Political uncertainty and market illiquidity : evidence from the Brexit referendum

(1)

Political Uncertainty and Market Illiquidity:

Evidence from the Brexit Referendum

By Derk-Jan Verhaak, 10384340

Master’s Thesis Finance, Quantitative Finance, Msc_FIN June 2018

Abstract

In this thesis, I give additional evidence on political uncertainty and its effects on illiquidity. For this, the Brexit referendum is used as an uncertain exogenous event. Different events related to this referendum and multiple categories for the companies are used to see how illiquidity changes over time and per category. These categories are multinational or national companies and companies that either need a high-skilled workforce or not. For this research, I analyse 422 companies of the London Stock Exchange for the period 2013 – 2016. Results indicate that there is a significant difference in effects found between national and multinational companies when looking before and after the referendum. The outcome of the difference suggests that national companies are more severely affected by the Brexit referendum than multinational companies. Next, other results reveal that an increase in illiquidity could only be seen after the referendum. Other events either had negative effects on illiquidity, implying a decrease in market uncertainty, or had no significant effects at all. Additionally, companies separated by the need of a high-skilled workforce do not show significant distinct effects, both in general and around the events.

(2)

Statement of Originality

This document is written by student Derk-Jan Verhaak who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are 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 contents.

(3)

Table of Contents:

Abstract... 1

Table of Contents... 3

I. Introduction... 4

II. Literature... 6

III. Theoretical framework... 9

IV. Fundamental background of the event... 11

V. Data source and sample... 12

VI. Methodology... 14

A: Variables B: Empirical Model VII. Results... 16

A: General changes in illiquidity B: Changes per category C: Changes per event VIII. Discussion... 30

IX. Conclusion... 31

References... 33

(4)

1) Introduction

In June 2016, The United Kingdom (UK) voted in a referendum to leave the European Union (EU), an event often referred to as the Brexit. This decision would result in the first country in history to withdraw from the EU since its formation. The effects of a Brexit could have

profound effects on both the UK and the EU: The UK holds, even before the EU was formed, strong economic relations with a number of the current member nations (Sandholtz & Sweet, 1998). Next, the UK is a long-established member of the EU and therefore has close policy- and economic ties with other member states. Numbers confirm this strong relationship; Ottaviano et al. (2014) show that over half of all the UK exports go to the rest of the EU and they argue that leaving the EU would likely impose substantial costs on the economy of the UK. Next, they argue it is difficult to predict the consequences and state that multiple scenarios should be considered concerning the severity of the effects.

Multiple scholars discuss this expected severity caused by the Brexit. Ramiah et al. (2016) find early-on evidence by examining the stock reaction after the referendum. They find that most sectors reacted negatively to the outcome, with the banking sector even showing an accumulative abnormal return of -15.37%. This negative reaction of the stock market on the Brexit referendum outcome is found by multiple scholars, who show that the event of the Brexit is significant to market participants (Bouoiyour & Selmi, 2018; Amewu et al., 2016). Besides these points, the Brexit outcome is not a predicted event, making it an exogenous shock (Cox & Griffith, 2018). All these factors together make the Brexit referendum a one-of-a-kind uncertainty shock with unprecedented effects. As these events are rare, relatively little research has been done about the effects of this political uncertainty. Therefore, in this thesis, I seek to utilise the Brexit

referendum to test these effects, if any, of political uncertainty on market illiquidity, a much-used measure related to uncertainty (Pasquariello & Zafeiridou, 2014; Ben-Rephael, 2017; Cox & Griffith, 2018).

The main objective of this thesis is to deliver new evidence on political uncertainty, its effects on illiquidity and explore new fields related to these variables. These new fields are represented via research on multiple events that are related to the Brexit and via certain

categories on which companies are separated. These categories are respectively multinational- and national companies compared and companies that require a high-skilled workforce compared with companies that require a low-skilled workforce. By analyzing it in this manner, I show how market uncertainty affects a company’s illiquidity for various timelines and categories.

This research analyses 422 stocks of the London Stock Exchange. Via event studies on several happenings that are related to the Brexit, the general effect on illiquidity is measured.

(5)

Next, a differences-in-differences estimation is used to seek the difference in effect between the multiple categories.

Studies about illiquidity and (political) uncertainty generally show that they are positively related (Asteriou & Siriopoulos, 2000; Irshad, 2017; Cox & Griffith, 2018; Gao & Qi, 2013). These studies normally cover elections or other events of which the consequences can be compared to other relatable events. In these studies, however, the Brexit has not yet been thoroughly examined, and the related events that I aim to research have never been researched before. Next, scholars merely focus on overall illiquidity or, sometimes, on illiquidity per sector. Other characteristics, such as being a multinational, have not yet been included in any research. By testing both on multiple events related to the referendum and several different characteristics of companies, I aim to make the first step towards a distinct view on uncertainty and its

consequences for various companies.

The main findings show that there is indeed a difference between categories: A significant difference in effects is found between national and multinational companies; this shows that national companies are more severely affected by the Brexit referendum than multinational companies. The companies separated by the need of a skilled workforce do not show any differences in illiquidity. Next, results find that the referendum itself has an immediate positive effect on illiquidity while the other events show either negative or neutral immediate effects. This indicates that, merely after the referendum uncertainty increases with the corresponding effects on the market. Lastly, illiquidity does return to pre-event levels again after a certain period. These findings are in line with the literature and theoretical framework. Namely, multinationals are much more flexible than national companies, as they can reconsider their place of business and exploit exchange rates (Rangan, 1998; Cumming & Zahra, 2016). Concerning the timing effects regarding illiquidity, it is logical that merely after the referendum an increase in illiquidity is immediately observable as this result is accompanied by a surprise negative outcome, which increases information asymmetry (Cox & Griffith, 2018).

Section 2 continues with the literature section. This section surveys the literature and empirical studies, discussing papers from previous years concerning information asymmetry and (political) uncertainty. Other sections include a theoretical framework, data source and sample, methodology, results, discussion and lastly, a possible conclusion.

(6)

2) Literature:

This literature section gives an overview of existing literature on both information asymmetry and (political) uncertainty. Hereby, the effects of these aspects regarding company performance is discussed. Next, it is elaborated how any sectorial differences between companies affect the companies’ stock.

In times of uncertainty, not all information is universally known. According to Zhang et al. (2013), this could lead to information asymmetry, which translates into a situation in which one of the parties in a transaction possesses superior information compared to the other party involved. From a market’s perspective, full liquidity potential can merely be reached when information is transparent and obtainable by all market participants. In times of uncertainty, however, this is merely not the case. Information asymmetry arises as some parties possess a greater material knowledge than other parties, which results in increased market illiquidity. Other scholars also show the positive relationship between illiquidity and information asymmetry stating that it has a significant effect on each other (Minović, 2016; Lambert et al., 2011; Tetlock, 2010; Duarte & Young, 2009).

Zreik & Louhichi (2017) examine information asymmetry by analysing the words that are used in an annual report. They relate this to the concepts information asymmetry, a market’s risk sentiment, and illiquidity to provide new insights concerning these topics. They find that uncertainty regarding information transparency leads to increased illiquidity. Also, they find that if the information that is shared by a company is both transparent and sufficient, illiquidity rates are generally lower, which again shows the positive relationship between illiquidity and information asymmetry. Besides these points, they discover a negative correlation between the measure of risk and the liquidity of stocks, showing that higher risk sentiment increases illiquidity too. This relationship between sentiment and liquidity has been confirmed by several other scholars (Liu, 2015; Degennaro et al., 2008).

Regarding political uncertainty, studies generally find a negative relationship between political- instability and uncertainty and stock- and bond prices (Asteriou & Siriopoulos, 2000; Irshad, 2017; Pástor & Veronesi, 2012; Gao & Qi, 2013). Asteriou & Siriopoulos test the effects of uncertain socio-political conditions on the general Greek index, the Athens Stock Exchange, and find that a higher socio-political instability correlates negatively with both economic growth and the stock market development. Even after controlling for factors associated with the stock market development, their results remain significant. Pástor & Veronesi (2012; 2013) state that political uncertainty increases the risk premium which results in declines in stock prices, higher volatility and increased correlation between stocks. Also, they find that, when the measured

(7)

political uncertainty is higher, the effects of uncertainty become more severe (Pástor & Veronesi, 2013). These findings are concurring with the results that scholars find when looking at

information asymmetry; namely, an uncertain political event causes the market to hold

insufficient information, leading to information asymmetry. Information asymmetry again leads to the several consequences that were noted earlier, including increased illiquidity. In addition, Jens (2017) discovers the field of uncertainty more thoroughly as he looks at the investments that companies make when in a case of political uncertainty. In his event study, he uses U.S.

gubernatorial elections as an exogenous variation in uncertainty. His results include a 5 to 15 percentage points decrease of overall investments of firms. The somewhat same results are found by Akey & Lewellen (2017), as they show that, as a result of an uncertain political event, firms that are more sensitive towards government policies experience larger consequential changes in value relative to less-sensitive firms. The researched variables vary from firm value and operating performance to a company’s investments and the stock’s volatility. These findings concur with other scholars’ research who research the relationship between political uncertainty and, for example, volatility. (Chau et al., 2014; Buraschi et al., 2014; Liu & Zhang, 2015). These papers could indicate that certain companies experience an increase in risk compared to other companies, which shows that characterizations of companies could be significant for other researches as well.

Other scholars extend this field of research by including the observable changes in illiquidity caused by political uncertainty. Scholars find that, generally, illiquidity increases when an uncertain event, such as an election, approaches (Pasquariello & Zafeiridou, 2014). However, in the months after the event, they actually find an increase in both trading volume and liquidity. According to hem, this effect argues for the existence of information asymmetry. Besides this, Ben-Rephael (2017) finds that, in periods of extreme market uncertainty, mutual funds reduce their aggregate holding in illiquid stocks. This phenomenon is caused by retail investors who withdraw their positions, with more emphasis on illiquid stocks. These findings correspond with the general hypothesis that illiquidity increases in times of uncertainty. Other scholars also research illiquidity and relate it to events such as the Brexit Referendum, the 2016 U.S.

Presidential Elections or even natural disasters (Cox & Griffith; 2018, Rehse et al., 2018; Chung & Chuwonganant, 2012). They generally find an increase in illiquidity in the period shortly after the event but do, again, not find persistent changes in illiquidity as a result of (political)

uncertainty.

Overall, uncertain events influence the markets in different ways. Though it may sometimes be ambiguous whether and in what way uncertainty affects illiquidity, there are still

(8)

some conclusions that can be made regarding this relationship. First of all, uncertainty and a lack of information generally cause illiquidity to increase, at least in the short term. Secondly, after a certain amount of time, markets generally return to their pre-event levels as more information becomes available. Lastly, different companies can react differently to the same uncertain events.

(9)

3) Theoretical framework:

In this theoretical framework, I highlight the thesis’ main focus, expectations, and theory regarding illiquidity and uncertainty. In this process, the research question and hypotheses are also formulated. Lastly, the concept of uncertainty is expanded.

Research on uncertain political events usually covers the whole market. This is done even though multiple scholars find, when zooming into distinct groups, different results per sector, showing the significance of differentiating groups (Jens, 2017; Akey & Lewellen, 2017). They find differences in effects for multiple factors such as a stock’s expected return or volatility. Because of these findings, I aim to add two groups that are expected to be affected differently by the events related to the Brexit. As was said in the introduction, this thesis looks at companies that are either multinational companies or not and at companies that are either in need of a highly skilled workforce or not. If these characteristics indeed influence the effects that the referendum has, then this is a first step in exploring the effects of these and other company characteristics on illiquidity. I formulate the following research question:

Research Question:

!" $ℎ&' $&() *+* 'ℎ, +--+./+*+'( 01 2,3'&+" )'024) 0" 'ℎ, 50"*0" 6'024 782ℎ&"9, 2ℎ&"9, &) & 3,)/-' 01 'ℎ, :3,8+' ;0', +" 'ℎ, <,3+0* =,'$,," 2013 − 2016?

Multiple scholars research which sectors would be affected most by the Brexit. Examples of these sectors are logistics, banks and financial services, real estate, and technology (Tielmann & Schiereck, 2016; Bouoiyour & Selmi, 2018; Schiereck et al., 2016; Ramiah et al., 2017). However, none of these scholars categorised these companies by its access to the international markets, even though this could significantly impact the effects of Brexit that are experienced by these companies. Related to this, Cumming & Zahra (2016) state that various multinationals are reconsidering the UK as a place of business; namely, political fluctuations could (indirectly) influence their profit margins. This statement is corresponding with a paper by Rangan (1998) who researches the flexibility of multinational companies. He finds that multinationals systematically exploit currency shifts to their advantage. Especially in this time of increased competitiveness, such flexibility is important. It could, in practice, result in companies crossing borders to avoid political and foreign-exchange risks. As multinational companies can easier change their place of business, they would quicker do so when the Brexit would threaten their profits. This would suggest that the effects of the Brexit are less severe for multinational companies, as they can change their place of business relatively easy if needed. Next, their

(10)

international characterisation makes them less vulnerable to country-specific risks. All points above create the following hypothesis that this thesis will research:

Hypothesis 1:

Eℎ, )'024 +--+./+*+'( 01 "&'+0"&- 20F<&"+,) 0" 'ℎ, 50"*0" 6'024 782ℎ&"9, +"23,&),) )+9"+1+2&"'-( F03, 'ℎ&" 'ℎ, )'024 +--+./+*+'( 01 F/-'+"&'+0"&- 20F<&"+,) +" 'ℎ, <,3+0* =,'$,," 2013 − 2016

&) & 3,)/-' 01 'ℎ, :3,8+' 3,1,3,"*/F.

Besides the researched multinationals in this thesis, another defined category is a company’s need for high-skilled workers. According to multiple scholars, technology became more advanced in the last centuries, which created a shift in labour as the demand for high-skilled workers became increasingly higher (Bresnahan et al., 2002; Falk & Biagi, 2017). To respond to this higher demand, countries import a part of their highly skilled workforce, creating competition for skilled employees between various developed countries (Bauer & Kunze, 2004). Countries’ policies could play a key role in attracting this workforce, as they are related to the possibilities and opportunities that these workers get. The Brexit referendum could influence these policies, and it could create a higher risk for the foreign workforce. These factors could lead high-skilled workers to pick another country instead of the UK. This would then again influence the companies in the UK that need the particular high-skilled workforce to be operative. Because of that, the Brexit could increase the uncertainty that these companies face, resulting in higher expected illiquidity. The following hypothesis can be formed:

Hypothesis 2:

Eℎ, )'024 +--+./+*+'( 01 20F<&"+,) +" ",,* 01 ℎ+9ℎ )4+--,* $034,3) 0" 'ℎ, 50"*0" 6'024 782ℎ&"9, +"23,&),) )+9"+1+2&"'-( F03, 'ℎ&" 'ℎ, )'024 +--+./+*+'( 01 20F<&"+,) +" ",,* 01 -0$ )4+--,* $034,3)

+" 'ℎ, <,3+0* =,'$,," 2013 − 2016 &) & 3,)/-' 01 'ℎ, :3,8+' 3,1,3,"*/F.

The third and last hypothesis concerns time-specific factors and is elaborated in the next section ‘Fundamental background of the event’. For now, risk and uncertainty are distinguished, as these are, according to multiple scholars, two things that need to be treated differently (Rehse et al., 2018; Knight, 1921). Knight (1921) researches this distinction and states that risk applies to situations where one does not know the outcome of a given situation but can measure the odds of certain outcomes. Uncertainty, on the other hand, refers to situations where not all the information is known to set accurate odds in the first place. Regarding the Brexit, I argue that this event has a significant uncertainty component. Namely, the event has no comparable events and has relatively many unpredictable consequences. This results in somewhat incalculable odds for this event, making it an uncertain event.

(11)

4) Fundamental background of the event:

Besides certain selected characteristics of companies, I also look at the timeline of the Brexit referendum. Namely, to see when exactly illiquidity changes, multiple events related to the Brexit referendum should be analysed. In that manner, the effect can be fully shown. For this, an official briefing paper by the House of Commons is used (Walker, 2018). A summary with the relevant dates is shown below. See Table 1.

Table 1: Dates & corresponding events related to the Brexit referendum

Source: Walker, 2018

Of these events, I select the following events by date: First, the main event, the Brexit referendum, is analysed, which translates into the 23rd_{of June, 2016. Then, three dates before the}

Brexit are analysed. These are 23 January 2013; 14 April 2015; and 17 December 2015. The first event is the announcement of the goodwill of Prime Minister Cameron for an in-out referendum on the UK’s membership of the EU. The second event is the receipt of a Royal Assent. Third and last, the date that Prime Minister Cameron confirmed the referendum to be on the 23rd_of

June, 2016 is used. Each date represents an event on which the referendum became more certain to happen. A hypothesis concerning these different events can be stated as well. Because the referendum outcome was a surprise, information asymmetry likely only occurs after the referendum, causing illiquidity to increase merely after the referendum as well. This is formulated with the following hypothesis:

Hypothesis 3:

Eℎ, )'024 +--+./+*+'( 01 20F<&"+,) 0" 'ℎ, 50"*0" 6'024 782ℎ&"9, 0"-( 2ℎ&"9,) )+9"+1+2&"'-( +" 'ℎ, *&() &1',3 'ℎ, 3,)/-') 01 'ℎ, :3,8+' H,1,3,"*/F &"* "0' &1',3 &"( 0'ℎ,3

(12)

5) Data source and sample:

For the analysis, I use stocks of the London Stock Exchange. Via the official site of the London Stock Exchange, all the listed companies are collected. The original dataset holds 2,026

companies. As not all these companies are used for this dataset, some adjustments have to be made, even before extracting the data. First, to dispose this research of ambiguity about the effects of Brexit, solely companies that are incorporated in the UK are analysed. Therefore, it drops companies that are merely listed on the London Stock Exchange, but incorporated somewhere else. An example of this is Toyota Motor Corporation, which is incorporated in Japan, making the effect of Brexit on this company relatively ambiguous.

511 data points deleted – Companies that are merely listed on the London Stock Exchange but are not incorporated in the UK.

The next adjustment concerns the market capitalisation; a very low market capitalisation will most definitely result in less trading going on, making the stock vulnerable for relatively small changes in volume. Therefore, companies with a market capitalisation below a benchmark of 10 million British Pound are dropped from the dataset. In this way, enough companies remain in the dataset, but companies with the lowest trade volumes are dropped from the dataset and will not make the results unnecessarily biased.

238 data points deleted – Companies that have a market capitalisation of 10 million British Pound or less.

As was stated in the theoretical framework, sectors have to be picked to analyse the effects I aim to research. Some sectors are also dropped as it is unclear whether they need high or low-skilled workers. The selected remaining sectors are Automobiles & Parts, Banks, Basic Resources, Chemicals, Construction & Materials, Financial Services, Food & Beverage, Health Care, Insurance, Oil & Gas, Personal Goods & Household Goods, Real Estate, Retail,

Technology, and Travel & Leisure. Below one can see which group fits in which category according to the need for low- or high-skilled workers. See below how each sector is assigned.

High-skilled workers needed most Low-skilled workers needed most Banks, Chemicals, Financial Services, Health

Care, Insurance, Oil & Gas, Real Estate, and Technology

Automobiles & Parts, Basic Resources, Construction & Materials, Food & Beverage, Personal Goods & Household Goods, Retail, and Travel & Leisure

(13)

275 data points deleted – Companies that are assigned to a sector which creates ambiguity about low- or high-skilled workers.

Of the companies that are left over, the 115 largest companies per discussed category are extracted, resulting in 460 companies in total to be analysed. The data for these 460 companies is obtained from Thomson DataStream, an Economic database that provides over 10 million economic time series for 162 markets. From here, daily data of a company’s price and turnover volume are obtained for all 460 stocks for the period between the start of 2013 until the end of 2016. After all the data is gathered, there is still some missing data: As companies are not always listed for a long period, there are some data points that are not stock listed yet at a certain time. Because of this, companies that are not stock listed from the 23rd_{of June, 2015, are dropped}

from the dataset.

38 data points deleted – Companies that lack data because they were not stock listed yet at the 23rd_{of June,}

2015.

A total of 422 companies are left over. These companies are split into four groups, either multinational or not, and either in need of high-skilled workers or not. The data is reshaped to get an observation for every company for every date. In this process, weekend days and holidays are also dropped from the dataset. In this way, every company gets a separate timeline, which gives this thesis the possibility to analyse the data correctly.

(14)

6) Methodology:

A. Variables

Multiple measures are used to analyse the effects of market uncertainty correctly. First of all, I elaborate on the thesis’ illiquidity measure, which is the dependent variable in the models. The illiquidity measure that was first introduced by Amihud (2002) is chosen as an illiquidity measure in this thesis. This measurement has the following formula:

!55!I_JK = 1 M_JK∗ HJKO PQ5M_JKO RST KUV

Here, !55!I_JK is the measure of illiquidity, H_JKO is the return on stock + on day * of year '. PQ5MJKO is the corresponding daily volume in dollars. Thus, the ratio of the absolute value of

HJKO over PQ5MJKO shows the absolute price change per dollar of daily trading volume. Lastly,

M_JK is the number of days for which data are available for stock + in year '. Illiquidity is used by multiple scholars as a measure of risk, as these two indicators are positively and significantly related (Amihud et al., 2006; Amihud et al., 2015; Hagströmer et al., 2013; Daley & Green, 2016). The relation between uncertainty and illiquidity is tested in this thesis. Scholars state that this relation creates the so-called illiquidity premium which shows that more illiquid stocks have a higher expected return (Chan & Faff, 2005). This indicates that market participants do include illiquidity in their investment decisions and is of importance.

Besides an illiquidity measure, I create multiple variables for every analysed event. These variables include various time variables, which indicate the number of days that have passed after an event. These are made to seek how long the estimated effects in illiquidity would last. Also, variables that indicate when a company was affected by the chosen ‘treatments’, either being a national company or being in need of a high-skilled workforce, are created. Lastly, multiple interaction variables are created between the time-dependent and treatment-dependent variables. In this manner, one can see how a particular event affects the treated companies differently than non-treated companies.

B. Empirical Model

I use two empirical strategies to test my hypotheses. First, I perform multiple event studies. With this, I show the general effect of multiple events that are related to the Brexit on illiquidity. Second, I estimate a differences-in-differences (DiD) model. A DiD method is handy here as it addresses both cross-sectional and time-dependent variation, which would normally cause a bias. By using a DiD, the effect of the treatment is isolated, meaning the different effects caused by

(15)

the Brexit referendum on different companies can be shown. For the DiD method, there are two control groups: These are multinationals and companies that are generally in need of low-skilled people. By examining these categories on multiple events, the results show when exactly a difference in illiquidity is found. Also, the results show how different companies are differently affected per event. With this all said, the following model can be constructed:

!--+._J,K = X_Y+ X_V:3,8+'_K∈[Y]+ X_^:3,8+'_{K∈[V,_]}+ X_`:3,8+'_K∈[a,VY]+ X_b:3,8+'_{K∈[VV,V_]} + X_{_}c&'+0"&-_J + X_a64+--_J+ X_d!"'c&'+0"&-_J+ X_e!"'64+--_J + f_J where !--+.J,K is the measure of illiquidity. :3,8+'K∈[Y;V,_;a,VY;VV,V_] are dummy variables that are

equal to one if the date is in the specified interval. For example, the numbers 1 and 5 mean the interval is between one and five days after the event. Going on, c&'+0"&-J and 64+--J are

dummy variables for respectively whether a company is a national company or not and whether a company requires high-skilled workers or not. !"'c&'+0"&-J are the interaction variables that

connects the time-dependent variables with the treatment group, which is the national companies in this case. !"'64+--J are the interaction variables between time-dependent variables and

companies in need of a high-skilled workforce. Lastly, fJ is the error term.

When looking at the model above, every indicator or control factor is included. There are also some alterations tested where some indicators are included, and some are excluded.

(16)

7) Results:

In this section, the hypotheses that were stated and whether political uncertainty has an impact on market illiquidity are tested. This results section is structured as follows: First, a general figure of illiquidity around the Brexit is shown per category. Then, general market illiquidity is

researched around the Brexit referendum and other events. After this, results are more

concentrated and are researched per category to spot changes between the mentioned categories around the referendum and other events. In that way, it is observable how illiquidity changes per event and per category.

Figure 1 shows the market illiquidity around the Brexit referendum. The tables describe illiquidity per category. Respectively clockwise, these are all companies, high skill &

multinationals, low skill & multinationals, high skill & nationals, and low skill & nationals. As these categories are researched in this thesis, this figure could give insight on the analyses. Looking at the figure that observes all the companies, one can see a spike of illiquidity on the 24th

of June, 2016, which is the event date. This spike, however, is relatively small and it does not particularly draw the attention. The illiquidity per category, described in the other tables, might be of more value here. As one can see, multinationals seem to experience, illiquidity-wise, little to no effect on the referendum date; in the 30 days observed, all multinationals remain in a 5-percent range. When one looks at the national companies in the dataset, spikes, though small, can be observed. It does not seem that the illiquidity is higher at the event date, though this would have been expected. More research is needed on this. Other helpful insights that are gathered from this figure are that national companies typically have higher illiquidity than multinationals. Skill, on the other hand, does not hand a clear conclusion yet. See below.

(17)

Figure 1: Market illiquidity around the Brexit referendum.

As can be seen below, illiquidity is measured 15 days before and 15 days after the Brexit referendum. See below for the results per category.

(18)

A. General changes in illiquidity:

With the help of Figure 1, it could be seen that plotted differences of the Brexit are relatively small. Therefore, Table 1 discusses the general differences in illiquidity around the Brexit referendum, when controlled for fixed effects. It will, for now, solely look at time-specific variables and their effect on illiquidity. These time-specific variables are named ‘Referendum Date - 0, [1,5], [6,10], [11,15], and [16,20]’. As can be seen in Table 1, the time-dependent variables are significant for small ranges. The first regression, for example, shows a significant effect of 0.004, meaning that on the event date, illiquidity was estimated to be 0.004 higher. This translates, with an average illiquidity of 0.274, into an increase of 0.146 percentage point. Other results that cover larger timeframes also show that illiquidity remains significantly higher after the particular event date. However, after a certain time, the illiquidity increase becomes less

significant. As can be seen in regression 8, all the effects except the last time variable are not significant. These results show that, after a certain time, illiquidity is not significantly different anymore. As was found by other scholars, illiquidity seems to return to its pre-event levels after a certain period. See table 1 below.

Next, tables 2, 3, and 4 describe the changes in illiquidity observable for the other events, which were described in the topic 4. As can be seen, the results differ per table. Table 2, in particular, shows highly negative and significant coefficients. That result is not corresponding with hypothesis 3, which states that illiquidity merely changes around the Brexit. Next, the sign of the coefficients is remarkable, as it indicates a decrease in illiquidity after the event, implying more certainty in the market. However, this increase in market certainty could be explained by many factors, as market participants might saw this happening as an opportunity instead of a risk. Tables 3 & 4 show less significance and show either positive coefficients or coefficients that are almost equal to zero. This respectively shows that illiquidity became either higher after the event, or that no real effects are observable. See tables 2, 3, and 4 below.

(19)

TABLE 1: Analysis of changes in market illiquidity: Brexit Referendum REGRESSION RESULTS I

.

Dependent variable: Illiquidity

1 2 3 4 5 6 7 8 Referendum 0.004*** _0.004 *** _0.004 _0.003 _0.002 _-0.001 Date 0 (0.001) (0.001) (0.002) (0.002) (0.003) (0.003) Referendum 0.005*** _0.006*** _0.006*** _0.007*** _0.006*** _0.005*** _0.002 Date [1,5] (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.002) Referendum 0.005*** _0.005*** _0.002 Date [6,10] (0.001) (0.001) (0.002) Referendum 0.006** _0.000 Date [11,15] (0.003) (0.002) Referendum -0.003* Date [16,20] (0.002) Observations 798 1,599 1,599 3,187 3,187 6,390 10,775 17,534

These estimations cover different timelines varying from 2 days-range to 40 days-range to analyse the effect on these different ranges.

Because the estimates are relatively low (as illiquidity is a low measure), all estimations have been multiplied by 100. To control for this, simply divide by 100. All estimations are controlled for fixed effects

The numbers in brackets and italic are the standard errors.

(20)

TABLE 2: Analysis of changes in market illiquidity: Premier Cameron agrees with a referendum REGRESSION RESULTS I

.

1 2 3 4 5 6 7 8 Event -0.039 -0.030 -0.030 -0.050 -0.144 -0.140 Date 0 (0.024) (0.038) (0.047) (0.069) (0.138) (0.130) Event -0.055** _-0.062** _-0.091*** _-0.096*** _-0.147*** _-0.237*** _-0.227*** Date [1,5] (0.029) (0.030) (0.028) (0.029) (0.001) (0.064) (0.060) Event -0.212*** _-0.287*** _-0.285*** Date [6,10] (0.051) (0.064) (0.060) Event -0.278*** _-0.274*** Date [11,15] (0.100) (0.002) Event -0.333*** Date [16,20] (0.002) Observations 659 1,994 1,994 3,332 3,332 6,013 10,030 13,381

(21)

TABLE 3: Analysis of changes in market illiquidity: Royal Assent REGRESSION RESULTS I

.

1 2 3 4 5 6 7 8 Event -0.001 0.001 0.001 0.000 0.001 0.001 Date 0 (0.001) (0.001) (0.002) (0.002) (0.003) (0.003) Event 0.001 0.001 0.001 0.001 0.003** _0.003* _0.003* Date [1,5] (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.002) Event 0.011*** _0.012*** _0.013*** Date [6,10] (0.002) (0.001) (0.002) Event 0.013*** _0.012*** Date [11,15] (0.002) (0.002) Event 0.009*** Date [16,20] (0.002) Observations 781 1,560 1,560 3,131 3,131 6,195 10,127 14,068

(22)

TABLE 4: Analysis of changes in market illiquidity: Brexit Referendum became certain REGRESSION RESULTS I

.

1 2 3 4 5 6 7 8 Event -0.000 -0.001 -0.000 0.000 0.001 0.003 Date 0 (0.001) (0.001) (0.001) (0.002) (0.002) (0.003) Event -0.000 -0.000 0.000 -0.000 _0.000 _0.002 _0.004*** Date [1,5] (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.002) Event -0.001 -0.000 0.002 Date [6,10] (0.001) (0.001) (0.002) Event -0.003* _0.001 Date [11,15] (0.002) (0.002) Event -0.000 Date [16,20] (0.002) Observations 783 1,564 1,564 3,126 3,126 6,279 10,226 17,326

(23)

B. Changes per category:

Now that I showed the general effects of illiquidity, results per category are shown; in Table 5, one can see the DiD estimations between national and multinational companies and in Table 6, the DiD estimations between companies in need of a high- or a low-skilled workforce are shown. First, Table 5 is discussed.

Two types of indicators are included in the regressions: First, a variable for the treatment is created named ‘Treatment National’. This indicates that a company is a national one instead of a multinational one. Next, multiple interaction variables are added between the time-dependent variables and the treatment variable. As can be seen, the treatment effect is significant for the 1-% level and positive for every analysis. This effect concurs with Figure 1, where it was shown that illiquidity was generally higher for national companies than it was for multinational

companies. Also, it can be seen that the interaction variables are positive and significant for the 1-% level for every regression that is done. These interaction variables show how the mean differences between the treated and non-treated group change after the event has passed. As it is positive, it shows that the effects of the event are more severe for national companies than they are for multinational companies. This was predicted in hypothesis 1 as well. One last remarkable finding is that the interaction variable remains positive and significant, indicating the difference in effect does not go back to pre-event levels, which was the case for overall illiquidity. See Table 5 below.

Table 6 is similar to Table 5, except for the fact that the variables included in this table are for companies that require a high-skilled workforce. As can be seen, both the effects of the treatment and the interaction variables lack significance. Overall, the treatment effect seems to be negative indicating that companies in need of a high-skilled workforce have, in general, lower illiquidity. However, the 95-% interval still includes positive values, meaning no convincing conclusions can be made out of this. The interaction variables are almost equal to zero, especially when standard errors are taken into account. Therefore, the effects are negligible, indicating that the referendum outcome has no difference in effect between the particular treated and non-treated groups. See Table 6 below.

(24)

TABLE 5: Differences-in-Differences estimation: Brexit Referendum REGRESSION RESULTS II

Dependent variable: Illiquidity Multinationals & Nationals

1 2 3 4 5 6 7 8 Referendum [Date 0] 0.001 (0.001) (0.001) 0.002 (0.002) -0.000 (0.003) -0.002 (0.004) -0.003 (0.004) -0.005 Referendum [Date 1, 5] 0.000 (0.001) 0.001 (0.001) -0.000 (0.001) -0.000 (0.002) -0.002 (0.001) -0.003 (0.002) -0.005** (0.002) Referendum [Date 6, 10] -0.002 (0.002) (0.002) -0.003 -0.005 ** (0.002) Referendum [Date 11, 15] -0.005 (0.004) -0.009 *** (0.002) Referendum [Date 16, 20] -0.012*** (0.002) Treatment National 2.928*** (0.426) 3.033*** (0.401) 3.030*** (0.397) 3.032*** (0.397) 3.030*** (0.395) 3.175*** (0.396) 3.153*** (0.381) 3.160*** (0.364) Referendum [Date 0] * Treatment National 0.008 *** (0.002) 0.008 *** (0.002) 0.010 *** (0.003) 0.010 *** (0.004) 0.010 ** (0.003) (0.007) 0.008 Referendum [Date 1, 5] * Treatment National 0.009 *** (0.002) 0.012 *** (0.002) 0.015 *** (0.002) 0.017 *** (0.002) 0.017 *** (0.002) 0.016 *** (0.003) 0.014 *** (0.003) Referendum [Date 6, 10] * Treatment National 0.015 *** (0.003) 0.019 *** (0.003) 0.017 *** (0.003) Referendum [Date 11, 15] * Treatment National 0.025 *** (0.005) 0.020 *** (0.003) Referendum [Date 16, 20] * Treatment National 0.019*** (0.003) Observations 798 1,599 1,599 3,187 3,187 6,390 10,775 17,534

Because the estimates are relatively low (as illiquidity is a low measure), all estimations have been multiplied by 100. To control for this, simply divide by 100. All estimations are controlled for random effects

(25)

TABLE 6: Differences-in-Differences estimation: Brexit Referendum REGRESSION RESULTS II

Dependent variable: Illiquidity High Skill & Low Skill

1 2 3 4 5 6 7 8 Referendum [Date 0] 0.005*** (0.001) 0.004 *** (0.001) 0.005 ** (0.002) (0.003) 0.004 (0.004) -0.003 (0.004) 0.001 Referendum [Date 1, 5] 0.004*** (0.001) 0.005*** (0.001) 0.007*** (0.002) 0.008*** (0.002) 0.006*** (0.002) 0.006*** (0.002) 0.003 (0.002) Referendum [Date 6, 10] 0.005** (0.002) 0.008 ** (0.004) 0.005 ** (0.002) Referendum [Date 11, 15] 0.009 (0.004) (0.002) 0.002 Referendum [Date 16, 20] -0.004 (0.002) Treatment High-Skill 0.011 (0.449) -0.075 (0.450) -0.075 (0.446) -0.082 (0.449) -0.081 (0.447) -0.207 (0.462) -0.180 (0.460) -0.193 (0.462) Referendum [Date 0] *

Treatment High Skill (0.002) -0.001 (0.002) -0.000 (0.003) -0.001 (0.004) -0.001 (0.005) -0.003 (0.006) -0.003

Referendum [Date 1, 5] *

Treatment High Skill (0.002) -0.001 (0.002) 0.001 (0.002) -0.002 (0.002) -0.002 (0.002) -0.000 (0.003) -0.002 (0.003) -0.002

Referendum [Date 6, 10]

* Treatment High Skill (0.003) -0.002 -0.005

*

(0.003) (0.003) -0.005

Referendum [Date 11, 15]

* Treatment High Skill (0.005) -0.005 (0.003) -0.003

Referendum [Date 16, 20] * Treatment High Skill

0.001

(0.003)

Observations 798 1,599 1,599 3,187 3,187 6,390 10,775 17,534

(26)

C. Changes per event

Besides the changes per category around the referendum, changes in illiquidity are researched for the different stated events that are related to the referendum. The following tables indicate this, as they analyse three major events before the Brexit. As was mentioned in topic 4, the first event represents the announcement of Prime Minister Cameron of him being in favour of an in-out referendum on the UK’s membership of the EU. The second and third events represent the receipt of a Royal Assent and the date that Prime Minister Cameron announced the referendum date to be on the 23rd_{of June, 2016. Merely the treatment regarding national companies is shown}

here, as no significant differences were found regarding skill when looking at the Brexit

referendum itself. The tables for skill can be found in appendix tables A.1, A.2, and A.3 with an elaboration.

Table 7 shows the effects on illiquidity around the announcement of Prime Minister Cameron being in favour of the Brexit referendum. As can be seen, the results are somewhat the same except for the interaction variable. The interaction variable is in almost none of the cases, with a few exceptions, significant. This translates into no significant changes in the mean

difference between treated and non-treated companies when looking before and after the event. That is quite logical: Market participants look at the risk and odds of certain events; as the referendum was still vague and far away, the Brexit had a relatively small change of actually happening, meaning no real differences in effect should be observable.

Tables 8 and 9 show the next events, which are the Royal Assent and the announcement of the referendum date. In both analyses, the interaction variables are insignificant, leading to the same results as table 4. This means that, until the actual outcome of the Brexit referendum, the events related to the Brexit did not have significantly different effects between multinational and national companies. See the tables below.

Overall, the results are in line with the expectations. The main findings can be found in Table 5, which shows that national companies experience a higher estimated change in illiquidity than multinational companies after the Brexit referendum, indicating that national companies are more heavily affected by the referendum than multinational companies. Also, the other tables find that the mean difference in illiquidity between national and multinational companies does not change after the observed events. This shows when the actual change in illiquidity between national and multinational companies could be seen, which is merely after the referendum and not around the events before the referendum.

(27)

TABLE 7: Differences-in-Differences estimation: Premier Cameron agrees with a referendum REGRESSION RESULTS III

1 2 3 4 5 6 7 8 Event [Date 0] -0.037 (0.031) (0.005) -0.004 (0.061) -0.040 (0.089) -0.063 (0.180) -0.180 (0.168) -0.176 Event [Date 1, 5] -0.020 (0.037) -0.003 (0.004) -0.045 (0.036) -0.052 (0.037) -0.088** (0.044) -0.217*** (0.084) -0.223*** (0.078) Event [Date 6, 10] -0.105 (0.066) -0.249 *** (0.084) -0.256 *** (0.078) Event [Date 11, 15] -0.267** (0.131) -0.264 *** (0.078) Event [Date 16, 20] -0.243*** (0.078) Treatment National 3.938*** (0.690) 4.083*** (0.698) 4.079*** (0.697) 4.178*** (0.692) 4.174*** (0.686) 4.234*** (0.698) 4.046*** (0.667) 4.657*** (0.264) Event [Date 0] * Treatment National (0.005) -0.004 (0.008) 0.014 0.025 *** (0.096) (0.140) 0.032 (0.282) 0.083 (0.264) 0.009 Event [Date 1, 5] * Treatment National (0.006) -0.009 (0.006) -0.008 -0.113 ** (0.056) -0.109 * (0.058) -0.144 ** (0.069) (0.130) -0.049 (0.122) -0.-10 Event [Date 6, 10] * Treatment National -0.262 ** (0.069) (0.131) -0.091 (0.123) -0.069 Event [Date 11, 15] * Treatment National (0.020) -0.029 (0.123) -0.020 Event [Date 16, 20] * Treatment National -0.213* (0.122) Observations 659 1,994 1,994 3,332 3,332 6,013 10,030 17,534

(28)

TABLE 8: Differences-in-Differences estimation: Royal Assent REGRESSION RESULTS III

1 2 3 4 5 6 7 8 Event [Date 0] -0.000 (0.001) (0.002) 0.001 (0.002) -0.000 (0.003) 0.001 (0.004) 0.002 (0.004) 0.003 Event [Date 1, 5] 0.000 (0.002) 0.000 (0.002) -0.000 (0.001) -0.000 (0.001) 0.002 (0.002) 0.003** (0.002) 0.005** (0.002) Event [Date 6, 10] 0.008*** (0.002) 0.013 *** (0.002) 0.014 *** (0.002) Event [Date 11, 15] 0.016*** (0.003) 0.017 *** (0.002) Event [Date 16, 20] 0.015*** (0.003) Treatment National 2.989*** (0.450) 3.072*** (0.440) 3.071*** (0.440) 3.191*** (0.436) 3.191*** (0.432) 3.194*** (0.411) 3.173*** (0.406) 3.174*** (0.389) Event [Date 0] * Treatment National (0.001) -0.002 (0.003) 0.002 (0.003) 0.003 (0.005) -0.001 (0.006) -0.004 (0.006) -0.005 Event [Date 1, 5] * Treatment National 0.001 (0.003) (0.003) 0.002 (0.002) 0.003 (0.002) 0.003 (0.002) 0.002 (0.003) -0.002 (0.003) -0.004 Event [Date 6, 10] * Treatment National 0.007 ** (0.003) (0.003) -0.001 (0.003) -0.003 Event [Date 11, 15] * Treatment National (0.004) -0.007 (0.003) -0.012 Event [Date 16, 20] * Treatment National -0.011** (0.005) Observations 781 1,560 1,560 3,131 3,131 6,195 10,127 14,068

(29)

TABLE 9: Differences-in-Differences estimation: Brexit Referendum became certain REGRESSION RESULTS III

1 2 3 4 5 6 7 8 Event [Date 0] 0.000 (0.001) (0.001) 0.000 (0.002) 0.001 (0.002) 0.001 (0.003) 0.001 (0.004) 0.002 Event [Date 1, 5] 0.000 (0.001) 0.000 (0.001) 0.001 (0.001) 0.001 (0.001) 0.001 (0.001) 0.002 (0.001) 0.003 (0.002) Event [Date 6, 10] 0.001 (0.002) (0.001) 0.001 (0.002) 0.001 Event [Date 11, 15] 0.001 (0.002) (0.002) 0.001 Event [Date 16, 20] 0.002 (0.002) Event National 2.906*** (0.397) 3.002*** (0.427) 3.002*** (0.414) 3.065*** (0.427) 3.065*** (0.424) 3.230*** (0.414) 3.229*** (0.419) 3.227*** (0.375) Event [Date 0] * Treatment National (0.002) -0.000 -0.002 (0.002) (0.003) -0.002 (0.003) -0.001 (0.004) 0.001 (0.007) 0.002 Event [Date 1, 5] * Treatment National -0.000 (0.002) (0.002) -0.001 (0.002) -0.001 (0.002) -0.002 (0.002) -0.002 (0.002) -0.000 (0.003) 0.002 Event [Date 6, 10] * Treatment National (0.003) -0.004 (0.002) -0.003 (0.003) -0.000 Event [Date 11, 15] * Treatment National (0.003) -0.008 (0.003) -0.002 Event [Date 16, 20] * Treatment National -0.005 (0.003) Observations 783 1,564 1,564 3,126 3,126 6,279 10,226 17,326

(30)

8) Discussion:

The analyses of this thesis showed how illiquidity changes after an uncertain political event. For this, the Brexit was used as an exogenous uncertainty shock. Interesting and expected

conclusions that support the already existing literature were found, but also remarkable results that were not necessarily expected were found. Possible issues that are related to the analysis are discussed here.

The first possible problem arises when looking at omitted variable biases; though this thesis controlled for fixed and random effects, there could still be other variables not controlled for that influence illiquidity when looking at the differences between categories. An example of this could be size; on average, multinational companies are much bigger than national companies. Maybe size is increasingly related to illiquidity around the Brexit referendum which would then partly explain the found differences in illiquidity. Though it is unlikely that the results would show drastic adjustments because of this, it should still be acknowledged.

Another point worth noticing is the dataset. On the London Stock Exchange, there are relatively many companies that have a low market capitalisation. Though this thesis controlled partly for this bias by dropping companies with a market capitalisation of 10 million British pounds or less, there were still companies with a market capitalisation that would be considered low. Because of this, relatively small changes in volume can cause a company’s volume to change with a high degree. That could make the research not generalizable.

Next, whether companies need a high-skilled workforce or not was based on sectors. However, in practice, not every company is the same and companies that are, for example, in the financial services sector can vastly differ in their needs for high-skilled workers. Therefore, the dataset could have been too broad, which resulted in no significant results found for the proxy high-skill.

Lastly, Bertrand et al. (2004) discover DiD estimates and state that, because of serial correlation, standard errors may be underestimated, leading to over-estimation of t-statistics and significance levels. However, the significance levels that were found in this thesis are relatively high, making it unlikely that this argument would change the thesis’ results.

Overall, the results that were found are trustworthy as it was controlled for several effects that could influence the model. The focus of interest and the main result, which is the difference in effect between multinational and national companies as a result of the referendum, is highly significant and isolated of effects that could influence it.

(31)

9) Conclusion:

This study contributed to the literature of political uncertainty and market illiquidity by examining how market illiquidity is affected by multiple events related to the Brexit referendum. I found that illiquidity increased significantly in the first couple of days after the Brexit referendum and returned to pre-event levels after that. Also, the announcement of premier Cameron being in favour of an in-out referendum decreased the illiquidity indicating a decrease in market uncertainty, which is not fully in line with the expectations. The other events do not directly show any significant differences in illiquidity, indicating that market participants did not react significantly to them.

The main results of this thesis were that national companies experience a more severe effect by the referendum than multinational companies do, which is concurring with hypothesis 1. Even when looking at a large timeframe, the difference in effects remained positive and significant, indicating that nationals are affected systematically higher by uncertain events. That is remarkable, as the general increase in illiquidity did not show a systematic increase after the referendum. It also shows that different characterisations of companies could directly influence how a company gets affected by events that might cause uncertainty. Next, no significant differences in effect are found between multinational and national companies when looking at other events related to the referendum. Apparently, the anticipation on the referendum itself did not cause the spike in illiquidity which could merely be seen after the unexpected outcome was recognized. This could indicate that expectations of the referendum were different than the results, causing no increase in illiquidity around the other events.

The results regarding the other category, whether a company requires a high-skilled workforce or not, were not significant. This shows that this variable might not be as important as expected for the illiquidity of the selected companies. Hypothesis 2 can, therefore, be rejected.

Lastly, hypothesis 3 regarding the timing of changes in overall illiquidity can be rejected, as there were multiple events before the Brexit referendum that showed a significant change in overall illiquidity after the particular event. However, when looking at the selected categories, merely changes could be seen around the Brexit referendum.

Overall, the regressions were relatively successful as the results presented new evidence on illiquidity and, in particular, on how certain events can influence different companies in various ways. It became clear that not only sectors but also other characteristics are important for the effects on illiquidity being measured. Out of the results that were found, two suggestions can be made for future research. Firstly, as expectations were different for the Brexit referendum than the outcome, market illiquidity sometimes showed little difference in the events related to

(32)

the Brexit. However, if the event had the same expectations as the outcome, would the effects on illiquidity show more quickly? And would events, that make the ultimate event more likely to happen, also experience increases in illiquidity? This could give more insight in how predictions and adaptations are formed in the market. The second suggestion concerns the proxy high-skill where no real effects were found. As was stated in the discussion, companies can differ vastly but still be in the same sector. Therefore, scholars should focus on looking at companies individually to estimate whether they need a high-skilled workforce or not. If it is selected per company, the high-skilled proxy is less vulnerable to biases, making the analyses more reliable.

(33)

References:

Akey, P., & Lewellen, S. (2017). Policy uncertainty, political capital, and firm risk-taking. Amewu, G., Mensah, J. O., & Alagidede, P. (2016). Reaction of global stock markets to

Brexit. Journal of African Political Economy and Development, 1(1), 40-55.

Amihud, Y. (2002). Illiquidity and stock returns: cross-section and time-series effects. Journal of

financial markets, 5(1), 31-56.

Amihud, Y., Hameed, A., Kang, W., & Zhang, H. (2015). The illiquidity premium: International evidence. Journal of Financial Economics, 117(2), 350-368.

Amihud, Y., Mendelson, H., & Pedersen, L. H. (2006). Liquidity and asset prices. Foundations and

Trends in Finance, 1(4), 269-364.

Asteriou, D., & Siriopoulos, C. (2000). The role of political instability in stock market

development and economic growth: The case of Greece. Economic Notes, 29(3), 355-374. Bauer, T. K., & Kunze, A. (2004). The demand for high-skilled workers and immigration policy. Ben-Rephael, A. (2017). Flight-to-liquidity, market uncertainty, and the actions of mutual fund

investors. Journal of Financial Intermediation, 31, 30-44.

Bertrand, M., Duflo, E., & Mullainathan, S. (2004). How much should we trust differences-in-differences estimates?. The Quarterly journal of economics, 119(1), 249-275.

Bouoiyour, J., & Selmi, R. (2018). Are UK industries resilient in dealing with uncertainty? The case of Brexit. The European Journal of Comparative Economics.

Bresnahan, T. F., Brynjolfsson, E., & Hitt, L. M. (2002). Information technology, workplace organization, and the demand for skilled labor: Firm-level evidence. The Quarterly Journal of Economics, 117(1), 339-376.

Buraschi, A., Trojani, F., & Vedolin, A. (2014). When uncertainty blows in the orchard:

Comovement and equilibrium volatility risk premia. The Journal of Finance, 69(1), 101-137. Chan, H. W., & Faff, R. W. (2005). Asset pricing and the illiquidity premium. Financial

Review, 40(4), 429-458.

Chau, F., Deesomsak, R., & Wang, J. (2014). Political uncertainty and stock market volatility in the Middle East and North African (MENA) countries. Journal of International Financial

Markets, Institutions and Money, 28, 1-19.

Chung, K. H., & Chuwonganant, C. (2012). Uncertainty, Fear, and Liquidity. Shared Growth and

Sustainable Development, 352.

Chung, K. H., & Chuwonganant, C. (2014). Uncertainty, market structure, and liquidity. Journal of

(34)

Cox, J., & Griffith, T. (2017). Political Uncertainty and Market Liquidity: Evidence from the Brexit Referendum and the 2016 US Presidential Election.

Cumming, D. J., & Zahra, S. A. (2016). International business and entrepreneurship implications of Brexit. British Journal of Management, 27(4), 687-692.

Daley, B., & Green, B. (2016). An Information-Based Theory of Time-Varying Liquidity. The

Journal of Finance, 71(2), 809-870.

DeGennaro, R. P., Kamstra, M. J., & Kramer, L. A. (2008). Does risk aversion vary during the year? Evidence from bid-ask spreads.

Duarte, J., & Young, L. (2009). Why is PIN priced?. Journal of Financial Economics, 91(2), 119-138. Falk, M., & Biagi, F. (2017). Relative demand for highly skilled workers and use of different ICT

technologies. Applied Economics, 49(9), 903-914.

Gao, P., & Qi, Y. (2013). Political uncertainty and public financing costs: Evidence from US gubernatorial elections and municipal bond market.

Hagströmer, B., Hansson, B., & Nilsson, B. (2013). The components of the illiquidity premium: An empirical analysis of US stocks 1927–2010. Journal of Banking & Finance, 37(11), 4476-4487.

Irshad, H. (2017). Relationship Among Political Instability, Stock Market Returns and Stock Market Volatility. Studies in Business and Economics, 12(2), 70-99.

Jens, C. E. (2017). Political uncertainty and investment: Causal evidence from US gubernatorial elections. Journal of Financial Economics, 124(3), 563-579.

Lambert, R. A., Leuz, C., & Verrecchia, R. E. (2011). Information asymmetry, information precision, and the cost of capital. Review of Finance, 16(1), 1-29.

Liu, S. (2015). Investor sentiment and stock market liquidity. Journal of Behavioral Finance, 16(1), 51-67.

Liu, L., & Zhang, T. (2015). Economic policy uncertainty and stock market volatility. Finance

Research Letters, 15, 99-105.

Minović, J. (2016). Asymmetric information influence on efficiency of capital market. Economic

analysis, 164-179.

Ottaviano, G. I. P., Pessoa, J. P., Sampson, T., & Van Reenen, J. (2014). Brexit or Fixit? The trade and welfare effects of leaving the European Union.

Pástor, L., & Veronesi, P. (2012). Uncertainty about government policy and stock prices. The

Journal of Finance, 67(4), 1219-1264.

Pástor, Ľ., & Veronesi, P. (2013). Political uncertainty and risk premia. Journal of Financial

(35)

Pasquariello, P., & Zafeiridou, C. (2014). Political uncertainty and financial market quality. Ramiah, V., Pham, H. N., & Moosa, I. (2017). The sectoral effects of Brexit on the British

economy: early evidence from the reaction of the stock market. Applied Economics, 49(26), 2508-2514.

Rangan, S. (1998). Do multinationals operate flexibly? Theory and evidence. Journal of International

Business Studies, 29(2), 217-237.

Rehse, D., Riordan, R., Rottke, N., & Zietz, J. (2018). The effects of uncertainty on market liquidity: Evidence from Hurricane Sandy.

Sandholtz, W., & Sweet, A. S. (Eds.). (1998). European integration and supranational governance. OUP Oxford.

Schiereck, D., Kiesel, F., & Kolaric, S. (2016). Brexit:(Not) another Lehman moment for banks?. Finance Research Letters, 19, 291-297.

Tetlock, P. C. (2010). Does public financial news resolve asymmetric information?. The Review of

Financial Studies, 23(9), 3520-3557.

Tielmann, A., & Schiereck, D. (2017). Arising borders and the value of logistic companies: Evidence from the Brexit referendum in Great Britain. Finance Research Letters, 20, 22-28. Walker, N. (2018). Brexit timeline: Events leading to the UK’s exit from the European Union. Wang, J. (1994). A model of competitive stock trading volume. Journal of political Economy, 102(1),

127-168.

Zhang, W., Shen, D., Zhang, Y., & Xiong, X. (2013). Open source information, investor attention, and asset pricing. Economic Modelling, 33, 613-619.

Zreik, O., & Louhichi, W. (2017). Risk sentiment and firms’ liquidity in the French market. Research in International Business and Finance, 39, 809-823.

(36)

Appendix A. Skill proxy per category:

As there are no immediate interesting results obtained in the DiD estimates between companies that either require a high-skilled workforce or not, the results are discussed in this Appendix, as they are still worth noting.

As can be seen in Table A.1, the effects on illiquidity around the announcement of Prime Minister Cameron is shown. The results generally state a negative effect, of both the high-skill treatment and the interaction variable between high-skill and time. However, all these results are insignificant, meaning no real conclusions can be made here. When looking at the next events, in tables A.2 and A.3, almost the same can be seen except for one exception: As the time span gets larger, results of the interaction variable get more positive and significant. This would indicate that companies with a need of a high-skilled workforce were eventually affected by the analyzed events. While almost no effect was found during the Brexit referendum, other events show that illiquidity did change earlier in the timeline of Brexit. Overall, this adds to the literature that illiquidity could already change before the actual results have been found. As these events show an increased chance of the referendum happening, and therefore a higher risk on uncertainty, the results here show that market participants anticipate on this. Future research should focus on events that have a negative prediction. In that way, the anticipation on the expected negative impact could influence illiquidity more drastically than is the case right now.