THE IMPACT OF LUXURY HOUSING ON NEIGHBORHOOD HOUSING PRICES IN TAIPEI CITY

Chun-Chang Lee¹⁺ --- Chih-Min Liang² --- Cheng-Huang Tung³ --- Yu-Jian Lu⁴

¹Professor, Department of Real Estate Management, National Pingtung University, Taiwan No. 51, Mingsheng East Road, Pingtung, Taiwan

²Associate professor, Department of Public Finance and Tax Administration, National Taipei University of Business, Taiwan

³Associate professor, Department of Computer Science and Information Engineering, National Pingtung University, Taiwan

⁴Department of Real Estate Management, National Pingtung University, Taiwan No. 51, Mingsheng East Road, Pingtung, Taiwan

Keywords:Luxury housing Difference-in-difference method Housing price Spillover effects Geographic information system Taipei City.

ARTICLE HISTORY: Received:9 May 2018. Revised:3 August 2018. Accepted:7 September 2018. Published:30 October 2018.

Contribution/ Originality:This suggests that luxury housing has spillover effects on neighborhood housing prices and, specifically, can help to increase neighborhood housing prices. Furthermore, the results showed that housing prices in the neighborhood in which the luxury housing was to be built were inversely proportional to the distance from the luxury housing.

1. INTRODUCTION

In recent years, Taiwan’s real estate market has been dominated by high-priced housing products, which are advertised as being built with the highest quality building materials and equipment in the best locations, with large floor areas, good planning and strong management. Moreover, due to the intense competition of the housing sales market, numerous luxury housing development projects are currently in progress. Thus, the effects of such luxury housing have become a focus of concern. With regard to high-priced housing, the Taipei City Revenue Service defines a house with a total transaction price above 80 million NTD as a “senior house,” which would more commonly be described as luxury housing. According to Lin and Jou (2005 ) luxury housing refers to houses with exceptional landscapes, locations and other rare space-related resources, or houses with a high total price, high floor area, high quality facilities and high profile. Almatarneh (2013 ) suggested that real estate developers distinguish between luxury housing and general housing through their advertisements of various housing characteristics, such as design, planning, and architectural concepts, thereby further strengthening the position of luxury housing in the minds of the public. The unique and irreplaceable features of luxury housing units have a very important impact on the real estate market.

The choice of a good living environment depends not only on the characteristics of the home itself, but also on the characteristics of the given neighborhood residential environment. DiPasquale and Wheaton (1996 ) indicated that when the appearance of a residential area becomes more beautiful, the quality of life in that neighborhood will become more attractive, resulting in positive spillover effects on neighborhood housing prices. Bourassa et al. (2004 ) proposed that attractive buildings increase the value of the real estate in the surrounding neighborhoods, and that attractive luxury housing may also attract more residents, including those with high incomes, as well as members of the political and business elite with high socioeconomic statuses. Moreover, people may select their places of residence to maintain their images by getting close to people with high incomes and social statuses (Cialdini et al., 1976 ). Ooi and Le (2013 ) argued that both newly developed buildings and luxury housing may have positive external effects or spillover effects on neighborhood housing. The architectural style of luxury housing is generally different from that of general buildings. In addition to their innovative and stylish appearances, luxury housing units may create price spillover effects through the beautification of the neighborhood environment and high sales prices, thereby affecting the average neighborhood housing price.

This study aims to investigate whether luxury housing has such spillover effects on neighborhood housing prices. It adopts the difference-in-difference method (DD) to analyze the impact of luxury housing on neighborhood housing prices before and after the obtainment of a construction license. This method can accurately estimate the impact of a single event on two groups. After eliminating the common growth trend of the treatment group and the control group, as well as other unobserved influential factors, the real impact of luxury housing buildings on neighborhood housing prices can be more effectively estimated.

The purposes of this study are: (1) to discuss the differences in housing prices between a neighborhood in which luxury housing is to be built and another, non-luxury neighborhood before the announcement of luxury housing construction, as well as the differences in neighborhood housing prices after the announcement of luxury housing construction; that is, to determine whether luxury housing spillover effects have a greater impact on the neighborhood housing prices in which the luxury housing is actually built than on the prices in other neighborhoods; and (2) to determine whether, after a decision to construct luxury housing has been made, there are different impacts on neighborhood housing prices based on the distance from the luxury housing. If the impact of the luxury housing on neighborhood housing price decreases with increasing distance from the luxury housing, it means that the luxury housing spillover effects will decrease over increasing distances from the luxury housing.

2. LITERATURE REVIEW

The impact of luxury housing on neighborhood housing prices can be explained by spillover effects and amenity effects. The quality of a living environment depends not only on the characteristics of a home itself, but also on the characteristics of the given neighborhood residential environment. Boyle and Kiel (2001 )  proposed that environmental externalities such as a neighborhood’s air quality, water quality, and unpopular land use are likely to affect neighborhood housing prices. Des Rosiers et al. (2002 ) found that beautification of landscaped curbs will increase neighborhood real estate prices by about 4.4%. Bourassa et al. (2004 ) argued that attractive buildings can increase neighborhood real estate prices by about 37% and that the unique and attractive improvements in a neighborhood can increase neighborhood real estate prices by about 27%. In contrast, the prices of housing in neighborhoods of poorer environmental quality can be reduced by 51%.

The prices of houses in a neighborhood with luxury housing are affected by that luxury housing because people are attracted by things of value higher than their intrinsic value and seek associate themselves with them. (Cialdini et al., 1976 ) called this effect the basking in reflected glory (BIRGing) effect, arguing that people associate with successful people to create a positive self-image and better evaluations from others. According to Tesser (1988 )  when people become closer to outstanding or superior people, they feel honored with improved self-evaluations. Similarly, luxury housing will also attract those with high socio-economic status to live in a given neighborhood 5 neighborhood effect (DiPasquale and Wheaton, 1996 ; O’Sullivan, 2000 ).

According to a review of the literature on empirical estimation methods, in the past, the hedonic price method was generally used for estimating the impact of improved neighborhood environment quality on housing prices. However, in recent years, where the occurrence of a specific event has improved the quality of a specific neighborhood environment to affect the neighborhood housing price, the estimation of cause and effect has typically been accomplished by using the difference-in-difference method. The advantage of the method is that it can review the impact of the announcement of a specific policy or the construction of a specific building on the neighborhood.

Kiel and McClain (1995 ) explored the impact of the establishment of waste incineration plants on the price of the housing in the North Andover area by setting the announcement times for the establishment of the incineration plants from 1979 to 1980 and comparing the housing prices as of 1978 and 1981. Residential housing outside of the perimeter of 3 miles from the incineration plants was the control group, while houses within the perimeter of 3 miles from the incineration plants were regarded as the treatment group. The research findings suggested that the difference-in-difference price was 11,863.9 USD.

Kavetsos (2012 ) analyzed the effect of the announcement of hosting the Olympic Games on London housing prices by using five regions hosting Olympic events as the treatment group and the rest of the regions of London as the control group. The results, estimated by using the difference-in-difference method, indicated that the housing prices in the five administrative regions (Greenwich, Hackney, Newham, Tower Hamlets, and Waltham Forest) increased by 2.1% to 3.3% after the announcement.

Dehring et al. (2007 ) explored the impact of the announcement that a new stadium would be constructed on the local housing prices in Dallas-Fort Worth. The sample data included the residential housing sales data for Dallas and Arlington from January 2004 to March 2005. The difference-in-difference method was adopted to analyze the impact of the announcement of building a new stadium on local housing prices. The research findings suggested that the average housing price decreased by 1.5% after the announcement that a new stadiums would be built as the tax burden on the local community was increased to subsidize the construction. Therefore, the local housing prices dropped in the short term.

Dubé et al. (2011 ) applied the difference-in-difference method to estimate of the impact of rapid bus transit service in Quebec on residential housing prices, finding that the residential housing prices did not change significantly. Heintzelman (2010 ) employed the difference-in-difference method in the analysis of CPA (community preservation act) implementation in communities across Massachusetts on local real estate prices by using communities implementing CPA as the treatment group and communities not implementing CPA as the control group. According to the estimation results, the local real estate prices decreased by about 1.5% after the implementation of CPA. Compared to many previous research methods such as the hedonic price method, the difference-in-difference method can eliminate the possible interference of unobserved factors on housing prices to avoid estimation bias, and thus demonstrate the real impact on neighborhood housing prices.

3. EMPIRICAL MODEL

3.1. Empirical Model Settings

The difference-in-difference method is a research method now widely used in the field of micro econometrics. The method assumes that the treatment group and the control group are the groups that will be compared both before and after a policy change, where the treatment group is the group affected by the policy change and the control group is a group not affected by the policy change. The purpose of using the method is to evaluate the impact of the policy. Through time delineation, that is, by examining certain periods before and after the occurrence of the specific event or policy announcement, as well as by comparing the differences between the control group and the treatment group. An analysis that uses the difference-in-difference method can effectively control for any systematic differences caused by potential selection bias (Wooldridge, 2009 ).

The estimation of luxury housing spillover effects can be made by comparing the differences in housing prices of a control group and a treatment group after the occurrence of a specific event, namely, the obtainment of a construction license for the luxury housing. In general, real estate developers can legally sell housing products after obtaining the relevant construction license, and can thus pre-sell the housing products before the completion of construction. Such a pre-sales method is very common in Asian countries and regions. Lai et al. (2004 ) and Munneke et al. (2011 ) used the time of obtaining the construction license as the delineation benchmark. This study also used the date on which the official construction license for the luxury housing in question was obtained as the time at which the decision to construct the luxury housing was made. With regards to studies classifying the treatment group and the control group by distance, Liang (2009 ) and Dempsey and Plantinga (2013 ) classified the groups according to a radius of 500 meters; Voicu and Been (2008 ) compared the housing prices in the research rea of a radius between 500 feet and 1000 feet (about 150~300 meters). This study classified houses located within the radius of 500 meters from the luxury housing as the treatment group and those outside the radius of 500 meters from the luxury housing as the control group in order to discuss the differences in the impact on housing prices in terms of the distance from the luxury housing. The settings of the empirical model used are shown as follows in Eq. (1):

In order to analyze whether the housing prices of sub-samples within the treatment group were different due to their different distances from the luxury housing, samples within the treatment group were divided in terms of their distances from the luxury housing as follows: distances within 100 meters, distances of 100 meters to 300 meters, distances of 300 meters to 500 meters.

3.2 Variable Selection Descriptions

Ihlanfeldt (2007 ), Matthews and Turnbull (2007 ) and Andersson et al. (2010 ) divided the residential housing attributes into basic attributes, including the hedonic price variables in the unit of household or building, and housing neighborhood environmental attributes, including environmental factors and the accessibility of public transport facilities. This study used the basic hedonic attributes as the internal attributes. Secondly, this study used Arc GIS software 10.1 to compute the nearest distances of the houses to the neighborhood environment attributes (e.g., MRT station, railway station, primary school, junior high school) as the external housing attributes. Thirdly, this study also explored the difference-in-difference attributes. The definitions of various variables are shown below in Table 1.

(1) Dependent Variables

This study used the natural logarithm of the housing transaction total price as the dependent variable ( In PRICE ). Cohen and Coughlin (2008 ), Kavetsos (2012 ) and Gibbons et al. (2013 ) argued that the total transaction price as the dependent variable has more complete economic implications as compared to the dependent variable of the transaction unit price in representing the housing price. Moreover, many scholars, including Sirmans et al. (2005 ), Lipscomb and Farmer (2005 ) and Zietz et al. (2008 ) have suggested that the logarithm of dependent variable price has been confirmed to be able to reduce the data heterogeneity and avoid the estimation by using the quantile regression approach. As such, this study adopted the log-linear function type as the regression model.  

(2)Basic Housing Properties

The residential area ( AREA ) is the actual registered area of the housing subject and is a continuous variable. In general, under the same assumed conditions, if the area is greater, the housing price will be higher. McMillen and Redfearn (2010 ), Landry and Jahan-Parvar (2011 ) and Abbott and Klaiber (2011 ) found that residential area has a positive impact on housing price. The housing age ( AGE ) refers to the number of years since the year that construction of the building was completed to the year of the transaction. It is also a continuous variable, and directly reflects the depreciation and replacement cost. Under the same assumed conditions, if the housing age is older, the facilities and building materials will deteriorate over time, resulting in a reduced housing price. Osland (2010 ), Carrillo (2012 ) and Bin and Landry (2013 ) argued that housing age and housing price were negatively correlated. It is expected that housing age has a negative impact on housing price. However, the depreciation due to housing age does not occur in a rigid linear fashion as the depreciation is faster in the beginning. If the regression model adopts the variable of housing age only, only the linear change caused by housing age is observed. Therefore, the addition of the variable of housing age squared ( AGES) is used to observe the non-linear change of depreciation. Kadish and Netusil (2012 ) and Bin and Landry (2013 ) argued that housing age squared and housing price were positively correlated. It is expected that housing age squared has a positive impact on housing price. 

The number of rooms (ROOM ) is related to the internal layout of the housing subject. It is a continuous variable. If the residential area remains unchanged, a greater number of rooms results in a higher likelihood of using the internal space. Sirmans et al. (2005 ), Hansen et al. (2006 ), McMillen and Redfearn (2010) and Araya and Iturra (2013 ) suggested that number of rooms and housing price were positively correlated. It is expected that the number of rooms has a positive impact on housing price. The number of living rooms (LIVEROOM) is related to the internal layout of the housing subject measured by the unit of living rooms. It is a continuous variable. When the number of living rooms is greater, it means the area of the house is larger, and the price of the house will be higher. This attribute will be directly reflected in the house purchasing cost. Hong and Lin (1999 ) and Huang et al. (2010 ) found that the housing price would be higher if the number of living rooms was larger, indicating that the number of living rooms and housing price were positively correlated. It is expected that the number of living rooms has a positive impact on housing price. The number of bathrooms ( BATH ) is also related to the internal layout of the housing subject. It is also a continuous variable. According to a review of previous studies on price models, it is generally accepted that the number of bathrooms has a positive impact on housing price (Anselin and Lozano-Gracia, 2008 ; McMillen and Redfearn, 2010 ; Baltagi and Bresson, 2011 ) . It is expected that the number of bathrooms has a positive impact on housing price.

The residential floor refers to the floor of the housing subject. It is set as a dummy variable. FLOOR1 means that the house is located on the first floor; otherwise, it is set as 0. Alternatively, FLOOR4  means that the house is located on the fourth floor; otherwise, it is set as 0. Houses on the first floor can usually be used as shops or for other commercial purposes, and thus there will be an effect of increasing the housing price. Therefore, it is expected that being located on the first floor has a positive impact on housing price. As for houses on the fourth floor, the pronunciation of the number “four” sounds similar to the word for “death” in Chinese, so people in Taiwan are generally less willing to live on the fourth floor. Therefore, the utility ratio of the fourth floor is generally lower (Yang and Su, 2011 ). Lin et al. (2011 ) argued that a location on the fourth floor thus has a negative impact on real estate price. It is expected, then, that a location on the fourth floor has a negative impact on housing price.

This study uses data from residential samples including buildings and apartments. In general, a building of six or more floors with elevators is regarded as a building while a building of five or fewer floors without elevators is regarded as an apartment. The housing type (TYPE) is set as a dummy variable. It is set as 1 for a building and 0 for an apartment. The building and apartment are common types of condominium housing in Taiwan. In recent years, most of the newly built buildings have steel structures. In contrast, apartments generally do not have steel structures. Moreover, apartments usually have only four or five floors. Buildings have more than six floors and also have elevators. Therefore, the construction costs of buildings are higher than those apartments (Lin and Ma, 2012 ). It is expected that prices of buildings are higher.  

(3) Neighborhood Environment Attributes

Concerning the distance to MRT station variable, this study considered MRT stations in Taipei City. Since the MRT was put into service, the cost of commuting time has been reduced and the accessibility of neighborhoods has been improved, making urban life more convenient and better; hence, MRT may have a significant impact on housing price. Ahlfeldt and Maennig (2010 ) found that increasing the distance to an MRT station may lead to a reduced housing price. It is expected that the nearest distance to an MRT station has a negative impact on housing price. Regarding the distance to a railway station (TRAIN) variable, this study considered several railway stations including Taipei Station, Wanha Station, Songshan Station, and Nangang Station. As a transportation node, a railway station can promote the development of commercial activities in the surrounding areas, and living facilities closer to railway stations are more numerous and convenient. Gibbons et al. (2014 ) found that the distance to a railway station is inversely proportional to housing price. It is expected that the nearest distance to a railway station has a negative impact on housing price. In terms of the distance to school ( PRIM , JUIN ) variables, the schools considered in this study included primary schools and junior high schools. As schools provide living and recreational space for residents, and provide families with children convenient and safe access to education, the distance to a school is regarded as one of the major factors to consider in purchasing a house. Gibbons et al. (2014 ) found that a housing price would be higher if it was closer to a school, as they were negatively correlated. It is expected that the distance to a school has a negative impact on housing price.

(4) Difference-in-Difference Attributes

As for the time when the luxury housing construction license was obtained ( TIME ) variable, the sample period of this study was from January 1, 2008, to December 31, 2011. The sample data were divided into samples from before a construction license was obtained and samples from after a construction license was obtained by the benchmark period for obtaining such construction licenses of July to December, 2010. It is a dummy variable. The value of the variable for samples during the period from January 1, 2008, to June 30, 2010 (that is, before a license was obtained) is set as 0, while the value of the variable for samples during the period from January 1, 2011, to December 31, 2011 (that is, after a license was obtained) is set as 1. The appearance of luxury housing is more attractive than that of general housing, and it can beautify the surrounding environment; therefore, it can improve the quality of a neighborhood environment to make the neighborhood more attractive (O’Sullivan, 2000 ). In addition, it can attract people of high social and economic status to move in. As a result, regular people and investors may follow the lead of celebrities or others of high socioeconomic status in choosing to live in the neighborhood or invest in the area. People may also have higher self-evaluations if they live in a neighborhood with luxury housing (Cialdini et al., 1976 ; Tesser, 1988 ) resulting in increased housing prices for the whole community. It is thus expected that obtaining a construction license for luxury housing has a positive impact on housing prices.

The luxury housing neighborhood (TREAT) variable refers to the number of houses that are really located in the radius of the luxury housing. It is a dummy variable. If the sample is located within the radius of the luxury housing (the treatment group), the value is set as 1; otherwise (the control group), it is set as 0. This study sought to explore the difference between samples within the radius of 500 meters from the luxury housing and samples outside the radius of 500 meters from the luxury housing in terms of housing price. The development of luxury housing may result in positive spillover effects on a neighborhood. Bourassa et al. (2004 ) argued that attractive buildings can increase neighborhood housing prices. The neighborhood environment can improve the neighborhood quietness and comfort to create a positive impact on neighborhood housing prices. In addition, based on the social attributes of Taiwanese society and the extent to which Taiwanese people typically seek to bask in the glory of the high status people and positive events of a neighborhood, it is expected that neighborhood luxury housing has a positive impact on housing price. The time of obtaining construction license × luxury housing neighborhood ( TIME * TREAT ) variable is an interactive variable referring to the interaction between the time of obtaining the construction license and being in the neighborhood of luxury housing. After obtaining the construction license, if a house is located in the luxury housing neighborhood, luxury housing can beautify the neighborhood environment to improve the neighborhood environmental quality. Moreover, the architectural design of luxury housing can improve the neighborhood housing landscape to result in a positive impact on the neighborhood. Ooi and Le (2013 ) explored the impact of new development projects on neighborhood housing prices. Kavetsos (2012 ) discussed the impact of newly built Olympic stadiums on neighborhood housing prices. Both of those studies applied the difference-in-difference method and predicted that newly developed or newly built buildings have a positive impact on neighborhood housing prices. It is expected that the time to obtain the construction license × luxury housing neighborhood has a positive impact on housing price.

4. DATA SOURCE DESCRIPTIONS AND DESCRIPTIVE STATISTICS

4.1. Data Source Descriptions

In this study, the residential transaction data source used was the “Taiwan Real Estate Transaction Bulletin” published by the Giga House website of the Taiwan Real Estate Transaction Center (Taiwan Real Estate Portal). The data period was from January 1, 2008, to December 31, 2011. Focusing on Taipei City, this study obtained a total of 10,930 samples. Taiwan Real Estate Portal started to provide a transaction information query service in 2005. The data were collected from well-known real estate agencies in Taiwan including Sinyi Realty, Pacific Realtor, Century 21 Real Estate LLC, Chinatrust Real Estate, H&B Real Estate, Dafengfu Real Estate, Good Morning Rehouse, ERA Taiwan Master Franchise, Ever Spring Real Estate, and Chiau Mau Realty, and the data were about transactions with a guarantee of performance. This study obtained the transaction data regarding buildings and apartments from the bulletin, including housing attributes such as the housing transaction price, floor area, housing age, and number of floors. As the data from the website were not limited only to housing transaction data, this study eliminated any data regarding non-residential housing transactions including transactions involving land, parking spaces, factories, offices, and stores, as well as any samples with incomplete information. This study eliminated the top and bottom 5% of samples of housing prices in various administrative regions to avoid the effect of outliers on the statistical computational results. Finally, 8,871 samples were used for the empirical analysis. According to luxury housing standards set by the Taipei City Revenue Service and the central bank of Taiwan,¹ -- that is, independent building, luxurious appearance, great location, good landscape, few households on each floor, parking space for each household, strict security and comprehensive management -- the total price of such a house should be 80 million NTD and above, and the unit price of each ping should be 1 million NTD and above. After cross-validation, this study collected real cases of luxury housing in Taipei City. During the sampling period from July to December 2010, the luxury housing cases with construction licenses were selected as the samples in this study. The five cases of luxury housing for this study were located in various districts of Taipei City, including the Zhongzheng, Songshan, and Da’an districts. The sources of the data on the luxury housing cases included the Ministry of Interior real estate transaction price query service website, the official websites of various construction companies and other housing information websites on the Internet.²

Table-1. Variable Descriptions

4.2. Descriptive Statistics

According to the descriptive statistics in Table 2, the average housing price was 13.978 million NTD. The average residential area (AREA ) was 31.63 ping (1 ping equals 35.58 sq. ft.). The average housing age ( AGE ) was 20.86 years. The average number of rooms ( ROOM ) was 2.47 rooms. The average number of living rooms ( LIVEROOM ) was 1.77 rooms. The average number of bathrooms ( BATH ) was 1.48 rooms. The residential floor being on the first floor ( FLOOR1 ) accounted for 6.0% of the sample houses, the residential floor being on the fourth floor ( FLOOR4 ) accounted for 21.0%, and the residential floor being on the other floors accounted for 73.0%. Regarding housing types ( TYPE ), buildings accounted for 56.0%, while apartments accounted for 44.0%. The average nearest distance to an MRT station ( MRT ) was 588.55 meters. The average distance to the nearest railway station ( TRAIN ) was 3,713 meters. The average distance to the nearest primary school ( PRIME ) was 401.92 meters. The average distance to the nearest junior high school ( JUNI ) was 586.60 meters.

Table-2. Descriptive statistics (N=8,871)

5. EMPIRICAL RESULTS

Regarding basic housing properties, the residential area estimation coefficient was 0.017, having reached the 1% significance level. This means that if the residential area is increased by 1 ping, the housing price will increase by 1.7%. The housing age estimation coefficient was -0.016, having reached the 1% significance level. This means that if the housing age is increased by one year, then the housing price will decrease by 1.6%. Its square estimation coefficient was 0.001, having reached the 1% significance level. This means that an older housing age can result in a lower housing price. However, the decrease is at the marginal decrease state. The number of rooms’ estimation coefficient was 0.027, having reached the 1% significance level. This means that an increase of one room can result in an increase of the housing price by 2.7%. The number of living rooms estimation coefficient was 0.082, having reached the 1% significance level. This means that an increase of one living room can result in an increase of housing price by 8.2%. The number of bathrooms estimation coefficient was 0.136, having reached the 1% significance level. This means that an increase of one bathroom would result in a housing price increase of 13.6%. The Floor1 estimation coefficient was 0.175, and the Floor4 estimation coefficient was -0.069, having reached the 1% significance level. This means that the price of Floor1 was higher than the average price for the rest of the floors by about 17.5%, while the price of Floor4 was lower than the average price for the rest of the floors by about 6.9%. The housing type estimation coefficient was 0.253, having reached the 1% significance level. This means the average building housing price was higher than the average apartment housing price by about 25.3%.

Regarding neighborhood environment attributes, the variable of distance to an MRT station estimation coefficient was -2.604e-5, having reached the 1% significance level. This suggests that an increase in the distance to an MRT station by 1 meter would result in a housing price reduction by about 0.026%. The distance to a railway station estimation coefficient was -2.408e-5, having reached the 1% significance level. This suggests that an increase in the distance to a railway station by 1 meter would result in a housing price reduction by about 0.024%. The distance to a primary school estimation coefficient was -8.204e-5, having reached the 1% significance level. This suggests that an increase in the distance to a primary school by 1 meter would result in a housing price reduction by about 0.082%. The distance to a junior high school estimation coefficient was -8.360e-5, having reached the 1% significance level. This suggests that increase in the distance to a junior high school by one 1 meter would result in a housing price reduction by about 0.083%.

Regarding difference-in-difference attributes, the estimation coefficient of the variable of when the luxury housing construction license was obtained was 0.202, having reached the 1% significance level. This means that housing prices were higher by about 20.2% after obtaining the luxury housing construction license. The estimation coefficient of the variable of luxury housing neighborhood was 0.101, having reached the 1% significance level. This suggests that housing prices in the luxury housing neighborhood were higher by 10.1%. The estimation value of the difference-in-difference was 0.106, having reached the 1% significance level. This suggests that the luxury housing (after obtaining the construction license) had positive spillover effects on neighborhood house prices by about 10.6%.

Model 3 categorizes samples in the luxury housing neighborhood (the treatment group) into three groups, including samples within the distance of 100 meters, samples between the distances of 100 meters and 300 meters, and samples between the distances of 300 meters and 500 meters, to test whether the impact of luxury housing varies with increasing distance from the luxury housing. The GLS regression results are as shown in Model 3, and its F statistic value was 765.499, having reached the 1% significance level, whilewas 65.0%. The estimation results of the coefficients of variables of the basic housing properties and neighborhood environment attributes were not significantly different from those of Model 2.

Regarding the interactive items of difference-in-difference, the estimation coefficients of the samples of 100 meters, 100 meters to 300 meters and 300 meters to 500 meters after obtaining the construction license were 0.490, 0.182 and 0.047 respectively. This suggests that the spillover effects of luxury housing after obtaining the construction license will increase the housing prices in general. However, the effects will gradually decrease over greater distances from 49.0% to 18.2% to 4.7%. Moreover, the effect in the case of samples at distances from the luxury housing in the range from 300 meters to 500 meters did not reach the level of significance.

6. DISCUSSION

The focus of a DD study is (interaction term coefficient). However, to make the estimation more efficient and avoid the possible selection error, covariates can be added. The covariates are selected according to the hedonic model, and the empirical results seem to be in line with prediction results.  

According to the difference-in-difference attribute regression results, the time to obtain the luxury housing construction license, luxury housing neighborhood, and time to obtain the luxury housing construction license × luxury housing neighborhood had significant and positive impacts on housing prices, as expected. The time to obtain the luxury housing construction license had a positive and significant impact on housing prices. Its estimation coefficient was about 0.202, suggesting that housing price after obtaining luxury housing construction license were increased by about 20.2%. The luxury housing had a positive impact on neighborhood housing prices. Its estimation coefficient was about 0.101, suggesting that the housing prices in the neighborhood of the luxury housing increased by 10.1%. The interactive term of the time to obtain luxury housing construction license and luxury housing neighborhood had a positive and significant impact on housing prices. The estimation coefficient was about 0.106, suggesting that the difference in the time to obtain construction license and luxury housing neighborhood and non-luxury housing neighborhood spatial difference was approximately 10.6%. The results are in line with the findings by Bourassa et al. (2004 ), Kavetsos (2012 ) and Ooi and Le (2013 ). Newly built or attractive buildings with positive effects on a neighborhood can improve that neighborhood’s environmental quality and comfort to facilitate the increase in neighborhood housing prices. The luxury housing can significantly beautify the surrounding environment and maintain tidiness, so that it can help improve the neighborhood’s environmental quality.

According to GLS estimation results, the difference-in-difference coefficients B₁ , B₂ , B₃ and were obtained. According to the descriptions of Hill et al. (2011 ) about difference-in-difference, changes in luxury housing after obtaining the construction license, luxury housing neighborhood and non-luxury housing neighborhood housing prices can be determined. Figure 1 suggests that the luxury housing neighborhood (the treatment group) housing prices will be higher than those of the non-luxury housing neighborhood (the control group). Moreover, regardless of construction decisions, there was a common trend that housing prices were increased over time. If the two increase at the same rate, after the construction decision, the neighborhood (the treatment group) housing prices will increase more rapidly than the housing prices of the non-luxury housing neighborhood (i.e., greater than the original trend). The interactive term estimated by using the difference-in-difference method could be used to calculate the housing price difference before and after obtaining the construction license. To get the difference in the impact of luxury housing, we need to convert the estimation coefficient by using log-linear function type for computation to get more accurate results.  

The log-linear function estimated coefficients can roughly compute and illustrate the percentage of change in a dependent variable caused by each unit of independent variable. The method is fast and easy. To more accurately illustrate the regression coefficients estimated by using the log-linear function, the regression coefficients can be input intoBy multiplying the value and the average housing price of the sample data, this paper computed the real change in housing prices caused by each unit of change in an independent variable. The computation of the log-linear function is as shown in Table 4. The intercept, TIME , TREAT , TIME*TREAT  GLS estimated coefficients were .

 with the average sample housing price, we obtained 460.3, 312.7, 148.5 and 156.2, respectively. This suggests that the intercept is the starting price without the influence of any other variables. The basic housing price is about 4.603 million NTD. In the case of the control group, after obtaining the construction license, the housing prices increased by 3.127 million NTD. Housing prices in the neighborhood of the luxury housing before obtaining the construction license were higher by about 1.485 million NTD. The difference before and after getting the construction license and the luxury housing neighborhood and non-luxury housing neighborhood difference, that is, the difference-in-difference  housing price,  was about 1.562 million NTD.

Figure-1. difference-in-difference estimation results

Table-4. Log-liner function conversion

Note: the value in the brackets refers to the price of converted result percentage multiplied by sample average housing price measured by the unit of NTD (10, 000).

7. CONCLUSION AND SUGGESTIONS

7.1.Conclusion

This study applied the difference-in-difference method by the time to obtain luxury housing constructio

Kiel and McClain (1995 ), Dehring et al. (2007 ) and Ooi and Le (2013 ), Bourassa et al. (2004 ) argued that attractive buildings can increase neighborhood housing prices. However, empirically, they only estimated the impact of buildings on neighborhood housing prices without deducting the location difference caused by time trends and geographic space from the estimation results, and thus may have overestimated the impact of such buildings. The difference-in-difference method applied in this study can explore the impacts of differences caused by the occurrence of an event and by geographic space on housing prices. Through the interactive variables of time and space -- that is, the interactive term of the time to obtain the luxury housing construction license and luxury housing neighborhood -- this study was able to provide more accurate estimation results.

According to the empirical results of this study, the housing price difference caused by the time to obtain the construction license and distance to the luxury housing was about 1.56 million NTD, suggesting that obtaining a construction license has a real impact on luxury housing neighborhood housing prices. That is, the average housing price increases by about 1.56 million NTD after obtaining a construction license. This paper further categorized the treatment group samples into those within 100 meters, those between 100 meters and 300 meters and those between 300 meters and 500 meters for estimation. The results suggested that housing prices increase in inverse proportion to the distance from the luxury housing neighborhood.

7.2. Research Limitations and Follow-Up Suggestions

Due to the limitation of data processing, this study only measured the nearest linear distance of a given house from the relevant public facilities without considering the real geographical environment (e.g., rivers and slopes) and urban road planning. If the real distances from the houses to the public facilities can be measured and the above geographic environment and urban roads can be considered in a follow-up study, then housing prices can be more accurately and realistically estimated.  

REFERENCES

Abbott, J.K. and H.A. Klaiber, 2011. An embarrassment of riches: Confronting omitted variable bias and multi-scale capitalization in hedonic price models. Review of Economics and Statistics, 93(4): 1331-1342.Available at: https://doi.org/10.1162/rest_a_00134.

Ahlfeldt, G.M. and W. Maennig, 2010. Impact of sports arenas on land values: Evidence from Berlin. The Annals of Regional Science, 44(2): 205-227.Available at: https://doi.org/10.1007/s00168-008-0249-4.

Almatarneh, R.T., 2013. Choices and changes in the housing market and community preferences: Reasons for the emergence of gated communities in Egypt: A case study of the Greater Cairo Region, Egypt. Ain Shams Engineering Journal, 4(3): 563-583.Available at: https://doi.org/10.1016/j.asej.2012.11.003.

Andersson, D.E., O.F. Shyr and J. Fu, 2010. Does high-speed rail accessibility influence residential property prices? Hedonic estimates from Southern Taiwan. Journal of Transport Geography, 18(1): 166-174.Available at: https://doi.org/10.1016/j.jtrangeo.2008.10.012.

Anselin, L. and N. Lozano-Gracia, 2008. Errors in variables and spatial effects in hedonic house price models of ambient air quality. Empirical Economics, 34(1): 5-34.Available at: https://doi.org/10.1007/s00181-007-0152-3.

Araya, D.P. and R.V. Iturra, 2013. Substitution bias and the construction of a spatial cost of living index. Papers in Regional Science, 92(1): 103-117.

Baltagi, B.H. and G. Bresson, 2011. Maximum likelihood estimation and Lagrange multiplier tests for panel seemingly unrelated regressions with spatial lag and spatial errors: An application to hedonic housing prices in Paris. Journal of Urban Economics, 69(1): 24-42.Available at: https://doi.org/10.1016/j.jue.2010.08.007.

Bin, O. and C.E. Landry, 2013. Changes in implicit flood risk premiums: Empirical evidence from the housing market. Journal of Environmental Economics and Management, 65(3): 361-376.Available at: https://doi.org/10.1016/j.jeem.2012.12.002.

Bourassa, S.C., M. Hoesli and J. Sun, 2004. The price of aesthetic externalities. Journal of Real Estate Literature, 13(2): 165-188.

Boyle, M. and K. Kiel, 2001. A survey of house price hedonic studies of the impact of environmental externalities. Journal of Real Estate Literature, 9(2): 117-144.

Carrillo, P.E., 2012. An empirical stationary equilibrium search model of the housing market. International Economic Review, 53(1): 203-234.Available at: https://doi.org/10.1111/j.1468-2354.2011.00677.x.

Cialdini, R.B., R.J. Borden, A. Thorne, M.R. Walker, S. Freeman and L.R. Sloan, 1976. Basking in reflected glory: Three (football) field studies. Journal of Personality and Social Psychology, 34(3): 366-375.Available at: https://doi.org/10.1037//0022-3514.34.3.366.

Cohen, J.P. and C.C. Coughlin, 2008. Spatial hedonic models of airport noise, proximity, and housing prices. Journal of Regional Science, 48(5): 859-878.Available at: https://doi.org/10.1111/j.1467-9787.2008.00569.x.

Dehring, C.A., C.A. Depken and M.R. Ward, 2007. The impact of stadium announcements on residential property values: Evidence from a natural experiment in dallas-fort worth. Contemporary Economic Policy, 25(4): 627-638.Available at: https://doi.org/10.1111/j.1465-7287.2007.00077.x.

Dempsey, J.A. and A.J. Plantinga, 2013. How well do urban growth boundaries contain development? Results for oregon using a difference-in-difference estimator. Regional Science and Urban Economics, 43(6): 996-1007.Available at: https://doi.org/10.1016/j.regsciurbeco.2013.10.002.

Des Rosiers, F., T. Marius, K. Yan and V. Paul, 2002. Landscaping and house values: An empirical investigation. Journal of Real Estate Research, 23(1-2): 139-162.

DiPasquale, D. and W.C. Wheaton, 1996. Urban economics and real estate markets. Englewood Cliffs: Prentice Hall.

Dubé, J., F. Des Rosiers, M. Thériault and P. Dib, 2011. Economic impact of a supply change in mass transit in urban areas: a Canadian example. Transportation Research Part A: Policy and Practice, 45(1): 46-62.Available at: https://doi.org/10.1016/j.tra.2010.09.002.

Gibbons, S., S. Machin and O. Silva, 2013. Valuing school quality using boundary discontinuities. Journal of Urban Economics, 75: 15-28.Available at: https://doi.org/10.1016/j.jue.2012.11.001.

Gibbons, S., S. Mourato and G.M. Resende, 2014. The amenity value of English nature: A hedonic price approach. Environmental and Resource Economics, 57(2): 175-196.Available at: https://doi.org/10.1007/s10640-013-9664-9.

Hansen, J.L., E.D. Benson and D.A. Hagen, 2006. Environmental hazards and residential property values: Evidence from a major pipeline event. Land Economics, 82(4): 529-541.Available at: https://doi.org/10.3368/le.82.4.529.

Heintzelman, M.D., 2010. The value of land use patterns and preservation policies. The BE Journal of Economic Analysis & Policy, 10(1): 1-26.Available at: https://doi.org/10.2202/1935-1682.2484.

Hill, R.C., W.E. Griffiths and G.C. Lim, 2011. Principles of econometrics. 4th Edn., Hoboken: John Wiley and Sons.

Hong, D. and C. Lin, 1999. A study on the impact of subway system and road width on the housing prices of Taipei. Journal of Housing Studies, 8: 47-67.

Huang, B., B. Wu and M. Barry, 2010. Geographically and temporally weighted regression for modeling spatio-temporal variation in house prices. International Journal of Geographical Information Science, 24(3): 383-401.Available at: https://doi.org/10.1080/13658810802672469.

Ihlanfeldt, K.R., 2007. The effect of land use regulation on housing and land prices. Journal of Urban Economics, 61(3): 420-435.Available at: https://doi.org/10.1016/j.jue.2006.09.003.

Kadish, J. and N.R. Netusil, 2012. Valuing vegetation in an urban watershed. Landscape and Urban Planning, 104(1): 59-65.

Kavetsos, G., 2012. The impact of the london olympics announcement on property prices. Urban Studies, 49(7): 1453-1470.Available at: https://doi.org/10.1177/0042098011415436.

Kiel, K.A. and K.T. McClain, 1995. House prices during siting decision stages: The case of an incinerator from rumor through operation. Journal of Environmental Economics and Management, 28(2): 241-255.Available at: https://doi.org/10.1006/jeem.1995.1016.

Lai, R.N., K. Wang and Y. Zhou, 2004. Sale before completion of development: Pricing and strategy. Real Estate Economics, 32(2): 329-357.Available at: https://doi.org/10.1111/j.1080-8620.2004.00094.x.

Landry, C.E. and M.R. Jahan-Parvar, 2011. Flood insurance coverage in the coastal zone. Journal of Risk and Insurance, 78(2): 361-388.Available at: https://doi.org/10.1111/j.1539-6975.2010.01380.x.

Liang, J.H., 2009. A comparative study of urban renewal spillover effects. Department of Natural Resources: Chinese Culture University. NSC 98-2410-H-034-025-.

Lin, C.C. and Y.C. Ma, 2012. An application of bayesian inference in the real estate market: A case study of taipei collective housing. Journal of housing studies. 21(1): 1-18.

Lin, J. and S. Jou, 2005. High-income housing and their building blocs in Taipei: A case study of luxurious condominiums in Xinyi Centre. Journal of Geographical Science, 40: 17-43.

Lin, T.C., W.Y. Lee and K.H. Chen, 2011. Condominium depreciation and building taxes. Journal of City and Planning, 38(1): 31-46.

Lipscomb, C.A. and M.C. Farmer, 2005. Household diversity and market segmentation within a single neighborhood. The Annals of Regional Science, 39(4): 791-810.Available at: https://doi.org/10.1007/s00168-005-0020-z.

Matthews, J.W. and G.K. Turnbull, 2007. Neighborhood street layout and property value: The interaction of accessibility and land use mix. The Journal of Real Estate Finance and Economics, 35(2): 111-141.Available at: https://doi.org/10.1007/s11146-007-9035-9.

McMillen, D.P. and C.L. Redfearn, 2010. Estimation and hypothesis testing for nonparametric hedonic house price functions. Journal of Regional Science, 50(3): 712-733.Available at: https://doi.org/10.1111/j.1467-9787.2010.00664.x.

Munneke, H., J.T. Ooi, C. Sirmans and G.K. Turnbull, 2011. Sequential sales of similar assets: The law of one price and real estate. Journal of Regional Science, 51(2): 355-370.Available at: https://doi.org/10.1111/j.1467-9787.2010.00691.x.

Neter, J., M.H. Kutner, C.J. Nachtsheim and W. Wasserman, 1996. Applied linear statistical models. 4th Edn., New York: McGraw-Hill.

O’Sullivan, A., 2000. Urban economics. 4rt Edn., Boston: McGraw-Hill.

Ooi, J.T. and T.T. Le, 2013. The spillover effects of infill developments on local housing prices. Regional Science and Urban Economics, 43(6): 850-861.Available at: https://doi.org/10.1016/j.regsciurbeco.2013.08.002.

Osland, L., 2010. An application of spatial econometrics in relation to hedonic house price modeling. Journal of Real Estate Research, 32(3): 289-320.

Sirmans, S., D. Macpherson and E. Zietz, 2005. The composition of hedonic pricing models. Journal of Real Estate Literature, 13(1): 1-44.

Tesser, A., 1988. Toward a self-evaluation maintenance model of social behavior. Advances in Experimental Social Psychology, 21: 181-227.Available at: https://doi.org/10.1016/s0065-2601(08)60227-0.

Voicu, I. and V. Been, 2008. The effect of community gardens on neighboring property values. Real Estate Economics, 36(2): 241-283.Available at: https://doi.org/10.1111/j.1540-6229.2008.00213.x.

Wooldridge, J.M., 2009. Introductory econometrics: A modern approach. Canada: South-Western Cengage Learning.

Yang, C. and S. Su, 2011. The impacts of housing price in yimby and nimby facilities. Journal of Housing Studies, 20(2): 61-80.

Zietz, J., E.N. Zietz and G.S. Sirmans, 2008. Determinants of house prices: A quantile regression approach. The Journal of Real Estate Finance and Economics, 37(4): 317-333.Available at: https://doi.org/10.1007/s11135-009-9296-x.

Footnote