Bias in the Pricing of Italian Motor Insurance

The researchers looked at around 20,000 motor quotes from nine leading insurers obtained through Italy’s leading price comparison website. One of their research questions was to establish if gender and birthplace directly influenced quoted premiums.

About half of Italian motor insurance GWP goes through aggregators. The cars were the two best sellers in Italy. Birthplace was Milan, Rome, Naples, Romania, Ghana and Laos, with residence being one of those three Italian cities.

A variety of techniques were used to harvest the 20,000 motor quotes, with a number of control features used. Two analyses were carried out – one for the cheapest quote and another comparing the top five quotes.

Three aspects of the regulatory context for Italian motor insurance are relevant here...

• the direct use of gender is prohibited;
• there is soft regulation to encourage insurers to avoid using nationality related factors such as birthplace and citizenship
• there is a dual duty to contract – consumers are obliged to buy cover, and insurers are obliged to offer cover to all drivers, regardless of their risk profile.

On gender, the research produced an intriguing result. They found no systematic gender bias. However, they did find sizable gender related price differences. This meant that some quotes for both men and women were clearly influenced by gender, but overall, the median for all quotes was zero. In other words, no gender was systematically at a disadvantage across the 20,000 quotes.

On birthplace, the results were clear. When the birthplace was overseas, quotes were higher. When the birthplace was Milan, the quotes were lower, as was the case with Rome over Naples. The researchers concluded that “birthplace related differences…exhibit patterns of systematic discrimination.”

Claims?

Now before any of you do yourself an injury by jumping up and down shouting “claims, claims, claims”, it’s worth noting that the researchers incorporated data from Italy’s bonus malus system into the 20,000 quotes used. That would of course reduce the influence of claims frequency by quite a lot, but not necessarily claims severity. Kilometres driven was also taken into account.

Do people born overseas, or in southern Italy have, on average, larger claims? Further research would establish the extent to which this is the case, but I think the outcome is likely to be middling, with some factors raising average claims and some lowering them. As the former head of insurance for Europe’s biggest vehicle fleet, it’s not an unconsidered opinion.

What we are left with then is research that points to bias in pricing systems in a large European insurance market. The ethnicity penalty report here in the UK found something similar. What such reports do is raise questions.

Firstly, have insurers been tracking these issues too? If yes, what has resulted from this and why is this not being shared? If no, then why not?

Secondly, have regulators been tracking these issues? The same yes/no questions apply.

Time to Think and Act Differently

Insurers should realise that their sector has become the case study par excellence for researchers in bias in digital decision systems. And as the body of evidence grows that there are problems with bias in pricing systems, this puts the ball pretty firmly in the sector’s court to respond. The narrative that ‘we do not discriminate’ and ‘we are not like that’ is failing to have traction. Insurers need to think and act differently.

(Note - to access the full report, go to the lead author's university webpage and look under 'publications'.)