The impact of Donald Trump’s Twitter usage on the US Dollar/Chinese Yuan Renminbi exchange rate volatility

The impact of Donald Trump’s Twitter usage on the

US Dollar/Chinese Yuan Renminbi exchange rate volatility

Bachelor Thesis

Abstract:

This study explorers whether statements by Donald Trump–the 45th _{and current President of the}

United States–on Twitter influence the US dollar/Chinese Yuan Renminbi exchange rate volatility. By using data from Twitter and Dukascopy, a selection of tweets by Trump is collected and thoroughly examined on its possible influence on the USD–CNY exchange rate volatility. All of these Tweets somehow concern Trumps ideas and opinions on the current economic and political frictions between the United States and China. Results show, that after all tweets that are considered ‘’negative,’’ volatility in the exchange rate increases. For “positive” tweets, no significant results were found. Also, there is no clear evidence for correlation between the content of the tweet and the direction the exchange rate moves in.

Author: Tobias Vollaers

Student Number: 10986413

Supervisor: Rui Zhuo

Bachelor Program: Economics and Business

Specialization: Economics

2

Statement of originality

This document is written by Tobias Vollaers, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the content.

3

Table of contents

1.

Introduction ………...4

2.

Literature review……… 6

2.1 Fundamentals of exchange rate determination

..………6

2.2 News and unanticipated event

………..

..………...7

2.3 New media

………..………..8

3.

Data analysis………10

3.1 Negative tweets

………..10

3.2 Positive Tweets

………...12

3.2 Direction of the exchange rate

……….12

4.

Discussion. .……….13

……….13

……….14

………15

5.

Conclusion……….16

Appendix ..………...17

Bibliography.………20

4

1. Introduction

From January 1, 2018, to June 2, 2018, the exchange rate between the US dollar (USD) and the Chinese yuan (CNY), or renminbi (RMB), had an average daily volatility of 0.255%. Since the 1970s, research has been done to examine what factors determine these short-term exchange rate fluctuations. Several studies have shown that both news announcements and unanticipated events mainly influence investors’ expectations, resulting in short-term exchange rate fluctuations. With the dawn of social media, the way this news and new information is distributed has changed dramatically. United States President Donald Trump, for instance, has over 52 million followers on Twitter, a social media platform, and has sent over 37,000 tweets; his social media posts thus have enormous range and are spread over the internet immediately. Currently, he is frequently using Twitter to build a case against the US–Chinese trade agreements, which he believes are unfair to the United States.

This thesis examines whether Trump’s Twitter usage affects the USD–CNY exchange rate volatility. It consists of a literature review and a data analysis. In the literature review, firstly all major fundamental exchange-rate-based models are discussed. Secondly, the role of news and unanticipated events in exchange rates will be discussed, and finally, a more recent approach on what impact “new media“ has on economic markets will be analyzed. “New media’’, stands for mass communication, using digital technologies like the internet. In the data analysis, a selection of tweets posted by Trump concerning the current US–China situation is thoroughly examined for its possible influence on short-term volatility in the USD–CNY exchange rate.

The internet and social media have changed the way information is gathered in the 21st century. New information is both more accessible and faster spread. Within these developments, Twitter—a social media platform on which users can publicly share small bits of information or messages—has developed itself into one of the foremost new media platforms. Users share this information through posts called “tweets.” Twitter currently has over 336 million active monthly users (Twitter - Statistics & Facts, 2018). To illustrate the relevance of Twitter, on the presidential election day of the United States in 2016, over 40 million election-related tweets were sent. As stated above, Trump himself has over 52 million followers: When Trump sends a tweet, over 52 million users receive this tweet on their timeline. Next to these 52 million people who get a notification on their timeline, anyone with a working internet connection can easily observe his tweets by visiting Twitter.

What happened on April 23, 2013, exemplifies the kind of influence Twitter and

unanticipated new information have on economic markets. The Associated Press, an American news agency, falsely reported on Twitter that President Barack Obama was injured in two explosions at the White House. In the minutes after this happened, the Dow Jones dropped 143.5 points, and the Stand & Poor's 500 index lost over US$ 136 billion. Once this news was revealed to be false, the market corrected itself almost immediately. Yet in this moment, for one of the first times, the influence of social media on economic markets became clear ("How Does President Trump’s Twitter Use Impact Forex, Markets And Stocks?", n.d).

Since Trump took office, he has repeatedly expressed his views and disbeliefs on the US– China trade agreements. On January 22, the US announced new tariffs on washing machines and solar cells, after which China announced “anti-subsidy” measures against US Sorghum. When the US then announced import tariffs of 25% on steel and 10% on aluminum, China reacted with similar

5 measures. This escalation of tariffs is considered the start of the so called “trade war” between China and the United States (Moyer, 2018). Since then, Trump has been frequently tweeting his views and opinions on China and the ongoing developments in trade between the two nations. In the literature review, several studies are presented which argue that news and

unanticipated events are both significant determinants for exchange rates. Other studies suggest that Twitter can be used to explain and forecast economic processes. The data analysis shows that following all Trump tweets considered “negative,” fluctuation increased. For the “positive” tweets, by contrast, only after half of the researched tweets did the USD–CNY exchange rate show greater fluctuation than would normally be expected. There seems to be no relation between the content of the positive tweet and the movement in the exchange rate.

This thesis is organized as follows. Section 2 presents a literature review. Section 3 consists of a data analysis of the data found on Twitter and Dukaskopy. Section 4 provides a discussion of the data, and section 5 outlines the conclusion of this thesis.

6

2.Literature review

2.1 Fundamentals of exchange rate determination

In 1973, after the Bretton Woods system of fixed exchange rates was abandoned, most

industrialized economies started floating their exchange rates again. This return to floating rates provided economists with information to empirically model exchange rate markets (Dornbusch, 1982). The most standard model is the flexible-price monetary model in which the exchange rate is determined by the relative price of domestic and foreign money, which is determined by demand and supply for the concerned currencies (Frenkel, 1976). This model assumes that “Purchasing power parity” holds, which implies the exchange rate equals a value that causes a basket of goods to be equally priced in two different countries. Other well-known models include Dornbusch’s (1976) sticky price monetary model and Frankel’s (1979) real interest rate differential model. Dornbusch did not assume for purchasing power parity to hold and considered expectations on future exchange rates to be a key determinant of current exchange rates, while Frenkel claimed that exchange rate movements are determined by differences in real interest rate levels between countries. Most of these models use different variables, with which they attempt to explain and predict exchange rate fluctuations. Despite the significant amount of theories and models on this subject, there is still no agreement on what the major factors are that can forecast exchange rate fluctuations most precisely. All these models have different assumptions and therefore in some way or another contradict each other.

Focusing on fundamentals on short term exchange rates, Fama (1965), argued with his efficient market hypothesis that financial markets are always efficient and its prices reflect all information known in the market. Therefore they rapidly adjust to any new relevant information. This information consists, among others, of future expectations of assets. Agents are considered rational in their decision making, what makes the market efficient. The Efficient market hypothesis is often linked with the random walk hypothesis (Meese and Rogoff, 1983). Meese and Rogoff stated exchange rates follow a random walk. Since new and future information is mostly unknown and unpredictable, and moves somehow randomly, movements in exchange rates also move randomly. Tomorrow’s rate change will only reflect tomorrows news and is completely independent of today’s rate. This theory seeks to explain the Random walk hypothesis. Meese and Rogoff demonstrated that none of the fundamental-based exchange rate models were able to forecast exchange rates more accurate than the random walk hypothesis. Most events that influence short term exchange rates happen random in some way, like accidents or news. Investors are considered rational and will react quickly to this new information, which leads to fluctuations. With the aid of technology and high speed computers, these fluctuations can be observed within minutes or even seconds. In the 21st century, the idea that the efficient market hypothesis and random walk hypothesis always hold is far less universal than it used to be. A strain of efficient market critics is called behavioral economics. It contradicts the efficient market hypothesis on the idea that agent’s beliefs and psychological states are decisive in economic decision making. These beliefs are shaped by agent's private information and therefore their own expectations, while they are bound by human limitations and complications. This implies agents do not necessarily have to be rational, what can create market inefficiencies (Scheifler,2000). This irrational behavior makes economics believe stock prices are partially predictable, at least as predictable as people’s reaction to new information is.

7 Often used arguments by behavioral economics in their critics on the efficient market hypothesis are for example the “internet bubble” and the “crash of 1987.” Events like these are clearly evidence of irrational behavior. This would imply markets cannot be efficient. Malkiel (2003) fights these critics by saying events like this are exceptions rather than rules. He believes markets are indeed never perfectly efficient, because in that case there would remain no incentives for investors to react to new information. Yet he believes that although markets in the very short term may not always be perfectly efficient, in the long run markets always converge to an efficient equilibrium. He stated that the idea that markets are almost fully efficient will never be abandoned. “If any $100 bills are lying around the stock exchanges of the world, they will not be there for long.”

Gyntelberg, Loretan, Subhanij and Chan, (2009) investigated the effect of this “private information” on exchange rate markets. By doing empirical research, they argue private information regarding stock markets helps to explain exchange rate fluctuations. All investors have different ways in how they collect their information, and interpreted this information in dissimilar ways. By using daily-frequency data for several financial markets in Thailand, they found significant evidence this private information affects exchange rates.

2.2 News and unanticipated events

Assuming that expectations have a direct influence on exchange rates, divergent research has been done on what impact news can have on these expectations and thus exchange rates. In 1980, Dornbusch developed the rational expectations model, by which he tested for “news form.” Assuming asset markets are efficient, all available information is reflected in exchange rates. Therefore, all deviations in exchange rates are completely due to news. In 1981, Frenkel wrote another paper in which he examined the volatility of exchange rates. He distinguished between anticipated and unanticipated events. His paper showed that “news” is a key factor in affecting exchange rates. In his research, he assumed the assets market clears almost directly and that unexpected events (like news) are almost immediately reflected in exchange rates. The coefficients with which he tried to measure the news were mostly significant. This finding is in line with the theory that current exchange rates reflect all expectations on future exchange rates; these

expectations thus arise from new information (Frenkel, 1982). Similar results from Edwards (1982) and Macdonald (1983) have provided support to Frenkel’s outcomes, and Bomhoff and Korteweg (1983) found that on six major European currencies, 50% of the variation in unanticipated

fluctuations on spot rates can be attributed to news factors.

In addition, Goodhart, Hall, Henry and Pesaran (1993) used an extreme high-frequency data set to investigate the effect of news events on the dollar–sterling exchange rate. They recorded 130,000 observations in an eight-week period. Initially, when they left out the effect of news, they found a fairly consistent variance of the exchange rate. When they altered the effect of news, their results changed dramatically, prompting them to conclude that changes in exchange rates are far from random. They also found that these fluctuations in short-term exchange rates are usually impermanent. In their research, the level and volatility of the exchange rates returned to their “pre-news” value rather quickly.

Frankel and Rose (1994) have emphasized the announcement effect of news. Using announcement data, one can clearly isolate the impact of that announcement by examining exchange rate data immediately before and after an announcement. Empirical results show these exchange rate effects typically disappear when the changes are measured on daily intervals.

8 Therefore, using announcement data usually has an advantage over longer-term data in empirical research on the effect of news on exchange rates.

Ito and Roley (1987) have examined how macroeconomic announcements in both Japan and the United States influenced the yen–USD rate. They consider in which country the

announcements had the strongest impact, finding that US economic announcements had

significantly more effect than Japanese economic announcements. Beck (1993), furthermore, has found that whenever the US government made announcements of a large budget deficit, the dollar appreciated against foreign currencies.

2.3 New media

As explained above, most research done on this subject before the 21st century found that news most likely does impact exchange rates. All researched papers reviewed in the literature yielded similar results regarding the significance of news and unanticipated events. In the last 10 years, as a result of both digitalization and globalization, news and the media have taken new forms. News and other types of information are being spread online both faster and in larger proportions. With the dawn of social media, anyone can share their beliefs and opinions with the entire world. These developments have drastically changed the way people can gather their private information, which has variously shifted the ways people form their beliefs, behaviors and expectations.

Both Bollen, Mao and Zeng (2010) and Nofer and Hinz (2015) have examined correlations between people’s moods, Twitter and stock market prediction. Bollen et al. argue that behavioral finance has proven that financial decisions are driven mostly by mood and emotion. It is therefore reasonable to assume public moods can drive stock markets in the same way news does. These mood states consist of different degrees of, for example, anger or happiness. They investigated whether measurements of large numbers of mood states gathered on Twitter show any correlation with the Dow Jones value over time. They conclude it does, and thus that Dow Jones value

predictions can be significantly improved by including data-analysis on Twitter. Nofer and Hinz created a sample of almost 100 million tweets published in Germany, looking for any correlation with the German stock market. In their first analysis, no significant results were found. In their second analysis, when they integrated the followers of those who tweeted, they did find significant results. They concluded that the spread these mood states have is vital for their impact on economic forces.

Russo, Papaioannou, and Siettos (2013) have attempted to forecast what impact Twitter can have on economic trends. He used Twitter to model and forecast the euro–USD exchange rate on a high frequency scale. Using time series and trading simulations, he found that, for recent years, Twitter can be used to forecast economic events. People’s beliefs, which can be data-mined by analyzing social media, can outperform the random walk hypothesis in exchange-rate determination. Asur and Huberman (2010) have demonstrated how social-media content can be used to predict economic outcomes. They used Twitter to forecast box-office revenues for movies. They claim to have constructed a linear regression with which they can predict the box-office revenues of movies in advance of their release. Results were gathered by analyzing sentiments in tweets and identifying their reach. They believe their work shows how social media can create a form of collective knowledge which, when properly used, can yield accurate determinants for future outcomes on a large range of topics.

In 2017, Malavar-Vojdovic researched whether tweets by Trump, containing information about American policy targeting Mexico, have had an impact on the daily USD–Mexican peso

9 exchange rate. In the period between June 16, 2015, and February 21, 2017, Malavar-Vojdovic analyzed 4,999 tweets posted by Trump. For most of this period, Trump was running his presidential campaign, in which he would repeatedly share his predominantly negative views on Mexico and the wall that needed to be built. By using a GARCH model and classifying negative tweets separately, he found that Trump’s daily negative tweets have indeed influenced the daily USD–Mexican peso exchange rate. Every negative comment on Twitter concerning Mexico made the daily average vitality of the exchange rate increase by 21.6%. He found that the Mexican peso mostly dropped after tweets by Trump. Also, on April 6, 2017, Mexico’s central bank chief stated the Mexican central bank had to alter course after several of tweets by Trump concerning Mexico had made the peso dive to historic lows. Thus, Mexico has clearly suffered under Trump’s Twitter usage.

10

Figure 1. 11 minute horizon of the USD–CNY exchange

rate. Impact is at point 6 (appendix, tweet 8)

3. Data analysis

The data used for this analysis is gathered from Dukasopy and Twitter. Via Dukascopy, a database is used consisting of the USD–CNY exchange rate per minute, hour and day from January 1, 2015, to June 2, 2018. The observed exchange rates consist of the opening rates, meaning the rates at the beginning of each period. On weekends, the exchange rate market is closed. The data therefore consists only of information on rates from Monday until Friday. Via Twitter, a selection is made of 15 tweets by Trump from March 5, 2018, to May 16, 2018 (Appendix). The time of when the tweets were posted is known by the minute. For both the tweets as the exchange rate, the GMT time zone is used to order the data correctly.

Firstly, leaving out the tweets, some data is found to examine the exchange rate differences before and during the Trump presidency. From January 1, 2015, to November 8, 2016 (the day Trump got elected), the average hourly exchange rate fluctuation was 0.027%. From November 8, 2016, to June 2, 2018, the average hourly exchange rate fluctuation increased to 0.0362%. This is a 33% increase relative to when Obama was still in office. From the day Trump got elected to the day he got inaugurated as president (January 20, 2017), the average hourly exchange rate fluctuation was 0.0434%. This is an increase of 60% relative to before he got elected. On November 8 itself, the dollar appreciated 0.549% relative to the Chinese Yuan. This is unusually high for a daily exchange rate. This data provides strong evidence that Trump's presidency has directly influenced the USD– CNY exchange rate volatility.

Via Twitter, 12 tweets of Trump are gathered and thoroughly examined on their possible influence on the USD–CNY exchange rate volatility (Appendix). All of these tweets somehow relate to the current frictions and trade negotiations between China and the United States. A distinction is made between “positive” and “negative” tweets. This distinction is made on basis of what Trump states in his tweets and what is believed he wants to predicate with these statements. 8 of the tweets are considered negative, 4 are considered positive. 3 of the tweets were written on either Saturday or Sunday and will therefore be left out of this analysis. By comparing the time of the tweet to the exchange rate movement right before and after the time of posting, fluctuations in the

exchange rate are to be observed. The average 1-min fluctuation for the USD–CNY from January 1, 2018, to June 2, 2018, is 0.004916%. Further 1-min fluctuations in this analysis will be compared to this value.

3.1 Negative tweets

On May 16, 2018, 01:09 PM, Trump posted 3 tweets concerning China (Appendix, tweet 8). Firstly he stated the Washington Post and CNN falsely reported on the trade negotiations with China, after which he said The United States has been losing billions of dollars on China and China’s demands should therefore be little. The tweet was posted on the 6th minute on the 11-minute horizon on Figure 1. Figure 1 shows exchange rate fluctuations rise almost immediately after impact. The average fluctuation in the exchange rate before impact is 0.00443% per

11 minute. After impact this increases to 0.00791%, implying an 78% increase in fluctuation. In the minute the tweet was posted, the exchange rate increases with 0.00141%. One minute later, the exchange rate drops 0.01382%, which is 2.81 times average (Table 1).

Because Twitter gives de time of the tweets to the minute precise, it cannot be observed whether impact is at the beginning of this minute or further into the minute. When a tweet would for example be posted at 01:09:59 PM, any possible fluctuations as a result of this tweet cannot reasonably be expected at the minute of impact. When the tweet is send on 01:09:01 PM,

fluctuations following this tweet can be expected at the same minute in which de tweet was posted, so it remains relevant. This will be accounted for in further analyzing by examining both the minute of impact and the minute after impact separately.

On Table 1, an analysis is made of 8 tweets that are considered “negative.” These tweets are analyzed similarly as the tweet above. In all analyzed tweets from table 1, the exchange rate

fluctuation rises after impact. Regarding the minute of impact, 4 out of 8 tweets show above average fluctuation during this minute. Out of the 4 tweets that do not show above average fluctuation at impact, 3 show significantly more fluctuation one minute after impact, namely 0.0138%, 0.01996% and 0.0095%. This is respectively 2.81, 4.06 and 1.93 times average. Only the rate concerning tweet 5 shows no real fluctuation, until 5 minutes after impact, when it drops 0.0173% (3.52 times average) in one minute.

Average change per min before impact(%) Average change per min after impact(%) % Change After/ Before Impact (%) Ratio Impact/ Average Impact +1 (%) Ratio Impact+1/ Average Tweet 1 0.0036312 0.0065377 +80.04% 0.0077371 1.573 0.0047374 0.964 Tweet 2 0.0052901 0.0068403 +29.30% 0.0028505 0.580 0.0095014 1.933 Tweet 3 0.0045347 0.0048509 +6.97% 0.0079267 1.612 0.0000000 0.000 Tweet 4 0.0034805 0.0061701 +77.28% 0.0034805 0.708 0.0061701 1.255 Tweet 5 0.0026951 0.0057703 +114.1% 0.0095118 1.9348 0.0014266 0.290 Tweet 6 0.0037001 0.0044589 +20.50% 0.0001581 0.032 0.0015811 0.321 Tweet 7 0.0069415 0.0078928 +13.70% 0.0038034 0.774 0.0199686 4.060 Tweet 8 0.0044300 0.0079187 +78.75% 0.0014140 0.288 0.0138255 2.812

Table 1. Analysis of 8 ‘’negative’’ tweets. Column 1 and 2 are calculated by taking the percentage change from 1 minute to the next,

for 5 minutes before and after the tweet, and averaging that. Column 3 is the percentage increase in fluctuation, and calculated by dividing after by before. Column 4 (Impact) is the percentage change of the exchange rate in the minute the tweet was posted. Column 6 (Impact+1 )is the percentage change of the exchange rate in the minute after the tweet was posted. Impact and impact+1 divided by average are ratios of the percentage change in the exchange rate of each minute divided by the average of 2018 (0.004916%).

12

Figure 2. 11 minute horizon of the USD–CNY exchange

rate. Impact is at point 6 (tweet 12, appendix)

3.2 Positive tweets

Also a selection of 4 tweets is made that contains “positive” tweets. Meaning Trump expresses himself on China in an approving and supportive way. These tweets were written from April 10, 2018, to May 8, 2018. Table 2 shows the data concerning these tweets. Only after 1 out of 4 tweets, the exchange rate increases in fluctuation. This differs from the “negative” tweets, where after 8 out of 8 tweets, the USD–CNY exchange rate increases in fluctuation. In addition, for 3 out of 4 tweets, neither the minute of impact nor the minute after impact show any fluctuation that can be

considered above average. This also differs from the results found on negative tweets, where for 7 out of 8 tweets, an above average fluctuation in either the minute of impact or one minute after impact is observed.

On May 8, 2018, Trump posted a tweet in which he stated he would be speaking to his friend, President Xi of China, after which he said good things were about to happen. On figure 2, the US–CNY exchange rate around this time is plotted. Again, impact is at point 6. In contrast to all negative tweets, it seems that the exchange rate stabilizes right after the tweet. Whether this is a result of his tweet cannot be concluded, but it is interesting to observe that this is a path that is observed in neither of the negative tweets.

Before After % Change After Before Impact Ratio Impact/ Average Impact +1 Ratio Impact+1/ Average Tweet 9 0.0030871 0.0037233 +20.60% 0.000000 0.000 0.0015912 0.324 Tweet 10 0.0031546 0.002900 -8.07% 0.0054174 1.101 0.0022308 0.454 Tweet 11 0.0048385 0.0016654 -65% 0.0017283 0.352 0.0003142 0.064 Tweet 12 0.0052499 0.0045582 -13.17% 0.0004558 0.093 0.0045582 0.927

3.3 Direction of the exchange rate

Heretofore, this data analysis focused only on whether there are fluctuations, not in which direction these fluctuations move. Regarding the “negative” tweets, there is much difference in whether the exchange rate increases or decreases. After 5 out of 8 tweets, the USD depreciates in the minute the tweet was posted. In the following 5 minutes, after 6 out of 8 tweets, the USD depreciates. For the positive tweets, 2 out of 4 show an appreciation of the USD at impact, the other 2 show a small depreciation. Regarding all 12 tweets, after 8 tweets the exchange rate starts moving in a different direction then they were just before impact. On basis of this information, their seems to be no hard evidence on correlation between the tweets and in what direction the exchange rate moves. Only for “negative” tweets, there are some minor hints it might lead to a depreciation of the USD more often than an appreciation.

13

4. Discussion

As discussed in the data analysis, for all eight tweets considered “negative,” an increase in the volatility of the exchange rate is observed just after the posting of the tweet. These are strong hints that investors actually respond to Trump’s tweets. “Positive” tweets seem to produce no indication of an increase in the volatility in the exchange rate. Also, there seems to be no or a minor correlation between the content of the tweets and the direction in which the exchange rate moves. This

discussion will try to explain these findings by assessing them from demographic, political, economic and cultural perspectives. Firstly, the specifics of the situation between China and the United States are elaborated. Secondly, the increased fluctuation is explained on basis of fundamental investing paradigms, and finally, the validation of announcement data is discussed.

4.1 China–US relations

Malavar-Vojdovic (2017) has concluded that Trump's tweets concerning Mexico have significantly influenced the USD–Mexican peso exchange rate. He found that the Mexican peso usually dropped after tweets regarding Mexico, while this data analysis found that the USD predominantly dropped after tweets concerning China. It is therefore believed there are some structural differences in Malavar-Vojdovic’s results concerning Trump's tweets on Mexico and this thesis’s results, which apply to China alone.

The first main difference between the US relation with Mexico and China is the exchange-rate regimes between both countries. While the USD–Mexican peso exchange-rate is floating, China changes its foreign exchange rate policy frequently. Right now, China does not allow the USD–CNY exchange rate to fluctuate more than 2.25%; if this tends to happen, the Chinese government will intervene to keep the rate within its bandwidth.

Secondly, it is questionable whether Trump’s opinions and social media usage can influence the USD–CNY exchange rate as much as it does the USD–MXN rate. The United States not only has a much larger economy than Mexico does, but they are also neighboring countries. These are both arguments for Mexico to be largely dependent on the United States, both economically and politically. It would therefore make sense for investors to react more drastically to Trump’s views and judgments on Mexico and, for example, the possibility of building a wall on the US-Mexico border.

China, nevertheless, is placed on the other side of the world, relative to the United States, and is one of the largest world economies. China is therefore in certain ways less connected with the United States than Mexico is. Furthermore, formal diplomatic relations between the US and China did not start until the 19th century. Both countries developed in isolation from one another for most of their history, which led to the rise of two distinct cultures. Today, this difference translates, for example, into the fact that only a small number of the Chinese people speak English, and almost no American people speak Chinese. Moreover, many Western companies and websites, like Twitter, are banned by the Chinese government. This ban of Twitter means that Trump’s tweets are not or are hardly available to Chinese agents and investors. This is alien to one of the assumptions of the efficient market hypothesis, which states that information must be available for all investors.

Altogether, China and the United States are two countries who continue to differ and are separated on many levels. The differences as described above are possible reasons for Trump his tweets to not reach or influence Chinese people and investors as it would, for example, people in

14 Mexico or in the European Union, each of whom are far more integrated with the United States, on certain levels. This divide presents a set of reasons for the CNY to react differently as did the Mexican peso as a result of Trump’s negative tweets

Although there are many differences between the US and China, as described above, on trade they are certainly more integrated. In 2016, China goods and services trade with the US estimated a total of US$ 648.5 billion. China is also the largest supplier of imports for the US, and is the US’s third-largest export market. The US is China’s largest import market and fourth export market (Office of the United States Trade Representative, 2018). The current frictions and disagreements on trade agreements between China and the US can hence have great impact on both countries. The stakes are considered high. Both countries’ actions presently seem reactive. Therefore, Trump’s tweets, especially concerning the US policy on this situation, could heavily affect investors’ expectations and thus the USD–CNY exchange rate. This might provide a possible

explanation for the larger increase in fluctuation of the USD–CNY after ‘’negative’’ tweets compared to ‘’positive’’ tweets. Negative remarks on China and trade agreements by Trump might make investors believe more US policy actions regarding trade tariffs are upcoming, which hence reshapes their expectations more drastically.

In conclusion, Trump’s Twitter usage seems not to have the same effect on China as it does on Mexico. While it had a predominantly negative effect on the Mexican peso, the CNY fluctuates in dissimilar directions. As explained above, this difference between China and Mexico may be due to the cultural and demographic differences between the United States and China and, thus, that Trump’s tweets do not reach and influence Chinese investors as much as Mexican investors. Nevertheless, some fluctuations were observed after the tweets, expected to be largely due to the economic dependence between the two countries.

4.2 Fundamentals in practice

The efficient market hypothesis (Fama, 1965) states that exchange rate markets are always efficient and that market rates reflect all information known in the market. It also states that investors are considered rational when they respond to new information. Related topics of interest to discuss are, firstly, the validity of categorizing Trump’s tweets as “new and relevant information,” and secondly, whether agents can be considered rational in their response to his tweets. Analyzing these matters might offer a clearer understanding of agents’ behavior and thus exchange rate fluctuations.

Trump is known for offering his opinion on Twitter, on many subjects. Not only is he not considered by many to be very nuanced in his tweets, but he also communicates falsehoods frequently. These factors contribute to the idea that Trump’s tweets cannot be considered very objective or a relevant source of information. On the other hand, he is one of the most powerful people in the world. His opinions are, or ought to be, relevant because of his role as the president of the United States. Whether or not his tweets can therefore be considered to be relevant information is not entirely clear.

This consideration is important in determining whether agents are considered “rational” in their response to the tweets. “Rational behavior” describes behavior that accords with reason and logic. When a source of information is based on neither reason nor logic, however, the rationality of one’s response to this information might also suffer.

Behavioral economics may provide a clearer solution to this issue. As unclear and un-objective Trump’s tweets can be, so may become agents’ responses to his tweets. Trump’s tweets

15 are full of his “moods and beliefs,” as described by Papaioannou et al. (2013). These typical

characteristics of a Trump tweet could make it significantly harder for investors not to let their own beliefs and psychological characteristics influence their behavior. Imagine, for example, an

independent news channel stating purely facts about a US government deficit. This objective presentation of information gives investors more opportunity to logically reflect on what the

consequences of this news could be and hence to adjust their investment strategy in a rational way. Discussing the rationality of investors is relevant in this case, for it might be able to explain, in part, the fact that there is much difference in the direction the exchange rate moves after Trump’s tweets. As discussed in the data analysis, there are some minor hints that negative tweets lead to more frequent depreciation of the USD, but there is no hard evidence in this conclusion. Different investors react thus dissimilarly to the same tweets. Behavioral economics would explain this by stating that all these investors have different beliefs and interpret these tweets differently. This less-rational behavior leads to both uncertainty and dissimilarity in which direction exchange rates move. An interesting, related topic is whether or not Trump is aware of the kind of impact his tweets might have on economic agents and thus economic markets. If he is aware, he might be able to compose his tweets in such a way as to accomplish a certain target. Malkiel (2003) has stated that stock prices are at least as predictable as people’s reaction to new information is. If Trump is aware of the way investors would interpret and react to his tweets, he could easily adjust his tweets in such a way that investors respond in the way he wants them to respond.

4.3 Announcement data

The results from the data analysis are found by using announcement data (Rose & Frankel, 1994), meaning that the exchange rate fluctuations only just before and just after the tweet were observed, 5 min on each side. As Rose and Frankel have described, using announcement data is easiest for analyzing the impact on exchange rates, because it isolates the effect of the tweet and ensures that as few as possible other factors can play a significant role as well.

The counterargument for using announcement data though, is that it does not take into account longer-term effects. Just as Goodhart et al. (1993) have observed, exchange rates can return quickly to their “pre-value.” Exchange rates fluctuate almost every second, implying the effect Trump’s tweets have on the USD–CNY exchange rate will mostly be offset by other fluctuations afterwards. The main question to ask here is whether it is still relevant to examine announcement data if the effects of these tweets is offset relatively quickly.

Despite the fact that the fluctuations of these tweets might be offset relatively quickly, indeed, the short-term fluctuations can still provide some interesting information. They offer important insight into the ways in which investors and agents react to new information. For example, the differences in the effect Trump’s Twitter usage on the USD–CNY and the USD–MXN exchange rates can provide economists with information on cross-country differences and economic behavior. Of course, for any kind of economic analysis or modelling, understanding economic agents is one of the most important conditions.

16

5. Conclusion

As previous research showed, ‘’news’’ and new information are determinants of exchange rates. When new information arises, economic agents react accordingly which is immediately reflected in the concerning exchange rates. In this thesis, the effect of Trump’s Twitter usage on the USD/SNY exchange rate volatility is analyzed.

Since the beginning of 2018, tensions are running high between the United States and China. Both countries are imposing tariffs on import products and are threatening with more tariffs. By using announcement data, 12 tweets of Trump consisting of his views and beliefs regarding this topic are analyzed, on their impact on the USD–CNY exchange rate volatility. Results show that for

“negative” tweets, in most cases, the exchange rate volatility increases almost immediately after the tweet is posted. For “positive” tweets, no significant results were found. Also, there is no clear evidence for any correlation between the content of the tweet, and the direction the exchange rate moves in.

In general, results seem not to be as one-sided as Malavar-Vojdovic’s findings on Mexico, who did find a significant correlation between the content of the tweet and the direction the exchange rate moves in. This is believed to be partly due to the cultural and demographic

differences between China and the US. The observed fluctuations following the ‘’negative’’ tweets though, are expected to be largely due to the economic dependence between the two countries. It is doubtful whether agents can be considered rational in their response to Trump’s tweets. Different agents seem to form dissimilar expectations based on the content of the tweet. By using announcement data, this thesis offers an interesting insight in these agent’s behavior and thus exchange rates.

For upcoming research, it would be interesting to examine whether statements or speeches by Chinese representatives influences the USD–CNY exchange rate in a similar way as Trump’s tweets do.

17

Appendix

Negative Tweets:

1. Delegation heading to China to begin talks on the

Massive Trade Deficit that has been created with our Country. Very much like North Korea, this should have been fixed years ago, not now. Same with other countries and NAFTA...but it will all get done. Great Potential for USA! 11:00 AM - 1 May 2018

2. When a car is sent to the United States from China, there

is a Tariff to be paid of 2 1/2%. When a car is sent to China from the United States, there is a Tariff to be paid of 25%. Does that sound like free or fair trade. No, it sounds like STUPID TRADE - going on for years!

10:03 AM - 9 Apr 2018

3.China, which is a great economic power, is considered a

Developing Nation within the World Trade Organization. They therefore get tremendous perks and advantages, especially over the U.S. Does anybody think this is fair. We were badly represented. The WTO is unfair to U.S.

2:32 PM - 6 Apr 2018

4.China has been asked to develop a plan for the year of a

One Billion Dollar reduction in their massive Trade Deficit with the United States. Our relationship with China has been a very good one, and we look forward to seeing what ideas they come back with. We must act soon!

18

5.We are not in a trade war with China, that war was lost

many years ago by the foolish, or incompetent, people who represented the U.S. Now we have a Trade Deficit of $500 Billion a year, with Intellectual Property Theft of another $300 Billion. We cannot let this continue!

11:22 AM - 4 Apr 2018

6.We are on the losing side of almost all trade deals. Our

friends and enemies have taken advantage of the U.S. for many years. Our Steel and Aluminum industries are dead. Sorry, it’s time for a change! MAKE AMERICA GREAT AGAIN! 12:10 AM - 5 Mar 2018

7.We cannot keep a blind eye to the rampant unfair trade

practices against our Country! 2:37 PM - 14 Mar 2018

8.The Washington Post and CNN have typically written false stories about our trade negotiations with China. Nothing has happened with ZTE except as it pertains to the larger trade deal. Our country has been losing hundreds of billions of dollars a year with China...

1:09 PM - 16 May 2018

...We have not seen China’s demands yet, which should be few in that previous U.S. Administrations have done so poorly in negotiating. China has seen our demands. There has been no folding as the media would love people to believe, the meetings...

1:09 PM - 16 May 2018

...haven’t even started yet! The U.S. has very little to give, because it has given so much over the years. China has much to give!

19 Positive Tweets:

9. So much Fake News about what is going on in the White

House. Very calm and calculated with a big focus on open and fair trade with China, the coming North Korea meeting and, of course, the vicious gas attack in Syria. Feels great to have Bolton & Larry K on board. I (we) are

10:38 AM - 11 Apr 2018

10.Very thankful for President Xi of China’s kind words on

tariffs and automobile barriers...also, his enlightenment on intellectual property and technology transfers. We will make great progress together!

6:30 PM - 10 Apr 2018

11.I will be speaking to my friend, President Xi of China, this

morning at 8:30. The primary topics will be Trade, where good things will happen, and North Korea, where relationships and trust are building.

11:22 AM - 8 May 2018

12.Our great financial team is in China trying to negotiate

a level playing field on trade! I look forward to being with

President Xi in the not too distant future. We will always have a good (great) relationship!

20

Bibliografie

Asur, S., & Huberman, B. A. (2010). Predicting the future with social media. In Proceedings of the 2010 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology-Volume 01 (pp. 492-499). Los Alamitos, USA: IEEE Computer Society.

Beck, S. E. (1993). The Ricardian equivalence proposition: Evidence from foreign exchange markets. Journal of International Money and Finance, 12(2), 154-169.

Bollen, J., Mao, H., & Zeng, X. (2011). Twitter mood predicts the stock market. Journal of Computational Science, 2(1), 1-8.

Bomhoff, E. J., & Korteweg, P. (1983). Exchange rate variability and monetary policy under rational expectations: Some Euro-American experience 1973–1979. Journal of Monetary Economics, 11(2), 169-206.

Dornbusch, R. (1976). Expectations and exchange rate dynamics. Journal of Political Economy, 84(6), 1161-1176.

Dornbusch, R. (1982). Exchange rate economics: where do we stand? In International Economics Policies and their Theoretical Foundations (pp. 557-599). Cambridge, USA: Massachusetts Institute of Technology.

Edwards, S. (1982). Exchange rates and ‘news’: A multi-currency approach. Journal of International Money and Finance, 1(1), 211-224.

Fama, E. F. (1965). The behavior of stock-market prices. The journal of Business, 38(1), 34-105.

Frankel, J. A. (1979). On the mark: A theory of floating exchange rates based on real interest differentials. The American Economic Review, 69(4), 610-622.

Frankel, J. A., & Rose, A. K. (1994). A survey of empirical research on nominal exchange rates (No. w4865). Berkeley, USA: National Bureau of Economic Research.

Frenkel, J. A. (1976). A monetary approach to the exchange rate: doctrinal aspects and empirical evidence. The Scandinavian Journal of Economics, 78(2), 200-224.

Frenkel, J. A. (1982). Flexible exchange rates, prices and the role of ‘news’: Lessons from the 1970s. In R. A. Batchelor, & G. E. Wood (Eds.), Exchange Rate Policy (pp. 48-100). London, England: Palgrave Macmillan.

Goodhart, C. A., Hall, S. G., Henry, S. B., & Pesaran, B. (1993). News effects in a high-frequency model of the sterling-dollar exchange rate. Journal of Applied Econometrics, 8(1), 1-13.

Gyntelberg, J., Loretan, M., Subhanij, T., & Chan, E. H. P. (2009). Private information, stock markets, and exchange rates. Retrieved from

https://www.imf.org/en/Publications/WP/Issues/2016/12/31/Private-Information-Capital-Flows-and-Exchange-Rates-26225

How Does President Trump’s Twitter Use Impact Forex, Markets And Stocks? (n.d.). Retrieved May 5, 2018, from https://www.fxcm.com/insights/president-trumps-twitter-impact-forex-markets-stocks/

Ito, T., & Roley, V. V. (1987). News from the US and Japan: which moves the yen/dollar exchange rate?. Journal of Monetary Economics, 19(2), 255-277.

MacDonald, R. (1999). Exchange rate behaviour: are fundamentals important? The Economic Journal, 109(459), 673-691.

MacDonald, R., & Taylor, M. P. (1992). Exchange rate economics: a survey. Staff Papers, 39(1), 1-57. Malaver-Vojvodic, M. (2017). Measuring the Impact of President Donald Trump’s Tweets on the

Mexican Peso/U.S. Dollar Exchange Rate. Ottawa, Canada: Department of Economics of the University of Ottawa.

Malkiel, B. G. (2003). The efficient market hypothesis and its critics. Journal of Economic Perspectives, 17(1), 59-82.

Meese, R. A., & Rogoff, K. (1983). Empirical exchange rate models of the seventies: Do they fit out of sample? Journal of international economics, 14(1-2), 3-24.

21 Moyer, L. (2018, April 4). A timeline of Trump's escalating trade war with the world. Retrieved May

11, 2018, from https://www.cnbc.com/2018/04/04/a-timeline-of-president-trumps-escalating-trade-war-with-the-world.html

Nofer, M., & Hinz, O. (2015). Using twitter to predict the stock market. Business & Information Systems Engineering, 57(4), 229-242.

Office of the United States Trade Representative. (n.d.). The People's Republic of China. Retrieved May 5, 2018, from https://ustr.gov/countries-regions/china-mongolia-taiwan/peoples-republic-china

Papaioannou, P., Russo, L., Papaioannou, G., & Siettos, C. I. (2013). Can social microblogging be used to forecast intraday exchange rates?. Netnomics: Economic Research and Electronic

Networking, 14(1-2), 47-68.

Shleifer, A. (2000). Inefficient markets: An introduction to behavioural finance. Oxford, England: OUP Oxford.

Twitter - Statistics & Facts. (n.d.). Retrieved April 25, 2018, from https://www.statista.com/topics/737/twitter/