fbpx

Services Family uses cookies to make this site simpler. Find out more about cookies 

Just another statistic? Causal Factors

How statistics can cost us money!

We spoke in the last two blogs about how it is easy to wrongly see two events in the progress of time as being seemingly connected.

COVID-19

The correct analysis of “A causes B” must show them having a “causal link”.

The Media is full of stories about young people being a factor in the spread of COVID-19 but has yet to prove a link.

Let us look at the issue in the round before getting to the nub of the problem for the politicians, journalists and yet the medics.

Party, party…

It is very important for society that our young people go out socialising, find friends, spouses, different points of view and cultures etc.

At the same time, increased testing capacity and university testing reveals a rise in positive cases in young people

= University LOCKDOWN!
 

But is it a sound decision? based on:

Positive tests go up, young people go out. – but, what’s the critical causal link?

Can we learn from insurance?

Insurance is very focussed on the cause of loss – these are called the causal links.

Just because B follows A in time it doesn’t necessarily mean that A caused B. So, it is essential to identify the cause of B.

For example, let us say your garden fence (B) has collapsed.

  • Case 1: The neighbour leaned on the fence (A) and it collapsed – insured or,
  • Case 2: The storm (A) blew the fence over – not insured.

Same outcome different causes.

This causal link analysis is extremely important in science and statistics.

Our students and COVID – a ‘fair’ response?

There could be many causal links to the rise of positive tests in students apart from just the volume of testing.

These could be;

  • the virus travels in beer (!),
  • alcohol weakens the body’s resistance to the virus,
  • young people do not socially distance very well wherever they are,
  • we are testing more young people from confined or multi-occupancy living space (University residences)
  • etc etc.

Add to this, the other important factor are the transmission mechanisms of the virus which are still not clear at all.

For some, the research now suggests (only suggests though because of insufficient data) that only 10% of sufferers may be infectious. There is also data suggesting that children appear not to be significant spreaders at all…

You see the problem? We do not have answers, the scientists do not yet have the answers and so we have to take a bet – literally.

So, here’s another observation and a clue:

Watch out for the word ‘could’ in any report!

When science talks of facts and presenters use the word ‘could’– “this could result in 500 deaths per day by x” ….

Normally, statistics are qualified by a measure of uncertainty/probability – 500 deaths with a 60% probability.

So, a given cause results in an outcome with a given probability.

“Could” means they don’t really know what the true risk is, but it would be real bad if……

This puts our decision makers into a difficult place. While the science has given them some facts, and some possible forecasts or outcomes, because they have to make a ‘political’ judgement.

Human nature being what it is decision makers will not want to be seen later to underestimate an issue, they need to play it safe.

In the worst case, they seek to reduce the risk or the probability to zero. This is known as the PRECAUTIONARY PRINCIPLE.

The ‘precautionary principle’ argues that public policy must avoid harm that may result from implementation. The harm need not be certain, but just scientifically plausible.

In simple English – something must be proven totally safe before use/doing it.

This approach still voiced by some politicians and activist groups, especially if they aren’t responsible for the decision. The Principle is often incompatible with societal norms like freedom of choice.

After all everything we do has some risk. You may remember the comedy catch line:

Life – a sexually transmitted terminal ‘disease’

So how does this all apply to insurance? Can you insure anything?

The answer is practically – YES!

According to LoveMoney:

  • Miley Cyrus insured her tongue for $1 million (£786k) when it became one of her ‘iconic’ features.

rolling stones in wales

  • The Rolling Stones’ guitarist Keith Richards took out insurance on his hands for $2 million (£1.6m).

You can read more from LoveMoney about celebrity insurance plays.

This can be done as insurance is A PROBABILITY PLAY by both the insurer and to an extent the customer (insured).

The industry prides itself on the ability to boil down facts, averages and trends in CAUSAL FACTORS into RISKS. These let insurers decide what (and who) they want to insure and how often the risks might happen (LIKELIHOOD) – the probability.

The premiums we pay are set by the probability of something happening (causal factor) and the financial impact (loss) of it happening. Plus business costs and profit margins of course.

So, what are the challenges for insurance in statistical probability?

An insurer looks at 1000 accident claims from their database. It sees that in 1 out of 10 accidents (likelihood) the driver and passenger are injured (risk) and these present the highest cost (loss).

I covered selection bias in my last blog. The insurance industry is very aware of selection bias and how poor statistics can cost us money.

In a recent FCA report, looking at unfair pricing methodologies, the FCA recognised that leading insurers have hundreds and hundreds of adjustment factors to try and ensure that the risks are accurately and fairly priced.

However, to keep the quote process quick, most insurers have used assumptions and use fewer, simpler questions and customer declarations to replace a real “drill down” into the individual customer’s circumstances.

The result is that the quote is so ‘averaged’ or statistically derived it doesn’t reflect your circumstances. Just a model of some key indicators which might not apply well to you and may not be applied at all if they actually understood “who you were”.

This approach works for the majority but, this is how statistics can cost us money – some customers are over-paying to compensate for others under-paying.

Our approach is to develop products to help specific customer groups.

Our first proprietary policy is for the UK Armed Forces and covers their unique life style! This is our tailored Military Home, Contents and Kit insurance.

Why? Well because a house behind security fences and with guards did not qualify for a 10% neighbourhood watch discount!

By better defining the risk the specific group faces, it allows better and more accurate pricing. In turn, this improves ‘fairness’ as more appropriate cover can be provided at the right premium.

If you enjoyed this blog on ‘how statistics can cost us money’, check out our last blog where we talked about the way statistics can be manipulated to tell a true story or support a misleading perspective.