2025 Variance Prize Honors Research on Credibility Theory and Machine Learning
Christophe Dutang, Giorgio Alfredo Spedicato, and Quentin Guibert have won the 2025 Variance Prize for their Variance paper, "Adjusting Manual Rates to Own Experience: Comparing the Credibility Approach to Machine Learning."
As insurers increasingly explore how machine learning can strengthen actuarial decision-making, understanding how these techniques complement established actuarial methods is becoming increasingly important. The 2025 Variance Prize paper demonstrates how classical credibility theory and modern machine learning can work together to improve ratemaking.
“By comparing traditional credibility methods with modern machine learning and transfer learning techniques, the authors demonstrate how established actuarial principles can be enhanced by data-driven approaches to improve ratemaking accuracy,” said Variance Editor-in-Chief Peng Shi. “The work provides valuable practical insights while building on the strong theoretical foundations of actuarial science, illustrating how innovation can complement rather than replace actuarial expertise. The Casualty Actuarial Society established the Variance Prize to recognize outstanding contributions that bridge actuarial research and practice—this paper is a worthy addition to that tradition.”
The authors are a Franco-Italian team. Dutang is an assistant professor at Laboratory Jean Kuntzmann and teaches at Ensimag, both in Grenoble, France. Spedicato is an actuarial data science manager at Leithà, an internal consulting firm within the Unipol Group in Italy. Guibert is an assistant professor at the Université Paris-Dauphine.
Established in 2007, the Variance Prize recognizes outstanding papers published in Variance that make significant contributions to actuarial research and practice. Conferred by the Variance Editorial Board, the prize includes a $5,000 cash award and reflects the CAS's commitment to advancing property and casualty actuarial science through trusted research. Authors are encouraged to submit rigorous, practice-oriented research to Variance.
Read the 2025 Variance Prize-winning paper, "Adjusting Manual Rates to Own Experience: Comparing the Credibility Approach to Machine Learning," in Variance.
