Trading and non-trading period Internet information flow and intraday return volatility
Introduction
The Mixture of Distribution Hypothesis (MDH) argues that the volatility of stock return is proportional to the rate of information flow at a given interval [1], [2], [3], [4]. Although this is a reasonable hypothesis linking the information flow and the “stylized facts” of stock return. Empirical validation is difficult in quantifying the intangible form of information as well as capturing the volatility of return. Recent studies show evidence that the Autoregressive Conditional Heteroskedasticity (ARCH) model can capture well the variation in the volatility process [5], [6]. In support of the argument of MDH, Lamoureux and Lastrapes [7] firstly introduce the daily trading volume as the proxy for information flow and show a significant explanatory power to the variance of daily returns. Wagner and Marsh [8] use the surprise volume (unexpected above-average trading activity) as the proxy for information flow and show that surprise volume can explain volatility persistence and excess kurtosis. Kalev et al. [9] employ the firm-specific announcements as the proxy for information flow and find that the inclusion of this variable in the conditional variance equation of the GARCH model reduces the volatility persistence. Although the above-mentioned empirical studies strongly support the MDH, they are subject to criticism. Firstly, the trading volume-based proxies may suffer from the nonstationary properties and endogeneity. Secondly, firm-specific announcement is not sufficient, which only represents one aspect of the information flow.
In a seminal work by Zhang et al. [10], they employ the number of information appeared in Baidu News as the proxy for information flow, make a comprehensive comparison with the existing proxy and find that this Internet information-based proxy can reduce the volatility persistence of the SZSE SME PRICE INDEX. Our paper is a naturally extension of their work and contribute to the literature in two aspects. Firstly, we extract the information from Baidu News for individual stock, rather than focus on the stock index. Observing the explanatory for individual stock provides additional evidence for MDH. Secondly, given to the availability of the Internet information, we divide the information into trading period information and non-trading period information, include each of them into the conditional variance equation of the GARCH model and observe the decreases of the volatility persistence.
This paper is also in line to a broad literature on using Internet information for financial economics research [11], [12], [13], [14], [15]. Specifically, Da et al. [16] use the search frequency in Google Trends as the proxy for investor attention and find that this direct proxy can predict stock prices in the next 2 weeks and long-run underperformance of IPO. In a similar way, Zhang et al. [17] use the search frequency of stock name in Baidu Index as the proxy for investor attention and show that the quantified variable can predict the abnormal return and trading volume. Preis et al. [18] analyze changes in Google query volumes for search item related to finance and find patterns can be viewed as “early warning signs” of stock market movement. Moat et al. [19] also document changes in Wikipedia contain early signs of stock market movement.
The rest of the article is organized as follows. Section 2 describes the data. Section 3 presents the mode setup as well as the empirical results. Section 4 is the concluding remarks.
Section snippets
Data description
The dataset comprises daily returns for 10 actively traded stocks in CSI 300 Index,1
Models
In line with previous empirical studies [7], [8], [9] and by making a direct comparison, we also adopt the GARCH (1, 1) model. Because the GARCH (1, 1) specification is a parsimonious characterization of conditional variance, which fits many financial time series quite well [20]. Therefore, the model can be expressed in the following equations: where represents the stock return in D2 and D4, respectively, which is calculated as the closing price
Concluding remarks
This paper connects to employing Baidu News as the proxy for information flow, incorporates this proxy into the conditional variance equation of the GARCH (1, 1) model and provides some evidence the MDH with trading and non-trading period Internet information for individual stock return volatility persistence. Although the empirical results show that the volatility persistence is most decreased when including the contemporary information, there is still some unexplained volatility persistence.
Acknowledgment
This work is supported by the National Natural Science Foundation of China (71532009, 71201112 and 71320107003).
References (22)
- et al.
ARCH modeling in finance: A review of the theory and empirical evidence
J. Econometrics
(1992) - et al.
Public information arrival and volatility of intraday stock returns
J. Bank. Finance
(2004) - et al.
Internet information arrival and volatility of SME PRICE INDEX
Physica A
(2014) - et al.
Open source information, investor attention, and asset pricing
Ecol. Modell.
(2013) Generalized autoregressive conditional heteroskedasticity
J. Econometrics
(1986)A subordinated stochastic process model with finite variance for speculative prices
Econometrica
(1973)- et al.
The stochastic dependence of security price changes and transaction volumes: implications for the mixture-of-distributions hypothesis
Econometrica
(1976) - et al.
The price variability-volume relationship on speculative markets
Econometrica
(1983) Transaction data tests of the mixture of distributions hypothesis
J. Finan. Quant. Anal.
(1987)Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation
Econometrica
(1982)
Heteroskedasticity in stock return data: volume versus GARCH effects
J. Finance
Cited by (45)
Insight into the driving force of environmental performance improvement: Environmental regulation or media coverage
2022, Journal of Cleaner ProductionCitation Excerpt :The number of media reports involving sample enterprises is screened by full-text retrieval, and the value is added with 2 to take the natural logarithm (Wang and Ye, 2015). New media coverage (MEDIA_nt) is taken from “Baidu News”, adding 2 to the number of news reports obtained by the search and taking the natural logarithm (Li et al., 2021; Shen et al., 2016). Online reprinting means higher media coverage, so repeated reports are not excluded.
Trading volume and realized higher-order moments in the Australian stock market
2020, Journal of Behavioral and Experimental FinanceFirm-specific information and systemic risk
2020, Economic ModellingThe impact of oil price changes on stock returns of new energy industry in China: A firm-level analysis
2019, Physica A: Statistical Mechanics and its ApplicationsCitation Excerpt :Several recent literatures find mixed evidence in that returns of firms in the same industry related to oil price changes [13–15]. First, the volatility of oil prices [16–18] are susceptible to investor fear gauge or internet information [19–21], exchange rate [22], political events [23], and so on. It has a direct or indirect impact on the production costs of firms and then firm stock returns respond differently to it.
Stock market response to information diffusion through internet sources: A literature review
2019, International Journal of Information ManagementBehavioral heterogeneity and excess stock price volatility in China
2019, Finance Research Letters