An analysis of sensitivity of copula model parameters in green project risks assessment

• Date: 2026
• Main Authors: Квашук, І. О., Кузнєцова, Н. В.
• Format: Article
• Language: Ukrainian
• Published: Інститут проблем реєстрації інформації НАН України 2026
• Subjects: copulas; mathematical models; parameter sensitivity; risks; green projects; confidence intervals; maximum likelihood method
• Online Access: https://drsp.ipri.kiev.ua/article/view/358481
• Journal Title: Data Recording, Storage & Processing

Given the increasing emphasis on the application of copulas in modeling green risks, a rigorous understanding of parameter estimation procedures has become critically important. This issue is particularly relevant for relatively small datasets, which often contain no more than several thousand observations. Since copulas explicitly model tail dependence and extreme events, they may be unstable to minor perturbations or slight data drift. Among the available estimation techniques, two of the most widely applied approaches are the Maximum Likelihood Method (MLE) and the Method of Moments (MoM). In finite-sample (non-asymptotic) settings, the behavior of these estimators may differ, necessitating a systematic comparative analysis. To assess the sensitivity of parameter estimates to small variations in input data, a bootstrap technique was employed. The resulting distributions of parameter estimates were analyzed both graphically and statistically, using measures such as the mean, variance, and confidence intervals. This methodological framework provides a foundation for further robustness and stability analysis in copula-based risk mode-ling. The empirical results indicate that both estimation methods yield comparable outcomes, with no statistically significant differences between them. Furthermore, neither method consistently produced uniformly narrower confidence intervals across parameters. These findings suggest that, in small-sample applications related to green risk modeling, the choice between Maximum Likelihood and Method of Moments should be guided by practical considerations rather than expectations of systematic superiority in estimator stability. Tabl.: 2. Fig.: 3. Refs: 21 titles.