DEGREE PROJECT
REAL ESTATE AND CONSTRUCTION MANAGEMENT BUILDING AND REAL ESTATE ECONOMICS
MASTER OF SCIENCE, 30 CREDITS, SECOND LEVEL STOCKHOLM, SWEDEN 2020
Social Movement Effects on the Market Economy
–The Impacts of Anti-Extradition Law Amendment Bill movement on Hang Seng Properties index
Junyuan Guo
TECHNOLDEPARTMENT OF REAL ESTATE AND CONSTRACTION MANANT ROYAL INSTITUTE OF TECHNOLOGY
DEPARTMENT OF REAL ESTATE AND CONSTRUCTION MANAGEMENT
Master of Science thesis
Title Author Department
Master Thesis number Supervisor
Keywords
Social Movement Effects on the Market Economy Junyuan Guo
Real Estate and Construction Management TRITA-ABE-MBT-20562
Kent Eriksson
Hang Seng Properties index, political turmoil, market performance
Abstract
The aim of this thesis is to analyze the effects of political instability on the market performance.
The Anti Extradition movement in Hong Kong will be the study object and its impact on the Hang Seng Properties Index will be tested. The market performance will be measured with the parameters market risk and risk premium. Two regression models will be built where the political event serve as dummy variables and categorized into relatively peaceful protest (PI!"), massive conflict (PI#") and election period (EP"). The results indicate that all political events cause increased market fluctuation, except for EP" variable (in the risk premium model) which had a market stabilizing effect. The conclusion that the real estate market is sensitive to political turmoil is drawn.
Acknowledgement
I want to show my gratitude toall my supervisors for useful guidance and helpful advices for the completion of this thesis. I would like to especially thank Kent Eriksson (my supervisor) for all the support and professional recommendations given to me. To my comrade Qinglin Ouyang that gave me valuable insights. To my girlfriend Caroline Yang who provided language support and helped me to correct my English grammar.
Examensarbete
Titel Författare Institution
Examensarbete Master nivå Handledare
Nyckelord
Effekter av sociala rörelser på marknadsekonomin Junyuan Guo
Fastigheter och byggnader TRITA-ABE-MBT-20562 Kent Eriksson
Hang Seng Properties index, politisk instabilitet, marknadsresultat
Sammanfattning
Syftet med denna avhandling är att undersöka hur en marknad reagerar under ett politiskt ostabilt läge. För att göra detta kommer påverkan av demonstrationerna (mot det kontroversiella lagförslaget) i Hong Kong 2019 på Hang Seng Properties Index att studeras.
För att mäta marknadsresultatet kommer parametrarna marknadsrisk och riskpremie att användas. Regressionsanalyser utförs och de politiska störningsmomenten är indelad i tre kategorier; mild protest, masskonflikt och valperiod. kommer att agera som dummyvariabler i analysen. Resultatet från analysen indikerar att alla typer av politiska störningsmoment ger en ökad marknadsvolatilitet. Därmed kan slutsatsen att marknader är känsliga mot politisk turbulens härleddas.
1. Introduction ... 1
1.1 Purpose ... 2
1.2 Research questions ... 2
1.3 Delimitations ... 2
2. Background ... 3
2.1 The general background for the 2019-2020 Hong Kong protest ... 3
2.2 The stock market’s behavior during the period of turbulence ... 3
2.3 Performances of subdivision of the real estate market ... 5
2.3.1 Retail ... 5
2.3.2 Housing ... 6
2.3.3 Private office and flatted factories ... 7
3. Literature review ... 9
4. Theory ... 12
4.1 Autoregressive model ... 12
4.2 Stationarity ... 12
4.3 GARCH Model ... 12
4.4 Country risk premium ... 13
5. Method ... 14
5.1 Market risk ... 14
5.2 Country risk premium ... 16
5.3 Econometric models ... 16
6. Data ... 18
7. Results ... 23
7.1 Market volatility results ... 23
7.2 Risk premium results ... 24
8. Discussion ... 26
9. Conclusion ... 28
References ... 29
Appendix ... 31
1
1. Introduction
Market volatility can be influenced by several factors, i.e., investment risks, inflation trends, industry, and sector factors. It is a multidimensional phenomenon with various aspects. One aspect that needs to be introduced and pinpointed: the role of social movements (i.e., protest demonstrations, regime changes, political events) in market volatility. When social movements evolve into political unrest, the market price will start to fluctuate. Political instabilities will cause investors to be more offish and more responsive to future actions. This precautious attitude can result in investors being strategic and withdrawal capital from the affected market.
Consequently, a negative spillover effect will be applied to global economic growth (Morales
& Andresso-O’Callaghan, 2019). This argues for the fact that the market is sensitive to societal changes, and its effect should not be underestimated.
A close correlation can be seen between political changes and the stock market. It can be more extreme cases such as transforming from a market economy to a socialistic economy, or it can be more causal administrative changes induced by the government (Kim & Mei, 2001). An illustrating example of how political events can alter market behavior (by increasing or decreasing market volatility) is the Stock Exchange of Hong Kong during the Sino-British negotiation (Chan & Wei, 1996). This previous knowledge supports the assumption that the market is sensitive to social or political disturbances. Therefore, the current on-going demonstration against the Extradition Law Amendment might with high possibility cause the same effect on the market.
Hong Kong is an ideal market to study political effects, especially when it involves the central communist party since it has a long history with instabilities. To understand why the Extradition Law Amendment raised so many emotions, one must look into the relationship between China and Hong Kong. A few years ago, Hong Kong introduced an anti-national education, which caused a sentiment of anti-China (Kwong, 2016). Therefore, the majority of the citizens identify themselves as “Hongkonger” while repelling being called “Chinese” (Lam
& Ng, 2020). Consequently, protesters targeted and destroyed China-affiliated stores or any brand associated with the mainland China (Abacus, 2019). This anti-China attitude will possibly influence the price development on the stock index because most of the protesters believe that only by damaging Hong Kong’s economy can the local government pay enough attention to the purpose of withdrawing the draft.
A research conducted by Guo and Mei (2004) also showed that market volatility could be affected by political events. The authors concluded that by investigating 22 countries, political instability (i.e., election or during transition periods) could increase the probabilities of financial crises. However, it should be noticed that political instabilities do not necessarily contribute to or induce a financial crisis. It can only destabilize the market, and there will not be constant market volatility. Besides, another illustrating example is the steady decline in the German stock market during world war I. The stock market fluctuations clearly reflected the political turbulences during that time. The decline was initiated during the start of the war and
2 reached a slump in 1922 when the German military lost control over Ruhr (Bittlingmayer, 1998). Another paper by Saad (2011) that concentrates on the Israeli market also concludes that on-going political intension can alter the market risk and risk premium.
1.1 Purpose
This thesis aims to investigate the effects of social movements, specifically how political instabilities can influence Hong Kong’s economy in the real estate sector. Its prior focus will be on the current demonstrations against the Extradition Law Amendment 2019 in Hong Kong and analyze its potential effects. If a correlation between political unrest and market performance can be distinguished, hopefully, this thesis can serve as a guideline for investors during turbulent periods.
1.2 Research questions
1. Can political instabilities affect a markets’ performance and to what extent can the market be affected?
2. What kind of political turbulences can impact the market volatility?
3. In order to measure the market performance, the parameters risk premium and market risk will be used. How will these parameters react to political turmoil?
1.3 Delimitations
The most significant limitation of this thesis is highly specific. Every movement and demonstration are unique and can intake different forms, for instance, local or global, large- scaled or small-scaled, which will result in various outcomes. Additionally, local factors such as cultural differences and geography also do play a crucial role when it comes to result.
Furthermore, due to the outbreak of Covid-19, the global market has been affected to varying degrees, yet the aftermath of this protest has not entirely stopped, if the timespan wholly covers the movement, then the final result will somewhat be distorted. Therefore, this thesis is only for the stage before the outbreak of Covid-19.
3
2. Background
2.1 The general background for the 2019-2020 Hong Kong protest
The direct trigger of the movement is a murder case in Taiwan. The murder victim Hiu-wing Poon and the suspect Tong-kai Chan, who are both Hong Kong residents, went to Taiwan as a couple on February 8th, 2018. Poon was killed in the hotel room on the 17th of the same month, and the body was abandoned the next day in the grass while Chan took a flight to Hong Kong after the incident.
After the murder was unveiled on March 13th, the Taiwan procuratorate made three requests for mutual legal assistance to the Department of Justice of Hong Kong. Since there was no mechanism for transferring the suspects, Taiwan’s request to send Chan back to Taiwan for the trail was ignored.
In this case, the Hong Kong Security Bureau proposed to amend the Fugitive Offenders Regulations, which was initiated by the Central Commission for Discipline Inspection of the Communist Party of China, allowing the Hong Kong Special Administrative Region Government to transfer suspects to jurisdictions such as mainland of China, Macau, and Taiwan. Taiwan’s authority was concerned that Taiwanese in Hong Kong would be extradited to the mainland of China as well after the revision of the law and stated that they will not transfer Chan under such law. As the draft allows extradition of suspects in Hong Kong for trail in mainland of China, opponents worry that it will weaken Hong Kong’s Status as an independent jurisdiction under “one country, two systems”. The controversy widened thereafter, and eventually became the fuse of the outbreak of the Anti-Extradition Law Amendment Bill Movement (BBC News, 2019).
2.2 The stock market’s behavior during the period of turbulence
Market volatility, in general, is caused by various events, it is rare where one certain event can be associated or causative for a decline or a raise of the market performance. Events as quantitative easing in the Eurozone, rate cuts in the announced by Federal Reserve and the bank reserve requirement cuts of the People’s Bank of China, etc. can also have impacts. What can be analyzed from the changes in the Hang Seng Properties Index during the period where the demonstration took place, is the fact that it played an essential role in order for the market volatility to either decrease/increase to the point it shows in Figure 1.
The Anti-Extradition Law Amendment Bill Movement is a social movement that broke out on June 9th. The discontent among citizens of Hong Kong is due to the fear of mainland China will politically interfere with the Hong Kong government if this law comes into action. Hong Kong citizens are afraid of losing democratic rights if mainland China intervenes. For this law, protesters protested in the form of demonstration, rally, sit-in, blocked roads, obstructed MTR
4 trains from closing, destroyed, sang, shouted, doxing, “three strikes,” etc. hoping that the Hong Kong government will withdraw the draft.
Figure 1 shows the performance of the Seng Properties index during the unrest. The data was gathered from Finance Yahoo!.
Figure 1. Hang Seng Properties Index performance during the Protest
The protests and disruptions on June 9th affected the Hang Seng Properties index slightly negatively. Initially, the index showed a small decline, which can be explained by weak results in mainly in retail sales. This drop is not significant enough to be noteworthy. It can simply be regarded as an effect of the broad commitment to the demonstration, as the trendline shifts back to a positive trendline after the June 17th when the tension is eased.
A sharp downward trend was shown on July 21th, and the Hang Seng Properties index performance was affected severely. The market volatility fluctuated considerably, and a stable trendline could not be distinguished. The primary reason behind this massive decline might be related to the escalation of the protests, from peaceful demonstrations in gradually turned into vandalism. This vandalism contributed to a conflict between anti-China protesters and the rest of the citizens that where neutral who felt that it was unnecessary to aggressive and act like outlaws. The local conflict peaked from the Yuen Long attack, where a small-scale civil war was declared between the protester and neutral citizens. Ever since, after the Yuen Long attack, the market performance dropped with approximately 7000 points.
The remaining periods of the demonstrations the Hang Seng Properties index did tend to show some recovery, however, as soon as a protest occurs, the market volatility is registered to be negative. Even with the government releasing policies that should stimulate the economy, for instance, the release of property-related policies on October 21st, the overall Hang Seng Properties index remained a dismal market performance. The second most significant decline of the index was after the shooting on November 11th, and a man was set on fire. This event is
5 also associated with the conflict between neutral citizens and anti-China protesters. This local conflict between these two parties causes most impacts on the market performance (which is reflected by the decline in index). Assumedly, it is the local events that have the most effects on the Hang Seng Properties index rather than external factors, and the external factors cannot at least affect the same extent in comparison to local instabilities.
Around early December, the protest movements gradually lost their social impact and not so much attention and resources were directed to the protestants. The index started to regain some positive climb-ups. However, it still had difficulties returning to the initial 44000 marks. The summative trendline for December is positive, even if small volatility is shown throughout the month. This small volatility should be considered as normal market fluctuation and not primarily as an after effect of the Hong Kong demonstrations. It was stabilized and fluctuated normally around the 40000 marks until the outbreak of Covid-19 in January, where the index fell back to 37000 marks.
2.3 Performances of subdivision of the real estate market
The presented data will include a comparison with the Hang Seng Properties index (HSNP), which will give a more comprehensive and relatable aspect of the overall real estate market performance. After analyzing the constituents of the different types of Hang Seng Properties index, it has been chosen as the benchmark for comparing in the preliminary phase. Hong Kong property sector can be categorized into four major groups: private domestic premises, private offices (Grade A, B, and C), private retail, and private flatted factories. The monthly statistics of the above are retrieved from Hong Kong Census and Statistics Department.
For the following four charts, price indices, and the average rents and rental indices of fresh/renewal lettings of each sector are compared with Hang Seng Properties index.
2.3.1 Retail
The government reported that June retail sales fell 6.7% from a year earlier but the rise of the HSNP in June was largely contributed by the retail sector. It is not in line with the expected performance. However, it dropped gradually after that as expected since the protesters occupied the streets, the shops along the streets were affected. Note that some of the stores are not targeting the locals while tourists were scared away. What can be noticed is the period from May 19th to August 19th showed the greatest fluctuations (from 0% to -5% and recovery back to 5%). From the period March 19th to July 19th, it can be concluded that the price index for retail followed the same trend as the HSNP, while the periods after it showed some deviations from the HSNP trend.
6
Figure 2. Trendlines for rental prices in the retail sector
2.3.2 Housing
The price indices went down mildly until November as expectations, and new home prices have fallen slightly more the rents. Despite the impact on housing prices in the short term, Hong Kong’s housing demand is still high in the long run, combined with a slightly increased transaction volume, which results in a small decrease. In addition, long-standing dissatisfaction with unaffordable housing has spread to the streets in recent protests. It began with the now- suspended extradition bill and turned into a demand for greater democracy.
Figure 3. Trendlines for rental prices in the private domestic premises -15%
-10%
-5%
0%
5%
10%
15%
Oct-18 Nov-18
Dec-18 Jan-19
Feb-19 Mar-19
Apr-19 May-19
Jun-19 Jul-19
Aug-19 Sep-19
Oct-19 Nov-19
Retail
HSNP Price indice
Average rents and rental indices of fresh and renewal lettings
-15%
-10%
-5%
0%
5%
10%
15%
Oct-18 Nov-18
Dec-18 Jan-19
Feb-19 Mar-19
Apr-19 May-19
Jun-19 Jul-19
Aug-19 Sep-19
Oct-19 Nov-19
Private Domestic Premises
HSNP Price indice
Average rents and rental indices of fresh lettings
7 2.3.3 Private office and flatted factories
Private office and flatted factories can be grouped together since they have the same overall trend; a significant drop can be seen in August and bounced back in September. However, both the increase and decrease in September and October, respectively, are more evident in Private office (+3.67%/-8.7%, +1.75%/-3.31%), which means the office sector is more volatile. It is worth noting that in October, the Grade A office’s rent showed the largest month-on-month decrease since July 2019. A survey conducted by Amcham shows that more than 80% of the respondents agreed that the HK protesters had affected their decision in investing in HK in the future, and 27% of them were considering relocating, of which Singapore won the highest vote (91%) as the primary destination. However, due to the adjacency to mainland China and the distrust of socialism or CCP, the corporations that aim at Chinese markets have less intention to leave. Still, the probability of moving offices to less rented areas could not be ruled out.
Additionally, the office market has been turbulent since street protesters deliberately sabotaged Chinese bank branches and retail stores suspected of having links to mainland corporate groups.
This has reduced the interest of mainland Chinese companies.
Figure 4. Trendlines for rental prices in private office section -15%
-10%
-5%
0%
5%
10%
15%
Oct-18 Nov-18
Dec-18 Jan-19
Feb-19 Mar-19
Apr-19 May-19
Jun-19 Jul-19
Aug-19 Sep-19
Oct-19 Nov-19
Private Office (Grade A, B and C)
HSNP Price indice
Average rents and rental indices of fresh and renewal lettings
8
Figure 5. Trendlines for rental prices in private flatted factories -15%
-10%
-5%
0%
5%
10%
15%
Oct-18 Nov-18
Dec-18 Jan-19
Feb-19 Mar-19
Apr-19 May-19
Jun-19 Jul-19
Aug-19 Sep-19
Oct-19 Nov-19
Private Flatted Factories
HSNP Price indice
Average rents and rental indices of fresh and renewal lettings
9
3. Literature review
Hong Kong housing market is sensitive to political change due to the uniqueness in its immovability, large quantities of money, and long-term nature compared to other assets (He, Myer & Webb, 1998). Property developers have been the leading capitalists in Hong Kong since the handover in 1997 under government permission. Unlike Shenzhen, the major city on the Mainland bordering Hong Kong to the south, Hong Kong did not have any new investment in social housing by developing the remain rural areas (Graeber & Hui, 2014). Combine with the “high land price policy,” which supplies land by auction (Li, 2016), thereby resulting in higher housing prices. The rent for under 40 square meters space has inflated 28.3% between 2013 and 2014 alone, and young generations are then forced to live in misery, which is also one of the causes of the occupy central movement that happened at the end of 2014 (Graeber
& Hui, 2014). It is likely that the economic frustration of purchasing property can be alleviated by such campaigns as they create a sense of ambivalence that deters tourists from visiting Hong Kong, thereby inducing a recession, which is reflected in the real estate market (Morales &
Andreosso-O’Callaghan, 2019).
Morales & Andreosso-O’Callaghan (2019) examined the occupy central movement in 2014 and found out that the property sector in the Hang Seng stock market experienced a substantial loss (approximately twice as the mean of all sectors) at the beginning of the occupation.
However, after the a short period of volatility in the Hang Seng Properties Index, it resiliently recovered to its original status before the demonstration and showed strong performance during the event. Conclusively, the demonstration altered the performance in all sectors despite the fact that the disturbances were short-lived. This is cohesive with the previously reviewed statement that political instability has a negative influence on the stock market.
Specifically, He (1998) studied the Hong Kong real estate market that was affected by Tiananmen square events by comparing non-real estate markets. A hypothesis that the Hong Kong property sector was more politically elastic than non-property sectors was put forward out of the concern that real estate investment isless liquid. A real estate property index that includes nine companies was examined, and the empirical results provided supportive evidence for the hypothesis (He et al., 1998). Is the Hong Kong real estate market vulnerable? Yes, to some extent. When the epidemic, Severe Acute Respiratory Syndrome (SARS), struck Hong Kong in 2003, the property value dropped by 8 percent with the fall of around 30 percent in transaction volume (Wong, 2008).
Why does the real estate market matter to Hong Kong so specifically? Hong Kong working class had made the dream for property ownership tangible through hard working between the 1960s and 1990s. The reality, however, confirms the “conventional wisdom” that the housing price can only increase at an increasing rate that grows faster than the increment of their salaries, making home ownership more desirable (Leung & Tang, 2011). Why specifically the impact of current social unrest on real estate market? As a Special Administrative Region (SAR) of China, it has a long history of campaign for democracy. Yet it seems not to be the root causes
10 leading to all mass actions and street demonstrations. Some of the protesters have been campaigning in the street because of the increasing housing prices and widening gap between rich and poor (Chan, 2014). With the Gini coefficient reaching 0.539 in 2016, Hong Kong has become one of the most uneven wealth distribution cities in the developed world. To some extent, the sky-high housing prices and declining opportunities drove the frequently occurred social movements (Augustin-Jean & Cheung, 2018).
Theoretically, a social movement has been divided into 8 stages: seeming normal as first stage, following up with exposing injustice, ripening conditions, taking off, losing heart, winning the majority, reaping success, and ends with consolidating achievements. Simply summarized this means that there is an underlying dissatisfaction towards an injustice that will burst out with the right conditions. Ultimately the social movement will either lead to a social change that will result in disclosure or end before a change is induced due to loss of faith in supporters.
Today in Hong Kong a protest against the Extradition Bill 2019 has occurred and has reached the stage 6, winning the majority (Tai, 2018).
The paper by He, Myer & Webb (1998) studied the Tiananmen Square Event’s impact on the wealth of Hong Kong real estate. From this study, the authors managed to identify different types of political events and sort each event into six separate categories. Each unit was regarded as a dummy variable. In other words, when a particular political event took place, the value of the dummy variable would be 1 and otherwise 0. In general two regression models will be run.
In the first model, the regression of return indexes will be considered as the dependent variable and the six dummy variables (corresponding to the six categories) as independent variables.
The second model consists of the regression of above and the Hang Seng Properties Index will serve as another independent variable. The results the of the second model in terms of R squared is greatly larger than the first one (0.2271/0.9714). This suggest or can be interpreted as the political events that occurred can only be explanatory for only a small proportion of the market performance fluctuations.
Saad (2011) also intends to investigate the effects of political tension on the Israeli stock market.
The regression models in this paper show similarities with the model that He, Myer & Webb proposed, both studies chose to include political instability as the dummy variables. For the independent variables, Saad selected however the Moody’s rating of Treasury bonds and the ratio between the standard deviations of Israel and the U.S., where the standard deviations will be measured in terms GDP growth. Note that the denominator should be a country that can function as a benchmark or baseline. Saad also presents a second regression model where it takes the election period into account, as the election itself in the investigated country should be regarded as a different kind of political instability.
There is a causal-in-mean relationship between the exchange rate, interest rate, the stock return and the financial sector in general. It is important to value this linkage during risk assessments, especially during financial crisis. Studies shows that it is mostly the long-term rates that have significant (both positive and negative) impacts on the stock market and the market volatility.
Additionally, both the interest rate and the exchange rate can be used for future market
11 performance predictions. For instance, recent studies show that US interest rate volatility can be used as a predictor of Australian bank stock return volatility. The exchange rate has a causal and bilateral relation to the stock prices. From a study made by Acikalin and Seyfettin Unal (2008), it is suggested that past historical exchange rates can be used to predict future stock market performance (Mouna and Anis, 2016).
For a precise evaluation of an emerging market and the assessment of an equity risk premium, the concept country risk and country risk premium are introduced by Damodaran (2003). In an already well-established and mature market for example the U.S., the risk premium can be simply estimated with historical risk premiums, however, this kind of approach is not suitable for emerging markets due to the lack in quantity of historical data. Therefor modifications are required. Damodaran gives the suggestion that country risk can be measured instead and its conversion to country risk premium can be regarded as an equivalent to a country’s equity risk premium. There are three approaches that Damodaran presents for the measurement of country risk premium; country bond default spreads, relative equity market standards deviations and default spreads + relative standard deviations. (The first mentioned method will be presented more thoroughly as it will be used in this thesis).
12
4. Theory
4.1 Autoregressive model
Time series variables are often dynamic and a way of describing this dynamic relationship is the usage of autoregressive model. An autoregressive model is where the current variable y is dependent on its past values and with a number of lagged values. Lagged variables can be understood as how many time periods into the past the current value is correlated to. For instance, the variable y is dependent on current and past values of itself up to q periods into the past. For instance, the model can be expressed as
𝑦$ = 𝛿 + 𝜃!𝑦$%!+ 𝜃#𝑦$%#+ ⋯ + 𝜃&𝑦$%&+ 𝑒$
where the coefficients 𝜃' is the lag weights and p represents a finite or infinite number of periods. This type of model is extremely useful when making forecasting analysis and capturing the correlation between current and past values of a variable in a time path.
4.2 Stationarity
Variables that are stationery means that the variances are constant and do not change over a period. The autocorrelation of the variables is only dependent on the time interval between each observation and do not look at a specific time mark. Stationarity implies that when estimating different subsets of observations with different time horizons, the same population quantity is estimated. Thus, the mean, the variance and the autocorrelations receive all the same value. Some characteristic features of stationary variables are that it only fluctuates around its mean value, it does not wander aimlessly around the mean (always returns to the mean value) and does not have an explicit trendline.
Figure 6. Time series of a stationary variable (left), a non-stationary variable that wanders (middle) and a non-stationary variable that trends (right). (Carter Hill, Griffiths & C. Lim, 5th edition, 2018)
4.3 GARCH Model
GARCH is a regression model, and in order for a correct application of GARCH, some requirements must be fulfilled. In the full name of GARCH, Generalized Autoregressive Conditional Heteroskedasticity, where “Heteroskedasticity” means the variances are not constant, i.e., they change over time, in other words, the time series is non-stationary.
13 Conversely, the term homoskedasticity means that all the observations follow the same variance. Thus, it is crucial to test whether the studied time-series data is stationary and weakly dependent. Variables’ stationarity can be confirmed by employing the Augmented Dickey- Fuller test (Dicky & Fuller, 1979; Cheung & Lai, 1995).
The Augmented Dickey-Fuller test is a unit root test to test stationarity in a higher order autoregression model (AR). It can be implemented in any econometric or statistical software, but often used in time series analysis. This test minimizes the chances of receiving unpredictable and biased results. Furthermore, it also ensures that the errors are uncorrelated by including sufficient amounts of lags. Mathematically speaking, when testing for the unit root in a high order AR where
∆𝑦$ = 𝛼 + 𝜃!𝑦$%!+ 𝜃#𝑦$%#+ ⋯ + 𝜃&𝑦$%&+ 𝑣$
against the alternative where 𝑦$ is stationery, can be considered as testing 𝐻(: 𝛾 = 0 against its alternative 𝐻!: 𝛾 < 0 if
∆𝑦$ = 𝛼 + 𝛾𝑦$%!+ 4 𝑎)∆𝑦$%)
&%!
)*!
+ 𝑣$
4.4 Country risk premium
Investing in a country or region, especially when the investment is exposed to numerous factors such as currency fluctuations, political instability, and adverse government regulations, country risk should be taken into consideration (Damodaran, 2003). The author suggests that in a developed market, as opposed to an emerging market, historical data can be used in a reliable way to estimate the standard error in the risk premium. However, under the topic of investigating the impact of political instability on the emerging stock market, the risk premium should be adjusted.
The method of Country Bond Default Spread from the paper by Damodaran (2003) will be selected in this thesis for the reason that the results of another two methods resulting in inappropriate data series. The risk premium for an emerging market should be generally higher than that of a developed market as it relatively bears more risks, making the yield of 10-year sovereign bonds in developed market lower than that of the emerging mark.
14
5. Method
This study intends to investigate the economic impact of the anti-extradition movement and its follow-up consequences on Hong Kong’s real estate market. Before investors entre a market, especially when there is a connection between the market and political factors, they usually consider the market risk and risk premiums of this market. An examination of whether these factors have changed or to what extent did they change is a way to reveal the market performance under the political turmoil.
5.1 Market risk
The main method that will be implemented for the retrograde analysis of the market changes in Hong Kong is the GARCH model. The GARCH model, shorts for Generalized Autoregressive Conditional Heteroskedasticity, will be implemented for an efficient and accurate representation of the market changes, expose the market volatility, and measuring the market risk.
The first step before implementing the GARCH model is to test the existence of stationarity.
Variables’ stationarity can be confirmed by the employing Augmented Dickey-Fuller test (Dicky & Fuller, 1979; Cheung & Lai, 1995). From the Hang Seng Properties index series shown in figure 1, no definite and clear upward or downward trend can be distinguished, and the trendline does not fluctuate around zero (a nonzero mean), this suggests that an augmented Dickey-Fuller test should be selected:
∆𝑦$ = 𝛼 + 𝛾𝑦$%!+ 4 𝛼)
&%!
)*!
∆𝑦$%) + 𝑣$
The formula outlined above is the Dickey-Fuller test with intercept and no trend.
As figure 1 exhibits a fluctuating behavior, it can be assumed that the variables might be non- stationary, therefore a check of stationarity is needed. In this case, the null hypothesis 𝑦$ has a unit root and is non-stationary, while the alternative hypothesis is that 𝑦$ is stationary around a nonzero mean. If the 𝜏 statistic is smaller than the critical value, the null hypothesis is rejected, and the stationarity of the variables is confirmed.
Secondly, it is also crucial to identify the optimal number of lags, therefore implementation of the Vector Autoregressive (VAR) model will be required (Hill et al., 2017).
A s-order vector autoregressive model, is
𝑦$ = 𝛼(+ 𝛼!𝑦$%!+ 𝛼#𝑦$%#+ ⋯ + 𝛼)𝑦$%)+ 𝑣$
15 Presumably, there would be s lags in the model. By using the data analysis tool in Excel, a regression model with several p-values based on the assumption of s lags can be obtained. If the p-values on certain lags are not acceptable, another regression without those lags needs to be run until all the p-values are below a certain level.
Lastly, before applying the GARCH model, the determination whether the observations show a heteroskedastic or homoscedastic behavior is required. Consequently, a test for the ARCH effects on the data is necessary. If there is an ARCH effect, then the observations are considered to be heteroskedastic and the implementation of the GARCH model is suitable. A GARCH model cannot be applied to observations that show homoscedastic behavior. Note here that, when calculating with the Hang Seng Properties index, it appeared that the trend has no ARCH effect, therefore this thesis will continue with the usage of return on the Hang Seng Properties index instead. The return on the Hang Seng Properties index is also representative of the analysis market risk, it has the same analytic equivalence to adjusted closing price.
For the GARCH (p, q) model:
𝑦$ = 𝛼 + 𝛽𝑦$%!+ 𝜀$ where
𝜀$êΩ$ ~ 𝑖𝑖𝑑 𝑁(0, ℎ$) and
ℎ$ = 𝜔 + 4 𝛼+ℎ$%+
&
+*!
+ 4 𝛽,𝜀$%,#
-
,*!
This formula shows that ℎ$ , the variance in day t, is dependent on a long run variance (𝜔), and separate summations of two different past values: the past squared error (𝜀$%,# ) value and the past variance value (ℎ$%+). The terms 𝛼+ and 𝛽, are to coefficient that represents the lagged term for each value, or as the weight of each parameter. The selected GARCH (p, q) is GARCH (1, 1) in this thesis. Consequently, the ℎ$ formula can be simplified to:
ℎ$= 𝜔 + 𝛼!ℎ$%!+ 𝛽!𝜀$%!#
What the formula essentially reveals is an estimation of today’s value (in this thesis, the value will be equivalent to the daily variance of the return on Hang Seng Properties index) as a function of yesterday’s value and yesterday’s error terms. This also explains why the GARCH (1, 1) model is selected and applied. The observations made in this thesis are daily. This means that it will be much more convenient and relevant to use yesterday’s data for the analysis of today’s value. For instance, if the parameters were defined to GARCH (2, 2), in order to analyze the current index value, error terms and variances from the day before and the day
16 before that will be needed (and so forth). The more the time elapses from the current value, the less credibility and relevance it will have. Therefore, specifying the GARCH (p, q) model to GARCH (1, 1) model provides a more optimal approach and also gives simple calculations.
5.2 Country risk premium
The following equation shows the calculation of country risk premium in Hong Kong.
𝐶𝑜𝑢𝑛𝑡𝑟𝑦 𝑟𝑖𝑠𝑘 𝑝𝑟𝑒𝑚𝑖𝑢𝑚./01 3/01
= 𝑌𝑖𝑒𝑙𝑑 𝑡𝑜 10 𝑦𝑒𝑎𝑟 𝑈𝑆 𝑡𝑟𝑒𝑎𝑠𝑢𝑟𝑦 𝑏𝑜𝑛𝑑
− 𝑌𝑖𝑒𝑙𝑑 𝑡𝑜 10 𝑦𝑒𝑎𝑟 𝐻𝑜𝑛𝑔 𝐾𝑜𝑛𝑔 𝑡𝑟𝑒𝑎𝑠𝑢𝑟𝑦 𝑏𝑜𝑛𝑑
The country risk premium obtained can then be incorporated with that of a selected mature market equity premium (the US market) to calculate the total equity risk premium in Hong Kong. The total equity risk premium is then presented:
𝑇𝑜𝑡𝑎𝑙 𝑒𝑞𝑢𝑖𝑡𝑦 𝑟𝑖𝑠𝑘 𝑝𝑟𝑒𝑚𝑖𝑢𝑚./01 3/01
= 𝑀𝑎𝑡𝑢𝑟𝑒 𝑚𝑎𝑟𝑘𝑒𝑡 𝑒𝑞𝑢𝑖𝑡𝑦 𝑝𝑟𝑒𝑚𝑖𝑢𝑚45 + 𝐶𝑜𝑢𝑛𝑡𝑟𝑦 𝑟𝑖𝑠𝑘 𝑝𝑟𝑒𝑚𝑖𝑢𝑚./01 3/01 5.3 Econometric models
The final step in the analysis process of impacts from the demonstrations is to distinguish how the political tensions and instabilities effect the real estate market. Three econometric models will be built to find if there is any correlation between an unstable period and volatile returns.
For clarification, the standard deviations are obtained from the GARCH model while risk premiums are from country risk method, and to examine the impact, a certain period will be chosen to see what the standard deviation and the risk premium will react.
Model 1
𝜎$ = 𝛼 + 𝛽! × 𝑟$+ 𝛽#× 𝐻𝐾𝐷$+ 𝛽6× 𝑃𝐼!$+ 𝛽7× 𝑃𝐼#$+ 𝜀$ 𝑅𝑝$ = 𝛼 + 𝛽! × 𝑟$+ 𝛽#× 𝐻𝐾𝐷$+ 𝛽6 × 𝑃𝐼!$+ 𝛽7× 𝑃𝐼#$+ 𝜀$
What is stated here are two regression models that both the market risk (𝜎$) and the risk premium (𝑅𝑝$) can, to some extent, be influenced by these four independent variables. The variable 𝑟$ is the discount window base rate published by the Hong Kong Monetary Authority.
𝐻𝐾𝐷$ represents the daily price of Hong Kong dollar to US dollar during the period. 𝑃𝐼!$ is a dummy variable that represents the political turbulences that Hong Kong faced. Note that only when a relatively peaceful protest or demonstration took place will the value of this dummy variable take 1. 𝑃𝐼#$ will take the value 1 when massive conflicts break out otherwise it will remain 0. Some examples of massive conflict can be aggressive protests, police engagement or civilian anarchy. 𝜀$ is an error term and functions as error marginal that the equations might have. 𝛼 and the 𝛽- values are all coefficients for the different variables.
17 Model 2
𝜎$ = 𝛼 + 𝛽! × 𝑟$+ 𝛽#× 𝐻𝐾𝐷$+ 𝛽6× 𝑃𝐼!$+ 𝛽7× 𝑃𝐼#$+ 𝛽8 × 𝐸$+ 𝜀$
𝑅𝑝$ = 𝛼 + 𝛽! × 𝑟$+ 𝛽#× 𝐻𝐾𝐷$+ 𝛽6× 𝑃𝐼!$+ 𝛽7× 𝑃𝐼#$+ 𝛽8× 𝐸$+ 𝜀$
Model 2 is a further development of model 1, the variables 𝑟$, 𝐻𝐾𝐷$, 𝑃𝐼!$, and 𝑃𝐼#$ are defined in the first model. The variable 𝐸$ presented the equations is also a dummy variable. This variable captures the effect that the election induced on the Hang Seng market. Hong Kong’s 2019 election period was between October 4th (nomination period) and November 24th (polling date). Since the election took place simultaneously along with the demonstrations, it is crucial to include its effects also. Note that 𝑃𝐼!$ and 𝑃𝐼$# represents political instabilities excluding the election turbulences. Model 2 encounters another different type of instabilities in Hong Kong in comparison with Model 1.
The purpose of adding two other variables, 𝑟$ and 𝐻𝐾𝐷$ , is that only a small proportion of market performance such as volatility and risk premium can be explained by political changes, that is, the R squared will be extremely low without such variables.
18
6. Data
The research methodology of this thesis is based on the calculations of risk premium and market risk, one is supported by the descriptive statistics for the Hang Seng Properties index, and the other one is the country risk premium plus the risk premium in U.S. stock market. The collected data sample from the Hang Seng Properties index will provide a stable framework for the estimation of market risk. Daily data of Hang Seng Properties index is gathered from Finance Yahoo! between 23rd January 2017 and 2020 22nd January 2020. There are 739 observations in total to ensure that optimal quantity of observations is available for the GARCH model analysis. Due to the outbreak of Covid-19 the protest never ended officially. As a consequence of Covid-19, the market performance was unsatisfactory globally. Therefore, the data collection’s end date is 22nd January 2020 (before the outbreak) to exclude the market effects of Covid-19.
Note that there are variations of closing days between the Hong Kong and the U.S. market due to different holidays, celebrations, and breaks. Therefore, there will be days when the data cannot match up, or missing data. This problem cannot be neglected as it brings out errors such as underestimating the market volatility if using an average price instead of real prices.
Additionally, missing data is also a sign of illiquidity. According to Janabi (n.d.), this mismatch and the lack of some data will require adjustments:
𝑃+ = `𝑃! $%!× 𝑃$9!
According to the formula above, the missing figure would be the geometric mean of the previous and following figures. This formula will balance out the mismatch and will provide reliable data for the standard deviation calculations.
In order to calculate the standard deviations, the GARCH model for return on both Hang Seng Properties index and S&P500, for comparison, will firstly be used.
The values for the tables below are calculated by implementing the GARCH (1,1) model. This model outbrings the values of the coefficients and its values will also validate the stated model.
This validation is required for the following step of analysis as the coefficients has a crucial role.
Table 1. Optimal parameters of the return on Hang Seng Properties index
Coefficients Estimate Std. Error t value Pr(>|t|)
omega 0.000134 0.000047 2.8371 0.004553
alpha1 0.169867 0.019055 8.9144 0
beta1 0.829133 0.015457 53.6401 0
19
Table 2. Optimal parameters of the return on S&P500
Coefficient Estimate Std. Error t value Pr(>|t|)
omega 0.000003 0.000003 0.9615 0.3363
alpha1 0.222757 0.23396 9.5213 0
beta1 0.742103 0.015172 48.9125 0
Table 1 and 2 represents the converted statistics for the Hang Seng Properties index and S&P500 respectively. The main highlight of the received values is that the coefficient values (more accurately speaking the estimate of the coefficients) are all statistically significantly, greater than zero. This means all the values of the parameter are plausible and there is a significant effect and move towards a specific direction.
Figure 7. Standard deviations of Hang Seng Properties index and S&P500
Figure 7 is a graphic representation of how the market performed between the January 2017 to January 2020. For this thesis the most interesting period will be around the 2019. In comparison to the S&P500, the Hang Seng Properties index showed much market fluctuations, it can be characterized with high peaks and steep downfalls, while S&P500 had a more stable and plateau-like market performance. Although the two markets in general varies greatly, during February 2019 to July 2019, both markets showed tranquility and the market reactions were cohesive to each other. On the other hand, the small peaks around August 2019 in both markets showed that the fluctuations may be not caused by the event we are about to investigate.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
0 0.1 0.2 0.3 0.4 0.5 0.6
2017- 01-24
2017- 03-24
2017- 05-24
2017- 07-24
2017- 09-24
2017- 11-24
2018- 01-24
2018- 03-24
2018- 05-24
2018- 07-24
2018- 09-24
2018- 11-24
2019- 01-24
2019- 03-24
2019- 05-24
2019- 07-24
2019- 09-24
2019- 11-24
Standard Deviation
Hong Kong
USA (right scale)
20 To test the effects of political instability on the Hong Kong stock market, the risk premium of the U.S. market will be collected. It will be mainly the changes of the risk premium that can reflect the political turbulences. The U.S. stock market’s risk premium will serve as a great guideline and reference point when analyzing the risk premium of Hong Kong. The risk premium for Hong Kong market will be calculated through Country Bond Default Spread method while US risk premium statistics will be directly retrieved from Market-Risk-Premia.
The changes in daily yields to maturity on Hong Kong and U.S sovereign bonds from 24th January 2017 to 22nd January 2020 are presented in Figure 8 below. The historical figures of 10-year treasury bonds of both Hong Kong and U.S. was retrieved from the website of wall street journal (wjs.com). Note that the overall trends of both lines are almost the same before the protest happened, thus the spread of them are relatively stable. However, as the demonstrations going on and the violent conflicts continue to escalate, the spread is narrowing as expected. The narrowed spread shows that the investors are constantly selling the Hong Kong government bonds, resulting in reduced demand so that the yield to maturity curve did not follow the trend it would have been.
Figure 8. Changes in daily yields to maturity on Hong Kong and USA treasury bonds.
The data of the independent variables 𝑟$ and 𝐻𝐾𝐷$ are both collected from the website of the Hong Kong Monetary Authority while the data of 𝑃𝐼!$ and 𝑃𝐼#$ may have some subjective factors. This is due to the fact that there is no absolute authority to count the specific days of peaceful march or violence between police and protesters.
0 0.5 1 1.5 2 2.5 3 3.5
2017- 01-24
2017- 03-24
2017- 05-24
2017- 07-24
2017- 09-24
2017- 11-24
2018- 01-24
2018- 03-24
2018- 05-24
2018- 07-24
2018- 09-24
2018- 11-24
2019- 01-24
2019- 03-24
2019- 05-24
2019- 07-24
2019- 09-24
2019- 11-24
Yield to Maturity
Country Bonds
Hong Kong USA
21 After careful analysis based on multiple media sources (both pro- and anti-Mainland China sources in order to remain an impartial picture), some criteria are set to help classify these two variables. To start with, most of the protests took place on Saturday or Sunday when the market was closed. The values therefore (1) are postponed to the next trading day (mostly the following Monday), which is reasonable as the market will automatically adjust according to the information available. Secondly, the classification of variables is based only on whether there is violence, that is, whether there is any protester or policeman injured, whether there is a shooting incident, or whether there is any death and so on, regardless of the amount of people on the streets. For example, on the first day of the outbreak of large-scale demonstration on 9th June, although there were 1.03 million (according to protesters)/ 0.24 million (according to the Hong Kong police) people going out, no violence occurred, which is then classified into the 𝑃𝐼!$. Thirdly, if there are peaceful rallies (participants including children and the elderly) and violence at the same time, the day will be placed into 𝑃𝐼#$. Note that when these two types of events occurred within a weekend, the next trading day would record both variables simultaneously. The specific results are presented in Appendix in page 31.
As for the election, it is crucial to have an expectation in foresight for each investor because many international companies are interested in the free business policies and system of Hong Kong instead of mainland of China. If the pro-China party wins the election, investor may think that the investment environment is not good and require more risk premiums. On the other hand, if pro-democracy party defeat their opponents, then this would be regarded as a success of the protests. Regardless of the outcome, people would tend to trade more frequently before and after elections, bullish buy in or bearish sell, which will undoubtedly increase the volatility.
Based on the above analysis, the week before 24th November 2019, the polling day, and two weeks after it would be marked as 1.
In the process of obtaining daily volatilities of the return on Hang Seng Properties index by GARCH model, three years of raw data were used to ensure the existence of ARCH effect.
Three years of data gives also a sufficient quantity which will provide a stable volatility framework. For the regression models, only the dates of interest will be analyzed. Therefore, the market returns from a specific time interval of 23rd January 2019 to 22nd January 2020 will be selected, preferably the period when the protest took place. If all three years of data was included in the regression the 𝑃𝐼!$ and 𝑃𝐼#$ would be irrelevant and meaningless. The selected time horizon will be one year, from 23rd January 2019 to 22nd January 2020 (the initial protest broke out at 9th June 2019). There will be time marginals at the beginning and the end in order to get as reliable and accurate results as possible. If the period is selected precisely on the start and the end of the protest the received results might be false positive or over representative.
The dates before and after the protest will hence serve as control and the results of the political turmoil effects will be more realistic. In total 246 observations will be adopted for the econometric models.
The descriptive statistics for both independent and dependent variables are given in table 3 below.
22
Table 3. Descriptive statistics
𝝈𝒕 𝑹𝒑𝒕 𝒓𝒕 𝑯𝑲𝑫𝒕 𝑷𝑰𝟏𝒕 𝑷𝑰𝟐𝒕 𝑬𝒕
Mean 0.108 0.037 2.509 7.832 0.159 0.278 0.061
Standard Deviation 0.081 0.003 0.288 0.022 0.367 0.449 0.240
Variance 0.007 0.000 0.083 0.000 0.134 0.201 0.058
Minimum 0.030 0.031 2.000 7.767 0 0 0
Maximum 0.424 0.045 2.750 7.851 1 1 1
23
7. Results
Empirical findings and analysis will be presented in this section. There are two main models/descriptors that helps to identify the Hang Seng Properties index’s market performance over 23rd January 2019 to 22nd January 2020.
7.1 Market volatility results
From the results of the models, the conclusion that a close relationship can be discerned between the market volatility and the political risks can be drawn.
Model 1
𝜎$ = 𝛼 + 𝛽! × 𝑟$+ 𝛽#× 𝐻𝐾𝐷$+ 𝛽6× 𝑃𝐼!$+ 𝛽7× 𝑃𝐼#$+ 𝜀$ Model 2
𝜎$ = 𝛼 + 𝛽! × 𝑟$+ 𝛽#× 𝐻𝐾𝐷$+ 𝛽6× 𝑃𝐼!$+ 𝛽7× 𝑃𝐼#$+ 𝛽8 × 𝐸$+ 𝜀$
To illustrate the political risk impacts on the Hang Seng Properties index, the regression model will be formulated in two ways. The first regression will have the standard deviation of HSNP as dependent variable and three independent variables: interest rate (𝑟$), Hong Kong’s foreign exchange rate to U.S. dollar (𝐻𝐾𝐷$), relatively peaceful protest (𝑃𝐼!$) and massive conflict (𝑃𝐼#$). The last two variables are also the dummy variables and intakes the values either zero (nothing happened) or one (ongoing conflict). The second regression will be exactly the same as the first regression model only adding an extra variable denoted as election period (𝐸$), which is also a dummy variable.
Table 4. Regression results of model 1, 𝝈𝒕
Coefficients Standard Error t Stat P-value
Intercept -10.296 1.731 -5.948 0.000
𝑟$ -0.178 0.016 -10.989 0.000
HKD% 1.384 0.222 6.223 0.000
PI&% 0.020 0.012 1.637 0.103
PI'% 0.024 0.010 2.426 0.016
An autocorrelation will be applied on the residuals in Table 4 during the regression analysis.
The adjusted R2 is estimated to 0.408, which means around 40.8% of the observed variations can be explained by the inputs of the regression model. From the table 4, the p-values suggest that effects of 𝑟$, 𝐻𝐾𝐷$ and 𝑃𝐼#$ are statistically highly significant (p-value < 0.05). Since the p-value for 𝑃𝐼!$ > 0.5, this variable is insignificant. This means that the peaceful protest that occurred in Hong Kong seems to have a slighter impact on the volatility of the return on Hang Seng Properties index.
