, the annual renewal always come back higher. So what they do? They move every year. Every single year, they go through the effort of calling few insurers to get marginally better price to avoid the inevitable hike.

I can only guess that if you are rich then you don't care - it's more likely you will stay with the same insurer because it's not worth your time to be looking elsewhere.

Richer insurees have more assets to go after if they are involved in a the 2-sigma events (eg killed someone while drunk driving; insurance can come after a wealthy person's assets).

It wouldn't be shocking that tesla drivers are expected to make fewer claims, based on the halo bias towards wealthy people.

But if you are the insurance of the Tesla driver, it's the other party (through insurance or directly) that will come after your driver, not you. You will just say to your driver "per our contract, I cover up to $X, and the rest, you will have to deal on your own". You're not responsible for getting that money out of him, if I'm not mistaken.

Perhaps it's because you need data to arrive at the correct pricing and if you price too high, you won't get that data. It may make sense to lose a little bit initially during the data gathering stage to more quickly arrive at an accurate pricing model than to overestimate and allow your competition to gather all the data.

It's probably not that difficult to create a model that will adjust the prices to reclaim the loss later. Starting low likely generates more sales overall.

Reply on news.ycombinator.com