Modelling higher moments and tail risk in finance

Le, Trung (2019) Modelling higher moments and tail risk in finance. Doctoral thesis, University of East Anglia.

Preview

PDF Download (1MB) | Preview

Abstract

This thesis aims to address three critical issues in modelling higher moments
and tail risk of financial returns. First, I propose applying the Mixed Data
Sampling (MIDAS) framework to forecast Value at Risk (VaR) and Expected
Shortfall (ES) under a semiparametric approach. The new models exploit the
serial dependence in short-horizon returns to directly forecast the tail dynamics
at the desired horizon. I examine the predictive power of the new models by an
extensive comparison of out-of-sample VaR and ES forecasts with the established
models for a wide range of financial assets and backtests. The MIDAS-based models
significantly outperform traditional GARCH-based forecasts and alternative
conditional quantile specifications, especially at the multi-day forecast horizons.
My analysis advocates models featuring asymmetric conditional quantile and the
use of Asymmetric Laplace density to jointly estimate VaR and ES.

Second, I carry out a comprehensive comparison of the forecasting ability and
economic importance of several prominent skewness models. My empirical analysis
advocates the use of information from option prices to forecast skewness. Option-implied skewness and a realized skewness model, which also uses information from
options, outperform two GARCH models and skewness forecasts derived from
conditional quantiles. I further propose a new skewness estimator that corrects
the option-implied skewness for skewness risk premium. The new estimator has
the highest information content on future skewness while it consistently leads to
the lowest out-of-sample forecast errors. A portfolio strategy that employs this estimator is superior to the “1/N” portfolio and to the strategies based on the rest of the skewness models considered.

Third, I investigate the role of conditional higher moments, up to the fourth
level, in an international portfolio allocation framework. The conditional moments
of return distribution are simultaneously approximated by a set of quantile
estimates using the law of total probability. My results reveal significant economic
gains to an international investor by jointly incorporating conditional higher
moments in the information set. The portfolio that employs both conditional
skewness and kurtosis outperforms the benchmark portfolio based on mean-variance
predictors and portfolio based on information up to only the third
conditional moment.

Item Type:	Thesis (Doctoral)
Faculty \ School:	Faculty of Social Sciences > Norwich Business School
Depositing User:	Zoe White
Date Deposited:	18 Jun 2025 14:23
Last Modified:	18 Jun 2025 14:23
URI:	https://ueaeprints.uea.ac.uk/id/eprint/99633
DOI:

Downloads

Actions (login required)

View Item