Content uploaded by Marc Sardy
Author content
All content in this area was uploaded by Marc Sardy on Jun 18, 2019
Content may be subject to copyright.
How Presidential Tweets Affect Financial Markets
Sasha Cove, Rollins College & Dr. Marc Sardy, Rollins College
Abstract
This paper examines the effect of President Trump’s Tweets and their impact on financial markets as measured
by the Market proxy of the S&P 500. Data has been collected on Trumps tweets since his presidency began until
fall 2018, Market data is based on daily S&P results for the daily returns. Our findings are that Trumps tweets
have an impact on financial markets even at the level of daily returns.
Introduction
The manner in which information is disseminated to the masses has evolved with the introduction of social
media platforms. Individuals are now receiving their daily news in the form of Facebook updates, Instagram posts,
and Twitter ‘tweets.' The average Twitter user can see around 1,000 tweets in the fourteen minutes that they spend
on Twitter in a day, meaning that users often spend only seconds on a given tweet (Pancer, E., & Poole, M. 2016).
Twitter, in particular, came to the foreground during the 2016 presidential election, as candidates used the platform
to increase their media coverage and garner public support. This method proved extremely successful for the
current president of the United States, Donald Trump, who achieved unprecedented media coverage through his
Twitter activity. His current position gives his tweets the power to influence public sentiment, and thus the
performance of the financial markets.
Social media use is becoming integral to politics, not only as a method of keeping up with political news, but
also as a mechanism through which individuals’ political views and involvement are being impacted. A survey
conducted by the Pew Research Center found that 25% of social networking site (SNS) users said that the sites
are ‘very important’ or ‘somewhat important’ to them in discussing or debating political issues with others, 25%
of users say that they have become more active in a political issue after discussing or reading about it online, and
that 16% of users say that they have changed their views about a political issue after discussing or reading about
it online (Rainie, L., & Smith, A. 2014 ). Furthermore, Social media’s use as a media outlet is will continue to
grow in importance in coming years, as social media is most used by the younger generations. As this generation
ages, they are beginning to comprise a growing portion of the voting population, indicating that social media will
maintain its relevance in politics. The use of social media in politics has become a method through which the
younger generation can be engaged; there is a strong positive correlation between social media use and youth
political engagement in users between the ages of 16 and 29 (Michael Xenos, Ariadne Vromen & Brian D. Loader
2014).
The emergence of the new hybrid media system has changed the ‘currency' of the media system as a whole. It
is no longer the ability to communicate to mass audiences, but rather the ability to attract attention from them in
a world where the average individual is constantly inundated with massive amounts of information on a day to
day basis. In our new ‘attention economy,’ media organizations and platform compete for views, as do the issues,
actors, causes, and factions that comprise our political sphere (Zhang, Y., Wells, C., Wang, S., & Rohe, K. 2018).
According to Zhang et al. in “Attention and amplification in the hybrid media system: the composition and activity
of Donald Trump’s Twitter following during the 2016 presidential election," attention's power is derived from its
four characteristics. It is necessary to change or mobilize the opinions of an audience, as it gives an actor access
to the ‘socialized communication' that lends an audience to its depiction. It can be a mechanism for translating
communication’s power into action; politically, it can be converted into civic engagement. It is a transferrable
currency, as the holder of attention has the power to direct said attention to another. Finally, it is powerful because
others perceive it as powerful; it holds a ‘popularity bias,’ as attention tends to beget attention. Zhang et al. also
introduced the idea of social media amplification, which are the actions of social media users that increase
measures of engagement or audience metrics, the likes, @replies, follower number, and retweets a Twitter account
receives. Twitter’s software uses these measures to engagement to determine what is ‘trending,’ what to feature
on the platform’s homepage and what will appear at the top of a user’s newsfeed. In this sense, the attention given
to a tweet is exponential; with each retweet the tweet is disseminated to the followers of the account that retweeted
it, causing more retweets, more likes, more attention.
The amount of attention an individual receives on social media has become an indicator of social status, of
‘worthiness,’ and this ‘worthiness’ is often translated to news-worthiness (Zhang, Y., Wells, C., Wang, S., &
Rohe, K. 2018). As a tweet begins to ‘trend,’ news platforms feel required to cover it, if not only to beat their
competitors to the punch. As such, the tweets of socially relevant individuals have been increasingly covered in
the news, providing the tweets with validity, and further increasing the attention the tweets receive.
Social media activity levels can be predictive of stock performance. A study researched the correlation
between intra-day responses of the stock market and the spread of news on Twitter. Unusually high spikes in
Twitter activity regarding 96 firms on the NASDAQ over 193 trading days were statistically compared with
Yahoo! Finance data of the related firms over the forty minutes following the spike. The study found a spike in
tweets per minute resulted in a spike in trading activity in the following forty minutes (Tafti, A., Zotti, R., & Jank,
W. 2016). For an individual, tweeting is fast and effortless but has no immediate consequences on financial
markets. However, if said individual is the president, the act of tweeting remains fast and effortless but has the
potential to cause immediate financial consequences. It becomes less necessary that there be a spike in activity
because the amount of people that are viewing presidential tweets is significantly larger, as is the content of the
tweet’s potential implications on the markets.
The concept that not everyone on Twitter holds the same influence on market sentiment is explored in a study
examining the influence of a financial community, a group of users whose interests align with the financial
markets. (Yang, S., Mo, S., & Liu, A. 2015) The community comprised a central group of 50 well-recognized
investment experts, 25 influential traders and 25 accounts associated with the seven financial news providers, and
their followers. By determining the sentiment of the central group’s tweets, disseminated by their followers, a
robust correlation was determined between the group’s sentiment and financial market movement measured by
lagged daily prices ((Yang, S., Mo, S., & Liu, A. 2015). It can be argued that because users are harvesting
information from these influential sources in order to make their daily trading decisions, the financial community
was exerting influence on the financial markets. It can be rationally inferred that the influence on public sentiment
that the financial community had can be compared to the influence that Trump’s Twitter has as a single central
group, amplified by both his supporters and opposers, paired with the resulting media coverage of his tweets.
As President, Trump’s tweets not only garner massive attention, but also have the potential to have significant
global implications. If an average individual were to tweet that increasing tariffs would be good for the country,
the tweet would have no relevance to the markets. However, if Trump were to tweet the a similar statement, there
is a significant chance that tariffs may in fact be increased, which would have a significant impact to the markets.
We propose a correlation between when Trump tweets about the economy and movements in the financial
markets.
H1. Tweets posted by Trump that are related to the economy are positively correlated to movements in the
S&P 500 and the VIX.
Early research on predicting financial markets was centered around the efficient market hypothesis (EMH),
which argues that financial market valuations incorporate all existing, new, and hidden information because
investors act as rational agents who seek to maximize profits (Mao, H., Counts, S., & Bollen, J. 2011). The
collective investment decisions of investors use the wisdom of crowds to determine stock market prices, driven
mainly by new information, such as news, rather than present and past prices. EMH has been challenged by
behavioural finance, as research has shown that stock prices do not follow a random walk and can be predicted to
an extent using early indicators such as Twitter feeds (Bollen, J., Mao, H., & Zeng, X. 2010). EMH is further
challenged by behavioural finance research which emphasizes the role of behavioural and emotional factors in
financial decision making (Mao, H., Counts, S., & Bollen, J. 2011).
Studies have determined that public mood state can be tracked from the content of large-scale Twitter feeds,
and that changes in public mood correlate to shifts in the Dow Jones Industrial Average (DJIA) values, which
occur with a lag of three to four days. (Bollen, J., Mao, H., & Zeng, X. 2010). The correlation between the positive
and negative mood of the masses on Twitter and stock market indices, including VIX, DJIA, S&P 500, and
NASDAQ have also been examined. Collective emotions were categorized by ‘mood words’ such as fear, worry,
and hope, and it was found that both positive and negative moods were positively correlated with all four indices
(Zhang, Fuehres, & Gloor. 2011). This result is relevant because VIX performance is strongly negatively
correlated with the DIJA, S&P 500, and NASDAQ, implying that people use more emotional words during times
of uncertainty, whether it is positive or negative in a financial context. The idea of public mood as a financial
market predictor can be furthered by looking at the sentiment toward specific stocks rather than general public
sentiment. A study tracked public sentiment towards 30 DIJA companies over a 15 month period and found a
strong statistical correlation between stock price returns and Twitter sentiment and volume of tweets regarding
the companies (Ranco, G., Aleksovski, D., Caldarelli, G., Grčar, M., & Mozetič, I. 2015).
Research has found that negatively worded tweets are liked and shared more often than positively worded
tweets, information which arguably furthered Trump’s media coverage during the election. A study conducted by
Pancer et al. found that during the 2016 election, posts relating to policy such as the economy, the environment,
or labour, decreased the post’s popularity by about 1,800 likes. Conversely, those concerning inflammatory
rhetoric increased popularity related to a 2,600 like increase, which they termed the content valence effect. The
study found that the popularity of Trump’s tweets was improved by mentioning disasters and religion, but
decreased when mentioning social issues, and that the content valence effect was present for Trump, but not for
Clinton during the election (Pancer, E., & Poole, M. 2016). This had a corruptive effect on the content being
tweeted at the time, an effect which has extended into the presidential term.
Reaching a larger audience is made possible by posting socially inflammatory content, and this has led to
tweets which have the power to cause shocks in the global financial markets. We propose a correlation between
the sentiment of Trump tweets relating to the economy and movements in the financial markets.
H2. The sentiment of Tweets posted by Trump that are related to the economy is positively
correlated to movements in the S&P 500 and the VIX.
Data
Our analysis is conducted on the six month time period of July 1, 2018 to December 31, 2018. In the
analysis we investigate the relationship between market data and Twitter data. The details of both are given in
the remainder of this section. The market and twitter data is available at
https://doi.org/10.6084/m9.figshare.7679363.v1.
Market data
The first source of data contains information on the performance of the financial markets, utilizing two
stock market indices, the S&P 500 and the CBOE Volatility Index (VIX), indicators of financial market
performance. We compiled the day-to-day data of both indices, each day’s open, close, and percent change, for
the six month period. This data is publicly available and can be downloaded for various Internet sources, such as
the Yahoo! Finance website (i.e. https://finance.yahoo.com/quote/%5EGSPC?p=^GSPC for the S&P 500). As
the markets are closed on the weekends, and Trump tweeted over the weekends, Saturdays and Sundays were
counted as a cumulative ‘day,’ with their market open being that Friday’s close, and their market close being
that Monday’s open; the percent change over the weekend was then calculated. The same method was utilized
for the five Market holidays that occurred during the sample time period, using the close of the day before and
the open of the day after in order to calculate the percent change over the holiday.
Twitter data
The second source of data is from Twitter, and consists of the Tweets posted by Donald Trump from
his personal twitter account @realDonaldTrump, and the sentiment of each Tweet. The data was collected from
Export Tweet (https://www.exporttweet.com/), and contained the date, content, retweets, and likes. In the 6
month period of study, Trump tweeted a total of 1,989 times. Saturdays and Sunday tweets were combined into
one ‘day’ to account for the lack of weekend market data. First, we determined total amount of tweets made per
day. Then, each tweet determined to be related to the economy was identified, a total of 284 economy-related
tweets, and the amount of economy-related tweets made per day was determined. The percent of tweets per day
that were economy-related was then calculated.
A Tweet sentiment was determined for each of the economy-related tweets. Sentiment was evaluated
according to the that of the author, Trump; each tweet was assigned one of three sentiment labels, positive,
neutral, or negative, determined by whether he had worded it in a positive, neutral, or negative tone. The daily
tweet sentiment was then calculated by assigning each sentiment label with a score; positive: +1, neutral: 0,
negative -1, and summing the sentiment of each economy-related tweet of the day. The more positive tweets
made in a day, the higher the score, and the more negative the lower; neutral statements had no effect on the
day’s cumulative sentiment score.
Methodology
Regression Analysis
We performed two regression analyses in order to determine the relationship between market
performance and Trump’s economy-related tweets.
Figure 1. Correlation between S&P performance and the % of economy related
tweets
Figure 2. Correlation between VIX performance and the % of economy related tweets
Sentiment Analysis
In order to examine how sentiment was viewed we used a measure of S&P movements and VIX
movements :
Table 1. The net impact of daily tweet sentiment on S&P 500 movement
Sentiment Total S&P Up Days S&P Down Days % Up % Down
Positive 75 42 33 56.00% 44.00%
Negative 20 9 11 45.00% 55.00%
Neutral 9 2 7 22.22% 77.78%
y = -0.0067x + 0.0013
R² = 0.0282
-4.0000%
-3.0000%
-2.0000%
-1.0000%
0.0000%
1.0000%
2.0000%
3.0000%
0.00% 20.00% 40.00% 60.00% 80.00% 100.00% 120.00%
S&P Performance
% Economy Related Tweets
y = 0.0594x - 0.0122
R² = 0.0322
-15.0000%
-10.0000%
-5.0000%
0.0000%
5.0000%
10.0000%
15.0000%
20.0000%
25.0000%
0.00% 20.00% 40.00% 60.00% 80.00% 100.00% 120.00%
VIX Performance
% Economy Related Tweets
Table 2. The net impact of daily tweet sentiment on VIX movement
Sentiment Total VIX Up Days VIX Down Days % Up % Down
Positive 75 34 41 45.33% 54.67%
Negative 20 9 11 45.00% 55.00%
Neutral 9 4 5 44.44% 55.56%
Results
Market moves do not seem to be independent of Trumps comments with bot the S&P Moves and VIX moves
showing low significance with regard to independence. So there seems to be some effects that are significant as
shown in our regression analysis of tweets vs S&P and VIX. The equation of S&P moves vs Trumps tweets about
the economy is:
S&P (move)= 0.013-.0067* (Tweets) with an R2 =.0282 (1)
While the explanatory power for the tweets are low there is a significant but negative effect on Trump tweets and
the effects on the S&P 500. There are several reasons as to why the results may not be very large. Firstly, we were
using daily returns and the instantaneous tweet information. There are a lot of things that happen on a daily basis
that can have an impact on the S&P. The effects of tweets can be diffused over the course of the day. Future study
will look at more accurate timing of market moves with regard to the timing of the tweets. Furthermore, it seems
that every time trump tweets up about the economy the S&P has a 56% change of positive movement and a 45%
chance of the S&P declining. With regard to negative Tweets, there is a 55% (10% margin of error) chance of a
down move in the S&P and a 45% (10% margin of error) of an up move.
The VIX data regression shows similar results with the VIX moving positively with Trump’s Economic tweets.
VIX (Move)= -0.0122+0.0594 (tweet) With an R2 of 0.0322 (2)
While the explanatory power for the tweets with respect to the VIX are low there is a significant but positive
effect of Trump tweets and the effects on the VIX. There are several reasons as to why the results may not be very
large. Firstly, once again we were using daily returns and the instantaneous tweet information. There are a lot of
things that happen on a daily basis that can have an impact on the VIX. The effects of tweets can be diffused over
the course of the day. Future study will look at more accurate timing of market moves with regard to the timing
of the tweets. Furthermore, it seems that every time trump tweets up about the economy the VIX has a 45% (10%
margin of error) chance of positive movement and a 55% (10% margin of error) chance of the VIX declining.
With regard to negative Tweets, there is a 55% chance of a down move in the VIX and a 45% (10% margin of
error) of an up move.
Discussion/Conclusion
Much of our findings are limited by the nature of daily S&P 500 returns. There is a gross mismatch between the
timing of the tweets and the daily results. If we had more frequent S&P results we would be more directly able to
see the instantaneous effect of the tweets on the market. That said, we do actually see an impact of tweets on the
market but it is considerably smaller then we might see with more micro-structured data.
Our results showed that the tweets related to economic issues have an impact on the market. Up tweets raise the
market 56% ±10% of the time and down tweets move the market 55%±10%. In addition to moves in the S&P, the
VIX is also affected by the tweets. The VIX is up when President Trump tweets about the economy. The market
Volatility goes up by 45%±10%. when President Trump tweets up about the economy. The market volatility also
changes when President Trump tweets down about the economy The VIX potentially moves down by 55%±10%.
or up by 45%±10%. Overall there seems to be a direct impact of tweets and market moves that are significant and
impactful.
References
Bollen, Johan ; Mao, Huina ; Zeng, Xiao-Jun, (2011), Twitter mood predicts the stock market, Journal of
Computational Science, 2(1), March 2011, Pages 1-8
Himelboim, Itai ; Smith, Marc A ; Rainie, Lee ; Shneiderman, Ben ; Espina, Camila, (2014) Classifying
Twitter Topic-Networks Using Social Network Analysis Social Media + Society, February Vol.3(1)
Mao, Huina ; Counts, Scott ; Bollen, Johan , (2011),Predicting Financial Markets: Comparing Survey, News,
Twitter and Search Engine Data, IDEAS Working Paper Series from RePEc,
Pancer, Ethan ; Poole, Maxwell, (2016) The popularity and virality of political social media: hashtags,
mentions, and links predict likes and retweets of 2016 U.S. presidential nominees’ tweets, Social Influence, 01
October Vol.11(4), p.259-270
Rainie, L., & Smith, A. (2014, February 18). Politics on Social Networking Sites. Retrieved from
http://www.pewinternet.org/2012/09/04/politics-on-social-networking-sites/
Ranco, Gabriele ; Aleksovski, Darko ; Caldarelli, Guido ; Grčar, Miha ; Mozetič, Igor (2015),The Effects of
Twitter Sentiment on Stock Price Returns, PLoS One, Sept, Vol.10(9), p.e0138441
Tafti, Ali ; Zotti, Ryan ; Jank, Wolfgang, (2016) Real-Time Diffusion of Information on Twitter and the
Financial Markets.(Report) PLoS ONE, August 9, 2016, Vol.11(8), p.e0159226
Vromen, Ariadne ; Xenos, Michael A. ; Loader, Brian, (2014) Young people, social media and connective
action: from organisational maintenance to everyday political talk, Journal of Youth Studies, 01 July, p.1-21.
Yang, Steve Y. ; Mo, Sheung Yin Kevin ; Liu, Anqi, (2015), Twitter financial community sentiment and its
predictive relationship to stock market movement, Quantitative Finance, 03 October Vol.15(10), p.1637-1656
Zhang, Xue ; Fuehres, Hauke ; Gloor, Peter A,.(2011) Predicting Stock Market Indicators Through Twitter
"I hope it is not as bad as I fear", Procedia - Social and Behavioral Sciences, Vol.26, pp.55-62
Zhang, Yini ; Wells, Chris ; Wang, Song ; Rohe, Karl, (2018) “Attention and amplification in the hybrid
media system: The composition and activity of Donald Trump’s Twitter following during the 2016 presidential
election, New Media & Society, September 2018, Vol.20(9), pp.3161-3182.
HRef: https://doi.org/10.6084/m9.figshare.7679363.v1. (checked 5/20/19)
HRef: https://finance.yahoo.com/quote/%5EGSPC?p=^GSPC (checked 5/20/19)
Href: https://www.exporttweet.com/ (Checked 5/20/19)
