Public information arrival: Price discovery and liquidity in electronic limit order markets
Highlights
► We study the impact of newswire messages on market quality. ► Negative messages convey more information than positive or neutral messages. ► Adverse selection costs around news are especially high around negative messages. ► Liquidity decreases around negative messages and increases for other messages. ► Market participants have different information processing capabilities.
Introduction
Technology and computers have not only changed trading in financial markets, but have also revolutionized the way financial news is disseminated and analyzed. As trading technology has advanced, news providers have kept pace and deliver news to traders around the world within a fraction of a second. News providers have also started to offer newswire products with machine-learning systems that cater to algorithmic traders. However, most news is still read by professional human traders who read newswires such as Thomson Reuters, Bloomberg, or Dow Jones on a regular basis. They spend a considerable amount of time and money on these information sources and emphasize the importance of the speed and accuracy of their news. Newswire messages represent much of the overall information and real-time information traders receive. The intraday impact of newswire messages is however still not well understood. It is not entirely clear whether newswire messages actually contain new information, whether traders act in advance or after the arrival of messages, and how newswire messages impact liquidity and price dynamics in a modern electronic limit order market.
This paper studies the impact of Thomson Reuters newswire messages on the intraday price discovery, trading activity, and liquidity of stocks traded on the Toronto Stock Exchange. The Toronto Stock Exchange is well suited for such an analysis. First, it is a highly automated electronic limit order book market comparable to many international exchanges. Second, in contrast to most European markets, there are no major second language news streams. Third, the Canadian market has a very low level of fragmentation during our observation period.
In this paper, newswire messages are clustered by sentiment. The differentiation between positive, negative, and neutral news enables us to investigate potentially asymmetric reactions to newswire messages based on their tone. Liquidity increases around positive and neutral messages and decreases around negative messages. Trading intensity increases around all types of newswire messages. In general, we find higher adverse selection costs around newswire messages. Negative messages are associated with significantly higher adverse selection costs than positive messages.
Traditional financial theory does not differentiate between positive and negative public information. However, psychological studies from the field of impression formation show that humans react stronger to bad news than to good news (cf. Soroka, 2006). Overall, our results suggest that participants’ possess different information gathering and information processing capabilities and that these participants react differently to good and bad news.
The remainder of the paper is structured as follows. Section 2 presents related literature. Section 3 gives an overview on the institutional structure of the Toronto Stock Exchange. Section 4 provides details on newswire messages, trade and order book data, and the sample selection. Section 5 introduces the research design and methodology. Section 6 provides the results and interpretation and Section 7 concludes.
Section snippets
Related work
The existing public information literature studies different types of public information, from unstructured media content to scheduled earnings announcements. The Thomson Reuters newswire messages are in-between these extremes. Ranaldo (2008) analyzes the intraday market dynamics of firm specific unstructured news at the Paris Bourse. The six months of news data is based on the Reuters alert system without ex ante measures news sentiment (i.e. positive, negative, or neutral). He finds a
Institutional details
The Toronto Stock Exchange (TSX) is Canada’s most important equity exchange and is operated by the TMX Group.1 The TSX is North America’s third
Data and sample selection
Our news data consists of Thomson Reuters NewsScope Content and is tagged using the Thomson Reuters NewsScope Sentiment Engine (RNSE).3 The RNSE real-time data stream is disseminated to approximately 370,000 Reuters screens worldwide. According to Thomson Reuters, they “deliver over 500,000 alerts and over two million unique stories a year”.
Price discovery
There are several methodologies to decompose stock price movements around information events into various components. The most common models are the MRR model introduced by Madhavan et al. (1997) and the model by Huang and Stoll (1997). More recent developments employ a variance decomposition into information-driven and noise-induced volatility. This is applied for macroeconomic news using a state-space model as in Hautsch et al. (2011) and rather concentrates on volatility-based price
Results and interpretation
The descriptive statistics for each firm in our sample are in Table 2. The average firm has 201 distinct news items and a marginally negative sentiment of −0.04. The average sentiment is in line with studies that report a bias towards negative media coverage (Soroka, 2006). The average market capitalization of a firm over the years 2005–2008 is approximately C$24bn. Market capitalization ranges from C$5.5bn to C$63bn. With roughly 63% of total market capitalization, our sample covers a large
Conclusion
In this paper, we analyze the impact of Thomson Reuters newswire messages on intraday price discovery, liquidity, and trading intensity at the Toronto Stock Exchange. In contrast to existing literature, we are able to cluster news based on message content. We split the news data into groups of news messages with positive, negative, and neutral sentiment which gives us the opportunity to study asymmetric reactions to news messages. News are not sorted based on ex post return measures but on ex
Acknowledgements
The authors thank Ike Mathur (the managing editor) and an anonymous referee for valuable comments and suggestions. Financial support from Stuttgart Stock Exchange and from the IME Graduate School at Karlsruhe Institute of Technology (KIT) funded by the Deutsche Forschungsgemeinschaft (DFG) is gratefully acknowledged.
References (24)
- et al.
A cross-exchange comparison of execution costs and information flow for NYSE-listed stocks
Journal of Financial Economics
(1997) - et al.
Stock return variances: the arrival of information and the reaction of traders
Journal of Financial Economics
(1986) - et al.
When machines read the news: using automated text analytics to quantify high frequency news-impled market reactions
Journal of Empirical Finance
(2011) - et al.
The impact of macroeconomic news on quote adjustments, noise, and informational volatility
Journal of Banking and Finance
(2011) - et al.
Market reactions to anticipated announcements
Journal of Financial Economics
(1991) - et al.
Market liquidity and volume around earnings announcements
Journal of Accounting and Economics
(1994) - et al.
The power of bad: the negativity bias in Australian consumer sentiment announcements on stock returns
Journal of Banking and Finance
(2010) - et al.
Is all that talk just noise? The information content of internet stock message boards
The Journal of Finance
(2004) - et al.
All that glitters: the effect of attention and news on the buying behavior of individual and institutional investors
The Review of Financial Studies
(2008) - et al.
Public information arrival
The Journal of Finance
(1994)
Asymmetric effects of federal funds target rate changes on S&P100 stock returns, volatilities and correlations
Journal of Banking and Finance
Cited by (103)
An MA-MRR model for transaction-level analysis of high-frequency trading processes
2024, Journal of Management Science and EngineeringLost in translation. When sentiment metrics for one market are derived from two different languages
2023, Journal of Behavioral and Experimental FinanceBreaking news headlines: Impact on trading activity in the cryptocurrency market
2023, Economic ModellingThe effect of overnight corporate announcements on price discovery
2023, Finance Research LettersForeign institutions and the behavior of liquidity following macroeconomic announcements
2022, Finance Research LettersHow does news sentiment affect the states of Japanese stock return volatility?
2022, International Review of Financial Analysis