The everyday blog of Richard Bartle.

10:07am on Tuesday, 4th January, 2005:

Sex and Insurance


The long-established practice of giving women less expensive car insurance in the UK received a boost yesterday when it was announced that the upcoming EU anti-discrimination laws on gender would not be applied in this instance. The Deputy Minister for Women and Equality (isn't that an oxymoron?), Jacqui Smith, is quoted as having said:

The rights we've enjoyed in the UK for nearly 30 years will now be enshrined in this directive for the benefit of all Europeans.

Rights? Enjoyed? We?

Let's suppose the figures showed that members of some ethnic minority were more likely to be involved in road accidents per unit distance driven. Could the insurance industry then charge members of this ethnic minority more for their car insurance? Well no, it couldn't: I recall many years ago when the law prohibiting racial discrimination in insurance was brought in (to great applause), and now financial companies are not allowed to give people services on less favourable terms on the basis of race. Why are they allowed to do it on the basis of gender?

Insurance is a tricky field. Almost any differentiator will pick up a statistical imbalance between groups of individuals. It may be that people who wear hats while driving have a higher incidence of accidents than people who don't. Should they be charged more for their insurance? Probably not, because it's the kind of person who wears a hat while driving who is the problem, not their actual wearing of a hat (which may actually reduce the chance of their having a prang from what it would be otherwise — it certainly acts as a warning to those in cars behind them on single carriageway roads that they can look forward to travelling no faster than 40mph for the next hour).

Some cases of differentiation are necessary, though. Life insurance, for example, would disappear as a concept if everyone were charged the same amount irrespective of their age. Now although people can't help how old they are, their age does change and it does happen to everyone. If we decide that discrimination on the basis of age is allowed, well, OK, at least we know it's going to happen to us some day.

The grey area is DNA-related insurance. Can insurance companies charge different rates depending on medical history, for example? If a woman's mother, aunts, grandmother and great-grandmother all died of breast cancer, isn't her risk of contracting it higher? Shouldn't she be paying higher premiums? The short answer: yes, and she is doing.

The thing is, you can't change your DNA (yet). When decisions are made on the basis of your DNA, there had better be a causal link between the DNA and what the decision concerns or it's wrongful discrimination. Statistically, women live for longer than men, and it looks like there is a genetic basis for this (although social factors also play a part, eg. more young men than young women get killed in wars). It may be harsh, but it's a fact of life. Between races, there is less variation in DNA than between sexes, but nevertheless there do seem to be genuine differences which could conceivably affect medical insurance premium levels. On the whole, though, the insurance industry seems to take the view that these even themselves out rather than add up (or at least that they're sufficiently small that the benefits in charging for them are outweighed by the costs of appearing to be racist).

Another thing you can't change is your social background. Poor people are sick more often than well-off people (or rather, they're sick with things that are likely to kill or disable them more often than are well-off people). Should people be discriminated against on this basis in insurance terms? Again, probably not unless it's directly related to what's being insured. If a geographic area is subject to subsidence, more poor people are likely to live there than rich people because the housing will be cheaper; however, the reason it's cheaper is because of the subsidence, therefore the poor people who live there can expect to have to pay a higher buildings insurance premium precisely because they are poor.

We can thus derive an acceptance criterion for the validity of insurance discriminators: it's OK to differentiate on the basis of DNA or social background if and only if there is a direct causal link between the DNA/background and that which is being insured (and even then it still might not be OK).

So, women drivers: there's a definite DNA component and probable social component. Are either of these causal?

Women are not better drivers for DNA reasons, at least not provably so. It might be argued that testosterone plays a part, and I'd go along with that; however, I'd note that testosterone doesn't cause all (or even the vast majority of) men to drive like lunatics.

There may be social reasons for men to drive more recklessly than women (peer pressure, macho culture, having a passenger who's nagging them the whole time, whatever). Are any of these causal, though? Is there a direct link between the social background of male drivers and road accidents? No, there isn't. Being a part of male culture doesn't make you drive too fast (although it may mean you drive more often on the kind of journeys that involve crashes). Therefore, men should not, in general, be charged more for motor insurance than are women.

Undeniably, a disproportionate number of road accidents involve men, particularly young men. Should middle-aged women be forced to pay more for their car insurance to subsidise these people? Of course not! But then, neither should I. Neither should I have when I was a young man, either, because I've never been involved in a car accident. Just because most accidents are caused by young men, that doesn't mean most young men cause accidents.

If you want to charge higher insurance premiums (hmm, premia?) to people who are likely to be involved in accidents, find the causal links. Look at the cars they drive, where they drive them, for how long they drive them, their personal claims history - anything that is causal (or indicative in a Bayesian way). Don't take the lazy way out and differentiate on the basis of an easy-to-make distinction between individuals that merely passes the "unfair subsidy" charge from one group to a sub-group.

It's unfair.

Latest entries.

Archived entries.

About this blog.

Copyright © 2005 Richard Bartle (richard@mud.co.uk).