A few questions on car insurance (for a couple)

SouthLondonUser

BeenThroughItAll wrote: »

You're the one making assumptions.

Having been involved in writing the quotation engines for some of the UK's first direct and later online insurers, and having worked with insurance datasets on a number of projects, I can assure you that the insurers make decisions based only on statistics. How each insurer weights their decisions will vary, but all of them are statistically based.

So are these "relationships": http://www.tylervigen.com/spurious-correlations

Were you the statistician who actually coded the pricing algorithms? Do you know for a fact that insurers had enough data to price reliably and to watch out for spurious correlations? Surely you have heard that correlation doesn't necessarily imply causation?

I remember an interview on the FT with the head of underwriting of either a challenger bank or a peer to peer lender . He said that people who filled in their application forms all in capital letters tended to show a different risk profile to those who didn’t. Causation? Chance? Should it be used in the underwriting?

Of course insurers run a business, not a charity, and they must make money. But I cannot think of many other industries where prices vary so wildly for no apparently explicable reason. Bank X may give me a mortgage while bank Y might not because they have different criteria; Bank X may change its terms and charge a different rate, or reject me altogether, if I apply 4 months from now. But no bank is going to quote me a 1.5% interest rate now, a 1.8% interest rate this evening, and a 1.65% rate tomorrow. How would you feel if the rate a bank charges varied so much? Would you think: “oh, well, the global economic situation and my own risk profile has certainly changed wildly over the last 6 hours, so these differences are justified” or would you wonder what substances the underwriters are on? Banks will charge different rates and fees, but there aren’t many cases where one bank can quote 1% and another 5%, whereas a 5x difference in prices quoted by insurers is not unheard of and certainly way more frequent.

Also, not all decisions are made on statistics. I tried to get quotes for a MV Agusta Turismo Veloce – it’s a motorbike made by a niche Italian brand. Insuring it turns out to be 2 to 3 times more expensive than insuring German or Japanese bikes which are more expensive and more powerful. MV has a very limited production, and the UK is not a core market; there are very, very few MVs in this country, so I simply do not buy the argument that statistics shows the MVs to be riskier than more powerful and more expensive bikes. It seems way more plausible that insurers simply try to rip off customers who are attracted by the appeal of a niche bike. Which is fine – it’s a free market, after all, but please don’t tell me it’s a decision driven by “statistics”.

It’s not a coincidence, after all, that insurers all over the world have been constantly fined for cartels, price collusion, etc.
Have you ever read “Weapons of math destruction?” It’s a book by an American mathematician, who explains, among other things, how the black box algorithms used by American insurers penalise the poorer parts of the population, e.g. by relying heavily on credit scores, and by penalising poorer drivers more than richer drivers for the same offences. But no, she’s wrong, because it’s all backed up by statistics, right?

SouthLondonUser

takman wrote: »

Insurance companies are under alot of pressure to be competitive especially since more and more people are comparing prices online. If insurance companies were overcharging for the risks involved then one company would be able to come in and undercut all their prices and get a massive increase in business.

This is the theory. This is what would happen in the perfect world of perfect markets and perfect competition assumed by microeconomics textbooks.
The real world is more complex. In the real world, companies can and do collude to fix prices. In the real world, companies love non-transparent pricing because it makes it harder for customers to shop around and compare.
Also, in the real world, insurers have less of an incentive to keep the repair costs down because they can make more money using that as an excuse to hike prices even more. The Spectator had an interesting story on this a couple of years ago: https://www.spectator.co.uk/2015/10/the-car-insurance-industry-is-a-disgusting-racket/

SouthLondonUser

Car_54 wrote: »

I think you have some of that diametrically wrong. It is (AFAIK) illegal to discriminate against a foreigner if he is an EU citizen. There is nothing illegal - or indeed wrong - about discrimination on the basis of choice of shirt or holiday destination.

Sure. It is common knowledge that this is what lenders are always doing, right? And, even if they aren't doing it now but started doing it tomorrow, it would be perfectly within their rights and no regulator would read them the riot act...

Car_54

SouthLondonUser wrote: »

Also, not all decisions are made on statistics. I tried to get quotes for a MV Agusta Turismo Veloce – it’s a motorbike made by a niche Italian brand. Insuring it turns out to be 2 to 3 times more expensive than insuring German or Japanese bikes which are more expensive and more powerful. MV has a very limited production, and the UK is not a core market; there are very, very few MVs in this country, so I simply do not buy the argument that statistics shows the MVs to be riskier than more powerful and more expensive bikes. It seems way more plausible that insurers simply try to rip off customers who are attracted by the appeal of a niche bike. Which is fine – it’s a free market, after all, but please don’t tell me it’s a decision driven by “statistics”.

It’s not a coincidence, after all, that insurers all over the world have been constantly fined for cartels, price collusion, etc.

I'd suggest two possible factors for your motorbike - (1) price and availablity of spares, or (2) more likely, just fear of the unknown.

As for cartels, have you any examples from the UK?

SouthLondonUser

Car_54 wrote: »

I'd suggest two possible factors for your motorbike - (1) price and availablity of spares, or (2) more likely, just fear of the unknown.

As for price, like I said, insuring more powerful and more expensive German or Japanese bikes was way cheaper.

As for parts, I'm not sure it would be particularly relevant for the non-comprehensive quotes I was getting. If I damage the bike, I wouldn't be covered. If others damage it, it would be the other party's insurer paying. Maybe if the bike is stolen, then found but in a bad shape? Not sure, tbh.

As for 2, yes, most likely, so that's a subjective call, not a scientific, statistically driven decision.

Car_54 wrote: »

As for cartels, have you any examples from the UK?

AFAIK UK motor insurers have never been formally fined for price-fixing like those of other countries.
The competition commission did, however, find that the practice of getting kickbacks from car hire operators, claims management firms etc is uncompetitive and inflates the cost of insurance policies. However, nothing was done about it! :mad:Ah, the power of lobbying...
There were a couple of stories on the Torygraph, one from 2014:
http://www.telegraph.co.uk/finance/personalfinance/insurance/motorinsurance/10681261/Car-insurance-Profiteering-insurers-must-cut-premiums-says-regulator.html
and another from a weeks ago: http://www.telegraph.co.uk/news/2017/07/30/telegraph-investigation-drivers-overcharged-motor-insurance/
I'm not sure if I can quote from the last article, since it's behind a paywall, but you get the concept; for example, insurers handling claims which are not their fault regularly overcharge rival firms.This is all in line with the other article from the Spectator I quoted.

I realise this may come as a shock to all the people on the forum who seem convinced that it's all driven by statistics and that if one firm overcharged it would be driven out of business, but, surprise surprise, this is what's been happening in the real world.

SouthLondonUser

BeenThroughItAll wrote: »

Having been involved in writing the quotation engines for some of the UK's first direct and later online insurers, and having worked with insurance datasets on a number of projects, I can assure you that the insurers make decisions based only on statistics.

PS To be clear, did you work on something like the IT infrastructure of a quotation engine, or on the actuarial data analysis and pricing?

takman

SouthLondonUser wrote: »

AFAIK UK motor insurers have never been formally fined for price-fixing like those of other countries.
The competition commission did, however, find that the practice of getting kickbacks from car hire operators, claims management firms etc is uncompetitive and inflates the cost of insurance policies. However, nothing was done about it! :mad:Ah, the power of lobbying...
There were a couple of stories on the Torygraph, one from 2014:
http://www.telegraph.co.uk/finance/personalfinance/insurance/motorinsurance/10681261/Car-insurance-Profiteering-insurers-must-cut-premiums-says-regulator.html
and another from a weeks ago: http://www.telegraph.co.uk/news/2017/07/30/telegraph-investigation-drivers-overcharged-motor-insurance/
I'm not sure if I can quote from the last article, since it's behind a paywall, but you get the concept; for example, insurers handling claims which are not their fault regularly overcharge rival firms.This is all in line with the other article from the Spectator I quoted.

I realise this may come as a shock to all the people on the forum who seem convinced that it's all driven by statistics and that if one firm overcharged it would be driven out of business, but, surprise surprise, this is what's been happening in the real world.

You seem to be mixing up two different issues. The statistical analysis is what the insurance company use to determine the likelyhood of someone being in a crash and what type of crash they are likely to be in. There is no mention in any article you have posted or anything I have seen that says they havnt been basing premiums on statistics.

The other issue is insurance companies being extremely inefficient in the way claims are handled incurring costs from multiple different companies during a claim. I don't think anyone on here would say insurance companies handled claims in the most financially efficient way they would and everyone knows about claim handling companies.

So the fact that repairs processes are expensive will mean that people likely to be involved in an accident will pay a higher premium which is based on the cost of claims in the past. So the premium is calculated statically but there are opportunities to reduce the costs of claims so everyone will pay less.

But the article quotes that if this was fixed it would only save everyone £8 a year on insurance. Which isn't much and seems abit low.

BeenThroughItAll

SouthLondonUser wrote: »

PS To be clear, did you work on something like the IT infrastructure of a quotation engine, or on the actuarial data analysis and pricing?

I worked on what these days would be called 'data science'.


One does not have to be a statistician to understand that the risk profiling was undertaken based on sound statistical principles.


To answer your earlier points, yes, of course I'm aware correlation <> causation. Insurance statistics don't work like that.

SouthLondonUser

takman wrote: »

You seem to be mixing up two different issues. The statistical analysis is what the insurance company use to determine the likelyhood of someone being in a crash and what type of crash they are likely to be in. There is no mention in any article you have posted or anything I have seen that says they havnt been basing premiums on statistics.

The other issue is insurance companies being extremely inefficient.

There is a difference between being inefficient, and explicitly and intentionally overcharging rival firms for repairs, in the full knowledge , and with the explicit intent, of driving up premiums!

Where is the £8 potential saving you quoted? Sorry, I must have missed that

SouthLondonUser

BeenThroughItAll wrote: »

I worked on what these days would be called 'data science'.
One does not have to be a statistician to understand that the risk profiling was undertaken based on sound statistical principles.

May I humbly ask how you have reached this conclusion?
The mathematician in the book I mentioned (weapons of math destruction) has very explicit examples of how American motor insurers do NOT use sound statistical principles. Yes, I am aware UK <> USA – the point is that statistical analyses can be, and are, done in non-robust, non-sound ways.
If you torture the data long enough, nature will confess…
Just saying that a statistical analysis was conducted does not guarantee it was sound. As a banal example, a statistical algorithm to select job candidates might easily conclude that the best candidates are white upper-class males who went to posh public (as in, expensive non-State) schools and then to Oxbridge. Guess what? The data shows that most business leaders fit this description. It’s sound statistical analysis! Or maybe not…

I'd also be interested in understanding what on earth causes insurance quotes to vary so wildly from one day to the next. I cannot honestly think of many other sectors where prices vary so wildly. Like I said, what would you think if a mortgage lender offered you a mortgage at 1% today, at 1.5% tonight and at 1.3% tomorrow?