Time-space optimization of risk concentration for rainfall parametric insurance policies

Parametric insurance policies are different from classical ones in that they do not involve a damage certification from an expert. On the contrary, the damage is automatically evaluated based on a parameter, which could be for example:

– More than 2mm of rain from 21:00 to 24:00, which triggers a concert cancellation in the Arena of Verona

– Less than 10 knots of wind from 8:00 to 13:00, which triggers the cancellation of a sailing race on Garda Lake

– Temperature higher than 37°C for 4h, which triggers a 10% loss of honey production because of bees’ death

The examples above are related to weather, but this kind of policies can range from catastrophic events such as earthquakes or floods to travel delays, provided the travel infrastructure (airport, railway company) can certify the correct arrival time.

Parametric products are fully automated, both in the liquidation process (damage certification) and in the buying process.

It may seem that the best scenario for an insurance company is to sell the maximum number of policies, but a high-risk concentration can harm profitability due to potentially large losses. If a big festival is insured for rainfall on a certain weekend in Verona, it would be better to insure a second event in another distant location or on another weekend rather than to insure another event for the same rainfall risk, in the same location and on the same time.

The insurance company must then develop a procedure to stop selling specific covers if the risk is too concentrated in the same area and/or at the same time, as betting companies do when too many people bet on the same result.

The students are asked to devise a basic rainfall risk management system, allowing to decrease the risk concentration for the insurance company in place and time, while maximizing the revenues. Given an exogenous stream of new policies, characterized by their coverage place and time, the system must identify the optimal risk configuration to maximize the revenues, subject to input constraints. The topic mainly applies concepts of mathematical optimization and statistics, using Python as a programming language to implement the risk management system.

Instructor: Giovanni Poccobelli (REVO, IT)


  • King Hang Mok, University of Göttingen
  • Dora Mokri, University of Szeged
  • Ahsen Naeem, University of Koblenz
  • Boróka Jankus, University of Szeged
  • Gianluca Cappellari, Università degli Studi di Verona

Where: Room 1.15b, Veronetta, Istituto Ex-Orsoline, via Paradiso 6, 37129 Verona, G-map: https://goo.gl/maps/cZ6EExRJ5Cqq7W819