1# RECORD 4412b66b7af7dc2e4784f04e96dda07a5d24ef902# TITLE: Buyers’ property asset purchase decisions: an empirical study on the high-end residential property market in Hong Kong3# YEAR: 2016.0 VENUE: nan DOI: 10.3846/1648715x.2015.11053224# AUTHORS: W. M. Jayantha; Jia-Ming Lau5# PAGES: 16 CANDIDATE TABLE PAGES (1-based): [10, 11, 12]6# ABSTRACT: Demand for luxury housing units from the upper and upper-middle income groups in Hong Kong has been increasing over the last few years. as the market cannot satisfy demand, some prospective buyers have turned their attention to “special” housing units. this research paper attempts to investigate buyers’ preferences for two types of “special” units, namely duplex units and adjoining flats. the study investigates the price premiums paid by the buyers, and examines the effects of these special units on property price. the study employs two hedonic price models: one measuring the buy- ers’ preference on duplex units and the other one measuring buyers’ preference on adjoining flats. The results show that buyers are willing to pay a larger premium for special residential units: HK$588/ft-super-2 more for a duplex unit and HK$934/ft-super-2 more for an adjoining property unit, respectively. furthermore, a relatively larger premium is found for adjoining flats compared to duplex units. This implies that a property unit, purchased as an adjoining flat can add more value to the property price (in terms of $s per sq. ft.) compared with being a duplex unit.789===== PAGE 1 (score 11, numeric-paren 1, table-words 0, var-words 10) =====1011Copyright © 2016 Vilnius Gediminas Technical University (VGTU) Press12http://www.tandfonline.com/tspm13International Journal of Strategic Property Management14ISSN 1648-715X / eISSN 1648-9179152016 Volume 20(1): 1–1616doi:10.3846/1648715X.2015.110532217* Corresponding author. E-mail: bsmjwadu@polyu.edu.hk18BUYERS’ PROPERTY ASSET PURCHASE DECISIONS: 19AN EMPIRICAL STUDY ON THE HIGH-END RESIDENTIAL 20PROPERTY MARKET IN HONG KONG21Wadu Mesthrige JAYANTHA a,*, Jia Ming LAU b22a Department of Building and Real Estate, The Hong Kong Polytechnic University, Hong Kong SAR, 23China24b Freelance author, Kowloon, Hong Kong SAR, China25Received 5 September 2014; accepted 14 January 201526ABSTRACT. Demand for luxury housing units from the upper and upper-middle income groups in 27Hong Kong has been increasing over the last few years. As the market cannot satisfy demand, some 28prospective buyers have turned their attention to “special” housing units. This research paper attempts 29to investigate buyers’ preferences for two types of “special” units, namely duplex units and adjoining 30flats. The study investigates the price premiums paid by the buyers, and examines the effects of these 31special units on property price. The study employs two hedonic price models: one measuring the buy-32ers’ preference on duplex units and the other one measuring buyers’ preference on adjoining flats. The 33results show that buyers are willing to pay a larger premium for special residential units: HK$588/ft2 34more for a duplex unit and HK$934/ft2 more for an adjoining property unit, respectively. Furthermore, 35a relatively larger premium is found for adjoining flats compared to duplex units. This implies that a 36property unit, purchased as an adjoining flat can add more value to the property price (in terms of $s 37per sq. ft.) compared with being a duplex unit.38KEYWORDS: Property market; Duplex unit; Adjoining flat; Property price; Hong Kong391. INTRODUCTION40Even though the housing affordability issue in 41Hong Kong has been a serious problem, especially 42among many low and medium income households, 43there has been an increasing demand for luxury 44housing units by the upper middle and high income 45households over the last few years. With the rising 46aspirations of this sector in society, along with in-47creasing national income (and per capita income), 48there is an increasing demand for so-called “spe-49cial” housing units (i.e. duplex units and adjoining 50flats). As the market cannot satisfy demand, some 51prospective buyers have turned their attention to 52these types of “special” housing units, for some as 53a way of getting into the market, for others as a 54way of increasing their investment and diversifica-55tion into this type of asset. The main reason for this 56significant increase in demand for special units is 57the existence of the wealth effect and the limited 58supply of these units in the property market (Fong 592009). The wealth effect is defined as “…an increase 60in spending or consumption that accompanies an in-61crease in wealth, or merely a perceived increase in 62wealth.” (Pigou 1943). According to Cheung et al. 63(2004), property equity has dominated the wealth 64portfolio of typical middle-class Hong Kong resi-65dents for a long time. Buying “bricks and mortar”, 66i.e. property, has been viewed as one of the safest 67stores of wealth as well as investment, according to 68Chinese tradition. These special housing units, in 69particular (i.e. duplex units and adjoining flats), are 70regarded as high quality durable investment goods 71as well as consumer goods. It is clear that people 72in Hong Kong have begun, on one hand, to appre-73ciate the characteristics of these property units 74in terms of prestige value and potential for asset 75appreciation, and, on the other hand, to embrace 76green living (Jayantha, Wan 2012). Hence, these 77special housing units are in high demand in the lo-78cal property market.798081===== PAGE 10 (score 42, numeric-paren 0, table-words 16, var-words 8) =====8283W. M. Jayantha et al.841085aspects of the properties, date of occupation of the 86building, etc. were obtained from Economic Prop-87erty Research Centre (EPRC) Limited and crossed 88checked with Centadata, which is established 89by Centaline Property Agency. EPRC provides 90a database that records all the registered prop-91erty transaction in the Land Registry since 1991 92in Hong Kong. The data provided in Centadata is 93based on the registered property transactions in 94the Government Land Registry. When necessary, 95a land search was obtained from the Land Registry 96for verification and supplementary purposes. Other 97property information is obtained through various 98government departments and site investigation. 99Location, geographical and neighbourhood informa-100tion of the selected buildings are gathered from the 101GeoInfo Map powered by the Lands Department. 102Floor plans of some of the selected properties were 103collected from the Building Records Access and 104Viewing On-line (BRAVO) System established by 105the Buildings Department. The building manage-106ment information employed in this research was 107obtained from the Database of Private Buildings 108in Hong Kong provided by the Home Affairs De-109partment. Recreational facilities associated with 110the chosen residential developments were verified 111through property agents’ and developers’ websites. 112Site investigation was also conducted to identify the 113views, accessibilities, etc. of the properties.1143.4. Statistical interpretations in the HPM115Empirical results of HPM (regression model) are 116generally interpreted and analyzed by using nu-117merous simple statistical techniques (i.e. t-test, 118F-test and adjusted R2). Our HPM has eleven in-119dependent variables, and the t-test is used to test 120whether each parameter of a variable is significant 121or not. This is done by comparing the estimated 122t-value with the critical t-value. An independent 123variable is said to be significant if the absolute 124t-statistic for that coefficient of the variable is 125greater than the critical value. That implies that a 126particular variable has a significant impact on the 127dependent variable. Another important statistic in 128the regression model is the F-test, which reflects 129the overall performance of the model. The overall 130model is said to be performing well or is statisti-131cally highly significant if the estimated F-statis-132tic is larger than the critical value. That means 133the independent variables included in the model 134jointly explain the dependent variable satisfacto-135rily. Finally, the explanatory power of the model 136is evaluated through the adjusted R2 value that 137ranges from 0 to 1; and the greater the value of 138it, the better the model. High values of R2 indicate 139that the independent variables selected, together 140substantially explain the behaviour of the depend-141ent variable (meaning that no significant variables 142are missing).143Table 2. Summary of descriptive statistics (Model 1)144Variable145Mean146STD147Min148Max149Price1509010.881511806.721522177.0015315472.00154FlatSize155770.01156202.62157556.001581229.00159FlatRoof16068.5016192.341620.00163292.00164AGE16514.561665.301675.7316825.24169Floor17017.141710.8817215.0017319.001744. EMPIRICAL FINDINGS1754.1. Results of model 1 (Duplex unit)176Table 2 presents the descriptive statistics of the 177variables used in model 1. Empirically estimated 178coefficients and the explanatory power of the mod-179el along with statistical measures are presented in 180Table 3. All the employed independent variables 181(except Orient) are found to be highly significant 182at conventional levels and also bear the antici-183pated theoretical signs of variables. The adjusted 184R2 value of 0.745 indicates that around 75 percent 185of the total variation of residential property price 186can be described by the explanatory variables, and 187this is quite a high value. The estimated F-value of 188231.95 (p-value is 0.000) also suggests that the set 189of explanatory variables chosen are jointly statisti-190cally significant and at least one of the independ-191ent variables is able to explain the variation of the 192property price in the model. Confidence interval 193levels (at 95%) were also tested for all the esti-194mates of independent variables8. All the estimated 195coefficients are within the range of 95 % confidence 196level, confirming the reliability of the estimates.197The most relevant and significant variable in 198the study is DUPLEX, and as anticipated, it is 199found to be statistically highly significant. The es-200timated t-value of 2.408 and p-value 0.016 suggest 201that the variable DUPLEX is highly significant at 2025 percent level. The coefficient of 588.54 of DU-203PLEX means a property being a duplex unit can 204add value and raise the property price by $588.54/205sq.ft. To put it differently, potential buyers are 206ready to pay $588.54 more per square feet for a 207duplex property unit than for a normal/typical 208unit. This tells us that if a residential building has 209duplex units, the value of the other units in the 210residential property increases significantly.2118 Not reported in the text, but available upon request.212213214===== PAGE 11 (score 53, numeric-paren 10, table-words 11, var-words 17) =====215216Buyers’ property asset purchase decisions: an empirical study on the high-end residential property market…21711218A duplex unit can enhance the living standard 219of the end user due to the protection of privacy 220among users through the separation of spaces into 221two floors. In addition, the property owner can en-222joy the prestige value of the duplex unit because 223of the limited supply of these units in the market. 224The result shows that the quality of housing prop-225erty in terms of prestige and living standard seems 226to be playing an important role in buyers’ decisions 227on purchasing a property. As duplex units are now 228well recognized housing units with these attrib-229utes, potential property buyers may be willing to 230pay a premium for these special housing units that 231enhance their prestige level and living standard.232All the other explanatory variables (except Ori-233ent) incorporated into the model are also statisti-234cally significant. The coefficient of 3.229 of Flat-235Size suggests a positive relationship between the 236saleable area (excluding flat roof area) and the 237property value per square feet. Buyers are willing 238to pay a higher price as they can enjoy a larger liv-239ing space. It reveals that a property with a larger 240perimeter and larger living area naturally tend to 241appreciate more rapidly (Cebula 2009). The result 242is consistent with previous research works (e.g., 243Mok et al. 1995; So et al. 1997; Chau et al. 2003). 244Variable FlatRoof is also found to be highly sig-245nificant (t-statistics of 3.734 and p-value of 0.000) 246with the coefficient of 4.056. This indicates that 247potential property buyers value a flat roof area as 248part of the property and are willing to pay $4.056 249for every 1 square foot of flat roof area. In other 250words, every 1 sq. ft. of flat roof area causes the 251property price (per saleable area) (excluding flat 252roof area) to increase by $4.056. The positive re-253lationship is obvious since the property price per 254saleable area by definition excludes the flat roof 255area. With a bigger flat roof size, the property 256owner can enjoy a larger open space for leisure 257and recreational activities, for example barbeque, 258resting, feeding pets, etc. Hence, the purchasers 259are willing to pay a premium for the added benefit 260of a flat roof area.261The negative coefficient of age (–45.506) im-262plies a negative correlation between building age 263and property price (per square feet) with a rate of 264$45.506/sq. ft. per year. The estimate is significant 265at 5% level with t-statistics of –6.948 and p-value 266of 0.000. This finding is consistent with Mok et al. 267(1995) and Choy et al. (2007) and Coulson and Lahr 268(2005). The age of a building is expected to link 269negatively with the property value as the build-270ing structure deteriorates physically with time. 271Potential buyers obviously pay only lower prices 272for older buildings as structural defects are gener-273ally found in aged properties requiring significant 274maintenance, so buyers pay only lower prices.275The coefficient of Floor has a positive sign with 276a value of 187.311, which means that the higher 277the floor level of the property is located, the higher 278the property price per square feet with a rate of 279$187.311. for each successive floor level. The result 280is significant at 5% level with t-statistics of 2.843 281and p-value of 0.005. This suggests that the buy-282ers are willing to pay more ($187.31 per square 283foot) for a unit at higher level to enjoy the better 284view and the cleaner air. Generally, higher level 285units enjoy cleaner air and better view, and fewer 286mosquitoes, compared with properties located at 287lower floor levels and this is totally true in Hong 288Kong. The result is in line with previous research-289ers such as Mok et al. (1995), So et al. (1997) and 290Chau et al. (2003).291Property buyers treat ‘orientation’ as a key fac-292tor when purchasing a property in Hong Kong. The 293reason being that the West direction is normally 294considered as the worst orientation because units 295facing West would be a lot warmer in summer, 296whilst on the other hand; units facing South di-297rection would enjoy a better ventilation (airflow) 298and better natural lighting. However, the results 299do not support this view. The variable Orient is 300found to be insignificant even though it bears the 301expected negative sign. This suggests that buyers 302do not take orientation into serious considerations 303in purchasing properties. The reason may be partly 304because some blocks in selected residential estates, 305for instance Heng Fa Chuen, are surrounded by 306other blocks of building or mountain and thus have 307a less exposure to excessive sunlight. The obstruc-308tion of sunlight may somehow reduce the negative 309impact on some of the units.310Table 3. Coefficients of the model 1 (Duplex)311Variable312Coefficient313t-statistic314Sig.315Constant3163465.965*3173.1113180.002319FlatSize3203.229*3215.4143220.000323FlatRoof3244.056*3253.7343260.000327AGE328–45.506*329–6.9483300.000331Floor332187.311*3332.8433340.005335Orient336–102.268337–1.3673380.172339ViewHw340–229.866**341–2.1903420.029343ViewMount344390.095**3452.5113460.012347ViewSea3481368.619*3498.5303500.000351DUPLEX352588.540**3532.4083540.016355Adj. R23560.745357F-value358231.950359*, ** indicate significant at 1% and 5% levels, respectively.360361362===== PAGE 12 (score 65, numeric-paren 14, table-words 15, var-words 14) =====363364W. M. Jayantha et al.36512366The view factor is another significant element 367in Hong Kong that influences purchasing decision 368of a property (see for example, Mok et al. 1995). 369Sea view is considered as the best, followed by the 370mountain view for many property buyers whilst 371units having high-way view or obstruction with 372adjacent buildings and structures are considered 373negatively. Findings of the model support this 374view. The negative coefficient (–229.866) of Vie-375wHw indicates that the possession of a highway or 376railway view lowers the property price (per sq. ft.) 377significantly. The result is significant at 5% level 378with t-statistics of –2.190 and p-value of 0.029. It 379is obvious that traffic noise and air pollution aris-380ing from the highway or railway cause an undesir-381able living environment to the end user leading 382to discounts in the property value. This finding is 383also in line with previous works (e.g., Nelson 1982; 384Theebe 2004; Brandt, Maennig 2011). On the other 385hand, the positive coefficient of ViewMount sug-386gests that possession of a mountain view enhanc-387es property price (per square foot). It is obvious 388that the tenants would appreciate a mountain 389view (compared to a highway or buildings view) 390as it is pleasing to have a green view to the eye 391of the end user, and hence they are willing to pay 392a higher price for such properties. The most ap-393preciated view by the buyers is the sea view; and, 394this is very clearly reflected through the result. 395Highly significant positive coefficient (1358.619) 396of ViewSea indicates that the possession of a sea 397view would increase the property price dramati-398cally, by $1358.619/sq.ft. Potential buyers would 399be willing to pay a larger premium if the unit pos-400sesses a sea view as the sea view can yield more 401satisfaction to the end users with a better living 402environment. The estimate is significant result at 4031% level with -statistics of 8.530 and p-value of 4040.000. This finding is also consistent with previous 405studies (e.g., Mok et al. 1995; Choy et al. 2007).406All these physical and locational attributes 407have been well documented in the literature with 408regard to the housing price. For example, floor 409area of the unit (Cebula 2009; Neill et al. 2007; 410Sirmans et al. 2005; Coulson, Lahr 2005; Chau 411et al. 2003; So et al. 1997; Mok et al. 1995), age 412(Choy et al. 2007; Neill et al. 2007; Coulson, Lahr 4132005; Sirmans et al. 2005; Mok et al. 1995), floor 414(Jayantha, Wan 2012; Chau et al. 2003; So et al. 4151997; Mok et al. 1995), orientation (Jayantha, Wan 4162012), view (Brandt, Maennig 2011; Theebe 2004; 417Mok et al. 1995; Nelson 1982).4184.2. Results of model 2 (Adjoining flat)419Descriptive statistics of the data series for the 420model are summarized in Table 4. The empiri-421cal estimates of variables of the model 2, includ-422ing t-statistics and other measures are shown in 423Table 5. The adjusted R2 (0.81) suggest that 81 424percent of the total variation of housing price (for 425saleable area) can be explained by the attributes 426of the model, meaning that is has a very high ex-427planatory power. Again, as in model 1, all the ex-428planatory variables (excluding Orient) are found 429be significant at conventional levels. The F-statis-430tic value of 266.09 indicates that the overall model 431performs very well and all the selected explanatory 432variables are jointly statistically significant.433The main emphasis of the model is on ADJOIN, 434which represents the special housing units (adjoin-435ing flats). As in model 1, it is found that variable 436ADJOIN (special unit) is highly significant and 437bears the expected positive sign. The estimated 438coefficient of ADJOIN is 934.079, which is much 439larger than (estimates for DUPLEX) that of in 440model 1. The t-statistic (2.513) along with 0.012 of 441corresponding p-value indicates that the ADJOIN 442variable is highly significant. Thus this variable 443Table 4. Summary of descriptive statistics (Model 2)444Variable445Mean446STD447Min448Max449Price4508852.524512555.754523608.0045321426.00454FlatSize455852.41456295.64457518.004581654.00459FlatRoof460100.01461177.254620.00463541.00464AGE46514.174668.554670.0046832.58469AgeSq470273.88471269.014720.004731061.51474Floor47532.854766.574770.2547842.00479Table 5. Coefficients of the model 1 (Adjoining)480Variable481Coefficient482t-statistic483Sig.484Constant4851786.530*4861.8544870.064488FlatSize4892.690*4905.3764910.000492FlatRoof4934.469*4944.4424950.000496AGE497–56.700*498–2.8334990.005500AgeSq5011.538*5022.6185030.009504Floor505110.119*5064.7635070.000508Orient509–101.477510–0.9995110.318512ViewOpen513596.154*5143.2265150.001516ViewMount517338.198**5181.9775190.048520ViewSea5211040.848*5226.2075230.000524ADJOIN525934.079*5262.5135270.012528Adj. R25290.810530F-value531266.086532*, ** indicate significant at 1% and 5% level, respec-533tively.534535536===== PAGE 13 (score 30, numeric-paren 0, table-words 5, var-words 19) =====537538Buyers’ property asset purchase decisions: an empirical study on the high-end residential property market…53913540exerts a strong positive impact on residential prop-541erty price. The coefficient of 934.079 suggests that 542a property being an adjoining flat can add val-543ue and raise the property price by $934.079 per 544square feet. Put differently, potential buyers are 545willing to pay $934.079 more per square feet if it 546is an adjoining property unit. This means if a resi-547dential building is with adjoining flats, the value of 548the other normal units in the residential property 549also increases significantly. Adjoining flats by their 550nature improve the living standard of the end user 551due to a wider view and extensive enlargement of 552floor spaces. In addition, great prestige feeling val-553ue that potential buyers have living in such a unit 554encourages them to pay extra money for the prop-555erty being an adjoining flat. This finding is in line 556with the results obtained from model 1. The find-557ings of both models indicate that potential prop-558erty purchasers value special units (Duplex units 559and Adjoining flats) and are ready to pay an extra 560amount of money for them. This is very much in 561line with findings of some of the previous stud-562ies. For example, Cebula (2009), Coulson and Lahr 563(2005) and Sirmans et al. (2005) find out that de-564sirable internal luxury housing attributes such as 565courtyards, number of bedrooms and bathrooms, 566stories in structure, garage car spaces, square 567footage, the presence of a deck, a pool/hot-tub, etc. 568are positively related with the real housing price. 569These special units (Duplex units and Adjoining 570flats) possess all these housing attributes. Similar 571to model 1, all the other variables (except Orient), 572are also found to be highly significant, implying 573the fitness of the model.5744.3. Difference in magnitudes of price effect 575between Model 1 and Model 2576In an attempt to compare these two models, the 577magnitudes of price effects between these two 578cases are analyzed in this section. Models 1 and 5792 clearly indicate that potential buyers value spe-580cial units (Duplex units and Adjoining flats) and 581are ready to pay a premium for them. Both models 582also show that special units exert a clear positive 583impact on property price.584The empirical results summarized in Table 6 585clearly show that the coefficients of special units in 586both models are highly significant. This suggests 587that potential property buyers are happy to pay 588an extra amount over and above typical units (a 589premium) for special units: duplex units and ad-590joining flats. Potential buyers normally tend to 591appreciate and value a better living environment, 592and hence are willing to pay a higher premium for 593these units. However, the magnitude of coefficient 594of ADJOIN is much larger than that in DUPLEX: 595588.054 and 934.079, respectively. That means po-596tential buyers are willing to pay a higher premium 597for adjoining flats compared to duplex units: a pre-598mium of HK$934/ft2 compared to HK$588/ft2. In 599other words, a property being an adjoining flat can 600add more value (i.e. HK$346/ft2) to the property 601price compared with being a duplex unit. The re-602sult indicates that buyers prefer an adjoining flat 603than duplex unit. It can be therefore concluded 604from the empirical results that an adjoining flat 605has a greater positive impact on the property price 606and is more preferred by homeowners in Hong 607Kong compared with duplex unit. This situation 608could be explained this way.609Though duplex units enhance the prestige im-610age and provide a house-like feeling for the own-611ers, it also has some disadvantages for the users. 612Duplex units normally comprise of two stories 613with a stair case, and staircase occupied some 614floor spaces of the unit. However, adjoining flats 615do not have this disadvantage. Also, older people 616may not feel safe in climbing stairs, so adjoining 617floor space would be safer and preferable. On the 618other hand, adjoining flats enhance the floor space 619usage and also possess a broader view. Thus, in-620vestors specially consider adjoining flats as a good 621investment.6225. SUMMARY AND CONCLUSIONS623The study aimed to investigate the buyers’ prefer-624ence for special housing units (duplex units and 625adjoining flats), and to examine the effects of these 626special housing units on property prices. This re-627search issue was investigated by adopting two 628models (cases). The empirical results suggest that 629special units can add value and raise the property 630price significantly. Potential buyers are willing to 631pay HK$588/ft2 and HK$934/ft2 more for a duplex 632unit and an adjoining property unit, respectively. 633In other words, the findings suggest that poten-634tial buyers are willing to pay a larger premium 635for special housing units. For example, one may 636be willing to pay HK$ 588,000 (model 1) and HK$ 637934,000 (model 2) more on a duplex unit and 638Table 6. Comparison of coefficients of special units 639(Duplex and Adjoining)640Coefficient641t-statistic642Sig.643Model 1 (Duplex)644588.5406452.4086460.016647Model 2 (Adjoining)648934.0796492.5136500.012651