I was talking to my friend who collates data for a large insurance company in the UK.
The idea is to group people in different ways and see which group is more/less likely to make a claim. The prices of the policies they offer are based on these findings, so for example, if you live in a rough area you will be charged more for a motor insurance policy (even if it’s just a 3rd party policy) because you are more likely to make a claim.
Here’s the rub though, they grouped people by star sign and found that there was a 20% difference in likelihood to claim between the best and worst group. Eager to discount astrology as bunkum I asked about the sample size, but this runs into 100,000’s, which seems quite large.
The insurance company does not use this info to price policies, but apparently does use more earthly reasons which show a smaller % difference.
Can anyone think of any reasons (other than our disposition being spelled out in the heavens) to account for this.
Who grouped people by star sign? Surely a large company didn’t waste time and resources doing this. Did the employees do it for fun on their own time? I suspect some sort of data error. That or your friend is pulling your leg.
One possibility, that you may be able to rule out, is that the sample size in the best and/or worst groups in particular are low, whilst there are a large number in the other groups. This would make some sense if the probabilities of being born at different times of the year varied.
In that case, you might get odd results in those cases in particular, hence the difference.
Also, it’s important to know how many claims are made in the period being considered (if the figures are taken over a year). If there are not very many, then you may get a cluster for one group in particular over the course of the year. How likely this is would depend on the number of cases.
Whether this applies depends if this is a yearly figure, or totals over a number of years.
Another possibility is that birthdate is somehow linked to other factors somehow. I’m not sure how this would work, but it might be worth checking. Of course, that wouldn’t neccesarily rule out astrology, as astrology could be what is causing the factors (if those are things that occur after birth, such as education), but it would provide an alternative explanation.
It depends on what the sample size means. Is that 100,000 claims or 100,000 insured drivers? You need several hundred claims for each star sign to make the statistical error less than 10%.
My second thought is that the star sign determines when (which month) you turn 16. If most people start driving immediately after turning 16, then people who turn 16 in winter or before a rainy season will face difficult road conditions with little experience.
The star sign also determines which month your license expires and needs to be renewed. But I have no idea how it can relate to accident stats.
It’s also possible that they are limiting the sample to people born in 1973, for example, and comparing star signs within that group. In that case, the star sign is correlated with age. Admittedly it’s hard to believe that 20-year old drivers are 20% more dangerous than 21-year old ones, but it’s a possibility.
It makes more sense the younger the age groups get, maybe theres a 20% difference between 16 and 17 year olds. Especially if the 17 year olds have mostly been driving for nearly a year.
Good idea, I’ll find out if they did just do the stats for 1 year.
This reminds me of the X-files where there was a real psychic selling insurance, and tried to persuade customer to buy by horrifically describing their deaths (“what if there’s a mangled body dripping blood out of a car, with your wife screaming …think of the children!”) and just scaring them off instead of selling to them