CAViaR and Cross-sectional quantile regression models to assess risk in S&P500 sectors

Abstract

The aim of this thesis is to investigate the performance of different models used in risk management to identify and control risks that may negatively impact company operations due to unpredictable events. More specifically, the object of this paper is the discussion of a cross-sectional quantile regression model (CSQR) and the CAViaR model, which is a time series quantile regression model. Additionally, two highly used models were added: the Historical VaR and the Normal VaR. The out-of-sample analysis performed from 2000 to 2021 to the 11 equally sector sectors within the S&P500 index, suggests that the cross-sectional models outperform the time-series models. The outperformance is evident even during periods of stressed market conditions such as the Financial Crisis and the Covid Pandemic. Similar results were found for the value-weighted sectors. The study concludes that the additional incorporation of data from other firms in the same sector allows the risk manager to a more efficient assessment of the market risk and faster adaptation to market shocks.