Use this URL to cite or link to this record in EThOS:
Title: Risk measurement with high-frequency data : value-at-risk and scaling law methods
Author: Qi, Jun
ISNI:       0000 0004 2722 0900
Awarding Body: University of Essex
Current Institution: University of Essex
Date of Award: 2011
Availability of Full Text:
Access from EThOS:
This thesis investigated reliable measures of market risk using high frequency data. The first part of the study, comprising of chapters 2,and 3, investigated the issue of risk measurement on a single time scale method basis. In Chapter 2, we explored the market risk measurement based on high-frequency measures of volatility with selected stocks in three different sectors. Parametric as well as non-parametric procedures are discussed. Furthermore, the backtesting results for comparing the risk forecasting performance of different risk measurements are also provided. Chapter 3 proceeds into the analysis of liquidity risk in high-frequency trading. We proposed an extended VaR measurement by incorporating the liquidity risk in intraday trading strategies when analysing limit order book data. The focus is on the integration of asymmetric information, upward or downward risks, into the input factors of forecast variables. We further find that there exists an asymmetric behavior of bid spread and ask spread between different trading volume. By taking account of the actual liquidity risk faced by investors with different trading size and positions, we proposed a liquidity adjusted intraday VaR (LAIVaR). We apply the bivariate analysis to investigate the asymmetric effect of the bid and ask side. The empirical results of liquidity adjustment show that the liquidity risk is a crucial factor in estimating VaR. The second part of the study, the analysis in Chapter 4, is extended to a multiple time scale framework by using an empirical scaling law method. This chapter proposed three new empirical scaling laws based on the original maximal price change (MPC) scaling law by Glattfelder, Dupuisy, and Olsen (2011). The most valuable property is the scale invariance which allows financial analysts to assess the maximum loss for any time interval. The strong evidence of the effectiveness of the new scaling law methods are provided in the model performance section. We find that the scaling law method with only one month in-sample data can already provide a good prediction of risk, whereas for the conventional VaR at least ten years length of data is needed to obtain a reasonable result.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available