Real Estate Investor Behavior in a Post Brexit World
- Lara van Zalingen
December 2018 MSc Thesis
Student number: 10400443
University of Amsterdam, Amsterdam Business School MSc Finance and Real Estate Finance
Abstract:
This thesis investigates the anticipation of real estate investors on the unexpected market shock caused by the Brexit. As the potential loss of passport rights and fear of potential trade barriers harm the financial industry, the UK-real estate sector is under severe pressure. As already examined, UK house prices are decreasing. This analysis aims to show the Brexit-effect on the indirect market as measured by the UK-REIT (Real Estate Investment Trust) market, as this investment vehicle enables one to invest in real estate, without directly buying property. Apart from this, the Brexit effect on the direct market is measured by Foreign Direct investment. Both markets show a significant negative ‘Brexit effect’ after the announcement date. Next, this thesis provides proof of an intra-market dependency between REITS and direct real estate investments. Moreover, the referendum-poll outcome per voting region in the UK is correlated with the change in Foreign Direct Investment, concluding that local markets know their degree of exposure to risks related to Brexit, and vote accordingly. These statements further underline the main conclusion of this thesis that markets are forward looking, and anticipate on political risk caused by Brexit. The analysis is based on a panel data regression, thereby pooling data of 18 UK-REITS, compared to an identical analysis on American REIT data, where no Brexit-effect is appearing. Next, two OLS regressions are performed on the direct transaction data received from Real Capital Analytics, one showing the inter market relationship and the second showing the relationship between voting behavior and direct real estate investments per region.
Statement of Originality
This document is written by Lara van Zalingen who declares to take full responsibility for the contents of this document.
‘I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.’
Acknowledgements
As this thesis will be the final work for my 6 years of studying at the University of Amsterdam, I would like to take this chance to thank some very important people. I am very grateful to my supervisor prof. dr. drs. Francke, for guiding me throughout the process. His way of guidance is especially appreciated, as he planted ideas in my mind, but let me figure them out myself. Also, I would like to thank M. Mackay for challenging my thoughts and for providing his bright insights regarding the topic. I am very grateful for my friend M. Jansen, for proofreading my thesis multiple times and for her great support as a dear friend. At last, I would like to thank my parents, for unconditionally supporting my decisions.
Table of Content
pages
1. Introduction 4
2. Literature 7
2.1 The impact of Brexit on financial markets 7
2.2 Property derivatives trading around Brexit 10
2.3 REITs and the underlying asset 11
2.4 Political behavior or herding? 14
3. Research questions and methodology 16
3.1 Panel data with fixed effects 17
3.2 Measuring volatility dynamics 18
3.3 Research section 2: direct investment analysis 20
3.3.1 REITs and the underlying property 20
3.3.2 Poll-outcome and the change in Foreign Direct Investments 21
4. Data and descriptive statistics 22
4.1 REIT market data 22
4.2 Data on direct real estate investments 27
4.3 Poll-model data 29
5. Results 31
5.1 Results research section 1: Panel regression 31
5.2 ARCH/GARCH estimation results 35
5.3 Results Research section 2: Direct investment analysis 37 5.2.1 Results for REITs and the underlying property 37
5.2.2 Poll-model results 38
6. Discussion 40
7. Concluding remarks 41
8. Bibliography 43
1 Introduction
The vote on June 23 for Great Britain to terminate the European Union membership is regarded the most impactful financial new story of the year 2016. Conflicting polls prior to the actual referendum date prove that this outcome was not expected, and the final result shook the political landscape and global financial markets.
The status of thousands of people as EU citizen residing in the UK became uncertain. Apart from this, the future of international trade-dependent businesses became at risk and the potential trade barriers threatened especially the future of the financial sector. Therefore, the Brexit vote resulted in increasing expected future supply and simultaneously decreasing expected future demand for real estate.
While the progress of formulating a ‘Brexit deal’ has been turbulent, according to the EU conference held in Brussels on November 25th, 2018, a deal has been put forward by prime minister May. All EU leaders have approved the ‘Brexit-deal’ as proposed during the conference. However, the next challenge is that the prime minister has to convince the British parliament to approve the deal, as many opposition parties do not agree with current terms of the Brexit1 yet.
These terms include trade barriers for the banking sector, as the financial authorities in Brussels are in the position to block British financial institutions’ access to the European market when the level-playing field is harmed. Meaning, in terms of tax, competition, state support, labour and environmental standards, British financial institutions should not differ from European standards. If they do, all European trade will be blocked. Moreover, the border with Ireland is regarded a bottleneck in negotiations. By using a transition period the British parliament is hoping to smoothen the effects of the implementation of a physical Irish border. When parties do not agree on terms during this transition period, the UK will be forced into a backstop where the British parliament cannot create new trade agreements with other countries. Apart from this, the UK has to pay 39 billion pound to the EU, as they are obligated to pay for the European government bills upon 2020. These payments will be spread out over multiple years. At last, the current Brexit – deal will not harm the status quo of EU members residing in the UK, as their legal passport rights will remain unchanged.2 When the British parliament does not approve the current Brexit deal,
there is no secondary plan, leaving the future of the UK uncertain.
1‘ Brexit’ meaning the outcome of the British referendum on June 23d, 2016 to leave the European Union. Terms as ‘Brexit’ and ‘referendum
outcome’ or ‘vote’ are used interchangeably throughout this thesis.
2 The fact that passport rights remain unchanged became clear after the timeframe chosen in this thesis (on 25th of November 2018). Therefore, the
potential loss of financial passport rights is seen as one of the main causes of uncertainty in the market in the observed time frame around Brexit in 2016. This will be referenced to multiple times in this thesis.
A recent article by Monfared and Pavlov (2017) examined the link between political uncertainty and real estate values. The authors document a drop in London property prices following the referendum outcome, specifically in those areas with many EU passport holders. Moreover, areas with highly educated residents experienced a sharper drop in housing prices, meaning markets are anticipating on the Brexit before it actually takes place. Therefore, expectations are that individual investors in the UK housing market will respond to the political risk potentially caused by the Brexit by lowering their investment (direct or indirect) in the London property market.
As Monfared and Pavlov found evidence for this political behaviour in the London residential property market, this thesis will focus on the effect of the Brexit announcement on CRE3 investor
behaviour. This analysis is divided between the indirect and direct property market, where UK-REITs represent the former, and transaction data on UK property represents the latter. Where Monfared and Pavlov (2017) conclude that markets are forward looking and incorporate information quickly, the hypothesis of this thesis is that investors respond to this political risk caused by Brexit and act accordingly by lowering their positions in UK real estate.
Whilst existing research explains the Brexit-effect on the UK stock market and on the (residential) property market, one research (Zimmermann, 2018) tried to prove the effect of Brexit on UK-REITS, and found no significant results. This thesis performs a different statistical procedure by applying a panel data regression on UK-REITs with several time dummies, and does find a significant negative Brexit effect. In order to identify a relationship between REIT returns and the shock caused by Brexit, results are compared to an identical model using American REIT returns. In the American model, no negative effect of Brexit is present, as results even show significant positive coefficients. In Research Section 2, covering the direct property market, an analysis is performed on the co-movement of REITs with the underlying property. These results are in line with the extensive literature regarding the topic and suggest that the REIT market co-moves with the underlying direct property. However, the predictive power of REITs with respect to the direct property market is not significantly proven.
In order to truly answer the question if the observed negative returns are caused by investors acting upon increased political risk, a final model is included. The model aims to shed light on markets’ anticipation upon the risks caused by Brexit. It is formally tested if local UK markets know their level of European dependency and the implied negative risks of Brexit, and vote
accordingly to remain or leave the EU. The results show that UK-regions that voted to ‘Remain’ within the EU, indeed experienced a stronger decline in direct property investments compared to regions that voted to ‘Leave’ the EU. This further suggests that markets are aware of their local exposure to Europe and the increased political risk caused by exiting Europe, and act accordingly when voting on Brexit. Together with the result of a negative Brexit-effect on REIT returns, this will be the main contribution to existing literature.
The remainder of this thesis is structured as followed. As the research question is built on several sub questions, a clear division has been made between the indirect and direct property market. The former is explained in Research section 1, covering the indirect REIT market, while the latter is explained in Research Section 2. This division is applied consistently in all relevant chapters. First, existing literature is discussed in Chapter 2. Chapter 3 will explain the main research questions: ‘Does Brexit negatively affect the indirect UK-REIT markets’ return?’ and ‘Does Brexit negatively affect the direct UK property market?’ followed by the applicable methodology for both research questions and sub questions. In Chapter 4, data is described accompanied by descriptive tables, where data for the indirect market is retrieved from GFM management and Compustat, and the data on direct real estate transactions is provided by Real Capital Analytics . Main results are discussed in Chapter 5, followed by a discussion regarding the research in Chapter 6, elaborating more on internal and external validity and providing suggestions for further research. At last, Chapter 7 provides concluding remarks to sum up the final results being: - Brexit has a significant negative effect on the indirect REIT market, - Brexit has a negative effect on direct real estate investments, - the REIT office market co-moves with the direct office market and, - the poll-outcome per region is correlated with direct real estate investments in that region. These statements sum up to the main conclusion and that real estate investors are forward looking and anticipate on the increased political risk caused by Brexit. This conclusion will be the main contribution to existing literature on the Brexit topic.
2 Literature review
In order to understand the results of this thesis in the right theoretical framework, existing literature and background information is provided in this chapter. First of all, the impact of Brexit is described, elaborating on the short term effects in stock and bond markets, followed by a more in-depth analysis on the effect on CRE. The succeeding paragraph provides background information on property derivatives, followed by a paragraph explaining REITs and their correlation with the underlying direct property. At last, an analysis in the field of behavioral finance is included to shed light on the degree of political and/or herding behavior among CRE investors when anticipating on Brexit. The remainder of the thesis is built on the assumptions as described in these four sub sections.
2.1 The Impact of Brexit on financial markets
In order to fully understand how and why Brexit could affect the British real estate and stock market sector, the following section provides a brief overview of relevant literature discussing Brexit’s impact and its’ severance.
The decision of the British public to terminate the membership in the European Union shook international markets. As the referendum polls showed the night before the actual vote took place, the public was still undecided. The unexpected outcome was a shock to the European financial markets, which have been turbulent ever since. The potential consequences, while still uncertain, increase the degree of political risk when investing in UK property. Mark Carney, governor of the Bank of England explained on November 28th how the Bank of England prepares for a potential catastrophic Brexit. The bank performed several stress tests and modelled all possible Brexit outcomes into different scenario plans. The worst case scenario being a ‘no deal, no transition period’ scenario. Under this scenario, sudden trade barriers and custom declarations will block international trade and investments. Mr. Carney explains the importance of slowly disintegrating, as a sudden cut off would create a serious market shock. Under the worst case scenario, GDP will decrease by 8% over the next five years, and housing prices will decrease with 33% (BBC). The Bank of England stated that compared to the situation before the Global Financial Crisis in 2008, all major banks in England have 3,5 times more assets in place, even when accounting for the effects of this worst case scenario. Stated differently, the British banking system will not fail to provide its’ services to customers, nor to other banks, when Brexit happens in the worst case scenario. While this statement must reassure British people and investors, the fact remains that there is a certain risk of a market shock causing housing prices to decrease with 33%. This risk
brings a high degree of volatility into the market, upon which investors react by adapting their risk/return strategy.
According to Sampson (2017), the British government committed to the Brexit without knowing the actual terms and conditions of leaving the EU. The true economic costs and benefits are still unclear today, and will depend strongly on the new partnership the country will engage with the EU and on how markets will respond to this. The first short-term effect of the outcome of Britain leaving the EU was the depreciated pound, which fell 7% over night (Lathan et al, 2017). Succeeding, real estate funds suspended redemptions as investors sought to reduce their holdings due to the increased possibility of a downturn in the UK- property market. Publicly disclosing the halting of redemptions could lead to a self-perpetuating cycle, as systems do not allow funds to flow in direction investors want them to (Lathan et al., 2017).
In total, the global stock market lost $2.8 trillion of its value in the first two trading days after the referendum. Despite this stock market decline, there has been substantial heterogeneity in losses across firms. Firms that are more dependent on imported intermediates, or that are more heavily oriented towards Europe perform significantly worse compared to the market as a whole. Larger firms however, seem to better react on the turmoil of the Brexit compared to small firms (Davies & Studnicka, 2018). These findings are supported by Oehler et al. (2017), who also argue the importance of firm-level internationalization when analyzing the Brexit effect. The authors show that in the UK stock market, firms with a lower degree of internationalization, as measured by domestic sales, realized more negative abnormal returns compared to firms with a higher level of internationalization. This internationalization factor was important as well in the recovery phase, as more internationalized firms recovered to normal returns faster (Oehler et al, 2017).
Another interesting study by Schiereck et al. (2017) compares the Brexit to the Lehman Brothers scenario in 2008. Share prices of the Royal Bank of Scotland declined with -18% after the Brexit vote, whilst shares of Barclays dropped with 17,7%, followed by Deutsche bank with -13.9%. Since the Lehman Brothers bankruptcy, stock prices of banks haven’t been under such severe pressure. The authors conclude that stock prices of banks reacted more heavily to the Brexit announcement then to the former banks’ bankruptcy. Moreover, EU banks experience an increase in CDS spreads after the referendum. However, the increase in CDS spreads is less severe than in 2008. Apart from this, Dhringra and Sampson (2016) show that Foreign Direct Investment into the UK declined with 22%. This in turn will lower future real income by 3.4% accordingly. The question arises what the reasons behind this market reaction are.
The most important concern regarding the Brexit is the potential loss of financial passport rights. This would harm the UK financial sector and will have an immediate effect on the real estate sector (Zimmermann, 2018). This argument is supported by La Pena (2016). According to La Pena, nearly two thirds of all non-European banks have their headquarters based in the UK. In the likely event that Britain will not sign the European Economic Area (EEA) agreement, which allows non- European companies to provide products and services directly to clients without local authorization, the financial sector will be harmed and this will cause companies to relocate their headquarters outside the UK. In fact, many banks have already moved. Nomura and Daiwa responded to the referendum outcome by moving their headquarters to Frankfurt. Goldman Sachs, HSBC, JP Morgan and UBS are next up in line, and are planning to move to Frankfurt, Paris or Luxembourg (Amaro, 2018). Along with banks moving their headquarters, the estimation is that approximately 83,000 jobs will be lost in the financial sector in the coming seven years post-Brexit, together with 149,000 jobs in supporting sectors. (Lathan et al. 2017). This will harm the real estate sector with a rising supply and reduced demand in the UK- office market as well as the UK- residential market. A report by JLL (2016) also shows evidence of a slowdown of UK property investments, partly caused by a UK commercial property downgrade in response to the vote. Moreover, increased disclosure requirements in financial reporting demonstrate the uncertainty about market movements and the increased risk to corporate performance (EY, 2017).
The effect of the Brexit vote on the residential real estate sector of London City is studied by Monfared and Pavlov (2017). The authors concluded that for the London areas with many EU passport holders, residential real estate prices dropped between the range of 1.9% and 3% in the 4 months following the Brexit vote. This indicates an immediate capitalization of the potential increase in real estate supply, following the fact that many EU passport holders will have to leave the UK. Also, the proportion of the population that is highly educated responded immediately to the Brexit vote. These two outcomes further support the statement in this thesis that markets are forward-looking an incorporate information quickly (Monfared and Pavlov, 2017).
As all the above literature explains, it is hard to assess the outlook for the European real estate market as the political implications of the referendum are not clear yet, but first signals of a disturbed real estate market show the presence of a true economic shock, rather than temporarily trading noise. This statement, together with the argument that markets are forward looking, will be essential for interpreting the remainder of this thesis.
2.2 Property Derivatives trading around Brexit
‘Investor Behavior’ is a very broad term, and in order to provide a clear interpretation of the change in investor behavior, a short explanation is provided on the change in property derivatives trading. Most common ways to hedge the exposure to the risk that comes along with investing in real estate is by the use of futures, options, swaps and property index notes. Swaps allow investors to take a position in a real estate sub-sector in which they are currently not investing. It provides a way to exchange returns between sub-sectors. For example, part of the returns of an office portfolio is exchanged for the returns of a residential portfolio for a predetermined timespan (normally up to three years). Another way to hedge risk is to go ‘long’ by replicating the exposure of buying property, or to go ‘short’; by replicating the exposure of selling property without actually owning that property. At last, property future contracts establish an option to buy or sell property for a predetermined price at a predetermined date, thereby eliminating potential risks like the effect of the Brexit vote. Therefore, it is expected that in the build up to the referendum, the trading activity in UK property futures would rise, as investors would be keen on hedging away the risks of a potential Brexit on their real estate investments.
Figure 1: Arca PRM Indicative Brexit Pricing of Property Futures (2016 Q2-2017Q1)
As can be seen from Figure 1, there was a sharp drop in the ‘UK all property’ future pricing following the Brexit announcement on the 23d of June 2016. In the following months, prices recovered to normal levels. The above graph shows indicative pricing of Arca, a property risk management company that originated the first property future contract in the UK in 2015. As this
product is only traded over the counter, the prices reflect the exercise price of a property future. Figure 1 indicates a 20 percent drop in pricing after the referendum, which only recovered partly in the following months.
Important in this analysis is that during this event and the timespan around the Brexit announcement, there was no trading activity at all in property futures. ‘As this product is relatively new to the market, many investors are not convinced yet and are unwilling to use this as an investment vehicle’ (Jon Masters, Arca). Currently, all IPD Property Futures on Eurex are ‘dead’, as there is no trading activity anymore. This goes against the prior suggestion that investors would be keen on this market opportunity in order to make an arbitrage profit by using property futures as an investment vehicle, or simply wanting to hedge away their risk.
2.3 REITs and the Underlying Asset
This thesis will focus on price movements in REIT returns, or Real Estate Investment Trusts, which enables to track price movements in CRE markets on a daily basis, without having to observe property transactions directly as they tempt to occur with a varying frequency only and among dissimilar assets. The following section will therefore provide a short explanation on REITs, followed by existing literature on the correlation with REITs and the underlying asset.
REITs are originated in 1960 in order to provide access to cash flows provided by real estate, without having to invest in property directly. It allows professional investors to buy or finance property in the same way as they would invest in other markets; through stocks. Hereby, investors can overcome industry specific hurdles such as low liquidity, high transaction costs, and lumpiness when investing in real estate (Ling & Naranjo, 1997). REITs are classified in multiple sectors. Sectors analyzed in this thesis are: Industrial, Office, Residential, Retail, Healthcare and Other (Storage). REITs may be registered with the Securities and Exchange Commission (SEC) and shares can be listed and traded on different stock exchanges. However, this is not obliged, as REITs may act as private companies as well. The fundamental difference with the stock market arises from the fact that the Internal Revenue Code (IRS) has determined that at least 75 percent of REITs income must come from real estate in the form of rent, interest, or sales of real estate assets. Furthermore, REITs avoid taxes at the corporate/entity level in exchange for compliance with the 90% payout of taxable income requirement. These two requirements are fundamental to REITs and must be taken into account when analyzing their returns. As can be seen in Figure 8 in the Appendix, REITs
returns are outperforming stocks and bonds since 2012. Next section will elaborate more on the drivers behind REIT returns.
Extensive research has been done in the macroeconomic variables found in stocks and bonds that are proven to affect REIT returns and risks (Chan et al., 1990; Ling & Naranjo, 1997; Karolyi & Sanders 1998). According to this field of research, REITs are originally seen as an hybrid blend between stocks and bonds in terms of exposure and risk. The intuition behind this argument is that REITs are subject to the same macroeconomic factors driving stocks. At the same time, the fixed nature of cash flows caused by the income property with long term leases, together with the high dividend yield REITs have to pay out to investors (90%), imply that REIT returns and volatility are also correlated with the same macroeconomic variables that affect bond returns (Clayton & MacKinnon, 2003).
In contrast to the view that REIT returns are driven by stock and bond returns, another field of research regards a ‘pure real estate factor’ to be present in the NAREIT (public market) returns as well as the NCREIF (private market) returns (Giliberto, 1990). The paper argues that both markets are significantly correlated after deleting the effects of stock and bond returns. This view is consistent with Barkham and Geltner (1995), who prove an existing underlying link between the de-lagged NCREIF returns and the de-lagged NAREIT returns. The NCREIF returns have to be de-de-lagged, since there is a fundamental difference between the direct and indirect property market. As mentioned earlier, the indirect (securitized) property market adapts much quicker to market shocks as the investment vehicle is more liquid than real property. For this reason, changing market fundamentals are often first observed in the REIT market after which the private CRE market will follow.
Since the REIT IPO –boom in the 1990s, REITs’ structures began to change. As the market became more mature, the link with the underlying property became more clear and made REITs ‘more like real estate and less like stock’ (Ziering et al., 1997). The expanded magnitude of the REIT sector together with a more sophisticated investor base, improved the information stream and helped to better track the underlying real estate’s performance. The extent to which this co-movement with the underlying real estate is present, is investigated by Clayton and MacKinnon (2003). The authors show that during the early 1980’s, 72 percent of the NAREIT volatility is explained by the large cap stock factor. In the 1990’s, this stock factor has diminished to only 9 percent of the volatility. Concurrently, they found strong evidence for a significant real estate factor emerging in the 1990s. These findings of a stronger correlation between REITs and direct property are supported by Hoesli and Oikarinen (2012), who state that in the short run securitized real estate
returns do not move the same way as the underlying direct property but amplify that this is caused by the earlier mentioned lag in direct properties’ reaction to changing fundamentals. When looking at the longer term, they argue that the returns in the direct property market strongly co-vary with the securitized REIT returns. ‘Real estate shocks are taking place at first in the REIT market, after which the direct market adjusts to these shocks.’ (Hoesli & Oikarinen, 2012).
Another important issue when analyzing the co-movement with REITs and the underlying asset is the use of leverage. Several studies use techniques to de-lever REIT returns in order to show the degree of correlation with the underlying asset (Vlaming; 2016, Ling & Naranjo; 2015, Giliberto; 1993, Geltner & Kluger; 1995, Horrigan et al.; 2009). The latter research developed a tradable REIT based portfolio which reflects the unlevered returns per property type within individual market segments defined by usage type and geographic markets, or the so-called pure-play portfolio. This pure-pure-play portfolio represents a liquid and transparent investment which allows to compare price movements with the underlying asset. Moreover, it presents interesting derivative trading possibilities for synthetic investment in direct property. The authors conclude that REIT markets are sufficient in predicting the direct property market when de-levered. Furthermore, the pure-play portfolio volatilities are similar and sometimes even lower than the volatilities of transaction-based private market indices (Horrigan et al., 2009). These findings are complemented by Ling and Naranjo (2015) who state that passive de-levered REIT portfolios outperformed their direct private CRE market benchmark significantly with 49 basis points. The authors tried to examine what other underlying factors were possibly determining this risk-return profile. According to Ling and Naranjo, other factors of influence are liquidity, dividend yields, interest rates’ term structure, inflation, default spreads and the common Fama and French risk factors. When including these control variables in their regression, the significance of lagged REIT returns in the private market return disappears. That is, the included asset pricing control variables have no predictive power regarding the private CRE market as they do not contain additional information useful in predicting these returns. Said differently, REITs returns simply react to asset pricing information more quickly than the direct CRE market due to greater liquidity and transparency only (Ling & Naranjo, 2015).
2.4 Anticipation on Political risk or herding?
Political risk is a common phenomenon, as it has affected most national stock markets. According to Jorion and Goetzmann (1999), political events have caused interruptions in the transaction markets in more than 25 countries in the past century. Another study by Beaulieu et al. (2006) investigated the effect of the 1995 Quebec referendum on stock market returns. Just like the event analyzed in this study, the polls in Canada for the Quebec referendum were indecisive until the day before the outcome. Financial markets could not have anticipated on the outcome of the referendum beforehand, thereby creating the perfect climate for a natural experiment. The article concluded that the uncertainty around the referendum had a significant impact on the Quebec firms’ stock returns, and that this effect varied across the observed firms determined by the structure of assets and the degree of foreign involvement.
The incorporation of political risk in the real estate market specifically, is analyzed in numerous papers. Badarinza and Ramadorai (2017) find that short-term price movements in London are explained by the degrees of political risk in foreign countries. The authors show that house prices in London neighborhoods with a high concentration of certain immigrants, react positively when the political uncertainty in that immigrants’ country of origin increases. For example, when the war in Iraq started, housing prices in London with many inhabitants from Iraq, increased. This is probably caused by the next flow of immigrants who prefers to live near London residents of the same origin, thereby causing a rising demand in that specific area followed by increased housing prices.
Another study performed by Chan and Wei (1996) found evidence for the existence of political behavior. The authors document the impact of political news on stock market volatility and returns in Hong Kong. Surprisingly, they found evidence for an increase in stock price volatility following a political shock, hence not for stock returns. Chan and Wei suggest the absence of the effect on returns is caused by ‘combined market wide and substitution effects caused by accompanying political news’. This argument might hold in the Brexit case as well, as the effect of Great Britain leaving the EU can have serial by-effects complementing the initially negative shock on the capital market. One example of these so called by-effects is herding behavior;
Herding is typically described as the tendency of investors to follow the actions of others rather than their own belief (Zhou & Anderson, 2011). Herding can drive stock prices away from their fundamental values, thereby creating profitable arbitrage opportunities for practitioners. Moreover, it contradicts the traditional asset pricing theory as stock prices are influenced by other factors then stock specific fundamentals. Especially in the REIT market, herding behavior shows its
appearance, since the market behaves differently than the general equity markets. A study by Anderson et al. (2010) explained how REITs are more volatile than other equity markets after an unexpected shock. The authors stated that after an unexpected monetary shock, REITs react twice as strong as broader equity markets during high-variance periods. Moreover, the authors argue that REITs show significant higher extreme risks, measured by value at risk and expected shortfall. Compared to the broader equity market, the REIT market shows strong herding behavior in the higher quantiles only. In lower quantiles on the other hand, investors seem to rely more on market fundamentals and are not herding. This is not the case in other equity markets, where this behavior is present in all quantiles (Zhou & Anderson, 2011). The authors concluded that herding in REITs is more likely to appear in down markets, than in up markets with rising returns.
The broad literature on herding behavior in REIT markets suggests that it might be the case that the UK-REIT market is overreacting after the political shock of Great Britain leaving the EU. The question remains whether investors perform political behavior and act on behalf of their own information and knowledge, or show market-wide herding behavior and stick to the consensus regarding the market. The latter might aggrandize the previous; nonetheless the true answer regarding this question is beyond the scope of this thesis. For now, the assumption is made that both factors are present to a certain degree in the Brexit scenario.
3 Research Questions and Methodology
This thesis is based on multiple statements as described in the Literature chapter. For one, it is assumed that the effect of Brexit will cause a true market shock in the UK property market. This market shock is caused by the loss of financial passport rights and potential financial trade barriers, followed by the relocation of many financial services companies, including their employees, which in turn will originate a demand shift. Secondly, it is assumed that the REIT market co-moves with the direct real estate market, and is sufficient in predicting the real estate market as a whole. It is further suggested that real estate investors anticipate upon the political shock, which might by influenced by herding behaviour. The aim of this thesis is to analyse the extent to which real estate investors are forward looking and anticipating on Brexit. The research question is divided into two sub questions measuring the Brexit effect on indirect versus direct real estate investments in the UK;
Research Section 1: Does Brexit negatively affect the indirect UK-REIT markets’ return?
Research Section 2: Does Brexit negatively affect the direct UK-property market? Measured by: 1. Total investment volume in the UK pre- and post-Brexit announcement
2. Foreign / Domestic investments ratio 3. REITs effect on Direct Investments
4. The relation Between poll outcome and decline in Direct investments per region
As both research questions are distinct of nature, multiple methodologies will be applicable. For Research Section 1, the main statistical procedure used is a panel data regression with REIT fixed effects. This statistical test is performed in order to predict the effect of the Brexit vote on REIT returns. In order to identify the cause of the shock and allocate it to the Brexit announcement, these results are compared to the American market by performing the same analysis on US-REITs. This approach is supplemented by an ARCH/GARCH model in order to measure the effect on return volatility. The volatility which, according to existing literature, is the driving force behind the market shock. Research Section 2 will display the Brexit effect on direct investments. More specifically, it will show the effect on Foreign Direct Investments and the foreign/domestic investment ratio. Next, the link with direct investments per region and the referendum poll-outcome in that specific region is investigated. Apart from this, the REIT markets co-movement with the underlying property is further analysed.
3.1 Panel data with fixed effects
The following paragraph will elaborate further on the panel data approach with fixed effects and discuss why this specific method is preferred over other methods. The use of panel data regressions is supported by Hsiao (2007) who mentions the advantages of a panel data regression in his paper. Panel data regressions often show a more accurate inference of model parameters compared to cross sectional data, due to higher degrees of freedom and more sample variability. Moreover, the use of panel data controls for omitted variables, since it contains information on both the intertemporal dynamics and the individuality of the observed REITs. At last, panel data generates more accurate predictions by pooling data rather than predicting the individual outcomes (Hsiao, 2007). These two features are especially important in the Brexit scenario, since a panel data approach accounts for REIT specific characteristics as sector, market, and investment horizon. Not accounting for these differences could cause a bias in the results. Also, pooling data rather than predicting individual REIT effects, will increase the degree of external validity, as it represents the shock in the UK-REIT market in total. For these reasons, the panel data approach is chosen as main statistical procedure in this thesis.
As a control variable, the UK interest rate is added in the regressions. The interest rate is chosen as the only control variable, since availability of daily data is limited and might be influenced by the Brexit as well. The British stock market for example, also experienced a downturn following Brexit. For this reason, macroeconomic control variables other than the British interest rate are not included in the model. The lack of control variables will result in a low R-squared, or model fit. However, in order to still identify the shock in UK-REITs returns to be caused by Brexit, an identical analysis is performed on US-REITs. Expectations are that the effect of Brexit will not reach the US, especially not in such a short time-frame. Other macro-economic factors affecting the real estate market in general will be captured in the US-REIT market. If these coefficients show different results, it is plausible4 to identify the shock to be caused by the Brexit announcement. A contrasting approach, as
used in other papers, is to create models with many macro – economic control variables. These models can include for example currency exchange rates, or other daily available data covering potential macro –economic effects. The reason why this method is not sufficient in this case, is because of data availability. Especially with respect to the property market, the availability of daily data is limited. Following this argument, it is preferred to use a panel data regression and compare the results to an identical analysis on the US market, thereby eliminating omitted variable bias and at the same time accounting for macro –economic drivers affecting the property market in general.
An important assumption for using fixed effects is that differences across REITs are uncorrelated with the regressors. As quoted by Green (2008): “The crucial distinction between fixed and random effects is whether the unobserved individual effect embodies elements that are correlated with the regressors in the model, not whether these effects are stochastic or not” (p.254). If there is sufficient reason to assume differences across REITs do not affect the dependent variable, the use of fixed effects is applicable. For this assumption a Haussmann test, a Pesaran test and a Breaush and Pagan Langrarian multiplier test are performed in order to confirm that fixed effects are applicable in this scenario. Following this assumption, the regression in Equation 1 is used for the panel data setup of measuring the effect of Brexit on UK-REITs returns:
(1) Yit = β1Xit + β2𝑖𝑡+ αi + εit
For this analysis, multiple timeframes around the Brexit announcement date (23d of June 2016) are analysed. This is done by creating dummy variables, each representing a different horizon varying from one month to four months before -and after the Brexit date; time 0. The time dummy variable will be used as the independent variable Xit, to measure the effect on the dependent variable; REIT return. This regression will be repeated four times, to capture the ‘Brexit effect’ after one month, two months, three months and four months’ time respectively. In Equation 1, αi (i…n) is the unknown intercept for each entity representing a REIT fund in this case, Yit is the dependent variable, the UK-REIT return with i = entity (or a single REIT), and t = time measured in days. Then, the effect of interest rates on REIT return is measured by β2𝑖𝑡. Next, εit measures the within -entity error. A last model is included as a falsification test, where the analysis of four months before and after Brexit announcement is repeated, only with a fake ‘Brexit’ date, on June 23d 2015, instead of 2016. Here, it is expected that Brexit will not have a negative effect on REIT returns.
3.2 Measuring volatility dynamics
In order to analyse the Brexit effect on the volatility of returns, and thereby the implied risk of
investing in UK-REITs, a GARCH(1,1) analysis is performed. In this analysis, the same panel setting holds for the GARCH(1,1) model where time- dummies M1, M2, M3, M4 are measuring the changing dynamics of volatility in different timespans around Brexit. Predictions are that volatility increases after the shock, with a strong left skewness caused by more persistent negative than positive returns. The Generalized Autoregressive Conditional Heteroscedasticity (GARCH) model is regarded as an appealing technique to capture this skewness, as well as possible volatility clustering and the variation
of volatility over time (Arowolo, 2013). Bollerslev (1986) first introduced the GARCH model as an extension to the ARCH model, a model which was originated in order to estimate the variance of the UK inflation. The reason behind this model was that the usual econometric assumption of a time invariant volatility is improbable. Since extensive literature suggests that the volatility in financial markets’ data has the tendency to cluster in times of high or low volatility, a new model for stochastic forecasting was found by Engle (1982); the Autoregressive Conditional Heteroscedastic model (ARCH). The model is serially uncorrelated, has a zero mean, and has non-constant variances and a constant unconditional variance. The important assumption regarding the ARCH model is that the (conditional) variance can change over time as a function of past errors. The GARCH model extends the ARCH model by allowing the variance to depend on its own lags and on the squared error. Moreover, the model allows for a more flexible lag structure and includes the long run variance. The GARCH(1,1) model estimates the variance, based on the long run variance, the lagged squared error, and the lagged variance. (Bollerslev, 1986)
(2) 𝜎𝑖𝑡2 = 𝜔𝑖+ 𝛼𝑖𝜀𝑡𝑖−12 + 𝛽𝑖𝜎𝑖𝑡−12
Key is to estimate the constants ω, α and β in Equation 2. The weights are (1- α – β, α, β) . The weights must sum to one. In the GARCH(1,1) model, the first number refers to the number of autoregressive lags or ARCH terms that appear in the equation and the second number refers to the number of specified moving average lags, or the number of GARCH terms. Sometimes, more than one lags are needed in order to find good variance forecasts. In this model however, a (1,1) approach is adopted. This is because the regression on which the model is based; the REIT return, is modelled as unpredictable. Therefore, the original regression only includes the intercept and is not based on lagged predictors.
Another implication of the GARCH model is the possibility to measure the persistence of variance movements. This persistence is calculated by the sum of the coefficients α and β. If that sum is large and close to 1, this means that changes in the conditional variance are persistent, indicating long periods of high volatility, or volatility clustering. At last, the estimators in ARCH and GARCH models are normally distributed in large samples, which is why confidence intervals can be constructed as the maximum likelihood estimate (Stock and Watson, 2015, pp. 187-203). The GARCH(1,1) model will be adapted in this thesis in order to analyse the change in volatility in REIT returns and to measure the persistence of this volatility.
3.3 Research Section 2: Direct Investment analysis
Previous paragraphs explained methods to empirically test the ‘Brexit effect’ on the indirect property market measured by UK-REITs’ returns. As stated in the start of Chapter 3 and again below, the second research question in this thesis covers the direct real estate market by measuring the impact of Brexit on direct real estate transactions in the UK.
Research Section 2: Does Brexit negatively affect the direct property market in the UK? Measured by: 1. Total investment volume per continent pre- and post-Brexit announcement
2. Foreign/Domestic investments ratio
3. The relation between poll outcome and decline in direct investments per region 4. REITs effect on direct investments
3.3.1 REITs and the underlying property
Aim of this secondary analysis is to show how investor behaviour is changing measured by the foreign /domestic investment ratio, total investment volume, nationalities of investors. Apart from this, the argument that the (lagged) REIT market can predict the direct real estate market is formally tested by OLS regressions, measuring effect of the REIT trading volume, the lagged REIT volume and a Brexit dummy on the direct real estate market. This analysis is performed for the real estate market in total, and for the separate property-type sectors. The rationale behind this approach is based on the findings of Ling and Naranjo (2015), who analysed the REIT markets’ co-movement with the underlying asset and concluded that adding control variables as liquidity, dividend yields, inflation and interest rates had no significant effect. Ling and Naranjo concluded that asset pricing control variables had no predictive power in explaining the relation between the REIT and the direct CRE market. For this reason, the model is restricted to the REIT volume level, its’ first and second lagged value and a Brexit dummy. Equation 3 shows the formula, where Direct.vol implies the total volume in direct property investments at time t, which is affected by the total REIT volume at time t, and the first and second lagged values of the UK-REIT market volume. At last, a Brexit dummy is included, which is one for all observations after June 23d, 2016. This model is repeated for the different real estate sectors; Office, Retail, Residential, Industrial, Healthcare and Other.
3.3.2 Poll-outcome and the change in FDI
Apart from modelling the REIT market co-movement with the underlying asset, it is interesting to see whether the referendum poll-outcome per region correlates with the change in direct investments for that particular district. It is assumed that regions within the UK that are highly exposed to the negative effects caused by Brexit, anticipate on these negative effects and vote ‘Remain’. Contrarily, regions with less exposure to negative externalities caused by Brexit are expected to vote ‘Leave’. As the Brexit is proven to already have had a negative effect on the overall property markets, it is interesting to evaluate the differences across regions. Therefore, it is formally tested by OLS if UK-regions voting ‘Remain’ experienced a sharper decline in Direct property investments compared to the ‘Leave’ counterparties. In this way we can prove that markets are indeed forward looking and anticipating on Brexit. For this analysis, the total investment volumes per district from one year before –and after the Brexit announcement are compared. These two aggregated volumes are compared with each other, and the percentage difference is taken as dependent variable. As the data is limited to quarterly observations on real estate transactions in the UK, the second quarter of 2016, where the Brexit is observed is the hypothetical year-end. In this setup, 2015Q3 - 2016Q2 represents the timeframe for the first year, and 2016Q3 - 2017Q2 represents the second timeframe, where a lower degree of total FDI in property is measured as an effect of political uncertainty caused by Brexit. In order to eliminate potential seasonal effects and to account for potential long lags in response time, the chosen time frame is one year around Brexit announcement. These percentage values are regressed against a dummy variable that equals one when the district voted ‘Remain’. Expected is that this coefficient will be negative, as the ‘Remain’ voters are facing a higher exposure to negative effects caused by Brexit. The following hypothesis provides an intuition;
H0: The change in Foreign Direct Investments in UK property for ‘Remain’ voters is higher (more negative) than the change in Foreign Direct Investments in UK property for ‘Leave’ voters.
This hypotheses is formally expressed as: (4) ∆𝐹𝐷𝐼 = 𝐷𝑢𝑚𝑚𝑦 𝑅𝑒𝑚𝑎𝑖𝑛 + ε𝑡
Data on the change in FDI per region is provided at the end of Chapter 4, including summary statistics on the average, minimum and maximum change in FDI after the Brexit vote.
4. Data and Descriptive Statistics
In order to interpret the results arising from the explained methodology, this chapter will explain all relevant data. First, the data for Research Section 1 is described, corresponding to indirect real estate measured by UK-REITs. Descriptive tables and summary statistics are provided, and explained in more detail. As described in Chapter 3, the REIT analysis is compared to an identical analysis on US-REITs. Therefore, descriptive tables on the US-REITs are included as well in this chapter. Next, for Research Section 2, the data on direct property investments in the UK is explained. Here, descriptive tables will show a clear drop in direct investment volume around the Brexit announcement. Apart from this, an overview of the poll-outcome per region is provided.
4.1 REIT market Data
For Research Section 1, data on the top-20 UK-REITs is retrieved from GFM asset management. The corresponding data on US-REITs is retrieved from the CRSP/Compustat database. The UK-REITs data set includes information on trading price and volume. Covered sectors are Retail, Residential, Industrial, Office, Healthcare, and Specialty; which in the UK case is a fund that invests in storage only. Healthcare and Specialty are underrepresented in this dataset. This might introduce a bias in the results. For this reason, an analysis on the different sectors in REITs is excluded, as the data set is not representative for the complete market. Moreover, one fund is not active anymore, and is therefore also excluded from the analysis. This might induce survivorship bias, on which will be elaborated more in a later chapter. All panel data regressions are using a time – dummy variable as independent variable, measuring the total effect on all 18 REIT returns in one month, two months, three months and four months before and after Brexit announcement. These variables are named M1, M2, M3 and M4 respectively. In descriptive Table 1 the number of observations representing trading days per regression is provided. For the US, the same analysis is performed measuring the effect in 50 US-REITs accordingly.
Table 1: Number of Observations per regression (UK-REITs, 2016) Descriptive Statistics
Variable Obs Mean Std.Dev.
M1 721 .523 .5
M2 1401 .524 .5
M3 2098 .528 .5
Figures 2 and 3 show the effect of the Brexit announcement on the independent UK-REITs. As can be seen from both figures, the effect differs in magnitude across funds. For funds R7 (LondonMetric Property) and R5 (Derwent London PLC), the shock was almost -30%. Other funds however, seemed to not respond at all. This can be explained by the sector the REIT operates in. For example, R17 represents Target Healthcare. This fund provides inflation linked dividends backed by care homes. Their investments are long term, as the healthcare sector often operates with very long term lease contracts. This is due to the fact that healthcare institutions are often built for the specific user type, which makes it hard to sell the property after expiry date of the lease contract. For this reason, there is no clear shock of the Brexit visible at Target healthcare REIT. Moreover, fund R1 representing A&J Mucklow group, shows a stable return around Brexit. This fund only invests in long – term industrial property located in the Midlands, which is less responsive to news compared to the Office, Retail or Residential property sectors. Moreover, as can be seen from the referendum polls in the graphical display (Figure 5 p.29), the Midland area voted to leave the EU, which might suggest that the Midlands area is less dependent on EU membership. The same argument holds for R12, Tritax Big Box fund, a REIT investing in storage property only, a sector which is likely not affected by a potential Brexit. In short; first descriptive tables show that Healthcare, Industrial and Storage funds seem to be less affected by Brexit than Office, Retail and Residential REIT funds. The funds that show a clear drop in returns are elaborated on in Chapter 5.1.
-. 3 -. 2 -. 1 0 .1 -. 3 -. 2 -. 1 0 .1 -. 3 -. 2 -. 1 0 .1 -. 3 -. 2 -. 1 0 .1 1/1/2016 7/1/2016 1/1/20171/1/2016 7/1/2016 1/1/2017 1/1/2016 7/1/2016 1/1/20171/1/2016 7/1/2016 1/1/20171/1/2016 7/1/2016 1/1/2017 R1 R10 R11 R12 R13 R14 R15 R16 R17 R18 R2 R3 R4 R5 R6 R7 R8 R9 R e tu rn Date Graphs by Fund
The included summary statistics in Table 2 and Figure 3 below show the REITs’ name and the corresponding real estate sector, the long run average returns, standard deviation, kurtosis and skewness. Descriptive tables for the US are enclosed in Table 3 and Figure 4.
Table 2: Summary Statistics UK-REITs. (2016)
REIT(UK) Fund # Sector Mean Return St.Dev Kurtosis Skewn.
A&J Mucklow Group 1 Industrial 0.000244 .02 5.596 -.104
Big Yellow Group PLC 11 Specialty 0.000468 .016 6.191 -.216
British Land Co. PLC 3 Land 0.000187 .014 29.677 -1.805
Custodian REIT PLC 4 Office 0.000125 .008 7.329 .108
Derwent London PLC 5 Office 0.000402 .015 65.672 -3.313
Empiric Student Prop. 6 Residential 0.000166 .01 14.45 .823
Hammerson PLC 13 Retail 0.000142 .015 29.744 .83
Hansteen Holdings 8 Industrial 8.13E-05 .016 58.506 -3.079
Intu Properties 9 Retail -0.00041 .015 12.243 -.456
Land Securities Group 10 Land 0.000134 .014 16.041 -.901
LondonMetric Property 7 Office 0.000243 .011 11.775 -.781
Primary Health 14 Healthcare 2.48E-05 .011 6.581 -.298
SEGRO PLC 15 Office 0.000289 .014 8.464 -.457
Shaftesbury PLC 16 Retail 0.00042 .012 6.55 -.252
Target Healthcare 17 Healthcare 9.42E-05 .012 3.599 -.144
Tritax Big Box REIT 2 Specialty 0.000216 .010 8.385 -.30
Workspace Group PLC 18 Office 0.000745 .019 11.388 -.763
Table 3: Summary statistics US REIT Market (2016)
CompanyName Mean St. Dev. Kurtosis Skewn.
A G MORTGAGE INVES 0.001702 .014 5.096 -.142
A G N C INVESTMENT 0.000784 .01 4.45 .197
ACADIA REALTY TRUS 1.47E-06 .012 3.765 .065
ACRE REALTY INVEST 0.001067 .05 5.233 .682
AGREE REALTY CORP 0.001393 .014 5.933 -.663
ALEXANDERS INC 0.000543 .013 5.381 .569
ALEXANDRIA REAL ESTATE 0.00087 .014 5.153 -.817
AMERICAN ASSETS 0.000524 .012 3.485 -.055
AMERICAN CAMPUS CO. 0.000735 .013 3.11 -.205
AMERICAN FARMLAND 0.001177 .024 15.973 1.446
AMERICAN HOMES 4 0.001205 .014 6.008 .443
AMERICAN RESIDENTI -0.00218 .021 6.188 1.249
AMERICAN TOWER 0.000388 .013 6.003 -.837
ANNALY CAPITAL MAN. 0.000787 .01 4.383 .036
ANWORTH MORTGAGE 0.001165 .012 6.439 .268
APARTMENT INVESTMENT 0.000495 .014 5.177 -.736
APOLLO COMMERCIAL 0.000499 .012 12.91 -1.603
APOLLO RESIDENTIAL 0.001508 .024 94.07 8.229
APPLE HOSPITALITY 0.000269 .012 5.839 -.703
ARBOR REALTY TRUST 0.00054 .013 4.911 .421
ARES COMMERCIAL REIT 0.001006 .016 5.715 .14
ARMADA HOFFLER PRO 0.001236 .014 4.401 -.182
ARMOUR RESIDENTIAL 0.00056 .014 5.765 -.371
ASHFORD HOSPITALITY 0.001338 .029 5.003 .393
AVALONBAY COMMUNITY -0.00012 .012 3.345 -.295
B R T APARTMENTS 0.001182 .013 11.007 1.336
BIOMED REALTY TRUST 0.0003 .001 2.656 .223
BLACKSTONE MORTGAGE 0.000846 .012 4.832 .201
BLUEROCK RESIDENTIAL 0.000878 .017 3.574 -.369
BOSTON PROPERTIES 0.000204 .013 4.173 -.1
BRAEMAR HOTELS & 0.000285 .032 32.626 2.842
BRANDYWINE REALTY 0.000862 .014 4.444 -.438
BRIXMOR PROPERTY 5.49E-05 .017 68.646 -5.741
C B L & ASSOCIATES 7.58E-05 .025 5.167 .279
C I M COMMERCIAL 0.000439 .022 7.924 .84
C Y S INVESTMENTS 0.000755 .013 4.39 .003
CAMDEN PROPERTY TRUST 0.000614 .012 3.882 -.341
CAMPUS CREST COMM. 0.000757 .004 4.288 .712
CAPSTEAD MORTGAGE 0.00113 .012 7.176 .88
CARE CAPITAL PROPERTY -0.00038 .016 5.988 -.918
CARETRUST REIT INC 0.001451 .016 3.161 -.072
CATCHMARK TIMBER -5.1E-05 .015 4.197 .015
CEDAR REALTY TRUST -0.00039 .016 3.296 -.102
CHATHAM LODGING TRUST 0.000354 .019 4.166 -.033
CHERRY HILL MORTGA 0.001642 .011 10.514 .027
CHESAPEAKE LODGING 0.000428 .017 4.981 .387
CHIMERA INVESTMENT 0.001645 .012 5.143 -.498
The data description in Table 3 for the US contains averages, standard deviations, the kurtoses end skewness’s of all observed American REIT funds. The American REIT market is chosen as a benchmark for the global REIT market, in order to identify the shock in the UK-REIT market to be caused by Brexit. The American REIT market shows a stable return around the Brexit announcement, as expected. The magnitude of a shock caused by Brexit is most likely not affecting the US, or other continents than Europe. The German or French real estate market however, could also experience a decrease in returns caused by the negative externalities of Brexit. Therefore, it is especially chosen not to compare the British REIT market to other European counterparties. The American REIT market however, is most likely not affected by Brexit, but is affected by more general market factors driving the real estate market in general, like global economic prosperity, increasing population etc. Therefore, the US-REIT market is observed in order to identify the shock the be caused by Brexit. When comparing both countries’ descriptive tables, the UK data shows a strong left skewness in most of the REITS. This skewness is less persistent in the US dataset, since there is no shock appearing in this timeframe.
4.2 Data on Direct Real Estate Investments
This paragraph will elaborate further on the applicable data for Research Section 2, covering data on direct real estate transactions in the UK. Data is obtained from Real Capital Analytics (RCA) and must remain confidential. The data set contains sector – and market specific information, allowing to analyze which markets in the UK gained or lost popularity among investors. The data set is based on quarterly information, which in this case is sufficient since property prices are proven to have a significant time lag in adapting to news due to the illiquidity of assets (Barkham & Geltner, 1995). Most importantly, the data set includes information on origins of the properties’ buyer and seller. This allows for an in-depth analysis in the change in direct investment volume per country. In this case, the MSCI format is adapted regarding the division in global real estate markets, as can be seen from Table 4 on next page. Countries are labeled as either; Emerging Markets, Europe Euro-countries, Europe non- Euro counties (including Great Britain), Developed Markets and North-America. A first impression of this dataset presents a clear drop in international trading volume in UK direct real estate (2016Q3).
Apparently, European non-Euro countries did not invest in the UK at all in the quarter around Brexit. As can be seen from Table 5, which shows the Foreign/Domestic investment ratio, British investors did not invest in their home country during times of Brexit. The high degree of uncertainty in the UK market is experienced most by British investors, as they are confronted with news items regarding Brexit every day. This might result in a negative home bias, which is not appearing in other countries. For example, North America seems less altered in its UK investment pattern. Apart from this, it can be seen that in the quarters after Brexit investment volume recovered to normal levels. This is against earlier expectations, as the degree of uncertainty in the market is still high as the Brexit terms and conditions are unclear, and the true effect remains unknown to investors. Direct investments from emerging markets however, seem to have grown drastically in the fourth’ quarter in 2017, which is caused by one outlier; a £2,2 bln. deal in the Industrial sector, sold from the US to China. The same argument holds for the second quarter of 2015; when the US embassy in Central London is sold to Qatar for £1.5 bln, to turn into a hotel. When accounting for these outliers, direct investment from emerging markets follow a stable pattern. As can be seen from Table 5, the Domestic investment ratio dropped around the second and third quarter of 2016, potentially caused by investor fear of a potential market collapse. Interesting to see is that during this period investments in London remained relatively high compared to the total Foreign Direct Investments (Table 6). This can be explained by the fact that London is seen as an international core investment market and whilst Brexit will potentially harm London as well, core real estate property will always be a relatively safe investment. Therefore expectations are that international investors will keep investing in London.
5,000,000,000 10,000,000,000 15,000,000,000
Direct Investments London
Outer London Inner London Central London 5,000,000,000 10,000,000,000 15,000,000,000 20,000,000,000 15Q2 15Q3 15Q4 16Q1 16Q2 16Q3 16Q4 17Q1 17Q2 17Q3 17Q4 18Q1 £ Time
Foreign v. Domestic Invesmtent Ratio
Foreign Domestic 5,000,000,000 10,000,000,000 15,000,000,000 20,000,000,000 15Q2 15Q3 15Q4 16Q1 16Q2 16Q3 16Q4 17Q1 17Q2 17Q3 17Q4 18Q1 Overige ontwikkelde markten Opkomende markten Noord Amerika Europa niet-eurolanden Europa eurolanden
Table 6: Direct investment in London. Source: RCA (2016) Table 5; Forgein/Direct investment UK. Source: RCA (2016)
4.3: Poll-model data
For this section, the identical RCA dataset is providing the basis for a secondary analysis regarding direct real estate investments in the UK, measuring the change in FDI per year and the vote outcome per region. Predictions are that markets know their local level of exposure to Europe and act upon this level of exposure when voting to leave or remain in the EU. In order to provide more intuition behind this statement, Figure 5 shows the final outcome of the referendum vote to terminate membership with the European, by region. In this graphical display, most parts of Ireland, Scotland, Wales and London voted to ‘Remain’ within the EU. It is suggested that these regions are more dependent on a European relationship. This relationship is potentially harmed by trade barriers, loss of financial passport rights and other negative effects caused by Great Britain leaving the EU. These increased risk factors have had a significant negative effect on UK real estate in general, however the degree of this negative effects differ among regions. This situation lends itself to discover whether local markets know their local level of exposure, and vote accordingly. This is formally tested by means of an OLS regression on the effect of voting ‘Remain’ on the difference in Foreign Direct Investment in UK property. The descriptive statistics on the change in FDI per region are provided in Table 7, followed by a descriptive table of the of the variables as used in the OLS regression, which equals the average difference across regions (Table 8).
Table 7: Change in direct investment volume 1yr before and after Brexit per geographic area (2015-2017)
Row Labels Total (15Q3-16Q2) Total (16Q3-17Q2) Difference (%) Vote
Aberdeen £ 293,434,781 £ 116,952,163 -60% Remain Belfast £ 106,771,618 £ 204,501,591 92% Remain Birmingham £ 2,887,600,731 £ 2,187,371,808 -24% Leave Birmingham/Midlands £ 946,378,049 £ 762,621,671 -19% Leave Bristol £ 1,272,538,650 £ 617,767,406 -51% Remain Cardiff £ 533,833,285 £ 493,921,421 -7% Remain Central London £ 17,904,751,971 £ 18,340,655,728 2% Remain East Midland Other £ 713,911,063 £ 703,321,439 -1% Leave East of England Other £ 2,199,564,692 £ 2,272,752,232 3% Leave Edinburgh £ 792,705,367 £ 1,174,671,066 48% Remain Glasgow £ 1,012,490,609 £ 599,024,245 -41% Remain Inner London £ 3,579,421,949 £ 1,903,956,744 -47% Remain Isle of Wight £ 33,715,000 £ 12,160,789 -64% Leave Leeds £ 706,636,969 £ 374,230,574 -47% Remain Liverpool £ 1,025,360,752 £ 402,085,969 -61% Remain Manchester £ 2,944,346,368 £ 2,109,802,224 -28% Remain Manchester/NW Other £ 672,484,584 £ 620,565,299 -8% Remain Milton Keynes - Suburban £ 316,976,101 £ 184,306,907 -42% Remain Newcastle £ 801,682,264 £ 342,431,868 -57% Remain North East Other £ 327,186,432 £ 437,066,759 34% Leave Northampton £ 438,358,430 £ 500,738,004 14% Leave Northern Ireland Other £ 205,005,185 £ 135,582,308 -34% Remain Nottingham £ 350,768,466 £ 405,474,620 16% Leave Outer London £ 6,431,497,153 £ 5,361,458,704 -17% Remain Reading £ 819,707,901 £ 279,988,715 -66% Remain Scotland Other £ 796,239,337 £ 412,992,537 -48% Remain Sheffield £ 607,331,858 £ 754,484,754 24% Leave Southampton £ 272,927,534 £ 207,095,951 -24% Leave Southeast Other £ 4,727,081,885 £ 4,443,329,483 -6% remain Southwest Other £ 1,097,089,820 £ 1,198,056,054 9% remain Wales Other £ 446,149,455 £ 393,011,107 -12% Leave Yorkshire/NE Other £ 732,666,688 £ 819,561,983 12% leave
Grand Total £ 59,475,053,752 £ 53,207,655,264 -11% Leave
Table 8: Descriptive Statistics Difference coefficient
Variable Obs Mean Std.Dev. Min Max
