5. Conclusion and Discussion 5.1 Research Discussion and Contribution
This study analyzes the relationship between macroeconomic news and market volatility from January 2007 to December 2017. The macroeconomic news used in this study is extracted from CMoney. The study applies the text mining techniques on news articles and collect news article related to local industries and global stock markets. A sentiment dictionary is constructed to quantify the wordings and tones of the news articles. The dictionary used in this study is integrated from the sentiment and financial warning dictionary established by Seng (2017), Chang (2009) and Lin (2013).
VIXTWN is used to represent investors’ sentiment, VIXTWN can also be known as the fear gauge. Financial news affect investors’ sentiment, hence, financial news affect market volatility.
The result of the study provide empirical support for our proposition that investors’
sentiment is associated with financial news and investors’ sentiment further affect market volatility. The study observes the impact of local industrial news and global stock market news on market volatility separately. Result shows that percentage of positive and negative words used in both types of news are correlated with VIXTWN.
When the percentage of positive words increases, investors’ sense of security increases and market volatility decreases. When the percentage of negative words increases, investors change their view on future market movements, therefore, market volatility increases. Moreover, the result suggests that investors are more sensitive to negative information, the coefficient of pessimistic mood is greater than the coefficient of optimistic mood. The optimistic mood and pessimistic mood of news affect the VIXTWN of the current day and the coming two days.
The second finding of the study is that news tone of news are associated with market volatility. News tone measures the strength of wordings used in financial news and it represents the overall direction of opinion for news published on the same day.
Results shows that news tones of local industrial news and global stock market news are negatively associated with VIXTWN. Positive news tone indicates an overall positive opinion presented by the news articles, positive information reduces market volatility. We consider the immediate and delayed impact of news tones on VIXTWN.
Global stock market news tones pose impact on VIXTWN of the current day and the coming two days. However, the delayed reaction of VIXTWN generated by local industrial news is more transient, local industrial news tone affect market volatility of the current day and the coming day.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Furthermore, the research analyzes the relationship between quantity of news and market volatility. Result suggests that the weighted average quantity of local industrial news is not significantly correlated with market volatility. However, average quantity of global stock market news is associated with market volatility. The information content carried by local industrial news are more complex and messy, its quantity do not reflect useful information content that investors gather from the news articles. The quantity of global stock market news is not as much as the quantity of local industrial news. News related to the global stock market usually carry more useful content and it represent the size of information that investors receive from the media.
The study includes four major sentiment variables in our models, more variables was originally included in the model to enrich the elements concerned in the model.
The difference between number of positive words and negative words used in the news articles was considered in our models. However, this variable give rise to the problem of collinearity and affect the result of the model. Therefore, we decide to exclude this variable from the model to ensure accuracy of results.
Moreover, we establish different models to consider the delayed reaction of VIXTWN caused by other sentiment variables. The regression result shows that the interaction between VIXTWN and other sentiment variables of three days before are not significant. Therefore, we include models measuring the interaction of VIXTWN and other sentiment variables of the current day, the day before and two days before.
The empirical result of this study proves the association between financial news and market volatility. The media communicate with investors through news articles and content of news article reveal signs of financial warning. Rational investors’ sentiment is affected by wordings and tone of news, their sentiment changes further affect their investment decision, hence, affect market volatility. The study also proves the delayed reaction of VIXTWN cause by the content of financial news.
‧
5.2 Limitation and Future Research Work
In this study, we use the name and abbreviation of local industries and stock markets to extract news that are related to our research. We expect to collect non-firm –specific, industry-wide or market-wide news articles from the database. However, the mass media sometimes mentions the industry that the firm belongs to, in firm-specific news articles. VIXTWN is a macroeconomic indicator, firm-specific news may not affect the overall market volatility. We should exclude firm-specific news when doing future research work. In addition, the abbreviations of industries and stock markets are identified by manual, some abbreviations that are not usually used may be missed. The list of abbreviation should be further enriched.
Moreover, many news articles describe different issues in a single article. Our research captures and counts all the positive words and negative words in the news. But the words that we count may not be describing events happened or will happen in that particular industry or stock market. Therefore, the number of positive words and negative words counted may be overstated. However, we believe that this problem is inevitable.
The study uses a sentiment dictionary to measure the value of the sentiment variables. The dictionary is enriched by integrating dictionaries constructed by different studies. Although the dictionary is more comprehensive, some Chinese proverbs, idioms and financial jargons are not included in the dictionary. Future researches should further enrich the sentiment dictionary used in this study. Short phrases, proverbs and financial jargons should be consider when measuring news tone. Apart from further enriching the dictionary, we could also score different words. Different wordings represent different level of severity. For example, financial tsunami and financial distress are both negative words, however, financial tsunami is more serious than financial distress.
In addition, we believe that issues of different industries would impact investors’
sentiment in different extent, we estimate the relative influence according to the market value of the industry. However, weighting of TAIEX may not be an accurate estimate of the relative influence of industry-specific news on market volatility. Indeed, we believe that it is difficult to use other methods to better estimate the relative influence of issues of different industries on market volatility. Furthermore, the weighting of TAIEX changes daily, the weighting that we use is not the timely relative market value of industries. Historical data of TAIEX industries weighting is not available in the TEJ database and website of TWSE.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Furthermore, we calculate the overall stock market sentiment variables by simple average. We cannot accurately estimate the relationship between events happened in other stock markets and local investors’ sentiment. In the future, we should establish methods that can better estimate the relative influence of issues happened in different stock market on Taiwanese investors’ sentiment. Suitable weighting should be used to calculate the overall stock market sentiment variables accurately.
In this study, we consider the relationship between optimistic mood, pessimistic mood of news and market volatility. We did not consider the relationship between percentage of neutral words and market volatility. Percentage of neutral words might be associated with market volatility, we could include the percentage of neutral words as one of the dependent variables in future studies.
Last but not least, we group the news according to the date that they were published.
However, the opening and closing time of Taiwan stock exchange is 09:00 and 13:30 respectively. News published after the closing time of the previous day and before the closing time of the current day should be classified as news of the current day. As news published after the closing time would not affect the market volatility of the current day, but they would affect the market volatility of the next day. We should consider the trading time of the exchange market when we identify the date of news.