Development of Spatial Decision Support Systems for ... · Development of Spatial Decision Support Systems for Residential Real Estate ... Applications of this type of SDSS are readily

Journal of Housing Research • Volume 9, Issue 1 135q Fannie Mae Foundation 1998. All Rights Reserved.

Development of Spatial Decision Support Systems forResidential Real Estate

Kim Peterson*

Abstract

Geographic Information Systems (GIS) can enhance the efficiency and effectiveness of decision makingin the residential real estate industry. They can organize, manage, and analyze information in waysthat were not possible with traditional information management systems. Although GIS are now usedto perform specific business functions, their use can be magnified and extended through the creationof enterprise-wide spatial decision support systems (SDSS).

This article provides a conceptual framework for the development of enterprise-wide SDSS. The firstpart of the article discusses the nature of real estate decision making and investment analysis, payingspecial attention to residential real estate. It also reviews different approaches to SDSS development.The second part of the article discusses enterprise-wide information architecture planning and speci-fies a conceptual framework for SDSS development. It then discusses issues related to technologytransfer and SDSS implementation.

Keywords: Geographic information systems; Spatial decision support systems; Residential real estate

Introduction

This article presents a conceptual framework for analyzing how Geographic InformationSystems (GIS) and spatial decision support systems (SDSS) can be effectively implementedin large, complex organizations. It argues that effective and efficient decision support re-quires consideration of enterprise-wide information technology (IT) issues and needs andthat decision support is best approached as part of an IT architecture planning process.

As a basis for developing a conceptual framework, this article discusses the nature of realestate decision making and emphasizes residential analysis. It then defines GIS, expertsystems (XS), and decision support systems (DSS) and discusses how they relate to SDSS.It also emphasizes the ways in which SDSS incorporate GIS functions.

This article then specifies a conceptual framework based on IT architecture planning prin-ciples. These principles are well founded, and the framework derived from them provides aflexible approach to specifying how SDSS can be developed to serve enterprise-wide decision-making needs.

* Kim Peterson is a research associate at the Institute for Urban Land Economics Research, Inc. The author wishesto thank Ayse Can of the Fannie Mae Foundation for her guidance, and two anonymous referees for their helpfulcomments on an earlier draft of this paper.

136 Kim Peterson

The Nature of Residential Real Estate Decision Making

Residential real estate decisions are made by a variety of actors pursuing a broad range ofobjectives. These actors include home buyers and renters, builders, brokers, bankers, andthe public agencies that provide physical networks and services such as streets, utilities,and schools. Many decisions have spatial dimensions and may be grouped according to func-tional area, recognizing that some functions are actor specific and some may be classified inmore than one function. The following functional areas are among the most important: (1)planning, strategy formulation, and research; (2) construction and development; (3) finance,including construction lending, equity investment, mortgage finance, public and private ven-turing, and property taxation and assessment; (4) property management; (5) risk manage-ment; (6) marketing; and (7) regulatory compliance.

Geographic information and expert systems have been developed to support residential de-cision making in each of these areas; selected examples illustrate their contributions (seeBelsky, Can, and Megbolugbe 1996; Thrall and Amos 1996). In the areas of planning andconstruction, expert systems have been designed to analyze land use laws and other legalissues related to location, including determining whether a proposed land use meets zoningand other land use regulations (Waterman 1985). GIS have been developed to assist in siteselection and location analysis for residential subdivisions (Barnett and Okoruwa 1993),and expert systems have been designed for managing construction activities (Levitt andKunz 1985). Several vendors offer computer applications for estimating construction costsbased on geographic location, which has done much to streamline the real estate appraisalbusiness.

In the area of risk management, GIS and SDSS have been developed to help mortgagelenders and insurers improve their underwriting procedures and price their policies. Theseapplications help to determine whether a property is located in an area prone to naturaldisaster (e.g., floods or earthquakes) and to calculate rates based on automated assessmentof neighborhood crime rates and distances to fire hydrants, fire stations, and police stations(Francica 1993; Kochera 1994). Applications have also been developed to automate the ap-praisal process (Can and Megbolugbe 1996; Robbins 1996; Rodriguez, Sirmans, and Marks1995) and to utilize detailed site measurements and sales data provided via computer net-works and CD-ROMs. When used to mark mortgage portfolios to market value, these sys-tems help lenders improve credit loss forecasts.

Mortgage lenders, originators, and consultants have adopted GIS for a variety of marketingfunctions (for examples, see Beaumont 1991a, 1991b; Clark 1993; Graham 1992; Hall 1993;Pittman and Thrall 1992), including identification of market areas for specific branch officelocations, estimation of market potential for particular mortgage products within those ar-eas, computation of expected capture rates, measurement of current market penetration,segmentation of markets based on geodemographics, selection of targets for direct mail ad-vertising, and location-allocation modeling to help determine the optimum number and lo-cations of branch facilities.

Finally, GIS have been developed to assess the compliance of lenders with the CommunityReinvestment Act (CRA), which stipulates that financial intermediaries meet the creditneeds of their entire community, including residents of low- and moderate-income neigh-

Development of Spatial Decision Support Systems for Residential Real Estate 137

borhoods. These systems provide visual displays of the geographic locations of mortgageloan applications, approvals, and denials and combine these data with thematic map over-lays showing census tracts by income level; the result is a visual display of patterns of serviceand underservice (Belsky, Can, and Megbolugbe 1996; Thrall, Fandrich, and Thrall 1995).

GIS and SDSS

The applications mentioned in the preceding section demonstrate that GIS can be efficientand effective tools for managing both business and geographical information. They providean important technology for the development of DSS that require spatial data as input orthat have spatial implications. But the analytical requirements placed on these systemsvary, depending on the specific knowledge domains they are designed to support. Some busi-ness areas can be served with existing commercial GIS functionality (e.g., mapping andsimple spatial queries), whereas some require the incorporation of XS modules and sophis-ticated algorithms. The following definitions highlight important features of GIS, XS, DSS,and SDSS and provide a conceptual basis for matching applications to functions.

Geographic Information Systems

A GIS is defined as a system of hardware, software, data, people, organizations, and insti-tutional arrangements for collecting, storing, analyzing, and disseminating informationabout areas of the earth (Dueker 1989). Requisite functions include the ability to create,edit, and delete geographically structured data; link locational and attribute data; performspatial analysis functions such as map overlay of multiple data themes; and display geo-graphic information.

Expert Systems

Expert systems are computer programs that apply artificial intelligence to narrow andclearly defined problems. They typically combine rules with facts to draw conclusions, relyheavily on theories of logical deduction, and are developed using heuristic methods or con-ventional computer programs (Ortolano and Perman 1990). The subject area of an XS iscalled its domain. The collection of facts, definitions, rules of thumb, and computationalprocedures that apply to the domain is called the knowledge base. The set of procedures formanipulating the information in the knowledge base to reach conclusions is called the con-trol mechanism (or inference engine).

Decision Support Systems

Numerous definitions and characterizations of DSS have been proposed. For purposes of thisdiscussion, a DSS may be viewed as an interactive computer-based system that helps deci-sion makers utilize data and models to solve unstructured problems (Sprague and Carlson1982). An unstructured problem is not susceptible to algorithmic solution because it has notarisen before, because its precise nature and structure are elusive or complex, or because it

138 Kim Peterson

is so important that it deserves custom-tailored attention (Simon 1960). In such situations,decision makers may need to research possible solutions and evaluate and modify the pos-sibilities until an acceptable solution is obtained. Accordingly, a central theme for the de-velopment of DSS is the notion that the role of the DSS is not to undertake any kind ofsearch for a problem solution but to assist the decision maker who undertakes that search(Davis and McDonald 1993).

Spatial Decision Support Systems

When a DSS is developed for use with a domain that includes spatial data (one or more dataattributes with a distribution over space), or when the solutions generated by the DSS havespatial dimensions, the analytical system may be referred to as an SDSS (Wright and Bueh-ler 1993). SDSS enhance decision making by supporting processes that can be characterizedas iterative, integrative, and participative (Densham and Goodchild 1989). This support isgiven through provision of models in a model base, a user-friendly interface, and easy accessto appropriate data (irrespective of location or format).

Characteristics specified by Densham and Goodchild (1989) can be used to formally defineSDSS: These systems are normally provided for a limited problem domain, integrate bothspatial and nonspatial data, typically facilitate the use of analytical and statistical modeling,and convey information to decision makers by way of a graphical interface. These systemsalso adapt to the decision maker’s style of problem solving and are easily modified to includenew capabilities (Keen 1980).

SDSS Evolution

Recent advances in GIS and SDSS have been driven in part by enhancements in the powerand affordability of desktop computing. This microchip-based technology includes free-standing desktop machines but may also include networks of machines linked together andto servers, which may be other microcomputers, workstations, minicomputers, or main-frames. Two evolutionary paths can be identified (figures 1 and 2).

Spatial decision support systems of the first path are evolving mainly from GIS (figure 1).These systems emphasize spatial data and higher-order spatial relationships (Goodchild1992), and are typically applied to problems in which spatial analysis is the primary focus.For example, an SDSS from this path is seen as ‘‘some device (software and hardware) thatprovides information to a decision maker, ideally in an interactive framework, that can assistwith a locational decision’’ (Fotheringham 1990, 1137). These systems are sometimes clas-sified according to whether they rearrange existing information or generate new informa-tion, and the latter are further differentiated according to the presence or absence of con-sumer choice (Fotheringham 1991). Spatial data handling is highly sophisticated in thesesystems, and ‘‘toolboxes’’ of spatial analysis functions are typically included. Because theyemphasize spatial analysis, these systems generally provide only limited DSS and XS com-ponents for nonspatial analysis and decision support. This narrowness of focus is reflectedin the broken line linkage between the GIS and decision support boxes shown in figure 1.


Figure 1. GIS-Driven Evolution of SDSS

Figure 2. DSS-Driven Evolution of SDSS

Applications of this type of SDSS are readily found in retail location analysis. For example,spatial interaction models of different types can be selected from menus in various GISsoftware packages (e.g., SPANS [TYDAC Technologies, Inc.] and TransCAD [Caliper Cor-poration]). Other examples include SDSS developed by public utilities to support asset op-timization decisions related to processing service requests, engineering design, load man-agement, and work management (Epstein and Odenwalder 1993). Applications have alsobeen developed for the provision of social services by state and local governments (Vachon

140 Kim Peterson

1993) and for land use planning in both public and private contexts (Budic 1994; Kinsey andAvin 1992), but the focus of all of these systems remains predominantly locational analysis.

SDSS of the second path (figure 2) are evolving mainly from or in concert with DSS. Thesesystems emphasize decisions in which spatial analysis is of only coequal or secondary im-portance. They are typically tailored to a specific knowledge domain and emphasize themanipulation and analysis of data relevant to that domain. High-level spatial data handlingprocedures and fine-grain geographic precision are not available in many of these systems,but they often provide access to, or are compatible with, sophisticated DSS and XS compo-nents that have been developed for the knowledge domain or problem area of interest.

Examples of SDSS of this second path may be found in a wide variety of application areasincluding land development suitability analysis (Han and Kim 1990; Lein 1990), construc-tion management (Levitt and Kunz 1985), comprehensive land use planning and regulation(Davis and McDonald 1993; Maidment and Evans 1993), identification of hazardous mate-rials and their handling in the case of accidents (Cooke 1992), allocation of territories tosales personnel (Bryan 1993), and the specification of product mix for alternative retail sitesbased on geography, traffic flows, and customer profiles (Kroh 1995).

Although SDSS of both paths address specific user needs, the first path holds limited promisefor significant near-term improvements in decision support. This limitation follows mainlyfrom the kind of analysis that is typically performed and the relatively smaller market inwhich this analysis takes place. In contrast, the problems addressed by SDSS of the secondpath are based mainly in business processes, not geography, and in many cases do not callfor the same kind of high-level geographical precision for which digitized maps and GIStoolbox functions are required. The market for this type of analysis is much larger, becausethere may be hundreds of businesses that can use this type of system for every public agencyor business dealing with locational decisions.

The evolution of SDSS along the second path will also be fueled by three features of desktopcomputing: the increasing power and affordability of desktop systems, the widespread adop-tion of these tools by businesses and individuals, and the increasing use of DSS for businessanalysis. Microcomputer-based spreadsheet and database management systems frequentlyfunction as DSS and have already been widely incorporated into organizational proceduresfor analyzing problems and making decisions. Some business schools even use enhancedspreadsheet software to teach economics and statistics courses, and millions of these DSShave been installed in businesses, schools, and homes. It follows that SDSS and desktopGIS products that can function in concert with these DSS are likely to benefit from theirwidespread popularity.

The importance of this compatibility is reflected in recent assertions that GIS provide a newparadigm for the organization of information and the design of information systems (Dan-germond 1995; Vonderohe et al. 1993). The central tenet of this paradigm is the use oflocation as a basis for restructuring information systems and developing new ones. The latestdesktop GIS provide for easy, seamless linking with spreadsheets and database managementsystems (DBMS), and a wide array of value-added vendors provide census data and other


information in spatially referenced formats. In addition, at least one vendor has released aspatial database engine (SDE) for use with Oracle database management systems (ESRI1995). This SDE makes highly accurate spatial data available as a common data type withincorporate and government database environments, meaning that GIS data can now be fullyincorporated with nonspatial data in a single relational database architecture.

The evolution of SDSS from DSS is also being promoted in the applications developmentenvironment. Some microcomputer-based GIS provide for the development of custom appli-cations to fit special user needs, and at least one vendor has designed its GIS desktop prod-ucts so they can be customized with the same programming software used to build spread-sheet and DBMS applications.1 This ability to customize will encourage tighter integrationof GIS functions with business-oriented DSS and facilitate creation of a broad array of SDSS.

SDSS Development and IT

Implementation of GIS and SDSS tools throughout a large, complex enterprise requirescareful planning, and specification of an IT architecture is implied. In developing this ar-chitecture it is important to account for the current enterprise-wide IT infrastructure, es-pecially the DBMS currently in use and important legacy systems.

The benefits typically cited for creating and implementing an IT architecture are substantialand include the following (Rosser 1995):

1. The ability to achieve interoperability—that is, different systems working together, es-pecially in sharing data

2. The ability to speed the implementation of new systems by having many choices alreadysettled and the skills and learning in place

3. Lower costs due to reduction of support effort as the number of products and processesis reduced

4. The general upgrading of system quality and increased ability to make modificationsmore easily in the future

5. Communication of a common direction throughout IT and end-user departments, a by-product of going through the architecture planning process

6. In the effort to reduce the total number of components and processes employed, anarchitecture may forestall the crisis of complexity that is projected for networked com-puting as more and more demands are made and expectations rise.

1 ESRI provides its own application development language for ArcView but also supports Microsoft’s Visual Basic.

142 Kim Peterson

A Conceptual Framework for SDSS Development

Previous efforts to develop conceptual frameworks for GIS and SDSS have grown out ofapplication areas as diverse as transportation research (Vonderohe et al. 1993), urban plan-ning (Innes and Simpson 1993), market analysis (Beaumont 1991a, 1991b), and environ-mental analysis (Abel et al. 1992), but all have shared a common core of concerns. The designof these frameworks have all taken into account that rapid changes in information technol-ogy are radically affecting the ways in which GIS and SDSS are built and used, that weshould rethink these systems completely rather than modify them in a piecemeal fashion,and that we need to focus on decision-making processes and organizational structures tounderstand how these systems may influence work practices, roles, responsibilities, and thenature of tasks.

The conceptual framework developed in this article is based on IT planning principles thatdefine and distinguish three types of information architecture: data, process, and networktechnology (Zachman 1987). The structure of this framework is shown with real estate–related examples in figure 3.2

This framework may be used to analyze the IT needs of an organization as a whole, in whichcase all enterprise areas and subareas would be evaluated. In the present case, it is used toorganize thinking about only those areas of the enterprise that deal with real estate decisionmaking and analysis. It provides guidelines for determining what data must be included inan enterprise-wide SDSS (column 1), what business analysis and decision-making processesmust be supported (column 2), and where these services must be provided in terms of bothorganization structure and geography (column 3). The three cells in the top row of figure 3list general classes of data, processes, and locations that the information architecture isintended to address. The cells in the second row list specific entities belonging to each class.

Data Architecture

Cell 1 of figure 3 lists the classes of data entities that are important to the business enter-prise as a whole or to specific departments. It is generally impractical or impossible to designinformation architecture for all data classes, so the decision regarding which subset of itemsto choose is based on the values and strategies of the business. Similar decisions to narrowthe focus of the architecture are made for the other general-level description cells (i.e., cells2 and 3 of figure 3). For example, there will probably be insufficient resources to automateall relevant business processes (cell 2), and out of the total list of locations in which thebusiness operates (cell 3), there are probably insufficient data processing resources to placehardware and software at every location.

At this most general level, an operative data question for a large residential lender mightbe ‘‘What things must we know about in order to make residential mortgage loans?’’ Classesof housing market participants come immediately to mind, including space producers, spaceconsumers, and infrastructure providers (Graaskamp 1980). Space producers include thosewho assemble capital, provide materials or expertise, and construct dwelling units on site;

2 Only the top two rows of Zachman’s architecture planning framework are provided here. The remaining four rowshave been omitted because they deal with IT design issues, which are beyond this article’s scope.


Figure 3. Framework for SDSS Architecture

DataArchitecture

Entities importantto the business

ProcessArchitecture

Processes thebusiness performs

NetworkArchitecture

Locations in whichthe business operates

General Level 1

Classes of Entities:

Space ProducersSpace Consumers

Infrastructure Providers

Features of UrbanEnvironments*

2

Classes of Processes:

Mortgage LoanUnderwriting

Property Management

Construction Lending

Market Research

3

Classes of Locations:

Office Locations(home and branch)

OrganizationDepartments

Specific Level 4

Specific Entities:

Anderson-Kiefer Builders

Empire Mortgage Company

Mortgage UnderwritingDepartment

5

Specific Processes:

Projection ofPopulation Trends

NCREIF Guidelinesfor Analysis

Proprietary Heuristics

6

Specific Locations:

Branch Office:Santa Rosa Plaza

Loan UnderwritingDepartment

Property ManagementDepartment

Note: Based on Zachman’s Framework for Information Systems Architecture (1987, p. 285).*Andrews (1980) identifies land use, physical, social/cultural, psychological, economic, and institutional/politicalenvironments. Features are associated with each. For example, features of an area’s physical environmentinclude topography, ground cover, soils, atmosphere, buildings, streets, and so on.

architects, mortgage bankers, lumbermen, lawyers, city planners, and apartment owners/operators are all included in this group; space consumers include households and individualsseeking to purchase or rent places to live. Infrastructure providers include enterprises thatcreate and maintain physical networks of streets, sewers, and utilities; they also provideservices such as education, police, and fire protection and administer operational systemsfor deed registration, government regulation, adjudication, and various forms of economicactivity.3

3 A similar perspective is provided in the analysis of institutional forces as important determinants of urban areadevelopment (Shlay 1988). The focus here is on developers, speculators, and financial organizations as classes ofentities. Financial organizations can in turn be segmented into subclasses of mortgage bankers, banks, insurancecompanies, pension funds, and so on.

144 Kim Peterson

A second general question might be ‘‘What things must we know about to establish andmaintain our branch offices?’’ This question is important because of the complexity of se-lecting locations, maintaining owned buildings, and evaluating the relative attractivenessof leasing space in other kinds of facilities (e.g., placing branch offices in supermarkets).4

Decision support may also be needed in the property management division if the enterprisemanages residential properties.

Data classes can also be organized in terms of urban environments, including land use,physical, social/cultural, psychological, economic, and institutional/political environments(Andrews 1982). For example, analysis of economic environments may focus on real estatemarkets as an important subarea. Data needed to support this type of analysis would includelevels and rates of growth for populations of individuals, households, and businesses; eco-nomic data related to interest rates, employment, and income; and demand parameters suchas household size, quantities of housing space demanded, expenditures per capita, and levelsof move-up demand (Myers and Mitchell 1993). Historical capture and absorption rates (byreal estate type) could also be required.

Whereas classes of entities are specified in cell 1, specific instances are implied in cell 4 ofthe framework. For example, the general class of space producers identified in cell 1 includeslocal residential builders and mortgage brokers. When managers in the firm’s mortgageunderwriting department specify an entity such as a home builder, what they have in mindis a developer or contractor that builds single-family homes. Similarly, within-enterpriseentities could include the mortgage underwriting department, the property managementdepartment, and construction lending.

Relationships among data types—the business rules or strategies that relate one entity toanother—are also specified in column 1 of figure 3. A business rule or strategy for a largemortgage lender might be ‘‘to make loans so as to satisfy CRA requirements.’’ Another re-lationship might be ‘‘to place new branch facilities in only those census tracts that are pro-jected to experience annual population growth in excess of x percent.’’

Process Architecture

As with the data architecture, each of the cells in the process architecture column of figure3 has a different meaning depending on level of generality, although each may be describedin terms of an input-output diagram. In the discussion that follows, general kinds of pro-cesses in which a large residential lender engages are discussed first, followed by specificprocesses related to mortgage investments in residential properties.

At the general level, process means class of business process in which to invest resources forautomation or decision support. This definition implies procedures such as mortgage loanunderwriting, property management for homes or apartment projects acquired through loandefault, construction lending, and housing market research. A large residential lender might

4 The classes of entities of interest to facilities managers may include transportation networks; fleets of vehicles;parking facilities; buildings and rooms; multibuilding heating, cooling, and electrical systems; the location of officefurniture and equipment (including computer terminals); and space leased to or from other firms.


choose to emphasize mortgage loan underwriting and a comprehensive program of housingmarket research but to defer support for procedures used in a construction lending process.

At the specific level, decisions would be made regarding the particular procedures to providein the decision support apparatus. A functional flow diagram might be specified in whichprocess would be a business process (not an information systems process), and inputs andoutputs would be business resources such as people, cash, or material. For a large mortgagelender, examples include processes for projecting population and employment trends in spe-cific housing markets under different scenarios and procedures for delineating market areasfor branch office locations. Many specific algorithms or procedures are available for sup-porting these processes. The National Council of Real Estate Investment Fiduciaries spec-ifies guidelines for conducting a variety of analyses, including market area delineation, de-mand analysis, supply analysis, absorption analysis, vacancy analysis, and market rentalrate analysis (Wincott and Mueller 1995).

Also important are the procedures, guidelines, and rules of thumb developed and currentlyemployed within the enterprise. These heuristics are likely to be based on proprietary datasuch as loan-to-value ratios for mortgages grouped by geographic area. Specific rules em-ployed by institutional investors would probably take a form similar to the following heu-ristics, which could be used for underwriting apartment building loans: (1) accept no projectswith fewer than 8 units, and prefer those with 24 units or more; (2) avoid outdoor balconyentrances (motel type); (3) value should be 6.5 to 7.5 times gross income; and (4) expensesshould range between 35 and 45 percent of gross revenue at 100 percent occupancy.

These analysis procedures may be provided in a number of specific applications (e.g., anSDSS for underwriting apartment project loans), although SDSS may also be structured astoolboxes or generators.5 Given the way in which XS and DSS are used to support decisionmaking for narrowly defined problem areas, specific SDSS make the most sense. Individualdecision makers would probably use only a small number of these specific SDSS, but becausemany of the decision procedures supported may use the same general types of data and mayneed to access the same proprietary and/or historical databases, these SDSS would be mostefficiently provided on a network through icons in a Windows-like graphical user interface.

Network Architecture

After deciding on the data entities and processes to be supported, attention can turn to thelocations at which system access will be provided. This consideration obviously interactswith data and process concerns, but it is likely to be decided more or less independently, aswith the decisions for the entities in cells 1 and 2.

At the general level, the list in cell 3 would indicate the classes of locations from which thelender operates and at which data and processing capabilities may be required. Examplesof network nodes at this level could include the home office and branch office locations. Also

5 Sprague (1980) differentiates between three different types of DSS—DSS toolboxes, DSS generators, and specificDSS—all of which are applicable in the spatial domain. DSS toolboxes are used to construct DSS generators, whichare used to develop prototype implementations quickly.

146 Kim Peterson

implied are departments within the organization itself, regardless of whether these depart-ments are geographically disparate.

At the specific level, a node would be seen as business units or aggregations of businessresources (e.g., people, facilities, responsibilities, etc.) at specific geographical locations. Thenodes shown in cell 6 include business units defined in terms of specific geography (e.g., thelender’s branch office at Santa Rosa Plaza) and organization structure (e.g., the loan un-derwriting department and property management department, both located at the homeoffice).

Technology Transfer

Implementation Issues

Given this conceptual framework for planning an enterprise-wide SDSS architecture, im-plementation issues may be briefly discussed. Considerable difficulty may be associated withimplementing any significant new computing system, and most of the enterprises that un-dertake information architecture planning are not successful (Spewak 1992). Critics contendthat failure occurs because information architecture planning proceeds from two deeplyflawed assumptions (Davenport 1994): (1) the architect has perfect control over componentsand their interaction, and (2) those components are immutable and understood throughoutthe organization. These assumptions may be reasonable in the construction of a building,but they are inconsistent with the realities of most IT environments.

Numerous obstacles have been cited as contributing to failures in enterprise architectureplanning, many of which relate to the structure of the IT department and to practices em-ployed in the planning process. For example, insufficient resources and failure to gain thefull awareness and support of top management are often cited (Spewak 1992). But morefundamental problems exist with respect to adoption of new technologies by the people whowill use them. These problems are addressed here.

Diffusion of Innovation

Technology transfer offers important insights for the planning and implementation ofenterprise-wide SDSS. It complements the architecture planning perspective that has beenused to develop this conceptual framework and offers important insights into largerorganization-wide IT issues. Many models of technology transfer are available, only one ofwhich is addressed here: diffusion of innovation.6

The guidelines proposed here follow from the work of Rogers (1983), who reviewed researchon innovations in many fields, and from the work of researchers who have focused on in-novations applied to land use planning (see, for example, Dueker 1987; Dueker and DeLacy1990; Innes and Simpson 1993). Five general guidelines are suggested to help assure effec-

6 Technology transfer has been widely studied by and is of special interest in the software engineering community,which continuously develops and attempts to apply new methodologies and management practices. In this area thenorm is constant change in approaches to developing and maintaining software, yet success in implementing newtechnologies has been limited.


tive diffusion of innovation (Rogers 1983): simplicity, observability of benefits, relative ad-vantage, ability to make small trials, and compatibility.

Simplicity. A technology must be understood by those involved in building and using it, andit must also be meaningful to them. The problem with many GIS and SDSS is that they areinherently complex and difficult to understand, especially those that are implemented asmultipurpose, multiuser systems. Some managers have tried to compensate for this com-plexity by defining GIS at the outset as tools for just one or two tasks. The often-expressedhope is that, as users become accustomed to the system, they will find more uses for it (Innesand Simpson 1993). This approach can be readily applied to SDSS implementations, becausethe DSS or XS modules that are provided in SDSS are typically designed to support a nar-rowly defined problem or issue.

Other GIS innovators attempt to introduce more complex multipurpose systems with theaid of an image or metaphor that captures the overall complexity of the system and conveysit as a simple idea. Researchers have found managers in urban planning agencies experi-menting with images for GIS that included ‘‘an architecture,’’ ‘‘an information engine, driv-ing applications from a database,’’ and ‘‘the ‘glue’ that binds departments together’’ (Innesand Simpson 1993).

Both approaches may be used to implement SDSS in large, complex enterprises. In bothcases, enhancing user understanding and building a simple, shared meaning of the systemthroughout the organization will help ensure effective diffusion of innovation.

Observability of Benefits. Success in adopting any new technology is more likely if its valuecan be seen and verified. In other words, those who adopt GIS or SDSS must know whatthey are getting and be able to assess its worth. Unfortunately, accurate prediction andassessment are difficult if there are relatively few routine activities that can be readilyautomated. Difficulties are especially likely with respect to the ill-structured real estateinvestment problems that SDSS are used to address, since a goal of SDSS is to providesupport for problems that are anything but routine (e.g., on the surface, apartment loanunderwriting may seem to be a routine task, but the range of issues is wide and varies fromone project to the next, requiring the provision of expert advice and accumulated knowledgeto analysts). In this context, it is likely that benefits will not be immediately observable assavings of time but rather as long-term benefits that accrue as a consequence of higher-quality real estate decisions.

Well-structured procedures are available for analyzing the consumption benefits of landinformation systems (see Stewart, Weir & Co. 1985) and GIS (see Budic 1994). These pro-cedures may be modified to assess the benefits of adopting SDSS. For example, one mayanalyze impacts on the separate factors of operational and decision-making effectiveness(Budic 1994). Operational effectiveness is typically defined in terms of accuracy of positionaland attribute data, availability of current data, data collection time, and accessibility ofmaps and tabular data contained in the GIS or SDSS. Decision-making effectiveness isdefined in terms of time needed to make decisions, explicitness of decisions, identificationand clarification of conflicts, communication and interpretation of information, and confi-dence in analyses generated with the GIS or SDSS.

148 Kim Peterson

Relative Advantage. Adopters of any innovation must believe that benefits will exceed costsin both monetary and human terms (Stinchcombe 1990). For implementation of GIS andSDSS, benefits and costs are likely to fall unevenly on different participants. Adoption islikely to be costly for individuals because the time required to learn the system must bediverted from directly productive tasks and because change generally induces stress. Thesecosts may be minimized, however, to the extent that the GIS or SDSS can be integrated withexisting desktop decision-making support frameworks such as spreadsheets and databasemanagement systems. If GIS and SDSS can be provided along with these nonspatial DSSin a user-friendly graphical interface (e.g., based on Windows, OS2, Macintosh, or Unix),then implementation may be easier.

Other cost savings may be realized if the enterprise can tap into government databases andGIS to access information for research and investment analysis. This approach is alreadybeing tried, with some local and state governments providing access to their GIS for infor-mation related to the preparation of environmental impact assessments, local economic con-ditions, and the laws and regulations governing land use. In addition, the U.S. Departmentof Housing and Urban Development has adopted desktop mapping software to help expeditefederal grants to cities and towns. Its Office of Community Planning and Development hasdistributed a Windows CD-ROM developed using desktop mapping software. The system isintended to eliminate much of the paperwork confronting grant applicants, urban planners,agency administrators, and individuals trying to understand how federal money is beingspent at the local level (Bowen 1994).

Ability to Make Small Trials. A fourth general guideline for technology transfer is that tech-nology should be introduced incrementally and that the changes introduced with it shouldbe reversible in the early phases. A complex technology that requires large-scale change atintroduction is unlikely to be implemented, and large-scale failures early on may derail thefunding of future implementation steps. This guideline suggests that linking GIS and SDSSwith existing desktop spreadsheets and DBMS may be an effective approach to implemen-tation. If access to enterprise proprietary or historical data files must be provided, or if paperdocument files must be converted to electronic format, these steps may require considerableresources and should probably be undertaken on a segmented basis (e.g., by geographicalregions).

Compatibility. To ensure its successful adoption, technology should be compatible with theculture, language, skills, practices, understandings, and organizational and social structuresof the organization that will use it (Innes and Simpson 1993). Again, compatibility with in-place desktop software should provide the best opportunity for successful adoption. Fortu-nately, desktop computing is well established in many large business organizations eitheras freestanding systems or via terminals attached to servers, minicomputers, or mainframesby means of networks, making adoption of GIS functions or SDSS promising from a com-patibility standpoint.

It makes sense to begin implementation by automating the existing tasks that DSS and XSmodules in the SDSS have been created to address. The next steps of applying the systemto new tasks or new ways of doing existing ones should probably follow from a formal plan-ning process, preferably enterprise architecture planning, as described earlier. For this pur-pose, steering committees have been used to build consensus for the implementation of GIS


in planning contexts (Innes 1992; Warnecke et al. 1992), and similar approaches should beconsidered for both GIS and SDSS in real estate investment contexts.

Organization-Wide IT Issues

In addition to these implementation issues, at least four organizational issues are likely toinfluence how decision support is provided throughout a large enterprise involved with resi-dential real estate and how GIS and SDSS are provided within this general framework.These issues include providing access to critical data stored in different types of enterprisedatabases, maintaining and analyzing historical data, distributing data and the processingto deal with them across multiple locations, and the adoption of object-oriented technology.

Access

Relative to column 1 in the conceptual framework (figure 3), information technology mustnot only store, maintain, retrieve, and distribute data that are acquired from externalsources and through ongoing collection and research efforts but also provide this kind ofaccess to historical enterprise information. Historical data include both proprietary andexternal data; proprietary data include, but are not limited to, operational and resourceinformation of all kinds, plus the transaction histories of enterprise customers and suppliers.For land use–related enterprises, external data include business data and census informa-tion on population and housing that are acquired from government agencies and other ex-ternal sources.

Decision-making processes may rely on critical data stored in different types of databases,on different systems, and in different locations throughout the enterprise, so building anenterprise data warehouse may become necessary. The principal objective of a data ware-house is to provide decision makers with easy access to business information, typically bycopying data obtained from operational systems and external information providers andthen loading them into separate integrated warehouse databases. These data are thenchecked for errors and reformatted for easy access and comprehension by users.

Warehouse databases are linked to desktop computers throughout the enterprise by meansof a network or system of networks (sometimes referred to as a wide area network, or WAN);decision makers can thus access any warehouse database from any computer on a networkthat is connected to it. A variety of data access software is available for processing queriesfrom the desktop and for copying results to other desktop software such as spreadsheets,databases, presentation graphics, and word processors for further analysis and reporting.

Historical Data

A second organization-wide issue related to data architecture (figure 3) involves maintainingand analyzing historical data. These data may be used for enterprise strategy formulation,problem solving, and research. Frequently, an organization has records of important trendsrelated to a decision at hand. These data may include proprietary records of project vacancyrates, rental rates, and property appreciation rates over time; historical census data and

150 Kim Peterson

national housing survey data may also be used to advantage. Analyzing these historical datato find patterns that illuminate the present is loosely described as ‘‘data mining’’ and hasbeen fueled in many enterprises by advances in quantitative analysis procedures, includingneural nets and abductive reasoning.7

Historical proprietary data may reside in paper document files as well as in magnetic format,which means that computer imaging technologies may be required. To implement a docu-ment image management system, enterprise paper files and documents must be selectedand scanned into digitized images, which are then compressed and stored on optical disk.As many as 20,000 pages of documents can now be stored on a single CD. Once stored, imagescan be retrieved in seconds and used for decision support or integrated into other informationsystems.

Distributed Processing

Given the specification of data, processes, and physical locations for enterprise-wide SDSS,system planners must decide on the computing hardware and software that will deliver GISand SDSS functions. This is a design issue and is technically beyond the scope of this article,but it influences the success of SDSS implementation and is therefore briefly discussed.

Three distributed application models may be considered (Rofrano 1992).

1. Distributed processing distributes functions or resources across two or more inter-connected processors; these processors can be any combination of mainframe computer,minicomputer, or microcomputer (programmable workstation). Distributed processing isa generic term that includes cooperative processing and client-server computing.

2. Cooperative processing describes an application whose functions are divided between amainframe processor and a programmable workstation. It is a term coined by mainframeusers and provides a host-centric view in which the workstation adds value to the main-frame by providing a better human interface and possibly additional processing.

3. Client-server computing is a term created by personal computer users to describe anapplication whose functions are divided between a programmable workstation and alocal area network (LAN) server (computer). It is a workstation-centric view in whichthe LAN server adds value to the workstation by carrying out work on its behalf. Someusers extend the definition to include enterprise servers, which are often mainframes orminicomputers that have taken on new roles. In this framework, the client is the ma-chine making the request and the server is the machine that does the work to satisfythe request.

7 Neural network software facilitates construction of complex mathematical models of the way a collection of braincells, called neurons, operates—that is, learn from experience, develop rules, and recognize patterns. The financialindustry has already adopted neural nets; traders and asset managers have been using these models for trendanalysis and pattern recognition, credit evaluation, and marketing. Abductive models are used in many of the sameproblem areas as neural nets; they employ mathematical functions to represent complex and uncertain relation-ships, and networks to break problems into manageable pieces. Unlike neural nets, the functions they use tocompute outputs from inputs may vary throughout the network.


Distributed processing issues are relevant to enterprise-wide SDSS because of the evolu-tionary path SDSS are expected to follow and because of the vast amounts of data a largereal estate investment enterprise is likely to process. The evolutionary path was specifiedearlier and is based in desktop computing and the proliferation of LANs. Vast quantities ofdata can be connected in this way, including operating histories of the real estate projectsthe enterprise has invested in, loan applications or payment histories and records of default,various types of business and economic data provided by vendors, and census data for pop-ulations and housing—all provided over time (e.g., data for the entire United States fromthe last three decennial censuses). These proprietary and external data caches may be ac-cessed on one or more warehouse database servers that are maintained at a central location(e.g., the home office) or at branch locations; all these machines may be connected througha variety of telecommunications networks.

Object-Oriented Technology

Object-oriented database (OODB) software was introduced about 10 years ago as a methodfor storing and retrieving various types of data used in computer-aided design and manu-facturing applications. The basic organizing principle in these systems is the packaging ofboth data and procedures into structures related by some form of inheritance mechanism.An OODB is therefore a database designed to store and retrieve these structures withoutreducing them to some arbitrary internal format such as columns and rows.

Object-oriented methodologies have been viewed as especially useful for spatial analysisbecause of their ability to accommodate the complexity of spatial objects and the relation-ships among them (Armstrong, Densham, and Bennett 1989; Dueker and Kjerne 1987).Object-oriented methodologies are also useful for developing prototype systems because theysupport a modular approach to application software development and provide a program-ming environment in which code reusability is an important feature. This efficiency couldbe especially beneficial for SDSS development in a large enterprise because it provides anefficient method for developing the wide array of area-specific SDSS that the enterprise mayneed. Object-oriented systems could also be used with large-scale imaging systems that maybe used to convert and manage proprietary paper document files.

But OODBs also have a negative side. Because they store complicated data and relationshipsdirectly rather than map them to relational columns and rows, OODBs may be difficult torelate to enterprise data warehouses and other relational databases. This drawback hassignificant implications with respect to SDSS, which may need to access both kinds of in-formation.8

The adoption of object orientation is currently a topic of controversy in the IT industry. Someauthorities criticize the ways in which relational and object-oriented concepts are beingintegrated (Darwen and Date 1995). Others argue that this controversy, although theoreti-cally necessary and important, is detrimental to strategy formulation relating to object ori-entation issues (Tasker and Von Halle 1995). Despite the controversy, object orientation is

8 Efforts are currently under way to address this compatibility issue, and object orientation is being investigatednot only with respect to object-relational problems but as a way to address issues of interoperability among multiple,heterogeneous databases in general (see, for example, Pitoura, Bukhres, and Elmagarmid 1995).

152 Kim Peterson

likely to have a major impact on systems development in many enterprises and to influencethe development of enterprise-wide SDSS.

Conclusions

The general conclusions to be drawn from this research are that real estate decision makingin general, and residential decision making in particular, can benefit significantly from thedevelopment of spatial decision support tools, including GIS and SDSS. Although efforts arewell under way to incorporate decision support and analytical functions into full-featuredGIS, a more promising approach follows from the evolution of SDSS from desktop DSS. Thisevolutionary path is faster and taps the considerable momentum supporting the develop-ment of desktop hardware and software, including integrated program suites that combinespreadsheets, databases, presentation graphics, and sophisticated query tools.

Numerous guidelines are available for planning this technology transfer. The principles ofinnovation diffusion suggest that provision of enterprise-wide GIS and SDSS for spatialdecision support is feasible. However, development and implementation processes must beconsidered within a larger context that includes issues related to the organization as a whole.The first of these issues involves providing access to proprietary and historical data thatmay be stored in different locations and in different forms throughout the enterprise. Thesecond issue involves the ability to maintain and analyze these data to identify hiddenpatterns. Third is the need to distribute data and processing to many different locations.And the fourth issue is the growing popularity of object-oriented technology. Taken together,these issues imply the need to consider data warehousing and imaging technologies, so-phisticated mathematical and statistical analysis procedures, distributed processing, thebenefits and drawbacks of object-oriented technology, and system development within thecontext of organization-wide IT budgets.

To develop and implement this kind of spatial decision support across multiple levels of alarge real estate enterprise, an information architecture planning process is required. Al-though this kind of planning is not without its pitfalls and problems, the benefits of effectiveand efficient development and implementation of enterprise-wide SDSS and GIS justify theeffort.

References

Abel, D. J., S. K. Yap, R. Ackland, M. A. Cameron, D. F. Smith, and G. Walker. 1992. EnvironmentalDecision Support System Project: An Exploration of Alternative Architectures for Geographical Infor-mation Systems. International Journal of Geographical Information Systems 6(3):193–204.

Andrews, Richard B. 1980. Situs Theory Part I: An Overview of Current Thought on Intra-Urban LandUse Location. Madison, WI: Center for Urban Land Economics Research.

Andrews, Richard B. 1982. Situs Theory Part III. Situs: Its Relation to Environment. Madison, WI:Center for Urban Land Economics Research.


Armstrong, Marc P., Paul J. Densham, and David A. Bennett. 1989. Object Oriented Locational Anal-ysis. Proceedings GIS/LIS ’89, 717–26. Orlando, FL: American Society for Photogrammetry and Re-mote Sensing.

Barnett, Albert P., and Ason A. Okoruwa. 1993. Application of Geographic Information Systems in SiteSelection and Location Analysis. Appraisal Journal 61(2):245–53.

Beaumont, John R. 1991a. GIS and Market Analysis. Geographical Information Systems: Principlesand Applications, ed. David J. Maguire, Michael F. Goodchild, and David W. Rhind, 139–51. New York:Wiley.

Beaumont, John R. 1991b. Spatial Decision Support Systems: Some Comments with Regard to TheirUse in Market Analysis. Environment and Planning A 23:311–17.

Belsky, Eric S., Ayse Can, and Isaac F. Megbolugbe. 1996. A Primer on GIS in Mortgage Finance. Paperpresented at conference, Research and Business Uses of Geographic Information Systems in Housingand Mortgage Finance. Fannie Mae Research Roundtable Series, Washington DC.

Bowen, Ted Smalley. 1994. Desktop Mapping to Help Cities Track Down HUD Resources. PC Week,September 26, p. 48.

Bryan, Nora Sherwood. 1993. What is ‘‘Business Geographics’’ Anyway? GIS World, February, pp. 48–49.

Budic, Zorica D. 1994. Effectiveness of Geographic Information Systems in Local Planning. APA Jour-nal, Spring, pp. 244–63.

Can, Ayse, and Isaac F. Megbolugbe. 1997. Spatial Dependence and House Price Index Construction.Journal of Real Estate Finance and Economics 14(2):203–22.

Clark, Rodney E. 1993. Market Area Analysis: Using Geocoded Customer Data and Demographics toDetermine Branch Market Areas. GIS in Business ’93 Conference Proceedings, 59–60. Boston: GISWorld.

Cooke, Donald F. 1992. Spatial Decision Support Systems: Not Just Another GIS. Geo Info Systems,May, pp. 46–49.

Dangermond, Jack. 1995. 1995 ESRI User Conference Emphasizes New Technologies. ARC News17(3):1–2.

Darwen, Hugh, and C. J. Date. 1995. Introducing the Third Manifesto. Database Programming andDesign, January, pp. 25–35.

Davenport, Tom. 1994. Maybe It’s Time to Move On. InformationWeek, September 26, p. 68.

Davis, James R., and Geoffrey McDonald. 1993. Applying a Rule-Based Decision Support System toLocal Government Planning. Expert Systems in Environmental Planning, ed. Jeff R. Wright, Lyna L.Wiggins, Ravinder K. Jain, and T. John Kim, 23–46. New York: Springer-Verlag.

Densham, Paul J., and Michael F. Goodchild. 1989. Spatial Decision Support Systems: A ResearchAgenda. Proceedings GIS/LIS ’89, 707–16. Orlando, FL: American Society for Photogrammetry andRemote Sensing.

Dueker, Kenneth J. 1987. Geographic Information Systems and Computer Aided Mapping. Journal ofthe American Planning Association 53(3):383–90.

154 Kim Peterson

Dueker, Kenneth J. 1989. An Overview of GIS Terms and Concepts. Paper presented at the conferenceDeveloping Geographic Mapping and Analysis Systems, University of Wisconsin–Madison, April 17–19, Madison.

Dueker, Kenneth J., and P. Barton DeLacy. 1990. Geographic Information Systems in the Land De-velopment Process. Journal of the American Planning Association 56(4):483–91.

Dueker, Kenneth J., and Daniel Kjerne. 1987. Application of the Object-Oriented Paradigm to Prob-lems in Geographic Information Systems. Proceedings: International Geographic Information Systems(IGIS) Symposium, ed. Robert T. Aangeenbrug and Yale M. Schiffman, 79–87. Arlington, VA: Associ-ation of American Geographers.

Epstein, Mark L., and Edgar P. Odenwalder. 1993. AM/FM/GIS Sparks Utility Reengineering. GISWorld, March, pp. 34–36.

ESRI. 1995. Oracle and ESRI Announce Joint Agreement—Spatial Database Engine Released. ARCNews 17(4):1–2.

Fotheringham, A. Stewart. 1990. One View of a Conference on SDSS: Are We Descended from theGolgafrinchams After All? Environment and Planning A 22:1137–40.

Fotheringham, A. Stewart. 1991. Some Random(ish) Thoughts on Spatial Decision Support Systems.Environment and Planning A 23:1699–1700.

Francica, Joseph R. 1993. Profiting from a Geographic Information System. GIS World, February, pp.75–84.

Goodchild, Michael F. 1992. Geographical Information Science. International Journal of GeographicalInformation Systems 6(1):31–45.

Graaskamp, James A. 1980. Fundamentals of Real Estate Development. Washington, DC: Urban LandInstitute.

Graham, Ken. 1992. Data Base Marketing with a PC-Based MCIF. Journal of Retail Banking 14(2):17–24.

Hall, Paul R. 1993. The Role of GIS in Targeted and Database Marketing for Packaged Goods. GIS inBusiness ’93 Conference Proceedings, 65–76. Boston: GIS World.

Han, Sang-Yun, and Tschangho John Kim. 1990. ESSAS: Expert System for Site Analysis and Selec-tion. In Expert Systems: Applications to Urban Planning, ed. T. John Kim, Lyna L. Wiggins, and JeffreyR. Wright, 3–13. New York: Springer-Verlag.

Innes, Judith. 1992. Group Processes and the Social Construction of Growth Management: Florida,Vermont, and New Jersey. Journal of the American Planning Association 58(4):440–53.

Innes, Judith, and David M. Simpson. 1993. Implementing GIS for Planning. Journal of the AmericanPlanning Association 59(2):230–36.

Keen, Peter G. W. 1980. Adaptive Design for Decision Support Systems. Data Base 12:15–25.

Kinsey, Ken, and Uri Avin. 1992. GIS in Action. Urban Land, March, pp. 18–21.

Kochera, Andrew. 1994. Home Characteristics and Property Insurance. Housing Economics 42(12):9–12.


Kroh, Todd. 1995. Texaco’s Next Logical Step beyond the Gas Pump. ARC News 17(4):8.

Lein, James K. 1990. Exploring a Knowledge-Based Procedure for Developmental Suitability Analysis.Applied Geography 10:171–86.

Levitt, Raymond E., and J. C. Kunz. 1985. Using Knowledge of Construction and Project Managementfor Automated Schedule Updating. Project Management Journal 16(5):57–76.

Maidment, David R., and Thomas A. Evans. 1993. Regulating the Municipal Environment Using anExpert Geographic Information System. In Expert Systems in Environmental Planning, ed. Jeff R.Wright, Lyna L. Wiggins, Ravinder K. Jain, and T. John Kim, 119–44. New York: Springer-Verlag.

Myers, Dowell, and Phillip S. Mitchell. 1993. Identifying a Well-Founded Market Analysis. AppraisalJournal 61(4):500–508.

Ortolano, Leonard, and Catherine D. Perman. 1990. Applications to Urban Planning: An Overview. InExpert Systems: Applications to Urban Planning, ed. T. John Kim, Lyna L. Wiggins, and Jeffrey R.Wright, 3–13. New York: Springer-Verlag.

Pitoura, Evaggelia, Omran Bukhres, and Ahmed Elmagarmid. 1995. Object Orientation in Multi-database Systems. ACM Computing Surveys 27(2):141–76.

Pittman, Robert, and Grant Thrall. 1992. Improving Real Estate Market Research. Real Estate Issues17(1):1–7.

Robbins, Michael L. 1996. Spatial Decision Support Systems for Automated Residential Property Valu-ation: A Discussion of an Actual Installed System. Paper presented at conference: Research and Busi-ness Uses of Geographic Information Systems in Housing and Mortgage Finance, Fannie Mae ResearchRoundtable Series, Washington DC.

Rodriguez, M., C. F. Sirmans, and Allen Marks. 1995. Using Geographic Information Systems to Im-prove Real Estate Analysis. Journal of Real Estate Research 10(2):163–72.

Rofrano, John J. 1992. Design Considerations for Distributed Applications. IBM Systems Journal31(3):564–89.

Rogers, Everett. 1983. Diffusion of Innovations. 3rd ed. New York: Free Press.

Rosser, Bill 1995. Research Notes, September 8, 1995. Gartner Group Continuous Services. Stamford,CT: Gartner Group.

Shlay, Anne B. 1988. Not in That Neighborhood: The Effects of Population and Housing on the Dis-tribution of Mortgage Finance within the Chicago SMSA. Social Science Research 17:137–63.

Simon, Herbert A. 1960. The New Science of Management Decision. New York: Harper & Row.

Spewak, Steven H. 1992. Enterprise Architecture Planning. Boston: QED Publishing Group.

Sprague, Ralph H. 1980. A Framework for the Development of Decision Support Systems. MIS Quar-terly 4(4):1–26.

Sprague, Ralph H., and Eric D. Carlson. 1982. Building Effective Decision Support Systems. EnglewoodCliffs, NJ: Prentice-Hall.

156 Kim Peterson

Stewart, Weir & Company. 1985. The D.G.S. Cost-Benefit Procedure: A Manual for Undertaking Cost-Benefit Analysis of Land Information Systems. Alberta, Canada: Alberta Bureau of Surveying andMapping.

Stinchcombe, Arthur L. 1990. Information and Organizations. Berkeley and Los Angeles: Universityof California Press.

Tasker, Dan, and Barbara Von Halle. 1995. Where the Wild Things Are. Database Programming andDesign, May, pp. 13–16.

Thrall, Grant Ian, and Paul Amos. 1996. GIS within Real Estate and Related Industries. Paper pre-sented at conference: Research and Business Uses of Geographic Information Systems in Housing andMortgage Finance, Fannie Mae Research Roundtable Series, Washington DC.

Thrall, Grant Ian, Judy Fandrich, and Susan Thrall. 1995. The Location Quotient: Descriptive Geog-raphy for the Community Reinvestment Act. Geo Info Systems 5(6):18–22.

Vachon, Michael. 1993. GIS Offers Limitless Potential to Human Service Organizations. GIS World,February, pp. 52–57.

Vonderohe, Alan P., Larry E. Travis, Robert L. Smith, and Victor Tsai. 1993. Functional Frameworkfor GIS and GIS-T Software. Section 2.3 in Final Report of NCHRP 20–27: Adaptation of GeographicInformation Systems for Transportation, pp. 21–36.

Warnecke, Lisa, John M. Johnson, Karen Marshall, and R. Steven Brown. 1992. State GeographicInformation Activities Compendium. Lexington, KY: Council of State Governments.

Waterman, Donald A. 1985. A Guide to Expert Systems. Reading, MA: Addison Wesley.

Wincott, D. Richard, and Glenn R. Mueller. 1995. Market Analysis in the Appraisal Process. AppraisalJournal 63(1):27–32.

Wright, Jeff R., and Kurt A. Buehler. 1993. Probabilistic Inferencing and Spatial Decision SupportSystems. In Expert Systems in Environmental Planning, ed. Jeff R. Wright, Lyna L. Wiggins, RavinderK. Jain, and T. John Kim, 119–44. New York: Springer-Verlag.

Zachman, John A. 1987. A Framework for Information Systems Architecture. IBM Systems Journal26(3):276–92.

Development of Spatial Decision Support Systems for ... · Development of Spatial Decision Support Systems for Residential Real Estate ... Applications of this type of SDSS are readily

Documents