4. EMPIRICAL RESULTS
4.3 REGRESSION ANALYSIS
In this section, the paper presents three regression results of three hypotheses, and explains how financial news affects investor sentiment. Four discussions will be described respectively in the following section.
4.3.1 Number of News Releases and TWVIX
The first hypothesis is to test the effect of importance of financial news on TWVIX. Two major independent variable of this issue is analyzed in equation (1).
Table 4-3-1 gives the regression result of equation (1). From the analysis, this study finds that NEWSNUM has a significant positive association with VIX as expected (the coefficient of NEWSNUM is 0.0337, p< 0.01). This suggests that the number of news released by media of a day motivates the level of investor sentiment.
When investor receives more information, which is financial news, there is a higher level of news coverage highlighting the importance of financial news and people are more likely to generate psychological react to news that increases the TWVIX.
Some control variables also have significant effect on VIX. TWSTOCK is negatively associated with investor sentiment (the coefficient is -0.0026, p< 0.01), which consistent with prior literatures3. Trading value of all securities and investor sentiment is positive related (the coefficient is 3.1888, p< 0.01), which indicates that prosperity of trading activity increases the level investor sentiment. INFL has positive effect and IPI has negative effect (the coefficient is 29.4397 and -0.0426, p<0.05 and p< 0.01), which is clearly that with better environment of macro economy, investors are pleased. But CCI has a positive effect and UEMP has negative effect on VIX, which is not as expected (the coefficient is 0.0328 and -1.882, p<0.05 and p< 0.01).
3 Fleming et al. (1995) study of VIX change and S&P 100 index returns. Fernandes et al. (2014) study of VIX and S&P 500 index returns.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
55
4.3.2 Length of a News Article and TWVIX
Another aspect of importance of news in this study is the length of news. From the table 4-3-1, TLHIT has a negative association with VIX (the coefficient is -10.473, p< 0.01), which is not consistent with expectation. The regression result shows that the more characters per news come with the lower TWVIX. The H2 is rejected. The explanation is that the longer length of one article may not actually convey more valuable messages proportionally. And too many characters in an article dilute the strength of sentimental expression. As a result, there is negative relation between TWVIX and the length of a financial news article.
‧
Table 4-3-1 the Association between TWVIX and Importance of Financial News Regressand: VIX
Variable Expected Sign Coefficient Std. Error
Intercept
40.1526***This table presents estimation results for equation (1), which is used to test hypotheses regarding the association between TWVIX and importance of financial news.
t-Statistics are presented in parenthesis.
*** Significance at the 0.01, ** Significance at the 0.05,* Significance at the 0.10 levels, respectively.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
57
4.3.3 Word Usage of News and TWVIX
Considering the effect of content of financial news, this study employs the hypothesis 3.a and 3.b to investigate the influence of positive and negative word on investor sentiment, respectively. The regression analysis result of equation (2) is given by the table 4-3-2. First, this research finds that the quantity of positive wording negatively affects the level of sentiment as expected (coefficient of PW is -0.0290, p<0.01). It can be interpret that if media use less positive words in financial news, investor are pacified and sooth the level of nervousness to the financial market.
Additionally, the regression result of negative words also shows a significant connection to TWVIX (the coefficient of NW is 0.0320, p<0.01). This indicates that using more negative words in financial news can provoke the level of investor sentiment. Comparing the coefficient of these two variables, this study also finds that negative words pose more impact than positive words. When the number of negative words per day is increased by 1, TWVIX will climb by 0.0320 contemporaneously.
But the same case of positive words only helps bring down TWVIX by 0.0290.
Investors are more sensitive to the negative terms than the positive.
Above these findings of the regression in equation (2), the TWVIX is significantly influenced by the words usage. And this study suggests that media can influence the level of investor sentiment through controlling the usage of sentimental words in financial news. The remainder of control variables in this model has been discussed in section 4.3.1. Their results are identical with equation (1), except IPI. It shows no significant relation with TWVIX in equation (2).
‧
Table 4-3-2. The Association between TWVIX and Words Usage in Financial News
Regressand: VIX
Variable Expected Sign Coefficient Std. Error
Intercept
-88.4380***This table presents estimation results for equation (2), which is used to test hypotheses regarding the association between TWVIX and words usage in financial news.
t-Statistics are presented in parenthesis.
*** Significance at the 0.01, ** Significance at the 0.05,* Significance at the 0.10 levels, respectively.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
59
4.3.4 Tone of Financial News and TWVIX
In this section, this study investigates the relation between the TWVIX and tone of financial news.
The study uses the measurement of tone by 𝑇𝑂𝑁𝐸 = (𝑃𝑜𝑠𝑖𝑡𝑖𝑣𝑒 𝑊𝑜𝑟𝑑𝑠 − 𝑁𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝑊𝑜𝑟𝑑𝑠)/(𝑃𝑜𝑠𝑖𝑡𝑖𝑣𝑒 𝑊𝑜𝑟𝑑𝑠 + 𝑁𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝑊𝑜𝑟𝑑𝑠) as a variable in equation (3). Furthermore, the model adds NEWSNUM and TLHIT, which is proved their relationships with the degree of investor sentiment, as control variables to control their effects on VIX.
The regression result of equation (3) is shown by table 4-3-3. The coefficient of
TONE is -9.4685 and the p-value is under 0.01.This demonstrates that the scale of
tone in financial news has a significant negative effect on VIX. The greater value of TONE means financial news is closer to optimism, so the negative relation to investor sentiment shows that optimistic opinion of financial news can decrease the degree of investor sentiment, otherwise pessimistic opinion increase the TWVIX.This important finding suggests there is evidence proved that investor on financial market actually concerns the financial news, and generates mentally changing after reading the news. The direction of tone has impact on investor sentimental status.
Once the financial news released today is full of pessimistic tone, there will be disturbance by news opinion raising the TWVIX, which measures the degree of investor sentiment.
Other variables in this model show the empirical results similar to previous models except CCI. The unemployment rate turns out to be insignificant in this model.
‧
Table 4-3-3. The Association between TWVIX and Tone of Financial News Regressand: VIX
Variable Expected Sign Coefficient Std. Error
Intercept
-77.0457***This table presents estimation results for equation (3), which is used to test hypotheses regarding the association between TWVIX and tone of financial news by press media.
t-Statistics are presented in parenthesis.
*** Significance at the 0.01, ** Significance at the 0.05,* Significance at the 0.10 levels, respectively.
‧
5.1 RESEARCH DISCUSSION AND CONTRIBUTION
This study investigates the relationship between TWVIX and financial news in Taiwan from Dec. 2006 to Dec. 2014. We use TWVIX, which is called a fear index as a proxy for investor sentiment in Taiwan. This study includes dictionary approach and text analysis in order to examine the content of financial news to see their effects.
To collect the test bed news, financial news is obtained from KMW data base and is filtered by financial related key words we set. By following the dictionary approach of Carretta et al. (2011), Davis et al. (2012) and Garcia (2013), this study employs a dictionary approach of building up words lists with two categories of positive words and negative words. The text analysis method is referred to Carretta et al. (2011), Davis et al. (2012) and Huang et al. (2013) employing a formula based on the counts of words for measuring tone of financial news.
From the empirical results, this research finds that the number of news released by media of a day has positive effect on TWVIX. It suggests that with more information provided by media, investor is more likely to be influenced and increases the degree of their sentiment. Besides, the length of financial news has a negative relation with TWVIX, which is not consistent to the expectation. The reason perhaps is that long articles with too many characters dilute the strength of sentimental expression.
Secondly, this study finds that the usage of words also has significant impacts on the degree of investor sentiment. The number of positive words of financial news on the same date decreases TWVIX. On the contrary, negative words decrease the TWVIX. So this study suggests that media can affect the level of investor sentiment through controlling the usage of sentimental words in financial news
Lastly, the study inspects the tone of financial news, which also can be referred