shi-ming yu(a2)

Upload: sdg20

Post on 09-Apr-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/8/2019 Shi-Ming Yu(A2)

    1/21

    7 th ASIAN REAL ESTATE SOCIETY CONFERENCESEOUL, KOREA 4-6 JULY 2002

    A Spatial Analysis of SingaporesIndustrial Property Values

    Shi-Ming Yu and Ann Basuki

    Department of Real EstateSchool of Design & EnvironmentNational University of Singapore

    A b s t r a c t :

    The Jurong Town Corporation (JTC) is a statutory body responsible for the

    industrialization of Singapore since its establishment in 1968. As the largestlandlord of industrial properties, its role includes the planning, development

    and management of some 30 industrial estates located in various parts of

    Singapore. Valuation is an important function to support the allocation,management and redevelopment of the industrial properties, especially when

    economic cycles and industries have increasingly shorter time spans.

    This paper examines the valuation of industrial properties, focusing on the

    spatial distribution of industrial property values. It covers the establishment ofa valuation model for all JTC industrial properties. The model will help

    provide indicative valuation of every industrial property under itsmanagement. The model is also built on GIS platform, which will allow for the

    spatial analysis of industrial property values. This will help to further improve

    on the planning and allocation of industrial uses as well as decision-making inthe redevelopment of certain industrial estates.

    Key words: industrial property values, automated valuation model, GIS, spatial analysis

  • 8/8/2019 Shi-Ming Yu(A2)

    2/21

    1 . I n t r o d u c t i o n

    Extensive studies on factors affecting property value, especially on the importance of location

    due to the spatial immobility of property, have been carried out. Various techniques have

    been used to examine the significance of location on determining property values,

    especially for residential and commercial properties. However, industrial property

    remains a relatively under-researched area in the literature. This study aims to fill this

    gap.

    In Singapore industrial space is provided by both the public and private sector. Public sector

    refers to public authorities that are government bodies such as Jurong Town Corporation

    (JTC) and Housing Development Board (HDB), while private sector refers to companies

    that had built industrial properties for owner-occupation and also industrial property

    developer firms. In terms of floor space, the private sector has 75 per cent of the market

    share in 1999 (Chow, et al., 2002).

    JTC is a statutory body responsible for the industrialization of Singapore since its

    establishment in 1968. As the largest landlord of industrial properties, its role includes the

    planning, development and management of some 30 industrial estates located in various parts

    of Singapore. A wide range of industrial facilities were developed by JTC, which include

    prepared industrial land, standard factory, workshops, flatted factory, Business Parks and

    Science Parks. In this study the focus is on the first three types of industrial property.

    From the planning point of view, the challenge for Singapore is the scarcity of land. Demandfor land will continue to increase as the economy grows and population expands. The 2001

    Concept Plan makes provisions for high value-added industries which will contribute more to

    Singapores economic growth. The plan has set aside sufficient land for industries in the

    future inspite of the land intensive nature of some of these industries. Figure 1 shows the

    allocation of land for industrial uses. These are concentrated mainly in a few outlying areas,

    namely Jurong (western part), Changi (eastern part) and Woodlands (northern part).

    2

  • 8/8/2019 Shi-Ming Yu(A2)

    3/21

    Figure 1. Concept Plan 2001

    Source: Urban Redevelopment Authority

    Valuation is an important function to support the allocation, management and redevelopment

    of the industrial properties, especially when economic cycles and industries have increasingly

    shorter time spans as shown in figure 2. With the primary concerns of accuracy and

    efficiency, automated valuation model has emerged as one way to address these needs.

    Figure 2. Economic Performance

    -10

    -5

    0

    5

    10

    15

    20

    1961

    1964

    1967

    1970

    1973

    1976

    1979

    1982

    1985

    1988

    1991

    1994

    1997

    2000

    GDPGrowth(%)

    Source: Singapore Department of Statistics

    3

  • 8/8/2019 Shi-Ming Yu(A2)

    4/21

    Although there is no perfect valuation method that is able to predict the actual value of a

    property precisely all the time, a good model should be able to estimate values that are as

    close as possible to the actual transacted values as frequently as possible. Statistical analysis

    provides a useful tool in analyzing the importance of physical, location and economic factors

    in value determination.

    This paper examines the valuation of industrial properties, focusing on the spatial distribution

    of industrial property values. It covers the establishment of a valuation model for all JTC

    industrial properties. The model will help provide indicative valuation of every industrial

    property under its management. The model is also built on GIS platform, which will allow

    for the spatial analysis of industrial property values. This will help to further improve on the

    planning and allocation of industrial uses as well as decision-making in the redevelopment of

    certain industrial estates.

    2 . L i t e r a t u r e R e v i e w

    The automated valuation model is one way to address the need for more objective andquantifiable valuations. Computer analysis of data has been widely used as it can greatly

    reduce the time to estimate a value, especially when mass appraisal is needed. The model is

    more thorough and systematic than human assessment. It is a systematic appraisal of groups

    of properties as of a given date using standardized procedures and statistical testing

    (McCluskey et al., 1997; Waller 1999). In the modeling, the aim is to derive a mathematical

    model, which represents supply and demand patterns for groups of properties.

    Accuracy in data collection and collation are essential to ensure that the data assembled is of

    sufficient quality. Compatible data format is equally essential to support the transferring and

    downloading of data from one system to another. The modeling techniques can take a number

    of forms, such as regression analysis, expert systems and neural networks.

    Regression method has been widely used especially in the US (Mark and Goldberg, 1988;

    Murphy, 1989; Ambrose, 1990; Fehribach et al., 1993; Ramsland and Markham, 1998;

    Panayiotou et al., 1999; Isakson, 2001). The common problem with regression analysis is

    4

  • 8/8/2019 Shi-Ming Yu(A2)

    5/21

    multicollinearity between explanatory variables that causes coefficient estimates to be

    unstable.

    Ridge regression technique was introduced as an alternative to multiple linear regressions

    when the data are highly correlated which result in unstable regression coefficient estimates

    (Hoerl and Kennard, 1970; Newell, 1982 and Ferreira and Sirmans, 1988). Ridge regression

    introduces biased estimates for these regression coefficients and therefore has a smaller mean

    square error than the multiple linear regressions. This will result in the regression coefficients

    being closer to their expected values and have the correct signs. To obtain the ridge

    regression solution one must determine how much bias must be introduced to ensure stable

    property characteristics, which can be achieved by examining a graph for each regression

    coefficient for differing extent of bias.

    The potential of expert systems to property valuation has been the subject of a number of

    studies in the past (Scott, 1988 and Nawawi et al., 1997). The systems consist of three

    components: a user interface, an inference engine and the knowledge base. The user interface

    allows the end-user to interact with the system. The inference engine contains the logic and

    reasoning, while the knowledge base contains the relevant data. The decision making process

    of an expert are transcribed into a series of if/then statements in the knowledge base. The

    systems will ask the end-user a series of questions, which the inference engine then uses to

    search the knowledge base that can lead to a conclusion. The potential problems of this

    system include maintenance of the knowledge base to adapt to the changing environment, the

    programming and debugging of the inference engine. Property values are very dynamic;

    therefore the knowledge base will need to be reprogrammed each time external factors cause

    property values to change. The system is time consuming and difficult to extract and

    accurately quantify specific knowledge from experts (Wyatt, 1996).

    Several studies have applied Neural Network methodology for valuation of residential

    property (Borst, 1995; Tay and Ho, 1995; McCluskey, 1996). Neural networks are a

    nonlinear methodology, trained by example to learn relationships from repeatedly presented

    input and output of data sets. It is self-maintaining, dynamic and automatically adapt to

    changes from exposure to new information. The models are conceptually difficult to

    understand and explain. It does not provide a traceable path on how a decision was made.

    Therefore many neural network critics present the black box argument. The proper setting of

    5

  • 8/8/2019 Shi-Ming Yu(A2)

    6/21

    neural networks model takes several iterations to find the set of parameters that best fit in

    application, thus can have a very long run times. As found in many studies, small changes

    can result in very different findings (Allen and Zumwalt, 1994, Worzala et al, 1995). The

    results are inconsistent between neural network packages, even between runs on the same

    software.

    Multiple regression analysis has been demonstrated, as being the primary technique used in

    the mass appraisal. Therefore in this study we will use multiple regressions for the automated

    valuation model. It is basically a hedonic model, attempting to disaggregate value into

    different contributing factors such as physical characteristics of the property. In order to get

    an accurate and effective model, all attributes should be properly accounted for.

    Ambrose (1990) examines the relationship between the asking price and the physical

    characteristics in northeast Atlanta during 1986 and 1987. The results show that the

    buildings size, the office space, the number of dock-high and drive-in doors, the presence of

    railway siding and the ability to build-to-suit office space are significant in determining the

    asking price. A study by Asabere and Huffman (1991) indicates that location, neighbourhood,

    lot size and time have affected the industrial land value in Philadelphia significantly and

    overzoning for industry has caused land with industrial use to have about 58 percent lower

    value compare to commercially zoned land. Fehribach et al. (1993) included financial traits,

    location attributes and market economic conditions as variables to determine the factors that

    influence industrial property prices in Dallas and Tarrant Counties from 1987 to 1991. The

    financial variables include industrial capitalization rate and prime rate. The location variables

    are the distance to major highway or airport and the economic variables include indices on

    employment, consumer price and industrial production. Linear relationship was assumed. The

    results indicate that building size, office space, number of dock-high doors, ceiling height,

    age, prime rate, county, distance to airport, type of tenant, industrial cap rate and the date of

    sale are significant variables in the explanation of the samples sales prices. The latest study

    by Lockwood and Rutherford (1996) indicates that physical characteristics, regional market

    influences and location factors are the primary determinants of industrial property values in

    Dallas/Fort Worth area over the 1987-1991 period. The national market factor and interest

    rate are found to have insignificant influence over industrial property prices.

    6

  • 8/8/2019 Shi-Ming Yu(A2)

    7/21

    A number of researches have applied Geographic Information System (GIS) to property

    valuation (Ishizuka, 1995; Gallimore et al., 1996; Wyatt, 1996; McCluskey et al., 1997; Yu et

    al.,2000 ). The studies have demonstrated the added value of the geographical display and

    locational analysis of property information; various datasets are required to be joined with the

    database in ArcView, which hold additional spatial information.

    3 . R e s e a r c h D e s i g n

    The study covers a period of 9 years from 1993 to 2001. In the early 1990s Singapore

    experienced strong economic growth due partly to high foreign investment (figure 3).

    Economic growth slowed in 1996 followed by a significant decline due to the Asian financial

    crisis in 1997.

    Figure 3. Net Foreign Investment Commitment

    0.0

    1,000.0

    2,000.0

    3,000.0

    4,000.0

    5,000.0

    6,000.0

    7,000.0

    8,000.0

    9,000.0

    10,000.0

    1993 1994 1995 1996 1997 1998 1999 2000

    Source: Singapore Dept. of Statistics

    NetForeignInvestment

    Commitment

    ($million)

    The performance of industrial properties follows the pattern closely. The industrial property

    price index increased steadily from 1993 and reached its peak in the third quarter of 1996

    (figure 4). Subsequently the price index fell steadily to its lowest in the third quarter of

    1999 before moving up slightly in 2000. Due to the heightened uncertainties in the global

    economic environment, the price index fell again almost to the level in the third quarter

    of 1999 in late 2001.

    7

  • 8/8/2019 Shi-Ming Yu(A2)

    8/21

    Figure 4. Industrial Property Performance

    0

    20

    40

    60

    80

    100

    120

    140

    160

    180

    93Q1

    Q4

    Q3

    Q2

    96Q1

    Q4

    Q3

    Q2

    99Q1

    Q4

    Q3

    Q2

    2002Q1

    PriceIndex

    0

    2

    4

    6

    8

    10

    12

    Vacancyrate(%

    )

    Price Index Vacancy

    Source: Urban Redevelopment Authority

    This study covers only workshops, standard factories and purpose-built factories belonging to

    JTC. Workshops have land area of less than 500 square metres, standard factories have land

    area starting from 500 square metres and purpose- built factories starting from 1000 square

    metres. These factories are sold by JTC to industrialists for 30 year or 30 + 30 year leases.

    Under the terms and conditions of the lease, the lessee has the right to reassign after a certain

    period of occupation. The buyer will take over the remaining lease, subject to the same termsand conditions as the original assignment. Excluding the outliers, there are 128 transactions

    for workshops, 360 transactions for purpose-built factories and 493 transactions for standard

    factories. Figure 5 presents the geographical distribution of the transaction data.

    Figure 5. Transaction Data Distribution

    8

  • 8/8/2019 Shi-Ming Yu(A2)

    9/21

    Table 1 lists the summary statistics of the land rates and floor rates for each of the three types

    of the industrial properties. Workshops have the highest mean for both land and floor rates

    while purpose-built factories have the lowest.

    Table 1. Summary of Transaction Data

    Workshop Standard Factory Purpose-built

    Land Rate(psm)

    Floor Rate(psm)

    Land Rate(psm)

    Floor Rate(psm)

    Land Rate(psm)

    Floor Rate(psm)

    Mean 2996.48 2151.74 980.14 1650.81 671.88 1146.09

    Standard Deviation 1290.92 753.99 514.06 578.12 609.09 566.22

    Range 6477.01 3355.57 2787.15 2519.42 4644.82 3366.89

    Minimum 689.66 689.66 162.01 468.05 34.25 190.07

    Maximum 7166.67 4045.23 2949.16 2987.47 4679.07 3556.96

    Secondary source data were obtained from the following government agencies:

    1. Monetary Authority of Singapore: Monthly Prime Rate

    2. Urban Redevelopment Authority: Quarterly Vacancy Rate

    3. Singapore Department of Statistics: Building Material Index, Unit Labour Cost,

    Unemployment Rate, Growth in Industrial production and Consumer Price Index

    (CPI)

    9

  • 8/8/2019 Shi-Ming Yu(A2)

    10/21

    Spatial Distribution Analysis

    Comparisons were made for transaction prices per square metre of the gross floor area (floor

    rate) and per square metre of land area (land rate). A deflation factor was used

    over the period of study to obtain the comparable prices.

    In the analysis, the transactions were grouped based on the industrial estate where the

    property is located. The largest industrial estate, Jurong, is further subdivided into eight areas.

    We use Arc View GIS to perform the spatial analysis, by applying the functions of map

    overlay and data aggregation. The relevant data sets were transferred into SPSS for non-

    parametric tests.

    We intend to capture the dynamics of the spatial distribution of industrial property values

    over time, specifically for the year 1995 and 2000. These periods were chosen since in 1995

    the price index was reaching its peak while in 2000, the opposite happened (figure 4).

    Regression Analysis

    Four groups of variables are included in the analysis. These are physical attributes, location,

    financial and economic variables.

    Physical Attributes

    Land area, building area and building age are included in the regression model. Other

    physical attributes, which have proven to be significant in previous studies, such as office

    space, number of dock-doors and ceiling height, were not included due to the unavailability

    of data.

    Location variable

    The variable estate is included to determine the effect of location on the transaction price. In

    the database only industrial properties in fifteen estates were transacted from 1993 to 2001.

    Since Jurong is the largest estate and most of the transactions were from this estate, we

    subdivided the estate into 8 areas as follow: Jurong East, Jurong Pier, Chin Bee, Joo Koon,

    Benoi, Gul, Pioneer and Tuas. A dummy variable is used to represent the various locations.

    10

  • 8/8/2019 Shi-Ming Yu(A2)

    11/21

    Financial variable

    As most real estate purchases are financed, prime rate is an indication of the cost an investor

    incurs through his acquisition.

    Economic variables

    Several indicators of economic activity were examined. These include GDP growth,

    unemployment rate, growth in industrial production, consumer price index (CPI), building

    material index, labour cost index and vacancy rate. Unit labour cost constitutes almost half of

    the overall weight in the manufacturing business cost index (JTC, 2001)

    Regression Model

    A separate model will be developed for standard factory, purpose-built factory and workshop.

    Based on the curve estimation and as proven in previous studies by Ambrose (1990) and

    Fehribach et al (1993), a linear relationship was assumed to exist between the dependent and

    independent variables.

    The linear equation is specified as follows:

    Property value = f(Land Area, Floor Area, Estates, Age, Effective Year, Prime Rate, GDP

    growth, unemployment rate, growth in industrial production, vacancy rate, Consumer Price

    Index, Building Material Index, Labour Cost index)

    4 . S p a t i a l D i s t r i b u t i o n

    The transacted price of a JTC property reflects the market value of the building subject to

    JTCs land lease conditions at the time of the transaction. A land rent is payable based on the

    initial contract when the industrial property was first assigned by JTC. For subsequent

    transactions or reassignments, the new buyer will continue to pay the contracted land rent. As

    a result, during recession the transacted price will be very close to the market value of the

    building, while during a boom, the price will take into account the profit rent for the land.

    The land rents are divided into two categories for outlying areas and urban and suburban

    estates. The land rents for urban and suburban estates are higher due to the higher permissible

    density of development.

    11

  • 8/8/2019 Shi-Ming Yu(A2)

    12/21

    12

  • 8/8/2019 Shi-Ming Yu(A2)

    13/21

    Figure 6. The distribution of floor rate (psm)

    Spatial Analysis of Industrial Property Values: A Case Study of JTC in Singapore7t h A s i a n R e a l E s t a t e S o c i e t y C o n f e r e n c e , S e o u l , 4 - 6 J u l y 2 0 0 2

    13

  • 8/8/2019 Shi-Ming Yu(A2)

    14/21

    Figure 7. The distribution of land rate (psm)

    Spatial Analysis of Industrial Property Values: A Case Study of JTC in Singapore7t h A s i a n R e a l E s t a t e S o c i e t y C o n f e r e n c e , S e o u l , 4 - 6 J u l y 2 0 0 2

    14

  • 8/8/2019 Shi-Ming Yu(A2)

    15/21

    Spatial Analysis of Industrial Property Values: A Case Study of JTC in Singapore7t h A s i a n R e a l E s t a t e S o c i e t y C o n f e r e n c e , S e o u l , 4 - 6 J u l y 2 0 0 2

    15

  • 8/8/2019 Shi-Ming Yu(A2)

    16/21

    A non-parametric technique, the Kruskal-Wallis test was applied to determine whether there

    was a difference in the average of land rates and floor rates in various estates for the three

    types of industrial properties. Almost all categories confirmed at 1 per cent significant level

    that the values were not distributed uniformly among the estates (Table 2).

    Table 2. Kruskal-Wallis test of Land Rates and Floor Rates

    Type Chi-square Degrees of freedom

    Land Rates (psm) Purpose-built 96.199* 17

    Standard Factory 87.693* 10

    Workshop 49.719* 7

    Floor Rates (psm) Purpose-built 28.638*

    *

    17

    Standard Factory 29.756* 10

    Workshop 24.626* 7* indicates significant at the 1 per cent level** indicates significant at the 5 per cent level

    Figures 6 and 7 present the spatial distribution of industrial property values (floor rate and

    land rate per square metre, respectively) for purpose-built factory, standard factory and

    workshop in 1995 and 2000. The spatial units are grouped based on their standard deviation

    from the respective means. The darker areas represent areas with higher values whereas the

    lighter areas represent those with lower values. The white areas are areas where no

    transactions were registered for that particular type of industrial property.

    The distribution shows that there is no distinct pattern with regard to the effect of location (or

    estate) on industrial property value. While there are different levels of value exhibited in

    different estates, these are likely due to the agglomeration effects of certain industries as well

    as the infrastructure of the estate rather than, say, nearness to transportation hub. This couldbe unique to Singapore given its small size and the well developed transportation network,

    which allows easy accessibility.

    The comparison between 1995 and 2000 also shows no distinct pattern in the distribution of

    values. In fact, the estates with the highest values in 1995 were different from those in 2000.

    This could be attributed to the changing environment as a result of new policies and

    initiatives by the government. Furthermore, given JTCs role in the industrialization of the

    Spatial Analysis of Industrial Property Values: A Case Study of JTC in Singapore7t h A s i a n R e a l E s t a t e S o c i e t y C o n f e r e n c e , S e o u l , 4 - 6 J u l y 2 0 0 2

    16

  • 8/8/2019 Shi-Ming Yu(A2)

    17/21

    country, the distribution of industrial property values is very much subject to policy and

    strategy changes.

    5 . R e g r e s s i o n R e s u l t s

    Table 2 presents the summary of the stepwise multivariate regression results. All the

    variables included in the models are significant at the 5 per cent significant level. The

    independent variables for the standard factory model explained 78 per cent of the variations.

    For the purpose-built factory, the model explained 70 per cent of the variations, while the

    model for workshop explained 68 per cent of the variation.

    For all the models, floor area and effective years were found to be the most significant in

    explaining the variance of the dependent variables. These are to be expected since the

    transacted price reflects closely the building value, which is determined largely by the floor

    area as well as the remaining lease term. The variable land area is more significant for

    purpose-built and standard factories and not for workshops. This is also expected since the

    variation in land area for workshops is much smaller than the other two categories. As with

    the spatial distribution, the regression results show that location, as represented by the

    different estates, is not amongst the most significant independent variables.

    Table 2. Summary of Regression Results.

    Purpose-built Factory

    IndependentVariables

    Adjusted R2 StandardizedCoefficients

    t Sig. VIF

    Floor Area .655 .865 20.623 .000 2.119Effective Year .672 .241 3.682 .000 5.145

    Prime Rate .688 .131 4.522 .000 1.010Land Area .695 -.105 -2.553 .011 2.028

    Age .699 -.150 -2.300 .022 5.137

    Location (Pioneer) .702 -.061 -2.089 .037 1.041Durbin Watson: 1.834

    Standard Factory

    IndependentVariables

    Adjusted R2 StandardizedCoefficients

    t Sig. VIF

    Floor Area .603 .441 9.599 .000 4.653Effective Year .670 .341 12.367 .000 1.675

    Building Material Index .708 .276 9.271 .000 1.958

    CPI .742 .199 6.712 .000 1.945Land Area .767 .383 8.120 .000 4.918

    Spatial Analysis of Industrial Property Values: A Case Study of JTC in Singapore7t h A s i a n R e a l E s t a t e S o c i e t y C o n f e r e n c e , S e o u l , 4 - 6 J u l y 2 0 0 2

    17

  • 8/8/2019 Shi-Ming Yu(A2)

    18/21

    Location (Kamp. Ayer) .770 .094 3.229 .001 1.853Prime Rate .773 .074 2.913 .004 1.424

    Location (Benoi) .776 -.053 -2.487 .013 1.018Location (Senoko) .777 -.047 -2.141 .033 1.066

    Durbin Watson: 2.114

    Workshop

    IndependentVariables

    Adjusted R2 StandardizedCoefficients

    t Sig. VIF

    Floor Area .308 .520 9.601 .000 1.162Effective Year .554 .543 9.700 .000 1.242

    Location (Jurong East) .613 .266 4.644 .000 1.296Building Material Index .652 .345 5.260 .000 1.700

    CPI .679 .217 3.360 .001 1.653Durbin Watson: 1.97

    For the financial variable, prime rate, it is significant for purpose-built and standard factories

    but not so significant for workshops due to the relatively lower prices for the latter.

    For the economic variables, Building Material Index and Consumer Price Index, they are

    significant in the standard factory and workshop models but not significant in the purpose-

    built factory model. This can be expected as the standard factories and workshops are built

    by JTC with similar design and the cost of construction would therefore affect the transacted

    price. On the other hand, purpose-built factories, as the name implies, are built by the

    industrialists for their own specific needs. Potential buyers of such purpose-built properties

    could therefore choose to build or buy when the cost of construction is low.

    Autocorrelation is measured by using the Durbin-Watson statistics. The correlation of each

    residual and the residual for the time period immediately preceding the one of interest was

    measured. For all the three models the value is close to 2, which indicates almost total

    randomness.

    The variance inflation factor (VIF) is calculated for each independent variable to detect

    collinearity amongst the variables. Neter, Wasserman and Kutners rule of thumb associated

    with VIF is that an independent variable with a VIF above 10 would indicate a severe effect

    on the regression model. None of the variables chosen for the models indicated a variance of

    inflation factor above 10. Therefore multicollinearity does not appear to be a concern.

    Spatial Analysis of Industrial Property Values: A Case Study of JTC in Singapore7t h A s i a n R e a l E s t a t e S o c i e t y C o n f e r e n c e , S e o u l , 4 - 6 J u l y 2 0 0 2

    18

  • 8/8/2019 Shi-Ming Yu(A2)

    19/21

    6 . C o n c l u s i o n

    This study has attempted to develop a valuation model for industrial properties owned by

    JTC, a public sector body who is also the largest industrial property landlord in Singapore.

    Transactions of purpose-built factories, standard factories and workshops over a 9 year period

    from 1993 to 2001 were analysed. The regression models for the three types of factory space

    produce an acceptable adjusted R2 of about 70%. The coefficients of the independent

    variables are expected with the most significant being the floor area and the remaining lease

    term. Both economic and financial variables are also significant in some of the models.

    However, the variable for location, which is defined by the different industrial estates, is not

    significant in all the three models. The spatial distribution of the transacted prices, using a

    GIS platform, also does not indicate any significant pattern. This could be unique to

    Singapore, given the small land size and the efficient network of transportation.

    ________________________

    Spatial Analysis of Industrial Property Values: A Case Study of JTC in Singapore7t h A s i a n R e a l E s t a t e S o c i e t y C o n f e r e n c e , S e o u l , 4 - 6 J u l y 2 0 0 2

    19

  • 8/8/2019 Shi-Ming Yu(A2)

    20/21

    R e f e r e n c e s

    1. Allen, W.C. and Zumwalt, J.K (1994) Neural Networks: A Word of Caution, unpublished working

    paper, Colorado State University

    2. Ambrose, Brent W. (1990) An Analysis of the Factors Affecting Light Industrial Property Valuation in

    Journal of Real Estate Research v. 5 (3), p. 355-370

    3. Asabere, Paul K. and Huffman, Forrest E. (1991) Zoning and Industrial Land Values: The Case of

    Philadelphia in AREUEA Journal, v. 19 (2) p. 154-160

    4. Buttimer, R.J., Rutherford, R.C. and Witten, R. (1997) Industrial Warehouse Rent Determinants inthe Dallas/Fort Worth Area in Journal of Real Estate Research, v. 13 (1), p. 48 - 55

    5. Borst, R. A. (1995) Artificial Neural Networks in Mass Appraisal in Journal of Property Tax

    Assessment and Administration, v. 1 (2)

    6. Chow, Yuen L., Ong, Seow Eng and Thang, Doreen, C.L. (2002) A Cointegration Approach to

    Understanding Singpores Industrial Space Supply in Journal of Property Investment and Finance, v.

    20 (2) p. 96-115

    7. Detweiler, John H. and Radigon, Ronald E. (1999) Computer-assisted Real Estate Appraisal: A Toolfor the Practicing Appraiser in The Appraisal Journal, Jul 1999, p.280-286

    8. Dreyer, Brent J. (1989) Artificial Intelligence: The AI MAI Appraiser in the Appraisal Journal, Jan

    1989, p.51-56

    9. Eichenbaum, Jack (1995) The Location Variable in World Class Cities: Lesson from CAMA Valuation in

    New York City in Journal of Property Tax Assessment and Administration, v.1 (3) p. 46-60

    10. Fehribach, Frank A., Rutherford, Ronald C. and Eakin, Mark E. (1993) An Analysis of the

    Determinants of Industrial Property Valuation in The Journal of Real Estate Research, v. 8 (3), p.365-376

    11. Ferreira, Eurico J. and Sirmans, G. Stacy (1988) Ridge Regression in Real Estate Analysis in The

    Appraisal Journal, July 1988, p. 311-319

    12. Gallimore, P., Fletcher, M. and Carter, M. (1996) Modelling the Influence of Location on Value inJournal of Property Valuation & Investment, v. 14 (1), p. 6-19

    13. Hoerl, A. and Kennard, R. (1970) Ridge Regression: Biased Estimation for Nonorthogonal Problems

    in Technometrics, p. 69-82

    14. Isakson, Hans R. (1998) The Review of Real Estate Appraisals Using Multiple Regression Analysis inJournal of Real Estate Research v. 15(1/2), p. 177-190

    15. Ishizuka, Teruo (1995) Computer-assisted Appraisal System using Geographic Information System

    in Journal of Property Tax Assessment and Administration, v.1 (3) p. 91-103

    16.

    17. Kling, John L. and McCue, Thomas E. (1991) Stylized Facts about Industrial Property Construction in

    Journal of Real Estate Research, v. 6 (3) p. 293-304

    18. Kowalski, Joseph G. and Colwell, Peter F. (1986) Market Versus Assessed Values of Industrial Land inAREUEA Journal, v. 14 (2) p. 361-373

    19. Lockwood, Larry J. and Rutherford, Ronald C. (1996) Determinants of Industrial Property Value in Real

    Estate Economics, v. 24 (2) p. 257-272

    20. Mark, Jonathan and Goldberg, Michael A. (1988) Multiple Regression Analysis and Mass Assessment: A

    Review of the Issues in The Appraisal Journal, Jan 1988, p. 89-109

    21. McCluskey, William J. (1996) Predictive Accuracy of Machine Learning Models for the Mass Appraisal of

    Residential Property in New Zealand Valuers Journal, July 1996, p. 41-47

    22. McCluskey, William, Deddis, William, Mannis, Adam, McBurney, Dillon and Borst, Richard (1997)

    Interactive Application of Computer-assisted Mass Appraisal and Geographic Information System in

    Journal of Property Valuation & Investment, v. 15 (5) p. 448-465

    Spatial Analysis of Industrial Property Values: A Case Study of JTC in Singapore7t h A s i a n R e a l E s t a t e S o c i e t y C o n f e r e n c e , S e o u l , 4 - 6 J u l y 2 0 0 2

    20

  • 8/8/2019 Shi-Ming Yu(A2)

    21/21

    23. McCluskey, William and Anand, Sarabjot (1999) The Application of Intelligent Hybrid Techniques for

    the Mass Appraisal of Residential Properties in Journal of Property Investment & Finance, v. 17 (3) p.

    218-238

    24. Murphy III, Lloyd T. (1989) Determining the Appropriate Equation in Multiple Regression Analysis inThe Appraisal Journal, Oct 1989, p. 498-513

    25. Nawawi, A.H., Jenkins, D. and Gronow, S. (1997) Expert System Development for the Mass Appraisalof Property in Malaysia in McCluskey, W.J. and Adair, A.S. (eds.) , Computer Assisted Mass Appraisal:An International Review, Ashgate, Aldershot

    26. Newell, Graeme J. (1982) The Application of Ridge Regression to Real Estate Appraisal in The

    Appraisal Journal, Jan 1982, p. 116-119

    27. Newsome, Bobby A. and Zietz, Joachim (1992) Adjusting Comparable Sales Using MultipleRegression Analysis The Need for Segmentation in The Appraisal Journal, January 1992, p. 129 -

    135

    28. Panayiotou, Panayiotis Andrea, Jenkins, Pastor David, Pattichis, Constantinos, and Plimmer, Frances(1999) Computer Assisted Valuation: Multiple Regression Analysis in Cyprus in Journal of Property

    Tax Assessment and Administration, v.4 (2) p. 27-52

    29. Ramsland, Maxwell O. and Markham, Daniel E. (1998) Market-Supported Adjustments Using MultipleRegression Analysis in The Appraisal Journal, April 1998, p. 181-191

    30. Scott, I (1988) A knowledge based approach to the computer-assisted mortgage valuation ofresidential property, unpublished PhD thesis, University of Glamorgan

    31. Tay, Danny P. H. and Ho, David K. H. (1995) Intelligent Mass Appraisal in Journal of Property Tax

    Assessment and Administration, v.1 (1) p. 5-25

    32. Thompson, Bob and Tsolacos, Sotiris (2001) Industrial Land Values A Guide to Future Markets? in

    Journal of Real Estate Research, v. 21 (1) p. 55-76

    33. Thompson, Bob and Tsolacos, Sotiris (1999) Rent Adjustments and Forecasts in the Industrial

    Market in Journal of Real Estate Research, v. 17 (1) p. 151-167

    34. Waller, Benie D. (1999) The Impact of AVMs on the Appraisal Industry in The Appraisal Journal, Jul1999, p. 287-292

    35. Ward, Richard D. (2001) Demonstration of CAMA in South Africa in Assessment Journal, v.8 (3), p.

    33-43

    36. Wheaton, William C. and Torto, Raymond G. (1991) An Investment Model of the Demand and Supplyfor Industrial Real Estate in Journal of the American Real Estate and Urban Economics Asosiation, v.

    18 (1) p. 530-547

    37. Worzala, Elaine, Lenk, Margarita and Silva, Ana (1995) An Exploration of Neural Networks and Its

    Application to Real Estate Valuation in The Journal of Real Estate Research v.10 (2), p. 185-201

    38. Wyatt, Peter (1996) The Development of a Property Information System for Valuation using a

    Geographical Information System in Journal of Property Research v. 13, p. 317-336

    39. Wyatt, Peter (1996) Using a Geographical Information System for Property Valuation in Journal ofProperty Valuation & Investment, v. 14 (1), p. 67-79

    40. Yu, Shi Ming, Siu, Kelvin K and Han, Sun Sheng, (2000) Using Response Surface Analysis in

    Computer-assisted Mass Appraisal to Examine the Influence of Location on Property Values in Journal

    of Property Tax Assessment and Administration, v. 5(4) p. 3 15

    Spatial Analysis of Industrial Property Values: A Case Study of JTC in Singapore 21