We do sometimes use actuarial data here, like the Social Security Administration’s tables, but I’m not sure how useful it is outside of very reliable data like mortality risks. All it is is a mass of correlations, and the most useful masses, lifestyle choices, may be commercially held.
That said, I could well be too pessimistic about the values.
Seeing as insurers have a commercial interest in their data being correct, the data they use should be of very high quality. Thus seeing what insights are really robust, like wearing a seatbelt is reducing the rate of death, should be useful.
I am an (almost qualified) actuary, working for a life insurance company.
I would love it if I had data of a very high quality. However, most insurance companies can’t use population statistics because of differences with underwriting standards (we don’t cover the very bad risks), target markets (we advertise in the Daily Slum, so only cover low socioeconomic classes, for example), and claim definitions (what is a disease in the population might not be a claim for the insurance company). So we use our own experience to modify the population stats. Very large companies might use entirely their own data.
Generally, there is not enough of it to be sure that it’s totally credible, especially when it comes to fine differences such as how much you smoke or drink. And that’s ignoring problems like non-disclosure. Age and Sex are easier, but there’s not much you can do about changing those, so it doesn’t help with the question at hand.
Of course, for some types of insurance, such as compulsory car insurance, there is more data to work with—I’ve never worked in general insurance, so I can’t comment on that.
People in business are under some pressure to be accurate, but not that much (they don’t actually have to be right, they just need to be more accurate than some of their competitors), and perhaps less so in businesses with high barriers to entry.
We do sometimes use actuarial data here, like the Social Security Administration’s tables, but I’m not sure how useful it is outside of very reliable data like mortality risks. All it is is a mass of correlations, and the most useful masses, lifestyle choices, may be commercially held.
That said, I could well be too pessimistic about the values.
Seeing as insurers have a commercial interest in their data being correct, the data they use should be of very high quality. Thus seeing what insights are really robust, like wearing a seatbelt is reducing the rate of death, should be useful.
I am an (almost qualified) actuary, working for a life insurance company.
I would love it if I had data of a very high quality. However, most insurance companies can’t use population statistics because of differences with underwriting standards (we don’t cover the very bad risks), target markets (we advertise in the Daily Slum, so only cover low socioeconomic classes, for example), and claim definitions (what is a disease in the population might not be a claim for the insurance company). So we use our own experience to modify the population stats. Very large companies might use entirely their own data.
Generally, there is not enough of it to be sure that it’s totally credible, especially when it comes to fine differences such as how much you smoke or drink. And that’s ignoring problems like non-disclosure. Age and Sex are easier, but there’s not much you can do about changing those, so it doesn’t help with the question at hand.
Of course, for some types of insurance, such as compulsory car insurance, there is more data to work with—I’ve never worked in general insurance, so I can’t comment on that.
People in business are under some pressure to be accurate, but not that much (they don’t actually have to be right, they just need to be more accurate than some of their competitors), and perhaps less so in businesses with high barriers to entry.