Quantifying Reserve Uncertainty Using Stochastic Multivariate Generalized Linear Model: A Case Study on Egyptian General Insurance Market | ||||
| مجلة البحوث المالية والتجارية | ||||
| Volume 26, Issue 3 - Serial Number 1, July 2025, Page 322-342 PDF (1.33 MB) | ||||
| Document Type: المقالة الأصلية | ||||
| DOI: 10.21608/jsst.2025.388686.2028 | ||||
 View on SCiNiTO | ||||
| Authors | ||||
شانا يوسف عبدالله  ; محمود السيد; داليا خليل
| ||||
| قسم التأمين و العلوم الاكتوارية كلية التجارة جامعة القاهرة مصر | ||||
| Abstract | ||||
| Accurate claims reserving is crucial for insurance companies as it directly influences risk assessment, pricing strategies, and overall financial position. Traditional univariate reserving approaches, which treat each line of business independently, fail to account for the interdependencies among business lines, leading to inaccurate reserve estimates and misallocated capital. This study applies two statistical frameworks - univariate generalized linear models (GLMs), and multivariate generalized linear models (MGLMs) - to estimate reserves and associated uncertainties. Triangular data will be used of three major lines of business - motor comprehensive, fire, and medical - from an Egyptian general insurance company during the period from 2015 to 2024. The Tweedie distribution is employed within these models to capture the complex characteristics of insurance claims data. The multivariate framework allows for the explicit modeling of dependencies between lines of business. Results implies the superiority of the multivariate approach in providing more robust and stable reserve estimates, This research highlights the importance of implementing multivariate models in modern actuarial reserving, demonstrating their effectiveness in reducing predictive uncertainty and improving regulatory compliance especially under standards like IFRS 17. | ||||
| Keywords | ||||
| Reserve uncertainty; diversification effect; stochastic reserving; generalized linear models; multivariate generalized linear models | ||||
|
Statistics Article View: 34 PDF Download: 33 |
||||
