Risk Measures Estimation Under Wasserstein Barycenter

TL;DR

This paper introduces a novel method for estimating multivariate risk measures using Wasserstein barycenters, demonstrating its effectiveness on US market indices during COVID-19, with promising results in both stable and volatile periods.

Contribution

It proposes a new approach for multivariate risk measure modeling using Wasserstein barycenters on location-scatter families, comparing it with copula-based VaR models.

Findings

Model performs well during COVID-19 market volatility.

Provides realistic VaR forecasts in volatile and stable periods.

Outperforms traditional copula-based models in certain scenarios.

Abstract

Randomness in financial markets requires modern and robust multivariate models of risk measures. This paper proposes a new approach for modeling multivariate risk measures under Wasserstein barycenters of probability measures supported on location-scatter families. Simple and advanced copulas multivariate Value at Risk models are compared with the derived technique. The performance of the model is also checked in market indices of United States generated by the financial crisis due to COVID-19. The introduced model behaves satisfactory in both common and volatile periods of asset prices, providing realistic VaR forecast in this era of social distancing.