The Center for Natural Hazards Research Thomas HarriotCollege ofArts and Sciences East Carolina University A 116 Brewster Bldg. @Greenville. NC 27858 Tel: (252) 328-5718. Fax: (252) 328-6'143 wlVw.ecu.edu/hazards • [email protected] Valuing Spatially Integrated Amenities and Risks in Coastal Housing Markets January 2006 Okmyung Bln* Assistant Professor, DepartmentofEconomics, EastCarolinaUniversity Tom Crawford Assistant r>rofossor, DepartmentofGeography, EastCarolina University Jamie Brown Kruse ProfessorofEconomicsand Director, Natural HazardsMitigation Research Center, EastCarolina University Craig Landry Assistant Professor, Dcpartmcnl ofEconomics, EastCarolina University 'Correspondlng authormay be reached at: Voice: 252-328-6820; Fax: 252-328-6743; Email: [email protected]. The authors thank Glenn Blomquist, and seminarparticipantsat theAssociation of Environmentaland Resource EconomistsSession at the Southern EconomicAssociation 2005Annual Meeting, fortheirhelpful comments. The authors also thank HankMontgomeryfor providing New HanoverCountyreal estate data and Paul Hindsleyforproviding valuable research assistance. This researchwas supported byEastCarolina University2005 Research DevelopmentGranl. Valuing Spatially Integrated Amenities and Risks in Coastal Housing Markets ABSTRACT Hallstrom and Smith (2005) argued thatcoastal amenities and risk arc so highlycorrelated thatconventional hedonic methods cannotidentifyseparateeffects. In this study,we construct a thrce-dirnensional measure of view, nccounting for natural topographyand builtobstructions, thatvaries independentofthe risk to disentangle these spatiall}fintegmtc(1 housing characteristics. A spatial hedonic model is developed to provide consistent estimates ofthewillingness to pay for coastalamenities and risk. Our findings suggestan alternative to the J-Iallstrom and Smithapproach thatincorporates the multipledimensions ofspatial attributes can besuccessful in isolatingrisk from amenities on thecoast. Keywords: coastal amenities, coastal hazards, hedonic prices,spatial regression JELClassifications:D12,Q24, Q26, R21. Valuing Spatially Integrated Amenities and Risks in Coastal Housing Markets I. Introduction Coastalareas in the U.S. have beenwitness togrowingpopulation and increased economicactivityin recent years. Population in thecoastal zonegrew37.20/0between 1970and 2000,andwhile thisgrowth is roughly proportional to the averagegrowth seen in interiol'parts ofthecountryover the same time pcdod, population densityin the coastal zone is two to three times higher (Colgan 2004). The contributionofcoastal areas toeconomic activityis ratherlargegiven the limited land area: whereas the coastal zone contains only 40/0 ofU.S. land area, economicacti.yityin the coastal zone contributed 11% to the U.S. economyin 2000 (Colgan 2(04).1 The relativelydensepopulation and abundant economic act.ivityin the coastal zoneis cxposed to vagades stemmingfrom the interactionofland,sea,and climate,placingcoastal developmentat considerable risk. Natural hazard risks in the coastal zone include storms Olllrrie~nesand nor'eastcrs on the castcoast),with associated flooding and wind damage, as weU as ewsion stemming fWI11 stonns and sca level rise. Recentpredictionssuggest thatwe are entering aperiod ofincreased stann activity (Goldenbergetal. 2001;\X1cbsteret al. 2005),and sea level is projected to rise 9 to 88centimetersovcr the nextcentury (Intergovernmental Panel on ClimateChange (IPCC) 2(01) creatingpotential problems for the coastal economy. Meanwhile, coastalpropertyvalues continue to soarand construction projects in the coastal zone expand.2 Do buyersand builders in the coastalzonecomprehend thediverse risks that characterize this landscape? Oneway to address this question iswith data on housingsales. The hedonic property model provides an intuitive 2n21ytical tool th2t can be used to exatnine theeffects ofhousingcharacteristicson prices. Ifcoastal properties in asingle housing marketvary in theirdegree ofriskofstorm/flood damage and buyers and sellers observe and react to thesevariations, then the hedonic price function will reflectvariations in propetty risk. From the perspectiveofthe household, marginal implicit prices from the estimated hedonic price model can beinterpreted as marginalwillingness to pay for housingattributes (Rosen 1974). A nUl11berofpapers have utilized this approach tovalue risk (Beron etal. 1997;Gayer, I-Iarnilton, and Viscusi 2000;Simmons, Kruse, and Smith 2000; Landry, Keeler, and Kriesel 2003; Binand Polasky2(04). 1ContributiontoeconomicactivityIsmeasuredbyemploymentandvalueadded(Colgan2004). 2Forexample,anarticlebyFrankNortonIntheRaleighNews&Observerreportsthataverageseilingpricesforresidential propertiesInWrightsvilleBeach,NorthCarolinahaveIncreased420percentsince2001,andthenumberofbuildingpermitsIn CarolinaBeachovertheprevious24monthperiodexceedIhenumberofpermitsIssuedovorthepast20years("BeachPricesRide Crest"29May2005). WrightsvilleBeachandCarolinaBeacharetwoofourstudyslles. 2 In coastal zones, however, riskisoften highlycorrelatedwith coastal amenities, such as water-frontage, distance to beach, and ocean view..Hallstrom and Smith (2005) claim that this correlation is perfect(orclose toperfect) and make usc ofa repeatsales model and adifference-in-difference framework tovalue changes in subjective risk assessmentdue toa near miss" hurricane event-which theyidentif}'as achange in u infonnaUon.3 In essence, both risk and amenities at<e spatially-delineated,and thedegreeofcorrelation between these spatialattributes can be high in the coastal setting. Nonetheless, risks and amenities maynot be perfectlyconelated-itdependson how the researcherdescribes and models agent perceptionsand thus how housingattributes are measured. Our findings :'luggestan altemative to the Hallstrom and Smith approach that incorporates the rnultipledirnensionsofspatial attributes can besuccessful in isolating risk from amenities on thecoast. This stud}' uses spatialeconometric methods and parcel measuresderived usinggeographic information systems (GIS) to provideseparateestimatesofthe marginal valueofboth coastal risks and amenities in the hedonic framework using the conventional modelingapproach-a parametric regression model utili:dngdata on housingattributes and asinglesales price for each propeny. Ourstudyarea iscomprisedofbeach communities in New I-hnoverCounty,North Carolina. The selectedcomrnunit.iesare among the most highlydeveloped areas in the region and Me at riskofflooding and shoreline erosion dueprimaril)' to severe weathereventssuch as tropical storms and nor'easters. \Y/e use Federal Emergency ManagementAgency (FEMA) flood maps to identifySpecial Flood I-fazarel Areas (1 chanceofflooding peryear),and use adichotomous classification for aU properties in outsample % accordingto this distiuction. The North Carolina DepartmentofEnvironmentand Natural Resources provided information on historical emsion rates (though for outsample there is novariation in this risk measure). Coastal atnenities that vary by p<lrccllocationare measured using spatialdata processing techniques implementedwithin GIS. \Y./e utilizeLIDAR (Light Detectionand Ranging) dataand temporal information on parcel development histories to createaquantitative measureofoceanview for each parcel. Ourviewmeasure is both spatiall}'and temporallyexplicit in thatitaccounts for the fact thatviews can change overspace:md t.ime as buildings and infrasu'ucturesurroundingapropenyare altered,4 Wle also quantify for each parcel the Euclidean distance to the beach/oceanshoreline. Given that the ocean view and distancemeasures need notbe perfectlycorrelated with risk rneasures, ourspatialeconometril;:analysis allows for identification ofthesedifferential effects. JAhighdegreeofcorreIalionbetweenrlsk'andcoastalamenitiesIsalsonotedIn81n,Kruse,andLandry(2005). Ifamenltlesare nollncludedIntheirmodel,themarginalvalueofrisktakesonthe~wrong~signduetothecorrelationwithamenllfes. However, IncludIngamenitymeasuresIntheirmodelfixestheproblem,suggesUn'gthaicorrelationamongriskandamenlllesIsnotperfect. 4OurdatasupporttheclaimthatoceanvIewscanchange. forexample,ofthe69housesthatsoldIn2001havinganoceanview, 18(25%)experiencedadecreaseintheamountvIewfrom2000to2001 duetonewlybuildstructures. ThemeandecreaseIn angularfieldofviewwas14.4degreeswithamaximumdecreaseof36.4degrees. 3 II. A Simple Behavioral Framework We take an approach similar to Hallstrom andSmith (2005) (hereafter HS) by consideringasimpleexpected utility maximization problem with only twooutcomes, hunicane (H) and no hurricane (NH),S Out"primary modifi~ationto their model is thatwe do not restrictamenities and risk to be represented bya onc dimensional attribute-distance from the beach, r..........<ls do HS. Flood and erosion risk, in fact, varyncmss multjplc physical dimensions, as doesview amenity.6 Distance frorn thebeach is included in our model as onlya measure ofpr<?ximity, presumably reflecting recreationalvalue (to theextent thatonecan control for variation in erosion risk along theshore). The household's expected utility function is: v =P(l~E)UII{tI,l;b,1IJ- /((tJ,l;b,I~E,io)- L(b,F,E,io}j (1) +(I -P(F,E))UNH{II,I;b,1IJ- /((v,r,b,F,E,io)), where the utility function is statedependent)p(r~E)is the subjective probabilityofadamage-producingevent, F represents location in a flood zone)E represents location in an erosion hazal'd zone (historical erosion rate as proxy)) fJrepresentsview amenity,his avectorofhousing characteristics,11/is income) H(.)is housingsales price (expressed in annual terms ifincome is expressed in annual terms), L(.)is the size ofthe monetaryloss due to storm netofinsurancecoverage, and iois the insurance rateperdollarofcoverage. Note thathazard risks and amenities arc represented bydiffel·entcovadates in (I). This specification differs ftom thatofHS) in which I~distance from the shore, represents both hazards and amenities. Generally)weassume thatFand E varyindependentlyofVand 1:7 In addition,weassume thathazard information is constantover the period ofouranalysis and therefore suppress the I-IS infonnation parameter,I. Undercompetitive marketconditions, buyers take the hedonic pricescheduleasgivcn and optimizeexpectcd utility through the choiceofhousingcharacteristics,with residual incorneleftover for consumptionofa compositecommodity. Assumingcontinuous fisk and amenity measures, in equilib~·illm we have the following: (2) (3) f>WeretaInthehurricane/nohurricane&tatesofnatureInkeepIngwiththeHSframeworkbutusehurricanetorepresentanyflood anderosIonproduclngweatherevent. I>Floodriskdependsuponlocalionwithinthefloodplain,whichvarieswithlatitude,longllude,andelevallon. ErosIonriskIncreases withproximitytothewaterfront,buttypicallyvariesalongthecoastline(asevidencedbyvariationInhistoricerosionrates). likewise,viewsvarywithlatitude,10ngllude,andelevation,andchangeoverlimeassurroundingdevelopmentevolves. lOurdataonlypartiallysupportthIscontention. WeareunabletoexamineerosionriskbecausethereIsnovariationInthe hIstoricalerosionrateestimatesforNewHanoverCounty. Assuch,distancefromtheshorewillbecorrelatedwitherosIonrisk becausethereIsnotadditionalvariationInerosionriskacrosstheshoreline. ThecaseofNewHanoverCountyIspeculiarand ratheruniqueIntheregard,asmostcoastalcountiesexhIbitvarlallonInthehistoricalerosionrate. 4 for fl::: lV',md j::: F,E. Equadol1 (2) indkates that the marginal implicit hedonic pricc fot" arnenities rcflects the expected amcnityvalue. Equation (3) indicates that the marginn} implicit price for risk attributes refl<;cts: (i) the perceived marginal change in theprobabilityofstol'ln damage,pfi (ii) theincremental utilitydifference across states, UH- UNH;and (iii) the marginal impactofrisk on the mngnitudeofloss,pUl/"J-j. Dividingby the expected value ofthe marginal utilityofincome produces measures ofwillingness to pay. Ifone is willing to assume that risk attributes affectonly the pt'Obabilityofloss and not the magnitudeofloss, orviceversa, equation (3) sin1plifiessomewhat.' We use the framework of(1) and the results in (2) and (3) toguideour formulation ofthe hedonic price model and interpret the parameterestimates. III. Study Area and Data Ourstudyarea encompasses fouf primarybeach communities- Carolina Beach,Figure Eight Island, Kure Beach,and \'hightsville Beach- along theAtlantic CoastofNew HanoverCounty, Nonh CaroHna (Figure 1). Ongoingdeveloprncnt has increased overallexposure ofprivateand social capitnlassets to natural disasters such as stormsurge flooding and shorelineet'Osion caused bysevereweatherevents. Over the period ofouranalysis a numberofhurricanes made landfall along the Nonh Carolinacoastline. Between 1995 and 2002,six named storms affected the southeasterncoastofNorth Carolina~Arthur,Berthaand Fmnin 1996, Bonniein 1998,and Dennis and Floydin 1999. Despite t~chigh level ofhurricaneactivity, structumllossesin the study area were relativelyrninor. Fotexample, duringhurricanes Dennis and Floyd (Floyd being the strongeststorm d~ringthis period atcategory 4) onlyone buildingwas destroyed in \X1rightsville Beach,while a more significantportion were threatened byerosion in otherareas~63properties on Figure Eight Island and 5each in Carolina Beachand Kure Beach (Rogers 2000). Given the relatively high level ofhurricane activity,we feel itis reasonable to assume thatbuyers and sellers wetccognizantof natural hazard risks when engaging in the housing marketduring the period ofOul' analysis. W/e assume the level ofinformation and pcrception.sofrisk arc constantover this time period. Moreovet, since direct impacts on the studyareawere relatively minorwe can be relativelyconfident that exogenousshiftsin housingsupplyare notlikely to playa role,and thus the hedonic propertyprice function should be telatively stable. \V./e utilizea rich spatial dataset thnt includes property parcel boundaries and agridded elevationsurfaceall stored within aGIS using the North CarolinaState Planecoordinatesystern (with horizontal and vertical units in feet and referencing the 1983Nonh American datum). Elevation surfaces weredcrived from LIDAR source data obtained from thc National Occnnicand Atmospherici\dministtation. Parcel data were 8Forexample,IfJaffectstheprobabilityofloss,bulnotthemagnItude-selfprotectionIntheparlanceofEhrlichandBecker (1972}-ltem(HI)willnotappearInequallon(3). Onthehand,Ifj affectsthemagnitudeofloss,butnot theprobablHly-what EhrllchandBeckerrefertoasselfInsurance-Items(I)and(II)willnotappearInequation(3). 5 obtained frorn the New I-hooverCountyTax Office aod contained information on sale yeal', sale price,and various structuralattributes. W/c selected a total of1,075single-family residential homes thatwere sold in our foul' studycommunities between 1995and 2002. I-lousesales prices are adjusted to 2002prices usinga conSUmel"price index fOl' housing. Theaverage sellingpricewas $297,968with aminimum sales priee of $30,298and a maximum of$3,575,000. The first and third (]uartileofthe sales priceis $110,000and $295,000, respectively. An innovation ofthis shH..Iyis our amenit-y measure ofocean viewcalculated usingGIS techniques. fI.'lost shtdies measurewaterview usingbinaryorordinal values (Le. total, partial,or noview) obtained from site visits, tax assessor mass appraisal data, or map-based interpretationofrelative location (Le.waterfront;::: total view,interior;::: partialornoview) (Benson etai. 1998; Fraserand Spencer 1998;Rinehartand Pornpe 1999; Seileretal. 2001; Bateman etal. 2002; Bourassaetal. 2004).9 Advances in geospatial technolo!:,')'and data availability haveenabled objective,quantitative measures ofspatialenvirol1lnentalattributes (Geogheganet aI. 1997;Batemanetal. 2002) includingviews. Research ingeographic informationscience, urban planning, and landscapearchitecture has developed novel approaches to representation and rneasurenlentofview (Fisher 1995,1996;Kidneret al. 2001;O'SullivanandTurner2001;Turner et al. 2001;Con and Goodchild 2002; Fisher-Gewittzmanetal. 2003; Llobera 2003;Jiang2005; KUlllsap etai. 2005), Behavioral research,aided in somecases by the use ofGIS, has related view characteristics to human behavioral patterns (Bishop 2003; Turner2003;Wienerand Franz 2004;Stamps 2005a, 2005b). Propcrtyvaluation studies have begun to adopt GIS-based approaches to view (Lakeet al. 2000;WingandJohnson 2003; Paterson and Boyle2002;Bishopet a1. 2004). Gurmeasure ofocean view,which we denoteas JiieJJJ,was created using acarefullyconstructed database subjected to a multi-step processingsequenceinvolvingcustomized programs. All processingwas implemented usingArc/Info GIS software. Next section describes in detail ourstratem' to quantifyocean vicw, About 32%ofthe properties (346 out ofthe 1075properties in ourdata) have apositivevalueofJiiew, and an average ofoceanvicwis about 18.37degrees in one miledistance. Anotherimportantcoastal amenity measure is the distance to nearest beach, Earlierstudies have shown that proximity to shoreline is highl}fdcsimblc in housingmarkets (Shabman and Bertelson 1979;fI.,mon, Gressel, and Mulkey 1984; Earnhart2001). The averagedistance to the ncarestbcach is 1,743 feet for tbc selccted properties. In addition,we utilho:ed aerial photographs tovisually interpret and constructdummyvariables for the existenceofpierand sound frontage. About 16% ofthe properties haveasound frontage and about two thirds ofthose homes have,a pier. 9ArecentartIclebyDavidA. FahrentholdInWaslllngfonPostreportsIhatNewHampshirehasembarkedonacontroversialquest toquantifyandthentaxthebeautyoflheirresidents'vistas. ThearUclereportsthatcurrentlylhereIsnoscIentificwaytovalueview ("N.H.PuisaPriceonPanoramas:PropertyTaxesSoarBasedonScenery~14November2005), 6 Ourstudy focuses on the flood hazard risks measured b}'location within SpecialFlood Hazard Area (SFHA). TheSFHA denotes the 100-yeal'floodplains (onepercentannual chanceofflooding). Flood risk areas are identified using the Flood Insurance Rate i\-1aps fmm FEMA. \Y/ecreate adummyvariable thatdenotes parcels located within the SFHA. R.isk measuresother than flood do notvarywithin the sample and are excluded from this study. Although ourbehavioral model in~ludeserosian risk) there is novariation in the erosion risk far the studyarea, and thuswedrop erosion risk in ourempirical model.lOSincewe cannot control for erosion risk,we mustinterpretourresults regardingproximity to the beach with caution, as proximitywill reflect thevalue ofaccess for recreation and leisure activities butalso exposme to risk from coastalerosion. In addition,we calculate thedistance ta nearesthurricane evacuation mute and also thedistance to nearest central business district (downtown.\X!ilmington). Distancesare measured as the Euclidean distance from the centroid ofthe property to the nearestedge ofa feature. Othet·variables used in this studyincludestructural characteristicsofpmpe1'tiessuch as age ofthe stt1.1ctU1'e, numberofbathrooms)sizeoflot)etc. A typical home is about22}'earsold and has 9,235square feet oflotsize. Approximatel}'90percentofthe homes have airconditioning,and 49 percent havea fit-eplace. About45 pel·centofthe homes are multistoryunits) and 8percent have hardwood floo1'~. Less than 5% ofthe homes haveadetachedgarageOfaswimming pool. Property tax appraisers rated about 150/0ofthe homes in"goodcondition". \Y/e use dumm}'variables to control for unobservable differences across townships. Alarge fraction ofhouse sales occurred in CaroHna Beach (40%) and Kure Beach (34%). Table Idefines thevariables usedin the hedonic price function and provides summarystatistics. Table 2displays pairwise correlationcoefficients for the amenity and risk variables included in out·anal)'sis. The estimated correlation between flood hazard (SFHA) and distance to the beach (BEACH) is not statisticallydifferent from zero in ourdataset. This empirical result is atodds with the theoretical rnodel of HaJlstrorn and Smith (2005), inwhich distance to the shore is assumed toc!osc!}' reflect flood risk. Since flood hazard varies across multipledimensions) the correlationwith distance to t.he shore is notstatisticall)' significantin ourdata. The correlationcoefficient forviewal11enity (VIE\XI) and distance frorn the shore is negatjve and different from zero) butisalsodifferent from negative one at p <0.0001. Thus,view amenity and proximity in the shore are notlinearlydependent. Likewise, and moreimportantly,viewand flood hazard, while positivelycorrelated,are not perfectlycorrelatedeither (p <0.0001). Ollrdata do notappear to be ill-conditioned with respect to flooding hazards and viewarnenities. \Y/eagain notc thatwcdo not considererosion risk in our analysis, as there is notvariation in·estimatesofthe historical erosion rate in New HanoverCounty. 11)NorthCarolinaDivisIonofCoastalManagementreportsthatthelongternerosionrateforlhestudyareaIsabouttwofeetperyear thatdonotvaryacrosssites(htlp:lldcm2.enr.slate.nc.usfMaps!chdownload.htm). 7 IV. Measuring the View Viewis an intuitive concept f01" virtuaUyaU sighted personsdue to everydaylived experience. Despite its simplicity,constructinga measure ofview requires definingaspecific viewattribute to measure among the many possiblecandidates. Viewarea is onesuch measureobtained [mm aJ!ieJIIshedwhich is defined as the total areavisible from an observation pointataspecified location. A tlieJlJShedneed notbespatially contiguous; total visiblearea maybe fragmented into multiple patches. An alternative 2nd ordct"measureis the afca ofthe largestsingle JJicJlJshedpatch, orsome measureofthedistribution oftJieJ1fsheripatch sizes. Descriptive measures ofpatch spatial configuration usingconcepts and techniques borrowed from. landscape ecology havealsobeen developed. Further,stratification ofJJie1JJShedarea by distance bands from the observation pointyields further area ordistributional measures thatdifferentiate between neat"and distant views. These and otherviewmeasures aredescribed in Batty(200I), Germino ctaJ. (200I), L10bera (2003), and Stamps (2005b). W/e selecteda faid}' straightforwardand intuith'e attribute to measure view. OurJJieJJJrneasure represents the scopeofaparcel's oceanview. Itidentifies all possible fields ofview from acentral pointand sums their angularwidths toyield the summed total field ofviewin degrees (00-3600). Priorvaluation studies have utiJjzed asubjectivescope mcaSUl"e as theviewattribute ofinterest. Forcoastal comrnunities, these measures arc typicallycontained in mass tax appraisal databases with values having been determined in the field bylocal tax asses~ors and l"epotted usingordinal codes (i.e" full, partial, or noview). \Y/e are unawareofanyvaluation studies thathave used quantitative measures ofscopesuch as ours. :Moreover,despite the abilityto create GIS-basedviewsheds nnd various 2nd orderviewmeasures,qunntitativc measures ofscope are largelyabsent in the GIS Iiteratul"e. Creationof?urviewmeasure required the constructionofannual topographic surfaces from LIDARdata. LIDAR instrumentation mounted on an airplane emits adense arrayoflaser pulses thatstrikesurface features and reflect back to the recording LIDAR sensor\vhich contains aglobal positioningsystem that corrects foraircraftpitch and roll. The time it takes for emitted pulses to retum to the sensor is used to determine surfnce elevations. Ourbase elevation layerwas agridded surface ata5-foot pixel resolution flown inSeptember2000 (Figure 2). This'topographic surface represents built structures,vegetation, and bare earth. This isespcciall}fimportant to realistically measure vicwsince these features act to obstructocean views. A recent hedonic studyby Paterson and Boyle (2002) utilizes an aerialviewshed measure createdvia GIS techniques; howevel", theirlneasure relied on aHbareearth" topography that does not realistically account for the presenceofobstructjng builtorvegetation features. To construct annual elevation sUffaces, separate layers were built for each yearusingairphotos and buildingheightinformation from tax assessor 8 data to add 01'subtractelevation fol' digitized buildingfootprints as necessal'Y, Atotal ofeightelevation surfaces were aeated-one fol' each yeaL11 POl"simplicity, we distill the methods used to create lJie1l'into two steps that involved:.(1)aeatinga parcel's viewshed (i.e. viewarea) at a I-miledistance from the parcel, and (2) calculationoftheviewshed's sumrned field ofview (Le. degrees). Creating thevie\Vshed requires positioningaviuual "observer"somewhere on the landscape. FOl"each parcel)we establishedan observerpoint at the centel'point ofthebuildingfootprint at an elevation 10 feet below the estimated maximum heightofabuildingwhich was obtained from the e1C'~'ation surfaces or from tax assessordata that reported the building's numberofstories,12 An ocean zone wasdemarcated usingavectOl"shm'e1inelayer that repl"esented theocean/beach interface. The targetareaof potentialviewwas limited to thearea ofoceanwithin one mile from the observerlocation. To derive viewshed, the "lineofsight" mytracing technique comrnonlyavailable in GIS softwarewas used toidentify every5-footpixel in the targetocean areavisible from the obset\'el'location (Figure 2). To create oUl' final tJieJl!measure (i.e;) the summed field ofviewin degrees), we constructed circulararc segmentsat theouter I-miledistance oftheviewshed,calculated theirindividual lengths, and summed the individual lengths. A hypothetical maxirnum summed arc length was the circumferenceofthe I-milecircle, orapproximately6.28 miles sincecircumference = 2 X pi X mdius. VieJ!Jis thereforederived as: tJielJl= (arclengtllmib / 6.28milcs) X 360". POl'example)an arclength of3.14miles translates to 1800. Given the linear natureofourstudyarea's shoreline;an approximate maximum possible value for lJieJJIwould be 180".D Figure 3iHusttates viewsheds and reports values for the 14m'measure for two cxmnplc properties. V. Methods Since the pioneeringwotkofRosen (1974), the hedonic price Inodel has been used extensively in environmental and natural resource economics as atool for non-marketvaluation. The model assurnes that values ofheterogeneousgoods are reflected in pl'iccdifferentials. Residential propertycan bedistinguished based upon structural) neighborhood, and envil'Onrnentalcharacteristics,and one can assume that utility IIAsanexample,tocalculateviewfortransacllonsIn1997,weusedthe1997elevatlonsurfaceforwhIchelevationsforbuild!ng footprintsofbuild!ngsbuiltIn1998,1999,and2000weresubtractedfromtheorigInalyear2000surface. TransactionsIn2002used Ihe2002surfaceforwhichelevaUonsforfootprintsofnewbulfdlngsbuiltIn2001 and2002wereaddedtotheorigInalyear2000 surface. 12WhilesomebuildIngsmayhavevlew!ngdecksontheroof,IhlsconventlonwasusedunderIheassumptlonIhatmostpeople wouldviewfromavantagepointabout10feetbelowthemaximumbuildIngheight. 13AnImportantconsIderationwiththIsmeasureIsthatItIspossibleforsomeoftheoceanareaInteriortoatermlnal1¥mllearc segmenttonotbepartoftheviewshed. Forexample,anobserverata3-storyresIdencelocatedone+blockfromtheoceanfrontrow mayhaveto"lookover~a2-sloryoceanfrontresidenceora15-ftsanddunetoseetheocean. ThiswouldvisuallyobstructIhenear shoreoceanareasothatviewableoceanareaonlybecomesvisibleatsomedistancefromtheshoreHne. ThIsunderscoresthefact thatourviewmeasureIsquaUtativelydifferentfromvlewshed. However,forourdatathetwomeasuresarehIghlycorrelatedtothe a extentthattheyareeffectively!denticalln quantitativesense. Th!swillgenerallybethecaseforothersellingswithlinear shorelines. WeelectedtouseviewduetoiteaseofInterpretation. 9
Description: