The
effect
of the Brexit on firm value
Abstract In this thesis I analyse the impact of the Brexit referendum outcome
1. Introduction
On February 20th, 2016 Prime Minister David Cameron of the United
Kingdom (UK) announced the Brexit referendum to assess the general interest in leaving the European Union (EU). Against expectations a majority of the British public voted to leave the EU. The outcome caused quite a number of firms to evaluate their presence in the UK. On January 18th, 2017, both the Hongkong and
Shanghai Banking Corporation (HSBC) and Union Bank of Switzerland (UBS) announced they were deliberating on moving 1,000 jobs each from the UK if the Brexit plans persevered. This amounts to 20% of the positions both banks provide in the UK (Barbaglia, 2017). The Swiss bank UBS Group AG is considering Frankfurt for its trading headquarters, whereas HSBC is planning on moving its employees to Paris (Barbaglia, 2017; Finch et al., 2017). UBS is not the only firm appraising possible cities for relocation in Europe: EasyJet, Standard Chartered, Citigroup, and Lloyd’s Banking Group have all announced to relocate their headquarters once the UK leaves the EU. Apart from headquarter relocations and employee movement; the Brexit has other effects as well. The South African investment group Brait announced on the 24th of March this year
that it cancelled their listing plans on the London Stock Exchange (LSE). The well-known Swiss food manufacturer Nestlé plans to move its Blue Riband Biscuits production line abroad and the French bank Credit Agricole is in the process of moving its EU government bonds trading platform back to France as of October 2017.
Firm relocation, headquarter relocation, and outsourcing have become more common with the increase of globalization. Lowered coordination costs and improved information technology (IT) stability cause companies to move their firm, or parts of it, abroad for various reasons (Baldwin and Evenett, 2015). The move of headquarters however, is less common. Yet, this is what most firms called for following the outcome of the Brexit referendum: moving their headquarters and moving to other countries. Many multinational companies are aware of the uncertainty UK’s leave from the EU will cause. It is unknown at this point in time if tariffs on import and export will be introduced, and how the movement of persons across state borders will be arranged.
Because regulations, trade agreements, and other post-Brexit restrictions are still unknown, many firms plan their possible move of operations and/or headquarters out of the UK. Chubb (the US insurer) is moving its headquarters to Paris, Frankfurt is attracting banks, and other institutions decided to move to Luxembourg, Dublin, Brussels, and Amsterdam (Ralph and Chassany, 2017). Their plan is to reduce the risk they face in order to mitigate the effects following the Brexit on firm performance. Several studies analyse uncertainty and firm performance, but there appears to be less focus on headquarter relocation, and even less on the combination of the three. In this study I will bridge the gap between these studies by combining the political uncertainty of, to my knowledge, the unique event of the Brexit (a country leaving a bloc), with the effect on firm performance, and the effect of headquarter relocation announcements.
Many multinational companies are familiar with political uncertainty, but prior to the June 2016 referendum, it was not believed to highly impact firms in the UK. Uncertainty, as mentioned by Knight (1921), refers to risks of which the existence is known, but their probability of occurrence is not. Firms are less able to recognize cause-and-effect relationships due to uncertainty (Duncan, 1972). The effect of uncertainty on firm performance is often studied, but has different outcomes. Some state the negative effects on firm performance (Lorenzi et al., 1979), others indicate that it depends on how the firm handles it, passively or with a more proactive stance (Duncan, 1972; Carbonara and Caiazza, 2010). Studies show that firms do invest less during uncertain times. They delay or reduce investments that are irreversible, hampering growth (Alesina and Perotti, 1996; Rivoli and Salorio, 1996; Brunetti and Weder, 1998).
Firms tend to perform better when they take on a flexible stance throughout the company, making firms more capable to handle unexpected rises and drops in product demand (Merschmann and Thonemann, 2011). A focused strategy helps firms deal with this (Smith and Grimm, 1987). In this study I will examine how the market perceives in what way the firm will handle itself in uncertain times.
Since many multinational production facilities and headquarters are located in the UK, the Brexit will result in various issues. The “hard Brexit” (or “clean Brexit”) the prime minister is currently1 heading for will result in leaving
the single market and customs union of the EU. It would mean no more tax and tariff free trade between EU countries, and no more free movement of people (Kierzenkowski et al., 2016). For now, the UK is trying to discuss its own deal with the EU. However, if negotiations regarding Article 50 fail, it will have to adhere to tariffs the EU will impose on the UK as a member of the World Trade Organization (WTO) (Morris, 2017). For many firms these are arguments for reconsidering their presence in the UK, as most of their exports and imports are to and from the EU, and a significant amount of their employees are from the EU. Staying in the UK may damage profits, and thus many are looking to relocate within the EU. This results in the following research question:
How does political uncertainty caused by the announcement of the Brexit influence performance, and how does the relocation of headquarters affect these events.
method will be used to examine announcements on HQ relocations, other actions announced to mitigate the Brexit effects, and the announcement that no actions will be initiated. To test if the mean of the abnormal returns are significantly different from zero I conduct a one-sample student’s t-test for the short-term and a Johnson’s skewness adjusted t-test for the longer-term. I find support for hypothesis 1, the Brexit announcement has a significant (at the 1% level) negative effect on firm value, indicated by a CAR of -3.48% over the three days of the event, slowly decreasing to a CAR of -3.18% (also significant at the 1% level) over the 11-day event window. I find a slight negative effect on firm value as a result of the Article 50 announcement. For the longer term a negative BHAR is calculated of -3.79% 12 months post event. HQ relocation and other action announcements do not report any significant results; the report of a firm initiating no action however, generates a positive CAR and BHAR of 1.16% over the five-day window (p<0.1) and 2.06% (p<0.01) over the 30 days post event respectively.
The structure of the paper is as follows. Chapter 2 presents a literature review, on which hypothesis are based. The data and methodology are described in chapter 3, in 4 the results are discussed, 5 shows robustness test, and 6 the limitations and points for future research. Chapter 8 concludes. 2. Literature review and hypotheses 2.1 Political uncertainty
The unexpected outcome of the referendum made many international firms reassess their plans for the future. The UK is set on leaving the EU, but when and how is still unknown. Firms were unable to observe how this would affect their sales and any future changes in policy and regulation. Carbonara and Caiazza (2010) explain uncertainty as a situation that causes firms to have incomplete and imperfect knowledge of the current status of their business and its future. Unfortunately there is no single foreseeable instance of cause and effect on the firm: it is a multifaceted problem. As a result, variable management cannot prepare for this.
unpredictable, as they might be changed in the near future. This can lead to lower private investments and increase savings, disturbing a country’s economic growth (Alesina et al., 1996). Firms are hesitant to invest in activities that are less or completely irreversible, and prefer to delay or abstain from them all together (Pindyck, 1991; Rivoli and Salorio, 1996). Research has shown that investment is often reduced in the year leading up to a political event with an unpredictable outcome, thus impacting the performance of the firm (Julio and Yook, 2012).
Apart from policy change and private investment, there are often changes in public investments that impact firms. Darby et al. (2004) show that with additional political uncertainty, there is a shift from public expenditure to governmental consumption, being inefficient as a result. Besides, they also show biased effects on government rulings relating to public spending. Research on macroeconomic volatility and uncertainty shows that the ambiguity causes uncertainty in financial markets, increasing volatility in exchange rates. This has shown to have a negative effect on public investment as well (Brunetti and Weder, 1998).
Reduction in aggregate investment is not the only negative aspect of exchange rate volatility. The anxiety firms and investors have can be observed in different ways in the financial markets. Changes in exchange rates increase the need to hedge this exposure, especially for firms with international trade. Economic agents start to exhibit more risk averse behaviour, shown in the price increase of options that cover the duration of elections (Kelly et al., 2016), the increase in risk premia (Pastor and Veronesi, 2013), and changes in stock prices (Baker et al., 2016). This last one is thought to be caused by stock prices becoming less informative in politically uncertain times. Firms are unsure how new policies may affect them, so less knowledge will be taken from stock prices and more from cash flows. Unknown legislative plans may cause firms to rely more on outside financing than investors; therefore under the informational view of investment, cash flow becomes more important (Durnev, 2010).
patent system, valid in 26 countries. The Brexit will cause the UK to be excluded from this unified system (Spinks, 2016). Research has shown that uneasiness regarding the enforcement has an adverse effect on firm investment and performance (Brunetti and Weder, 1998).
This thesis will focus on the effect of uncertainty caused by the Brexit on firm performance. This event will most likely manifest itself in the various ways just discussed. Furthermore, I will study the uncertainty that is spurred on by the announcement invoking Article 50, as this, at this moment in time, has the largest foreseeable impact on future plans, policies, and regulations. For an overview of the various announcements see Table I below.
Table 1. Announcement dates
This table shows the dates and their corresponding announcements (Event) in relation to
the Brexit.
Date Event
20 February 2016 The UK prime minister David Cameron announces the Brexit
Referendum
24 June 2016 The outcome of the referendum is published, with majority of
votes in favour of leaving the EU.
29 March 2017 PM Theresa May signs letter to invoke Article 50, starting a hard
Brexit from the European Union.
08 June 2017 PM Theresa May calls for an election, hoping to strengthen her
position in the House of Commons. She did not increase her party’s seats, but went from 331 to 318, leaving her stance weakened.
08 December 2017 Brexit divorce bill announced to be between €40 bln and €45 bln.
and Caiazza, 2010). Furthermore, a proactive stance will help top management to anticipate uncertainty and find and maintain the balance between effectiveness and efficiency (Veliyath, 1992). Even though firms will do their best to stay ahead of all the changes, it remains a multidimensional problem, making it near impossible to oversee every factor (Carbonara and Caiazza, 2010).
It will come as no surprise that numerous descriptions exist in which way management should handle uncertainty. Firms that turn to a more innovative or contingent strategy outperform firms that follow a more unfocused strategy (Smith and Grimm, 1987). Showing more flexibility during turbulent times, especially in the supply chain, has proven to help firms navigate large oncoming changes (Martínez Sánchez and Pérez Pérez, 2005; Merschmann and Thonemann, 2011). Managers that network more during these moments, especially within the firm, add to a better performance as well (Sawyerr et al., 2003). The consensus is that firms that try to mitigate the effects during unstable times outperform firms that neglect taking action in one or multiple ways (Smith and Grimm, 1987; Sawyerr et al., 2003).
to communication and IT advances firms are becoming more fragmented. As some parts, such as production, are offshored, some of the essential roles or business unit headquarters move with it (Baldwin and Evenett, 2015). Birkinshaw et al. (2006) have also stated that a lot of variation exists between business unit HQs. Whereas corporate HQs are often relatively uniform in their make-up, a business unit HQ tends to be much smaller. The team represents the different occupations within that business unit and supporting roles. For the latter the size and scope tend to differ greatly from one unit to another. Large multinational firms often have a geographical HQ as well, which handles the specific issues in that region, and is more aware and able to adapt to cultural differences.
Previous studies on headquarter relocation often look at within-country relocations. These studies mainly focus on the spatial distribution of headquarters, how they are dispersed over cities within a country, and how much HQs are concentrated within a specific city (Baaij, van den Bosch, and Volberda 2004). International relocations bring a lot more uncertainty with them (Slangen and Beugelsdijk, 2010), and are often carried out due to pressures within the firm (Forsgren et al., 1995; Baaij et al., 2004; Brouwer et al., 2004; Birkinshaw et al., 2006). Furthermore, up until 2005 there were few international relocations within Europe as EU law restricted this (Baaij et al., 2004).
the new host country environment as more attractive. Birkinshaw et al. (2006) report strong evidence that this is a reason to move business unit HQs abroad. Chan et al., (1995) show that a headquarter relocation is often perceived as good news by investors. The decision is frequently seen as a positive net present value (NPV) project (Alli et al., 1991). This headquarter relocation does not necessarily have to be perceived as a flee from unfavourable conditions. Management may identify strategic assets, such as increased knowledge sharing, a better connection to international financial markets, and more expertise (Baaij et al., 2015). In today’s knowledge-economy these are crucial for a firm’s survival.
I conjecture that firms moving away from these uncertain variables as a way to mitigate their exposure will perform better than firms staying in the UK. Especially as the market will likely perceive this as an endeavour leading to better market conditions (Birkinshaw et al., 2006; Baaij et al., 2015). This leads to my second hypothesis:
H2: The announcement of a headquarter relocation will ameliorate the negative effect of political uncertainty on firm performance.
3. Data and methodology
3.1 Data
Barber and Lyon (1996), including these firms in the dataset would affect the calculation of the estimation window, thus firms listed for less than two years were excluded from the sample. The Europe Stoxx 600 includes firms from 17 different countries within the EU, including the UK. In this sample all but one country is present, Luxembourg did not have any firms with sufficient data. Furthermore, the UK is the best represented country within the sample, as the “Total Market UK” selection within DataStream is included in sample as well2.
For a visual representation of actual stock return and expected stock return see Figure 1 on the next page.
Figure 1. Stock return and expected return
This figure shows the stock return (in blue) and the corresponding expected return (y-axis) for all firms over the dates (x-axis) in the sample. Where the expected return is calculated using the market model, with the alpha and beta parameters estimated in the pre-event window.
Table 2. Firm distribution over countries
This table shows the distribution of the number of firms in the sample (Firms within sample), the percentage of firms coming from a country within the sample (%), and the corresponding market index used for that country (Local market index). The firms are obtained from the
Euro Stoxx 600 and the total market UK.
Country Firms within sample % Local market index Country Firms within sample % Local market index United Kingdom 531 55,78% FTAL Denmark 22 2,31% OMXCBGI
France 83 8,72% SBF 120 Finland 16 1,68% OMXHPI
Germany 72 7,56% DAX 30 Belgium 14 1,47% BEL 20
Switzerland 46 4,83% SMI Norway 12 1,26% OBX
Sweden 44 4,62% OMXSPI Ireland 9 0,95% IEOR
Italy 32 3,36% FTSE
Italia Austria 7 0,74% WBIEUR The Netherlands 31 3,26% AEX Portugal 3 0,32% PSI-20
Spain 28 2,94% IBEX 35 Czechia 2 0,21% FPXAA
Total 952 100,00%
To control for extreme outliers observed in the D/E, B/M, and P/E ratios I winsorize the data. These outliers may skew the results, however as I want to include all observations I did not omit them, and chose to set the values equal to the observation at the 97,5th percentile for all three, and to the 2,5nd percentile as
well for the D/E ratio. See Table 3 above for an overview of the descriptive statistics. 3.1.1 HQ relocation data Using newspaper articles and press releases I examine which firms were planning on relocating during or after the Brexit, and to what extent. During the early days after the referendum many firms made bold statements, stating they would pull half or even all of the positions out of the UK. A year later many firms have come back on these statements and reduced the amount of workers they
Table 3. Descriptive statistics
This table shows the descriptive statistics of the sample. Where Variables denotes the various variables in the sample, N is the number of firm observations, Mean shows the mean of the variable, St. Dev. its standard deviation, Min and Max show the lowest and highest observations respectively, and 25%, 50%, and 75% show the values at these percentiles. The variables are split in Firm characteristics where the market capitalization (Market cap) (in million Euro), the debt-to-equity ratio (D/E) (winsorized at the 5% level), book-to-market
(B/M), and price-to-earnings (P/E) (both winsorized at the 97,5th percentile) are given. For the
Stock characteristics the natural log of the stock return is given in percentage (Stock return),
and the intercept (Alpha) and slope (Beta) estimated using OLS to calculate the expected return are shown.
Variables N Mean St. Dev. Min 25% 50% 75% Max
plan to relocate. See for example cosmetics group l’Oréal Paris, which announced its reassessment of its stay in the UK in July 2016, putting the search for a new UK HQ on hold (Bloomberg, 2016), but has since renewed its search for a new building for its London HQ (Buckley, 2017). Many UK based companies are slowly starting to orient themselves on relocation post-Brexit. Several are considering setting up a European hub or subsidiary in the EU, or moving their EU HQs there. Frankfurt for banking, high frequency traders (HFTs) are looking at Amsterdam, and insurance firms are considering Luxembourg and Ireland. Yet none of the firms have shown any concrete plans to the outside world or to investors.
For the relocation announcements a sample of 11 firms, that have made a credible announcement of their plans to relocate their HQ out of the UK, is collected. In addition to relocation announcements I gather briefings made on other actions to mitigate the Brexit effects, indicated by 19 firms. For the purpose of robustness testing reports of firms indicating that the Brexit will not impact them are gathered, 22 firms report this. These dates, and the days around it, will be used to examine the impact of the announcements on the return of the firm.
Information on the firms in the initial sample on their plans to move their HQs is collected. I find that many firms attempt to remain as non-descript as possible regarding their Brexit plans. Using Reuters and Bloomberg I obtain 29 firms who confirmed their Brexit plans. These announcements were made either by the CEO or other members of the top management team, or by an official spokesperson of the firm, this to ensure it would be credible information and not speculation. Of these 29 firms, 11 made definite HQ relocation announcements. This sample consists of firms such as: EasyJet, Barclays bank, and Glencore. The remaining 18 firms made an announcement to move at least 3% of employees out of the UK, or announced other impacting actions such as Nestlé moving a production line.
3.2 Methodology
I use stock market prices at several days using an event study. I use the methodology as suggested by MacKinlay (1997) to identify the value effect of the Brexit announcement, as one can expect the market to respond rapidly. Studying the stock prices before and after the announcement of the event will give an indication of how the market responds to the uncertainty of the Brexit. Furthermore, by looking at the abnormal returns over a longer period one can identify if this was an event only around the announcement or if investors are more hesitant about future firm performance in the long run. On March 29th
2017 Theresa May confirmed the leave once more by putting the Article 50 negotiations in motion, starting the two-year departure process from the EU officially. Important dates in this research are thus, 24th of June 2016, when the results were announced, and 29th of March 2017. 3.2.1 Abnormal return Initially the interest is on in the effect of announcement of the referendum on the value of the firm. I want to study if investors perceive the announcement and subsequent actions related to the Brexit will harm the firm. Investors of a particular firm may decide to sell their stock if they expect the Brexit to negatively influence firm value, or perhaps signal to management the need to take action. This sale reduces the firm’s stock price. To examine if this transpires the cumulative abnormal return around the event window will be analysed. To begin I will calculate the stock return, 𝑅!!!"#$%#&"&'= ln 𝑝!" 𝑝!"!! , (1) where 𝑅!" is the continuously compounded return on the stock for firm 𝑖 at time 𝑡, 𝑝!" is the price of the stock for firm 𝑖 at time 𝑡, and 𝑝!"!! is the same price for firm 𝑖 but at time 𝑡 − 1. Next the return on the stock will be employed to estimate the abnormal return as follows: the difference between the actual and the expected return,
where 𝑎𝑟!" denotes the abnormal return for firm 𝑖 at time 𝑡, 𝑅!" is the previously mentioned actual return, 𝐸(𝑅!") is the expected (or normal) stock return for firm 𝑖 at time 𝑡. To estimate the expected return I will apply the market model as described by MacKinlay (1997). The market model is estimated using a linear regression with the stock return as the dependent variable and the return on the market as the independent variable, the linearity follows from the assumption of joint normal asset returns. The expected return is measured over the estimation window, which runs from 𝑇! to 𝑇! (see Figure 2 next page), where 𝑇! is at 120 days prior and 𝑇! is 40 days prior to the event date. There is no overlap between the event window and the estimation window. Overlap between the dates measuring the expected return and the event could cause the measured event to influence the expected return. The expected return using the market model is measured as follows,
𝐸 𝑅!" = 𝛼!+ 𝛽!"𝑅!" + 𝜀!" (3)
where 𝐸 𝑅!" denotes the expected return on the stock of firm 𝑖 at time 𝑡, 𝛼! and 𝛽!" are the intercept and slope parameters of the market model, estimated in the pre-event period, for firm 𝑖 for both parameters and at time 𝑡 for the slope (𝛽), 𝑅!" is the return on the market portfolio for a specific security at time 𝑡. For the return on the market I use the market index for the country the firm is registered in. For example, Royal Dutch Shell is a Dutch company, thus for the return on the market the Amsterdam Exchange Index (AEX) is used. Finally, 𝜀!" is the zero mean disturbance term, which is expected to be zero 𝐸 𝜀!" = 0 with a variance of 𝜎!! (𝑣𝑎𝑟 𝜀
!" = 𝜎!!).
As mentioned earlier, a linear regression, an ordinary least squares (OLS) regression is used to obtain the parameters of the expected return over the estimation window. After obtaining these parameter estimators they are employed to determine the expected return, which in turn is applied to calculate the abnormal return. For the formulas of these parameters, see Appendix A.
3.2.2 Cumulative abnormal return
There are multiple ways to employ the abnormal return in an event study. Both Ritter (1991) and Barber and Lyon (1996) discuss the use of cumulative abnormal return (CAR) versus the use of the buy-and-hold abnormal return (BHAR) on the longer term. This will be further discussed in paragraph 4.2. For the short term I will use the CAR. To analyse the announcement effect the three event windows surrounding the announcement are examined. The CAR denotes the aggregate abnormal return over this period, and can be expressed as follows, 𝐶𝐴𝑅!" = 𝑎𝑟!" ! !!! , (4)
where 𝐶𝐴𝑅!" describes the sum of the abnormal return for firm 𝑖 over period 𝑡. The event window is denoted by 𝜏, and ranges from 𝑡 = (−1, +1), 𝑡 = (−2, +2), or 𝑡 = (−5, +5).
I use the CAR to examine whether the actual return is statistically significantly different from the return one could have expected without the announcement. A student’s t-test is employed to analyse this. Thus, to examine whether the mean CAR is different from zero. 𝐻! in this case would be:
Figure 2. Timeline
This figure shows the timeline over which periods the variables are estimated. The alpha (intercept) and beta (slope) used to calculate the expected return are estimated over the estimation window, the CAR is estimated over the event window and the BHAR is estimated over the post-event window.
Estimationwindow windowEvent Post-event window
𝑇! 𝑇! 0
𝐻!: 𝐶𝐴𝑅 = 0
The equation for a one-sample student’s t-test is as follows,
𝑡!"# = 𝐶𝐴𝑅!"/(𝜎(𝐶𝐴𝑅!") 𝑛) (5)
where 𝐶𝐴𝑅!" denotes the sample mean, 𝜎(𝐶𝐴𝑅!") represents the sample standard deviation, and 𝑛 the number of observations (Barber and Lyon, 1996). The results will be presented in chapter 4. 3.2.3 Buy-and-hold abnormal return Both Ritter (1991) and Barber and Lyon (1996) state that the 12-month CAR is not equal to the annual BHAR, as a null hypothesis test of the 12-month CAR being zero is equal to a test of the monthly AR being zero, not a test of the annual AR being zero. The BHAR is defined as the buy-and-hold return in the firm minus the buy-and-hold expected return, and can be defined as follows, 𝐵𝐻𝐴𝑅!" = [1 + 𝑅!"] ! !!! − 1 + 𝐸 𝑅!" ! !!! , (6)
where 𝐵𝐻𝐴𝑅!" is the BHAR for firm 𝑖 over the event period 𝜏. This is estimated by the product of 1 + 𝑅!", which stands for one plus the return of the stock of firm 𝑖 at time 𝑡, minus the product of 1 + 𝐸(𝑅!"), which denotes one plus the expected return of firm’s stock 𝑖 at time 𝑡. The BHAR shows the actual return a holder of firm 𝑖’s stock would have received if the investor bought the stock and kept it for a period of time, minus the return one would have expected him to receive. I use the BHAR to examine if any abnormal returns are a temporary event that stabilizes in the months after the announcement, or if the effect on the firm value is lasting.
often present because of monthly rebalancing in compounded returns on reference portfolios but not on the sample firm.
As a reference portfolio I use the market portfolio. As stock returns are often positively skewed (Lyon et al., 1999) I adjust for this skewness by applying Johnson’s modified t-test. One may argue that the sample is sufficiently large to apply the central limit theorem (CLT), indicating the sample includes enough observations that their mean and finite standard deviation are approximately normal (Chen, 1995). However Jegadeesh and Titman (1993) demonstrate that momentum is often present in stock returns, with stocks which exhibited positive returns in the previous period outperforming stocks with previous negative returns. Therefore I apply Johnson’s skewness adjusted t-test, reducing the probability of a type I error (Chen, 1995).
To analyse the announcement effect the returns are studied. First from the event date to the post-event dates, ranging from 𝜏 to 𝑇! as shown in Figure 1, to observe if the abnormal return deviates from the expected return.
Then, as the market may overreact, the BHAR up to 12 months post-announcement is examined as well, as I conjecture the market to rebalance itself. To test whether one can infer that the sample mean is different from the expected mean I use a one-sample Johnson’s skewness adjusted t-test, with 𝐻! defined as: 𝐻!: 𝐵𝐻𝐴𝑅 = 0 The equation for the BHAR Johnson’s skewness adjusted t-test is as follows, 𝑡!"#$%& = 𝑥 − 𝜇! + 𝜇! 6𝑠!𝑁+ 𝜇! 3𝑠! 𝑥 − 𝜇! ! 𝑠! 𝑁 ! ! (7)
where 𝑡!"#$%& denotes the skewness adjusted t-test statistic, 𝑥 denotes the sample mean, 𝜇! is the zeroth central moment, or the total probability of the sample, 𝜇! the sample third central moment or its skewness, the BHAR standard deviation is shown as s, and 𝑁 indicates the number of observations (Johnson, 1978; Chen, 1995).
4. Results
I first analyse if any variables have unexplainable correlations, Table 4 below shows an overview of the correlation matrix. There is some positive correlation between the stock return and the market return, as the stocks are also included in their local market index. The correlation between the expected return and market return is to be expected, as the market return is used to calculate the expected return (see formula (3)). Alpha and beta influence the expected return as well.
4.1 Short-term value effects Brexit announcements 4.1.1 Referendum outcome
To continue, the CAR is calculated for the event window around the referendum outcome and the announcement of Theresa May invoking Article 50, declaring a hard Brexit. To continue the abnormal returns are calculated as indicated in formula (2) and summed (as indicated in formula (4)) to obtain the CAR. The outcomes indicate that there is enough evidence to reject 𝐻! and accept
Table 4. Correlation matrix
This table shows the correlation between the sample variables. The Stock return (𝑅!") and the
Market return (𝑅!) are the natural logs of the daily returns, the Expected return (𝐸(𝑅!")) is the expected daily return on the stock, calculated using the estimated Alpha (α) and Beta (β )
parameters using an OLS regression. The Market capitalization (𝑀!"#), Debt-to-equity (D/E),
Book-to-market (B/M), and Price-to-earnings (P/E) ratio are shown below as well.
𝐻!, meaning that one can infer that there is a difference between the actual return and the expected return3. As expected, the referendum outcome has a negative effect on firm value. The CAR shows a significantly lower mean return than what could have been the normal return had the announcement not taken place. The average stock return for a firm is reduced by an aggregated 3,48%, with a standard deviation of 8,78% over the three-day event window. The CAR dropped slightly over the five and 11-day event windows (from -3,24% to -3,18%) whereas the standard deviation increased from the five-day to 11-day window (8,57% and 10,79% respectively). The standard deviation, which shows the spread from mean of the sample, tells us that even though the mean CAR for the three day event window is -3,48%, significant variation might be present between firms. The increase from the five-day to the 11-significant variation might be present between firms. The increase from the five-day window can be explained by it simply being a longer time period, thus allowing the sample to deviate more.
To put this into perspective consider the following simplified example of the announcement effect on the firm value of a company. On the Europe Stoxx 600 exists a firm A. Firm A has 1.000.000 stocks outstanding, with a value of €250,- prior to the referendum outcome. So prior to the outcome A’s firm value is equal to €250mln (€250×1.000.000). In the three days surrounding the announcement firm A’s value decreased with the same percentage as the mean of the sample, leading to a stock price of €241,3 (€250× 1,00 − 0,0348 ), or a firm value of €241,3mln (€241,3×1.000.000), impacting the firm significantly. Since the median market cap of the sample is 3710,31mln, this would mean a decrease of 129,12mln for a firm of that value. For a visual representation of the average actual stock return and the average expected return around the event window see Figure 3 below.
Figure 3. Average return and average expected return around event
This graph shows the average return over 952 firms and the corresponding average expected return over all firms. Returns can be found on the y-axis and dates on the x-axis.
On the days surrounding the 24th of June a significant negative effect for UK firms
can be observed, with a CAR of -5,56% over the three-day event window, remaining above 5% over the other event windows as well, at the 1% level significance level. The rest of the EU does not experience this as significantly, with the three-day and five-day event CAR (both significant at the 1% level) being -0,85% and -0,75% respectively. The market seems to have rebalanced on the 11-day event window, with no significant gains or losses.
value effect, with a CAR of -1,48% over the first event window, and dropping afterwards.
Third, I split the sample based on the book-to-market ratio. I split the sample by the median again. A B/M ratio above one is said to be indicative of value stocks. As the book value of the firm is higher than the market value of the firm the stock is undervalued. A ratio below one could indicate that investors are willing to pay more for a stock than the value of its assets. It could also mean the firm does not possess many physical assets, such as technology firms. Overvalued firms seem to experience a larger negative effect on firm value than high B/M firms. The former group presents a CAR of -5,55% at the three-day window, whereas the latter a CAR of -1,63%, both significant at the 1% level. Next I analyse the sample based on a low price-to-earnings and high P/E ratio. A high stock price compared to earnings per share may indicate a future growth in a firm’s earnings. Undervalued firms or firms that performing better than past periods may have a low P/E ratio. Dividing a sample based on P/E ratio, however, is only slightly indicative, as the P/E ratio is best compared within industries. Results show that low P/E firms exhibit a significantly lower CAR (-6,04%) than high P/E firms (-1,32). All results can be found in Table 5 further on.
4.1.2 Article 50 announcement
I hypothesized a negative effect on firm value for the announcement by Theresa May that she signed the Article 50 clause, initiating a hard Brexit. There was a slight, but significant negative effect on the stock return in the three days surrounding this announcement. This effect however, stabilised during the longer event windows. There could be various reasons for this. The market may already have expected a hard Brexit, so this news only confirmed what was already accepted. There might have been investors considering this the best move, based on the belief that action should be pursued more aggressively, or simply because they have confidence in May’s ability to negotiate a good deal for the UK. Yet another reason might be that investors were less focused on this announcement, at least not as focused as they were on the referendum outcome. Results are presented in Table 6 further on.
Table 5. CAR referendum outcome
In this table the results of a one-sample student’s t-test are presented. I test if the CAR for the referendum outcome (24th of June 2016) is significantly different from zero. The intervals used are
(-1,+1), (-2,+2), and (-5,+5), Obs. denotes the observation of number of firms, the observed mean
CAR within the sample can be found under Mean, with the standard error in brackets below, and the corresponding t-statistic is shown under t-stat. The t-test is performed on the entire sample (All) first, I continue by splitting the sample in multiple ways. Below All you can find the sample split by countries (UK only and Rest of EU), split by market capitalization (Low market cap vs. High market
cap), B/M ratio (Low book-to-market vs. High book-to-market ratio), and P/E ratio (Low price-to-earnings vs. High price-to-price-to-earnings ratio). ***p<0,01, **p<0,05.
All Referendum outcome
Interval Obs. Mean t-stat Obs. Mean t-stat
(-1, +1) 952 -0,0348*** (0,0028) -12,2385 (-2, +2) 952 -0,0324*** (0,0028) -11,6753 (-5, +5) 952 -0,0318*** (0,0035) -9,0958 UK only Rest of EU (-1, +1) 531 -0,0556*** (0,0046) -12,0158 421 -0,0085*** (0,0021) -4,0939 (-2, +2) 531 -0,0522*** (0,0045) -11,6979 421 -0,0075*** (0,0023) -3,2986 (-5, +5) 531 -0,0547*** (0,0055) -9,9118 421 -0,0030 (0,0033) -0,9156
Low market cap High market cap
(-1, +1) 475 -0,0549*** (0,0044) -12,3599 477 -0,0148*** (0,0033) -4,4627 (-2, +2) 475 -0,0523*** (0,0043) -12,0616 477 -0,0013*** (0,0032) -3,9009 (-5, +5) 475 -0,0543*** (0,0055) -9,9155 477 -0,0094** (0,0041) -2,2910
Low book-to-market ratio (B/M) High book-to-market ratio (B/M)
(-1, +1) 450 -0,0555*** (0,0048) -11,5036 502 -0,0163*** (0,0030) -5,4306 (-2, +2) 450 -0,0517*** (0,0047) -10,9412 502 -0,0152*** (0,0029) -5,1754 (-5, +5) 450 -0,0550*** (0,0060) -9,2117 502 -0,0110*** (0,0037) -2,9941
Low price-to-earnings ratio (P/E) High price-to-earnings ratio (P/E)
Table 6. CAR Article 50 announcement
In this table the results of a one-sample student’s t-test are presented. I test if the CAR for the Article 50 announcement (29th of March 2017) is significantly different from zero. The intervals used are
(-1,+1), (-2,+2), and (-5,+5), Obs. denotes the observation of number of firms, the observed mean CAR
within the sample can be found under Mean, with the standard error in brackets below, and the corresponding t-statistic is shown under t-stat. The t-test is performed on the entire sample (All) first, I continue by splitting the sample in multiple ways. Below All you can find the sample split by countries (UK only and Rest of EU), split by market capitalization (Low market cap vs. High market
cap), B/M ratio (Low book-to-market vs. High book-to-market ratio), and P/E ratio (Low price-to-earnings vs. High price-to-price-to-earnings ratio). ***p<0,01, **p<0,05, *p<0,1.
All Article 50 announcement
Interval Obs. Mean t-stat Obs. Mean t-stat
(-1, +1) 952 -0,0012** (0,0006) -1,9321 (-2, +2) 952 0,0006 (0,0008) .7846 (-5, +5) 952 0,0010 (0,0013) .7699 UK only Rest of EU (-1, +1) 531 -0,0011 (0,0009) -1,2469 421 -0,0012* (0,0008) -1,6095 (-2, +2) 531 0,0004 (0,0012) 0,3578 421 0,0009 (0,0010) 0,8738 (-5, +5) 531 0,0011 (0,0019) 0,5871 421 0,0009 (0,0017) 0,5014
Low market cap High market cap
(-1, +1) 475 -0,0011 (0,0010) -1,1189 477 -0,0019** (0,0007) -1,8137 (-2, +2) 475 0,0010 (0,0014) 0,7339 477 0,0003 (0,0009) 0,3084 (-5, +5) 475 0,0012 (0,0020) 0,5944 477 0,0008 (0,0016) 0,4895
Low book-to-market ratio (B/M) High book-to-market ratio (B/M)
(-1, +1) 450 -0,0005 (0,0010) -0,5146 502 -0,0017*** (0,0007) -2,6134 (-2, +2) 450 0,0028** (0,0014) 1,9630 502 -0,0013* (0,0009) -1,3646 (-5, +5) 450 0,0043** (0,0022) 2,0104 502 -0,0020* (0,0015) -1,3707
Low price-to-earnings ratio (P/E) High price-to-earnings ratio (P/E)
4.2 Long-term value effects Brexit announcement As Barber and Lyon (1996) indicated, the CAR is not the ideal method for analysing returns over the longer term. The BHAR is more accurate as it is less subject to biases as new listing bias, compounding bias, and the skewness bias. To make a start each observation of the actual and expected return is increased by the value of one, ensuring that there will be no negative values. To calculate the product of each return and expected return a new variable is created. In this variable the total of the natural log non-negative returns for each firm is stored. I then replaced these natural logs with their exponent, creating the product of the return and expected return. As formula (6) indicates the expected return was subtracted from the actual return, resulting in the BHAR for each firm. In contrast to the CAR I run a one-sample Johnson’s skewness adjusted t-test to analyse if the mean BHAR is different from zero. This is the case for the first month, however not in the way I expected. I conjectured the market would stabilize itself after the initial shock had passed, but the opposite seems to be the case. The first and second months are the only month for which there is not enough evidence to reject 𝐻!, all other months, up to one year after the referendum outcome show a lasting, significant negative effect on firm value. It is interesting to note the insignificant but positive returns for the second month. Analysing the data for the second month does not show one impacting event, but positive returns over several weeks. No clear cause for this could be found. The outcome of the latter months support my first hypothesis, “H1: The outcome of the Brexit implies a negative effect on firm value”. An overview of the means and t-statistics for the BHAR can be found in Table 7 below.
4.3 HQ relocation results
4.3.1 Short-term HQ relocation announcement value effect
The announcement of a HQ relocation does not appear to result in a positive effect on firm value, however there is not enough evidence to reject 𝐻! thus I am unable to infer that the CAR is different from zero. Nonetheless, this finding may support the hypothesis that HQ relocation announcements mitigate the negative effect on firm value. A note of caution is due however, as the sample
Table 7. BHAR Brexit announcement
In this table the results for the Johnson’s skewness adjusted one-sample t-test, analysing if the BHAR is statistically different from zero, are presented for the post-event period of the
Brexit announcement (24th of June 2016). One month is counted as 30 days post
announcement, with results up to 12 months after the event. The BHAR is calculated as one plus the product of the stock return minus one plus the product of the expected return. The number of firm observations are found below Obs., the mean BHAR for a period under
Mean with the corresponding standard error between brackets, and the t-statistic is given by t-stat. The initial sample is split based on firms in the UK (UK only) and firms in other EU
countries (Rest of EU). ***p<0,01, **p<0,05, and *p<0,1.
All Referendum outcome
Period Obs. Mean t-stat Obs. Mean t-stat
As shown in Table 8, Glencore the natural resources producer and trader experienced the lowest CAR over the three-day event window, with -4,88%, whereas the Royal Bank of Scotland experience the highest CAR, experiencing a CAR of 2,95%. During the data-gathering phase I observed that many firms did not announce any plans, even though there was considerable speculation in the media. I anticipate that this is due to the significance such an announcement has on the firm as a whole. Part of this might be because many firms may still hope that the UK’s actual exit from the EU is going to fail. Theresa May is not standing as solid as she did at the initiation of the Brexit.
Table 8. CAR results and dispersion for HQ relocation announcements This table presents, on the left hand side, the one-sample student’s t-test results for an announcement of top management or a spokesperson on behalf of the firm stating the plans to relocate its HQ out of the UK. Where Interval denotes the event window, Obs. the number of firm observations, Mean the mean with its corresponding standard error in brackets below, and t-stat the t-statistic. On the right hand side the dispersion of the CAR over the three-day event window across firms can be found, with Mean and St. Dev. indicating the mean and standard deviation of the sample, followed by the Min, 25%, 50%, 75%, and Max indicating the
minimum, 25th, 50th, 75th, and maximum observed CARs. Firm reports the two firms
with the lowest and highest CAR.
Note: no significant t-test results were obtained.
Headquarter relocation CAR dispersion
Interval Obs. Mean T-stat Variable Firm
(-1,+1) 11 (0,0081) -0,0019 -.2303 Mean -0.0018 (-2,+2) 11 (0,0082) -0,0007 -.0857 St. Dev. 0.0267 (-5,+5) 11 (0,0094) -0,0079 -.8479 Min -0,0488 Glencore 25% -0,0029 50% 0,0032 75% 0,0010
Max 0,0295 Royal Bank
She had to negotiate a deal with a hard line Northern Irish party (Democratic Unionist Party (DUP)), who agreed to support her if, the border between Northern Ireland and Ireland becomes an actual border again. While May did indicate she would put effort into this, it currently seems as if she is trying to push it in such a way that the border between Northern Ireland and Ireland remains an open border, and thus undermining her position even more. This would form some way around import barriers and tariffs. Firms seem to be quietly getting their affairs in order while at the same time they are waiting until something actually happens, unless some concrete plans are made they contemplate to sit it out.
One reason for this biding is the lengthy negotiations. In 2020 the term of Theresa May will have ended, and elections may result in a new prime minister who has a different opinion on the Brexit. The first referendum has been won with a slight difference after all (51,9% voted leave, versus 48,1% voting remain) (Hunt and Wheeler, 2017). Furthermore one should take into account that some firms are so intertwined with their place of birth, that this would be considered an internal hurdle on a different level to relocate out of the UK. Take for example the UK luxury car manufacturer Rolls Royce; one cannot phantom them leaving the UK, or the Dutch beer breweries Heineken and Grolsch. Heineken is almost synonymous with Amsterdam, and whereas Grolsch did move from Grollo, they are still located in the region Twente in the Netherlands. Moving HQs from a place that, in a way, defines the company would have more impact on a firm than can be predicted beforehand (Voget, 2011).
described in paragraph 3.2.3, but up to three months instead of one year, as observations dropped too low after the third month. This is due to the announcements often being made half way through 2017. The results of the t-test on the BHAR being equal to zero are shown in Table 9 below. As with the CAR results there is not enough evidence to reject 𝐻!, thus I am unable to deduce that the BHAR is significantly different from zero. The results for the 60 and 90 days do show a slight positive mean, but as they are not significant at any of the customary significance levels I am unable to assume the positive effect the announcement has on firm value. This supports my second hypothesis, “H2: The announcement of a headquarter relocation will ameliorate the negative effect of political uncertainty on firm performance”.
Table 9. BHAR results and dispersion for HQ relocation announcements This table presents, on the left hand side, the one-sample Johnson’s skewness adjusted
t-test results for an announcement of top management or a spokesperson on behalf of
the firm stating the plans to relocate its HQ out of the UK. Where Interval denotes the event window, Obs. the number of firm observations, Mean the mean with its corresponding standard error in brackets below, and t-stat the t-statistic. On the right hand side the dispersion of the BHAR over the 30-day post-event window across firms can be found, with Mean and St. Dev. indicating the mean and standard deviation of the sample, followed by the Min, 25%, 50%, 75%, and Max indicating the
minimum, 25th, 50th, 75th, and maximum observed BHARs. Firm reports the two firms
with the lowest and highest BHAR.
Note: no significant t-test results were obtained.
Headquarter relocation BHAR dispersion
Interval Obs. Mean T-stat Variable Firm
The dispersion between returns is larger over the 30-day period post-announcement than it is over the three-day event window. Interestingly Glencore showed the highest BHAR (13,14%) over the 30 days post-event, even though it showed the lowest CAR on the three-day event window. The airline firm EasyJet reported the lowest results, with a BHAR of -13,95% over 30 days.
4.4 Other action results
4.4.1 Short-term value effect of other action announcements
The announcement of an employee move or production line move is less drastic than the announcement of moving headquarters. In the CAR results for the action announcements I observe a slight negative but insignificant effect on the mean CAR, indicating, as with the HQ relocation announcements, that the
Table 10. CAR results and dispersion for other action announcements
This table presents, on the left hand side, the one-sample student’s t-test results for an announcement of top management or a spokesperson on behalf of the firm stating plans to undertake action as a response on the Brexit uncertainty. The sample includes firms announcing to relocate at least 3% of their UK employees or actions as moving production lines. Where Interval denotes the event window, Obs. the number of firm observations,
Mean the mean with its corresponding standard error in brackets below, and stat the
t-statistic. On the right hand side the dispersion of the CAR over the three-day event window across firms can be found, with Mean and St. Dev. indicating the mean and standard deviation of the sample, followed by the Min, 25%, 50%, 75%, and Max indicating the minimum, 25th, 50th, 75th, and maximum observed CARs. Firm reports the two firms with the lowest and highest CAR.
Note: no significant t-test results were obtained.
Other action announcement CAR dispersion
Interval Obs. Mean T-stat Variable Firm
CAR stabilized itself somewhat around the announcement of the move of employees and/or other concrete plans. As Smith and Grimm (1987) and Sawyerr et al. (2003) described, firms that take action to mitigate the effects of uncertainty often outperform firms that neglect this. The Hungarian Airline Wizzair shows a CAR of -10,61% after its announcement to half its UK expansion plans, whereas UK insurer Hiscox enjoys a CAR of 3,24% after its announcement to set up a EU subsidiary in Luxembourg. See Table 10 on the previous page.
4.4.2 Long-term value effect of other action announcements
One month after the reports that action would be taken, a significant positive BHAR was shown. The two months after the event did not show this significant result, however adding one more month did. The final two months report a negative BHAR, significant for the sixth month. Prudential, UK’s largest
Table 11. BHAR results and dispersion for HQ relocation announcements
This table presents, on the left hand side, the one-sample Johnson’s skewness adjusted t-test results for an announcement of top management or a spokesperson on behalf of the firm stating plans to undertake action as a response on the Brexit uncertainty. The sample includes firms announcing to relocate at least 3% of their UK employees or actions as moving production lines. Where Interval denotes the event window, Obs. the number of firm observations, Mean the mean with its corresponding standard error in brackets below, and t-stat the t-statistic. On the right hand side the dispersion of the BHAR over the 30-day post-event window across firms can be found, with Mean and St. Dev. indicating the mean and standard deviation of the sample, followed by the Min, 25%, 50%, 75%, and Max indicating the minimum, 25th, 50th, 75th, and
maximum observed BHARs. Firm reports the two firms with the lowest and highest BHAR.
***p<0,01, **p<0,05
Other action announcement BHAR dispersion
Interval Obs. Mean T-stat Variable Firm
insurer by value shows a negative BHAR in the 30 days after its announcement of the plans to relocate its fund management division M&G to Dublin or Luxembourg. The French bank Credit Agricole, on the other hand, experienced a positive BHAR in the period after its announcement to repatriate its government bond trading platform bank to Paris, see Table 11 above. The BHAR results, as with the CAR, might be because the announcement of an employee move is less drastic than HQ relocation. Initiating other actions might be a way for management to show that they are aware of the effects the Brexit may have on the value of the firm. However, this theory does not hold for the entire six months. Announcing only an employee move will ensure that they do not burn any bridges in the UK, while simultaneously building a bridge to the EU. If the Brexit does proceed, it will be easier to relocate the firm if it is established in some form in a EU country. Banks often do this because of passporting rights, which is when the presence of the bank or firm in one EU country grants them the rights to sell their products and/or services throughout the EU. If the UK were to leave this agreement, they will lose their passporting rights as well (Finch, 2016).
5. Robustness
So far I find insignificant short and long-term results for HQ relocation announcements and a few significantly positive long-term results for reports of other actions undertaken. Initially I conjectured this was due to action being taken and top management showing their awareness of the threat the Brexit poses to the firm. For the purpose of further robustness of this study I analyze firms that announced not to undertake any action as a result of the Brexit, or reported that the Brexit would not affect their firm. On the short-term and long-term I find a positive, and in the five-day event window, and all four months post announcement, significant effect (see Table 12 and 13 below).
them a good investment candidate. Lastly, the notion that management has surveyed the potential risks has a similar effect as undertaking action, namely it indicates the firm is aware of its risks and possibilities, and knowingly doing nothing they still did something.
6. Limitations and future research
As naturally occurs with most studies, this thesis has some limitations. The first and foremost being that we are unable to observe all the effects Brexit will or will not have, as Brexit negotiations are still ongoing. When negotiations are concluded and the effects of the Brexit become more certain it would be of interest to repeat this study with more announcement dates.
Table 12. CAR results and dispersion for no action announcements
This table presents, on the left hand side, the one-sample student’s t-test results for an announcement of top management or a spokesperson on behalf of the firm stating the Brexit does not impact the firm or the firm plans to not undertake any action as a result of the Brexit. Where Interval denotes the event window, Obs. the number of firm observations,
Mean the mean with its corresponding standard error in brackets below, and stat the
t-statistic. On the right hand side the dispersion of the CAR over the three-day event window across firms can be found, with Mean and St. Dev. indicating the mean and standard deviation of the sample, followed by the Min, 25%, 50%, 75%, and Max indicating the minimum, 25th, 50th, 75th, and maximum observed CARs. Firm reports the two firms with the
lowest and highest CAR.
*p<0.1
No action announcement CAR dispersion
Interval Obs. Mean T-stat Variable Firm
The current sample of firms announcing relocations or other actions is relatively small, making the sample and tests prone to biases due to outliers, and outcomes may not be representative of the entire market. It would also be of interest to replicate this study with a benchmark that is less affected by these events. For example by analysing market and stock return for firms in Asia and the US, to determine if there are any negative shocks in the sample caused by global or other events. Furthermore this sample consists mainly of large European firms, as they often trade internationally. There might be small European firms as well who trade with the UK, on which the Brexit may have a more or less significant effect.
Table 13. BHAR results and dispersion for no action announcements
This table presents, on the left hand side, the one-sample Johnson’s skewness adjusted t-test results for an announcement of top management or a spokesperson on behalf of the firm stating the Brexit does not impact the firm or the firm plans to not undertake any action as a result of the Brexit. Where Interval denotes the event window, Obs. the number of firm observations,
Mean the mean with its corresponding standard error in brackets below, and t-stat the t-statistic.
On the right hand side the dispersion of the BHAR over the 30-day post-event window across firms can be found, with Mean and St. Dev. indicating the mean and standard deviation of the sample, followed by the Min, 25%, 50%, 75%, and Max indicating the minimum, 25th, 50th,
75th, and maximum observed BHARs. Firm reports the two firms with the lowest and highest
BHAR.
***p<0,01, *p<0,1
No action announcement BHAR dispersion
Interval Obs. Mean T-stat Variable Firm
7. Conclusion
In this thesis I attempt to determine the effect of Brexit announcements on firm value. Uncertainty is often difficult to deal with for a firm; however, studies have shown that the negative effects that often come with uncertainty can be mitigated through taking action. I showed, based on a sample of 952 firms, that the announcement of the outcome of the referendum had a significant negative effect on firm value. I examined this by observing the CAR on the short term, with event windows ranging from three days, to five, and 11 days. For the longer term I calculated the BHAR, showing negative results for up to one-year post announcement. These results were both significant on the short term and the long term, supporting my first hypothesis.
To test the positive effect of management taking action, I examined the announcements of HQ relocation. I estimated the CAR and BHAR for 11 firms, which announced their plans to relocate. Both the CAR and BHAR were not significant for any of the periods, indicating that I am unable to state they are significantly different from zero. This might be an indication of the mitigating effects of taking action, supporting hypothesis two. As firms might consider not to take irreversible actions by announcing their HQ relocation, but do want to build a bridge to the EU in case the Brexit does proceed, some firms announced a significant employee move, the move of production facilities, or other methods of ensuring a solid foothold in the EU. Estimating this effect using the CAR and BHAR gave a significant positive result for the BHAR over the first month, and insignificant positive results for most of the other periods. On the other hand I find significant positive results in the four months after firms reporting that no action will be undertaken, or that the Brexit will not affect the firm. I hypothesise this might be because investors are looking to reinvest in a firm not affected by the Brexit, the costs of the firm initiating action might be higher than the expected negative effect on value of the Brexit, the analysis performed by management indicates they have considered their options, or a combination of the above; they simply do not expect Brexit to affect their firm.
Appendix A. OLS estimators of the expected return parameters The estimators of the parameters for the expected return are: 𝛽! = (𝑅!"− 𝜇! !! !!!!!! )(𝑅!"− 𝜇!)/ 𝑅!"− 𝜇! ! !! !!!!!! (A1) 𝛼! = 𝜇! − 𝛽!𝜇! (A2) 𝜎!!! = 1 𝑇!− 𝑇! − 2 𝑅!"− 𝛼! − 𝛽!𝑅!" ! !! !!!!!! (A3) where 𝜇! = 1 (𝑇!− 𝑇!) 𝑅!" !! !!!!!! and 𝜇! = 1 (𝑇!− 𝑇!) 𝑅!" !! !!!!!!
the OLS estimators use the 𝑅!" and 𝑅!", the return in the estimation period spanning from 𝑇! to 𝑇! (See Figure 1).
Appendix B. CAR results by sector
Table B1. CAR results by sector
In this table the results of a one-sample student’s t-test are presented. I test if the CAR for firms across various sectors for the referendum outcome (24th of June 2016) is significantly different from zero. The intervals used are (-1,+1), (-2,+2), and (-5,+5), Obs. denotes the observation of number of firms,
the observed mean CAR within the sample can be found under Mean, with the standard error in brackets below, and the corresponding t-statistic is shown under t-stat. I divide the firms by sector,
resulting in seven sectors which make up the sample: Financial, Services, Technology, Industrial
goods, Consumer goods, Basic materials, and Healthcare. ***p<0,01, **p<0,05.
Referendum outcome
Financial Consumer goods
Interval Obs. Mean t-stat Obs. Mean t-stat
(-1, +1) 312 -0,0444*** (0,0060) -7,3432 104 -0,0216*** (0,0078) -2,7738 (-2, +2) 312 -0,0388*** (0,0053) -7,3797 104 -0,02185*** (0,0078) -2,7912 (-5, +5) 312 -0,0465*** (0,0054) -8,6681 104 -0,0211** (0,0108) -1,9551
Services Basic materials
