The Impact of MTR Stations on Housing Price in Hong Kong Stephanie Ng Advisor: Prof. Siodla

Transcription

1 The Impact of MTR Stations on Housing Price in Hong Kong Stephanie Ng Advisor: Prof. Siodla Abstract: The MTR subway system in Hong Kong expanded from 55 stations to 93 stations in the past two decades, serving more than 5 million passengers everyday. Due to the high reliance of residents MTR in providing public transportation to the city, it would be interesting to see to what extent do residents in Hong Kong value convenient access to an MTR station. In this paper, I analyze the impact of new MTR stations on housing prices in Yuen Long and Ma On Shan neighbourhoods. Using Rosen s Hedonic Pricing Model and difference-in-difference method, I find that, controlling for year and structural characteristics, neighbourhoods that gain access to MTR stations encounter an increase in housing prices of at least 8 percent. These results suggest that access to public transportation is highly valued in Hong Kong. Keywords: Housing Price, Hong Kong, Hedonic Pricing Model, Infrastructure, Transportation, Subway

2 Acknowledgement I would like to thank my advisor, Professor Jim Siodla for his patient supervision throughout the course of this thesis. I am also grateful to Professor Samara Gunter and Professor Randy Nelson for their guidance and feedback that c Lastly, I would like to thank my friends at Colby College and my family for endless support during this journey.

3 I. Introduction Articles about housing prices in Hong Kong seem never to leave the headlines of major newspapers. Hong Kong s housing market has been a hot topic not only because it catches the attention of the government and overseas investors, but also because it fundamentally impacts the day-to-day lives of average people: our cost of living, standard of living, and even family planning decisions are affected by the housing market. According to data collected by Midland Realty, an average 1000 square-foot apartment in Hong Kong that cost 320,000 USD in 2003 would cost 1.4 million today. 1 The 14 th Annual Demographia International Housing Survey: 2018 conducted by the Performance Urban Planning Organization assesses housing affordability by calculating the median housing price divided by gross pre-tax annual median household income. Result shows that among the 26 severely unaffordable major housing markets, Hong Kong is the least affordable with a Median Multiple of The city has also been the least affordable market since at least Scholarly interest in this topic has produced a number of econometric analyses that explore the internal and external factors of housing prices in Hong Kong. Tse, Ho and Ganesan (1999) study on the supply and demand of housing in Hong Kong. They specifically looked into the effects of population growth, transaction volume, and the inflation rate on housing prices in Hong Kong. As an elaboration on the pre-existing findings, my paper focuses on one of the key factors that drive housing prices in Hong Kong: transportation. Specifically, I look at how the expansion and development of the Mass Transit Railway ( MTR ) has affected housing prices in Hong Kong. 1 Source from Midland Realty, Hong Kong 2 MTR Corporation Annual Report 2017, page 2

4 MTR, which serves as one of the most important modes of transportation for commuters in Hong Kong, expanded from 55 stations to 93 stations across Hong Kong since Since the completion of three more stations in December 2016, the MTR now spreads across all eighteen districts in Hong Kong. The MTR is also renowned for its efficiency and cleanliness. During the morning peak hours, 8-car trains with a capacity of 2,500 passengers each run at 2-minute intervals, carrying around 75,000 passengers per hour per direction on the Tsuen Wan Line. Because the trains run at frequent intervals for 19 hours per day, MTR Corporation takes a 48.4% market share of Public Transport Market. 2 With such rapid development in MTR over the past two decades, it would be interesting to see whether or not MTR plays an important role in determining housing prices in Hong Kong. Specifically, I look at whether or not there s a housing price appreciation in the neighbourhood due to an opening of a MTR station in the neighbourhood. My hypothesis is that in an event where an MTR station is introduced to a neighbourhood, housing price has to increase to offset the lower opportunity cost that the neighbourhood enjoys compared to other distant locations. Second, if there is an appreciation, I am interested to see the trend of such appreciation, whether housing prices slowly adjust to the better access and convenience of having an MTR station, or immediately capitalize upon the opening of an MTR station. II. Literature Review A number of scholars that have explored the relationship between housing prices and transportation. The well-established model of Alosno (1964), Muth (1969) and Mills (1967) gives a strong foundation for understanding the fundamental forces 2 MTR Corporation Annual Report 2017, page 2

5 that explain the overall urban structure. The model assumes a city with a fixed population and a given income level living around the Central Business Districts ( CBD ). It also relies on the condition that consumers must be equally well off at all locations, achieving the same utility regardless of where they live in the city. When commuting cost increases, the disposable income of household decreases. The only way to keep consumers at all locations equally well of is to decrease the housing price per floor space as distance increases. In other words, a lower housing price compensates for the disadvantage of higher commuting costs at farther locations. The model also incorporates the time cost component of the commuting cost, directly drawing the tradeoff relationship between commuting costs and housing prices. In contrast to looking at overall housing demand and city structures (where housing is homogenous), the Hedonic Pricing Model by Rosen (1972) accounts for the specific attributes of the apartments. The function can be estimated by using data that are available to homebuyers, such as square footage, number of bedrooms, and floor level. Researchers have used the Hedonic Pricing Model in estimating the value of certain characteristics of housing. Ball (1973) and Dewees (1976) looks into the importance of accessibility attributes while Anderson and Crocker (1971) look into the importance of externalities and neighbourhood characteristics. Mok, Chan and Chow (1995) use the hedonic price model to look at private properties in Hong Kong and they found that the elasticities of housing attributes obtained from the Box-Cos analysis indicate that the valuation of a property is sensitive to changes in housing traits. Bajic (1983), more specifically, looked into the relationship between transportation and housing prices, and found that the residential values near a rail line were $2,327 higher than elsewhere.

6 As we have witnessed massive transportation improvements in the past two decades in Asia, researchers have paid more attention to the effect of transportation on housing prices in Asia. Sun, Zheng and Han (2013) study on the effect of subway lines on housing prices in Chengdu and find that housing prices are 7% to 14% higher within 1.5km around the subway station than outside the stations. Diao, Leonard and Ling (2016) also did similar research on how the opening of the new Circle Line affected housing prices in Singapore, and concluded that it increases housing prices by 8.6% in the treated zone relative to houses in control zones. Though many scholars have looked into the effects of transportation on housing prices, research on the relationship between transportation and Hong Kong housing prices is fairly limited. Ho, Tse, and Ganesan (1997) look into the influence of transportation on housing price. Yiu and Wong s (2005) paper is the most recent publication that explores the relationship between transportation and housing prices in Hong Kong. Specifically, they investigated the effect of a newly built tunnel on housing prices in Hong Kong and found that there were positive price expectation effects well before the completion of the tunnel, indicating a positive effect on housing prices. However, despite the rapid development of the MTR transportation system and the surge in housing prices experienced in the city, there is not any recent research exploring the impact of new MTR stations on housing prices in Hong Kong. This study attempts to estimate the impact of newly opened MTR stations on local housing prices.

7 III. Methodology and Data This paper uses Rosen (1974) s Hedonic Pricing Model ( HPM ). The basic premise of HPM is that price of the good, housing in this case, is determined by both internal characteristics of the good and external factors. An improvement in the characteristics may increase the value of the good, meaning when transportation improves, the value of the property increases. In this study, most data will be extracted from Centadata, a real estate information system that contains comprehensive information on past property transactions in Hong Kong. I collect transaction date, transaction price, rentable square footage, useable square footage, apartment name, age of the building, floor level and flat information from the Centadata online platform. Figure 1. CBD, Treatment Zone and Control Zone location in Hong Kong Source: Planning Data from Town Planning Board Survey Base Map from Lands Department In this study, Yuen Long is chosen as the treatment zone and the Gold Coast area is chosen as the control zone. Both the treatment zone and control zone are located on the Western side of the New Territories area, and are of similar distances

8 from the CBD ( Commercial Building Districts ) in Kowloon and Hong Kong Island. Central is the CBD on the Hong Kong Island and Tsim Sha Tsui is the CBD in Kowloon. Yuen Long (treatment zone) and Gold Coast (control zone) is 36.2 km away from Central on the Hong Kong Island and 31.6 km away from Tsim Sha Tsui in Kowloon respectively. Gold Coast is 30.2 km away from Central on the Hong Kong Island, and 26.6 km away from Tsim Sha Tsui in Kowloon. On December 20 th 2003, the Yuen Long MTR station was opened together with 9 other stations on the West Rail line. The West Rail line, which has a total travel distance of 35.7 km, connects many of the neighbourhoods in the Western part of the New Territories and the Hung Hom station in Kowloon. Because the line does not pass by any neighbourhood near the Gold Coast area in the Western part of the New Territories (refer to figure 1), the Gold Coast area is chosen as the control zone. The Gold Coast area is also chosen because of its similar distance to CBD and similar neighbourhood characteristics to Yuen Long. Table 1. Summary of observations in Yuen Long and Gold Coast district Yuen Long Gold Coast Total (Treatment zone) (Control zone) Prior to 20 th Dec After 20 th Dec Total To narrow the confounding factors that might influence housing prices in these neighbourhoods, I limit the transactions that occur before and after the MTR expansion to within two years. Table 1 shows from 20 th Dec 2001 to 20 th Dec 2003, there are 169 observations in total, 97 from Yuen Long and 72 from Gold Coast. For the post period, which is from 21 st Dec 2003 to 21 st Dec 2005, there are 191 observations in total, 122 from Yuen Long and 69 from Gold Coast. The data includes

9 360 observations in total from Yuen Long and Gold Coast in the Western side of the New Territories. For the treatment zone, Yuen Long, I collect all transaction data from Sun Yuen Long Centre. The property, which is situated next to the Yuen Long MTR station, has a total of five buildings ( blocks ). The data contain around 30 to 50 observations per block in the Sun Yuen Long Centre. For the control zone, Gold Coast area, I collect all transaction data from Hong Kong Gold Coast. The property has a total of 20 blocks and I collect data from 8 of the blocks, with each block providing around 10 to 30 observations. Table 2 shows summary statistics for Yuen Long and Gold Coast area. The rentable square footage area is slightly larger in Gold Coast, but actual useable square footage area in the two neighbourhoods is very similar. The building age of both properties is also very close, as both were built in the 1990s. Table 2. Summary Statistics for Yuen Long & Gold Coast MTR expansion to Yuen Long Yuen Long (Treatment zone) Gold Coast (Control zone) Total Mean SD Mean SD Mean SD Housing Price (USD thousands) Housing Price (Log form) Average floor level Rentable Square Footage Useable Square Footage Building Age Garden View 92% % % Sea View 0% % % # of bedrooms Observations N=191 N=169 N=360 The average floor levels for the observations are th in the treatment zone and th floor in the control zone. They have similar average floor level because in both areas, buildings have around 27 to 29 floors in total. The biggest difference

10 between the two properties is that Gold Coast is by the sea and hence, many of the apartments have a sea view, whereas in Yuen Long, most of the apartments have a garden view. To check whether or not some neighbourhoods may encounter more or less impact of MTR stations on housing prices, I also select a treatment zone and control zone in the eastern part of the New Territories. The treatment zone that I look at is Ma On Shan ( MOS ) station in Shatin district. The MOS line, which consists of 9 stations, was opened on Dec 21 st Sai Kung is chosen as the control zone because it has the most comparable geographical location as Ma On Shan, and the MTR station does not reach that area. Table 3. Summary of observations in MOS and Sai Kung district MOS Sai Kung Total (treatment zone) (control zone) Prior to 21 st Dec After 21 st Dec Total Table 3 shows that there are in total 248 observations for studying the effect of MTR expansion to Ma On Shan (MOS) location, where 144 observations are collected from the treatment zone, MOS, and 104 are collected from the control zone, Sai Kung. Of the 144 observations collected from MOS, 66 transactions were made prior to 21 st Dec 2004 and 78 were after 21 st Dec Of the 104 observations collected from Sai Kung, 45 transactions were made prior to 21 st Dec 2004 and 59 were after 21 st Dec Table 4 shows that the structural features slightly differ between MOS and Sai Kung. Sai Kung apartments have a relatively smaller square footage, are older, and consequently less expensive. Though both are located near the coastline, Ma On Shan also has more apartments that enjoy a sea view than in Sai Kung. That is because

11 Table 4. Summary Statistics for MOS and Sai Kung MTR expansion to MOS MOS (Treatment zone) Sai Kung (Control zone) Total Mean SD Mean SD Mean SD Housing Price (USD thousands) Housing Price (Log form) Average floor level Rentable Square Footage Useable Square Footage Building Age Garden View 34% % % Sea View 38% % % # of bedrooms Observations N=148 N=104 N=248 apartments are relatively taller in Ma On Shan, and developers took advantage of the view and have designed more flats to face the sea (see Appendix 3). Though Sai Kung is near the coast, apartments are centered near the town center rather than surrounding the sea. However, by using difference-in-difference method, I can account for the differences in these features. IV. Estimating equation and Identification Strategy The standard hedonic pricing model predicts that housing prices are determined by internal and external characteristics of the housing. Scholars have conducted studies on different characteristics that contribute to the housing price of a specific apartment or house. In the equation below, β! (k = 1,, k) shows the marginal change in the unit price of the kth characteristic x! of the apartment, where P! is housing price.! P! = α! +!!! β! X!" + μ! (1)

12 Common characteristics in a hedonic model include accessibility features, structural features and neighbourhood features. Accessibility variables may include the number of shopping malls in the district, distance to key shopping malls and distance to the CBD. Neighbourhood variables can include quality of the schools in the district and population density. Structural variables include actual floor area, number of rooms, floor level and age of the apartment. Sirmans et al. (2005) suggest that age of the building and square footage show up the most frequently in hedonic models, and in my paper, I will take these features into account. In this study, I test for changes in housing prices that correspond to the timing of the MTR expansion by estimating the following difference-in-difference equation. The equation also incorporates elements from the hedonic price model. The first specification is:! log (P! ) = α + β! Treatment! + β! Postperiod! + β! Treatment! Postperiod! + γ! log S!! +!!! γ! S!". +ε! (2) Where P! refers to the transaction price, Treatment is a dummy variable that takes a value of one if the apartment is located in the treatment zone, where it is less than 1 km away from the MTR station. Because residents living within 1 km of the MTR station will have a walking distance of less than 10 minutes to the station, they will most likely capitalize the amenity into housing values. For residents living beyond 1 km from the station, they may be partially benefited from it because walking distance to the station may be too high that it may be better off for residents to wait for alternative transportation such as buses or mini buses. In this study, we will be focusing on the apartments that likely have direct benefits from the opening of a MTR station. Postperiod is a dummy variable that takes a value of one if the transaction is made after 20 th December, 2003 (after the opening of the Yuen long MTR station). I am primarily interested in the coefficient estimate of β!, which

13 describes the change in housing price in Yuen Long due to the opening of the Yuen Long MTR station. Similar to the standard hedonic pricing model, there are structural effects and building effects that contribute to the housing price. S! (k = 1,, k) specifically refers to the kth structural characteristic S! of the apartment. In this paper, I take into account five structural characteristics: building age, useable square footage, floor level, number of bedrooms, and views of the apartment in estimating the housing prices. I also take the log for usable square footage S!, so that for one percent increase in usable square footage, it results in γ! percent change in housing price, holding all other factors fixed. Year is the year of sale. On a block level, Sun Yuen Long Centre (block 1) in Yuen Long share very similar amenities with Sun Yuen Long Centre (block 2). They share the same security system, as well as the same access to the swimming pools and recreational facilities. Therefore, it is not necessary to control on a block level. The only differences may be the age of the building, which is already captured as one of the structural characteristic variables. The second specification is:! log (P! ) = α + β! Treatment! + β! Year! + β! Treatment! Year! + γ! log S!! +!!! γ! S!". +θf. e. +ε! (3) In equation (2) I break down the percentage change in housing price by year. Year dummy variables takes the value of 1 if the transaction is made within that specific year. Treatment variable still takes the value of 1 if the apartment is located in the treatment zone. By separating transactions by years, we could observe whether or not there are differences in the percentage change in housing price. There are a few identifying assumptions in this model. First, it assumes that over the 2-year pre-opening period housing prices in the treatment zone and control zone follow similar trends. It also assumes that there are no significant changes in the availability and quality of public goods in the two areas that would impact

14 households decisions in buying or selling their apartments. The model omits the impact of being furnished or damaged on housing prices. Because our main treatment and control areas are both in the Western part of the New Territories, and are similar neighbourhoods, it is not likely that the impact of furnished home or damaged home would cause a positive or negative bias. V. Results V.I Impact in Western Part of the New Territories Figure 4 shows the trend in housing price per square foot in USD from 21 st December 2001 to 21 st December 2005 for Yuen Long and Gold Coast. The longdash line represents the date of the opening of the Yuen Long MTR station. Though the figures simply display the changes in housing prices per square foot without controlling for other factors, we can see that generally Yuen Long has a higher housing price per square foot after Figure 2. Housing prices per Square Foot in Yuen Long and Gold Coast from 2002 to 2005 Housing price per s.f (USD) /1/2002 1/1/2003 1/1/2004 1/1/2005 1/1/2006 Year Yuen Long (Treatment) Gold Coast (Control) Source: Centadata, transactions made between 21 st Dec 2001 and 21 st Dec Notes: Transactions in Yuen Long location are drawn from Sun Yuen Long Centre. Transactions in Gold Coast are drawn from Hong Kong Gold Coast. Vertical long-dashed line: Opening date of the Yuen Long MTR Station; Vertical short-dashed line: Announcement date

15 The short-dashed line represents the date, 19 th May 2003, when news media announced the expected opening date of the West Rail Line. There is not a large difference between the two neighbourhoods during the window from 19 th May 2003 to 20 th Dec 2003, suggesting that residents might not have perceived the potential benefit of the Western Rail Line upon the announcement date. However, the housing prices per square footage between Yuen Long and Gold Coast widened after Dec 20 th This gap seems to suggest that the benefit of the access to the new MTR station was capitalized into housing prices. Table 5: The Impact of the Yuen Long MTR Station on Housing Prices (1) (2) (3) (4) VARIABLES Model 1 Model 1 Model 2 Model 2 YL * ** *** (0.0427) (0.0388) (0.0537) (0.0369) postperiod *** *** (0.0394) (0.0234) YL*postperiod * *** (0.0549) (0.0326) Year *** *** (0.0571) (0.0291) Year (0.0506) (0.0254) Year ** 0.176*** (0.0504) (0.0254) YL*Year (0.0810) (0.0410) YL*Year * 0.118*** (0.0724) (0.0364) YL*Year * ** (0.0682) (0.0344) Floor ** *** ( ) ( ) Building Age *** *** (0.012) ( ) Log (square footage) *** 0.918*** (0.0661) (0.0538) No. of bedroom * (0.0263) (0.0214) Garden View (0.0373) (0.0303) Sea View * * (0.0433) (0.0354) Constant 2.855*** *** (0.0379) (0.473) (0.0370) (0.416) Observations R-squared Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

16 Table 5 shows the results for Yuen Long MTR station. The specification in Column (1) does not control for the structural characteristics of the apartment, and shows that Yuen Long experienced a 7.2 percent increase in housing prices compared to Gold Coast after the Yuen Long MTR station was introduced. Column (2) shows that after controlling for structural characteristics of the apartment, which include usable square footage, floor level, building age, number of bedrooms, and views of the apartment, Yuen Long had an additional 10.8 percent increase in their housing prices compared to Gold Coast. This effect is statistically significant at the 1% level and is more precisely measured after controlling for a variety of characteristics. Columns (3) and (4) break down the variable, YL postperiod into YL Year2004 and YL Year2005, allowing us to observe the impact of the Yuen Long MTR station in those two years. Without controlling for the structural characteristics, housing prices in Yuen Long encountered a 12.1 percent increase in 2004 and 12.3 percent increase in 2005 relative to those in Gold Coast. After controlling for the structural characteristics, the opening of the Yuen Long MTR station is associated with an 11.8 percent increase in housing prices in 2004 and an 8.41 percent increase in 2005 compared to Gold Coast. The coefficients also become statistically significant. In all four models, the coefficient for YL Year2003 is not statistically significant, meaning the difference between Yuen Long and Gold Coast housing prices had kept the same until 2004, soon after the Yuen Long MTR station was opened. This supports the identifying assumption of the model that over the 2- year pre-opening period housing prices in the treatment zone and control zone follow similar trends.

17 V.II Impact in the Eastern Part of the New Territories To understand if the percentage change in housing price due to the opening of MTR station is exclusive to the location in Yuen Long, I also include another treatment zone and control zone in the Eastern Part of the New Territories in this study. Figure 3 shows a gap in housing prices in the two neighbourhood, with Ma On Shan (treatment zone) having a higher housing price per square footage than in Sai Kung (control zone). The long-dashed line on Dec 21 st 2004 represents the opening of the Ma On Shan MTR station while the short-dash line on Jan 6 th 2004 represents the news release of the expected opening of the station. Compared to the opening of the Yuen Long station, homebuyers in Ma On Shan appear to capitalize the expected benefit of the MTR station before the opening at the end of December Within the window from the announcement date and the opening of the station, housing price per square footage in Ma On Shan location surged up while the housing price per square footage increased steadily over the period in Sai Kung. Nevertheless, prior to 2004, the neighbourhoods showed similar trends in housing prices. Figure 3. Housing prices per Square Footage in Ma On Shan and Sai Kung from 2003 to 2006 Housing Price per s.f (USD) /1/2003 1/1/2004 1/1/2005 1/1/2006 1/1/2007 Year Ma On Shan (Treatment) Sai Kung (Control) Source: Centadata, transactions made between 21 st Dec 2002 and 21 st Dec Notes: Transactions in MOS location are drawn from SunShine City and Bayshore Towers. Transactions in Sai Kung are drawn from Sai Kung Garden, Sai Kung Tower and Kam Po Court. Long dashed line: Opening Date of the Ma On Shan MTR station; Short dashed line: Announcement Date

18 The housing price trend due to the opening of the Ma On Shan MTR station is similar to the result in the Yuen Long MTR station, where we see a bigger effect within a year after the opening of the MTR station. Table 6: The Impact of the Ma On Shan MTR Station on Housing Prices (1) (2) (3) (4) VARIABLES Model 3 Model 3 Model 4 Model 4 MOS 0.600*** 0.329*** 0.521*** 0.226*** (0.0566) (0.0339) (0.0885) (0.0362) postperiod 0.135** 0.167*** (0.0580) (0.0253) MOS*postperiod ** (0.075) (0.0329) Year *** (0.0892) (0.0322) Year ** 0.221*** (0.0867) (0.0316) Year * 0.252*** (0.0873) (0.0315) MOS*Year *** (0.115) (0.0414) MOS*Year *** (0.113) (0.0410) MOS*Year *** (0.114) (0.0412) Floor *** *** (0.0014) ( ) Building Age ** *** (0.0032) ( ) Log (square footage) 1.117*** 1.115*** (0.0734) (0.0608) No. of bedrooms * (0.0354) (0.0295) Garden view ** *** (0.0193) (0.0160) Sea view *** 0.129*** (0.0251) (0.0210) Constant 4.768*** *** 4.736*** *** (0.0437) (0.3989) (0.0686) (0.331) Observations R-squared Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Table 6 shows the result for the effect of the Ma On Shan MTR station on housing prices. After controlling for structural characteristics, the variable MOS Postperiod in column (2) shows that Ma On Shan had an additional 6.1 percent increase in housing price compared to that of Sai Kung after the opening of the MTR station. In

19 variable MOS Year2005 and MOS Year2006, we also see that the first year of post-opening of the MTR station shows a higher appreciation in housing prices. In 2005, there was an additional 18 percent increase if the apartment is located in Ma On Shan instead of Sai Kung, and an additional 12.8 percent increase in While apartments face similar trend in housing prices after the opening of the MTR Station, the perceived benefit of Ma On Shan MTR station seems to be larger in Ma On Shan than in Yuen Long. In 2004, before the opening of the station, apartments in Ma On Shan experienced an additional 16.4 percent increase in housing prices relative to Sai Kung as seen in Variable MOS Year2004 in Column (4). The positive and statistically significant coefficients of the Year variables also suggest that housing prices were on a rise from 2003 to The earlier announcement date for the opening of the station may explain the price appreciation in There may also be a positive upward bias in the effect of the MTR station in Ma On Shan due to the renovations of the shopping malls that took place in Ma On Shan. This may explain why the Ma On Shan Station faced a higher housing appreciation than in Yuen Long Station. Nevertheless, the results from Ma On Shan station are consistent with the results we found in Yuen Long, where apartments see an additional 10 to 13 percent increase in housing prices when an MTR station is established nearby. V.III Additional Robustness Check Another concern for the model may be that quarterly effects may impact the changes in housing prices. Because weather changes are not extreme in Hong Kong, it is not likely to have a strong impact on transaction prices and volume. Therefore, I also check on whether or not controlling on a quarterly basis would impact my results for Yuen Long and Gold Coast. Housing prices in Yuen Long increased by an additional

20 11.8 percent in 2004 and 8.3 percent in 2005 after the opening of the new Yuen Long station when I control on a quarterly basis (Appendix 1). Housing prices in Ma On Shan increased by an additional 18.6 percent in 2005 and 13.0 percent in 2006 after the opening of the new Ma On Shan Station when I control on a quarterly basis (Appendix 2). VI. Conclusion The analyses in this paper deal with two specific issues. The first is whether or not access to an MTR station is capitalized into housing values. Second is whether housing prices slowly adjust to the added convenience of having an MTR station, or immediately capitalize upon the opening of an MTR station, causing a significant jump in housing price in the affected neighbourhood. In both treatment zones, housing prices increase by 10 to 13 percent after an opening of the MTR station. Both eastern and western part locations in the New Territories of Hong Kong experienced a jump in housing prices in the first year after the MTR was opened, indicating that households immediately realized the benefits and convenience of better access to other areas of the city, including the CBD. Compared to the opening of the Yuen Long MTR station, the opening of the Ma on Shan station caused an even faster reactionary increase in housing prices. The data also demonstrates that residents in Ma On Shan faced a higher overall increase in housing prices than the Yuen Long station. More studies are required to examine the possible explanations for the differences in housing price increases. One possible explanation may be the relative pricing of other modes of transportation in these neighbourhoods. It may play a role in determining how much homebuyers value the apartments in these locations.

21 Nevertheless, the results suggest that access to public transportation is highly valued by homebuyers.

22 References Alonso, William. Location and Land Use. 1964, doi: /harvard Brueckner, Jan K. Lectures on Urban Economics. MIT Press, Diao, Mi, et al. Spatial Difference-in Differences (SDID Models for Impact of New Mass Rapid Transit Line on Private Housing Values. SSRN Electronic Journal, 2016, doi: /ssrn Estimating the Drivers of Hong Kong s Housing Price Short-Run Dynamics. Hong Kong Monetary Authority, 3 Apr Hong Kong Monetary Authority, Mills, E.s. An Aggregative Model of Resource Allocation in a Metropolitan Area. American Economic Review, 57, 1967, pp Mok, H.m, et al. A Hedonic Price Model for Private Properties in Hong Kong. Journal of Real Estate Finance and Economics, 10, 1995, pp Muth, John F. Rational Expectations and the Theory of Price Movements. Econometrica, vol. 29, no. 3, 1961, pp , doi: / Rosen, Sherwin. Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition. Journal of Political Economy, vol. 82, no.1, 1974, pp , doi: / Sun, Cong, et al. The Effects of Subway Construction on Housing premium: A micro-data Analysis in Chengdu s Housing Market. Proceedings of the 17 th International Symposium on Advancement of Constrcution Management and Real Estate, Aug. 2013, pp , doi: / _46 Tse, Raymond Y.c., et al. Matching housing supply and demand: an empirical study of Hong Kongs market. Construction Management and Economics, vol. 17, no. 5, 1999, pp , doi: / Yiu, C.y. and S.k. Wong. The effects of Expected Transport Improvements on Housing Prices. Urban Studies, vol. 42, no.1, 2005, pp , doi: /

23 Appendix 1. Results for Yuen Long and Gold Coast area (with quarterly fixed effect) VARIABLES Log(HpriceUSD) YL *** (0.0368) Year *** (0.0297) Year (0.0255) Year *** (0.0255) YL* Year (0.0409) YL* Year *** (0.0362) YL* Year ** (0.0343) Floor *** ( ) Building Age *** ( ) Log (square footage) 0.907*** (0.0539) room (0.0213) GardenView (0.0303) SeaView * (0.0353) Q (0.0175) Q (0.0194) Q ** (0.0195) Constant 1.736*** (0.416) Observations 360 R-squared Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

24 Appendix 2. Results for Ma On Shan and Sai Kung (with quarterly fixed effect) VARIABLES Log(HpriceUSD) MOS 0.230*** (0.0363) Year *** (0.0323) Year *** (0.0317) Year *** (0.0314) MOS* Year *** (0.0418) MOS* Year *** (0.0412) MOS* Year *** (0.0412) Floor *** ( ) Building Age *** ( ) Log (square footage) 1.103*** (0.0623) # of bedroom (0.0301) Garden view ** (0.0162) Sea view 0.127*** (0.0210) Q (0.0196) Q (0.0196) Q * (0.0201) Constant *** (0.339) Observations 247 R-squared Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

25 Appendix 3. i. Floor plan of Sun Yuen Long Centre ii. Floor plan of Gold Coast Area

26 iii. Apartments at Ma On Shan location iv. Apartments at Sai Kung location