Claim amount

Hi in the insurance motor princing competition
For example for policy pl025631
I can t understand how come the claim amount can get bigger than the vh-value and how come after the accident the pol-no-claim-discount doesnt raise by 0.2.

Hi @raphael_sebbah

That’s an interesting point about the claim_amount Vs vh_value. This is just how it is for the vast majority of the claims I believe. I would be surprised if it’s try for all claims since you will find that in the training data there are some claims that are very very large.

Regarding the pol_no_claim_discount column, yes this is unfortunately a side effect of this being a real dataset. Please see this discussion for more.

Sorry i did not understand the answer
for 380 the claimamount is bigger thanthe value of the car
is it because for example he caused damage to another car or its a mistake?

Essentially yes. But please correct me if I’m wrong but you’re asking:

How can the claim_amount be larger than the vh_value if the worst possible claim just means you lose your car?

If that is a correct interpretation then the answer is that the claims in this dataset include third party liability claims which is exactly the damage caused to a third party. Could be a car, could be other things like a wall or a street or even a person.

In fact, most claims are of this type. You can see this explained under the pol_coverage column in the data dictionary found under the resources tab.

1 Like

Yes, the damages caused to other property or people is covered, that is what “third party liability” is.

The reason you need to buy insurance even if you own a 25 yo piece of shit car is third party. You could crash into a parked lamborhini and be liable for the replacement of THAT car. Depending on the country, You could also injure a pedestrian and need to pay for medical treatment for that person for years to come.


Thanks that was my question :grinning:
Another question since the claim amount is homoginized and there is no clear indication what is called a claim (with malus) and what is a small damage (window broken for ex with no malus) is it possible for you to give us or a number for ex 200 ,under which we are sure its not a claim, or to add a field which will indicate what is the kind of the loss cause please?

I would love to know this detail myself :thinking: however this was not provided to us by the data provider so I can’t say anything for certain.

Having said that, I think online you can find heuristics that might help you build that taxonomy yourself. Though I have no concrete sources to offer.