Please use this identifier to cite or link to this item: http://hdl.handle.net/10397/81918
Title: Implied volatility and its information content of future news
Authors: Chen, Sipeng
Advisors: Li, Gang (AF)
Cheng, Agnes (AF)
Keywords: Options (Finance)
Securities -- Prices
Stock price forecasting
Issue Date: 2019
Publisher: The Hong Kong Polytechnic University
Abstract: My study examines the information content of implied volatility (IV) and finds out that implied volatility(IV) can predict future realized volatility (RV) through the channel of predicting future news. There are two pieces of supporting evidence in my results. The first evidence is the "RV-relevance" of news, which refers to the fact that RV can be explained by the concurrent news in a batch of time-series regressions for each individual firm. The cross-sectional average of the regression coefficients is significantly positive with a t-statistic over 35, and more than 90% of the stocks have positive coefficients. The second evidence is the "IV-predictability" of news, which refers to the fact that the future news can be forecasted by the IV in a batch of time series regressions for each individual firm. The cross-sectional average of the regression coefficients is significantly positive with a t-statistic about 20, and around 65% of the stocks have positive coefficients. Both pieces of evidence hold up robustly in different measures of news and in most kinds of news. More specifically, I devise two measures to quantify the news. One is the news intensity (N) by counting the news occurrence(s) within a month. The other one is the news volatility (NV) which measures the total news impact magnitude, as defined by the sum of squares of the Composite Sentiment Score(s) (CSS) within a month. IV has a greater prediction power for NV than for N. Consistently, NV is greater in RV-relevance than N. Furthermore, I apply different news classification methods to find: which kind of news is more RV-relevant; which kind of news is more IV-predictable; and which kind of news drives the RVt-IVt-1 forecasting relation the most. In terms of timing predictability, the unscheduled news is more RV-relevant while the scheduled news is more IV-predictable. The scheduled news drives the forecasting relationship more. In terms of news formats as proxy for different information roles of media, both news-flash (proxy for information dissemination role) and full-article (proxy for information creation role) are similarly strong in RV-relevance. But news-flash is more IV-predictable, and thus drives the forecasting relation more. In terms of different news content groups, the overall results suggest that the strength of the IV-predictability is monotonically increasing with the strength of the RV-relevance. Thus, I can select all the news content groups that have high RV-relevance to form new measures of news. These new measures of news drive the RVt-IVt-1 forecasting relation significantly more than the original news measures of all the news. Eventually, through a two-stage mediation analysis, I am able to quantify the strength of the predictability through different kinds of news channels over the total predictability of IVt-1 on RVt. The best mediation model captures that about one-third of the total predictability comes from the news channel. JEL Classification: G12, G14, G17
Description: 2, 122 pages : color illustrations
PolyU Library Call No.: [THS] LG51 .H577P AF 2019 Chen
URI: http://hdl.handle.net/10397/81918
Rights: All rights reserved.
Appears in Collections:Thesis

Files in This Item:
File Description SizeFormat 
991022289508403411_link.htmFor PolyU Users168 BHTMLView/Open
991022289508403411.pdfFor All Users799.76 kBAdobe PDFView/Open
Show full item record
PIRA download icon_1.1View/Download Contents

Page view(s)

17
Citations as of May 6, 2020

Download(s)

5
Citations as of May 6, 2020

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.