our News

The relationship between news articles and the behaviour of the stock market has received steady interest by the scientific research community over the last 40 years. With the emergence of effective text analysis technology, and recently with the onset of semantic processing, efforts to capture this relationship have been intensified with (among many others) the following interesting results.

In independent publications, Fung (2), Klopchenko (3), Larkin (4), Liang (5; 6), Queck (7), Shegal (8), Schumaker (9; 10), Takehashi (11), Wutrich (12) and Yu (13) have all reported that using news as a additional input significantly improves the reliability of stock price prediction models. With the exception of Fung and Liang who both used Chinese language messages as input, all input was English language news. Queck quantifies the improvement at 8% over a model using numerical data only. Larkin concludes that correctly classified news could be coupled with traditional price and volatility inputs to give a market participant a sizable advantage.

Both Fung, Schumaker and Thomas report profitable results when using a news based system for simulated trading. Schumaker reported a 3.60% return in simulated trading on 500 stock over a five week period. In a much smaller simulation Yu made a 5.11% profit in two months. Thomas claims a return up to 30% provided that enough text data is available to train his classifiers. But when detection of abnormal market conditions suffices, Robertson (14) reports a staggering 80% reliability, leading him to concluded that forecasting the market reaction to news is a viable option.

The time window in which the asset markets responds to news is investigated by Ederington (15; 16), Gidofalvi (17) and Nofsinger (18). Ederington concludes that for interest rate and foreign exchange markets, prices respond within 10 second to the release of scheduled macroeconomic news and the changes are completed within 40 seconds. Volatility on the other hand is slower in its response, which may last op to several hours. Nofsinger reports for S&P options volume a 2 hour delay after scheduled economic news is released, but agrees with Elderington in the near immediate increase in volatility. Gidofalvi reports a definite predictive power for the stock price movement in the interval starting 20 minutes before and ending 20 minutes after news articles become publicly available.

The nature of the news has also been found relevant to Blasco (19), Boyd (20) and Liang (5) who independently report “bad news” to have a more profound impact on the markets than positive messages. In addition, Boyd and Liang also find the general market conditions to be of influence . Boyd reports that information about interest rates dominates during expansions and information about future corporate dividends dominates during contractions. Liang concludes that in the bear market, bad news influences the market more severely than good news and recommends that the impact of news should be addressed separately for bear or bull markets.

Moreover, also the time of day has been found relevant in researching the correlation between market price and news. Robertson (21) reports that there is an increased likelihood of events that lead to increased volatility in the news at the beginning of the day.

Mittermayer (22) also reports that a categorization of press releases is able to provide additional information that can be used to forecast stock price trends and has proved this by significantly outperforming a trader randomly buying and shorting stocks immediately after the publication of press releases. He warns however that an adequate trading strategy is essential for the results of the news analysis to be fully exploited. He concludes that the most promising way for improving the performance of new analysis systems is to make use of a priori domain knowledge. This is further supported by Schumaker (5) who yields the best results using named entity recognition over traditional “ bag of words” methods.

Admittedly, many of the listed studies vary widely in approach and statistical relevance. However, they would appear to support the notion that modern stock markets are at the least not completely informational efficient. This means that for those that have the technology to be kept aware of all information about their portfolio, the future stock prices will reveal themselves to be far from random – thus gaining a definitive competitive advantage.

References

1. Paulos, John Allen. Recession forecast if steps not taken. A Mathematician reads the newspaper. 1995.

2. Fung, Gabriel Pui Cheong, Yu, Jeffrey Xu and Lam, Wai. News Sensitive Stock Trend Prediction. Advances in Knowledge Discovery and Data Mining. s.l. : Springer Berlin / Heidelberg, 2002.

3. Kloptchenko, Antonia, et al. Combining Data and Text Mining Techniques for Analyzing Financial Reports. Intelligent Systems in Accounting, Finance & Management. s.l. : John Wiley & Sons, Ltd., 2004.

4. Larkin, Fiacc and Ryan, Conor. Good News Using News Feeds with Genetic Programming to Predict Stock Prices. Genetic Programming. s.l. : Springer Berlin / Heidelberg, 2008.

5. Liang, Xun. Impacts of Internet Stock News on Stock Markets Based on Neural Networks. Advances in Neural Networks – ISNN 2005. s.l. : Springer Berlin / Heidelberg, 2005.

6. Mining Stock News in Cyberworld Based on Natural Language Processing and Neural Networks. Liang, Xun and Chen, Rong-Chang. s.l. : IEEE – Neural Networks and Brain, 2005.

7. Predicting impact of news on stock price An evaluation of neuro fuzzy systems. Quek, C., Cheng, P. and Jain, A. s.l. : IEEE Congress on Evolutionary Computation, 2007.

8. Sehgal, Vivek and Song, Charles. SOPS Stock Prediction using Web Sentiment. Proceedings of the Seventh IEEE International Conference on Data Mining Workshops. s.l. : IEEE Computer Society, 2007.

9. Textual Analysis of Stock Market Prediction Using Financial News Articles. Schumaker. 2006.

10. Schumaker, Robert P. and Chen, Hsinchun. Textual Analysis of Stock Market Prediction Using Breaking Financial News The AZFinText System. s.l. : ACM, 2009.

11. Takahashi, Satoru, et al. Analysis of the Relation Between Stock Price Returns and Headline News Using Text Categorization. Lecture Notes In Artificial Intelligence. s.l. : Springer-Verlag Berlin, Heidelberg, 2007.

12. Daily stock market forecast from textual web data. Wutrich, B., et al. s.l. : IEEE International conference on systems man and cybernetics, 1998.

13. Zhai, Yuzheng, Hsu, Arthur and Halgamuge, Saman K. Combining News and Technical Indicators in Daily Stock Price Trends Prediction. Advances in Neural Networks – ISNN 2007. s.l. : Springer Berlin / Heidelberg, 2007.

14. Can the Content of Public News be used to Forecast Abnormal Stock.pdf. Robertson, Calum, Geva, Shlomo and Wolff, Rodney. s.l. : Seventh IEEE International Conference on Data Mining, 2007.

15. Ederington, Louis H and Lee, Jae Ha. How Markets Process Information News Releases and Volatility. The Journal of Finance. 1993.

16. —. The Short-Run Dynamics of the Price Adjustment to New Information. Journal of Financial and Quantatative anlysis. 1995.

17. Gidófalvi, Győző. Using News Articles to Predict Stock Price Movements. San Diego : University of California, 2001.

18. Nofsinger, John R and Prucyk, Brian. Option volume and volatility response to scheduled economic news releases. The Journal of Futures Markets. 2003.

19. Blasco, Natividad, et al. Bad news and Dow Jones make the Spanish stocks go round. European Journal of Operational Research. 2005, 163.

20. Boyd, John H., Hu, Jiann and Jagannathan, Ravi. The Stock Market’s Reaction to Unemployment News Why Bad News Is Usually Good for Stocks. The Journal of Finance. 2005, 2.

21. What Types of Events Provide the Strongest Evidence that the Stock Market is Affected by Company Specific News. Robertson, Calum, Geva, Shlomo and Wolff, Rodney. s.l. : Fifth Australasian Data Mining Conference, 2006.

22. Forecasting Intraday Stock Price Trends with Text Mining Techniques. Mittermayer, Marc-André. s.l. : Hawaï International Conference on System Sciences, 2004.

23. Graham, Michael, Nikkinen, Jussi and Sahlström, Petri. Relative importance of scheduled macroeconomic news for stock market investors. Journal of Economics and Finance. 2003.

24. Mittermayer, Marc-André and Knolmayer, Gerhard F. Text Mining Systems for Market Response to News A Survey.pdf. s.l. : University of Bern, 2006. 184.

25. Does Company Specific News Effect the US, UK, and Australian Markets within 60 minutes. Robertson, Calum, Geva, Shlomo and Wolff, Rodney. Sydney : Australasian Finance and Banking Conference, 2006.

26. News Aware Volatility Forecasting. Robertson, Calum, Geva, Shlomo and Wolff, Rodney. s.l. : Sixth Australasian Data Mining, 2007.

27. Integrating Genetic Algorithms and Text Learning for Financial Prediction. Thomas, James D and Sycara, Katia. 2002.

28. Tumarkin, Robert and Whitelaw, Robert F. News or Noise Internet Postings and Stock Prices. Financial Analysts Journal. 2001.

29. Classify Unexpected News Impacts to Stock Price by Incorporating Time Series Analysis into Support Vector Machine. Yu, Ting, et al. s.l. : International Joint Conference on Neural Networks, 2006.

30. Melvin, Michael and Yin, Xixi. Public Information Arrival Exchange Rate Volatility and Quote Frequency. s.l. : Arizona state University, 2000.

31. Szeto, K.Y. and Fong, L.Y. How Adaptive Agents in Stock Market Perform in the Presence of Random News A Genetic Algorithm Approach. Intelligent Data Engineering and Automated Learning — IDEAL 2000. Data Mining, Financial Engineering, and Intelligent Agents. s.l. : Springer Berlin / Heidelberg, 2000.