`
`Richard A. Borst, Chief Information Officer Cole Layer Trumble Company, USA
`William J. McCluskey, Senior Lecturer, University of Belfast, Northern Ireland
`
`Introduction
`The ultimate goal of any Computer Assisted Mass Appraisal (CAMA) system or project is to
`achieve ajurisdiction-wide set of equitable values. This is achieved most often via the ful
`market appraisal of all properties. Appraisal professionals are well aware that locatiorûl effects
`must be talcen into account in CAMA systems in order that accurate fiutI market valuations are
`achieved. In our research we consider the definition of location as it relates to property appraisal
`and then proceed to describe a practical approach to incorporating locational effects into a CAMA
`system.
`
`Treatment of Location
`Location could be defined as an economic characteristic of real estate composed of immobility,
`constant change, and elements of special distribution. Location is an economic concepteven
`though a location can be described in physical and legal terms. Locational analysis involves the
`thorough study of use, environment, time and anticipated patterns of change.
`
`Residential markets are complex phenomena, consisting of an extremely heterogeneous stock as
`well as diverse producers and consumers of widely varying réquirements and financial
`capabilities.
`
`The idea of thinking of residential markets as a unitary market is in many ways too simplistic,
`when one considers the discontinuities across locations and dwellings. The task of dividing a
`large market into submarkets raises several theoretical and methodological questions (Bourassa et
`al, 1997.) Typically a submarket can be defined as a set of dwellings that are reasonably close
`substitutes for each other, but relatively poor substitutes dwellings in other submarkets (Grigsby
`et al, 1987.) This definition leads to difficult questions about how to identify close substitutes
`and levels of aggregation or indeed disaggregation. In practice these questions are often
`answered in an ad hoc manner, using pre-defined or otherwise convenient geographical
`boundaries as the basis for determining and defining submarkets. In some other cases statistical
`techniques can be applied to determine whether a priori submarkets are in fact distinct. In this
`research we seek to define a refined ad hoc approach commonly referred to as neighborhood
`delineation. Our goal is to apply the tools of database analysis coupled with Geographic
`Information Systems to achieve a semi-automatic means for accounting for location effects in
`CAMA models.
`
`Use of Neighborhoods in CAMA Systems
`One of the most frequent methods of accounting for location influences in CAMA systems is the
`use of one or more neighborhood codes in the CAMA database. It can take the form of a single
`field containing a reference to a contiguous homogenous region. In such cases the jurisdiction to
`be valued is delineated into n contiguous regions, each of which is given a unique identification
`(NBHD). It can be two fields, one as previously described and another to identify m groups of
`similar, but not necessarily contiguous neighborhoods (NB}IDGRP). Thus a given parcel will
`have fields that indicate it is in NBHD 57 and NBHDGRP 4. Often the NBHD variable is
`
`
`
`formatted for future maintenance by allowing it to be split into 99 sub-neighborhoods as future
`changes may dictate.
`
`The use of such fields occurs in the modeling, calibration and valuation processes. Models are
`constructed that include NBHD or NBFIDGRP variables to account for-variations in value due to
`location effects. An example of a predictive model employing NBHD is given in the following
`expression:
`
`ESP = b0 * size * b110' *
`
`Where ESP is Estimated Selling Price and NBFID is a binary variable
`
`The model is, of course quite limited, but it illustrates one mechanism for using NBFID in a
`model structure.
`
`Systems that identi' and adjust comparable sales as part of the valuation process also use NBHD
`or NBFIDGRP to limit the search to areas that have been predefined as "comparable" via the
`location variables. For example the scheme might be to search the NBI-ID of the subject parcel
`first to see if there are sufficient comparable sales to estimate value. If there are, the algorithm
`computes an estimate of market value based on the given sales. If there were not, the algorithm
`would search the NBI-IDGRP to obtain the sales needed for the analysis.
`
`Traditional Neighborhood Delineation Process
`Appraiser uses a variety of tools and infonnation sources to accomplish neighborhood
`delineation. Regardless of the process used to identi& neighborhoods, they are ultimately
`translated to map sheets to create a visual display and to ensure that all geographic areas have
`been accounted for.
`
`Factors considered in neighborhood delineation include at least the following:
`
`Zoning
`Land use
`Lot size
`Available utilities
`View
`Building style
`Building size
`Building age
`Quality of construction
`Sale price
`
`There are more factors that could be listed, but generally the data the appraiser needs to perform
`neighborhood delineation could be found in a typical CAMA database coupled with ownership
`maps and preferably an aerial photo. Establishing neighborhoods frequently involves driving the
`streets of the jurisdictions with ownership maps and CAMA record printouts to make in-field
`determinations of changing neighborhood patterns.
`
`273
`
`
`
`Data Sources
`The study are is a dormitory town located approximately 7 miles north of Belfast. It has a
`population of bpproximately 30,000. Over the pastten it tas seen rapid growth in residential
`development. Residential properties re jiredominantly owner occupied and further characterized
`by significant numbers of semi-detachedand detached dwellings.
`
`The data used for the purposes of this research were supplied from two government agencies, the
`Valuations and Lands Agency (VLA) and Ordinance Surveys of Northern Ireland (OSNI.)
`
`The VLA is responsible for determining the values used for property tax purposes on
`approximately 670,000 residential properties and 66,000 commercial properties. To fulfil the
`statutory fUnction with regard to property tax assessment the VLA has a comprehensive database
`incorporating a range of specific attributes for each property. In relation to this work, information
`on the following variables was supplied:
`
`Sale price
`Sale date
`Gross floor area
`Number of bedrooms
`Presence of garage
`Presence of central heating
`Effective age
`Property type
`Property class
`
`OSNI is responsible for the creation and maintenance of the topographic archives of Northern
`Ireland. Digital information is available in vector form for 1:1250, 1:2500, 1:50,000 and
`1:250,000 scales. Color and monochrome raster information is also available. The vector digital
`database consists of 190 feature codes or themes capable of handling points, lines and polygons.
`The following diagram illustrates the range of themes available:
`
`Top Level
`
`Communication
`
`jWater Features
`
`Buildings
`
`Topo Areas
`
`Fresh
`
`General
`
`Perimeter
`
`Tidal
`
`Communal
`
`Boundaiy
`
`Rail
`
`Water
`
`Roajis
`
`274
`
`
`
`OSNI supplied vector maps at 1:1250 scales for the study area incorporating the following
`themes:
`
`Buildings
`Plot boundaries
`Geographical wards
`Roads
`Railways
`
`The Integration of CAMA and GIS
`The research was performed on a personal computer in which a GIS (Aicview) was integrated
`with a Relational Database (ORACLE) to allow for the convenient analysis of characteristic data
`in a GIS session. The following figure depicts the architectûre used:
`
`Personal Computer
`
`Arc/View
`t
`
`ODBC
`
`i:
`
`Oncle
`
`275
`
`
`
`This achieves integration between the graphic shapes and the CAMA database as depicted in the
`following figure:
`
`Joining Two IS Technologies
`
`CAMA
`
`OIS
`
`The linkage between the GIS and the CANtA databases is shown as Parcel ID. Other possibilities
`forjoining the data include parcel address as well. Once the data are joined, the concept of an
`"Analyst's Workbench" can be achieved. Constricting user friendly applications via the
`customization tools available from the GIS vendor does this. In this case the Avenue
`programming extension was employed to allow for neighborhood analysis and modeling without
`extensive training in the GIS product itself. The following figure illustrates this concept:
`
`276
`
`
`
`The custom tools et allows
`the analyst to construct
`"what f scenarios" and
`establish NBHD based on
`the available sales and
`properly characteristic data
`linked to the available GIS
`graphic database.
`
`Cus to m
`
`The following sequence of figures depicts updating NBHD directly to the ORACLE database:
`
`L
`
`OW
`
`J1A
`STOFW
`
`STUma
`f00
`U41I
`
`4
`
`za.
`z
`
`i
`1%
`POW.6Mf G
`
`The Analyst Has Created A Thematic Map Showing Size of Parcel Shaded by Color
`Figure Also Depicts ¡he CAMA Data for an Individual Parcel in Question
`
`277
`
`
`
`@D&ìSf
`
`berh @0
`
`Stied the held to classify on
`
`NO H D_P
`
`NOTECOI
`NOIECD2
`RE STRI Cn
`RES1RICTZ
`RE STR] Cfl
`TO PO i
`
`Stied The method Sor dasslfying
`
`Equal counts
`ranges
`
`IEqual
`
`I
`
`The Figure Depicts the Interface as the Analyst is About to Classjfy the Map by NBHD
`The Class (fier icon ön the Custom Toolset is employed in the process, allowing the Analyst to
`perform this function without extensive knowledge of the GIS
`
`*
`
`w
`
`Figure Dep ic/s the Interface after the Map Has Been Classed by NBHD
`
`278
`
`
`
`PIpS
`
`E?I®MflS
`
`@0 Cod
`
`sIiI;dropda
`
`Figure Depicts the Interface as the Analyst is about to Update NBHD
`
`Map Showing Updated Neighborhoods
`
`Depicts the updated NBHD Map after using the Class j/ìer Icon
`
`279
`
`
`
`Results
`Establishing NEI-ID and performing the subsequent modeling a calibration is ongoing at the time
`of this submassion. The toots described herein wilt be used to value the properties in the subject
`area. Actual results will be presented.at the conference in September of 1997.
`
`References
`
`Grigsby, W., Baratz, M., Gaister, G. and Maclennan, D. (1987), The Dynamics ofNeighborhood
`Change and Design, Progress in Planning, 28 (1), 1-76
`
`Bourassa, S.C., Hainelink, F., Hoesli, M. and MacGregor, B.D. (1997). Defining Residential
`Submarkets: Evidence from Sydney and Melbourne, Rea! Estate Research Unit, Working paper
`No. 3, University of Auckland.
`
`280
`
`I....1.4.
`
`