Time-variation in the impact of news sentiment

Abstract

Asian Finance Association Conference (2013) for constructive and helpful comments. I would also like to gratefully acknowledge the financial support from the Curtin Business School journal publication support fund. Utilising firm-specific news sentiment data provided by Thomson Reuters News Analytics, I construct aggregate measures to examine the relationship between news sentiment and stock market returns over the period 2004-2010. I find a highly significant relationship between aggregated measures of news sentiment and stock returns that fluctuates by time and by industry. I identify a link between the time-variation of news sentiment impact and industry beta, and determine that levels of investor sentiment (proxied by VIX) play an important role in explaining this variation.

Investor sentiment, flight-to-quality, and corporate bond comovement

“…and is widely used as investor sentiment proxy (e.g., Kurov, 2010;Kaplanski and Levy, 2010;Da et al, 2015;and Smales, 2015). A high value of VIX corresponds to low investor sentiment.…”

News media analytics in finance: a survey

“…These aspects allow researchers to study the impacts of news content aggregated over varying scales of time and entity. For example, researchers have studied the influence of news at the firm level (Tetlock et al ., ; Ferguson et al ., ), industry level (Li et al ., ; Smales, ), market level (Tetlock, ; Wei et al ., ), commodity level (Clements and Todorova, ; Maslyuk‐Escobedo et al ., ), between currencies (Nassirtoussi et al ., ), across countries (Griffin et al ., ) and at the global market level (Uhl et al ., ). Time horizons vary from intraday (Groß‐Klußmann and Hautsch, ; Ho et al ., ), daily (Tetlock, ; Garcia, ), weekly (Lu and Wei, ; Sinha, ), monthly (Ammann et al ., ; Cahan et al ., ) and over several years (Hillert et al ., ; Kraussl and Mirgorodskaya, ).…”

“…For formation periods exceeding one day, these measures may either be computed daily before being averaged over the entire formation period (Uhl, ; Sinha, ), or pooled together at once (Cahan et al ., ). It is also common for each document to be weighted by some measure of relative importance such as story relevance (Ho et al ., ; Shi et al ., ), or, in the case of aggregations over sector and market indices, by firm market capitalisation (Smales, ).…”

“…The market‐level effects of uninformative news and firm‐level effects of informative news was followed by an amalgamation of these two approaches: the cross‐sectional aggregation of firm‐specific news content into market‐ and sector‐level news indices (Smales, ). This subset of the literature examines the response to informative news content over different collections of assets, primarily at the market level.…”

Building News Measures from Textual Data and an Application to Volatility Forecasting

“…Riordan et al (2013) suggest that negative newswire messages from Reuters NewsScope Sentiment Engine, compared to positive ones, are associated with higher adverse selection costs, are more informative, and have a more significant impact on high-frequency asset price discovery and liquidity. Smales (2015) use Thomson Reuters News Analytics sentiment scores to create aggregate daily news sentiment indicators and find that positive and negative news result in above and below average returns, respectively, and that neutral news days are indistinguishable from days without news. Allen et al (2017) use the Thomson Reuters News Analytics data set to construct a series of daily sentiment scores for the Dow Jones Industrial Average (DJIA) stock index component companies, and study the relationship between these financial news sentiment scores and the stock prices of these companies using entropy metrics, which permit an analysis of the amount of information within the sentiment series, its relationship to the DJIA and an indication of how the relationship changes over time.…”