The effect of stock prices on housing prices :eEvidence from Hong Kong & Singapore

(1)

Form for Thesis

Your name: Hans Redgate

Your student number: 10846581

Track (within Economics and Business): Finance and Organization

Field: Finance: Real Estate, Stock prices.

Number of credits: 12

Research Title:

The effect of stock prices on housing prices: Evidence from Hong Kong & Singapore

Assigned supervisor (to be filled in by thesis coordinator): Name of supervisor: Ryan van Lamoen

(2)

Statement of Originality

This document is written by Student [Hans-Peter Alexander Redgate] who declares to take full responsibility for the contents of

this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text

and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

Real estate and stocks are major asset classes which comprise the net worth of a considerable number of individuals and corporate entities across most developed economies. The real estate market is typically not liquid and heterogeneous while stock markets exhibit characteristics of higher liquidity and changeability. Movements attributable to either will impact the overall well-being of the economy, an individual’s wealth or a firm’s valuation. This study examines the link between Hong Kong and Singapore stock prices and residential real estate prices.

Using quarterly data from 2005 to 2018 the following control variables are introduced i) GDP ii) interest rates iii) CPI alongside stock prices which was adopted as the main explanatory variable. A number of statistical tests were utilised. A time series analysis supplemented by checking the stationarity of the variables was undertaken. Once this and the multiple regression analysis had been compiled the set of tests concluded with a Granger causality test using a vector auto-regression model (VAR). The results collected support the hypotheses that a positive relationship does exist between stock and house prices in both countries.

(4)

Contents- 1. Introduction 2. Literature Review

• 2.1 What theories can describe the connection between stock prices and house

prices?

• 2.2. Empirical evidence on the relationship between stock markets and housing • 2.3 What are other determinants affecting the stock market and housing market? • 2.4 Hypotheses

3. Hong Kong & Singapore real estate survey 4. Data and Methodology

• 4.1 Data (Hong Kong) • 4.2 Data (Singapore) • 4.3 Descriptive Statistics • 4.4 Methodology 5. Results • 5.1 Correlation Analysis • 5.2 Stationarity of Variables

• 5.3 Linear Regressions for Hong Kong & Singapore • 5.4 Dynamic Model

• 5.5 Granger Causality Test

6. Conclusion

7. Limitations of Research 8. References

(5)

1. Introduction

Stocks and real estate display distinct traits. Property is a heterogeneous market; it is typically difficult to find two units the same, while aligning buyers with sellers is often subject to a time lag. Alternatively, the stock market exhibits more liquidity and prone to fluctuations. Notwithstanding these disparities, both present investment opportunities and impact household wealth and represent the major asset classes of households in most developed economies.

Investors have noted this unpredictability of holding stocks and other asset classes in their portfolio; this has led to a higher investment in real estate (Quan & Titman, 1996). It was suggested that real estate as an investment yields high returns with negligible risk (Domain et al, 2015). Domain et al (2015) also showed in their study that ownership of residential real estate decreases portfolio risk while diversified real estate ownership contributes to risk reduction. This area of research is important since it can help individuals and corporate entities make improved investment choices in their portfolio management. Similarly, understanding the relationship may help government policy-makers in defining housing policy and housing regulations across both private and public sectors.

For this study, an analysis of how changes in stock prices affect housing prices is conducted. More specifically, it aims to elucidate if a significant relationship is traceable from stock prices to housing prices while controlling for other factors. Residential real estate (where individuals reside) and not the commercial sector (such as offices, factories or retail outlets) will be considered.

An understanding of how this low covariance between stocks and residential real estate behaves can help us better examine the risk and return trade-offs with such investments (Domain et al, 2015). With more people investing abroad, the Asian market is increasingly relevant with strong growing economies such as Singapore and Hong Kong. Both city states are particularly interesting from a real estate perspective as noted by Chau et al (2001) the Hong Kong market is comparatively liquid.

Hong Kong and Singapore showcase, the city state niche in real estate research which is observable from the researchers quoted in this study. While this thesis can add to this repository of knowledge what comes to mind as this study evolves is that other city states in other time zones can be investigated. The five small states of the Persian Gulf: Abu Dhabi Bahrain, Dubai, Kuwait and Qatar exhibit some rather different economic characteristics to

(6)

Hong Kong & Singapore except land area. Brunei and Luxembourg city could also merit consideration. From a methodological perspective, the investigation of dynamic effects and Granger causality is introduced into the time series.

Drawing on previous studies, the key deliverable is an empirical study of Hong Kong and Singapore to address the main research question of this thesis:

The effect of stock prices on housing prices: Evidence from Hong Kong & Singapore

In support of the main research question, the following are considered: What theories describe the connection between stock markets and housing markets. Secondly, what is the empirical evidence linking stock prices with housing prices and lastly what are the other determinants of house prices. In addition, a survey is included examining the particular characteristics of Hong Kong & Singapore relating to real estate.

The thesis is structured as follows: After the introduction, the literature gathered will be reviewed for the determinants of house prices, including stock prices. In the following section the survey is captured thereafter a review of the empirical methods, results and discussion of the results are presented. Then the conclusion is compiled together with the main findings of this study.

2. Literature Review

The connection of stock prices and house prices inspires significant research interest in both higher educational and real estate circles. Discussion associated with different methods of analyses and diverse aspects, across both developed economies and some emerging economies is apparent. It is contextual to take the movement of house prices as being triggered by changes in both demand and supply factors, just like the pricing of any asset, good or services according to economic theory. Demand-side factors typically mean macroeconomic conditions – such as income growth, unemployment rate, and interest rates and inflation. Additionally, for city economies with significant migration movements the availability of labour is also a significant factor on the demand side. From a supply side perspective, the land area available for building on and the conversion rate /lead time of going from planning proposal to build completion is important.

(7)

The first theory to be discussed is termed the “wealth effect” which regards houses as consumer goods. It is described by Yuksel (2016) and Sim and Chang (2006). House price fluctuations are triggered by demand and supply. So if the value of stocks increase, investors will perceive their wealth to be growing, this sentiment leads to an increase in their demand for housing and so resulting in an upward pressure on house pricing. The long-term nature of the housing market means it is typically static in the short term as constructing houses is a lengthy process. This is likely to magnify the demand pressure further and accelerate house price inflation.

The second theory is referred to as the “credit price effect” which takes account of the net asset position of the firm in particular that real estate would comprise a significant proportion of the companies fixed asset portfolio. A rise in the value of real estate on the firm’s balance sheet would overall increase the value of the firm. A higher valuation would be one factor increasing the propensity of the company to be able to borrow at more favourable interest rates and therefore invest more and increase a company’s creditworthiness. The investments are taken to generate positive net present values and increase the firm’s valuation and share price. Both effects have been discussed in other studies including Batayneh & Al-Malki (2015) and Lean (2012). So, the credit price effect is synonymous with real estate affecting the share price.

Apparent from the two mechanisms, is the wealth effect which is related to residential real estate and therefore homes for families and individuals while the credit effect is aligned with corporate entities. This study excludes corporate real estate so the credit-price effect is unlikely to be applicable.

2.2. Empirical evidence on the relationship between stock markets and housing

Quan & Titman (1996) utilised data from a cross section of developed economies to analyse the interaction surrounding stock prices and price movements in commercial properties. The study concluded that the link between commercial building prices and stock prices was not statistically significant in 16 out of 17 countries. In the same study Quan & Titman (1996) made the following observations on Singapore and Hong Kong:

“In the smaller Asian countries, like Singapore and Hong Kong, the factors inducing a negative relation between stock returns and real estate prices may be less important. First, the stock market fluctuations in these economies seem to be due more to demand side effects (e.g.,

(8)

increased trade with China) than cost side effects (e.g., reduced labor costs). Second, these small, relatively open economies, experience increased capital inflows when their investment opportunities improve. Hence, the increased demand for office space that corresponds to an economic boom is not likely to be offset by higher interest rates. (This is especially true in Hong Kong where the interest rates are essentially the same as the U.S. rates because of the fixed exchange rate between Hong Kong and U.S. dollars.) “(P 5-6)

Abelson et al (2005) pointed out in a study looking at data in Australia between 1970 and 2003 that stock price movements were associated with house price changes. Sim and Chang (2006) used an econometric model to test for the link between stock and real estate prices using data from South Korea and obtained results that support the “credit price effect” mechanism.

Sutton (2002) used an econometric model called a vector autoregressive model (VAR), which included house prices, national income, interest rates and stock prices and chose 6 countries with a long availability of quarterly housing data. These countries were Australia, Canada, Ireland, the Netherlands, the United States and the United Kingdom. Sutton via the VAR noted how house prices were affected by a 10% rise in stock prices. This resulted in approximately a 1% rise in house price in Canada, Ireland and the United States; instead for Australia and the Netherlands it was a 2% hike over three years. While the biggest increase was noted in the UK with a 5% increase over three years.

Liow (2004) conducted a Singapore based study looking at the short-term and long-term interactions of the stock and property markets. For a short-term horizon, residential real estate prices had a greater impact on stock and property stock prices, but in the long run results suggested a contemporaneous connection of stock prices (property stock prices) and real estate prices.

2.3 What are other determinants affecting the stock market and housing market?

Other factors also influence the two markets as illustrated above at the commencement of this section. However, a deeper dive into the available literature on all these variables is to be limited and the decision is to focus on those (macro-economic) variables which have been designated as control variables in this study. Control variables are there to make sure we isolate as much as possible the effect of stock prices on residential retail prices.

(9)

Interest rates even from a layman’s perspective are visible from the house price debate; it is a topic of everyday mainstream conversation and of monthly financial importance in triggering a significant cash outflow for many households. Similarly, when a property is purchased a mortgage is typically needed; when interest rates are lower, more individuals are likely to buy a dwelling. Abelson et al (2005) found in the long run that real mortgage rates are a significant determinant of house prices. Sutton (2002) via the VAR model showed a 1% decrease in short term interest rates involved a rise in house price across a range of half of one per cent to one per cent over four quarters, while it is weaker for drops in long term interest rates.

Nakajima (2011) discusses the user cost theory of housing and rents. User costs are the cost of owning a house instead of renting so changes in the costs thereof should affect the house price. There are five types of user cost including interest rates. Poterba (1984) and Himmelberg et al (2005) explain various components. House dwellers are required to pay property tax as part of the user costs. Secondly, in some countries; interest payments are tax deductible this occurs in Singapore for the interest portion and Hong Kong. Next homeowners need to pay for repairs and maintenance. Lastly, future house price fluctuations, meaning inflation is reflected in the user cost. Interest rates are an important user cost category with respect to housing demand. A lower interest rate would typically result in more demand for houses. A steeper interest payment would be linked with the user cost rising and a more expensive house. Chung (2012) using quarterly data from 1984 to 2009 noted that interest rates have a negative relationship in affecting the residential housing prices while stock prices a positive relationship.

Economists often distinguish between nominal and real interest rates. The nominal interest rate would be the rate quoted on a house buyer’s mortgage agreement if say the prospective owner had negotiated a fixed rate deal with the mortgage lender. However, it is the real interest rate, the inflation adjusted rate which is more critical. This relationship is captured by the Fisher equation (1930).

r = i - π -where r denotes the real interest rate, i denotes the nominal interest rate, and π denotes

the inflation rate

Harris (1989) investigated the impact of real rates of interest on housing prices and concluded that the real interest rate is the underlying characteristic affecting changes in house prices, because the nominal rate lags behind in reflecting adjustments in expectations. In this study, the nominal interest rate is used in the data analysis.

(10)

2.3.2 Gross Domestic Product (GDP)

In addition to the interest rate, GDP is also used as a control variable in this study. GDP is one of a number of similar measures that relate to national income. National Income being the total value of all goods and services both within a country and income from abroad while GDP is specific to within the country. Overall, the measures indicate the state of the general economy.

Lam (2016) in a study of Hong Kong residential housing prices used GDP per capita which is the GDP divided by the number of people in the population. Lam identified that previous studies had rarely considered the role of the household and introduced the variable GDP per household because of contracting household size. He reported a reduction to 2.9 persons in 2014 from 3.6 persons in 1989. This is important for policy-makers in emphasising the need for smaller-sized flats available for single person occupancy. Lam noted in the results that larger-sized units where more price sensitive to changes in GDP. Previously, Cheng & Fung (2015) noted when using a GDP per capita measure as a variable that housing prices are positively related to GDP.

Away from Hong Kong, Sutton (2002) and Englund & Ioannides (1997) reported that growth in GDP was positively related to changes in house prices. Englund & Ioannides (1997) found the rate of growth of GDP had a strong predictive capability. Such that a 1 percent faster GDP growth in one year gave a 0.77% house price growth tomorrow.

2.3.3 Other determinants

A well-established metric in the housing market is the price to rent ratio, it is typically calculated as the ratio of house prices to the annual rental costs when comparing like for like units in the same location. The ratio is helpful in understanding real estate and broader macro-economic trends and reflects the importance of the trade-off between the two sub markets of rental and purchase. Hargreaves (2008) noted that house prices had risen more rapidly that rents since 2000 in many developed economies. Locally, in New Zealand Hargreaves (2008) also reported that inward net migration was a key driver in rental demand and from the research concluded rents are helpful in forecasting house prices. Chung (2012) also reported in a Hong Kong study that housing rental was an important determinant of residential real estate prices.

A relationship between inflation and house prices is applicable if one considers that any good with a limited supply will generate inflationary pressure. Ceteris paribus if the central

(11)

bank of a country printed money to increase the money supply and the number of houses did not change then inflation and house prices would increase. In the real economy inflation interacts with other factors like interest rates as noted by Harris (1989). For the Hong Kong hypotheses, the Consumer Price Index (CPI) is introduced as an additional control variable.

2.4 Hypotheses

To address the research question, 3 hypotheses can be deduced from the reviewed literature. Firstly, whether stock prices affect housing prices in both countries?

This thesis will investigate the relationship using returns from both the stock and housing market. The two hypotheses derived are:

Hypothesis 1: House prices in Singapore are affected by changes in Stock prices Hypothesis 2: House prices in Hong Kong are affected by changes in Stock prices The third hypothesis will be to determine which of the two models better predicts these changes, a comparison of both linear models will help to conclude this. A more informed answer should allow for investors to make safer real estate investments.

Hypothesis 3: One of the models will better explain this relationship.

3. Hong Kong & Singapore real estate survey

Hong Kong and Singapore are both Asian city states which exhibit similarities both culturally and economically. The two islands have similar population sizes and density with a large proportion of their populations being Chinese. Each is an advanced economy with a similar GDP and lacking in natural resources, which includes the land area. Geo politically both are key hubs for the larger country neighbours of China, Indonesia and Malaysia respectively.

Table 1- Characteristics of countries

Characteristic Unit Hong Kong Singapore

Land Sq. Km 1073 709.2

Population M 7.19 5.88

GDP B$US 334.1 305.8

(12)

Both countries have similar mixes of housing with both public and private interventions observable. From a public perspective, each government has implemented substantial housing programmes via the Hong Kong Housing Authority (HA) and the Housing Development Board (HDB) which is incumbent in Singapore. The main objectives of these organisations are schemes which develop a supply of rental properties and build subsidised housing which is made available to lower income families and individuals. Both were principled to provide a minimum standard of housing for all residents. Over time conditions and specifications of public housing have improved. The private housing market in both city states is comprised of individual residential buildings and blocks of flats sometimes referred to as a condominium which is also a multi-unit building where individuals own or rent individual units. In 2017 80% of Singaporeans live in government-built flats. The 2016 Hong Kong Census identified 53.2% of the population resided in private housing. A fixed exchange rate is applicable in Hong Kong in 2018.

4. Data and Methodology

In this section, the sources and quality of data will be explored. Sections 4.1 and 4.2 have been established to explain Hong Kong and Singapore separately. Section 4.3 will then give the descriptive statistics of both datasets, for a better understanding of the statistical properties. Finally, Section 4.4 will address the methodology of this paper. In regard to the added Consumer Price Index (CPI) variable, the investigation with this control variable was only conducted for Hong Kong. This was due to inconsistencies of the CPI series and availability of data across both countries.

4.1 Data (Hong Kong) 4.1.1 Property Index

The first variable which will be discussed is the Property Price index, the chosen index used in this investigation was the Centa-City Index (CCI). The Centa-City index collates data on a monthly basis using all Contract Price Data transactions logged with the Land Registry. Property price movements are from the previous month. With regards to the quality of the published data from the CCI, it is noted that Centaline Property Agency Limited has over 20% of the property agent market in Hong Kong. With this statistic, we can assume the index to be a reliable source and provides the most recent transaction data in the market.

(13)

The Property Price index comprises a number of constituent housing estates. With the base period of July 1997 being used. The CCI is calculated using the formula:

𝐶𝑒𝑛𝑡𝑎 − 𝐶𝑖𝑡𝑦 𝐼𝑛𝑑𝑒𝑥 (𝐶𝐶𝐼)𝑓𝑜𝑟 𝑎 𝑚𝑜𝑛𝑡ℎ

= 𝑇𝑜𝑡𝑎𝑙 𝑚𝑎𝑟𝑘𝑒𝑡 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑐𝑜𝑛𝑠𝑡𝑖𝑡𝑢𝑒𝑛𝑡 𝑒𝑠𝑡𝑎𝑡𝑒𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑚𝑜𝑛𝑡ℎ 𝑇𝑜𝑡𝑎𝑙 𝑚𝑎𝑟𝑘𝑒𝑡 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑐𝑜𝑛𝑠𝑡𝑖𝑡𝑢𝑒𝑛𝑡 𝑒𝑠𝑡𝑎𝑡𝑒𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑝𝑟𝑒𝑣𝑖𝑜𝑢𝑠 𝑚𝑜𝑛𝑡ℎ ∗ 𝐶𝐶𝐼 𝑓𝑜𝑟 𝑡ℎ𝑒 𝑝𝑟𝑒𝑣𝑖𝑜𝑢𝑠 𝑚𝑜𝑛𝑡ℎ

(Centa Property Index, 2018)

The total market value of the estates is equivalent to the total saleable area and adjusted unit price. To understand the adjusted unit price, unit price is defined as being equal to the transaction price per square foot. The adjusted unit price of an estate is the price average once factors such as views, direction, floor levels for a given period are omitted (Centa Property Agency, 2018).

Monthly data was collected for the following time period 2006-2018, which was then averaged to obtain quarterly data from 2006Q1–2018Q1. Quarterly changes between periods were then computed afterwards for the regression.

4.1.2 Independent Variables (Hong Kong)

The independent variables used for the Hong Kong analysis were: 1. Stock Price (Hang-Seng Index)

2. Gross Domestic Product (GDP) 3. Interest Rate (Interbank)

4. CPI (High Expenditure Level)

All the data used for the variables are changes in quarterly data from the period 2006 Q1 to 2018 Q1.

The main explanatory variable used in this investigation was stock prices gathered from the Hang-Seng Index (HSI). The HSI is a free-floating adjusted market value weighted index using a selection of companies. The weighting is measured via the total market value of outstanding shares. The index is split into four areas: Commerce and Industry, Utilities, Property and Finance.

(14)

𝐶𝑢𝑟𝑟𝑒𝑛𝑡 𝐼𝑛𝑑𝑒𝑥 = Σ[P(t)∗ 𝐼𝑆 ∗ 𝐹𝐴𝐹 ∗ 𝐶𝐹]

Σ[P(𝑡₋₁)∗ 𝐼𝑆 ∗ 𝐹𝐴𝐹 ∗ 𝐶𝐹]∗ 𝑌𝑒𝑠𝑡𝑒𝑟𝑑𝑎𝑦

′_{𝑠 𝐶𝑙𝑜𝑠𝑖𝑛𝑔 𝐼𝑛𝑑𝑒𝑥}

The HSI has 25% of its stock market capitalization comprised of real estate companies and non-property companies involved in the investment and development of real estate (Chau et al., 2001). Daily prices were collected from DataStream and then transformed into quarterly returns.

For GDP current market prices were used, this data was compiled from the Census and Statistics department of the Hong Kong government. GDP is calculated using a production approach analysing the economic activity of Hong Kong (Census & Statistics department, 2018); it is an aggregate measure of the total value of the net output of all resident producing units. The interest rate used was the HIBOR (Hong Kong Interbank Offered Rate) which is published by the Hong Kong Monetary Authority. The interbank rate is the rate of interest that is charged on short term loans made between banks. Similarly, data was sourced daily from DataStream and then converted to a quarterly measure.

CPI measures the changes in the price level of consumer goods over time. The rate of change for CPI is generally used as a proxy to determine how inflation is affecting consumers. The different types of CPI published by the Census and Statistics Department, reflect how consumer prices change for households with different levels of expenditure. A high expenditure pattern of CPI was selected for this investigation, as this best reflects the costs of purchasing houses. Quarterly data was collected from the Census and Statistics department in Hong Kong, where quarterly changes were then computed.

4.2 Data (Singapore)

All quarterly data for Singapore were collected for the period 2005 Q4-2018 Q1.

4.2.1 Property Index Singapore

The property index which was used for Singapore is available from the Urban Redevelopment Authority (URA). The index is computed using a hedonic regression model, with data compiled from transaction prices and units sold by licensed developers. Hedonic regression analysis looks into the deconstruction of house values into different components and examines how the composite sum of the element contributes to the overall price (Jiang et al.,

(15)

2014). Quarterly data was downloaded from the Singapore government Department of Statistics.

4.2.2 Independent Variables

The independent variables used for the Singapore analysis were: 1. Stock Price (Hang-Seng Index)

2. Gross Domestic Product (GDP) 3. Interest Rate (Interbank)

The stock index used was the Strait Times Index (STI), it is a capitalization-weighted stock index. The STI is Singapore’s benchmark stock index, which includes the largest 30 companies in the country. The interest rate used was the Interbank rate set by the Monetary Authority of Singapore. Both sets of daily data were found on DataStream, and quarterly changes in data were computed. For the final control variable GDP, current market prices were collected. GDP was used to determine the income level of Singapore. Quarterly data was collected from the Singapore Department of Statistics.

4.3 Descriptive Statistics

Table 2 below describes the descriptive statistics for the quarterly returns of Hong Kong. Both the stock and property index yield similar average returns, both at 2.026% and 2.025% respectively. The volatility of the Stock index (10.3%) is much higher than the Property index (4.77%), which relates back to how investment in stocks are shown as being more risky and unpredictable (Quan & Titman, 1996). The interbank rate displays the highest volatility of 37.9% and a large range, this could possibly be associated with the time period chosen (Pre-Financial Crisis to Post (Pre-Financial Crisis in 2007-2008). GDP shows a low return of (1.6%) and volatility of (6.6%). The CPI variable is seen to have a low standard deviation of (0.7%) and a mean of (0.7%), which implies that there is very little change on a period to period basis

(16)

Table 2 – Descriptive Statistics (Hong Kong)

Variable Observations Mean Std. Dev. Min Max

ΔStockIndex 48 0.020261 0.103608 -0.3040719 0.2842592

ΔPropertyIndex 48 0.020254219 0.0477733 -0.1495458 01503129 ΔInterbankRate 48 0.016273 0.379326 -0.8591763 1.465976

ΔGDP 48 0.0160162 0.066415 -0.1221902 0.1191856

ΔCPI 48 0.0075632 0.007645 -0.0067961 0.0227173

Table 3 below shows the descriptive statistics of Singapore, where 49 periods were examined from 2005Q4 to 2018Q1. The average returns recorded for both stock and property index are 1.34% and 4.43% respectively. Despite the lower average return for the stock index, the maximum return is still higher than the property index. Higher volatility (8.95%) as expected was exhibited for the stock index as with added risk, higher compensation is needed. The property index shows a standard deviation of 4.3%. The interbank rate again was shown to have the highest volatility out of the variables with 37.3%. GDP presents the lowest standard deviation of (3.06%) and a return of (1.66%) .

Table 3- Descriptive Statistics (Singapore)

Variable Obs Mean Std. Dev. Min Max

ΔStockIndex 49 0.013425 0.0894915 -0.3370072 0.2658519 ΔPropertyIndex 49 0.0443442 0.0431578 -0.1511036 0.1594963 ΔInterbankRate 49 0.0109538 0.3737742 -0.5592808 1.020229

(17)

4.4 Methodology

Time series analysis was the main research method deployed, and alongside time series data comes the importance of identifying the stationarity of variables. A stationary time series will allow for better forecasting. Unit root tests will check for constant mean, variance and autocorrelation across the series. To check for stationarity, an Augmented Dickey Fuller Test with no lags was conducted for each variable. A further review for 2 periods of lagged differences was undertaken on the main explanatory and dependent variable to ensure no unit root existed.

To answer the hypotheses, multiple linear regressions will be tested for the relationship between house and stock prices. In the base model, GDP and interest rate will be used as control variables when monitoring for changes. A further regression, with CPI used as an additional control variable was introduced for Hong Kong.

Linear regression model for Singapore:

𝐻𝑃 = 𝛽₀+ 𝛽₁𝑆𝑡𝑜𝑐𝑘𝑃𝑟𝑖𝑐𝑒_𝑡+ 𝛽₂𝐺𝐷𝑃_𝑡+ 𝛽₃𝐼𝑅𝑡 + 𝑒

Linear regression model for Hong Kong:

𝐻𝑃 = 𝛽₀+ 𝛽₁𝑆𝑡𝑜𝑐𝑘𝑃𝑟𝑖𝑐𝑒_𝑡+ 𝛽₂𝐺𝐷𝑃_𝑡+ 𝛽₃𝐼𝑅𝑡 + 𝑒

Where :𝐻𝑃 = 𝐻𝑜𝑢𝑠𝑒 𝑃𝑟𝑖𝑐𝑒, 𝐺𝐷𝑃 = 𝐺𝑟𝑜𝑠𝑠 𝐷𝑜𝑚𝑒𝑠𝑡𝑖𝑐 𝑃𝑟𝑜𝑑𝑢𝑐𝑡, 𝐼𝑅 = 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑅𝑎𝑡𝑒, 𝑡 = 𝑓𝑜𝑟 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑝𝑒𝑟𝑖𝑜𝑑

Linear regression model for Hong Kong with added control variable (CPI):

𝐻𝑃 = 𝛽₀+ 𝛽₁𝑆𝑡𝑜𝑐𝑘𝑃𝑟𝑖𝑐𝑒_𝑡+ 𝛽₂𝐺𝐷𝑃_𝑡+ 𝛽₃𝐼𝑅𝑡 + 𝛽₄𝐶𝑃𝐼𝑡 + 𝑒

Additional research will be carried out to investigate for dynamic effects in the model. OLS will be ran on the dynamic model in order to investigate for 2 lagged periods of the Property Index in both Hong Kong and Singapore.

Dynamic Model for Hong Kong:

(18)

Dynamic Model for Singapore:

𝐻𝑃 = 𝛽0+ 𝛽1𝑆𝑡𝑜𝑐𝑘𝑃𝑟𝑖𝑐𝑒𝑡+ 𝛽2𝐺𝐷𝑃𝑡+ 𝛽3𝐼𝑅𝑡 + 𝛽4𝐻𝑃𝑡−1+ 𝛽5𝐻𝑃𝑡−2+ 𝑒

If dynamic effects were present in the dataset, then a Granger analysis via VAR regression approach will be performed. A VAR analysis will allow us to better capture the linear interdependencies among multiple time series. Akaike and Bayesians checks will determine the optimal lag structure to use in the autoregressions. The Granger analysis will then determine which variables and lagged variables are significant and jointly significant in explaining the Property Index.

5. Results

This chapter will discuss the results and findings of the study. It will begin by analysing the bivariate correlation between variables of both Hong Kong & Singapore in section 5.1. The next section 5.2 will be looking into the stationarity of the variables while Section 5.3 will be discussing the results of the linear regression models. Finally, Section 5.4 and Section 5.5 will be discussing the results from the dynamic model and Granger analysis.

5.1 Correlation Analysis

5.1.1 Correlation Matrix (Hong Kong)

Examining the correlation between variables, will allow us to determine the type of relationship which exists. Table 4 below shows a correlation matrix between the variables. We can see that the Property Index and Stock Index are positively correlated, they are significant at the 0.01% significance level. GDP and interest rate were also shown to be significantly correlated with the same significance level. The Interbank rate shows a small positive correlation with the stock index and a zero correlation with the property index. The correlation of GDP with the stock index is positive and nearly zero when compared with the property index. CPI is shown to have low and insignificant correlations with all the variables in the model.

(19)

Table 4- Correlation Matrix (Hong Kong)

*** 1% Significance Level ;** 5% Significance Level;*10% Significance Level

Table 5- Correlation Matrix Singapore

Table 5 shows the bivariate correlation matrix for Singapore. The property index is significantly correlated to the stock index at the 0.01% level. The interbank rate has a nearly zero correlation with the stock index and a 0.1248 correlation with the property index. GDP is significantly correlated with the Property index at the 0.01% level, while it has a near zero correlation with the interbank rate and 0.1024 positive correlation with the stock index.

StockIndex PropertyIndex Interbank Rate GDP CPI Stock Index 1 Property Index 0.5120*** 1 Interbank Rate 0.0001 0.1248 1 GDP 0.0315 0.2088 0.3706*** 1 CPI 0.1140 0.0239 -0.0597 -0.0041 1 Stock Index Property Index Interbank Rate GDP Stock Index 1 Property Index 0.4669*** 1 Interbank Rate 0.0609 0.1248 1 GDP 0.1024 0.4425*** 0.0746 1

(20)

5.2 Stationarity of Variables

5.2.1 Stationarity test for all variables, no lags

To test for the stationarity (unit root) of the variables, an Augmented Dickey Fuller test was performed. Firstly, the stationarity of all the variables were checked with no lags for Hong Kong. Results are shown in table 6.

Table 6: Zero Lags Stationarity Test (Hong Kong), n=47

Test Statistic, Z(t) P Value, Z(t)

PropertyIndex -4.172*** 0.0007

StockIndex -5.378*** 0.0000

InterbankRate -5.411*** 0.0000

GDP -6.775*** 0.0000

All variables in the Hong Kong model were shown to be stationary at the 1% significance level. This can be seen from the interpretation of the test statistic, where the t-values were seen to fall within the 1% confidence interval across the variables. All the variables can be assumed to exhibit 3 characteristics of constant mean, variance and autocorrelation, when no lagged effects are tested for.

The same test was conducted on the variables for Singapore to check for stationarity, results are shown below in table 7.

(21)

Table 7 Zero Lags Stationarity Test (Singapore), n=48

Test Statistic, Z(t) PValue, Z(t)

PropertyIndex -5.269*** 0.0000

StockIndex -4.642*** 0.0001

InterbankRate -7.180*** 0.0060

GDP -3.586*** 0.0000

*** 1% Significance Level; ** 5% Significance Level; *10% Significance Level

All the variables of the Singapore model can be viewed as being stationary at the 1% significance level, which shows a constant mean, variance and autocorrelation across the time series with zero lags.

5.2.2 Stationarity of Property Index and House Index with lagged differences

In regard to stationarity of the dependent and main explanatory variable, an additional investigation was done considering lagged differences. To decide how many lagged periods to include, 2 lagged periods were chosen to see how robust the conclusions are about the unit root. Table 8 and Table 9 illustrates these results for the Stock Index and Property Index of Hong Kong.

Table 8: Stock Index with 2 lagged periods, for n=46 & 45

StocIndex(L1) -4.934*** 0.0000

StockIndex(L2) -4.356*** 0.0004

(22)

Table 9 Property Index with 2 lagged periods, for n=46 & 45

As seen from the tables, both stock and property index variables are stationary at the 1% significance level for both lagged periods. The subsequent tables 10 and 11 are for checking the stationarity of the two variables in Singapore.

Table 10: Stock Index with 2 lagged periods, for n=47 & 46

StockIndex(1) -4.899*** 0.0000

StockIndex(2) -3.900** 0.0020

*** 1% Significance Level ; ** 5% Significance Level;*10% Significance Level

Table 11- Property Index with 2 lagged periods, for n=47 & 46

The results show that the Singapore Stock index is stationary for both periods at the 1% significance level. The property index is significantly stationary at the 1% level for 1 lag period, while with 2 lag periods, the variable is only stationary at the 5% significance level. Stronger evidence can be drawn for the housing variable in Hong Kong being more stationary than in Singapore.

PropertyIndex(1) -5.324*** 0.0000

ProeprtyIndex(2) -5.846*** 0.0000

PropertyIndex(L1) -5.324*** 0.0000

(23)

5.3 Linear Regressions for Hong Kong & Singapore

Table 12- Linear Regressions for Hong Kong and Singapore

ΔPropertyIndex (Hong Kong) ΔPropertyIndex (Singapore) ΔPropertyIndex (Hong Kong) ΔStockIndex 0.2446526 (0.061)*** 0.203 (0.056)*** 0.2443 (0.061)*** ΔInterbankRate -0.0053559 (0.017) 0.0080498 (0.013) -0.0044947 (0.017) ΔGDP -0.0457108 (0.101) 0.5545588 (0.166)*** -0.0468061 (0.1019) ΔCPI 0.6183366 (0.8108) N 48 49 48 Adjusted R2 0.2199 0.3390 0.2124 F value 5.42 9.20 4.17***

(Standard errors are included in the parentheses)

As discussed in the methodology, a linear regression of the independent variables was ran against the property index. This was to check for the relationship between stock and house prices. The adjusted R^2 value of both models for Hong Kong and Singapore were 0.21999 and 0.3390 respectively. The adjusted R^2 value gives the total explained variance of the model, which in this case is 21.99% and 33.9%. To justify the low R^2 value in the model for both countries, it can be attributed to whether a type 2 error is present for sample size, time periods investigated and other outliers with the data. Quan & Titman (1996) conducted a time series investigation for the same countries but with different periods and have also observed low values of explanation. A further analysis looking at the significance of variables is also important to fully suggest whether the model is valid.

(24)

The F statistic for the model of Hong Kong and Singapore are given as F(3,44)= 5.42, p(F)> 0.0029 and F(3,45)= 9.20, p(F)>0.001. This implies that the variables used are jointly significant and improve the fit of the model. Upon closer inspection of the coefficients, it can be seen that stock prices were significant in explaining this change. In both models, a positive relationship is exhibited between the two variables. Interest rates were shown to be insignificant in explaining the changes with house price, a near zero relationship is observed for both models. The coefficient GDP was inconsistent across both models, in Singapore GDP was significant in explaining the changes in house price with a positive relationship of 0.5545, but with Hong Kong the coefficient was insignificant and a negative relationship was seen at -0.05.

An extra regression was done for Hong Kong, with CPI added to the basic model. Due to inconsistencies and availability of CPI data across both countries, this regression was only conducted for Hong Kong. The adjusted R2 is similar to the original model, the addition of the new variable CPI gives the value 0.2124. This indicates that the total explained variance of the model is 21.24%. The F statistic is F(4,43) = 4.17, p(F)> 0.0061, which indicates the variables are jointly significant and help in improving the fit of the model. Further inspection on the coefficients, indicate that stock index has remained significant at the 1% level in explaining changes with house prices. The other variables have remained insignificant in explaining the changes in house prices, it was noted that both the Interbank Rate and GDP still exhibit a near zero relationship. The new variable CPI was seen to be insignificant in the model, showing a positive relationship of 0.618336.

(25)

5.4 Dynamic Model

Table 13- Dynamic Model for Hong Kong and Singapore

ΔPropertyIndex (Singapore) ΔPropertyIndex (Hong Kong) Stock Index -0.0574 (0.0938) 0.1834 (0.0646) *** PropertyIndex_L2 -0.1752 (0.1417) -0.2001 (0.1570) GDP 0.5687 (0.1630) -0.0663 (0.0946) PropertyIndex_L1 0.7400 (0.2048)*** 0.4665 (0.1444)*** InterbankRate 0.0151 (0.0139) -0.004 (0.0164) CPI 0.253478 (0.7974) N 47 46 Adjusted R2 0.5405 0.3557 F 9.65 5.14

(Standard errors are included in the parentheses)

(*** 1% Significance Level ;** 5% Significance Level;*10% Significance Level

To investigate whether house prices can be predicted by previous periods, a dynamic model with two lagged periods of the dependent variable was suggested. As seen from table 13, dynamic effects are important in determining house prices for both Hong Kong and Singapore. House prices are explained by its lagged variables in both countries, the single lag period of the property index significantly explains the property price at the 1% significance level. The addition of the lagged variables has also improved the explanatory power as compared to the original OLS model, the new adjusted R2 value of the model are 0.3557 and 0.5405 respectively. This shows that the total variance of the model explained is 35.57% and 54.05%. Stock price remains significant at the 1% level in explaining house prices in Hong Kong, despite a decrease in the coefficient. However, in Singapore the previous positive

(26)

relationship held between the stock and housing index has become negative. The stock index has also been shown to be insignificant in explaining changes. To justify the changes in the significance levels and coefficients of the model, a combination of high serial correlation and presence of exogenous variables can influence this (Achen, 2001).

5.5 Granger Causality Test

5.5.1 Determining Optimal Lag Structure

As seen from the previous section, dynamic effects were present in the model. Hence, further research into the multiple time series was needed to infer for causality between stock and house prices. A Granger causality analysis was conducted via a vector autoregression approach (VAR). Firstly, an optimal lag structure has to be determined for the autoregressions. Akaike (AIC) and Bayesian (SBIC) checks were performed to test for the fit, and these results are shown in Table 14 and 15 below for both Hong Kong and Singapore respectively.

Table 14- Akaike and Bayesian tests Hong Kong

*** 1% Significance Level; ** 5% Significance Level; *10% Significance Level, Full table included in the Appendix

Lags Df AIC SBIC

0 -13.6916 -13.4888

1 25 -14.2924 -13.076

2 25 -15.8631 -13.6329***

3 25 -15.9014 -12.6574

(27)

Table 15- Akaike and Bayesian tests Singapore

As seen from the tables, the optimal lag structure for Hong Kong at the 1% significance level was 4 and 2 lags respectively for both Akaike and Bayesian test. For Singapore, the optimal lag structure at the 1% significance level was a 1 lag order for the Bayesian test and 4 lags with the Akaike test.

5.5.2 Granger Causality test (Hong Kong)

With the determined lag structure, a Granger causality test can be used to check for causal relationships. Tables 16 and 17 includes the Granger causality outputs for both lag structures in Hong Kong.

Table 16- Granger Results Hong Kong (2 lags)

Equation Excluded F df Df_r Prob>F

Property Index StockIndex 6.0778*** 2 35 0.0054 Property Index InterbankRate 3.6302** 2 35 0.0369

Property Index GDP 0.48418 2 35 0.6203

Property Index CPI 0.68818 2 35 0.5092

Property Index All 3.1489*** 8 35 0.0086

Lags Df AIC SBIC

0 -8.95748 -8.79689

1 16 -10.266 -9.4636***

2 16 -10.2887 -8.84336

3 16 -10.1412 -8.05352

(28)

Table 17-Granger Results Hong Kong (4 lags)

Property Index StockIndex 2.3425* 4 23 0.0850

Property Index InterbankRate 1.1302 4 23 0.3669

Property Index All 1.6467 16 23 0.1340

Degrees of freedom (df) on these tables are denoted for the lag structure. The results shown are to check for Granger causality in each variable individually and all added variables jointly in the VAR. As seen for the 2 period lag structure all variables jointly explain house prices at the 1% significance level, F(8,35)= 3.1489. This was not the case with the four period lag structure though. Newey tests were considered when checking for the fit of individual lagged variables, as this test corrects for autocorrelation in the error term. F test statistics were computed afterwards to see for the significance of individual variables. Table 18 below illustrates these results.

(29)

Table 18- Newey F tests 2 Lags; F(2,35) 4 Lags; F(4,23) Stock Index 7.06*** 1.71 Interbank Rate 3.40** 2.11 GDP 0.61 0.41 CPI 0.66 0.71

As seen from the results, the Stock Index does have a significant causal relationship with house prices. The F statistic F(2.35)= 7.06 is significant at the 1% level in explaining this. In addition, the interbank rate also significantly causes changes in house prices at the 5% signficance level, F(2,35)= 3.40. From all these results for Hong Kong, additional robustness checks can be determined. The VAR model with the 4 period lag structure does not better explain house prices as compared to the 2 period lag structure, this is illustrated both by the variables jointly and individually.

5.5.3 Granger Causality Test (Singapore) Table 19- Granger Results Singapore (1 lag)

Property Index StockIndex 5.1242** 1 43 0.0287 Property Index InterbankRate 1.033 1 43 0.3151

Property Index GDP 4.8487** 1 43 0.0331

Property Index All 4.2323** 3 43 0.0104

(30)

Table 20- Granger Results Singapore (4 lags)

In both lag structures, it can be inferred that all lagged variables do jointly support house prices. The results were shown to be significant at the 1% level at F(12,28)= 4.2355 in the 4 period lag structure. For the structure with 1 lag period, it was illustrated to be significant at the 5% level at F(3,43)= 4.2323. Again, a Newey test was used to correct for autocorrelation when comparing for the fit of individual lagged variables. Table 21 below displays these results.

Table 21- F-Test results (Newey)

2 Lags; F(1,43) 4 Lags; F(4,28)

Stock Index 1.64 1..30

Interbank Rate 0.91 1.00

GDP 3.18 5.87***

The added lag variables were only shown to be significant with 4 lags for the GDP variable. GDP was shown to be significant in causing house prices at the 1% level, F(4,28)=5.87.

6. Conclusion

The main goal of this paper was to establish whether changes in stock prices has effects on housing prices in Hong Kong & Singapore. It started by examining this relationship using a bivariate correlation analysis and static linear regression models. The results captured by the correlation analysis and linear regression models illustrates that a positive relationship

Property Index StockIndex 1.5655 4 28 0.2109 Property Index InterbankRate 1.3639 4 28 0.2716

Property Index GDP 5.7203*** 4 28 0.0017

(31)

does exist between stock and house prices in both countries. Additionally, GDP was also shown to have a positive relationship with house prices both of which were apparent from the Cheng & Fung (2009) and Sutton (2002) studies.

Dynamic effects were apparent in the time series analysis and do contribute to changes in house prices. A Granger causality test was used to check for linear interdependencies between the multiple time series. Newey tests were also undertaken to identify which

individual variable demonstrated causality with house prices. The results of the lagged stock price variables in Hong Kong showed a causal relationship with house prices, both

individually and jointly with other variables. In Singapore both lag structures display a joint causal relationship with house prices. However, stock prices individually do not exhibit this causality.

Overall, the results are indicative of changes in stock prices influencing house prices for both countries. Hence, the first two hypothesis in the study can be accepted. This positive relationship reveals limited diversification benefits for investors, although there are benefits that will allow for better speculation and prediction of house price movements. The dynamic effects models have added a new dimension and offer additional insights into examining causality, making it hard to select a particular country for investments. Hence, the third hypothesis can’t be accepted. Finally, the research findings may provide some useful information for government and non-governmental policy-makers in refining housing policies.

7. Limitations of Research

There are some limitations that should be considered when interpreting the results of this study. A small sample of data raises type 2 error issues, one suggestion for a more detailed analysis is to use daily or monthly time series data instead of a quarterly series. This will involve adding more observations and increasing the variability in the data, which in turn could help to improve the accuracy and reliability of results. This study also used the

interbank rate as the interest rate, another alternative could be to use mortgage rates given by commercial organisations such as banks. Mortgage rates are the interest rates charged to home buyers. An alternative for measuring GDP is to use a GDP index, as index prices are adjusted for inflation and standard price measurements.

From the results of the Granger causality model, it was also clearly shown that other determinants such as GDP affect house prices and vice versa. Further analysis into the

(32)

dynamic effects of these interdependencies could facilitate better assessments in changes of house prices and other variables such as the unemployment rate.

(33)

8. References

Abelson,P.,Joyeux,R.,Milunovich,G.and Chung,D.,2005.Explaining house prices in Australia:1970-2003*.The Economic Record.Vol.81:96-103.

Achen, C.H.,2001. Why lagged Dependent variables can suppress the explanatory power of other independent variables.:1-42

Batayneh K.I and Abdullah M Al-Maliki,2015.The Relationship between House Prices and Stock Prices in Saudi Arabia: An Empirical Analysis.International Journal of Economics and Finance.Vol.7 No 2.

Census and Statistics Department- GDP index, 2018 [Online] Available at: https://www.censtatd.gov.hk/hkstat/sub/sc250.jsp [Accessed on 10/06/2018] Centa Property Agency- Property Index,2018 [Online] Available at:

http://www1.centadata.com/cci/cci_e.htm [Accessed on 10/06/2018]

Central Intelligence Agency- The World Factbook,2018 [Online] Available at:

https://www.cia.gov/library/publications/the-world-factbook/ [Accessed on 09/06/2018]

Chau, K.W, Macgregor, B. D. and Schwann,G.M.,2001.Price discovery in the Hong Kong Real Estate Market. Journal of Property Research.Vol.18 issue3:187-216.

Cheng,A.C.S.and Fung,M.K.,2015.Determinants of Hong Kong’s housing Prices. Journal of Economics, Business and Management.Vol.3,Issue 3:352,355.

Chung,K.H.K.,2012. Determinants of Residential Property Prices in Hong Kong: A

Cointegration Analysis. International Research Journal of Finance and Economics Issue 96 :55-62.

Domain,D.,Wolf,R,and Yang H.F.,2015.An assessment of the risk and return of residential real estate. Managerial Finance Vol.41,Issue 6:591-599.

Englund,P.and Ioannides,Y.M.,1997.House Price Dynamics: An international empirical perspective.Journal of Housing Economics.Vol.6,Issue 2:119-136.

Fisher,I.,1930.The Theory of Interest. New York Vol.43.

Harris, J.C.,1989.The Effect of Real Rates of Interest on Housing Prices.The Journal of Real Estate Finance and Economics.Vol.2,Issue 1: 47-60.

Hargreaves,B.,2008.What do rents tell us about House prices?International Journal of Housing Markets and Analysis.Vol.1,Issue1:7-18.

Himmelberg,C.,Mayer,C.,and Sinai,T.2005.Assessing high house prices:Bubbles,

fundamentals and misperceptions. The Journal of Economic Perspectives.Vol.19,Issue 4: 67-92.

(34)

Jiang,L.,Phillips,C.B. and Yu, J.2014. A new hedonic regression for Real Estate prices applied to the Singapore Residential Market.Cowles Foundation Discussion Paper No.1969 Lam,K.C,2016.The Responsiveness of Hong Kong Private Residential Housing Prices. International Journal of Economics and Financial Issues.Vol.6,Issue 1:26-36.

Lean,H.H.,2012.Wealth effect or credit-price effect? Evidence from Malaysia. Procedia Economics and Finance.Vol 1:259-268.

Liow,K.H.,2006.Dynamic relationship between stock and property markets. Journal Applied Financial Economics.Vol.16,Issue 5: 371-376.

Nakajima,M.,2011.Understanding house-price dynamics.Federal Reserve Bank of Philadelphia Business Review Q2: 20-28.

Poterba,J.M.,1984.Tax subsidies to owner-occupied housing: an asset-market approach. The Quarterly Journal of Economics.Vol.99,Issue 4:729-752.

Quan,D.C and Titman,S.,1996. Commercial real estate prices and stock market returns: An International Analysis.

Sim,S.-H.and Chang,B.-K.,2006. Stock and real estate markets in Korea: Wealth or credit-Price Effect.Journal of Economic Research.Vol 11: 99-122.

Sutton,G.D.,2002.Explaining changes in house prices. BIS Quarterly Review part 6:46-55. Yuksel,A.,2016.The relationship between stock and real estate prices in Turkey: Evidence around the global financial crisis. Central Bank Review 16:33-40.

(35)

9. Appendix:

Table 14- Akaike and Bayesian tests Hong Kong

Table 15- Akaike and Bayesian tests Singapore

Lags LL LR Df p-value AIC SBIC

0 306.215 -13.6916 -13.4888

1 344.434 76.438 25 0.000 -14.2924 -13.076

2 403.989 119.11 25 0.000 -15.8631 -13.6329***

3 429.831 51.685 25 0.001 -15.9014 -12.6574

4 462.403 65.144 25 0.000 -16.2456*** -11.9879

Lags LL LR Df p AIC SBIC

0 20 5.543 -8.95748 -8.79689

1 250.998 90.909 16 0.000 -10.266 -9.4636***

2 267.496 32.996 16 0.007 -10.2887 -8.84336

3 280.177 25.364 16 0.064 -10.1412 -8.05352

(36)

Property Index StockIndex 6.0778*** 2 35 0.0054

Property Index InterbankRate 3.6302** 2 35 0.0369

Stock Index Property Index 3.3109*** 2 35 0.0482

Stock Index InterbankRate 0.04929 2 35 0.9520

Stock Index GDP 1.41 2 35 0.2577

Stock Index CPI 4.1904*** 2 35 0.0234

Stock Index All 2.4736 8 35 0.0306

Interbank Rate Property Index 0.2532 2 35 0.7777

Interbank Rate Stock Index 0.28504 2 35 0.7537

Interbank Rate GDP 1.0663 2 35 0.3552

Interbank Rate CPI 0.03071 2 35 0.9698

Interbank Rate All 0.58373 8 35 0.7842

GDP Property Index 0.26347 2 35 0.7699

GDP Stock Index 2.1689 2 35 0.1294

GDP Interbank Rate 0.67048 2 35 0.5179

GDP CPI 0.19323 2 35 0.8252

GDP All 1.5907 8 35 0.1632

CPI Property Index 1.7765 2 35 0.1841

CPI Stock Index 0.44959 2 35 0.6415

CPI Interbank Rate 1.2493 2 35 0.2992

(37)

CPI All 2.5113 8 35 0.0285 *** 1% Significance Level ;** 5% Significance Level;*10% Significance Level

Property Index StockIndex 2.3425* 4 23 0.0850

Property Index InterbankRate 1.1302 4 23 0.3669

Property Index All 1.6467 16 23 0.1340

Stock Index Property Index 2.3843* 4 23 0.0809

Stock Index InterbankRate 0.33806 4 23 0.849

Stock Index GDP 0.48876 4 23 0.7439

Stock Index CPI 1.0575 4 23 0.3998

Stock Index All 1.6028 16 23 0.1471

Interbank Rate Stock Index 0.28512 4 23 0.8846

Interbank Rate CPI 0.32321 4 23 0.8595

GDP Property Index 2.1594 4 23 0.1058

GDP Stock Index 1.4096 4 23 0.2622

GDP CPI 0.3916 4 23 0.8125

GDP All 2.1317* 16 23 0.0476

CPI Property Index 0.31194 4 23 0.8670

CPI Stock Index 0.9377 4 23 0.4598

CPI Interbank Rate 0.66953 4 23 0.6197

CPI GDP 2.8731** 4 23 0.0457

CPI All 1.7523 16 23 0.1069

(38)

Table 19- Granger Results Singapore (1 lag)

Property Index StockIndex 5.1242** 1 43 0.0287

Property Index Interbank Rate 1.033 1 43 0.3151

Property Index GDP 4.8487** 1 43 0.0331

Property Index All 4.2323** 3 43 0.0104

Interbank Rate StockIndex 0.22594 1 43 0.6370

Stock Index Interbank Rate 3.3161 1 43 0.0756

Stock Index Property Index 51.785*** 1 43 0.0000

Stock Index GDP 0.80061 1 43 0.3759

Stock Index All 27.401*** 3 43 0.000

GDP Stock Index 0.73283 1 43 0.3967

GDP Housing Index 1.4801 1 43 0.2304

GDP All 0.86615 3 43 0.4660

(39)

Table 20- Granger Results Singapore (4 lags)

Property Index StockIndex 1.5655 4 28 0.2109

Property Index Interbank Rate 1.3639 4 28 0.2716

Property Index GDP 5.7203*** 4 28 0.0017

Interbank Rate StockIndex 1.5807 4 28 0.2069

Stock Index Interbank Rate 1.7292 4 28 0.1716

Stock Index Property Index 3.7963 4 28 0.0137

Stock Index GDP 1.3184** 4 28 0.2874

Stock Index All 6.8729 12 28 0.0000

GDP Stock Index 0.78992 4 28 0.5416

GDP Housing Index 2.1167 4 28 0.1052

GDP All 0.2531** 12 28 0.0376

(40)