The increasing of data available in finance has substantially grows over the years, this large amount of data bring new challenges in how to model and forecasting financial time series. This session is devoted to discuss some advances in high-dimensional time series data and its application in finance as well as encourage other researchers to dive, propose and develop new procedures to deal with high-dimensional financial time series. The session will be divided into five presentations of 20 minutes each one (including Q&A). This IPS will be of great interest to all ISI WSC 2023 attendees interested in time series, business statistics, and computational statistics. Members of the ISBIS, IASC, and the ISI SIG on Data Science will be particularly interested in the talks and discussion of the IPS. [This session if co-organized by Carlos Trucíos and Paulo Canas Rodrigues.]

Title 1: Robust Estimation of High-Dimensional cDCC Models

Authors: Eduardo Gabriel Pinheiro (Murabei Data Science) and  Luiz K. Hotta - Speaker  (University of Campinas).

Abstract: Volatility plays an important role in many economic and financial applications and the number of assets has increased considerably in recent years. The cDCC model has been one of the most commonly used models and composite likelihood has been used to estimate  high-dimensional cases. The paper shows that this estimator can be strongly affected by additive outliers, one of the most frequent types of outliers. A robust method is proposed and it is shown that its performance is better than that of the traditional nonrobust estimator both by simulation study and by backtesting  on real-life stock return data.

Title 2: Modelling and Forecasting Covariance Matrices: A Simple Model with Stochastic Volatility Latent Factors.

Authors: Giorgio Calzolari (University of Florence) and Roxana Halbleib - Speaker (University of Freiburg).

Abstract: This paper proposes a simple approach for forecasting large dimensional covariance matrices with the help of a latent factor structure with stochastic volatility components applied to realized covariance matrices. The factor structure together with the conditional Wishart distribution automatically assures positive-definiteness, symmetry and thick-tails, captures the commonality in the dynamics and the long-persistence of the autocorrelation of realized (co)variances within a unified parsimonious framework with no parameter constraints. The Factor Autoregressive model we propose profits from what alternatives suffer, namely the curse of dimensionality: higher the matrix dimension, higher the efficiency of the estimates and forecasts. The model has a non-Gaussian non-linear state-space representation that we estimate by maximum likelihood together with a non-Gaussian filtering technique. Monte Carlo simulations provide evidence for the accuracy of the estimates and the comprehensive empirical application to DJIA components proves the usefulness of the model to accurately forecast large dimensional covariance matrices.

Title 3: Modeling Multivariate Positive-Valued Time Series with Financial Applications.

Authors: Nalini Ravishanker - Speaker (University of Connecticut), Chiranjit Dutta (University of Connecticut) and Sumanta Basu (Cornell University).

Abstract: We describe and compare two approaches for Bayesian analysis of vector positive-valued time series, with application to analyzing financial data streams. The first approach consists of a flexible level correlated model (LCM) framework for building hierarchical models. The LCM framework allows us to combine marginal gamma distributions for the positive-valued component responses, while accounting for association among the components at a latent level. We use integrated nested Laplace approximation (INLA) for fast approximate Bayesian modeling via the R-INLA package, building custom functions to handle a latent vector AR model. The second approach builds a logarithmic vector multiplicative error model (log-vMEM) assuming multivariate Gamma errors, and imposing hierarchical lag structures to guarantee a  sparse model. This model will enable us to study dynamic interactions among different intraday volatility measures for different financial assets.

Title 4: Tail risk measures for ESG and nonESG companies.

Authors: Alessandra Amendola - Speaker (University of Salermo) and Vincenzo Candila (University of Salermo).

Abstract: The attention of the literature on the environmental, social, and governance (ESG) themes has enormously increased over the last few years. Nowadays, the consciousness toward social responsibility is grown up and has motivated the surge of new research lines, where the main differences among ESG and nonESG companies, concerning their economic and/or statistical indicators, became a relevant issue.
The main aim of the current proposal is to investigate the performances of ESG and nonESG indexes in terms of the resulting tail risk measures such as the Value-at-Risk (VaR) and Expected Shortfall (ES). Moreover, the impact of relevant predictors (macro-economic variables, implied volatility, etc.) in the effcient estimate of the sustainability indexes volatility and tail risk measures are investigated. Further issues related to vast dimensionality, mixed sampled frequency, asymmetric effects and co-movement are also addressed in this framework.

Title 5: Forecasting Value-at-Risk and Expected Shortfall in Large Panels: A Robust General Dynamic Factor Approach.

Authors: Carlos Trucíos - Speaker (University of Campinas) and Marc Hallin (Université Libre de Bruxelles)

Abstract: Beyond their importance from the regulatory policy point of view, risk measures such as Value-at-Risk (VaR) and Expected Shortfall (ES) play an important role in risk management, portfolio allocation, capital level requirements, trading systems, and hedging strategies. However, due to the curse of dimensionality and the presence of extreme observations, their estimation and forecast in large portfolios is a difficult task. To overcome these problems, we propose a new procecedure based on filtered historical simulation, the general dyamic factor model and robust volatility models. The new procedure is applied in US stocks and the backtesting and scoring results indicate that both VaR and ES estimated using our proposal outperform several existing alternatives.

Organiser: Dr Carlos Trucios 

Chair: Dr Paulo Canas Rodrigues  

Speaker: Prof. Roxana Halbleib 

Speaker: Prof. Alessandra Amendola  

Speaker: Luiz Hotta 

Speaker: Prof. Nalini Ravishanker 

Speaker:  Dr Carlos Trucios