Skip to main content
European Insurance and Occupational Pensions Authority
News article2 October 20201 min read

Start of a European-wide comparative study on diversification in internal models

Diversification models

Today, the European Insurance and Occupational Pensions Authority (EIOPA) launched a European-wide comparative study on diversification in internal models. The objective of the study is threefold:

  • to gain an overview of the current approaches in the market and, on best effort basis, analyse and compare the levels of diversification
  • to facilitate a better understanding of modelling dependencies, aggregation and resulting diversification benefits and
  • to enhance quality and convergence of supervision on diversification in internal models.

Diversification effects depend on a variety of factors, such as the level of correlations, tail dependencies, number of risk factors, shape of underlying distributions and the dependency structures. These factors may operate at different levels which increases the complexity.

The study will therefore be carried out in two phases to balance complexity and completeness. The first phase of the study, starting early October 2020, focusses on top-level risk dependencies between market, credit, life, non-life, health, and operational risks. To complete the understanding of diversification effects, in combination to the respective risk profiles, the lower level inter risk dependencies will also be assessed in the second quarter of 2021 in the second phase of the study.

Undertakings using an internal model are expected to take part in this study, which comprises a quantitative and qualitative questionnaire. For undertakings applying the exact same correlation settings and aggregation structure as the standard formula, the first phase is limited to a subset of the qualitative questionnaire. Finally, the questionnaires are accompanied by detailed technical specifications, including examples, in order to adequately fill out the templates, given the bespoke internal models.

Read more about the study

Learn more about the study


Publication date
2 October 2020