Sunday, 2 December 2012

Sex discrimination

So, it is in the news that insurance companies (e.g. car insurance, life insurance, etc) are no longer allowed to price based on gender. i.e. sex discrimination laws now apply to insurance.

Once again I find myself feeling that the world has gone mad. And once again I find myself defending an industry I normally dislike.

Women campaigned for equality, quite rightly. They suffered from arbitrary and prejudiced discrimination. It is clear that such discrimination was wrong.

But ultimately, true equality is not sensible. People are not equal. It is a simple fact of life. Some people are smarter than others. Some are stronger. Some are healthier. And, in some cases, there are statistical differences that are specifically gender related. We recognise this in sports - where there is a clear difference and it is the case that a mixed gender sport would leave only one gender winning, we split the events. We even do this for people with disabilities.

The way insurance companies work is that they use statistics. Information on previous claims and on characteristics of the insured person which allow them to determine risk and hence price of insurance. They use a lot of factors, including age, health, history of claims or accidents, type of car, all sorts. They are keen to make their risk assessment as accurate as possible. They have found, just like sporting events, that gender is a factor, in both life insurance and motor insurance.

Now, this discrimination is not an unfounded prejudiced bias, but a statistical bias. It is one of the few areas where the bias is scientifically based and a reflection of reality. Forcing them to ignore gender has the impact of increasing prices for women, which I am sure is not the intention of the activists that brought about sexual equality in the first place.

But this then leads to interesting issues. What happens as more and more insurers work using in-car monitoring equipment and so price your policy based on your driving style. This is totally un-biased, but if, as is likely to be the case, they find that statistically people priced in this way get lower prices if they are women than men, will they be forced to somehow bias the pricing to make it balance? You would think insurers could say "we don't care what sex you are, just how well you drive", but that is, in effect, what they did already, and that has been banned.

There are, in some areas, anti age discrimination rules. Imagine if life insurance was forced not to consider age? That would make it impossible to get life assurance as they would have to work on the assumption that you are 105 already. After all, if not, everyone would wait until they are old to get life assurance.

There are times when I agree with Prof Farnsworth on this...

8 comments:

  1. Well, the way I've heard it explained is: discrimination is defined as the act of judging someone based on their membership of a group -- and it's discrimination even if that judgement is based on good statistics.

    The classic example (and it is a bit Godwin's-law-ey) is: black people, statistically, commit more crimes. Therefore, if you're an employer, it makes sense to avoid employing anyone who's black.

    So we all instinctively recoil from that. But it doesn't seem to be quite the same in the insurance case, does it? I wonder what shapes our different reactions here. One possibility is that we instinctively stick up for underdogs, and that outweighs the principle of discrimination (think, positive discrimination). Another is that we might see a causal link between not giving people jobs and their going on to commit crimes, whereas it's less likely that expensive car insurance would make you a worse driver...

    ReplyDelete
  2. "Now, this discrimination is not an unfounded prejudiced bias, but a statistical bias."

    But it's likely not a fair statistical bias. Undoubtedly it's true that women have less accidents than men. But that's likely NOT _because_ they are women, it's because statistically they tend to use their cars differently and are less likely to be a "boy racer" (of any age!).

    It's the same flawed logic as this -

    Imagine the "dangerous drivers club of great Britain" set up in your town and proceeded to have lots of accidents. Now statistically it would then be _entirely accurate_ to say it's a proven fact that people who live in that town tend to have more accidents and so should be charged more. And yet that's clearly hardly fair on those who are not members of that club :)

    I think the same thing applies here. Statistically woman have less accidents. But that doesn't mean that it's remotely fair to say that because you are a man you will have more accidents so should be charged more because being a member of that group isn't the _cause_ of the accidents.

    Charging depending on age and experience is more fair because it's much more certainly a fact that someone of 17 is going to have an accident than someone of 40 on an individual basis...

    ReplyDelete
    Replies
    1. Interesting argument, and probably valid to some extent in some cases, but the insurance price is based on chance of claim and loss, statistically. Yes, there are cases where your postcode changes the price of your insurance, which does indeed seem crazy are arbitrary, but the statistics are there, and pricing based on them is just as valid as anything else if they turn out to be a consistent indicator of future risk. If gender is a statistical predictor of risk then surely it is valid for determining price.

      Even your contrived case of dangerous drivers club. It will be an indication of future risk as being in that town I have a higher chance of being or becoming a member of that club than if I am not, even if my personal case is not like that.

      What matters if if the statistics are a valid indicator of future risk, or not.

      Whether car insurance is different for men and women because of the underlying differences or statistical anomaly is one thing, but I expect in life insurance it is much that just statistics, just as in sport. For life insurance I suspect that gender is validly an important factor just as you concede age is.

      Otherwise, even age, for car insurance, is unfair as there are always exceptions.

      Delete
    2. John Burton's logic seems self defeating and can be applied to itself.

      "But it's likely not a fair statistical bias. Undoubtedly it's true that women[older drivers/people with more experience] have less accidents than men[young drivers/people with less experience]. But that's likely NOT _because_ they are women[older drivers/people with more experience], it's because statistically they tend to use their cars differently and are less likely to be a "boy racer" [attentive]."


      "I think the same thing applies here. Statistically woman have less accidents. But that doesn't mean that it's remotely fair to say that because you are a man you will have more accidents so should be charged more because being a member of that group isn't the _cause_ of the accidents.

      Charging depending on age and experience is more fair because it's much more certainly a fact that someone of 17 is going to have an accident than someone of 40 on an individual basis..."

      Why is it remotely fair if I'm a 17 year old who drives extremely carefully has to pay more than someone of 40 (who due to experience may actually pay less attention to the road). Because I a member of a group (17 year olds) I shouldn't be charged more because being a member of that group isn't the _cause_ of the accidents.

      It is about statistically measurable _risk_. Unless it was seen that insurance companies just decided with no statistics that males/young people/short people/people with brown eyes were to be charged more _purely_ because they were male/young/short/brown eyed - then there is no discrimination.

      Delete
  3. Actually, pondering a little more.

    As people have said, being classed the same as others in a group because you are a member of that group is indeed a key feature of the problem with "bad" discrimination which has led to all of this in the first place.

    It is plainly unfair as an individual may indeed be very different to a stereotypical view of a particular group and treating them as such is not right.

    This leads to all sorts of interesting points.

    1. The issue is usually that one is treated as being compliant with a bias, stereotypical, and usually very incorrect view of that group. Treating all people of a certain race as unstrustworthy, taht sort of thing. If the viewpoint is statistically correct, then that is not quite the same problem, but still an issue for people that do not comply with the statistical mean for the group.

    2. Insurance is not about picking on people, it is about predicting risk. Treating people in a certain group as being higher or lower risk is sensible, commercially, if that is a valid predictor of risk. i.e. if the past characteristics of the group are consistent in to the future. If, for whatever reason at all, women have less accidents (or cost less money on them) than men, and continue to do so, then gender is a factor that can predict risk, and so should be, like any other factor to predict risk, used. It does not matter what the cause of the correlation is, as long as it is a good prediction of future risk.

    3. In order to be less unfair, you have to have better ways to predict risk. You need more narrowly defined categories so that you can carefully tune prices to very precise risk. That way careful drivers do indeed get lower insurance. But that only works if you have more categories not less. This move removes one of the categories used in assessing statistics and so risk, i.e. gender. Removing differentiators makes this less fair. It means treating all people in the same group and not in smaller and more accurate groups. Tarring us all with the same brush, which is the very problem we had in the first place.

    Ultimately, it is not clear what people want. The only way to be truly fair with insurance is to have a time machine and so price insurance based on actual future risk. This would mean people that were going to have an accident could not afford insurance, and those that will not can have insurance almost free. Whilst technically totally fair, people would, I am sure, moan about this a lot, and rather that things are averaged out more, and that there is more "treating all people like me the same". In short, you can't win.

    ReplyDelete
    Replies
    1. I do find insurance odd, and dislike it mostly. This funny "I'm going to bet something bad will happen, so if it does happen I have money, and if it doesn't I'm alright ,other than being drained every month of $sumofmoney

      The time machine argument is actually kind of possible in the future with the medical industry and DNA sequencing, and is a worry in this sector.

      If you go for medical insurance, instead of weighing the risks against the general population and other generalised factors such as age,race and gender, an insurance company could say "give us a DNS sample and we'll tell you your premium". So they discover you have a 90% chance of this disease/cancer at age 45+ (or not susceptible to anything) and give you your premium accordingly. Suddenly, as RevK says you would have one class of people that can't afford insurance unless they are rich (in which case you probably don't need it because you're rich) and another who get nice,cheap insurance (because they almost certainly don't need it). Insurance becomes pointless! (yay!) :)

      So yes, you want some differentiation in order for some groups of people not to pay a "tax" on the others (older drivers not paying for the inexperience of 17 year olds, the general driving styles of females not paying for the different driving styles of males). But once you're sequencing DNA or sticking a GPS in the car you are moving away from a "bet" (insurance is a negative bet after all, I bet something bad will happen) and into true discrimination (you have the wrong genes for our medical policy, your driving skill is too low for us to insure you etc)

      Delete
  4. We shall fight for his *right* to have babies!

    ReplyDelete
  5. FWIW

    1: In absolute numbers, women have more claims than men
    2: Those claims tend to be for minor damage (parking dings), whereas male drivers tend to have major damage claims.
    3: For the last 30 years, women under 25 have had equal or higher cost damage claims to men under 25.

    The single largest group of claims (more than almost everything else put together) is for reversing accidents in supermarket carparks. It says more about carpark designs than I care to mention.

    ReplyDelete