MassAppraisalMethods€¦ · Kauko CarolineMcPherson: “kauko_c000” — 2008/6/5 — 16:55 — page iii — #3 MassAppraisalMethods Aninternationalperspectiveforpropertyvaluers

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page i — #1

Mass Appraisal Methods

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page ii — #2

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page iii — #3

Mass Appraisal Methods

An international perspective for property valuers

Edited by

Tom KaukoDepartment of GeographyNorwegian University of Science and Technology (NTNU)TrondheimNorwayandOTB, Delft University of TechnologyDelftThe Netherlands

Maurizio d’AmatoArchitecture and Planning DepartmentTechnical University Politecnico di BariBariItaly

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page iv — #4

This edition first published 2008© 2008 Blackwell Publishing Ltd

Blackwell Publishing was acquired by John Wiley & Sons in February 2007. Blackwell’spublishing programme has been merged with Wiley’s global Scientific, Technical, and Medicalbusiness to form Wiley-Blackwell.

Registered officeJohn Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ,United Kingdom

Editorial offices9600 Garsington Road, Oxford, OX4 2DQ, United Kingdom2121 State Avenue, Ames, Iowa 50014-8300, USA

For details of our global editorial offices, for customer services and for information about howto apply for permission to reuse the copyright material in this book please see our website atwww.wiley.com/wiley-blackwell.

The right of the author to be identified as the author of this work has been asserted inaccordance with the Copyright, Designs and Patents Act 1988.

All rights reserved. No part of this publication may be reproduced, stored in a retrievalsystem, or transmitted, in any form or by any means, electronic, mechanical, photocopying,recording or otherwise, except as permitted by the UK Copyright, Designs and Patents Act1988, without the prior permission of the publisher.

Wiley also publishes its books in a variety of electronic formats. Some content that appears inprint may not be available in electronic books.

Designations used by companies to distinguish their products are often claimed astrademarks. All brand names and product names used in this book are trade names, servicemarks, trademarks or registered trademarks of their respective owners. The publisher is notassociated with any product or vendor mentioned in this book. This publication is designed toprovide accurate and authoritative information in regard to the subject matter covered. It issold on the understanding that the publisher is not engaged in rendering professional services.If professional advice or other expert assistance is required, the services of a competentprofessional should be sought.

Library of Congress Cataloging-in-Publication DataMass appraisal methods: an international perspective for property valuers/editedby Tom Kauko, Maurizio d’Amato.

p. cm. — (Real estate issues)Includes bibliographical references and index.ISBN-13: 978-1-4051-8097-9 (hbk.: alk. paper)ISBN-10: 1-4051-8097-8 (hbk.: alk. paper)1. Real property—Valuation. I. Kauko, Tom. II. d’Amato, Maurizio.

HD1387.M369 2008333.33′2—dc22

2008006134

A catalogue record for this book is available from the British Library.

Set in 10/13 pt TrumpMediaeval by Newgen Imaging Systems Pvt Ltd, ChennaiPrinted in Singapore by Fabulous Printers Pte Ltd

1 2008

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page v — #5

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page vi — #6

Books in the series

Greenfields, Brownfields & HousingDevelopmentAdams & Watkins9780632063871

Planning, Public Policy & PropertyMarketsEdited by Adams, Watkins & White9781405124300

Housing & Welfare in SouthernEuropeAllen, Barlow, Léal, Maloutas &Padovani9781405103077

Markets and Institutions in RealEstate & ConstructionBall9781405110990

Neighbourhood Renewal and HousingMarketsEdited by Beider9781405134101

Mortgage Markets WorldwideBen-Shahar, Ong & Leung9781405132107

The Cost of Land Use DecisionsBuitelaar9781405151238

Urban Regeneration in EuropeCouch, Fraser & Percy9780632058412

Urban SprawlCouch, Leontidou & Petschel-Held9781405151238

Real Estate & the New EconomyDixon, McAllister, Marston & Snow9781405117784

Economics & Land Use PlanningEvans9781405118613

Economics, Real Estate & the Supply of LandEvans9781405118620

Development & DevelopersGuy & Henneberry9780632058426

The Right to BuyJones & Murie9781405131971

Economics of the Mortgage MarketLeece9781405114615

Housing Economics & Public PolicyO’Sullivan & Gibb9780632064618

International Real EstateSeabrooke, Kent & How9781405103084

British HousebuildersWellings9781405149181

Forthcoming

Building Cycles & Urban DevelopmentBarras9781405130011

Transforming the Private LandlordCrook & Kemp9781405184151

Housing Markets & Planning PolicyJones & Watkins9781405175203

Towers of CapitalLizieri9781405156721

Property Market Analysis & ForecastingMcGough & Tsolacos9781405121224

Affordable Housing & the Property MarketMonk & Whitehead9781405147149

Property Investment & FinanceNewell & Sieracki9781405151283

Housing Stock TransferTaylor9781405170321

Real Estate Finance in the New EconomicWorldTiwari & White9781405158718

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page vii — #7

Contents

Contributors ixForeword xvPreface xxiAbbreviations xxiii

1 Introduction: Suitability Issues in Mass Appraisal Methodology 1Tom Kauko and Maurizio d’Amato

PART I MASS APPRAISAL PRACTICE ANDRECOMMENDATIONS 25

2 Data Issues Involved with the Application of AutomatedValuation Methods: A Case Study 27John F. Thompson, Jr

3 The Modified Comparable Sales Method as the Basis for a PropertyTax Valuations System and its Relationship and Comparison toSpatially Autoregressive Valuation Models 49Richard A. Borst and William J. McCluskey

4 Automated Valuation in the Dutch Housing Market: TheWeb-Application ‘MarktPositie’ Used by NVM-Realtors 70Dree Op ’t Veld, Emma Bijlsma and Paulien van de Hoef

5 Using Fuzzy Numbers in Mass Appraisal: The Case of theBelarusian Property Market 91Maurizio d’Amato and Nikolai Siniak

PART II CURRENT ADVANCED METHODS 109

6 Mass Appraisal, Hedonic Price Modelling and UrbanExternalities: Understanding Property Value Shaping Processes 111François Des Rosiers and Marius Thériault

7 Residuals Analysis for Constructing ‘More Real’ Property Value 148Małgorzata Renigier

8 The Hierarchical Trend Model 164Marc K. Francke

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page viii — #8

viii Contents

PART III EMERGING METHODS 181

9 Developing Mass Appraisal Models with Fuzzy Systems 183Marco Aurélio Stumpf González

10 Utterly Unorthodox Modelling for the Purposes of MassAppraisal: An Approach Based on Patterns and Judgments 203Tom Kauko

11 Rough Set Theory as Property Valuation Methodology:The Whole Story 220Maurizio d’Amato

PART IV COMPARISON OF TOOLS USING A SET OFSPECIFIC CRITERIA 259

12 Technical Comparison of the Methods Including Formal Testingof Accuracy and Other Modelling Performance Using Own DataSets and Multiple Regression Analysis 261Richard A. Borst, François Des Rosiers, Małgorzata Renigier,Marco Aurélio Stumpf González, Tom Kauko and Mauriziod’Amato

13 Property Market Classification and Mass Appraisal Methodology 280Maurizio d’Amato and Tom Kauko

PART V CONCLUSION 305

14 Automated Valuation Methods, Empirical Modelling of Value,and Systems for Market Analysis 307Tom Kauko

Glossary 321Index 325

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page ix — #9

Contributors

Emma Bijlsma has a longstanding career as a statistician and developer ofmarket information and decision support systems with both TNO (DutchNational Organisation for Applied Scientific Research) and MomentumTechnologies. She is one of the founders of that company. Emma wasresponsible for much of, or more accurately, the heart of, the estimationmethodology behind ‘MarktPositie’, the web-based valuation tool appliedby members of NVM (Netherlands Organisation of Real Estate Agents).

Richard A. Borst has been engaged in managing the design, development andimplementation of computer-based real property information systems since1973. He was president of North America’s largest mass appraisal firmwhileat the same time maintaining his contributions to the technical aspects ofmass appraisal systems. His technical background is evidenced in a numberof published articles and conference presentations. He introduced artificialneural networks to the assessment community in 1990. He was appointedto a three-year term in 1997 as a Visiting Research Fellow at the Universityof Ulster, Belfast, Northern Ireland. During this tenure he collaborated withmembers of the faculty at the University performing research in the fieldsof valuation modelling and the application of location effects in the modelstructuring and calibration process. He obtained a Doctor of Technologyfrom the University of Ulster, Northern Ireland, a Master of Science inIndustrial Engineering from the State University of New York at Buffaloand a Bachelor of Engineering Science, with honours, from the ClevelandState University.

Maurizio d’Amato is Associate Professor at the 1st Faculty of Engineer-ing, Technical University Politecnico di Bari, Italy, where he teaches realestate valuation. He completed his undergraduate work in Economics at theUniversity of Bari and worked for several banks in real estate finance beforeentering the doctoral program in Planning, specializing in valuation meth-ods, at the Politecnico di Bari. After completing this program, he served ascontract professor in Real Estate Valuation for three years. During this timehe received research grants from the Italian Council of Research (CNR) forprojects undertaken at the University of Florida in 1997, 1998 and 1999 andat the University of Alicante Spain in 2000. He has been scientific director ofthe Real Estate Center of Italian Association of Real Estate Counselor (AICI).He is scientific director and founder of Real EstateMarket Observatory of theTechnical University Politecnico di Bari. He has been a Member of ExpertCommittee of Tecnoborsa. He was appointed Fellow Member of the RoyalInstitution of Chartered Surveyors in 2004.

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page x — #10

x Contributors

François Des Rosiers is an economist and urban planner and holds a PhDdegree in Urban and Regional Planning Studies from the London Schoolof Economics, UK (1984). Since 1976, he has been teaching Urban andReal Estate Management at the Faculty of Business Administration, LavalUniversity, where he is a full professor and currently in charge of the Urban&Real EstateManagement Program. In addition to his teaching and researchtasks, François Des Rosiers has served over the past twenty years as a con-sultant and adviser for several public, semi-public, institutional and privatebodies and organizations, both in Quebec, Canada and abroad and has beeninvolved in various policy issues through numerous committees, commis-sions, panels, boards, etc., at the national, provincial and local levels. Hismain research area deals with hedonic price modelling and the measure-ment of urban externalities. Professor Des Rosiers’s academic realizationshave earned him several awards and manuscript prizes. In July 2007, he waselected Fellow of the Academy for Social Sciences of the Royal Society ofCanada.

Marc K. Francke graduated in 1994 in Econometrics at the Vrije Univer-siteit Amsterdam. Mark Francke worked at the Amsterdam Tax Authoritiesoffice during 1994–2000, where he developed models for mass appraisalof real estate. In 2001 he was co-founder of OrtaX, a company that spe-cialized in mass appraisal for local government and housing corporations.In 2006 he defended his PhD dissertation entitled ‘Marginal Likelihoodin State-space Models, Theory and Applications’. In 2006 he became apart-time Assistant Professor in the Econometrics Department at the VrijeUniversiteit Amsterdam. He has published in the Journal of Property TaxAssessment and Administration, the Journal of Business and EconomicStatistics, the Journal of Real Estate Finance and Economics and the Journalof Econometrics.

Marco Aurélio Stumpf González is Professor and Researcher at CivilEngineering Post Graduate Programme at Universidade do Vale do Rio dosSinos (UNISINOS), where he has been teaching Real Estate Valuation since1996. His research interests cover housing, sustainable buildings, economicsof building renewal, market analysis, taxation of real estate, valuation andfinancial analysis of real estate, and alternative property valuation tech-niques (especially fuzzy rules and artificial neural networks models). In 1993he obtained MSc in Engineering at Universidade Federal do Rio Grande doSul (UFRGS), with major in Real Estate Valuation. In 1998 he became a PhDcandidate at UFRGS (Civil Engineering), and completed his dissertation in2002.

Tom Kauko has been Associate Professor in Urban Geography at the Nor-wegian University of Science and Technology (NTNU) since 2006. Hisresearch interests cover housing market analysis; evaluation of planning

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page xi — #11

Contributors xi

and urban regeneration; locational quality in housing consumption; insti-tutional, evolutionary and behavioural property research, and alternativeproperty valuation techniques. In 1994 he obtained his MSc in Land Sur-veying at Helsinki University of Technology (HUT), with a major in RealEstate Economics and Valuation. During 1995–96 he worked as a plannerfor the research department of National Land Survey of Finland in Helsinki(Maanmittauslaitos). During 1996–97 he participated in a course in physi-cal planning at the Centre for Urban and Regional Research of HUT (YTK).He subsequently moved to Utrecht University, Faculty of Geographical Sci-ences, where he completed his dissertation in June 2002. From 2001to 2006he worked as a researcher at OTB, Delft University of Technology, where hestill has visitor’s affiliation.

William J. McCluskey is presently a Reader in Real Estate and Valuationat the University of Ulster. He has held various international positionsincluding Visiting Professor of Real Estate at the University of Lodz, Polandand Professor of Property Studies at Lincoln University, Christchurch,New Zealand. His main professional and academic interests are in thefields of real estate valuation and more specifically ad valorem propertytax systems, local government finance, computer assisted mass appraisal(CAMA) modelling and the application of geographic information systems(GIS). Within this context he has been involved in a number of interna-tional projects advising on ad valorem property tax issues including Jamaica,Northern Ireland, Bermuda, Poland, Kosovo, Tanzania and South Africa.

Dree Op ’t Veld has been responsible for the design and development of aconsiderable number of web-based market information and decision supportsystems in the Netherlands. Throughout his career the emphasis has beenon the analysis and modelling of regional economic and demographic phe-nomena for policy purposes by government officials or firms. Thematicallyhis interest nowadays is in housing and real estate markets and valua-tion on the one hand, and labour demand and supply on the other. Hereceived his Master’s degree in urban geography and planning, specializ-ing in statistical analysis and econometric modelling from the Universityof Nijmegen, the Netherlands. After a twenty year career at TNO – theDutch National Organisation for Applied Scientific Research – he startedthe company Momentum Technologies in 2000 together with a number ofcolleagues.

Małgorzata Renigier (aka Renigier-Biłozor) has been Assistant Professor inthe Department of Real Estate Management and Regional Development atthe University of Warmia and Mazury, Olsztyn, Poland since 2005. Hermajor fields of research interest comprise systems of real estatemanagement,value forecasting, non-linear analysis inmodelling of real estate value, influ-ence analysis of stochastic factors on the real estate value, and application

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page xii — #12

xii Contributors

of artificial intelligence (AI) in real estate management. In 2000 she obtainedher MSc in the faculty of Geodesy and Space Management at the Universityof Warmia and Mazury. During 2001 she began her doctorate studies at theDepartment of Real Estate Management and Regional Development at thesame university. In 2004 she obtained her PhD in Geodesy and Cartogra-phy. In 2006 she received a prize from the Polish Minister for Building andTransport for her PhD dissertation. Since 2006 she has been a member ofthe board of the Scientific Society of Real Estate. From 2004 to 2007 she hasbeen a co-author of the programme concerning creation of a managementsystem of real estate sources owned by local government, of the committeeof scientific research.

Nikolai Siniak (aka Mikalai) is Associate Professor at the Department ofEconomy and Management of Enterprises at the Belarusian State Techno-logical University (BSTU), Minsk, where he has been working since 1998.His scientific interests cover economics, valuation and management of realestate, restructuring of enterprises, optimization of production programmes,and simulation and formulation of economic problems on economy. AtBSTU he obtained the following degrees: Diploma in Mechanical Engineer-ing (1995), PhD in Economical science (1998), and Diploma of AssociateProfessor in Economics (2005). He has developed grounding methods for afurniture factory production programme, the concept of real estate valuationfor enterprises, and real estate market analysis. He has more than 70 publi-cations to his credit. His teaching activities comprise the courses ‘Economyof enterprise’ and ‘Methods of branch property valuation’.

Marius Thériault is a geographer and Professor at the Graduate Schoolof Land Planning and Regional Development, Laval University, Quebec,Canada, where he has been teaching GIS, spatial analysis, applied statistics,applied computing and climatology since 1979. His research activities areabout combining GIS, spatial statistics, GPS and spatio-temporal modellingto further studies related to real estate markets, mobility behaviour, acces-sibility to urban amenities, environmental impacts assessment and urbanplanning. From 2000 to 2006 he was Director of the Land Planning andRegional Development Research Centre at Laval University. He is currentlyinvolved in several research networks and peer reviewed journals in Canadaand Europe. Findings of his research are published in more than 40 journalsrelating to urban economics, housing, geography, geographical informationscience, transportation planning, computer science, environment, healthand regional development.

John F. Thompson, Jr has over 30 years of experience in designing, develop-ing and implementing computer-assisted mass appraisal (CAMA) systems.He has been responsible for numerous systems development and applicationprojects, including management and direction of the Company’s systems

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page xiii — #13

Contributors xiii

activities on CAMA implementation projects in 25 states in the USA andseveral provinces in Canada. His background includes computerization ofa variety of valuation manuals, including state manuals in 10 states, andexperience in modelling and valuation analysis in scores of communities. Inaddition, he has provided training and consulting services to assessment staffthroughout the USA. He has also written articles for various InternationalAssociation of Assessing Officers (IAAO) publications and contributed sev-eral chapters to the IAAO Property Valuation Assessment textbook, editedCAMAnewsletters, and presented papers on a variety of topics at assessmentconferences. His academic training includes appliedmathematics, computersoftware, operations research, and environmental engineering. He receivedboth a BS and an MS in mathematics from Rensselaer Polytechnic Institutein 1973.

Paulien van de Hoef studied econometrics at the University of Groningen,the Netherlands. After graduating, she specialized in the development ofcomputer systems for decision support in a number of fields, amongst themmarketing and media. Since 2004 she has worked at Momentum Tech-nologies, where she plays an important role in the analysis of the housingmarket and the development ofweb-basedmarket information and valuationsystems.

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page xiv — #14

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page xv — #15

Foreword

This book provides an authoritative outline of current mass appraisaltechniques being used internationally and in-depth research into state-of-the-art developments that are likely to permeate the industry over thenext decade. The editors indicate that the ‘book explores the possibilitiesfor developing mass appraisal methods, following two different arguments:firstly, that the performance and feasibility of appraisalmethods be comparedand evaluated with regard to a set of technical criteria, and secondly, thatdifferences in the suitability of methods also have to do with the particularcontext where application takes place.’Each chapter’s author(s) select a particular aspect of mass appraisal and

provide empirical analysis on a variety ofmodelling techniques. The researchpublished within the various chapters has addressed the book’s three coreobjectives: firstly, to understand the problem of advancing mass appraisalmethods/expertise from both the perspective of scientific debate and thepractical feasibility; secondly, to evaluate a set of heretic and orthodoxmeth-ods based on a set of specific criteria, partly technical/practical and partlyinstitutional; and finally, to establish an international platform for broadernetworking within this realm. The research has clearly met the objectivesand has taken this area of ‘mass valuation/appraisal’ to the next level interms of international significance. The various chapters provide a rigor-ous analytical framework within which various modelling paradigms aredeveloped.In the first chapter, Kauko and d’Amato set out to explore the possibil-

ities within this under-theorized problem area. In essence, mass appraisalmay be defined as a systematic appraisal of groups of properties using stan-dardized procedures where the methodologies normally being applied referto large groups of properties rather than to a single property. Currently, tworelated modelling traditions exist, both of which deploy multiple regres-sion analysis (MRA) for estimation, the model-driven hedonic approach, andthe data-driven statistical approach. The authors contend that MRA-basedhedonic price models may not be suitable for capturing all the necessaryinformation involved in the value formation process, and the literature onhow to develop the value modelling tools further is at best evolving. Onthe other hand, there are property markets where data are not available, or,even if such data are available, it may not be organized into a property databank. In this case, other types of automated valuation methods (AVMs) witha less deterministic relation between property value and property attributesmay be of assistance. Although the problems are highlighted, MRA remainscurrently the most important theoretical framework in mass appraisal.

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page xvi — #16

xvi Foreword

The authors suggest that when examining current trends of valuationmodelling research, a likely scenario would be that price research is goingtowards spatial tools on one hand, and on the other hand towards pragmatictools. In terms of this trend, according to the authors, there are two differentarguments concerning advancement: firstly, following the orthodox view, toremain within equilibrium modelling and sophisticated econometric tech-niques in order to align with the academic tradition, and secondly, followingamore heretic/heterodox view, to only consider the practical aspects of accu-racy and feasibility which would permit looking beyond hedonic modellingextensions towards computer simulation, artificial intelligence andmachinelearning paradigms.Thompson, inChapter 2, provides a review of property data issues and then

develops amass valuation case study of a jurisdiction in theUSA. The authorrightly contends that the performance of automated models and valuationmethodologies are related to a number of factors, the principal one being thequality of the data and its fitness for use in the modelling/valuation process.This applies not just to the attributes, but also the coding of the sales or mar-ket transactions as to their appropriateness for use in themarket analysis andvaluation processing. Data is crucial to the mass appraisal process and oftenthe sources of data need to be investigated so as to have confidence in theirquality, accuracy and applicability. Often agencies (private and government)may have captured information of limited use as it was not collected forvaluation purposes and may be incomplete. The author also suggests thatthe performance of models varies depending upon the nature of the popu-lation(s) of property being modelled, the ready availability of market dataon property in setting asking prices, the availability of housing (numbersof properties available for purchase) and overall frequency of sales activity.These all feed into the overall uniformity/efficiency of the market and willimpact on the performance statistics generated against the sales transactionbase to evaluate the performance of the AVM.Borst and McCluskey in Chapter 3 state that the traditional comparable

sales method of valuation (CSM) is widely used in the USA, the UK andindeed many other countries for the valuation of residential properties. Inthese jurisdictions, there is a clear and identifiable relationship betweenCSM as practised by mass appraisers and the recent developments in spa-tially aware valuation models. The authors of this chapter take the positionthat the CSM and modified comparable sales method (MCSM) are certainlyamong the best, if not the best, methods for mass appraisal. Other potentialtechniques for best practice would include the use of a well structured linearor non-linear model calibrated for an entire jurisdiction, market segmenta-tion models, the ‘response surface’ method, and a collection of advancedmodelling techniques including geographically weighted regression (GWR).

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page xvii — #17

Foreword xvii

This chapter provides a detailed comparative analysis of a selected subset ofthese techniques to the CSM and MCSM.In Chapter 4, Op ’t Veld, Bijlsma and Van de Hoef explain the charac-

teristics of the Dutch housing market and the role the realtors or estateagents play in the valuation process. Clearly, this is an important role as theagents are involved in the sale of approximately 90% of all dwellings soldin the Netherlands. This system is heavily reliant on quality data which isprovided by and shared amongst the participating agents. Through a web-application, an automated valuation of any house can be provided given itsunderlying data bymeans of a hedonic pricemodel. The system is designed toallow all member-offices of the Dutch association of realtors, NVM, to shareinformation on the houses they have on offer for sale through the web-basedTIARA-communication system. ‘MarktPositie’ is essentially the valuationframework that is used by realtors to value one house at a time. However, inits present form it cannot readily be used for the automated mass valuationof the entire housing stock of a municipality or region. This is primarilybecause the data for the entire housing stock of a municipality are not com-prehensive. For this system to work within the mass appraisal context forall dwellings, data on properties that have not been sold would need to becollected.The authors of Chapter 5 (d’Amato and Siniak) consider that the property

market is often viewed as inefficient due to its heterogeneity, illiquidityand high transaction costs, and considerably more inefficient in those tran-sitional democracies of Eastern Europe where confidence in property datais relatively low. As a consequence, property valuation techniques need todeal with a higher level of uncertainty than, for example, the stock mar-ket. The authors attempt to demonstrate the application of ‘fuzzy logic andnumbers’ to property valuation and property investment appraisal. It is gen-erally accepted that the use of fuzzy logic techniques can assist in deriving amathematical solution in cases of high levels of uncertainty. This researchindicates an interesting comparison between two different approaches inrelation to uncertain information; these are, the most probable selling price,and the most possible selling price.Des Rosiers and Thériault in Chapter 6 provide a detailed exposition of

the application of hedonic modelling for estimating the implicit prices ofattributes pertaining to heterogeneous goods. The authors take the view that,given the highly complex nature of real estate which consists of buildings,land and neighbourhood characteristics, it soon became a prime target forhedonic price modelling where it was applied for various purposes, amongwhich was the estimation of the marginal contribution of urban externali-ties – namely environmental ones – to market values. This chapter looks atthe rationale and conceptual framework underlying the hedonic approach;

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page xviii — #18

xviii Foreword

issues around the functional form issue which are addressed with partic-ular reference to the measurement of proximity effects on house values;innovative ways to account for a major determinant of property prices,namely accessibility to urban services; the problems of spatial dependenceare investigated and alternate modelling procedures designed at handlingspatial autocorrelation are discussed.InChapter 7, Renigier claims that residuals are phenomenawhich fail to be

explained by mathematical models because of a lack of the relevant statisti-cal data, incomplete theoretical and practical knowledge about themodelledarea and/or because of the dynamic, frequently unforeseeable, tempo ofchange of the study area. The author has put forward a number of consider-ations to provide grounds for the claim that integration of geo-deterministicinference (represented by the geostatistical model) and geo-stochastic infer-ence (represented by maps of residuals), linked to space valuation, makes itpossible to dynamically diagnose and characterize spatial phenomena andto make rational forecasts (and, consequently, planning) of changes in spaceand the real estate market.Francke in Chapter 8 presents a time series model for selling prices of

houses, called the hierarchical trend model (HTM). This statistical approachhas been successfully used in the Netherlands for the valuation of approxi-mately one million houses for property tax purposes. In essence, a structuraltime-series model for house prices is described that has proven its value foralmost a decade. The author argues that the main strengths of the HTM arethe modelling of the time dependence of the selling prices and the sophis-ticated way of modelling the housing characteristics. In addition, the HTMalso addresses the problem of spatial dependence of the selling prices, but ina rather straightforward way with some parameters varying over time andother parameters remaining constant over time.In Chapter 9, González argues that there are several deficiencies with the

use of regression techniques in mass appraisal, for example, real data haveseveral sources of error or imprecision, such as the lack of correct spec-ification of model format, multiple simultaneous relationships among theexplanatory variables, and rather imprecise transitions between submarkets.As an alternative, the author has approached the problemby developingmoreflexible and comprehensive models such as those based on fuzzy systems.Research does, however, conclude that fuzzy systems on their own may notreplicate market characteristics and therefore they are applied in associationwith other techniques, such as artificial neural networks (ANNs) and geneticalgorithms (GAs), performing hybrid systems.In Chapter 10, Kauko deals with two modelling techniques; firstly, the

self-organizing map (SOM, also known as the Kohonen Map), and secondly,the analytic hierarchy process (AHP, also known as the Saatymethod of elic-itation). While being fundamentally different, the two methods potentially

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page xix — #19

Foreword xix

complement more conventional methods of data analysis. The research dis-cussion shows the possibilities and limitations for using the two proposedapproaches for ‘heretic’ mass appraisal. Arguably, the author contends thatthe approaches put forward are better suited for situations where the marketplace is affected by frictional factors, monopoly price or other circumstances,when various externalities or simply human behaviour are suspected to havean effect on values, but the extent of which is yet to be comprehended andassessed in monetary terms.d’Amato in Chapter 11 presents a summary of the application of rough

set theory (RST) to mass appraisal. The author has applied the methodologyto a small sample of real property transactions in the residential real estatemarket of Bari, Italy. Essentially, RST’s application to property valuationdoes not require a model or underlying assumptions. The valuation processis directly obtained from observation, in this case the valuation process relieson an internal knowledge only. The chapter then provides a comparison ofMRA and RST carried out on a sample of 7107 observations located in theCatawbaCounty inNorth Carolina. The results of thework demonstrate thepotential that RST has in providing relatively accurate predictions of value.In Chapter 12, the authors (Borst, Des Rosiers, Renigier, González, Kauko

and d’Amato) test various modelling paradigms on data with a view todiscerning the most ’appropriate’ in terms of predictive accuracy. Eachresearcher has used their own data sets and attempted to evaluate differentapproaches in terms of measures such as error, absolute error, coefficient ofdispersion (COD), etc.In Chapter 13, d’Amato and Kauko propose a protocol for the applica-

tion of mass appraisal valuationmethodology based on selected criteria. Theauthors contend from the empirical research so far, it can be concluded that amass appraisal methodologymay be rigorous in itself, but institutionally notsuitable for the application in a given property market context. An analysisof market features should therefore be the premise for the application of aspecificmass appraisal valuationmethodology whereMRAmay be justified,or the premise for the use of an alternative method in an emerging market.The final chapter (14) by Kauko provides a critical review with brief syn-

opses of the various chapters. In this chapter, the author suggests that thefuture direction of research may be towards the use and application of a vari-ety ofmethods in different propertymarket contexts, in order to better definea paradigm of suitability of a method for each property market.The contribution that this book provides to the existing body of knowl-

edge lies in its applicability to both industry and academia. The use of massvaluation techniques does not lie exclusively within the domain of propertytax assessment departments. The techniques outlined in this book are beingaggressively developed by financial institutions in their quest to provide fastand efficient AVM solutions in real time. The book provides an insight into

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page xx — #20

xx Foreword

the traditional/conventional or pragmatic techniques as well as some of theother more ‘academic’ and cutting edge approaches. The discipline of massappraisal has been evolving over the last 50 years and clearly, as this bookdemonstrates, considerable development has already taken place.

Professor Alastair AdairUniversity of Ulster

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page xxi — #21

Preface

Many of us have an urge to explore the unknown – in real estate marketmodelling too. Which method is best in something, and why? This questionis currently overlooked due to one reason or another: inertia in research orga-nization and, in general, not being a particularly trendy topic. This havingbeen said, we set out to explore the limits of the methodological aspect ofmass appraisal. We have amassed a group of creative yet systematic people,who have not written the last word on the topic yet.The potential of informal meetings cannot be overstated; the first time

we – the Editors of this book – met was in the European Real Estate Soci-ety (ERES) conference in hot Alicante, Spain, in the summer of 2001; atthat meeting, we also met two other contributors of this volume: namely,François Des Rosiers and Nikolai Siniak. The first time we truly begansketching the idea of a book on methods was two years later in the cor-responding conference in Helsinki, Finland, where we had a talk after asauna, by the lake in the forest. Then we continued with a meeting in Bari,Italy, in November 2004, and the final meeting in a seminar in Delft, theNetherlands, in November 2006. By that time we had managed to recruit aset of new enthusiastic members, many of whom where committed enoughto embark on a book project. This last of the meetings proved so fruitful sowe arranged a new meeting in Delft in the year 2007 too.Alongside this growth process, we have gathered a headstrong group of

innovative thinkers, many of whom are co-authors in this book. Togetherwith these colleagues, we comprise a wide cultural variety with large dis-ciplinary differences, but what is common to us is a belief in this line ofinquiry. We still think it is an exciting endeavour and now also work welldone. We would like to thank all the contributors, reviewers, and the RoyalInstitution of Chartered Surveyors (RICS) andOTB for their support, and Pro-fessor Alastair Adair, Professor Alexis Tsoukias, Professor Marco Simonotti,ProfessorNeil Crosby, andmany others for their useful help and suggestions.We also wish to acknowledge the financial assistance of the Dutch govern-

ment through the Habiforum Program for Innovative Land Use and DelftUniversity of Technology through the Delft Centre for Sustainable UrbanAreas.Last but not least, we would like to thank our spouses for their patience,

which is rapidly diminishing each time we start writing a book.

Tom KaukoMaurizio d’Amato

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page xxii — #22

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page xxiii — #23

Abbreviations

ACI Average consistency indexAHP Analytic hierarchy processAI Artificial intelligenceANN Artificial neural networksAVM Automated valuation methodCAMA Computer-assisted mass appraisalCBD Central business districtCDU Condition, desirability and usefulnessCI Consistency indexCOD Coefficient of dispersionCOV Coefficient of variationCR Consistency ratioCSM Comparable sales methodCV Contingent valuationEGLS Estimated generalized least squaresFAGNIS Fuzzy automatically generated neural inferred systemFIS Fuzzy inference systemFRBS Fuzzy rule-based systemsGA Genetic algorithmGAM Generalized additive modelGBA Gross building areaGIS Geographic information systemsGSRA Global response surface analysisGWR Geographically weighted regressionHTM Hierarchical trend modelHVTL High-voltage transmission lineIAAO International Association of Assessing OfficersIAS International Accounting StandardsIFRS International Financial Reporting StandardIVS International Valuation StandardLISA Local indicators of spatial autocorrelationLVRS Location value response surfaceMAC Main activity centresMAPE Mean absolute percentage errorMAVT Multi-attribute value treeMBV Market basket valueMCDM Multi-criteria decision methodMCSM Modified comparable sales methodMIDAS Market Information Database and Application System

Kauko Caroline McPherson: “kauko_c000” — 2008/6/5 — 16:55 — page xxiv — #24

xxiv Abbreviations

ML Maximum likelihoodMLP Multi-layer perceptronMLS Multiple listing servicesMRA Multiple regression analysisNBA Net building areaNFS Neuro-fuzzy systemsNGBA Net/gross building areaNVM Netherlands Organization of Real Estate AgentsNZ Non-zeroO–D Origin–destinationOLS Ordinary least squaresPCA Principal component analysisPPMRA Piecewise parabolic multiple regression analysisQOL Quality of lifeREM Real estate managementRG Repertory gridRMSE Root of mean square errorRSA Response surface analysisRST Rough set theorySA Spatial autocorrelationSAR Spatial autoregressionSARS Simultaneous autoregressive specificationSEM Spatial expansion methodSFLA Square feet of living areaSLM Spatial lag modelSM Spatial managementSOM Self-organizing mapSP Sale priceTSK Takagi–Sugeno–Kang (rules)UNECE United Nations Economic Commission for EuropeVICS Value influence centresVTR Valued tolerance relationWTA Willingness to acceptWTP Willingness to pay

Kauko Caroline McPherson: “kauko_c001” — 2008/6/6 — 10:23 — page 1 — #1

1Introduction: Suitability Issues inMass Appraisal Methodology

Tom Kauko and Maurizio d’Amato

Note: In Italian academia, the system of output registration requires jointly publishedarticles to include an indication of the parts written by each author. Although thework was carried out in strict cooperation between the two authors, we may approx-imate that most of the second, third, fourth and fifth sections were written by TomKauko while most of the first, sixth and seventh sections were written by Mauriziod’Amato.

Background

Currently, quantitative methods in social and economic sciences areundergoing a renaissance; after at least three decades of being dubbed bor-ing and stale, the methods have resurged into the mainstream again. Thisis particularly true for spatial sciences and human geography. A variety ofreasons for this can be noted: an improvement in computing technology,better data registers and availability, and the launch of end-user friendlysoftware. One more specific reason is the need to standardize the proceduresof property valuation and mass appraisal. This is a consequence of economicglobalization – today a homogenous valuation/appraisal method is requiredall over the world. The International Accounting Standards/InternationalFinancial Reporting Standard (IAS/IFRS) and the International ValuationStandard (IVS) address this important issue. Because of the expansion ofbusiness activity across national borders it is becoming increasingly indis-pensable to propagate consistent approaches to the valuation of real estatefor accounting, banking activity, stock exchange (bourse) listing and securedlending purposes (Mansfield and Lorenz, 2004). This is articulated in theGuidance Note n. 13 of IVS (I.V.S.C., 2005) 1.7 as follows: ‘The Development


2 Mass Appraisal Methods

of Mass Appraisal Systems for property taxation should follow recognizedscientific standards in statistical applications. . .’, which in turn has ledto particular emphasis being given to a multiple regression analysis (MRA)approach.

On top of the scientific and business motivations above, one can also findsome planning and policy relevant grounds for the use of empirical modellingtechniques and automated valuation methods (AVM). It is widely recognizedtoday that the development of urban areas requires sustainability. It is alsoincreasingly being recognized that this concept is multi-dimensional, involv-ing ecological, environmental, economic, social and cultural aspects such asviability and quality-of-life (QOL). With the help of mass appraisal methodsit is possible to extract key dimensions of viability and QOL from a complexset of micro-level housing market data. This requires the use of high qualitydata cross-sections and a robust modelling tool. This could be characterizedas an innovative socio-economic and geographic approach to the analysis ofmarkets and price formation – the methodology is quantitative by definition,but because of the practical applicability aspect involved, it is not restrictedto elegant formal modelling protocols. Currently, however, comprehensivemass appraisal systems – possibly involving AVMs – are limited to prop-erty tax applications in a few countries, notably the USA and Denmark. Wealso anticipate, in line with de Soto (2000) that an AVM can play differentroles in different economies, in which case formal or non formal economieswith stronger or weaker institutions require different approaches to massappraisal methodology.

Mass appraisal may be defined as a systematic appraisal of groups of prop-erties using standardized procedures. Mass appraisal methods normally referto large groups of properties rather than to a single property. The accurateassessment of the value of a predefined set of properties – or one particu-lar property – indirectly, using a model, for a given practical purpose, isthe main target of these methods (e.g., McCluskey et al., 1997; Gonzálezet al., 2002a,b). Several contributions have addressed the importance of massappraisal, exploring the relationship between the property value, the prop-erty characteristics and urban social and economic problems. The marketbehaviour is influenced by the property prices, the high durability of theproperty asset and by the fixed geographic location (Robinson, 1979; Harvey,1996). Hedonic price modelling has been proposed to define an econometricrelationship between the price and the property characteristics, particu-larly in a residential context. Arguably, the standard MRA-based hedonicprice models may not be suitable for capturing all the necessary informationinvolved in the value formation process, and the literature on how to developthe value modelling tools further is (or at least should be) evolving. On theother hand, there are property markets where data are not available or, evenif such data are available, it may not be organized into a property data bank.


Suitability Issues in Mass Appraisal Methodology 3

In this case other kinds of AVMs with a less deterministic relation betweenproperty value and property attributes may help. Although the problems arehighlighted, at the moment MRA remains the most important theoreticalframework in mass appraisal.

The quantitative, MRA-based methodology may be referred to as the‘orthodox’ approach to mass appraisal valuation. Several methods have beenapplied in the last decade defining a new approach to property mass appraisalvaluation. In this work, these methods and methodologies are dubbed‘heretic’ because of their different theoretical basis from the MRA, the dom-inant approach to mass appraisal. Model-free estimation techniques, such asneural networks and fuzzy logic, have been introduced to bring some flexibil-ity to the property value calculations, without neglecting the mathematicalrigour. In doing so, the value model becomes more powerful than its formal,regression-based and completely crisp counterpart. Pattern recognition is yetanother relatively untried approach within this realm. Indeed a number ofcontributions here offer ingenious and pragmatic, if not totally transparent,modelling methodology (see e.g., Jenkins et al., 1999; McCluskey and Anand,1999).

In this book we set out to explore the possibilities within this under-theorized problem area. We explore the possibilities for developing massappraisal methods following two different arguments: one, that the perfor-mance and feasibility of appraisal methods may be compared and evaluatedwith regard to a set of technical criteria; and two, that differences in thesuitability of methods also have to do with the particular context whereapplication takes place. Our objectives are first, to understand the problemof advancing mass appraisal methods/expertise from two different points ofview: the scientific debate and practical feasibility; second, to evaluate aset of methods based on a set of specific criteria, partly technical/practicaland partly institutional; and third, to establish an international platform forbroader networking.

Orthodox approaches to mass appraisal valuation

Two related modelling traditions exist today, both of which deploy MRAfor estimation: the model driven hedonic approach, and the data drivenstatistical approach. Hedonic price models comprise the most frequentlyapplied models in the valuation practice as well as in monitoring the hous-ing market. In these models the variables are usually of two basic types:internal physical (i.e., house and plot specific, structural) and external loca-tional. On top of that there may be additional variables, most notably sometype of inflation control (e.g., Miller, 1982). The purpose of the develop-ment of the hedonic price model was to make possible econometric analysis



of large databases of price and other recorded information describing thenature of the property and its vicinity, and possibly some specific (other)circumstances of the transaction. A more practical or a theoretical statisti-cal approach, especially the regression analysis-based value/price-modellingtradition, has been applied in order to provide tools for valuation conductedby the public and private sectors in many countries with convenient landinformation infrastructure (i.e., readily available digital register informationwith the possibility of multiple spatial aggregation).

There are several MRA-based house price studies that are made either withor without the formal hedonic price theory underpinnings. As shown by amultitude of studies (see e.g., reviews by Ball, 1973; and Lentz and Wang,1998) the measures of success have been the model fit (indicated by the totalcorrelation coefficient, the R-squared statistic), the model significance (theF-test at a given level of significance), whether each independent variablehas the anticipated sign of price association (indicated by the partial cor-relation coefficients) and whether each independent variable is statisticallysignificant (indicated by the t-statistic and a given risk percentage level).The standard error of the model and various test statistics are also applied asformal criteria of modelling performance. Recently, hedonic modelling hasbeen applied successfully for constant-quality price indices (Hoesli et al.,1997a); determining rental values (Hoesli et al., 1997b); estimating the dis-turbance effect of traffic (Wilhelmsson, 2000); and estimating implicit prices(Laakso, 1997).

In the empirical hedonic modelling literature locational proxy variablesmay be defined in various ways (cf. surveys by Ball, 1973; Miller, 1982;Laakso, 1997; Lentz and Wang, 1998; see also Thériault et al., 2005, forsophisticated accessibility measures). Kang and Reichert (1991) constructeda locational quality dummy based on levels of price per m2 living space.Similarly, McCluskey and Anand (1999) used a solution where the locationwas captured with a seven-valued categorical ‘ward’-variable, with valuesbased on mean transaction prices for that area.

Adding location to the already complex analysis of residential differen-tiation has become easier with the help of modern geographic informationsystems (GIS). There are obviously operational problems with the hedonicmodel, such as lack of suitable variables and data, and the issue of spatial res-olution, that is choosing a wrong level of spatial aggregation. This brings ustowards GIS-aided analysis, with the benefit of data visualization and stor-age, the possibility to construct more efficient accessibility measures andspatial analysis. An additional point is that the GIS-technology makes value-modelling applications more user-friendly. For GIS-demonstrations withinproperty valuation, see Orford (1999); Rodriguez and colleagues (1995); Wyatt(1995, 1996a,b, 1997); Lake and colleagues (1998); and Ding and colleagues(2000). See also Bible and Hsieh (1996) for analysis of apartment rents


Suitability Issues in Mass Appraisal Methodology 5

Furthermore, a GIS also helps in correcting the spatial discontinuity problemwhen the spatial dimension is incorporated in the analysis more explicitlythan in the standard hedonic model (Orford, 1999, pp. 63–67; see also Wyatt,1995, 1996a,b).

In a GIS only a few spatial statistical methods can be integrated. Whenthe spatial analysis system is ‘loosely coupled’, the statistical analyses areperformed separately from the GIS interface. Both packages thus have theirown function in the system: the raw data are transported to a GIS for storage,visualization and generation of variables for analysis of locational externali-ties, most notably proximity indicators. Then, the latter may be exported toa separate statistical package where, for instance, standard regression analy-sis or more sophisticated multi-level regression analysis (see below) may beundertaken. Finally, the analysed data may be imported to the GIS again, forstoring and visualizing. Then, the output of a GIS becomes a surface indicat-ing the spatial effect of a given variable on price. Studies by Orford (1999);Lake and colleagues (1998), Rodriguez and colleagues (1995) and Bible andHsieh (1996) manage to show that, in one way or another, the importance oflocation is great (see also Wyatt, 1996a,b, 1997). In general, the GIS-aidedproperty value analysis is feasibly linked to broader register information(Federal land cadastre of Russia, 2001).

Today a variety of advanced spatial techniques enhance the possibilitiesof handling location in the hedonic-based house price analysis. Namely, theconsideration of sub-markets and spatial drift may improve results substan-tially (see Orford, 1999). This requires an appropriate routine of handling thenon-linearity and dynamics prevailing across space. State-of-the-art methodsinclude multi-level specifications and spatial expansion models. Kriging isan advanced spatial statistical method that builds on parametric MRA. Whenthe standard hedonic MRA principle is �Xm ⇒ Ym, the kriging principle is�Yn ⇒ Ym (e.g., Chica Olmo, 2007).

In multi-level specifications, each externality effect is measured at anappropriate level. In order to add some efficiency into the (hedonic) valuemodel, the variation in house prices is decomposed between different spa-tial scales. In the case of property valuation applications, the appropriatelevels may be neighbourhood, street and property levels. A major advantageof this specification is the ability to differentiate between compositionaland contextual effects of location on house prices, in other words of spatialvariations in the housing stock and the place itself (Orford, 1999, 2002).

In the spatial expansion model, the contribution of a housing character-istic to the price is allowed to change over space. This reflects a series ofinterrelated sub-markets with sliding boundaries. Many applications usesuch a specification where parameters vary in order to cope with the spa-tial heterogeneity of a housing market (e.g., Geoghegan et al., 1997; see alsoWilhelmsson, 2002).



We can note related contributions: Gillen and colleagues (2001) discussautocorrelation when it is either a function of distance separating proper-ties in space (isotropic) or a function of distance and direction (anisotropic).Riddel (2001) improves the hedonic approach by allowing for interactionbetween space and time in a dynamic modelling framework, as environmen-tal quality may induce market disequilibria. Hierarchical trend modelling,with hedonic underpinnings using a Kalman filter, has been proposed byFrancke and Vos (2004).

In many situations exogenous factors or lack of information constrainindividuals to participate in segments of a larger market (Michaels and Smith,1990). However, standard hedonic analysis has not ignored market segmen-tation completely (Simonotti, 1998; Salvo, 1999). In principle, a hedonicregression cannot detect zonal boundaries, only the significance of the direc-tion and coefficient of the effect of the value factors, as well as the accuracyand explanatory power within the total sample of observations. One way ofclarifying the issue is to use dummy variables (Laakso, 1997). Another solu-tion is to calculate hedonic quality ranks for each observation (Rothenberget al., 1991, pp. 380–385). A third option is to construct separate modelsfor separate subsets of the data, with each subset, usually comprising alltransactions within a region, having its specific hedonic equation (i.e., thepartitioning approach). Hence the data are split into different segments,which are either a priori predefined or somehow synthesized. Also, if seg-mentation in a theoretical sense is ignored, the partitioning approach may bejustified (see Needham et al., 1998).

Demand side segmentation, that is, collective preferences according tomembership of an a priori defined ethnic or socio-economic group, isoften studied with the specified two stage procedures of hedonic mod-elling. If the functional form is curvilinear we may derive willingness topay (WTP)/willingness to accept (WTA) estimates based on the shadowprices and specified demand side data. (If the equation is linear, the secondstep in the two-step hedonic approach cannot be taken.) The targets thenare the marginal WTP estimates (i.e., demand functions) for each relevantcharacteristic by groups (e.g., Bökemann and Feilmayr, 1997).

Integration between the adjustment grid methods and regression anal-ysis was developed by Colwell, Cannaday and Wu (1983, cited in Kangand Reichert, 1991), in order to integrate the ordinary least squares (OLS)estimation of adjustment factors to the standard method. Here the mul-tiplicative percentage grid adjustment method is considered a particularlypromising option according to Kang and Reichert (1991). Another exampleof earlier innovations is the Stein rule (Knight et al., 1993). Here, the ideais to improve the estimation accuracy by the following procedure: first addall variables to the model and estimate the partial correlations (i.e., betacoefficient, marginal adjustment factors) with OLS. Then, remove the vari-ables for which it is believed that beta equals zero, in other words the

MassAppraisalMethods€¦ · Kauko CarolineMcPherson: “kauko_c000” — 2008/6/5 — 16:55 — page iii — #3 MassAppraisalMethods Aninternationalperspectiveforpropertyvaluers

Documents