The Stock Market Reaction to Presidential Tweets in the Case of the US-China Trade War

Table of Contents

List of figures

List of tables

List of abbreviations

List of symbols

1 Introduction

2 Literature review
2.1 Financial markets’ reaction to new information
2.2 The impact of macroeconomic announcements on U.S. stock markets

3 Event Study
3.1 Event Review - the US-China trade war
3.1.1 Emergence and sources of conflict
3.1.2 Development and imposed tariffs
3.2 Methodology and theoretical framework
3.2.1 Assumptions
3.2.2 Event definition and selection criteria
3.2.3 Event window, post-event window and estimation window
3.2.4 Expected return models and (cumulative) abnormal returns
3.2.5 Null hypothesis and statistical validation
3.3 Data description

4 Discussion of results
4.1 Events’ impacts on aggregated stock market valuations
4.2 Events’ impacts on stock markets on a country level
4.2.3 U.S. American stock market - S&P 500
4.2.4 Chinese stock market - Hang Seng Index
4.2.5 German stock market - DAX
4.3 Overview and conclusion
4.4 Quality of the study and limitations

5 Conclusion and areas of future research

Appendix A - U.S. trade in goods with China (2017-2019)

Appendix B - Selection, classification and listing of tweets

Appendix C - T-distribution table

Bibliography

Abstract

In times of an increasingly digitalized world, behavioral changes in society do not spare high-ranking politicians and decision makers. In some cases, those changes in behavior can have unforeseen yet considerable consequences. By making use of the renowned event study methodology, this paper scrutinizes the impact of acting U.S. president Donald Trump’s Twitter activity on international stock markets. In particular, a select set of ten short messages posted in the context of the present US-China trade dispute is analyzed with regard to the U.S. American S&P 500, the Chinese Hang Seng Index (HSI) and the German DAX. Highly significant market reactions, both positive and negative, are found for the HSI and the examined markets’ aggregate, while Trump’s native market showed the least responsiveness to his tweets. Apart from that, the obtained results suggest a fairly rapid processing of new information and thus adjustment of prices.

List of figures

Abbildung in dieser Leseprobe nicht enthalten

List of tables

Abbildung in dieser Leseprobe nicht enthalten

List of abbreviations

Abbildung in dieser Leseprobe nicht enthalten

List of symbols

Abbildung in dieser Leseprobe nicht enthalten

1 Introduction

Ever since Donald Trump’s official inaugural address as the 45th president of the United States of America in January 2017, he has been cherishing a very polarizing and distinct way of leading the world’s largest economy compared to his more recent predecessors. That leadership style is not least characterized by his preferred yet - considering his position - rather uncommon way of communicating to the outside world, namely his extensive use of microblogging service Twitter for presidential announcements and commentaries of any nature¹. Amongst others, a particularly high activity can be observed in conjunction with the rising political and economic tensions between the United States and China that have been intensifying over Trump’s course of presidency. The dispute between the two economic superpowers - it comprises various topics such as the United States’ massive and long-standing trade deficit and alleged intellectual property theft - ultimately lead to the imposition of a series of mutual tariffs worth hundreds of billions of U.S. dollars. And while Trump regularly keeps his followers updated about his thoughts, claims and the statuses of trade negotiations, stock markets around the globe seem to react heavily to the developments of what is referred to as the US-China trade war. The question arises whether those unscheduled and seemingly impulsive short messages can be a causal explanation for recent stock market movements.

The imposed tariffs and the persisting gridlock in negotiations for a trade agreement undoubtedly have direct economic implications for either one of the two countries and a majority of domestic companies, respectively. From a broader perspective, the dispute might also affect parts of the macroeconomic landscape in the medium to long run. Research by Charbonneau and Landry (2018, p. 1), for instance, suggests “considerable changes in trade flows and sectoral output reallocations” as well as “modest impacts on long-run aggregate prices and output levels” as a consequence of proposed and already imposed tariff schemes. While both the release of scheduled macroeconomic announcements (see chapter 2.2) and the Twitter sentiment around users (for example Ranco et al., 2015) were found in prior analyses to be able to directly impact financial markets, to my best knowledge, no research has been conducted yet regarding the potential effects of unscheduled social media announcements or commentaries of high- ranking officials on the latter. An exception that further underlines the topic’s increasing relevance is formed by major U.S. investment bank J.P. Morgan that recently created an index tracking the effects of presidential tweets on U.S. bond yields (Alloway, 2019).

This paper aims at filling the still persisting void in scientific research by examining international stock markets’ reactions to presidential tweets as the leading information source and sentiment indicator of US-China trade negotiations. For this purpose, an event study is conducted that analyzes the effects of ten deliberately selected tweets of acting president Donald Trump on the three representative stock market indices S&P 500 (U.S.), Hang Seng Index (HSI, China) and DAX² (Germany). Depending on the implicit information content, the tweets are categorized into the three classes positive, negative and neutral. In addition to the mere reaction of the markets to each class of tweets, patterns amongst the former are identified and possible explanations are proposed.

The remainder of this paper is organized as follows: In chapter two, a brief literature review is given concerning the markets’ reaction to new information, the theory of efficient markets and the role of macroeconomic announcements in asset price building. Chapters three and four are the main focus of this paper and primarily consist of said event study: After recapping the emergence and economic implications of the trade war, the study’s theoretical framework is elucidated that in turn builds the foundation for an in-depth discussion of the obtained results. Finally, chapter five concludes and identifies areas of further research.

2 Literature review

In the following section, a brief review is given about the existing literature regarding both the impact of new information on financial markets in general and that of macroeconomic announcements on stock markets in particular. Doing so has a twofold purpose: First, the previous findings lay the foundation for comprehending the underlying price building mechanisms and model assumptions referred to in chapter three of this paper. Second, they point out consent and dissent among researchers and hence aid in assessing the findings of this study. For the sake of clarity, chapter 2.2 focuses on the U.S. market only.

2.1 Financial markets’ reaction to new information

As early as in 1970, American economist Eugene E. Fama formulated the so-called efficient-market hypothesis (EMH) in its classic form; a theory regarding asset price building that is as much controversial in academic science as it is still relevant in today’s financial world. According to the EMH, investment decisions can be made under the assumption that prices of financial assets at any time fully reflect all available information relevant to their valuation.³ Fully reflect, in that context, refers to the equilibrium expected return on a security being a function of - and only of - the associated risk of holding it. The latter, in turn, is “conditional on some relevant information set” (ibid., p. 384), i.e. all publicly available information that helps in determining an asset’s relative risk. In other words, the efficient-market hypothesis implies that prices of securities always represent a rational assessment of intrinsic value and future payoffs (Andersen and Bollerslev, 1998). Changes in asset prices should therefore solely reflect and be attributable to the release and processing of new information, that is, information which is not anticipated by market participants in advance of their release.

The EMH is itself based on and consistent with the theory of random walk, a financial theory first seized on by French broker Jules Regnault in 1863 and later taken up by, amongst others, French mathematician Louis Bechalier (1900), American professor Burton Malkiel (1973) and above-mentioned Eugene Fama (1965). The latter, in a condensed form of his doctoral dissertation, put it as follows: Pursuant to the idea of random walk, prices of securities move in an entirely random manner around their intrinsic value - or equilibrium price - that is in turn determined by its earnings potential. All occurring stock price movement in the absence of relevant new information, hence, must merely be attributable to underlying uncertainty in the real world that gives rise to disagreement among market participants about an asset’s “correct” intrinsic value. By implication, those discrepancies between intrinsic values and actual prices have to be random rather than systematic, otherwise traders would take advantage and thus immediately neutralize that very systematic behavior. Aforementioned Bechalier (1900, p. 1), early proponent of the idea of random walk, wrote with regard to the determination of the fluctuation of asset prices:

“[...] [They are] subject to an infinite number of factors: it is therefore impossible to expect a mathematically exact forecast. Contradictory opinions in regard to these fluctuations are so divided that at the same instant buyers believe the market is rising and sellers that it is falling. ”

An important implication of the above theories is that in such efficient or “ideal” markets, where prices serve as an accurate signal for investment decisions, none of the participants can outperform the market in a consistent manner. Making use of the likes of fundamental or technical analysis⁴ can thus not be expected to generate constant risk-weighted excess returns over the market under the assumption of efficiency.

Albeit findings were partly lacking statistical significance or conditional on specific assumptions, the theories of random walk and EMH, respectively, were directly or indirectly confirmed by a large number of researchers over time, for instance Radner (1979), Eun and Shim (1989), Metcalf and Malkiel (1994), Rubinstein (2001), Malkiel (2005) and Yen and Lee (2008), as reported by Sewell in his 2011 historical review of the theory. Equally, a considerable amount of literature - he determines about half of all reviewed papers to reject the EMH - has evolved that challenges the implications of efficient and rational markets. Grossman and Stiglitz (1980), for example, argue that markets can never be perfectly informationally efficient, as the process of information­gathering is costly for market participants. If prices perfectly reflected all information available, informed investors would obtain no compensation for their efforts, and there would hence be no incentive to acquire information in the first place. Lakonishok et al. (1994, p.1541) conclude that value strategies, i.e. trying to select stocks with low prices relative to their fundamental values such as earnings or dividends, yield higher returns (excess returns) by exploiting the “suboptimal behavior of the typical investor” rather than accepting higher fundamental risks. De Bondt and Thaler (1985), Chopra et al. (1992) and Haugen (1995), moreover, all point out certain forms of overreactions by market participants to new information; an argument that is however firmly rejected by Fama (1998) who argues that said overreactions are just as common as underreactions and therefore results of chance. He furthermore emphasizes the importance of the underlying methodology and states that “most long-term return anomalies tend to disappear with reasonable changes in technique” (ibid., p. 283).

In conclusion, the last five decades of research yielded a number of different approaches as to the exploration of asset price behavior and the role of new information within financial markets. While there are equally arguments supporting and rejecting the three forms of efficient markets, it is important to stress the theoretical nature of efficiency following the above definition. Strictly speaking, the EMH must almost certainly be false in a real-world scenario that is subject to uncertainty (cf. Sewell, 2011). Its fundamental idea, however, still appears to be of high relevance in the modern, interconnected financial world.

2.2 The impact of macroeconomic announcements on U.S. stock markets

By definition, the release of macroeconomic information has the potential to constitute a news event; namely, in case the actual announcement figures differ from the average market participants’ expectations. Following the theory of market efficiency, a majority of the conducted research in the field indeed concentrates on the “news component” of announcements (see for example Cutler et al. (1988); Aggarwal and Schirm (1998); Kim et al. (2004) and Andersen et al. (2007)).

Macroeconomic announcements hold a special position within capital markets. Unlike other financial metrics like exchange or interest rates, those figures can typically not be observed by market participants, but are instead released in predefined intervals. Aggarwal and Schirm (1998, p. 85) refer to this as “conditions of imperfect information”. In combination with rational expectations, changes in such variables can therefore be expected to influence prices of various assets. Following Lott, Jr. and Birz (2011), the channel by which macroeconomic and other economically relevant events could affect stock prices can be shown through the fundamental price equation of a security:

Abbildung in dieser Leseprobe nicht enthalten

where *t is the security’s price at time t, Ct+3 is the expected future cash flows at time t + τ and kt+3 is the future discounting factor at time t + τ as a function of the risk-free interest rate and the respective risk premium.

Macroeconomic announcements, in particular, may reflect useful new information about the state and performance of the domestic economy (Aggarwal and Schirm, 1998) and thus affect market participants’ expectations about future economic activity. Those expectations, in turn, determine the price of the security, as economic conditions can affect both cash flows and the discount rate. Empirical research provides us with divided results as for the role of various macroeconomic announcements in asset price building. Cutler et al. (1988), who examine the impact of seven kinds of macroeconomic announcements and other world events⁵ on the U.S. stock market between 1926 and 1985, report that macroeconomic news alone can only explain around one third of the total variance in stock returns, even though most of the analyzed variables had a significant influence. Aggarwal and Schirm (1998) capture significant asymmetrical effects of trade balance announcements (solely) for the first half of their examination period (1985-83): Equity prices were more responsive to relatively small announcement surprises than they were to large ones. The authors hypothetically link those findings to government and central bank objectives and according policies. Kim et al. (2004) investigate the impact of scheduled government announcements on U.S. financial markets between 1986 and 1998. Consistent with the efficiency hypothesis, they conclude that financial markets do not react in any manner to the sheer release of such information but rather to the implied news content. In accord with DeGennaro and Zhao (1998), they furthermore report the stock market to be most responsive to price-related news such as CPI and PPI, while little evidence is found in the context of trade news and other non-price news about the economical state. These findings are partly confirmed by Flannery and Protopapadakis (2002), who determine the CPI and PPI to be the only variables influencing the returns of the market portfolio, apart from money aggregates such as M1 (real money supply). However, they report specific real factors to affect the returns’ conditional volatility⁶, namely the trade balance, employment reports and housing starts⁷. Interestingly, two common measures of aggregate economic activity, the real gross national product (GNP) and industrial production, did not exhibit any significant effects at all. On the basis of a broader approach, Savor and Wilson (2013) examine the hypothesis of stock returns being predictably higher on announcement days of economic news, i.e. CPI and PPI, employment figures and interest rate decisions by the Federal Open Market Committee (FOMC). Not only do they find significantly higher stock returns, but also a ten times larger Sharpe Ratio as a measure for the risk-adjusted performance of assets between 1959 and 2009.

There are several approaches as for the explanation of mixed findings among researchers. Kurov et al. (2019) report a significant pre-announcement drift of asset prices that numerous studies were not considering and which might indicate information leakage. Andersen et al. (2007) determine the stage of the business cycle to play an important role for the equity markets’ response to macroeconomic news: They report bad news to exhibit the expected negative effect on equity markets during contractions, but a positive one during expansions which, if averaged across those cycles, might explain the partly insignificant or contradictory findings. Most important in light of this paper might be the argument that Lott, Jr. and Birz make, who were the first to report a significant effect of GDP growth on U.S. stock markets in their 2011 study. They argue that not the act of releasing information is the crucial factor, but rather the public interpretation of the latter whose dependence on economic conditions and overall complexity might lead to inconsistencies. In order to control for this, they (successfully) use newspaper headlines as a proxy for investors’ interpretations and an indicator for the relevance of information releases.

The vast amount of previous research is clearly split over the role of various macroeconomic figures in asset price building. Several studies report real economic variables such as GNP or GDP not to influence stock returns in any significant way. For price-related news, by contrast, a considerable number of researchers were able to capture significant effects in relative terms. Announcements in the context of trade balances, industrial production or unemployment rates yielded heterogenous results. From an overall perspective, it appears that macroeconomic factors play a vital, although not necessarily an outstanding role in asset price building. Especially the development towards more sophisticated and somewhat innovative approaches such as that of Lott, Jr. and Birz (ibid.) seems to yield compelling evidence for that.

3 Event Study

An event study is a comparably simple means of measuring the impact of economically relevant events on the valuation of firms or entire markets, respectively. Due to their straightforward application, event studies can be conducted for a wide area of possible events such as companies’ earnings announcements, the release of macroeconomic data variables and even unusual occurrences like natural disasters or the death of politically important decision makers, as for instance examined by Niederhoffer (1971). The main intention of this event study is to scrutinize the general ability of presidential tweets to move stock markets or, in other words, to answer the question whether market participants do perceive the former as news in the sense of the above definition. For this purpose, a select set of ten tweets - henceforth occasionally referred to as events - is analyzed in terms of its impact on three representative and globally relevant stock market indices: The U.S. American Standard and Poor’s 500 (S&P 500), the Chinese Hang Seng Index (HSI) and, finally, the German DAX.

3.1 Event Review - the US-China trade war

Donald Trump (1988, para. 2) first publicly declared his affection for tariffs as an instrument of protectionism in the 1980s, when stating to “believe very strongly in tariffs” in an interview regarding Japan’s trade practices and growing wealth at the time. While trade sanction announcements against China played a substantial part in the election campaign preceding his presidency, trade protectionism is now considered a core element of his economic policy.

3.1.1 Emergence and sources of conflict

On August 14, 2017, president Donald Trump (p. 1) wrote in a memorandum to the Office of the United States Trade Representative⁸ (USTR):

“China has implemented laws, policies, and practices and has taken actions related to intellectual property, innovation, and technology that may encourage or require the transfer of American technology and intellectual property to enterprises in China or that may otherwise negatively affect American economic interests. ”

He subsequently instructed the government agency to determine whether to investigate those circumstances under Section 301 of the U.S. Trade Act of 1974⁹. Said investigation was initiated shortly thereafter, and yielded the formal argumentative basis for the first round of tariffs on behalf of the United States. Amongst others, USTR (2018) confirmed both regulatory and administrative restrictions on foreign investments on the part of China with the purpose to pressure companies into transferring technology, as well as government backed cyber-attacks into U.S. commercial networks. By gaining unauthorized access to a wide range of sensitive information for over a decade, China had committed theft of intellectual property and other confidential business information. In consequence of those allegations, Trump regularly refers to the U.S.’ massive and long­standing trade deficit - it accumulated to a total of $375.4 Bn. by the end of 2017 and even $419.5 Bn. by the end of 2018 (United States Census Bureau, n.d.; see Appendix A) - as a result of unfair trade practices exercised by China.

China strongly rejects these accusations as “groundless” and itself accuses the United States of violating the rules of the World Trade Organization, as declared, amongst others, by Chinese Commerce Ministry spokesman Gao Feng at a news conference in Beijing (2018).

3.1.2 Development and imposed tariffs

Since the beginning of 2018, when Trump initiated first direct trade measures against China, mutual tariff rates are essentially on a continuous rise. Apart from some occasional tariff cuts, such as China’s Most Favored Nation (MFN¹⁰ ) tariff cut on consumer goods, automobiles and IT products on January 1, 2019, the trade dispute has been characterized by U.S. Section 301 tariffs and according retaliation on the part of China. Figure 1 below illustrates the development of the average tariff rates¹¹ on Chinese and American imports since the beginning of February 2018, when Donald Trump imposed first additional tariffs on solar panels and washing machines.¹²

Figure 1: Development of average mutual tariffs over all industries

Abbildung in dieser Leseprobe nicht enthalten

Note: The figure show the development of average mutual tariff levels of China and the United States since February 2018. The tariff rates are averaged across industries and projected for the remainder of 2019 based on recent government proposals.¹³

While, during the first six months, U.S. tariffs slightly increased by 0.7 percentage points (pp; from 3.1% to 3.8%) and Chinese tariffs even decreased by 0.8 pp from 8.0% to 7.2%, the subsequent half-year yielded an increase of 8.6 pp and 11.1 pp for U.S. and Chinese tariff rates, respectively. After a clearly visible plateau phase between October 2018 and June 2019, where the two parties seemed to converge and make considerable progress in closing a trade agreement, another wave of mutual tariffs followed after alleged renegotiation attempts by China. The latest announcements from August 23, 2019 - they include another increase of U.S. Section 301 tariffs as well as the re-imposition of previously suspended Chinese tariffs on automobiles - would lead to tariff rate peaks of 24.3% (U.S.) and 25.9% (China) by mid-December 2019.

Underlined are those numbers by the respective trade volumes: as of the end of June 2019, the United States have reportedly exported 52.0 Bn. worth of goods to China, while importing 219.04 Bn. The totals for the year 2018 amounted to 120.1 Bn. and 539.7 Bn. worth of goods, respectively (United States Census Bureau, n.d.; see Appendix A for a detailed listing of monthly trade figures between 2017 and 2019).

3.2 Methodology and theoretical framework

Unless indicated otherwise, this event study essentially follows the methodology outlined in Campbell, Lo and MacKinlay (1997).

3.2.1 Assumptions

The essence of any event study is the assumption that its effects are immediately reflected in asset prices, that is, the general validity of the efficient-market hypothesis in the semi­strong form (Fama, 1970), as defined in chapter 2.1. It should, however, be emphasized that by extending the examined time period beyond the actual date of the event, I allow for a gradual price adaption and the formation of possible patterns in the news processing. The select events are considered unforeseen, so that the abnormal returns constitute the market reaction to those very events, and it is assumed that there are no other confounding effects in the market (Sitthipongpanich, 2011). A fourth assumption is that the analyzed events are exogenous with regard to changes in market valuations, meaning that the events are the causal factor for potential price changes rather than vice versa (Campbell et al., 1997).

3.2.2 Event definition and selection criteria

In a first step, the events of interest are defined as a select set of 10 tweets, posted by acting U.S. president Donald Trump in the context of the US-China trade war.

[...]

¹ The U.S. President maintains an official Twitter account under the username “realDonaldTrump” with a following of around 64,5 million users as of September 23, 2019 (Twitter, 2019).

² “DAX” refers to “Deutscher Aktienindex“, the German equivalent to “German stock index”.

³ Fama defined three forms of market efficiency: the weak form, according to which security prices consider all past price information, the semi-strong form, where prices consider all publicly available information and adjustments happen immediately, and the strong form which claims that prices incorporate both all public and private information. Unless otherwise indicated, this paper, in principle, refers to the semi-strong form of market efficiency.

⁴ While fundamental analysis considers basic economic figures and ratios of a company in order to determine its financial health and the according fair value, technical analysis aims at predicting future prices by identifying patterns in past price behaviour and data.

⁵ 5 Cutler et al. define world events as major non-economic events such as presidencies or political conflicts, which also affected risk premia of assets and should hence be important for their pricing. By including those events, they are able to explain about half of the occurring variance in asset prices.

⁶ The conditional volatility, or conditional variance, is the variance of variables (returns) conditional on the value of at least on other variable, i.e. a specific information set (here: subsequently mentioned macroeconomic information).

⁷ “Housing starts” depicts an economic indicator that, for a specified time period, aggregates the number of privately-owned houses for which construction has started.

⁸ The Office of the United States Trade Representative (USTR) is a U.S. government agency that is primarily responsible for the development and recommendation of trade policy to the president, as well as the conduction of trade negotiations on various levels.

⁹ The Trade Act of 1974 was essentially passed with the purpose to strengthen economic growth, foster domestic employment and reduce trade barriers. Section 301, in particular, provides the U.S. with the authority to impose trade sanctions on foreign countries in order to resolve trade disputes and enforce trade agreements. It applies in the case of trade agreement violation or engagement in unfair trade practices by the trading partners (International Trade Administration, 2018).

¹⁰ MFN is a political status with respect to international trade that guarantees the nominee trade conditions, including tariffs rates, equal to any other country holding “most favored nation” status .

¹¹ The tariff rates are averaged over all industries, products and trade volumes and projected for the remainder of 2019 on the basis of announced increases in tariffs and estimates for the trade volume.

¹² It should be noted, however, that those additional tariffs were imposed in general and not solely on China.

¹³ The calculations for average tariff rates stem from Brown (PIIE, 2019) with data originally obtained from the International Trade Centre and announcements by China’s Ministry of Finance.