Hedonic Valuation of the Spatial Competition for Urban Circumstance
Utilities: Case Wuhan, China
Bin Zheng Lina Huang Yaolin Liu* Sch Resour Environ Sci Sch Resour Environ Sci Sch Resour Environ Sci Wuhan Univ Wuhan Univ Wuhan Univ Wuhan, P. R. China 430079 Wuhan, P. R. China 430079 Wuhan, P. R. China 430079
Abstract
It has generally accepted Alonso’s theory about the allocation of different land uses
of commerce, resident and industry in urban area. A bunch of researches have provided their aspects of the theme of the relationships between urban circumstances and urban land uses in either the influence of one or several designate circumstance factors on different land uses, or the comprehensive analysis of the influence of all kinds of circumstance on one selected land usage (e.g. residential use). There is still not a wholly analysis about the influence of all kinds of spatial characteristics, available for the location selection of different land uses. That’s why this research selects to engage in a study on the difference among “consumer preferences” to the location amenities in the city. Here we regard the behavior as “spatial competition of the locations”. Hedonic regression model (HRM) analysis is employed as the basic framework of the research. Tabular comparison of HRM parameters performed with principal components analysis (PCA) and Geographic Information Science (GIS) provides all necessary numerical investigation and spatial analysis until to the finally results. The research can be helpful for putting forward to a further integrated investigation on the relationship between urban circumstance and real land use values. Keywords: hedonic regression method; urban land value; circumstance utility; locational characteristics; principal components analysis; GIS
1 Introduction
It has come to common ground that urban land is a multi-attribute utility comprising diverse, heterogeneous characteristics [1-3]. As a major component of the property value, its most significant characteristics are neighborhood characteristics and spatial fixity [1, 4-6]. More specifically, each unit of land has its unique bundle of attributes including accessibility to central business district (CBD), nearby transport facilities supplies, environment amenities, and its neighboring properties, etc. All these characteristics to a large extent relate to the location of the plot. The activities above the land should concern the best location to get the best utilities. Since the activities for different land use purposes have different requirements to these characteristics and have different ability of taking advantage of the location, the will of adopting for use for the same plot of land will be different for different activities, which behaves as the diversification of bidding will in the market condition. We defined that as the spatial choice behavior of different urban land uses. This notion is the key to understanding the innate nature of urban land utility (or use value) and the market price it commands. In empirical applications, hedonic regression method (HRM) is perhaps the most important model to be involved in the researches concerning all kinds of characteristics of the property or land and its price. As mentioned above, the utility of land can be looked as the composite of a bunch of non-numeraire goods such as environment, centricity, landscape, etc. In the HRM, the total utility function he can
12nachieve is described as in a kind of the form of [7-9] and the potential land user in
auction would like to pay for every unit of land with particular value of n attributes or characteristics
U(z,z,K,z,y)-1-
http://www.paper.edu.cn
z∈Rn. In addition to the properties of land z, where y is all other goods consumed. By means of
HRM the implicit prices of every sub utility is revealed from observed prices and the specific amounts of characteristics associated with them, to which Bockstael and McConnel [10] called “Pure Willingness to Pay” in their manuscript.
Among the researches concerning the influence of location characteristics on property value, location to the central business distribute (CBD) or commercial places [11-14], the supply of public services [15-17], transporting improvement [12, 18], environment amenity [6, 19-22] and neighborhoods [4, 5, 23] are the most popular topics. Some others have paid their emphasis on some special attributes such as the influence of neighboring group home [24], or mix land use property [25, 26], etc. These researches have well described the influences of all kinds of location characteristics on property value. But in these investigations the indices of the HRM models were carefully selected and each of them can only figure out the influences of several characteristics. In the papers of Wen [27] and Zheng [28] the property value was investigated as a multiple criteria determined result and more characteristics are involved in but only residential properties were considered in their researches. In the earlier studies of urban land value concerning the difference of land use [3, 29, 30], it always attempted to figure out how the land value diversified in spatial domain or the temporal change of the land value, while the whys - the influences of all kinds of location characteristics, except for the location to CBD or some kinds of “centricity”, were always out of the considerations. For the current, there is still not a wholly HRM analysis available for different land uses. So the question of how different urban land uses (e.g. resident, commerce and industry[31, 32]) coexist in the same urban region and compete for the most suitable location through the free market whereas without disobeying the heterogeneous characteristics of urban land will be the main topic of this paper.
This paper sketches a model-analysis of the differentiation of bidding will for urban circumstances among different land uses, based on the hedonic hypothesis that urban land are valued for their utility-bearing attributes or characteristics for different use purposes. The market price of land and the location characteristics are employed in HRM models for each individual land use, and the contributions of the variables in deciding the bidding price are calculated and compared. Since the considerations in bidding for a plot in real situation are sophisticated and each of them can always behave as a combined thought or balance of different location characteristics, in this research, we employ a technique of principal components analysis (PCA) to simulate the decision process. With PCA applied in HRM, the major contributes of different location characteristics in such thoughts or balances are deduced and be explained as the location competition among different land uses under the bidding market considerations.
2 Methodology
2.1 The data
The hedonic models are estimated using the data of land market of Wuhan city during the period from 1992 to 2000 when it was experiencing a period from chaos to transparent in the marketlization process of urban land in China [33, 34]. The data come from Wuhan bureau of land administration. With careful adjustment and eliminating for any mispricing [35], total 912 plots of land sell record are employed in the research, they are tabular summarized in Tab. 1.
Table 1 Case summaries of the samples
Landuse
raw price N Mean Minimum Maximum Std. Error of
Mean
Mean Minimum
Commercial use
274335.583663.9710593.850.577625.28624.16
Industrial use
113307.7649
61.551218.418.947465.52784.12
Residential use
Total
525 912400.6112 369.570459.77 59.772784.65 10593.812.08307 16.900685.8393 5.63454.09 4.09logarithmic price -2-
http://www.paper.edu.cn
Maximum Std. Error of
Mean
9.270.04585
7.110.06125
7.93 9.270.02316 0.02218In this research, we adopt a geographic information system (GIS) address matching procedure. All parcels are geocoded. GIS is then used to calculate distances from every sold plot to the addressed factors or assign their belonging to the right area in digital maps. Then these distances are employed in the regressions.
2.2 Variables specification and description
Table 2 variable description for hedonic models
Variable Description of the variables Variance (1) Normal location indices in logarithmic form LN_D_CENT Distance to the city center (in the conception) Logarithmic distance LN_D_SCEN Distance to the sub center (in the conception) Logarithmic distance LN_D_COLL Distance to colleges Logarithmic distance LN_D_ROAD Distance to road Logarithmic distance LN_D_EMPO Distance to emporiums Logarithmic distance LN_D_HOSP Distance to hospitals Logarithmic distance LN_D_GYMN Distance to gymnasia Logarithmic distance LN_D_LIBR Distance to libraries Logarithmic distance LN_D_TELE Distance to telecommunication service store Logarithmic distance LN_D_POST Distance to post office Logarithmic distance LN_D_BUSS Distance to the nearest bus stop Logarithmic distance LN_D_MSCH Distance to middle schools Logarithmic distance LN_D_PSCH Distance to primary schools Logarithmic distance LN_D_NSCH Distance to nurse schools Logarithmic distance LN_D_COST Distance to coach stations Logarithmic distance LN_D_RAST Distance to railway stations Logarithmic distance LN_D_DOCK Distance to docks Logarithmic distance
(2) Limited location indices in direct distance indicate the binary attributes of distance and catchment L_D_PARK Distance to green area(In the catchment of 2 kilometers) distance L_D_WATE Distance to the waters(In the catchment of 1 kilometers) distance
(3) Numerical indices
N_POPU Population density Variable (person/km2)
(4) Graded dumb variables indicate the classification
0 to 2, 2 means the best condition
G_APOL Air pollution situation
without pollution
0 to 4, 4 means the best condition
G_NPOL Noise pollution situation
without pollution
G_TERR Terrain condition 0 to 2, 2 means the best G_GEOL Geology condition 0 to 3, 3 means the best
(5) Binary dumb variables of (0/1 set) indicate the belongingness to the catchment C_PA_300 In the direct catchment of parks (inside 300 meters) 1is in and 0 means not C_WA_300 In the direct catchment of waters(inside 300 meters) 1is in and 0 means not C_MR_500 In the direct catchment of main road(inside 500 meters) 1is in and 0 means not C_RO_30 In the direct catchment of road(inside 30 meters) 1is in and 0 means not C_RA_500 In the direct catchment of railway (inside 500 meters) 1is in and 0 means not
2.3 The hedonic price regression and principal component analysis (PCA) To explore the different desirability of a location quality among various land uses we use the equilibrium price function of hedonic price models including both spatial-associated variable and individual characteristics of each section’s demand. In this paper we adopt a semi-log hedonic
-3-
http://www.paper.edu.cn
regression model [22, 25, 36] including quasilinear regressor and dumb variables in the description of log regressand. The general form of the equation is as following: Where
log(pi)=ξ0+αT×Li+βT×Ai+γT×Di+μi (6)
pi is the bid price for the ith land unit, ξ0 is the constant in the regression, α, β, γ are the
nβ∈RmlLAα∈Rλ∈R, , ; i is the logarithmic regressor vector, i is the parameter vectors with
nDnormal regressor vector and i is dumb variable vector, here it also satisfies the form of α∈R,
2
β∈Rm, λ∈Rl; T is the transposition function, and μi~N(0,σ).
To settle the problem of inter-correlation among variables of the models, a principal component analysis was performed. Principal components analysis is a well-established technique used to transform data in the input multi-dimensional attribute space to a new multivariate attribute space of lower dimension to eliminate data redundancy [37-39]. Independent compressed components can be produced by PCA and used for hedonic price regression.
3 Result and discussion
3.1 PCA transform and variable contribution estimate for hedonic models In PCA, the normalized n×pdata matrixXof observed indices is looked as a space with n vectors
xi(i∈n). Where xi is composed with p dimensions for each. In geometry it is possible to
transform this space into a collection of new orthogonal intersected vectors without any information loss. By this mean the problem of self-correlation is settled [39]. Since each new vector i contributes to (or contains) a number of the information of theX. When most of the information of the space can be expressed with less Primary Components (PC), the improvement of the calculating process of HRM is achieved.
In this research, it regards that when the cumulative percentage for the contributing observations reaches 85%, the result can be accepted. The contributions of the PCs are tabular listed in the following table (Tab. 3(A)). The contributions of the original variables to these PCs are also accumulated (Tab. 3(B)).
Table 3 Statistic of the contributing of the Primary Components (PC) and variables
(A) Contributing of PCs (%)
PCs Average Commercial Residential Industrial 1 31.364 2 9.080 3 7.223 4 5.835 5 4.910 6 4.557 7 3.887 8 3.632 9 3.584 10 3.108 11 2.410 12 2.304 13 1.927 15
(85.64)
(B) Contributing of the Variables
Variables Average Commercial Residential Industrial
0.281 0.029 -2.491
2.504 2.167 1.560 8.393 5.519 6.458 2.328 4.262 -6.110 2.829 -2.872 6.814 2.569 -1.924 5.055 4.759 0.869 3.918 -1.709 0.007 1.030 -6.163 1.456 -4.037 0.294 -0.608 -1.750 3.341 1.670 -2.902 0.007 3.386 -8.619 6.224 8.132 10.827 9.129
7.393 13.697
9.904 9.190 10.375
z(i∈n)
24.103 23.630 43.068 N_POPU -2.7109.770 9.864 9.535 L_D_WATE -0.8529.235 7.157 6.676 L_D_PARK 7.2346.505 6.924 6.463 G_TREE 2.1775.799 5.463 5.439 G_NPOL -1.0804.734 4.850 4.508 G_GEOG -1.1704.382 4.206 3.651 G_APOL 1.6623.782 4.132 3.069 C_WA_300 1.9773.239 3.745 2.765 C_PA_300 -1.7992.899 3.306 (85.17) C_MR_500 2.4842.763 2.979 C_RA_500 -1.1152.549 2.598 C_RO_30 3.2072.409 2.348 LN_D_CENT 10.9072.172
LN_D_SCEN 8.991LN_D_COLL 8.668
2.005 2.031
14 1.820 2.160
-4-
http://www.paper.edu.cn
16 LN_D_ROAD-0.6654.106 -4.153 5.459 (86.33) (85.40)
17 LN_D_EMPO9.9518.498 10.252 12.711 18 LN_D_HOSP 8.6915.431 5.960 11.300 19 LN_D_GYMN9.1006.619 6.642 10.908 20 LN_D_LIBR 9.9075.659 7.352 11.499 21 LN_D_TELE 12.2997.678 8.955 11.371 22 LN_D_POST 1.534-0.819 1.332 5.364 23 LN_D_BUSS 7.6198.529 7.538 11.874 24 LN_D_MSCH5.9446.412 2.917 9.592 25 LN_D_PSCH 6.5564.196 6.904 8.321 26 LN_D_NSCH8.3008.484 7.677 10.467 27 LN_D_COST 11.5247.665 8.316 11.345 28 LN_D_RAST 6.9155.743 5.522 12.308 29 LN_D_DOCK0.7110.257 -0.527 1.215
The accumulation of the distributions of each variable can also be regarded as the average willingness in different sections to pay for that particular location amenity. As mentioned in the former analysis, when different potential users exist, the total willingness to pay can be expressed as the highest bidding price one can afford in competitions. Thus the contributions listed in Tab. 3(B) can be looked as the result of the competition on location characteristics among different landuse sections. The more accumulation the variable contribute to the ultimate land price, the more the land users from the section care about the characteristics represented by that variable. Measure such contributes from Tab. 3(B) in histogram chart makes the competition more significant (Fig. 1).
3.00Contribution of the Variables2.001.000.00average-1.00commercialresidentialindustrial-2.00ln_d_collg_npolg_apolln_d_librln_d_centln_d_postln_d_empoln_d_gymnln_d_costln_d_rastl_d_parkln_d_teleln_d_dockn_populn_d_mschln_d_roadg_geogln_d_scenln_d_hospln_d_bussc_mr_500g_treec_pa_300ln_d_pschc_wa_300ln_d_nschc_ra_500l_d_watec_ro_30 Fig. 1 Chart compare of the competition of location amenities
In Fig. 1, the columns represent the contributions of the variables and being near to the zero axis means
less sensitive to the location characteristic declared by the variable, vice versa. Form which we can roughly put forward the following discussions on the investigation:
Discussion 1: Industrial section looks more sensitive to location characteristic because the variations of its linear offset are the most obvious. But its location selection does not necessary be right near to the
-5-
http://www.paper.edu.cn
environmental amenities. Besides neighboring to roads or be with a good terrain condition also show none positive relation to their needs.
Discussion 2: The land users from residential section behave not as interested in the distance to road as those from commercial section do. While the figure also shows that the location immediate neighboring to road is still regarded to be deserved to compete for residential section. That can be reasoned by the commercial value of the first floor of the roadside residential buildings.
Discussion 3: The distance to the city center is one of the most important factors contributing to the land price bidding for all three kinds of land uses.
Discussion 4: Besides the distance to post office (be not sensitive to commercial section), distance to road (be near to does necessary means high land value for residential section), all the distance degressive variables are competed, no matter its logarithmic distance or directed distance degressive. Population density, grad of terrain, geology, pollution, and being immediate catchment of main road, railway and waters appears less considerate in competition. 3.2 Description of the most significant variables
It is observed that not all variables manifest significant contribution in description the PCs (Tab. 3(B)). That can be explained by the existence of redundancy information of the original data. Though the PCs are orthogonal intersected to promise non-linear and have reduced most of the redundancy by eliminating some insignificant Components, the redundancy is still be kept since all variables attend the description of PCs. In the description of PCs, those variables contribute less significant are regarded as carrying inefficient information and such redundancy should not be counted in that process. In the following table (Tab. 4) we keep the most significant signals by setting a strict criterion that only the variable with more than 75% of the maximum contribution to one PC is allowed to left in PC’s description. The accumulation of the number of PCs each variable contributing to (Tab. 4(A)) and the contributions (weighted by the contribution of PCs) of every variables in absolute value (Tab. 4(B)) are tabular listed in Tab. 4.
Table 4 Statistic of the contribution of the major effective variables for PCs
Variables
(A) Statistic in Count
All cases Commercial Residential
1 2 1 2 1 1 3 4 3 2 3 3 1 1 1 3 0 1 1 1 1 2
Industrial
All cases
1 3 0 3 2 1 5 2 2 3 3 3 2 1 2 2 2 1 1 1 1 3
(B) Statistic in Values (absolute) Commercial Residential Industrial
1 0.000 6.098 5.616 1.490 1 2.769 5.270 7.408 3.068 1 6.755 3.315 0.000 3.011 1 1.675 3.917 6.231 3.469 2 4.295 1.465 5.321 5.512 1 3.006 1.506 3.201 3.016 2 4.584 4.711 6.455 5.190 1 3.179 8.862 5.300 2.698 0 3.385 5.407 2.749 0.000 1 3.605 3.706 4.458 2.393 3 3.678 4.932 3.918 5.304 1 2.751 3.341 5.002 2.223 1 9.915 6.408 9.247 11.074 1 7.948 1.610 6.468 10.810 1 8.488 3.379 3.809 11.045 1 2.936 4.141 3.683 2.540 1 8.398 0.000 6.905 11.218 1 7.556 0.965 1.183 10.854 1 8.155 6.508 6.120 11.016 1 8.544 7.147 7.194 10.826 1 7.435 6.013 6.069 9.620 1 3.554 4.756 3.830 3.573
N_POPU 0 L_D_WATE 1 L_D_PARK 1 G_TREE 1 G_NPOL 2 G_GEOG 1 G_APOL 3 C_WA_300 1 C_PA_300 2 C_MR_500 2 C_RA_500 2 C_RO_30 2 LN_D_CENT 2 LN_D_SCEN 1 LN_D_COLL 2 LN_D_ROAD 2 LN_D_EMPO 2 LN_D_HOSP 1 LN_D_GYMN 1 LN_D_LIBR 1 LN_D_TELE 1 LN_D_POST 2 -6-
http://www.paper.edu.cn
LN_D_BUSS 2 LN_D_MSCH 2 LN_D_PSCH 1 LN_D_NSCH 1 LN_D_COST 2 LN_D_RAST 1 LN_D_DOCK 1 2 2 2 1 3 2 2
2 1 1 1 2 2 2
1 1 1 1 1 1 1 7.883 2.333 8.126 7.409 7.403 2.468 8.240 6.093 9.865 10.507 2.009 4.266 2.953 8.998
2.189 10.595 1.072 9.705 5.650 9.734 5.759 11.494 9.180 10.817 4.725 10.542 7.930 3.110
In Tab. 4(B) the absolute transform is used to eliminate accumulative counteract in statistics. Then weighted accumulation is employed to represent the total importance of the variables. With the selective variable collection, it makes the comparison appears more sharp and information redundancy is avoided at the expense of giving up some less important information. The statistic of Tab. 4(B) is also expressed in chart (Fig. 2).
12.00averagecommercialresidential10.00industrialContribution of the Variables8.006.004.002.000.00ln_d_collg_npolg_apolln_d_librln_d_centln_d_postln_d_empoln_d_costln_d_rastl_d_parkln_d_teleln_d_gymnln_d_dockn_populn_d_mschln_d_roadg_geogln_d_hospln_d_bussc_mr_500g_treeln_d_pschc_wa_300ln_d_scenln_d_nschc_pa_300c_ra_500l_d_watec_ro_30 Fig. 2 Chart compare of the competition of the most effective location amenities
The competition effect in Fig. 2 is magnified comparing to Fig. 1. From this chart we can come to
following discussions:
Discussion 5: Industrial section appears the most sensitive to location since the figure shows that the location variables of industrial section provide the most contributions. Residential section ranges the second and commercial section ranges the third.
Discussion 6: residential section and commercial section have similar appearance of the needs on some location characteristics, for example, the distance to hospital, the distance to libraries, and the distance to telecommunication service stores, etc.
An abbreviated investigation on the relationship of the counts statistic and value statistic of the accumulative contributions in Tab. 4 is performed.
Table 5 Summary of the behaviors among different landuse sections
behavior All cases Commercial Residential Industrial Total amenity demands 43 52 55 32 Accumulated cost 159.090 135.531 146.670 195.946 Average competition cost 3.700 2.606 2.667 6.123
-7-
http://www.paper.edu.cn
We accumulate the statistical counts in Tab. 4(A) for each section and list the result in the first row of Tab. 5 as the total amenity demands of different sections. In Tab. 4 only the most significant variables contributing to PCs is counted in. Whenever one more time a variable participates in the description of a PC, the person who buy the land could have express a kind of location amenity demand. The demand of the amenity for the same variable can repeated, because the amenities of a location characteristic could be more than one (e.g. be near to park can mean better view, better air condition and a nice place for recreation, etc. ). In the same way the statistical values in Tab. 4(B) is accumulated to stand for the total competition cost of different sections. The accumulate value means that when a standard plot with one unit of the location characteristics for each, how much the people in different section would like to afford in monetary. Then we can get the average competition cost by accumulated cost divide total demands. Generally, less average competition cost, more competition ability one possesses. Meanwhile, less amenity demands one has, more choices one possesses.
Then from Tab. 5 we can put forward following four topics for discussion:
Discussion 7: industrial section has the least demands and residential section has the most demands if only concerning those significant spatial factors.
Discussion 8: industrial section has to pay for more for the location goods, which means that the industrial section have to use less competition location to lessen the pressure of high cost. While commercial section needs to pay for the least for the same location characteristics.
Discussion 9: industrial section has to pay for more than double the amount of the average cost for the same unit of location goods, and commercial section is slight privilege to residential section. Discussion 10: when treating with all samples, the figure out of the average competition ability provides no help for distinguish the difference between commercial section and residential section. 3.3 OLS regression of the Z-scored/adjusted values of observed samples With the PCA transform, we can get a set of adjusted value from the original observed factors; which are the source for former analysis of PC matrixZ. As mentioned above, the Z-scored value set has settled the problem of self-correlation and is ready for OLS analysis.
In Tab.6 there are detail list of all available variable can apply to OLS models analysis. And in each models, the list follows the sequence of from the positive most significant one to negative most significant one. The sequence will help to better understanding the different selective tendencies in spatial competition.
Table 6 A sorted indicate list for the hedonic regression models
All cases
Commercial
Residential Industrial (positive) (positive) (positive) (positive) LN_D_ROAD N_POPU G_TREE LN_D_RAST G_TREE LN_D_GYMN C_PA_300 LN_D_CENT LN_D_COST LN_D_COST LN_D_SCEN LN_D_LIBR LN_D_SCEN LN_D_ROAD LN_D_HOSP G_TREE C_PA_300 C_PA_300 LN_D_CENT L_D_WATE N_POPU
(negative)
LN_D_ROAD C_RA_500 LN_D_NSCH
C_RA_500 C_RO_30 N_POPU C_WA_300 LN_D_TELE LN_D_DOCK (negative) (negative)
LN_D_CENT LN_D_EMPO LN_D_COLL LN_D_CENT LN_D_LIBR (negative)
LN_D_EMPO LN_D_ROAD C_WA_300 LN_D_SCEN LN_D_LIBR L_D_PARK LN_D_MSCH C_PA_300 -8-
http://www.paper.edu.cn
C_RO_30 LN_D_GYMN N_POPU LN_D_COST
3.4 Visual map comparison on the hypotheses
A visual comparison can provide immediate view and understandable prove for the analysis, that always makes the result more confident. ArcGIS software makes it feasible to process a spatial comparison by treating all the sample plots.
In the formal discussion, it concluded that the residential section holds the competition ability between commercial section and industrial section. Thus it is proposed to use residential plots to develop the reference map for comparison. Besides, the sample statistic also indicates that residential plots take 57.56% amount of the total samples. Generally more sample plots can create smoother baseline surface for improving the precision of the comparison.
It is generally believed that the urban land value behavior as homogeneity zones and thiessen polygon can provide better simulation for that [28]. In Fig. 3 the baseline surface is deduced from residential sample plots with cost allocation function provided by ArcGIS software. The cost matrix is worked out as former description. The dark orange color means high land value in the residential use. The residential plots are expressed as green squares and industrial plots are market as blue triangles.
Fig. 3 Locations (commercial & industrial) vs. Surface (residential)
It is observed that commercial section tend to occupy the center location in the urban area. On the contrary, industrial section prefers the outer locations where the location good are less competed as mentioned in the formed analysis. While for residential section itself, it is observed that the plots spread in the whole urban area and location in the center always have the higher land value. The phenomena match the discussion 7 that industrial has much less demand for the location amenities.
-9-
http://www.paper.edu.cn
Besides, it is also observed that the commercial section tend to be more “aggressive” in the high land value locations or flourishing locations alongside the commercial centers. In these center places there exists obvious competition between commercial section and residential section. On the contrary industrial section appears less competition ability. That just well proves the discussion 8 and 9. The following summary table provides the statistic prove for the spatial comparison discussion (Tab. 7):
Table 7 a summary of the map comparison
classify Ⅰ Ⅱ Ⅲ Ⅳ Ⅴ
mean value
N
(residential) commercial industrial Total 2078 10 0 10 1462 11 0 11 702 47 2 49 418 80 27 107 222 118 78 196
266
107 373 Total 428 In Tab. 7, the whole urban area is classified into five classes according to the natural break of
residential land values. In each classified sub area, the numbers of cases from commercial section and industrial section are accumulated and listed in the table. It clearly indicates that in the three highest residential land value areas there are almost no industrial plots except for the only two in the third class.
4 Conclusions
It has generally accepted Alonso’s [40] theory about the allocation of different land uses of commerce, resident and industry in urban area. The majority of the early studies showed their interests in the relationship between land uses and the centricity in the cities or road accessibility in the urban area [41-43] while few other location amenities were investigated. When the influence of various location amenities stepped in the attention of numerous of researches, the dissimilarities among different land use sections are then ignored. This situation brought difficult in Imperfect land market.
On the data of Wuhan city, we develop the HRM models with selective 29 variables. Then PCA approach is introduced for better explanation for the hedonic models and GIS application is involved for distance measurement. By means of figure comparison of the results of the model calculation, we put forward ten discussions for the investigation. Among them discussion 1 to 4 provide a summary of the difference among land use sections on their behavior concerning the variables of location characteristics. Then in discussion 5 and discussion 6, we regard that industrial section appears more location sensitivity, residential section and commercial section seem to have similar appearance with residential use to be little more sensitivity. In discussion 7 to 10 we attempted to provide the explanations on those differences. With these ten discussions on observed results, especially with the later four discussions, it is not difficult to conclude that the competition among different land use sections finally led to the choice of the location amenities, which match the hypothesis prompt in the former investigations. Tabular analysis also led to a result summary of the OLS models. Later a visual comparison in GIS provides confident evidence for discussion 7 to 10, which further support the other discussions.
In this research, it considered that the result from PCA approach can provide more sufficient explains on the spatial competition behaviors and can be of more significant in description than ordinary OLS analysis. The answer may lie in that OLS regression depends on a constant value. When this constant is too large, the bias tends to make the result become very difficult to explain. There is no bother in PCA approach necessary to handle this problem. On the other side, PCA approach is an “information-loss” process. Although the result from PCA analysis can provide the ideal matches with either the theory formula deducing result or visual comparison in GIS, there is still no enough evidence to answer the two information losses: one is the 15% information loss happened in retrieving the PCs (Tab. 4); the other one happens when retrieving major effective variables for PCs (Tab. 5). Whether the “principal components of the principal components” is sufficient to handle the whole hedonic price regression still needs demonstration. Besides, we also found that PCA may not properly handle the dumb variables in discussion 4. In the further research, more experiments should be processed to make confident explain.
-10-
http://www.paper.edu.cn
References
[1] Paul F. Wendt (1957) Theory of Urban Land Values, Land Economics, 33(3): pp. 228-240. [2] M. A. Qadeer (1981) The Nature of Urban Land, American Journal of Economics and Sociology, 40(2): pp. 165-182.
[3] Peter F. Colwell & Henry J. Munneke (1997) The Structure of Urban Land Prices, Journal of Urban Economics, 41(3): pp. 321-336.
[4] Eddie C.M. Hui, et al. (2007) Measuring the Neighboring and Environmental Effects on Residential Property Value: Using Spatial Weighting Matrix, Building and Environment, 42(6): pp. 2333-2343. [5] Raymond Y. C. Tse (2002) Estimating Neighbourhood Effects in House Prices: Towards a New Hedonic Model Approach, Urban Studies, 39(7): pp. 1165-1180.
[6] C.Y. Jim & Wendy Y. Chen (2007) Consumption Preferences and Environmental Externalities: A Hedonic Analysis of the Housing Market in Guangzhou, Geoforum, 38(2): pp. 414-431.
[7] Sherwin Rosen (1974) Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition, Journal of Political Economy, 82(1): pp. 34-55.
[8] J. Walter Milon, et al. (1984) Hedonic Amenity Valuation and Functional Form Specification, Land Economics, 60(4): pp. 378-387.
[9] James N. Brown & Harvey S. Rosen (1982) On the Estimation of Structural Hedonic Price Models, Econometrica, 50(3): pp. 765-768.
[10] Nancy E. Bockstael & Kenneth E. McConnell (2007) Hedonic Models of Heterogeneous Goods: A Theoretical Guide to Empirical Models, in Nancy E. Bockstael and Kenneth E. McConnell (eds.), Environmental and Resource Valuation with Revealed Preferences, Springer Netherlands: Dordrecht, The Netherlands. pp. 151-187.
[11] Yan Song & Jungyul Sohn (2007) Valuing Spatial Accessibility to Retailing: A Case Study of the Aingle Family Housing Market in Hillsboro, Oregon, Journal of Retailing and Consumer Services, 14(4): pp. 279-288.
[12] Greg T. Smersh & Marc T. Smith (2000) Accessibility Changes and Urban House Price Appreciation Journal of Housing Economics, 9(3): pp. 187-196.
[13] Cheng Ri Ding (2001) The Benchmark Land Price System and Urban Land Use Efficiency in China, Chinese Geographical Science, 11(4): pp. 306-314.
[14] Michael S. Johnson; Wade R. Ragas (1987) Cbd Land Values and Multiple Externalities, Land Economics, 63(4): pp. 337-347.
[15] Hoong Chor Chin & Kok Wai Foong (2006) Influence of School Accessibility on Housing Values, Journal of Urban Planning and Development-ASCE, 132(3): pp. 120-129.
[16] Randall Reback (2005) House Prices and the Provision of Local Public Services, Journal of Urban Economics, 57(2): pp. 275-301.
[17] Elizabeth A. Wilman (1988) Pricing Policies for Outdoor Recreation, Land Economics, 64(3): pp. 234-241.
[18] C. Y. Yiu & S. K. Wong (2005) The Effects of Expected Transport Improvements on Housing Prices, Urban Studies, 42(1): pp. 113-125.
[19] C. Y. Jim & Wendy Y. Chen (2006) Impacts of Urban Environmental Elements on Residential Housing Prices in Guangzhou (China), Landscape and Urban Planning, 78(4): pp. 422-434.
[20] Liisa Tyrväinen & Antti Miettinen (2000) Property Prices and Urban Forest Amenities, Journal of Environmental Economics and Management, 39(2): pp. 205-223.
[21] Liisa Tyrväinen (1997) The Amenity Value of the Urban Forest, Landscape and Urban Planning, 37(3-4): pp. 211-222.
-11-
http://www.paper.edu.cn
[22] Fanhua Kong, et al. (2007) Using Gis and Landscape Metrics in the Hedonic Price Modeling of the Amenity Value of Urban Green Space: A Case Study in Jinan City, China, Landscape and Urban Planning, 79(3-4): pp. 240-252.
[23] George Galster, et al. (2004) Supportive Housing and Neighborhood Property Value Externalities, Land Economics, 80(1): pp. 33-54.
[24] Peter F. Colwell, et al. (2000) The Effect of Group Homes on Neighborhood Property Values, Land Economics, 76(4): pp. 615-637.
[25] Yan Song & Gerrit-Jan Knaap (2004) Measuring the Effects of Mixed Land Uses on Housing Values, Regional Science and Urban Economics, 34(6): pp. 663-680.
[26] John W. Matthews & Geoffrey K. Turnbull (2007) Neighborhood Street Layout and Property Value: The Interaction of Accessibility and Land Use Mix Journal of Real Estate Finance and Economics, 35(2): pp. 111-141.
[27] Hai Zhen Wen, et al. (2005) Hedonic Price Analysis of Urban Housing: An Empirical Research on Hangzhou, China, Journal of Zhejiang University SCIENCE, 6A(8): pp. 907-914.
[28] Bin Zheng, et al. How Does Micro Circumstances Influence the Urban Residential Land Value - a Study Case Danyang, China, In Wei Min Ju and Shu He Zhao (eds.), Geoinformatics 2007: Remotely Sensed Data and Information, vol. 6752 pp. 67522W: SPIE.
[29] David Dale-Johnson & W. Jan Brzeski (2001) Land Value Functions and Land Price Indexes in Cracow, 1993–1999, Journal of Housing Economics, 10(3): pp. 307-334.
[30] Michael T. Rancich (1970) Land Value Changes in an Area Undergoing Urbanization, Land Economics, 46(1): pp. 32-40.
[31] GB137-90 (1991) Standard for Classification of Urban Land and for Planning of Constructional Land (in Cn), Issued by: M. C. P. R. C., China Planning Press, Beijing, China.
[32] CJJ46-91 (1992) Classification Code for Usage of Municipal Terra (in Cn), Issued by: M. C. P. R. C., China Architecture & Building Press, Beijing.
[33] Qingshu Xie, et al. (2002) The Emergence of the Urban Land Market in China: Evolution, Constraints and Perspectives, Urban Studies, 39(8): pp. 1375-1398.
[34] Fulong Wu (2001) China's Recent Urban Development in the Process of Land and Housing Marketisation and Economic Globalisation, Habitat International, 25(3): pp. 273-289.
[35] GB/T 18508-2001 (2002) Regulations for Valuation on Urban Land (in Cn), Issued by: G. A. Q. S. I. Q. P. R. C., Standards Press of China, Beijing, China.
[36] Jacqueline Geoghegan (2002) The Value of Open Spaces in Residential Land Use, Land Use Policy, 19(1): pp. 91-98.
[37] Xia Li & Gar-On Yeh (2002) Integration of Principal Components Analysis and Cellular Automata for Spatial Decisionmaking and Urban Simulation, Science in China Ser. D Earth Sciences, 45(6): pp. 521-529.
[38] Jacques Bénasséni (2005) A Concentration Study of Principal Components, Journal of Applied Statistics, 32(9): pp. 947-957.
[39] William F. Massy (1965) Principal Components Regression in Exploratory Statistical Research, Journal of the American Statistical Association, 60(309): pp. 234-256.
[40] William Alonso (1964) Location and Land Use: Toward a General Theory of Land Rent, Harvard University press, Cambridge, Massachusetts.
[41] F. Stuart Chapin (1965) Urban Land Use Planning, University of Illinois Press, Urbana. [42] J.N. Jackson(eds.). (1972) The Urban Future: A Choice between Alternatives. In (eds.), Urban and Regional Studies. Vol. 3, Allen & Unwin: London.
[43] A.W. Evans (1987) Urban Economics: An Introduction, Blackwell, Oxford.
-12-
因篇幅问题不能全部显示，请点此查看更多更全内容