Uncertainty Quantification and Global Sensitivity Analysis for Economic Models

Daniel Harenberg , Stefano Marelli (1, 2) , Bruno Sudret (2, 3) , Viktor Winschel (4)

1 ETH Zürich - Eidgenössische Technische Hochschule - Swiss Federal Institute of Technology [Zürich]
2 Chair of Risk, Safety and Uncertainty Quantification [ETH Zurich]
3 GdR MASCOT-NUM - Méthodes d'Analyse Stochastique des Codes et Traitements Numériques
4 Department of Economics

We present a global sensitivity analysis that quantifies the impact of parameter uncertainty on model outcomes. Specifically, we propose Sobol’ indices to establish an importance ranking of parameters and univariate effects to determine the direction of the impact. Sobol’ indices are based on variance decomposition and are able to fully capture non-linearities and identify interactions between parameters. Univariate effects are computed as conditional expectations. We employ the state-of-the-art approach for global sensitivity analysis by constructing a polynomial chaos expansion of the model from a limited number of evaluations. Sobol’ indices and univariate effects are then calculated analytically from the polynomial coefficients. We apply this analysis to a standard real-business-cycle model and compare it to traditional local sensitivity analysis approaches.

Domaines: Applications [stat.AP]; Méthodologie [stat.ME]

Daniel Harenberg, Stefano Marelli, Bruno Sudret, Viktor Winschel. Uncertainty Quantification and Global Sensitivity Analysis for Economic Models. Quantitative Economics, 2017, ⟨10.2139/ssrn.2903994⟩. ⟨hal-01902025⟩