Investigation of NYSE High Frequency Financial Data for Intraday Patterns in Jump Components of Equity Returns - PowerPoint PPT Presentation

Loading...

PPT – Investigation of NYSE High Frequency Financial Data for Intraday Patterns in Jump Components of Equity Returns PowerPoint presentation | free to download - id: 1c28f8-ZDc1Z



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Investigation of NYSE High Frequency Financial Data for Intraday Patterns in Jump Components of Equity Returns

Description:

Summarize research from Econ 201FS: Research Seminar and ... announcements, Engle et. al (1990), Chaboud, Chernenko, Howorka, Krishnasami, Liu, Wright (2004) ... – PowerPoint PPT presentation

Number of Views:35
Avg rating:3.0/5.0
Slides: 18
Provided by: peterlound
Learn more at: http://econ.duke.edu
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Investigation of NYSE High Frequency Financial Data for Intraday Patterns in Jump Components of Equity Returns


1
Investigation of NYSE High Frequency Financial
Data for Intraday Patterns in Jump Components of
Equity Returns
  • Senior Honors Thesis Proposal
  • Peter Van Tassel
  • Duke University
  • Durham, North Carolina
  • 10 September 2007

2
Agenda
  • Summarize research from Econ 201FS Research
    Seminar and Lab on High Frequency Financial Data
    Analysis for Duke Economics Juniors
  • Motivation
  • Data
  • Non-parametric statistics BNS and LM
  • Intraday patterns in Jump Components of Equity
    Returns
  • Moving forward, open discussion in regard to
    developing the current hypothesis

3
Motivation
  • Well documented U-shaped patterns in intraday
    equity return volatility dates back to Wood,
    McInish Ord (1985)
  • Literature on FX volatility detects similar
    patterns, particularly in response to
    macroeconomic announcements, Engle et. al (1990),
    Chaboud, Chernenko, Howorka, Krishnasami, Liu,
    Wright (2004)
  • Our hypothesis from the background literature is
    that jump components of volatility are driven by
    new information and that more information will be
    released at the start of the trading day
  • Our results indicate that statistically
    significant jumps are concentrated around the
    start of the trading day

4
Data
  • High frequency data from TAQ
  • Arranged with the help of Tzuo Hann Law
  • Focus on the SPY as a proxy for the market
    portfolio
  • Will consider individual equities such as PEP,
    KO, BMY
  • Price series begins on 1 Jan 2001 and ends on 31
    Dec 2005, with 771 observations per day and 1241
    total days
  • Sampling frequency used for the purpose of this
    research is 17.5 minutes unless stated otherwise

5
SPY Level and Return Plots
6
The Lee-Mykland Statistic
  • Stock price evolution modeled as
  • µ(t)dt drift term, s(t)dt geometric Brownian
    motion, dJ(t) non-homogenous Poisson-type jump
    process
  • LM Statistic

7
LM applied to SPY Data
8
LM Applied to Individual Stocks
9
Sampling Frequency
  • Admittedly, the LM statistic is not an authority
    in the non-parametric literature
  • One problem we find is stabilization at different
    sampling frequencies
  • We select a sampling frequency of 17.5 minutes in
    an attempt to alleviate problems related to
    market micro-structure noise

10
Realized and Bi-Power Variation Alternate
Approach and Similar Results
  • Model for stock price evolution
  • µ(t)dt drift term, s(t)dt geometric Brownian
    motion, dLj(t) pure jump Lévy process with
    increments
  • Returns defined as
  • M is our within day sampling frequency

11
Realized and Bi-Power Variation
  • Realized Variation
  • Bi-power Variation

12
Intraday averages
  • Define intraday realized variation as,
  • Define intraday bi-power variation as,
  • Also consider,

13
Plots for SPY Data
14
Plots for Individual Stocks
15
BNS Statistics
  • Tri-power variation defined as,
  • Huang and Tauchen (2005) recommend,

16
More plots
17
Conclusions
  • LM statistic indicates that jumps are more
    concentrated at the start of the trading day
  • Realized variation and Bi-power variation plots
    also indicate that the jump component is more
    prominent in the first 35 min of the trading day
  • Future Work
  • Investigate whether the jump component measured
    early in the day is helpful in forecasting
    volatility,
About PowerShow.com