Value-at-risk Quantile Forecast using GARCH-EVT and HAR-EVT Model

Citation

Aridi, Nor Azliana and Tan, Siow Hooi and Chin, Wen Cheong (2022) Value-at-risk Quantile Forecast using GARCH-EVT and HAR-EVT Model. In: Postgraduate Social Science Colloquium 2022, 1 - 2 June 2022, Online.

[img] Text
aug 2022.pdf
Restricted to Repository staff only

Download (1MB)

Abstract

Background - Extreme value theory (EVT) is one possible approach to identify and manage the extreme risks by using the estimation of high quartile and tail probabilities. Previous research on forecasting the value-at-risk (VaR) and expected shortfall (ES) based on EVT are widely analyzed using the standard GARCH class models which consider the daily returns’ data. Since the turn of the millennium, high frequency data has become more accessible, stimulating an increase of financial economics activities. A new approach has been introduced to use the realized variance (RV) measures to estimate the latent conditional variance precisely by using high frequency intraday returns’ data. Purpose - The EVT defines extreme losses mainly based on two methods which are Block Maxima (BM) and Peak over Threshold (POT). BM approach follows a Generalized Extreme Value (GEV) distribution whilst the POT follows a Generalized Pareto Distribution (GPD). Both EVT approaches has their own advantages and disadvantages. However, BM approach is not particularly suited for financial time series because of volatility clustering existed in stock price series. This research will use the POT method to estimate the VAR and ES. Design/methodology/approach - The sample of data collection was obtained from the Bloomberg and Olsen data provider which involved daily returns and intraday 5-minutes returns of Brazil Stock Market (BOVESPA) from 2014 – 2021. The two-stage conditional EVT approach proposed by McNeil and Frey (2000) was used to forecast the VaR and ES. The first stage involves the filtering procedure of the original daily return series with the standard GARCH, asymmetric EGARCH, long memory FIEGARCH model and the original realized variance with the standard HAR models. The second stage of the data analysis is to apply the EVT approach on these standardized residuals to model the tail distribution and forecast the VaR and ES. Findings/Expected Contributions - The result showed that risk models based on the realized variance provides better in-sample performance compared to the standard GARCH models which used daily returns. Research limitations - Since the COVID-19 pandemic created an unpredictable level of extreme volatility in the financial markets, an intensive extreme values analysis need to be done to anticipate risk exposed to investors in stock market during uncertain market events. The limitation is to identify how is the implications of Covid 19 to investors in stock market with a limited sample data within 2 years (Dec2019 – Dec2021). The data need to be updated and analyzed quickly. Originality/value - The study on forecasting extreme events using combination of GARCH-HAR-EVT model is limited. This study aims to add new finding into the literature review of stock market volatility forecast.

Item Type: Conference or Workshop Item (Paper)
Uncontrolled Keywords: Value-at-risk, Expected shortfall, GARCH model, HAR model, Extreme value theory
Subjects: H Social Sciences > HB Economic theory. Demography > HB3711-3840 Business cycles. Economic fluctuations
Divisions: Faculty of Management (FOM)
Depositing User: Ms Rosnani Abd Wahab
Date Deposited: 15 Aug 2022 03:24
Last Modified: 15 Aug 2022 03:24
URII: http://shdl.mmu.edu.my/id/eprint/10392

Downloads

Downloads per month over past year

View ItemEdit (login required)