Pricing Auto Insurance Policies | In Practise

Pricing Auto Insurance Policies

Former Chief Underwriting Officer, P&C, Generali Germany

Learning outcomes

  • How insurance companies use generalized linear models to price auto insurance
  • How insurers adopt telematics and barriers to wider adoption
  • Which data points carry the most weight when pricing risk
  • How pricing for insurance policies will evolve over the next decade
  • Advantages that disruptors such as Root Insurance could have over legacy insurers
  • Why the accident frequency hasn’t reduced as vehicle technology improves
  • Potential impact of a 20% lower accident frequency on the insurance industry

Executive Bio

Monika Sebold-Bender

Former Chief Underwriting Officer, P&C, Generali Germany

Monika joined the insurance industry in 1994 and has been involved in underwriting motor insurance for over 25 years. She has experience working for many leading European insurance companies, leading AXA Direct in Germany in the late 90’s before running motor insurance at Westfalische Provinzial, one of the largest regional insurance companies in Germany. Monika was then Head of Underwriting P&C globally at Allianz SE before running Cosmos Direkt, a leading German auto insurer, where she was managing over 3bn EUR of gross written premiums. Monika has experienced C-Level positions across Europe and is on the Board of associations and other businesses such as MS Amlin, ERGO Group AG, and Europ Assistance. Read more

View Profile Page

Monika, can you provide a short introduction to your background in the insurance business?

I have been in the insurance industry since 1994. I have worked for several insurance companies, such as AXA, Allianz, Generali and ERGO, for big international corporates. But I’ve also worked for a regional company, called Provinzial, which is a public insurer in Germany. I am a trained non-life actuary and, therefore, I was always focused on non-life business. I covered several different functions and roles and I always reported to C-level or was myself at C-level. The first time I was promoted to a board, as an executive member, was 2005. Today, I am non-executive independent director for two boards.

My first love in the insurance industry was the motor business because it was the first one where we really faced challenges and where we had new methodologies applied and my knowledge, as an actuary, was highly appreciated. I could apply most reasoned methods and statistical methodologies.

Can we take a step back, into the 90s, and maybe you could share a bit of context to exactly how motor insurance was underwritten, back then?

I think that’s quite interesting, especially in Germany. Until 1994, motor insurance was strictly regulated. Every tariff had to be approved by the regulator and there was a basic structure for the pricing provided by the association for all motor insurers. You weren’t allowed to differ from that, except for your own expenses as an insurance company, if you could prove that your expected claim per risk was lower than the market. You really had to prove it over a long time period. The market was pretty uniform and the main distinguishing element was the cost structure of the different companies.

It was mainly about acquiring the customer and the cost of the claims and processing the claims, rather than the actual price?

Yes. There was a pretty huge difference between the cheapest and the most expensive because their cost level was very different. There was – and still is – a mutual end of the German market, called HUK-COBURG and it has a cost level, an expense level, of 7.5%. The most expensive ones have a cost level of almost 30%. This cost level was already in place in the 90s and explains the difference between the prices in that area. It was a very competitive market, given the circumstances, and competition exploded after 1994.

What happened after 1994, in terms of the data and the models that you were using?

The model didn’t really change much. There were just a few additional criteria, such as age of the driver. But the competition and the reduction of prices were enforced and the market lost a huge amount of money. It was the first time that the underwriting side had produced combined ratios about 120% for some insurers. It was a big loss-making period, until the end of the 90s.

Why was that?

It was because the competition was so harsh and tough and the insurers were not prepared for that. They just gave a flat discount. They gave the competency to trust the price to the sales organization which just reduced the price by 10%, 20%. On the other hand, there was a reduction in frequency but not as fast as the price was reducing.

What exactly were the models that you were using, back in the 90s, to price the policy?

There were mainly three criteria that were applied. Firstly, the type of car, with its horsepower and so on. Secondly, the region in which the car was driven. Thirdly, the bonus-malus system. These three criteria were provided by the association and are still provided today, by the association, on statistics generated by the whole market.

How did the industry change to the models used today? What were the big changes, over the last 10 to 20 years?

First of all, all companies started to create their own actuarial departments. When I started, in 1994, everybody asked why my company hired and recruited an actuary for motor insurance; nobody needs an actuary for motor insurance. It’s just one plus one, equals two. They started to hire actuaries to set up generalized linear models, to adapt the IT systems to be more flexible and reactive to the market. Today, almost every company in the market applies GLM models, partly based on their own data and also using the data from the association – which still provides a lot data – to improve and enrich their own data. They partly use the association’s model, adapting it to their own data to some specific criteria which are important for the company.

Today, I think all companies, in the German market, use generalized linear models and some even try to use artificial intelligence and more sophisticated models which are not based on linear interdependency.

Does the UK or US market, typically, use GLM models as well?

Yes. At the moment, GLM is the state of the art model which is applied in all markets that I know of. Also, in the Asian markets, in India, for example, it is applied. It strongly depends on how sophisticated and how detailed the regulatory framework is; what is allowed and what is not allowed. For example, in the US, you still have some states where you have to ask for approval from the regulator. In India, there are strict regulations about which price you are allowed to charge your customer. There are sometimes restrictions which constrain how you apply what you know by generalized linear models. But as far as I can see, almost all insurance companies use generalized linear models to analyze their own portfolios and to know which part of the portfolio is profit making, which customers they should try to acquire and so on.

For portfolio management, I think most insurers in the world – I don’t know Africa so well – use generalized linear models. It is state of the art methodology.

Could we walk through it? Let’s say I’m taking out a policy, you are the insurer and you are underwriting my policy. What would be the approach to building a GLM?

Sign up to read the full interview and hundreds more. No credit card details required.
Sign up to read


Pricing Auto Insurance Policies

December 1, 2020

Sign up to listen to the full interview and hundreds more. No credit card details required.
Sign up to listen