SMU Research Data Repository (RDR)
Browse
36_Estimating_and_Forecasting_Intraday_Volatility.pdf (391.72 kB)

Estimating and forecasting intraday volatility

Download (391.72 kB)
thesis
posted on 2020-05-15, 08:46 authored by Xuna GAO

The purpose of this study is to investigate stock volatility and forecasting performance of different volatility models over high-frequency intervals. The multiplicative component model that decomposes the conditional variance into a daily component and a periodicity component is studied with different specifications. This model is applied to 30 stocks. For the daily component, both parametric and non-parametric measures are considered. 12 models that capture the long memory feature of volatility are examined. Our results show the HAR-MEM model with overnight jump and the HAR-MEM model have the best forecasting performance among 12 models, and adding an overnight return term improves model’s forecasting ability. Periodicity component is captured by the proportion of summation of intraday volatility to summation of daily volatility over some time period. In comparison with the literature, our specification of periodicity component has slightly better forecasting performance in the first 2-hour volatility.

History

Document type

  • Master thesis

Degree awarded

MSc in Economics

Year degree awarded

2013

Supervisor(s)

TSE Yiu Kuen

Committee members

YANG Zhenlin; Anthony S. TAY

SMU School

  • School of Economics

No. of pages

81

Usage metrics

    School of Economics Theses

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC