Brand Loyalty Analysis of Mainstream News Channels

Influence of candidate Trump’s tweets on stock prices of news channels’ holding companies

The 2016 Election marks one of the first elections in which social media plays an important role in changing the political discourse and impacting election results. Notable within last election cycle was candidate Trump extensive uses of the social network Twitter to express his opinions about various major news channels, either applauding or questioning the legitimacy of their integrity and quality. In this article, using Twitter data, trading statistics, and sentiment analysis, we seek to determine if his polarizing opinions have any impact on the monetary values of the major news channels CNN and Fox News. We investigate whether the sentiments in general tweets and candidate Trump’s tweets from 7/15/2015, when candidate Trump entered the election, to 11/09/2016, when candidate Trump won the election, as well as their popularities, have any correlations to the stock trading statistics of the holding companies of three major media companies: Fox Broadcasting Company, Time Warner Cable, and MSNBC.

Data Description


We used four datasets for our analysis: candidate Trump’s tweets, Twitter sentiment as a whole (from Hedonometer.org), a daily sample of tweets about specific topics, and several media companies’ stock trading data during the 2016 election period. We focused primarily on the media companies CNN and Fox News. We originally considered MSNBC, but determined that it did not have enough data to be significant.

We investigated Trump’s Twitter content, the time he tweeted and the attention and reactions he received in Figures 1 and 2.

Figure 1. Breakdown of Trump’s Tweets by Posting Time and Company

Figure 2. Tone about Different News Channel Before and After Presumptive Nomination — The percentage of negative words in candidate Trump’s tweets about the two news channels.

The data are separated by before and after the presumptive nomination, as well as by mentioned news channel.

Before his nomination, candidate Trump’s tweets about Fox News were slightly more negative than his tweets about CNN. After his presumptive nomination, however, candidate Trump’s tweet about CNN became considerably more negative (over 4 times higher than that from Fox). It’s very interesting to see this drastic attitude change of candidate Trump before and after nomination.

Figure 3. Tweet Content and Popularity over Time

Figure 3 describes the average and total retweets candidate Trump received throughout the period leading up to the 2016 Election. It is notable that after candidate Trump’s presumptive nomination as the Republican candidate, both his daily average retweets and daily total retweets started to rise at a much higher rate. In addition, after his nomination, the average numbers of retweets of tweets about CNN rose considerably at a larger pace and exceed those about Fox news. In other words, after his nomination, his tweets about CNN received more responses than those about Fox News.

The second dataset we looked at was the trading data of the holding companies of CNN and Fox News during the election period, including their trading volume, price changes, and daily price spreads.

Figure 4. Time Warner Inc stock price changes

Over the course of the election, CNN’s holding company experienced frequent price changes—the largest being a $7.85 intraday increase. Interestingly, these price changes smoothed out dramatically after the presidential election. This could have been due to uncertainty about the outcome, support/ criticism of CNN’s coverage, or other factors.

Figure 5.  Twenty-First Century Fox Inc stock price changes

Unlike with CNN, Fox News’ holding companies’ intraday price change do not seem to have decreased after the election. Fox’s overnight price variation has decreased, however. This is consistent with the fact that the majority of candidate Trump’s tweets occur in non-trading hours, although these tweets did not cease after the election.

The third dataset is the general tweets about the two news channels and candidate Trump. We used LIWC to access the sentiments in tweets we acquired by scraping the top tweet pages about relevant topics on a daily basis.  

Figure 6. Text Sentiment Analysis distribution

As we can see from Figure 6, all kinds of sentiment were more spread out compared to the average happiness on Twitter measured by Hedonometer. It can be explained by our small sample size and the controversial topics brought around candidate Trump.  From the last row graphs, we can see that anger is measured in a larger scale compared to the other two negative emotions, which can be an indicator of the general emotion against candidate Trump and mainstream news channels involved.

Figure 7. Average negative emotions on Twitter over the election period

Clearly, the tone regarding CNN didn’t coincide with the tones around two presidential candidates – tones around two candidates fluctuated more and the most spikes were seen around the topic of Trump and CNN.  Interestingly, we see a decrease in negative emotion in tweets around CNN but increase in the ones around both candidates approaching the election date. On the other hand, the anger in tweets on Trump and Fox decreased during the same time period. This observation coincides with the switched tones in Trump’s tweets about the two news channels.

Analysis


We have three main questions to drive our analysis:

How does the atmosphere on Twitter (regarding certain keywords) correlate to changes in the media companies’ stocks?

We found that in a sample tweets from the runup to the election, when people when people were sadder about Hillary and CNN, CNN’s stock price was more volatile (measured by high-low price spread). When people were sadder about candidate Trump and Fox News, Fox News’ stock price was also more volatile. These both imply that when the Twitter community feels sadder about more liberal things, CNN’s stock is more volatile; and similarly, when the Twitter community feels sadder about more conservative things, Fox News’ stock is more volatile. These two observations show that a sad atmosphere on twitter is correlated with changes in the respective more liberal or conservative media companies’ stock price.

Figure 8 pairs: Correlation between general tweeter sadness and stock prices volatilities measured by daily price spreads

How does the negative emotion in candidate Trump’s tweets correlate to changes in the media companies’ stocks? 

Negative (overnight) tweets about CNN, correlate to overnight price increases in CNN’s stock, whereas positive (overnight) tweets about Fox News, correlate to overnight price increases in Fox News’ stock. These effects could possibly be explained by the psychological principles of ingroup favoritism and prejudice. Ingroup favoritism is “the tendency to respond more positively to people from our ingroups than we do to people from outgroups” (opentextbc.ca/) and “prejudice is an affective feeling towards a person or group member based solely on that person’s group membership” (wikipedia.org/). Therefore, if candidate Trump’s tweets more negatively about CNN, then it should be expected that CNN’s viewers will oppose candidate Trump’s opinion since CNN’s primary audience is liberal and candidate Trump is conservative. Inversely, if candidate Trump tweets more positively about Fox News, then it should be expected that Fox News’ viewers will approve of candidate Trump’s opinion since Fox News’ primary audience is conservative and candidate Trump is conservative.

Figure 9: Correlation in media’s stock price and the emotion of candidate Trump’s tweets

How does the popularity of candidate Trump’s tweets correlate to changes in the media companies’ stocks?

Even though candidate Trump tweeted more about Fox News than CNN, his tweets about CNN were more retweeted (both on average and in absolute value). There is a slight positive correlation between the responses to candidate Trump’s tweets about CNN and CNN’s overnight stock price change. Even though the majority of candidate Trump’s tweets about CNN are negative, it seems that Trump tweeting about CNN may actually help CNN’s stock price. There is an essentially zero correlation, however, between the responses to candidate Trump’s tweets about Fox News and Fox News’ overnight stock price change. This implies that candidate Trump’s tweets impacted CNN more than Fox News even though candidate Trump was more likely to be positive toward Fox News than CNN.

Figure 10: Correlation between stock and popularity of candidate Trump’s tweets

Conclusion


From our findings, we are able to answer our three main questions. Firstly, sadness on Twitter about liberal or conservative issues is correlated with increasing price volatility in the respective media companies’ stock. Secondly, the negativity of candidate Trump’s tweets concerning CNN positively correlates to overnight price increases in CNN’s stock. Inversely, the positivity of candidate Trump’s tweets concerning Fox News positively correlates to overnight price increases in Fox news’ stock. Lastly, we found that candidate Trump’s more retweeted tweets correlate with more positive overnight price changes, whereas the correlation was zero for Fox News.

Drawbacks


Our research faces several drawbacks. Foremost, CNN and Fox News do not have individual stock tickers. Instead, their stock is part of the parent companies Time Warner (TWX) and 21 Century Fox (FOX), respectively. This means that we were unable to prove causation (we instead show correlation). Thus, price changes could be due to other events that does not relate to candidate Trump’s tweets. Furthermore, two-thirds of candidate Trump’s tweets occurred during after-hours and therefore, we could not measure immediate price changes as data on pre-market changes are less readily available. Secondly, our sentiment analysis software had difficulty parsing candidate Trump’s tweets. For example, it believes that “I won every poll from last night’s Presidential Debate – except for the little-watched @CNN poll.”, is a positively-toned Tweet towards CNN.

Additional Questions


We believe there is more research that could be done in the area of how politicized social media impacts media companies. One area is how viewership (size, demographics, time, etc) changed during the course and over the advent of candidate Trump’s tweets. We are unable to access this data from Nielsen due to their prohibitive pricing. Another question is whether different demographics respond to candidate Trump’s tweets differently. A third question is whether other political candidates’ (domestic or foreign) social media activities have similar effects, such as Justin Trudeau or Emmanuel Macron.


Research, writing, and infographics by Trang Ngo, Jacob Ryan, Astrid Weng and Joseph Xiong

Leave a Reply

Your email address will not be published. Required fields are marked *