Investor Attention and Twitter

I. TableofContents

I. TableofContents

II. ListofTables

III. ListofFigures

IV. List of abbreviations

1 Introduction

2 Theory and literature review
2.1 Financial attention
2.2 Direct measurements of attention
2.2.1 Search Volume Index
2.2.2 Yahoo Finance search
2.2.3 TheBaidu-lndex

3 Empiricalstudy
3.1 Data and methodology
3.2 Results

4 Conclusion

V. References

ListofTables

II. ListofTables

Table 1: Regression results for ASVI as dependent variable

ListofFigures

III. List of Figures

Figure 1 SVI for the term “Donald Trump”

IV. List of abbreviations

Abbildung in dieser Leseprobe nicht enthalten

1 Introduction

“[...] a wealth of information creates a poverty of attention and a need to allocate that attention efficiently among the overabundance of information sources that might consume it.”¹

Within the last few years, seeking information through social media has increased rapidly. Nowadays, the social media platform Twitter is not only used to follow tweets of celebrities, but also to keep up with the latest news. According to Statista Twitter has more than 330 million active users at present. Moreover most of these active users are located in the United States.² One of the most famous and influential twitter users is the 45th president of the United States, Donald J. Trump. The president is a fervent user of Twitter since he is posting several tweets a day. The fact that in December 2020 his Twitter account @realDonaldTrump has reached 88.5 million followers shows indeed that based on his political power and authority he reaches a large audience via social media. Furthermore, his tweets appear in the media. Thus, people like Donald Trump are known as super hubs, influencers, or alpha users.³ His Twitter channel enables him to share his thoughts in an effective manner with his community. Trumps’ way of communicating is characterized by straightforward statements that can affect him by mentioned people or companies in a good or negative way.

Several studies confirmed the impact of twitter as an information intermediary for professional and retail investors. In fact, news circulates across Twitter faster than it is absorbed in financial markets. The results of Tafti et al. suggest that Twitter can signal an imminent spike in trading activity for a company's stock. Therefore, the platform represents a way for investors to gather valuable information.⁴

However, in the contemporary world, attention is a quite limited source and Investors have limited attention.⁵ The attention of retail investors has been studied in more detail in previous work. Thereby, the attention of retail investors was verified as a direct proxy for measurement. In the work of Da et. al. it was demonstrated that the direct measure SVI provided by Google Trends can be used to measure retail investor attention. In their study they have proved that the direct measure SVI, provided by Google trends, can be applied to measure the attention of retail investors. Moreover, they occupied the forecasting of IPO stock returns using Google search volume.⁶

Using the SVI, the impact of whether Donald Trump's company-related tweets capture investors' attention has also been studied by previous researchers. Gehde-Trapp and Mohapatra determined that Trumps tweets cause a spike in attention and an impact on financial markets.⁷ However, Kaissar indicates that the tweets do not trigger financial attention.⁸ Therefore, it is not possible to accurately determine a priori whether presidential tweets trigger the investors' attention. Thus, the presented discrepancies are the motivational drive for the following thesis.

This thesis focuses on the impact of Trump's company-related tweets on the attention of retail investors. In this context, the main research question is whether Trump's tweets clearly have a triggering effect on retail investors attention and whether this financial attention can be measured with the help of direct proxies.

This thesis is organized as follows. The first chapter deals with the underlying theory and findings from previous research. Here, the first part presents a precise and clear definition of the concept of financial attention. Thereby, it is relevant to distinguish that it only refers to financial attention and not to attention per se. In the second part of the first chapter, direct and indirect proxies are presented and critically examined. As the research question relates to measurements for direct proxies, various direct attention measures are presented and evaluated. In addition, previous work examines how direct proxies can be applied as a measure of IPO events. The second chapter deals with data collection and methodology. In this process, company-related tweets from Donald Trump's Twitter account are collected and classified into a positive or negative tone. After that the SVI for the affected companies is searched on Google Trends. Subsequently, using a regression analysis in excel, the empirical relationship is examined. After describing and highlighting the results, the conclusion critically discusses the findings and offers suggestions forfuture work.

2 Theory and literature review

This chapter focuses primarily on the fundamentals and existing evidence from previous research.

2.1 Financial attention

Trump has already created a lot of attention during his inauguration. His campaign therefore took advantage of a communications environment that offered sensationalism, novelty and understated statements in various media. Consequently, Trump has attracted more attention from journalists as well. This enabled Trump to spend less on advertising than his competitors.⁹ This already shows that Trump's news arouse an increase attention. However, the issue is to determine whetherfinancial attention can also be measured.

On the stock market, processes and transactions have improved dramatically due to the Internet. Therefore, the efficiency of the stock market is due to the efficient dissemination of information.¹⁰ The fact that investors have only a limited time to weigh all stocks¹¹ and that attention is costly, attention on financial information must be efficiently placed on those that will later yield the most profit. The costs of attention encompass information processing costs, time, and opportunity costs. The benefits of financial attention, conversely, include the gains from trading.¹² When an investor is forced to allocate his attention across different stocks, his ability is negatively correlated with the attention demands of other stocks. This is called the limited attention hypothesis.¹³

Investors can gather information through Twitter since the SEC allowed companies to disseminate information through Twitter. After the approval, algorithms were developed to identify which tweets are of interest to investors based on the huge volume of previously potentially relevant information.¹⁴

However, since not all information is openly available to anyone, it is necessary to determine whether the stock market is efficient. There are two types of efficiencies that are relevant here. Allocation efficiency, which deals with the optimal allocation of scarce resources, and information efficiency, which reveals which and how much information is disclosed by the pricing process. An efficient market is characterized by the fact that all available information is contained.¹⁵

Early literature on people decision making shows that in many situations people tend to avoid information that would be beneficial for that decision.¹⁶ One possibility with which individuals can control their decisions is the determination whether to acquire information. In the research by Karlsson et al. it is presented that decisions about acquiring information are probably connected to the internal psychological treatment of information and the hedonic effect of information on a person’s utility.¹⁷ Also, they did some research indicating that the influence of news depends not only on the objective, but also on psychological factors. People seem to receive greater benefits from positive outcomes and greater disutility’s from negative outcomes when they feel personally responsible for the outcomes.¹⁸

Barber and Odean also studied attention-based decision making. In their work, they tested a model of decision making in connection with the purchase of common stock. In this context, investors are faced with the search problem of which shares to buy. Thus, without the help of a computer, it is quite impossible for most investors to evaluate the merits of all available common stocks. When selling, however, investors consider only the shares they already own.¹⁹

That investors are indeed limited in their ability to take time to seek information is also shown by the result of Dellavigna and Pollet's study. They found that investors' attention is more distracted on Fridays than on other days of the week.²⁰ As a consequence, they are unable to absorb all the information.

To summarize these claims retail investors often trade in a suboptimal way, lowering their expected returns by acting excessively. Unlike individual investors, rational investors recognize the limitations of buying stocks. They know that attention-based information may be included in the price.²¹

Since it is now established that financial attention is associated with the search for appropriate information and that this is dependent on many individual factors, input is now based on various measures of attention.

Investor attention can be measured by using different proxies. Indirect proxies contain advertising expenses²², the number of analysts following²³, trading volume, 1-day returns, and media attention measured by the news.²⁴ But there is no evidence that all these information is used by investors. Indirect proxies have the assumption that investors paid attention when the return or turnover changed dramatically or the company was mentioned in the news. This requires ensuring that the investor really reads the news and that there are no other factors influencing the return at the same time.²⁵ Another criticism of indirect proxies is that causality cannot be inferred unambiguously.²⁶

A direct proxy for investor attention is Google search volume. Da et al. applied SVI in their paper for a variety of reasons. One reason is that users use a search engine mainly to gather information. Among all search engines, Google represents the largest share of users. Another relevant reason is that when someone searches for a particular search term directly, a certain level of attention can be assumed. Thus, the aggregate search frequency in Google is a direct and unique measure of attention.²⁷ Another direct proxy for investor attention, especially in China, is the Baidu Index.²⁸ Further notable direct proxies would be the logins of investor online accounts²⁹ and Yahoo Finance.³⁰

2.2 Direct measurements of attention

Since the advantages of direct proxies have already been explained, this chapter examines three direct proxies in detail.

2.2.1 Search Volume Index

Google publicizes the SVI of search terms via the Google Trends. More accurate Google Trends was implemented in 2006 and includes search volume data since 2004 for different languages and different regions. In addition, Google Trends data is normalized. Thus, for each data point, it is divided by the total number of searches in the corresponding geographic area and time period. This determines the levels of relative demand. Otherwise, locations with the highest search volume would always rank at the top. Afterwards, on a scale from 0 to 100, the results appear that are aligned in comparison to all search queries for all topics. The data of Google Trends reflects the search queries that users make on Google very detailed. The reason is that Google filters out certain search queries. This happens, for example, because Trends only analyzes data for popular search terms and terms with low search volume therefore appear as 0. It also removes repeating searches from the same person within a short period oftime.³¹

SVI can be extracted as a CSV file, which allows data to be obtained and processed in a user-friendly way. Figure 1 illustrates an example for the search term "Donald Trump". The figure shows the evolution of the search volume index for the search term "Donald Trump" over the last 90 days. On the y-axis are the values for SVI's and on the x-axis the time. There is a spike on November 7, 2020 and January 7, 2021. That increased number of searches for the term can be attributed to the end of the presidential election. On November 7, 2020, the US media declared that Biden has won. The spike on 07 January 2021 can be traced back to the blocking of Trump's Twitter account after the storming of the Capitol.³²

[...]

¹ Da/Engelberg/Gao (2011), p. 1461.

² Cf. https://www.statista.com/statistics/242606/number-of-active-twitter-users-in-selected- countries/

³ Cf. Anger/Kittl (2011), p. 1.

⁴ Cf. Tafti/Zotti/Jank (2016), p. 15.

⁵ Cf. Da/Engelberg/Gao (2011), p. 1462.

⁶ Cf. Da/Engelberg/Gao (2011), p. 1497.

⁷ Cf. Gehde-Trapp/Mohapatra (2020), p. 317.

⁸ Cf. https://nirkaissar.com/trumps-tweets-arent-stock-tips/

⁹ CfWells/Shaha/ Lukito/Pelled/Pevehouse/Yang (2020), p.666.

¹⁰ Cf. Zhang/Shen/Zhang/Xiong (2013), p. 613.

¹¹ Cf. Barber/Odean (2008), p. 785.

¹² Cf. Sicherman/Loewenstein/Seppi/Utkus (2016), p. 864.

¹³ Cf. Corwin/Coughenour (2008), p. 3032.

¹⁴ Cf. Brans/Scholtens (2020), p.2.

¹⁵ Cf. Zhang/Shen/Zhang/Xing (2013), p. 613.

¹⁶ Cf. Ehrlich/Guttmann/Schoenbach/Mills (1957), p.99.

¹⁷ Cf. Karlsson/Loewenstein/Seppi (2009), p. 96.

¹⁸ Cf. Ibid., p. 99.

¹⁹ Cf. Barber/Odean (2008), p. 813.

²⁰ Cf. Dellavigna/Pollet (2009), p. 744.

²¹ Cf. Barber/Odean (2008), p. 790.

²² Cf. Yan/Chemmanur (2009), p. 64.

²³ Cf. Irvine (2003), p.449.

²⁴ Cf. Barber/Odean (2008), p. 788.

²⁵ Cf. Da/Engelberg/Gao (2011), p. 1462.

²⁶ Cf. Gehde-Trapp/Mohapatra (2020), p. 308.

²⁷ Cf. Da/Engelberg/Gao (2009), p. 1462.

²⁸ Cf. Zhang/Shen/Zhang/Xiang (2013), p. 614.

²⁹ Cf. Sicherman/Loewenstein/Seppi/Utkus (2016), p. 892.

³⁰ Cf. Lawrence, Ryans, Sun, Laptev (2016), p.2.

³¹ Cf. https://support.google.com/trends/answer/4365533?hl=de&ref_topic=6248052

³² Cf. https://www.bbc.com/news/world-us-canada-55597840