What is insured?


I gave a quick look at the “claim_amount” variable and I have a few questions around the “what is insured in Europe ?” theme.

Your car’s value ?
The value of the car you hit?
The life-long medical treatment of the pedestrian you hit?

The following two things that have me thinking

a) Claims over 10 000 euros (0.17%) are very rare, so I assume you don’t have to pay for people’s medical treatment. But the claims above 100 000 euro arent for Lamborghinis : the cars were worth 1113$ and 25000$. What were they claiming?

b) There are some very small claims (6% of all claims are under 100 euros). Why would you bother making such a small claim?

cheers and thanks for hosting this competition

1 Like

Hi @simon_coulombe

In short, we don’t know why a claim was made, we only have the claim amount.

The claim_amount column :dart:

The claim_amount column is how much you (the insurance company) has to pay out for a particular accident. We do not have a breakdown of why a claim was made and the only proxy we have for it is the amount of the claim. You are correct that in rare cases it could be medical (and very expensive), and in other cases it could be very minor damage, such a bump in the passenger door that has to be fixed.

What does the dataset cover :thinking:

Broadly speaking this dataset covers 4 types of policies, going from the lowest coverage that only covers Third Party Liability concerning motor insurance, to the highest which covers theft and so on. To quote from the data dictionary, the description of the pol_coverage column is:

There are four types of coverage: Min, Med1, Med2 and, Max, in this order. Min policies cover only Third Party Liability claims, whereas Max policies covers all claims, including Damage, Theft, Windshield Breaking, Assistance, etc. The two Med policies represent intermediate coverage.

For those unfamiliar with the terminology, Third Party Liability claims will concern only the cost that the policy holder has caused someone else (the third party). So if the policy holder causes an accident resulting in a claim of €1000, then the claim_amount will display 1000 for the policy holder in the dataset.

Your specific questions :question:

To answer your specific questions:

  1. Your cars’ value is how much your car is worth. This is the vehicle replacement value in Euros.
  2. The value of the car you hit is not included in the dataset.
  3. Some policies may include very large claims that could be medical but this is very rare and we cannot be certain.

Please let me know if these do not answer your question and have fun!


Thanks Ali for the detailed breakdown.
This is pretty much how it works in Canada. I wasnt sure what it would look like in Europe, with free healthcare and all :slight_smile:



Thanks for answering questions! I have one related follow-up – can “claim_amount” encompass multiple claims in one accident year? E.g. for a given policy if, in year 3, it had two claims, one in February for €500 and one in October for €700, would claim_amount just say 1200?


Hi @alan_feder

That’s correct! We don’t have enough granularity to know how many claims result in the total annual claim_amount and if there are multiple accidents, indeed the value is the sum of all claims for that year.

1 Like