WPCt !D"`2.݄ZF}@˸}wkulaC#U ͸3S3zB{ %y&z d0la*NM1JE=( O$`ƈ Pp:H9sHGa;53jtpjQhiCZh?@'gJ%}#MM,gFL>U^JyX2|ݫ,'>&#+,'SR1V.8-0l "QOhn}U+QO=ʼMA3zmƖZ6 :69j66Rj7}oi3a( R$új,C\} bȪU\\<qs Nj;#PsFW+ap |Wu/CP2S 7BCrEXSSk-F2 T\-XHL !μ )T 0~ 0xp 0k 0kS 0 0~X 0 0 0 01 0 ( 02 01P 0C 0V! 0# 0# 0$ 01% 0% 0& 0k' 0D( 0(U L) %) 0z) 0V* :+ 0=K+ 0F+ 0P+ 1, 0, 1- 0.F/ 0O2E73 093Ud@3U@4 0C^4E4 05f669 D+: ; 0O; 0E< AYc<<UN>>d?BDnG}JfCLaELbYLU$NMa NUNNUNkNUNNO2PU)NQ/RUB&VhVjVP-W}Y mxZsZ\UN]UN]{M^d_dUNaf!f}gnMiPkXknmdo Cufuav B*v?vU@Gvfvavfvav B"vfvavUNvUN4wfwawfwawWw 1{{n{ 04}~n ~#!n/2n:=XEw@4nnn nn  UN)wnnnnnnnɡṇϥץnߥnnnn nn!$n,/n7:BnJMnUnX[cnknnvyyynnnnnnnnnnnnnnnn&)n14n<?nGJnRUn]n`cknsvn~nnnnnnnnnnn n  n nnn&)n1n47?nGJRnZn]n`c k ns v"n~"$n$&n&(n(**n*,n,n.00n02n244n46n68n8n;= =n=?n ?#An+A.Cn6C9EnAEDGnLGOInWInZK]MeMnmMpOpOnxOn{Q~SSnSUnUWnWYnY[n[]]n^`d`nCdFfUNNf 0fngi 1miU<$j$j 1m`jjdVkUNnUN$oUNrororoUNooUNpppppppW\pdpppppppqrV& 8Document[8]Document Style0..8` ..` V8Document[4]Document Style.. . V 8Document[6]Document Style8..V 8Document[5]Document Style0..V/8Document[2]Document Style 2A.3  Ԁ   V& 8Document[7]Document Style0..0` ..` zU :Right Par[1]Right-Aligned Paragraph Numbers..2I.3  Ԁ..0..zh :Right Par[2]Right-Aligned Paragraph Numbers..` ..2A.3  Ԁ..0` ..` V?8Document[3]Document Style.. 21.3  Ԁ   z{ :Right Par[3]Right-Aligned Paragraph Numbers..` ..`  ..P 21.3  Ԁ` ..` 0 .. z :Right Par[4]Right-Aligned Paragraph Numbers..` ..`  .. .. 2a.3  Ԁ .. 0..z :Right Par[5]Right-Aligned Paragraph Numbers..` ..`  .. ..h..2(1)3  Ԁ..0h..hz :Right Par[6]Right-Aligned Paragraph Numbers..` ..`  .. ..h..h..2(a)3  Ԁh..h0..z :Right Par[7]Right-Aligned Paragraph Numbers..` ..`  .. ..h..h....2i)3  Ԁ..0..z :Right Par[8]Right-Aligned Paragraph Numbers..` ..`  .. ..h..h....p..2a)3  Ԁ..0p..pVX8Document[1]Document Style  @..^  2I.3  Ԁ     Ԉ l2:Technical[5]Technical Document Style.. 2(1)3  Ԁ. l2:Technical[6]Technical Document Style.. 2(a)3  Ԁ. l/%:Technical[2]Technical Document Style 2A.3  Ԁ   .. l,!:Technical[3]Technical Document Style 21.3  Ԁ   .. l(!:Technical[4]Technical Document Style 2a.3  Ԁ   .. l:0:Technical[1]Technical Document Style  2I.3  Ԁ     .. l1:Technical[7]Technical Document Style.. 2i)3  Ԁ. l1:Technical[8]Technical Document Style.. 2a)3  Ԁ. <p`(4Line Draw 12cpiFull-PER& 8BibliogrphyBibliography0....fp2Doc InitInitialize Document StyleS !    I. 1. A. a.(1)(a) i) a)S ($0 ($0 0 (($0 0 0   A_ekqwDocumentDocument StyleI.1.A.a.(1)(a)i)a)jo4Tech InitInitialize Technical StyleS #  1 .1 .1 .1 .1 .1 .1 .1 S CuyTechnicalTechnical Document Style11.11.1.11.1.1.11.1.1.1.11.1.1.1.1.11.1.1.1.1.1.11.1.1.1.1.1.1.1x?t2PleadingHeader for numbered pleading paper %  &V(   LXXX?h''*dE*??h''*dE*?HH1HH2HH3HH4HH5HH6HH7HH8HH910111213141516171819202122232425262728  .+('2V$ uU!   <i:BibliographyCoursey,DonL.,Hovis,JohnJ.andSchulze,WilliamD.,"TheDisparityBewteenWillingnesstoAccept̀* `(CG TimesScalable) `(CG TimesScalable($     ; -$  *XԀ  3    ݀ExceptionsincludeHoehnandLoomis(1993),Hoehn(1991)andHanemann(1984).SeealsoHanemannand  Kanninen(forthcoming)foradescriptionofmoregeneralspecifications.(X!2$ uU!   *XԀ  0    -$  *XԀ  5    ݀m *HoehnandRandall(1987)offeramodelthatpredictspeoplewillunderstateWTPandoverstateWTAwhenthey  lacktimetothinkoracleardefinitionofthecommodity.CrockerandShogren(1991)outlineamodelinwhichthecommodityiswelldefined,butunfamiliar.Evenwithadequatetimetothinkaboutthequestion,therespondentsneedtoinvestinlearningabouttheunfamiliargoodandthuswillsystematicallyoverstateWTP.#* mz#  uU!   ݀  6  B XԀ#\X BH#k X\ForadiscussionofissuessuchasstrategicbehaviorandfreeridinginCVsurveyssee,forexample,the  Winter1994issueoftheNaturalResourcesJournal.#\X k#B X\#\X B~#'dxd Level 1 Level 2 Level 3 Level 4 Level 5('2V$ uU!   ($ (    ) 2341M << deUU 73w7XX73w#XXX7X## 3XX#XX 37XXXX#XXX7X## 3XX#(6)`WTP_i~=`C^P(X_i)`+`Y_i-[Y_i^{1-theta(X_i)}+({1-theta(X_i)}over{1-lambda(X_i)})(E_{high}^{1-(X_i)}-E_{low}^{1-(X_i)})]^{1over{1-theta(X_i)}}`+` `` &d9 Z 6Times New Roman Regular (5)`WTP_i`=`C^P(X_i)`+`Y_i`-`[Y_i^{-#(X_i)}`+`({1-(X_i)}  over{(X_i)})(E_{low}^{-#(X_i)}-E_{high}^{-#(X_i)})  ]^{-1over{#(X_i)}}`+` (O5)WTPMM=ioCMM~CP(XMMdi)DYMMiL[YMMhMM>h#MM|h(MMhX9iMMh)MMi;(  1Y   (W XMM Pi )a } }( }XMMe 8i })+ )c ( EMM hMM_ h#MM h(MM hX 9iMM$ h)MM lowJ  EMM.hMMrh#MMh(MMhX9iMM7h)MMhigh])]MMMMM1MM(h#MMfh(MMhX9iMMh)3  -$  *XԀ  8    ݀)*Duringthelabexperiment,theparticipantsweredividedintogroupsandeachindividualwasaskedtowritedown  hisorherWTP.Ifthesumofthegroup'sWTPwasgreaterthanthecostoftheadditionaltrees,theparticipantsactuallyhadtopaytheamounttheybid.#*)z#ԀOnlyoneofthegroupsactuallyendeduppayingmoneytheothergroups D didnotcollectivelybidenoughtocoverthecostoftheextratreesandtheexperimentendedafter5trials,aspertheinstructions. d -$  *XԀ  7    ݀Nosignificantdifferencebetweenthebidsfromthetwoexperimentswasfound.InthemodifiedSmithAuction,  respondentsweretoldthetotalnumberofhouseholdswhowouldbeaskedtocontributeandthetotalcostofthenewtrees.Threepossibleoutcomeswereexplainedtoeachrespondent.First,ifthesumofthepaymentswaslessthanthecost,thenthehouseholdspaidnothingandnoadditionaltreeswouldbeplanted.Second,ifthesumofthepaymentsequaledthecost,theneachhouseholdpaidtheamounttheybid,andthetreeswouldbeplanted.Finally,ifthesumofthepaymentsexceededthecost,theneachhouseholdwouldpayafractionofwhattheybidsothatpaymentsequaledthecostofthenewtrees.  -$  *XԀ  14    ݀i *Specifically,thelikelihoodfunctionwasestimatedusingaNewtonRaphsonmaximizationtechniqueandthe  covariancematrixwasestimatedusingtheprocedureofBerndt,Hall,HallandHausman.SeeJudgeetal.(1980)orGreene(1993).#* i{# Table_M X -$  *XԀ  a    ݀LikelihoodRatioteststatistic)*anddegreesoffreedom#*)#ԀforthenullhypothesisthattheCVandSMdatasetsshould  bejointlyestimated.&d9 Z 6Times New Roman RegularTABLE A&@ Z 6Times New Roman Regular&d9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular  -$  *XԀ  20    ݀MitchellandCarson(1989)defineconvergentvalidityas"thecorrespondencebetweenameasureandother  measuresofthesametheoreticalconstruct...Inconvergentvalidityneitherofthemeasuresisassumedtobeatruermeasureoftheconstructthantheother.(p.204)" ( -$  *XԀ  16    ݀ThetestwasmadeusingaLikelihoodRatiotestcomparingCVWTA(2)andCVWTA(3).Thechisquared  teststatisticwas11.7with5degreesoffreedom.&d9 Z 6Times New Roman Regular  -$  *XԀ  15    ݀o *Theequationsusedinthelikelihoodfunctionarehighlynonlinearintheparameterstobeestimated,andtheCES  andCRRAequationsdidnotconverge.Tofacilitatetheestimationofmorecomplexmodels,futurestudiescouldbedesignedwithwidervariationinthebidspaceandtheattributespacefortheenvironmentalcommodity.Themeasureofincomeusedintheestimationalsodeservesmoreattentionbothinthiscalibrationmodelandinotherdemandmodels.Becausewillingnesstopayisasmallinrelationtototalincome,thestatisticalestimationprocesscouldbeimprovediftheanalysiswasbasedonsomefractionofincome.Forexample,onecouldreplaceincomewiththehouseholdbudgetfordiscretionaryspending.Ofcourse,determiningsuchabudgetisnotatrivialissue.#* o{#\  `*Times New RomanTT -$  *XԀ  1    ݀SeeFederalRegistervol.59.no.5(January7,1994)page1146. F -$  *XԀ  10    ݀Forexample,inequation(5)#=#intercept+#hhsize*HHSIZE+#gradsch*GRADSCH+#school*SCHOOL+  #parkview*PARKVIEW+#fincollege*FINCOLLEGE.)*Notethattheparametersandhavedifferentinterpretations z thanand#,sothecoefficientsontheseparametersarenotcomparable.#*)[# -$  *XԀ  11    ݀Specificationsofmodels(1)and(2)includingfincollegewererejectedduetomulitcollinearityproblems.U i -$  *XԀ  12    ݀AvaluesofCPandCALBIDwerecalculatedusingindividualcharacteristicsandthecoefficientestimatesin  Table2.ThemeansoftheindividualestimatesarepresentedinTable3withtheirstandarddeviations. Q3wQXXQ3wԀSTACKALIGN{  #XXXQX## 3XX#(7)&`WTP_i~=~C^P(X_i)`+`Y_i``e^{ln(Y_i)~+~{(1alpha(X_i))}over{alpha(X_i)}  Ѐ`(ln(E_{low})``ln(E_{high}))}`+`#(8)&`WTA_i`=`C^A(X_i)`+`e^{ln(Y_i)~+~{(1-alpha(X_i))}over{alpha(X_i)}`(ln(E_{high})`-`ln(E_{low}))}~-~Y_i`+`F}&d9 Z 6Times New Roman Regular(9 Z 6Times New Roman Regular q -$  *XԀ  13    ݛ *SeeMcConnell(1990)foradiscussionofassumptionsaboutthescalefactorandtheconditionsunderwhich  CameronandJames'modelisidenticaltoHanemann's.#* |# (O6)WTPMM=iCMMCP(1XMMi){YMMi_[YMM h1MMEhMMhMMh(MMhX89iMMPh)MMiv(  . 1  H ( XMM Pi )0 }1 } }F }(~ }XMM 8i })T ) ( EMMD h1MM| hMM hMM h(MM$ hXm 9iMM h)MM* high EMMh1MMhMM hMMIh(MMohX9iMMh)MMulow).]pMM1MM}h1MMhMMhMM9h(MM_hX9iMMh)|  -$  *XԀ  2    ݀Forexample,seeCameron(1992),EomandSmith(1994)andBlackburnetal.(1994).Oneexceptioniswork  bySchulze,MeClellandandLazo(1994)whoproposetransformingthebidsfromopenendedCVsurveyswithaBoxCoxspecificationuntiltheyfitanormaldistribution.(9 Z 6Times New Roman Regular  -$  *XԀ  4    ݀j *McConnell(1990)developsthevariationfunctionasachangeintheexpenditurefunctions.#* j{# s -$  cXԀ  13    ݛl *SeeMcConnell(1990)foradiscussionofassumptionsaboutthescalefactorandtheconditionsunderwhich  CameronandJames'modelisidenticaltoHanemann's.#c |#ԫ<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P  -$  *XԀ  18    ݀ThevaluesofCAandPREDBIDarethemeansofalltheindividualvalues.Theindividualvalueswere  calculatedusingthecoefficientestimatesinTable6andtheindividual'scharacteristics.CAandarelinear z combinationsofthevariablesinTable6.<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P y(Jy7y)yWTPMM84iyNyCMMP+y(cyXMM4iy)7yyYMM 4i?yyeMM<lnMMq<(MM<Y iMM<)MMH< TMM(MM1MM1 MMu MM (MM X& yiMM> )MMd )MM7 MMy (MM X iMM )MM <(MM <lnMM" <(MMH <E lowMM <)MM8 <MM <lnMM <(MM <EJ highMM <)MM<)>yy|(J|8|)|WTAMM87ij||CMMyA|(|XMM_7i|)|?|eMM?lnMM?(MM!?YdiMM|?)MM? PWMM](MM1MMMMMMA(MMgX|iMM)MM)MMMM(MM)XriMM)MM. ?(MMT ?lnMM ?(MM ?E highMM ?)MM ?MM. ?lnMM ?(MM ?E lowMMj ?)MM ?) | |YMM 7i>||F dTable_A&0 d d) `(CG TimesScalableTable_CTable_D dTable_E&O Z 6Times New Roman Regular(x-  Z 6Times New Roman RegularTable_GTable_H M -$  *XԀ  17    ݀@ *@XMX @Ԁ#@ X@XM#Alternatively,thebidscouldbecalibratedbyfirstestimatingequation(8)withoutasystematicbiastermand  usingtheseestimatestocalculateanuncalibratedpredictedWTA.ThenthepredictedWTAbidcouldbecalibratedbyestimatingCAinaseparateequationandsubtractingitfromtheuncalibratedWTA.#* @{#ԀThesamechoiceactuallyexists j forcalibratingopenendeddata.OnecouldeithersubtractestimatedbiasfromtheactualCVbid,aswedid,orusethevaluesofand#fromTable2tocalculateapredicatedWTP.  CRight ParRight-Aligned Paragraph NumbersI.A.1.a.(1)(a)i)a)3|w$<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P( U$  <p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7PHP LaserJet 5/5M - Standard,,,,0<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P N -$  *XԀ  9    ݀Unfortunatelythelabexperimentdatasetwastoosmalltoestimatethecalibrationmodel.Totesttheaccuracy  ofthelabbids,weestimatedequation(5)withCPspecifiedasthefunctionofanintercepttermandadummyvariable z forparticipationinthelabexperimentusingadatasetcombiningthesurveyandlabexperimentdata.Theresultssuggestthatthelabexperimentbidswerenotsubjecttosystematicbias. |$}$E~$$E$E$${$e$e$a"t<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P  -$  *XԀ  a    ݀PREDBIDwascalculatedforeachindividualfromtheequation: X e5vw/` p # `E"<yzdGZ e " Ѐwherewasspecifiedasafunctionofanintercept,FEEL3,FEEL8,REACT11,SUBST,EXQUALSHand d  BSTCHNCasinTable6.WTAwascalculatedforeachrespondentandthemeanispresentedinthistable.TheuncalibratedWTAisbasedoncoefficientsfromthismodelestimatedwithoutasystematicbiasparameter(theresultsarenotshowninTable6).ThecalibratedresultsarebasedonthecoefficientsinTable6.(x-  Z 6Times New Roman Regular<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P <p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P STACK{(1)~~U(Y_iITALBOLDwtp_i,`E_{high};`X_i)~=~U(Y_i,`E_{low};`X_i)#  (2)~~U(Y_i+ITALBOLDwta_i,`E_{low};`X_i)~=~U(Y_i,`E_{high};`X_i)}<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P n(Jn1n)hnUn(nYMMx)innwtpMM)in,LnEMM)highzn;nXMM$)iDn)nrnUn("nYMM)in,nEMMD )low n;& nXMM )i n)|(J|2|)h|U|(|YMMx7i||wtaMM7i|,L|EMM7lowT|;|XMM7i|)|L|U|(|YMM\7i||,|EMM 7high |;& |XMM 7i |)<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P(9 Z 6Times New Roman Regular<jj:Default ParaXXXQc\  `$Times NewRomanQXXXQc\  `$Times NewRomanQԫ<p`(4Line Draw 12cpiFull-PEdp@h'@* `(CG TimesScalablell-PEXXw P7XP) `(CG TimesScalablell-PEc P7P$35;AGMU]e1I.A.1.a)(1)(a)(i)(a)\  `$Times NewRoman# e37=CIQYag1.a.i.(1)(a)(i)1)a) STACK{(3)~~italBOLDwtp_i~=~f(Y_i,`E_{high},`E_{low},`(X_i);`X_i)# `` (4)~~italBOLDwta_i~=~g(Y_i,`E_{high},`E_{low},`(X_i);`X_i)} %n(]n3n){nwtpMMo)innfn(nYMMK)ikn,nEMM )highn,nEMMw)lown,Unn(nXMMU)iun)n;nXMMW )iw n)|(J|4|)h|wtaMM\7i|r|g|(|YMM^7i~|,|EMM 7high|,$|EMM7low,|,h||(|XMMh7i|)|; |XMMj 7i |)(9 Z 6Times New Roman Regular(x-  Z 6Times New Roman Regular&O Z 6Times New Roman Regular&O Z 6Times New Roman Regular(x-  Z 6Times New Roman Regular MWTA~=~e^{Log(Y)~+~{(1-alpha)}over{alpha}`(Log(E_{high})`-`Log(E_{low}))}~-~Y#$XXjWTAXXjXXjeMM:LogMM(MMJYMM)MMXMMO(MMO1MM3OMMwOMMO)MM3MM#(MMgLogMM3(MMwEhighMM )MM MM# LogMM (MM3 Ew lowMM )MMQ )XX jXX jY g -$  *XԀ  19    ݀Theanalysisincludedvariablestoaccountforissuessuchaswhetherthetravelcoststudyincludedtraveltime,  verygeneralmeasuresofsitequality,thetypeofCVquestionandothercharacteristicsofthestudy. uU!   ݛX\XXCXMXXX\Ԁ  )XQhXXCXMGXMXX)XQh̀@F F AConsistentMethodforCalibratingContingentValueSurveyData  CarolMansfieldSchooloftheEnvironmentDukeUniversityBox90328Durham,NC277080328(919)6138028  Acknowledgments:IwouldliketothankWilliamEvansforallhishelp.IamalsogratefultoMaureenCropper,JohnHorowitz,GlennHarrison,RandallKramer,KerrySmith,andtwoanonymousrefereesfortheirmanyusefulsuggestionsandDavidBrookshire,DonaldCoursey,RichardBishop,ThomasHeberleinandMichaelWelshforsupplyingmewiththeirdata.Anyremainingerrorsaremyown.  ! #)XQhXXGXM#7XMXX)XQhAbstract    Thispaperoutlinesastatisticalmethodforcalibratingthedatafromcontingentvalue(CV)surveysderivedfromtheassumptionthatindividualsmakeconstrainedutilitymaximizingdecisions.IndividualsreactdifferentlytoCVquestions,andaneffectivecalibrationmethodmustaccountforthis.#)XQhXX7XMz#7XMXX)XQhԀThemethodproposedinthispaperallowsustodeterminetheinfluenceof 0  individualcharacteristicsonbias,asdistinctfromtheirinfluenceonthepreferenceparameters.Toillustratethelogicofthisapproachaspecificfunctionalformforindividualpreferenceswasusedtoderiveclosedformanalyticalexpressionsforanindividual'swillingnesstopay(WTP)andwillingnesstoaccept(WTA).Thesefunctionsallowsystematicdeviationsinindividualresponsestobeexplicitlymodeled#)XQhXX7XM#7XMXX)XQhbyprovidingastructuralinterpretationoftheerrorterm#)XQhXX7XM`#7XMXX)XQh.The 0  frameworkisappropriateforbothopenendedanddichotomouschoicedata.WeillustratethisapproachwiththreeseparateCVdatasetsusingbothtypesofresponses.Ineachcase,thecalibratedCVresponsesarecloseinvaluetoresponsesderivedfromsimulatedmarketexperiments.#)XQhXX7XM#GXMXX)XQh   8)Qh)XXdd8 1.Introduction     #)XQhXXGXMU #@XMXX)XQhContingentvalue(CV)surveysareusedtoestimatetheeconomicvalueofnonmarket  goods,especiallyenvironmentalgoods.AmajorconcernwithCVsurveysisthepotentialforwhathavelooselybeencalledhypotheticalandstrategicbiasesintheanswerstoCVquestions.Foravarietyofreasons,oftenindividualspecific,anrespondent'sanswertoaCVquestionmaydifferfromhisorhertruevalueforthegood.Toaddressthepotentialproblemofinaccuratebids,theinitialversionoftheproposedrulesfortheOilPollutionActof#)XQhXX@XM, #GXMXX)XQh1990#)XQhXXGXM #@XMXX)XQhcalledforallCV @  valuestobedividedinhalf.K +  1      ׀#)XQhXX@XM #GXMXX)XQhԀTheprovisionwasintendedasachallengetopractitionersto   developamethodforcalibratingthedatafromCVsurveys.  ThispaperoutlinesastatisticalmethodforcalibratingthedatafromCVsurveysderivedfromtheassumptionthatindividualsmakeconstrainedutilitymaximizingdecisions.Themethodallowsustodeterminetheinfluenceofindividualcharacteristicsonbias,asdistinctfromtheirinfluenceonthepreferenceparameters.Toillustratethelogicofthisapproachaspecificfunctionalformforindividualpreferenceswasusedtoderiveclosedformanalyticalexpressionsforanindividual'swillingnesstopay(WTP)andwillingnesstoaccept(WTA).Thesefunctionsallowsystematicdeviationsinindividualresponsestobeexplicitlymodeled#)XQhXXGXM #@XMXX)XQhbyprovidinga   structuralinterpretationoftheerrorterm#)XQhXX@XM$#.Theframeworkisappropriateforbothopenended "  anddichotomouschoicedata.  Therandomutilitymodelframeworkfocusesattentionontheerrorterm,specificallyonthepossibilitythatthereisanindividualspecific,systematiccomponenttotheerrorterm T(#& thatisrelatedtobiasinCVresponses.Evidencefromexperimentaleconomicsandpsychologyfindsthatrespondentsmayreactdifferentlytothesamesurveyquestionorlaboratoryexperiment.Aspectsofthereactionmaybecorrelatedwithobservablecharacteristics,suchaseducationorage,whileotheraspectswillappearrandomtotheresearcher.ThusisitimportantforcalibrationtechniquestoallowfortheinfluenceofindividualcharacteristicsontheexistenceandthesizeofanybiasinCVresponses.  WeillustratethisapproachwiththreeCVdatasetsusingdatafrombothopenendedanddichotomouschoiceresponses.TheparticularapplicationswereselectedbecausecomparablesetsoflaboratoryorsimulatedmarketdataexistforeachofthethreeCVdatasets.Thisallowsacomparisonbetweentheresultsfromtheproposedcalibrationmodelandthelaboratoryorsimulatedmarketdata.Ideally,thecalibratedCVresultsshouldbecomparedtovaluesfromactualmarkettransactions,ratherthantheresultsfromlaboratoryorsimulatedmarketexperiments,whichmayalsobebiased.ThecalibrationmodelproposedinthispapercanbeappliedtosimulatedmarketdataaseasilyasCVdata,andfortwoofourdatasetsweareactuallyabletoestimatewhetherthesimulatedmarketdatasuffersfrombiasedresponses,aswell.  TheapproachderivedheredoesnotrequireadditionaldatabeyondtheCVsurveyitselftoimplement,thusitcanbeusedtocalibratedatameasuringuseornonusevalues.OthercalibrationtechniquesforCVdatarequireactualmarketdatafromweaklycomplementary goodsortheidentificationofasurrogatemarketgoodforthenonmarketgoodvaluedinthe P(#& CVsurvey.U +  2      ׀Butinmanycases,especiallywherenonusevaluesareimportant,itmaybe   impossible,ifnotcontradictory,todefinetheappropriatesetofweaklycomplementarymarketgoods.Instead,weinterpretthetaskofdevelopingacalibrationmodelforCVresponsesasalogicalproblemthatconsiderswhethertherearesufficientmodelrestrictionsandsampleinformationtoidentifythepreferenceparametersanddistinguishthemfromsourcesofbiasinCVresponses.  Whilethechoiceoffunctionalformforutilityisanimportantmaintainedassumptionconditioningtheresultsderivedfromthisapproachtocalibration,similarissueshavebeenroutinelyaddressedinmodelingconsumerdemand.Agreatdealofworkindemandanalysishasfocusedondevelopingstatisticalmodelsthatcanbeusedtotestdemandtheory.Inmuchofthiswork,researchershavebeenforcedtomakeassumptionsaboutthefunctionalformforeitherthedirectortheindirectutilityfunctions.However,evenwhenthefunctionalformsarerestrictive,theresultingestimatescanbeinformativeandprovideafoundationforfutureresearch(seeDeaton(1986)forareview).Furthermore,failuretoaccountforresponsebiasintheestimationprocesswillyieldparameterestimatesthatareacompositeofpreferenceandresponsebiaseffects.Becauseresponsebiasesmaybepositiveornegative,theestimateswillbedifficulttointerpret.  Thispaperisorganizedasfollows.Section2developsamodelofCVbidsandderivesstructuralequationsforWTPandWTA.Thecalibrationmodelisappliedtothethreedata p&!$ setsinSection3.TheresultsarediscussedinSection4,andSection5containsideasforfurtherresearch. 2.AModelforCalibration P     AtypicalCVsurveydescribesanenvironmentalgoodandthenproposesachangefor 0  betterorworseinsomefeatureofthatgood.Therespondentsareaskedtodecidehowmuchtheywillpayfortheimprovementorhowmuchcompensationtheyrequireifthechangeisfortheworse.Tothinkabouttheresultsfromthiselicitationprocess,assumeeachrespondentreceivesutilityfromtwogoods:theenvironmentalgood(E)thathastwolevels,EhighandElow,  andaHicksiancompositegood,representedbyincome(Y).Inthisframework,anindividual'struemaximumwillingnesstopay,wtpi,andminimumwillingnesstoaccept,wtai, p satisfytheequalities:5X^[ZJFz ,H p @Xdddddddd@EU4Hq 4whereXiisavectorofindividualcharacteristicsandattitudes. D 5fhg_OKz|~w p @Xdddddddd@E.z"wQ "?  Foragivenutilityfunctionwithparametervector,theseequationscanbesolved \  explicitlyforwtpiorwtai !  ЀC b $#eCGXMXX)XQh  MostCVstudiesrelyonbidfunctionsassumedtobelinearinobservedcharacter#)XQhXXGXM'#GXMXX)XQhistics., +  3      ׀ %O!& Byselectingaspecificfunctionalformforutility,explicitclosedformsolutionsforWTPandWTAcanbederived.#)XQhXXGXM'#@XMXX)XQhԀThesestructuralequationswillallowustodecomposetheindividual'sbid ($* intoapreferencebasedcomponentthatisourestimateofWTPandWTAandabiastermthatidentifiessystematicdeviationsfromtheassumptionsofthemodel.#)XQhXX@XM(#GXMXX)XQh h   Wechosetoestimatetheequationsbasedonarandomutilitymodel(RUM)whereWTP(orWTA)istreatedasarandomvariable.TherandomutilityapproachtoanalyzingCVdatawaspopularizedbyHanemann(1984).FromHanemannandKanninen(forthcoming),"onewantstoformulateastatisticalmodelfortheCVresponsesthatisconsistentwithaneconomicmodelofutilitymaximization.(p.4)"ThecalibrationequationsdevelopedbelowexploitthelinkbetweentheeconomicandstatisticalmodelsthatisthefoundationoftheRUMframeworktoidentifydeviationsinCVbidsfromtrueWTPorWTA.Ourmodelusesthedirectutilityfunction,ratherthantheindirectutilityfunction(Hanemann(1984))oravariationfunction(McConnell(1990))W +  4      ,becausethecalibrationmethodwasdevelopedinconjunctionwithefforts  toestimatetheparametersofutilityfunctionsassociatedwithWTPandWTAresponses(Mansfield(1996)).However,onecoulddevelopsimilar,closedformsolutionsforWTPandWTAfromindirectutilityfunctions.  Twosourcesoferror,systematicandrandom,maycauseanindividual'sbidtodifferfromtheamountheorshewouldactuallypayforthegoodifamarketexisted.#)XQhXXGXM*#GXMXX)XQhSystematicoveror  understatementofWTPandWTAmightoccurduetofactorssuchastheamountoftimetheindividualhastoanswerthequestion,thewordingofthesurvey,orthestructureoftheexperiment.HoehnandRandall(1987)andCrockerandShogren(1991)developtheoreticalmodelsofCVbidsthatpredictdeliberateoverorunderstatementofWTP./Z +  5      ׀Inaddition,thereisa h& largeliteratureontheincentivepropertiesofvarioussurveyandexperimentalformatsandthelikelihoodforstrategicbehavior.0J +  6      ׀Forexample,Bohm(1984)hypothesizesthatindividualswho 8"* favortheactionproposedintheCVsurveymightpurposelyinflatetheirWTPiftheydidnotbelievethesurveywouldactuallybeusedtodeterminetheamounttheyhadtopay.Horowitz(1993)discussesthepotentialformisunderstandingsbetweentheanalystandtherespondenttocontributetosystematicbiasinresponses.  #)XQhXXGXM/#@XMXX)XQhFurthermore,whetherandbyhowmuchindividualbidsdifferfromtheirtruevaluewill P  dependontherespondents'characteristics,attitudesandinterpretationofthesurvey.#)XQhXX@XMV4#GXMXX)XQhԀ#)XQhXXGXMZ5#@XMXX)XQhEvidence    thatindividualswillreactdifferentlytoidenticalincentiveschemescanbefoundinexperimentssuchasAndreoni(1995)ontheprovisionofpublicgoods.#)XQhXX@XM5#@XMXX)XQhԀHerrigesandShogren(1996)found   thatlocalresidentsandrecreationistsexhibiteddifferentanchoringbehaviorinasurveyvaluingwaterqualityimprovementsinanIowalake.Studiesfromthepsychologyliterature,reviewedinKrosnick(1991),indicatethattheresponsestrategyanindividualusestoanswerasurveyquestionmaybeafunctionofthehisorherpersonalcharacteristics.Anadhoclinearspecificationdoesnotallowtheanalysttodistinguishtheinfluenceofacharacteristic,suchaseducation,onpreferencesfromtheinfluenceofthatcharacteristiconthepropensityofrespondentstosystematicallyinflateordeflatetheirCVbids.Becauseindividualcharacteristicsandattitudesmayaffectbothpreferencesandtheresponsestrategyanindividualadopts(forexample,systematicoverstatement),weattempttodecomposetheinfluencesofrespondentattributesonthesetwoelementsofaCVbid.#)XQhXX@XM6#GXMXX)XQh "   Beyondthesesystematicinfluences,theCVbidswillbesubjecttorandomerrorduetounobservedheterogeneityintherespondentsandtheinabilityofthevariablesavailabletotheanalysttoperfectlymeasurethepertinentattitudesandcharacteristicsoftherespondents.  Tocapturesystematicvariationsinindividualbidsasystematicbiasparameterisaddedtoequations(3)and(4)inadditiontoarandomerrorterm.Equations(5),(6),(7)and(8)arethebidfunctionsderivedfromthreeutilityfunctions:CES,constantrelativeriskaversion(CRRA),andCobbDouglas.  @)$4 *w.Addd Xdd Xdd X(#(#w,9dd ,cdd +  #  # CES:(Y#+(1)E#)1/# * d * ' 'm58D95%!`|ig `EMgB|o+m +! +CRRA:(Y(1)/(1))+(E(1)/(1)) *   * ' f 'm56RT5%!`| `EM oN +m +! f +CobbDouglas:lnY+(1)lnE *   * '=  'm5PV]5%!`|  `EM=9&u+m"=   "  Ineachequation,CPorCAmeasuresthesystematiccomponentoftheerrortermwhile Y  andF,whichhavemeansofzero,measuretherandomcomponent.Theparametersoftheutility   functiondeterminethecharacteristicsoftherespondent'spreferences,andalloftheparameterscanbespecifiedasafunctionofindividualcharacteristicsandattitudes.Forexample,intheCESutilityfunctionincomeelasticityequals1/#,whilefortheCRRAfunctiontheincome I elasticityisafunctionofand.#)XQhXXGXM:#GXMXX)XQh a  3.ThreeApplicationsoftheCalibrationModelUsingCVData 1"    Thissectiondescribestheresultsfromthreeapplicationsofthecalibrationmodel.The # firstdatasetisfromBrookshireandCoursey(1987)valuingthedensityoftreesinaneighborhoodpark.TheauthorsconductedanopenendedCVsurveyandalaboratoryexperimentforthecommodity.TheothertwoCVdatasetsareWTPandWTA#)XQhXXGXMC#@XMXX)XQhdichotomous '!#! choicedatafromastudyvaluingdeerhuntingpermitsbyBishop,HeberleinandWelsh(BHW,seeWelsh(1986)andBishopandHeberlein(1990)).#)XQhXX@XMF#GXMXX)XQhԀInthisexperiment,datafromdichotomous *%% choiceCVbidswerecomparedtobidsfromidenticalSMexperimentsforaspecialoneday  ,Y'' permittohuntdeerintheSandhillWildlifeDemonstrationAreainWisconsinpriortotheopeningoftheofficialWisconsinhuntingseason. CalibratingOpenEndedWTPData  8   AlistofthevariablesandsummarystatisticsfortheBrookshireandCourseyCVandlabexperimentdatacanbefoundinTable1.TheWTPquestionaskedtherespondentstheirWTPfortwoincreasesinthenumberoftreesplantedinanewneighborhoodparkfrom200to225andthen250.ThefirstpartoftheexperimentconsistedofaCVsurveyandamodifiedSmithAuction(withoutactualpaymentofbids)conductedthroughdoortodoorinpersoninterviews.Wepooledtheresponsesintoonedatasetwith170observations.; +  7      ׀Thesecondcomponentof X  theiranalysisconsistedofalaboratoryexperimentinvolving27respondentsconductedatalocalschoolusingthemodifiedSmithauctionwithupto5repeatedtrials.: +  8      ׀Thelaboratorydata (x provideabenchmarkforthecalibration.}  +  9        Ѐ  Asdiscussedinsection2,thesystematicbiascoefficientmaybedeterminedbyfeatures H oftheexperimentitself,inadditionaltocharacteristicsoftheindividualrespondents.Unfortunately,becausetheBrookshireandCourseydatasetincludesalimitednumberofvariables,thisapplicationispresentedsimplytoillustrateoneuseofthiscalibrationmodel. 0    Table2reportstheresultsfortheCVWTPdatafromthreespecificationsderivedfromCESutility(equation(5))andonespecificationfromCRRAutility(equation(6)).Here,#,, h andCParemodeledaslinearcombinationsoftheobservedcharacteristics.L +  10      ׀Several  specificationsarereportedtoillustratetheimpactofchangingthespecificationonthesystematicbiascoefficientsandtheresultingcalibratedbids.M +  11      ׀ P    #)XQhXXGXMF#@XMXX)XQhSincetheaimofthisexerciseistopredicttrueWTP,itmightbemoreappropriateto    judgespecificationsusingatestbasedonmeansquarederror,oranoncentralFtest,ratherthanthestandardFtest.SeveraltestsofthistypearedescribedinWallace(1972)andWallaceandToroVicarrondo(1969),includingatestthatfocusesonforecastingtheconditionalmeanofthedependentvariable,ratherthanthevector.#)XQhXX@XMdQ#GXMXX)XQhԀForexample,inTable2onecouldcomparethe X  restrictedmodelCES(3)toeitherCES(1)orCES(2).UsingCES(1)astheunrestrictedmodel,theFstatisticis0.32with(4,153)degreesoffreedom.BoththestandardFtestandanFtestwithnoncentralityparameterofonehalffailtorejectCES(3).  Thebiasterm,CP,ispositiveforalltheparticipantsintheCVsurveyregardlessofwhich H modelisused,suggestingthattheparticipantssystematicallyoverstatedtheirWTP.Whenthebiasparameterisnotspecifiedasafunctionofothercharacteristics,thetwodifferentassumptionsaboututility,CES(3)andCRRA(4),yieldsimilarestimatesofthesystematicbiascoefficient(CPintercept)about$10.Asexpected,duetothelimitedqualityoftheinformation " available,noneofthevariablesusedtopredictsystematicbiasinCES(1)and(2)inTable2aresignificant.  Tocalibratethedata,weattempttodecomposetheindividualbidsseparatingtheportionofthebidthatisrelatedtopreferencesfromtheportionthatmightbeattributedtobias.InTable3thedataiscalibratedbycalculatingtheamountofbiasforeachindividualanswer,Cip,and #, subtractingthisfromtheamounttheindividualactuallybid,BIDi,toarriveatthecalibratedbid, %X . CALBIDi.SoCALBIDi=BIDiCiPforeachindividualiintheexperiment.Thefirsttworowsof p&!0 Table3presentthemeanandstandarddeviationofthebidsfromtheactualCVsurveyandlabexperimentdata.TheotherfourrowscontainthemeanandstandarddeviationofCPiand h CALBIDifromthefourdifferentmodelsestimatedinTable2.O +  12      ׀IfCALBIDiwaslessthan  zero,thenitwassetequaltozero.(ThusthemeanoftheCALBID'sinTable3willnotequalthemeanofBIDminusthemeanofCP.)#)XQhXXGXMgS#@XMXX)XQhInmodels(1)and(2)ofTable3themeanvalueofthe P  systematicbiastermisgreaterthantheactualbidsmadebymanyoftheparticipants.ThehighmeanvaluesforCPinmodels(1)and(2)mayresultfromspecifyingCPasafunctionof  p  respondentcharacteristicsnoneofwhicharesignificantinTable2andallofwhicharepositive.#)XQhXX@XM\#GXMXX)XQh @    ThemeansoftheCALBIDi'srangefromalowof$2.45inmodel(2)toahighofover X  $10inmodels(3)and(4).(Inmodels(3)and(4),CPwasnotspecifiedasafunctionofany  observedcharacteristics.)Theactualmeanbidfromthelabexperiment,$8.48,fallswithintherangeofestimatedCALBID's.#)XQhXXGXM]#GXMXX)XQh   CalibratingDichotomousChoiceCVSurveyData  `   #)XQhXXGXM_#@XMXX)XQhԀ#)XQhXX@XMW`#7XMXX)XQhԀTheBishop,HeberleinandWelsh(BHW)experimentconsistedof#)XQhXX7XM`#7XMXX)XQhbothWTPandWTA  #)XQhXX7XMa#7XMXX)XQhCVsurveysandsimulatedmarket(SM)surveysadministeredby#)XQhXX7XMa#7XMXX)XQhmail#)XQhXX7XMb#7XMXX)XQh.#)XQhXX7XMHb#@XMXX)XQhԀThesurveysvalued 0  specialonedaydeerhuntingpermitsthataredistributedfreeeachyear#)XQhXX@XMb#7XMXX)XQhto150hunters#)XQhXX7XM?c#@XMXX)XQhԀ#)XQhXX@XMc#7XMXX)XQhthrougha " lottery#)XQhXX7XMc#@XMXX)XQhԀheldbytheWisconsinDepartmentofNaturalResources.#)XQhXX@XM9d#GXMXX)XQhԀHunterswhohadlostthelottery P$ weresenttheWTPquestions.Halfofthesehunterswereofferedachancetoactuallypurchaseadeerhuntingpermitforasetprice.Theotherhalfreceivedasimilarhypotheticaloffer.TheWTAquestionsweresenttotheluckyhunterswhohadwonpermitsinthestatelottery.Thesehunterswereofferedachancetoselltheirpermitsbackforafixedpriceagain,halfreceivedarealoffer,whilehalfreceivedahypotheticaloffer.Inalltheexperiments,thepriceofthepermitwasvariedoverthesample.  #)XQhXXGXMd#@XMXX)XQhIndichotomouschoicequestions,respondentsarepresentedwiththeproposedchangein '(#2 thepublicgoodandthenofferedtheoptionofeitherpayingorreceivingsomefixedamountofmoneytosecurethechange.Theamountoftheofferisvariedoverthesample,andpeoplemustsimplyanswer"yes"or"no"totheCVquestion.#)XQhXX@XM>g#GXMXX)XQhԀ#)XQhXXGXMh#@XMXX)XQhUsingdichotomouschoicedata,theRUM  frameworkistypicallyestimatedusingeitheramethodsuggestedbyHanemann(1984)orCameron(seeHanemannandKanninen(forthcoming),p.6).#)XQhXX@XMi#Wechosetousethemethod P  outlinedinCameronandJames(1987)andCameron(1988),GXMXX)XQhwheretheoutcomeofthechoice    processistreatedasrandomvariable#)XQhXXGXMj#)X)XQh.#)XQhX)-k#GXMXX)XQh$S$$X$ +  13      ׀Thedifferenceisthatthismodelpredictsthatan & v  individualwillanswer"yes"toaWTPquestioniftheofferedamountislessthanorequaltotheindividual'strueWTPplusthesystematicbiastermandanerrorterm.Assumingtheresponseerrorisindependentlyandnormallydistributedwithmeanzero,theresultinglikelihoodfunctioncanbeestimatedusingamaximumlikelihoodtechnique.<$ +  14          BidfunctionsconsistentwithCobbDouglaspreferences(equations(7)and(8))wereusedtoconstructthelikelihoodfunction.HJ +  15      ׀ConsideringfirsttheCVWTPresultsinTable4  #)XQhXXGXMpk#@XMXX)XQhcolumn6#)XQhXX@XM[o#GXMXX)XQh,#)XQhXXGXMo#@XMXX)XQhthemodel#)XQhXX@XMo#@XMXX)XQhpredictsthattheCVdataareconsistentwithtrueWTP#)XQhXX@XM5p#GXMXX)XQhԀthebiasterm,C,is N insignificant.ComparingtheCVWTPresultswiththeSMresultsincolumn5providessupportforthisconclusion.Incolumn7equation(7)wasestimatedwithadatasetcombiningtheCVandSMWTPdata.Comparingcolumn7withcolumns5and6wecannotrejectthehypothesisthatthedatacanbejointlyestimatedatthe5%levelusingaLikelihoodRatiotest.Notsurprisingly,theestimatesfortheexpectedvalueofWTParequiteclose:$31intheSMand$35intheCVsurvey(ascalculatedbythesurveyauthors,seeBishopandHeberlein(1990)). n&   TheCVWTAestimatesinTable4column3suggestthattheCVbidsoverstatetrueWTA$$$"$thesystematicbiascoefficientispositiveandsignificant.ComparingtheCVresultswiththeSMresultsincolumn2,theestimatedvaluesofremaincomparable,butthebiasterm  andstandarderror(Cand%)aremuchlargerintheCVestimation.Inthiscase,wecanrejectthe 8 hypothesisthatthedatacanbejointlyestimatedatanyconfidencelevelusingaLikelihoodRatiotestcomparingtheresultsfromthejointCV/SMdatasetincolumn4withcolumns2and3.#)XQhXXGXMp#@XMXX)XQhIn    linewiththeseresults,theexpectedvaluesofWTAcalculatedbyBishopandHeberleinfortheCVandSMexperimentsarenotnearlyascloseinvalue$153intheSMand$420intheCVsurvey.Onedisturbingresultisthe#)XQhXX@XMu#@XMXX)XQhԀpositiveandsignificantsystematicbiascoefficientintheSM @  WTAmodel.TheSMexperimentofferedrealcashpaymentsfortheparticipants'deerhuntingpermits.ThusonemightexpectthebidsfromthisexperimenttoreflecttrueWTA.#)XQhXX@XMw#@XMXX)XQhԀHowever,in  thefullerspecificationsthatincludeindividualcharacteristicspresentedinTable6,theSMsystematicbiascoefficientisinsignificant,whiletheCVsystematicbiascoefficientremainssignificant.#)XQhXX@XM^x#GXMXX)XQh H   ThepreliminarytestssuggestthattheCVWTPdatadonotneedtobeadjustedforsystematicbias,whiletheCVWTAdataneedtobecalibrated.Table5liststhevariablesusedtoestimatethecalibrationequationsalongwiththeirmeansandstandarddeviations.Theparameter 0  wasspecifiedasafunctionoftherespondents'feelingsaboutdeerhunting,ameasureofthe " numberofsubstitutesthehuntersfelttheyhadforhunting,andthequalityofdeerhuntingatSandhill.Thesystematicbiascoefficientwasmodeledasafunctionoftherespondents'feelingsabouttheirrighttohunt,theirreactionstothesurveyandeducation.#)XQhXXGXMy#@XMXX)XQh  Table6reportstheresultsfromreestimatingtheWTAmodelsinTable4specifyingthe 8"* parametersasfunctionsoftheindividuals'observedcharacteristics.TheCVandSMdatasetswereanalyzedseparatelyusingthreedifferentspecificationsoftheCVdataandonespecificationoftheSMdata.ThelasttwocolumnsofTable6,CVWTA(3)andSMWTA,havethesamespecification.TheseresultsconfirmthattheCVbidsoverstatedtrueWTA(CAinterceptispositive '(#2 andsignificantinCVWTA(3)).However,themodelnowsuggeststhattheSMdataaccuratelyrepresenttrueWTA(CAinterceptisnotsignificantlydifferentfromzerointheSMWTAresults). *%6 WhenCAisspecifiedasafunctionofindividualcharacteristicsandattitudesinCVWTA(1)and ,`'8 (2),noneofthesystematicbiascoefficientestimates#)XQhXX@XM|#@XMXX)XQharesignificant#)XQhXX@XM#@XMXX)XQh.Howeveronecannotreject  thehypothesisthatthesystematicbiasparameterestimates#)XQhXX@XMԀ#7XMXX)XQhinCVWTA(2)#)XQhXX7XM#@XMXX)XQhԀarejointlysignificant h atthe10%level.F +  16      #)XQhXX@XMҁ#@XMXX)XQhԀAsfarasthespecificationof,EXQUALSHissignificantinallthree  models,whileFEEL3andSUBSTaresignificantinmodels(2)and(3).#)XQhXX@XM#GXMXX)XQh 8   #)XQhXXGXM#@XMXX)XQhCalibratingtheresponsestotheclosedendedquestionsisslightlymoredifficultthanfor P  openendeddatabecausewedonothaveadirectobservationoftheindividual'sminimumWTA.Rather,weinferminimumWTAbyestimatingaWTAfunction.Inthiscase,thecalibratedvalueforWTA(PREDBID)iscalculatedfromtheexpressionforWTAderivedfromCobbDouglaspreferencesusingthecoefficientestimatesfromTable6.(SeeTable7formoredetails.)ThePREDBIDcalculatedwiththecoefficientsfromTable6shouldrepresenttrueWTAbecauseanysystematicbiasintheCVbids,andthusinourestimatesoftheutilityfunctionparameters,shouldbecapturedbythebiasparameter.q$ +  17       (x #)XQhXX@XM#GXMXX)XQh  Themeansofalltheindividuals'valuesforCAandPREDBIDbasedontheresultsin  Table6arecontainedinTable7.[ +  18      ׀ThefirstrowisthepredictedvalueofWTAbasedon H estimatesfromamodelthatwasidenticaltothemodelincolumns(1)(3)inTable6,exceptthatitcontainednosystematicbiascoefficient.ThisprovidesanuncalibratedestimateofWTA.ThenextthreerowscontainthecalibratedvaluesofWTAandCA,whilethelastrowcontains 0  predictedWTAfortheSMdata.Accordingtothismodel,mostoftheparticipantsintheCVsurveyoverstatedtheirWTAthemeanofthepredictedWTAfromtheSMdataof$110.94islowerthantheuncalibratedmeanWTAof$799.80.ThemeansofthePREDBID'sfromthethreeCVmodels(whichrangefrom$160to$195)arelowerthantheuncalibratedWTA,butstill  ( higherthanthemeanofthePREDBIDfromtheSMestimate.AgainthethreedifferentCVmodelsprovidethereaderwithasenseofhowPREDBIDandCAvaryunderdifferent h specifications. 4.Discussion P     CVandSMresponsesfromthreedatasets,BrookshireandCoursey'sWTPfortreesand    WTPandWTAvaluesfromBHW'sdeerhuntingpermitdata,havebeenconsideredinevaluatingthecalibrationmodel.TheBrookshireandCourseyWTPCVbidsandtheBHWWTACVbidswerebiasedupwards,whiletheBHWWTPCVbidswereunbiased.  Inthecontextoftheliteratureonrevealedpreferencemethods,bidsfromtheBHWexperimentforhuntingpermitsshouldcapturerecreationalusevalue.#)XQhXXGXMd#7XMXX)XQhInametaanalysisof287  benefitestimates,Walsh,JohnsonandMcKean(1990)#)XQhXX7XM#GXMXX)XQhԀcomparedtheresultsfromtravelcostand (x CVestimatesofrecreationalusevalue.~ +  19      ׀#)XQhXXGXMI#GXMXX)XQhAccordingtotheiranalysis,CVsurveysproducelower  valuesthantravelcostmodels,butdichotomouschoiceCVvaluesareclosertothetravelcostestimatesthanCVestimatesusinganopenendedquestionformat.#)XQhXXGXMG#7XMXX)XQhUsingdatafromstudies ` valuingavarietyofquasipublicgoods,#)XQhXX7XMt#GXMXX)XQhԀCarsonetal.(1996)#)XQhXXGXM #7XMXX)XQhexaminetheratioof#)XQhXX7XMb#7XMXX)XQhCVtorevealed  preferenceestimates#)XQhXX7XM#7XMXX)XQh.#)XQhXX7XM2#GXMXX)XQhԀTheyfoundthatacross46comparisonsbetweenCVandsimulatedmarket 0  orexperimentaldata,includingtheWTPdatafromBHW,therewasaclosecorrespondencebetweenthevaluesfromthetwomethodologies.  Fordeerhuntingpermits,ourcalibrationmodelpredictedthatneithertheCVnortheSMWTPresultswerebiased,andthisresultwasreinforcedbythefactthattheCVdatawerestatisticallyequivalenttotheSMdata.ThusourresultsconfirmthefindingsofCarsonetal.Ontheotherhand,weestimatedasubstantialupwardsbiasintheCVWTAresults.ThedifficultyinmeasuringWTAthroughCVsurveysiswellknownandneithermetaanalysisincludeddatafromWTAstudies.ThepositiveandsignificantbiastermfortheWTACVbidsconfirmstheresultsfoundinothercomparisonsofWTACVandSMdata,suchasFisher,McClellandand '(#2 Schulze(1988)(seeMansfield,VanHoutvenandHuber(1997)foradiscussionofthedifficultiesinmeasuringWTA).  RespondentstotheBrookshireandCourseystudyweredrawnfromtheneighborhoodsurroundingthepark,sotheirbidsshouldreflectbothrecreationaluseandaestheticvaluefortheextratrees.TheresultsfromthecalibrationmodelpredictthattheWTPbidswerebiasedupwards.ThestudywasnotincludedinthemetaanalysisonquasipublicgoodsperformedbyCarsonetal.,howevertheauthorsnotethat"someCVestimatesclearlyexceedtheirrevealedpreferencecounterparts,thereforeoneshouldnotconcludethatCVestimatesarealwayssmallerthanrevealedpreferenceestimates.(p.93)"ThefactthatthebidsfromthesimulatedmarketexperimentwerelowerthanthebidsfromtheCVsurveysuggestthatthecalibrationmodelcorrectlyidentifiedtheupwardsbiasinthebids.  Forthisstudy,wedeliberatelychosedatasetsthatincludedbothCVandSMcomponentsinordertoprovideabenchmarkagainstwhichtojudgethecalibratedCVresults.Twoissuesrelatedtothisdecisionshouldbeemphasized.ThefirstissueisthattheSMbidsthemselvesmaynotaccuratelymeasuretrueWTPorWTA.TheBrookshireandCourseySMexperimentswereconductedatthelocalhighschool,anditisunclearhowtheparticipantsinterpretedtheexercise.Forexample,itispossiblethattheSMbidsunderstatedWTPiftherespondentsdidnotbelievethatthemoneywasactuallygoingtobeusedtopurchaseadditionaltrees.EventheBHWSMexperimentwasprobablyconsideredunusual,especiallybytheWTArespondents,sincetheyearlylotteryforpermitshadneverbeforeincludedopportunitiestobuyorsellthepermits.ThusacomparisonbetweentheSMbidsandthecalibratedCVbidsisatestofconvergentvalidity.E +  20       8"*   ThesecondissuerelatestotheapplicationofthiscalibrationmethodtoCVbidsthatmeasureprimarilynonuseorexistencevalues.Whennonusevaluesdominate,itismoredifficulttoimaginehowonewouldcollectbenchmarkdatawithwhichtoconfirmtheaccuracyofthecalibrationresults.Onecouldeasilyestimatethecalibrationmodelandasystematicbias '(#2 coefficientusingdatafromCVsurveysmeasuringnonusevalues.Howwellwouldthecalibrationmodelperformwithnonusevalue?ItseemsreasonabletosupposethatCVdatafornonusevaluesmightbenoisierthanCVdataforusevalue,sinceingeneralindividualsmaybelessfamiliarwithgoodsforwhichnonusevaluesarethedominantsourceoftotalvalue.#)XQhXXGXMu#7XMXX)XQhFurthermore,itmaybemoredifficulttoidentifyandmeasuretheindividualcharacteristicsthat P  determinethesizeofnonusevalues.#)XQhXX7XM#GXMXX)XQhThuswhilethiscalibrationmethodcanbeappliedtoany    CVdataset,properspecificationmaybemoredifficultwhenasignificantportionofthevalueisnonuse.$  $ 5.ConclusionsandFutureResearch  X    #)XQhXXGXMѢ#@XMXX)XQhDespitethecontroversysurroundingCVsurveys,theyareoftenemployedtoestimatethe  benefitsofnonmarketenvironmentalgoods.TheresultsfromCVsurveyswillvaryinqualitydependingonthecircumstancesofthesurveyimplementation,includingtheexpertiseoftheanalystsandthebudgetforthesurvey.ThedevelopmentofacalibrationtechniqueforCVdatawouldprovideameasureofthereliabilityof$the$dataandtheabilitytoadjustbiased$$$resutls$$results$$$.$$  #)XQhXX@XM+#7XMXX)XQhThemodelproposedinthispaperprovidesthebasisforasimpleandinexpensivewayof  isolatingbiasandcalibratingtheresponsesfromaCVsurvey.#)XQhXX7XM#GXMXX)XQhԀThemethoddoesnotrequire 0  additionaldatabeyondtheCVsurveyitself,|$allowingthecalibrationof|$}$|$|$|$nonuse|$}$~$bothuseandnonusedata.~$WhetheranindividualunderoroverstateshisbidinaCVsurveyisrelatedtotheindividual'scharacteristicsandhisreactiontotheformatofthesurvey.Thusitisimportanttoexaminetheissueofcalibrationattheleveloftheindividual.Thiscalibrationmethodallowsustoseparateouttheeffectofindividualcharacteristicsonsystematicbiasfromtheeffectofthesecharacteristicsontheparametersoftheutilityfunction.  #)XQhXXGXM#7XMXX)XQhThemostchallengingfeatureoftryingtocalibrateCVdataisfindingabenchmark %X . againstwhichtojudgetheresults.TotestthiscalibrationmethodweuseddatasetsforwhichlaboratoryorSMbenchmarksexisted.#)XQhXX7XMl#7XMXX)XQhThisanalysissuggeststhatonlytheBHWWTPCVdata '(#2 producedunbiasedvalues.Incontrast,thecalibrationmodelpredictsthattheresponsesfromtheothertwoCVsurveyswetestedoverstatedtrueWTPandWTA.#)XQhXX7XM#7XMXX)XQhForthesedatasetstheresultsof *%6 thecalibrationmodelareencouragingtheresultsfromthecalibrationmodelcorroboratedthe ,`'8 generalpatternobservedfromcomparingtheCVdatawithlaboratoryorSMdata.#)XQhXX7XMǬ#GXMXX)XQh    ThepowerofthecalibrationmodelcouldbeimprovedbyabetterunderstandingofhowindividualsanswerCVquestions,includingthetraitsorattitudesthatprovokeindividualstogivemoreorlessaccurateanswersandvariablesthatmeasurethesetraitsorattitudes.Thisisespeciallyimportantforquestionsofnonusevalue.Ofcourse,suchresearchwouldalsoimprovethedesignofCVsurveys,aswell.Tofacilitatetheestimationofmoreflexiblefunctionalforms,futurestudiesmightalsowanttoincludemorevariationinthebidspaceandintheattributesoftheenvironmentalcommodity.       @'Table1dCourseydaو@ BrookshireandCourseyWTPDataSet@Meansand(StandardDeviations)*`_` dd9dd cdd .A(#(#`, dd ,P dd ,dd",dd"+  ," `  Jx,Variable  (  '  'Description ' ( 'CVSurvey󀀀  (  Ѐ     LabExperiment  (  Ѐ 6,   x J Jx6INCOME '0   'Monthly,aftertax '0  '2241.35(776.77)   p 2196.22(717.10) 6, p x J Jx6PARKVIEW '  'Dummy=1iffullorpartial   viewofpark 'x  '0.31(0.46)  x  0.52(0.51) 6,x  x J Jx6HHSIZE '0  'Householdsize '0  '3.21(1.22)    3.07(1.07) 6,  x J 3x6SCHOOL '8  'Dummy=1forchildattending 8! elementaryschoolnexttopark 'h# '0.47(0.50)  (x% 0.44(0.51) 6,(x' x 3 Jx6FINCOLLEGE '( 'Dummy=1ifgraduatedfrom ) college 'p* '0.32(0.47)  p, 0.33(0.48) 6,p. x J x6GRADSCH '(/ 'Dummy=1ifattendedsome (0 graduateschoolorreceivedadvancedprofessionalortechnicaldegree '3 '0.19(0.39)  5 0.41(0.50) 6,7 x  Jx6BID '`8 ' '`9 '17.15(19.96)  P; 8.48(10.02) 9a,P= x J 9MEDIANBID 7a*>a 7 7a*?a 710 .a!@a .5 C9!A  a axCN '!PB ' '!PC '170  !PD 27$!PE x a $  "0F $$$w$G XGXM@'Table2 w @ NonlinearLeastSquaresEstimatesofBrookshireandCourseyData@WTPModelswithBiasTerm@ParameterEstimatesandStandardErrors*de d d dd P dd dd"dd"_`(#(#,dd ,dd",dd",dd",dd ,dd"+  ," ( @x,   CES(1)   CES(2)   CES(3)       CRRA(4) 6,  x @ x6intercept  Y  Є0.85(2.08)  #  Є0.38(2.34)  # Є5.63(4.43)  # intercept  Y 0.25(1.19) 6,# x  x6hhsize  e Є0.71*(0.43)  /  Є0.86(0.54)  /  0.01(0.42)  /  hhsize  e 0.06(0.21) 6,/  x  x6gradsch  q  Є18.14*(7.55)  ; ! Є14.97*(5.48)  ; # Є14.94*(6.65)  ; % gradsch  q & 1.55*(0.92) 6,; ( x  x6school  }  ) Є2.54*(1.51)  G + Є4.29(5.13)  G - Є1.68(1.38)  G / school  }  0 1.06*(0.60) 6,G 2 x  x6parkview   3 Є1.94(2.47)  S 5 Є3.05(6.02)  S 7 0.02(0.84)  S 9 parkview   : Є0.16(0.34) 6,S < x  x6fincollege  =   >   ? 3.45(4.21)  _A fincollege  B Є1.41(2.21) 6,_D x  x6#intercept  *E Є0.29(1.27)  kG Є0.41(1.67)  kI 0.50(0.95)  kK intercept  *L 0.75(1.13) 6,kN x  x6#hhsize  6O 0.02(0.02)  wQ 0.03(0.03)  wS Є0.01(0.02)  wU hhsize  6V 0.06(0.21) 6,wX x  x6#gradsch  BY 2.42*(1.30)   [ 2.36(1.67)   ] 1.70*(0.95)   _ gradsch  B` 1.41*(0.84) 6, b x  x6#school  Nc Є0.04(0.07)  e 0.06(0.23)  g Є0.07(0.07)  i school  Nj 1.10*(0.60) 6,l x  x6#parkview  Zm 0.01(0.10)  $o 0.07(0.26)  $q Є0.06*(0.04)  $s parkview  Zt Є0.14(0.34) 6,$v x  x6#fincollege  fw 0.39(0.31)  0y 0.45(0.47)  0{ Є1.07(1.87)  0} fincollege  f~ Є1.29(1.73) 6,0 x  x6CPintercept 'r '1.34(9.27) '<" '0.81(8.37) '<" '10.48*(2.07) '<" 'CPintercept  r 10.54*(1.71) 6,< x  x6CPhhsize ' ~ '2.15(1.86) '!H " '1.63(1.73) '!H " ' ' ~" '   ~  6, ~ x  x6CPgradsch '#! '1.38(3.97) '#T"" ' '#!" ' '#!" ' '#! ' ?5#!" x  x?CPschool ' %# '2.59(4.48) '%`$" '2.75(4.43) '%`$" ' ' %#" ' ' %# ' ?5 %#" x  x?CPparkview ''% '4.04(3.76) ''l&" '3.71(3.71) ''l&" ' ''%" ' ''% ' ?5'%" x  x?CPfincollege '%)' ' '%)'" '4.77(2.43) ')x(" ' '%)'" ' '%)' ' ?5%)'" x  @x?N  1+) 170  1+) 170  1+) 170  1+)   1+) 170 6,1+) x @ x6R2 s,* (AdjR2)  =-+ 0.38(0.32)  =-+ 0.39(0.33)  =-+ 0.38(0.33)  =-+   s,* 0.38(0.34)$=-+ x   $*Significantatthe10%level.  .Z- wGXMX G@'Table3  @LLBrookshireandCourseyData@ UncalibratedandCalibratedWTPResponses  MeansandStandardDeviations*fgdddd dd"dd"dd"dd dd"de(#(#,dd",|dd",|dd"+  ," `  Jx,   ( Variable    Mean(Std.Dev.) =h,   x J Cx=UncalibratedWTPBids̀CVSurvey(N=170) 2h!` h 2BID 2h!0 h 217.15(19.95) Fa5!`  x C h cxFUncalibratedWTPBidsLabExperiment(N=27) 2a!h a 2BID 2a!h a 28.48(10.02) Nh=!x  x0 c a JxNModel(1)CalibratedWTPCESUtility(N=170) <2' 0h <CP  P 54.56(73.06) 1'@  x J Zx1 (h 0  (CALBID 2h!X!h 26.53(14.02) Nh=!H# x Z h JxNModel(2)CalibratedWTPCESUtility(N=170) <2''*h <CP  ( 164.16(144.74) 1'* x J Zx1 (h * (CALBID 2h!h+h 22.45(11.15) Nh=!X- x Z h xNModel(3)CalibratedWTPCESUtility(N=170) <2'P1,h <CP A7 2 (\$@10.48(\$@A10.48 TJ73 (\$@10.48 x  (\$@ ZxT (h $t3 (CALBID 2h!4h 210.01(17.63) Nh=!6 x Z h xNModel(4)CalibratedWTPCRRAUtility(N=170) <2'#$:.h <CP A7 !T; Gz%@10.54Gz%@A10.54 TJ7!T< Gz%@10.54 x  Gz%@ ZxT (h "< (CALBID 2h! #p=h 29.98(17.61)1'%$`? x Z h 1  %` ? XG XGXMGXMX G@@'Table4  @@ ProbitEstimatesofBHW#)XQhXXGXM#7XMXX)XQhWTAandWTP#)XQhXX7XM##)X)XQhn#)XQhX)GXMXX)XQhDataSets  @@ ParameterEstimatesandAsymptoticStandardErrors*q=> d ddd"|dd"|dd"fg(#(#q,4dd",dd",dd",dd",dd",dd",dd"+  ," p ax, '8" '(2) 8 SMWTA   ( (3) 8 CVWTA   (  (4)WTA a  (  (5) 8  SMWTP +! ( a +(6) 8 CVWTP   ( (7)WTP 6, ( x a Jx6  @  0.99(3.2E3)  0  .999(1.0E4)  0  0.99(6.4E7) a 0  0.997(2.6E3) +!0 a +0.999(2.6E4)  0  0.997(6.4E7) 6,0  x J Jx6c    96.00(44.02)   ! 454.44(183.93)   # 177.75(38.16) a  % Є18.71(37.56) +! 'a +Є34.57(86.08)   ) Є29.91(38.67) 6, + x J Jx6%  @ , 117.16(27.84)  0 . 319.81(192.07)  0 0 195.28(50.81) a 0 2 65.79(29.21) +!0 4a +89.99(70.19)  0 6 75.72(30.17) 6,0 8 x J ax6LRTestStatistic(DF)? +  a       8:   H ;   H < 25.54(3) a 8>  +!H ?a +  H @ 0.72(3) 6,8B x a Jx6#G XGXMy#95%confidencelevelGXMX G  C   D  @6 E HzG@7.82HzG@@7.82 <a-+F HzG@7.82 HzG@ < +!Ga + @6 H HzG@7.82HzG@@7.82 J@6I HzG@7.82 x J HzG@ xJN 5+ XJ  Q@70Q@570 UK)XK  Q@70 Q@  Q@68Q@U68 VL)XL  Q@68 Q@ @a@138@a@V138 [aL*XM @a@138 @a@  @Q@69@Q@[69 bX6XN  @Q@69 @Q@a  K@55K@b55 VL)XO  K@55 K@ _@124_@V124 RH0XP _@124 x _@ axRLoglikelihood B8 pQ (\"B-36.27(\"BBЄ36.27 f\-pR (\"B-36.27 (\"B 333333A-34.40333333AfЄ34.40 f\-pS 333333A-34.40 333333A \(T-83.44\(TfЄ83.44 baS-pT \(T-83.44 \(T q= ף?-31.69q= ף?bЄ31.69 si:pU q= ף?-31.69 q= ף?a Q5-21.97Q5sЄ21.97 f\-pV Q5-21.97 Q5 (\K-54.02(\KfЄ54.02H><pW (\K-54.02 x a  (\K H#G XGXM#  `W XXX@@'Table5@@::"BHWWTADataSet,VariableNames,MeansandStandardDeviations,andDescriptions*ij dd4dd"dd"dd"dd"dd"dd"dd"=>X%X%,Sdd ,9dd",dd",dd +  *aa  Ax* VARIABLENAMES )aad aa ) WTA d  SM )aa.aa ) WTA d  CV )aa.aa ) VARIABLEDESCRIPTION S >4!d  x A aa @x>Ѐ Answers:StronglyAgree=4toStrongly p  ЀDisagree=1  :   -#p  x @ @x-  FEEL1  |$  3.70(0.67)  F  3.65(0.69)  F  Therighttohuntshouldnotbeboughtorsold -#|$ x @ x-  FEEL3   0 2.00(1.06)  R  2.18(0.98)  R  DeerhuntingissoimportanttomethatIwouldbewillingtopayalotofmoneyforanopportunitytohuntdeer -#R  x  @x-  FEEL8   <  1.60(0.82)  ^   1.37(0.67)  ^   ThereareanumberofotherthingsIwouldjustassoondoasdeerhuntduringthistimeofyear 1'^   x @ x1Ѐ Answers:DefinitelyTrue=4toDefinitely H  ЀFalse=1  j !  -#H " x  x-  REACT1  T# 1.91(1.12)  v% 1.82(0.90)  v' Ithoughtyourofferwaslikeagame,whatIdidwouldn'treallymatter. -#T( x  x-  REACT5  `) 3.03(1.13)  *+ 3.00(1.17)  *- Ididn'tliketheofferbecausetherighttohuntisnotsomethingthatshouldbeboughtorsold -#*/ x  x-  REACT8  l0 3.07(1.21)  62 2.49(1.20)  64 Iwonderedwherethemoneyforthisstudycamefrombecausetaxdollarsshouldn'tbeusedforthiskindofresearch. -#66 x  x-  REACT11  x7 2.07(1.20)  B9 1.72(0.94)  B; ASandhilldeerpermitisn'treallyworththatmuch,evenifthereisachancetotakeatrophybuck,becausethereisalwaystheregularseasonthefollowingweekend -#d > x  x-  REACT13  N? 1.23(0.57)  pA 1.53(0.97)  pC IrefusedyourofferbecauseIwasafraidtheStatewouldincreasehuntingfees. -#ND x  x-  REACT14  ZE 2.37(1.24)  |$G 2.28(1.12)  |$I Thisoffermademeworryabouthuntingbecomingarichman'ssportinWisconsin 6,|$K x  Ax6 OtherVariables :  fL   fM   fN  -#fO x A @x-  HIGHSCH  P 0.44(0.50)  rR 0.41(0.50)  rT Graduatedfromhighschool -#U x @ @x-  MORESCH   !V 0.41(0.50)  !~X 0.44(0.50)  !~Z Someeducationbeyondhighschool -# ![ x @ `x-  SUBST  # \ 2.89(0.86)  #!^ 2.82(0.91)  #!` Consideringalloftheactivitiesyoucouldpotentiallydo,howmanysubstitutes # a doyouhavefordeerhunting?Inotherwords,ifyoucouldn'tgodeerhunting, #!b howmanydifferentactivitiesaretherethatyouwouldenjoydoingjustasmuch?1=many,2=some,3=onlyafew,4=none -#v%#d x ` x-  EXQUALSH  &`$e Ѐ0.44(0.79)  '*%g 0.59(0.97)  '*%i IfyouweregoingtohuntatSandhillthisyear,doyouthinkthatthequalityof &`$j deerhuntingatSandhillwouldbehigher,lower,oraboutthesameasotherplacesyouhuntdeer?0=higher,1=aboutthesame,2=lower -#L(%l x  x-  BSTCHNCE  )6'm 0.59(0.50)  X*(o 0.71(0.46)  X*(q IfeetthatSandhillismybestchancetoshootatrophybuckinWisconsin1=yes,0=no -#X*(s x  x-  #) G#@ )ACCEPT +B)t ЀOFFER   ! #) @#@ )  .-*v 0.59(0.50)  d, *x 0.21(0.41)  d, *z #) @7#@ )Answered"yes"toCVquestion#) @#@ ) -#+B){ x  x-  N#) @? # >4 p.,|  Q@70Q@>70 ^T)p.,}  Q@70 Q@  Q@68Q@^68 5+)p.,~  Q@68 Q@ 5G )Numberofobservations(p., x    (@@'Table6ProbitEstimatesofBHWWTAData@@ ParameterEstimatesandAsymptoticStandardErrors*mn d dSdd 9dd"dd"dd ijX%X%,@dd ,(dd",(dd",(dd",(dd"+  ," ( @x,   CVWTA(1)   CVWTA(2)   CVWTA(3)   SMWTA 6,  x @ x6INTERCEPT  :  1.62*(0.19)    1.58*(0.16)   1.60*(0.18)   1.61*(0.13) 6, x  x6FEEL3  F  0.03(0.03)    0.04*(0.02)    0.04*(0.02)    0.01(0.02) 6,  x  x6FEEL8  R  Є0.02(0.04)    Є0.001(0.05)    Є0.02(0.05)   " 0.03(0.04) 6, $ x  x6REACT11  ^  % Є0.02(0.03)  ( ' Є0.03(0.03)  ( ) Є0.02(0.03)  ( + Є0.03(0.02) 6,( - x  x6SUBST  j . 0.05(0.04)  4 0 0.06*(0.03)  4 2 0.06(0.04)  4 4 0.02(002) 6,4 6 x  x6EXQUALSH  v7 Є0.09*(0.03)  @9 Є0.10*(0.02)  @; Є0.10*(0.02)  @= Є0.08(0.07) 6,@? x  x6BSTCHNCE  *@ Є0.05(0.07)  LB Є0.04(0.06)  LD Є0.07(0.05)  LF 0.05(0.04) 6,LH x  x6CAINTERCEPT  6I Є44.43(172.23)  XK Є200.74(158.71)  XM 184.14*(37.80)  XO 53.33(50.40) 6,XQ x  x6CAFEEL1  BR 50.26(41.20)  d T 66.01(36.44)  d V   BW  6,BX x  x6CAREACT1  NY Є25.21(42.55)  p[   N\   N]  6,N^ x  x6CAREACT5  Z_   Z` 36.53(36.28)  |$b   Zc  6,Zd x  x6CAREACT8  fe 8.05(24.23)  0g   fh   fi  6,fj x  x6CAREACT13  rk 16.48(28.30)   <m   rn   ro  6,rp x  x6CAREACT14  !~q   !~r 11.87(28.43)  "H t   !~u  6,!~v x  x6CAHIGHSCH  #!w 40.86(82.60)  $T"y Є12.80(97.48)  $T"{   #!|  6,#!} x  x6CAMORESCH  %#~ 82.79(80.42)  &`$ 26.79(92.07)  &`$   %#  6,%# x  x6%  '% 73.11(50.31)  (l& 74.62(48.58)  (l& 116.86(45.45)  (l& 104.06(27.82) 6,(l& x  @x6Loglikelihood  *' Є19.72  *' Є18.42  *' Є24.27  *' Є26.83$*' x @  $*Significantatthe10%level.  d, * GXMX G@@'Table7 X @@SS"BHWData@@I I UncalibratedandCalibratedWTAResponsesMeansandStandardDeviations*op dd@dd (dd"(dd"(dd"(dd"mnX%X%,Sdd",9dd",dd"+  ,"  Jx,Ѐ  ( Variable +  a       p Mean(Std.Dev.) 3h"p  x J x3UncalibratedWTA(N=68) 3)' p ,,0h 3PREDBID     799.80(849.26) ' p x x' h  `   )h0 h ) Ih80  x h JxIModel(1)CalibratedWTÀCVSurvey(N=68) <2'x bh <CA   P  191.41(55.51) 1'@  x J Jx1 h 0  PREDBID )h h )160.32(174.02) Eh4! x J h JxEModel(2)CalibratedWTACVSurvey(N=68) <2' dh <CA  X! 183.19(73.70) 1'H# x J Jx1  t# PREDBID %$ %190.76(211.86) Ah4!& x J  axAModel(3)CalibratedWTACVSurvey(N=68) 6h)'P*fh 6CA Ih<(+h Gzg@184.14Gzg@I184.14 bXE(, Gzg@184.14 x a Gzg@h Jxb ! $, !PREDBID  D- 195.23(206.79) =hh,4/ x J q=PredictedWTASMSurvey(N=70) )hh2hh )PREDBID )hh3hh )110.94(110.42)1'%5 q hh 1  |5 XXXX  @%#)XQhXXGXM # IXMXX)XQhReferences   0   Andreoni,J.,"CooperationinPublicGoodsExperiments:KindnessandConfusion,"The  AmericanEconomicReview,September1995,85(4),891904.(#(# 0   Bishop,R.C.,andHeberlein,T.A.,"TheContingentValuationMethod,"inR.L.JohnsonandG.V.Johnson,eds.,EconomicValuationofNaturalResources:Issues,Theory,and P  Applications.Boulder,CO:WestviewPress,1990,p.81104.@ (#(# 0   Blackburn,M.,Harrison,G.W.,andRutstrm,E.E.,"StatisticalBiasFunctionsandInformativeHypotheticalSurveys,"AmericanJournalofAgriculturalEconomics,December1994,76 `  (5),10841088. (#(# 0   Bohm,P.,"RevealingDemandforanActualPublicGood,"JournalofPublicEconomics,1984, 0  24,13551. (#(# 0   Brookshire,D.S.andCoursey,D.L.,"MeasuringtheValueofaPublicGood:AnEmpiricalComparisonofElicitationProcedures,"AmericanEconomicReview,September1987,77,  55465. (#(# 0   Cameron,T.A.,"ANewParadigmforValuingNonMarketGoodsUsingReferendumData,"JournalofEnvironmentalEconomicsandManagement,1988,15,35579.`(#(# 0   Cameron,T.A.,"CombiningContingentValuationandTravelCostDatafortheValuationofNonmarketGoods,"LandEconomics,August1992,68,302317.0(#(# 0   Cameron,T.A.andJames,M.D.,"EfficientEstimationMethodsfor'ClosedEnded'ContingentValuationSurveys,"TheReviewofEconomicsandStatistics,1987,69,26976.P(#(# 0   Carson,R.T.,Flores,N.E.,Martin,K.M.,andWright,J.L.,"ContingentValuationandRevealedPreferenceMethodologies:ComparingtheEstimatesforQuasiPublicGoods,"Land   Economics,1996,72(1),8099.!(#(# 0   Crocker,T.D.,andShogren,J.F.,"PreferenceLearningandContingentValuationMethods,"inDietz,F.J.,vanderPloeg,F.,andvanderStraaten,J.,eds.,EnvironmentalPolicyandthe $" Economy.Amsterdam:NorthHolland,1991.% #(#(# 0   Deaton,A.,"DemandAnalysis,"inGriliches,Z.andIntriligator,M.D.,eds.,Handbookof `'"% Econometrics,Vol.3.Amsterdam:NorthHolland,1986.P(#&(#(# 0   Eom,S.Y.andSmith,V.K.,"CalibratedNonmarketValuation,"unpublishedmanuscript,NorthCarolinaStateUniversity,1994. (#(#  ,`'* Ї0   Fisher,A.,McClelland,G.H.andSchulze,W.D.,"MeasuresofWillingnesstoPayversusWillingnesstoAccept:Evidence,ExplanationsandPotentialReconciliation,"inG.L.Peterson,B.L.Driver,andR.Gregory,eds.,AmenityResourceValuation:Integrating  EconomicswithOtherDisciplines.StateCollege,PA:Venture,1988,pp.12734.p(#(# 0    (#(# 0   X   Greene,W.H.,EconometricAnalysis,2ndEdition.NewYork,NewYork:Macmillan,1993.P (#(# 0   Hanemann,W.M.,"WelfareEvaluationsinContingentValuationExperimentswithDiscreteResponses,"AmericanJournalofAgriculturalEconomics,August1984,66,33241. p (#(# 0   Hanemann,W.M.andKanninen,B.,"TheStatisticalAnalysisofDiscreteResponseCVData,"inBateman,I.J.andWillis,K.G.,eds.,ValuingEnvironmentalPreferences:Theoryand @  PracticeoftheContingentValuationMethodintheUS,ECandDevelopingCountries. 0  Oxford:OxfordUniversityPress(forthcoming). (#(# Herriges,J.A.andShogren,J.F.,"StartingPointBiasinDichotomousChoiceValuationwith  FollowUpQuestioning,"#)XQhXXIXMw'#IXMXX)XQhJournalofEnvironmentalEconomicsandManagement,    January1996,30(1),112131.0   Hoehn,J.,"ValuingtheMultidimensionalImpactsofEnvironmentalPolicy:TheoryandMethods,"AmericanJournalofAgriculturalEconomics,1991,73,28999.`(#(# 0   Hoehn,J.andLoomis,J.,"SubstitutionEffectsintheValuationofMultipleEnvironmentalPrograms,"JournalofEnvironmentalEconomicsandManagement,July1993,25(1),56 0 75. (#(# 0   Hoehn,J.andRandall,A.,"ASatisfactoryBenefitCostIndicatorfromContingentValuation,"JournalofEnvironmentalEconomicsandManagement,September1987,226247.@(#(# 0   Horowitz,J.,"ANewModelofContingentValuation,"AmericanJournalofAgricultural   Economics,December1993,75(5),126872.!(#(# 0   Judge,G.G.,Griffiths,W.E.,Hill,R.C.,Ltkepohl,H.,andLee,TsoungChoa,TheTheoryand #! PracticeofEconometrics.NewYork:JohnWileyandSons,1985.$"(#(# 0   Krosnick,J.A.,"ResponseStrategiesforCopingwiththeCognitiveDemandsofAttitudeMeasuresinSurveys,"AppliedCognitivePsychology,1991,5,213236.`'"%(#(# 0   Mansfield,C.A.,"DespairingOverDisparities:ExplainingtheDifferenceBetweenWillingnesstoPayandWillingnesstoAccept,"DukeEconomicsDepartmentWorkingPaper,1996. (#(#  ,`'* 0   Mansfield,C.A.,VanHoutven,G.,andHuber,J.,"GuiltbyAssociation:CompensationandtheBriberyEffect,"DukeEconomicsDepartmentWorkingPaper,1997. (#(# 0   McConnell,K.E.,"ModelsforReferendumData:TheStructureofDiscreteChoiceModelsforContingentValuation,"JournalofEnvironmentalEconomicsandManagement,January p 1990,18,1935. (#(# 0   Mitchell,R.andCarson,R..UsingSurveystoValuePublicGoods:TheContingentValuation @  Method.Washington,D.C.:ResourcesfortheFuture,1989.0 (#(# 0   Schulze,W.,McClelland,G.,andLazo,J.,"MethodologicalIssuesinUsingContingentValuationtoMeasureNonUseValues,"PaperpreparedforDOE/EPAWorkshop,May1920,1994,Herdon,VA. (#(# 0   Wallace,T.D.,"WeakerCriteriaandTestsforLinearRestrictionsinRegression,"Econometrica,   July1972,40(4),689698. (#(# 0   Wallace,T.D.andC.E.ToroVizcarrondo,"TablesfortheMeanSquaredErrorTestforExactLinearRestrictionsinRegression,"AmericanStatisticalAssociationJournal,Dec.1969,  16491663. (#(# 0   Walsh,R.G.,Johnson,D.M.andJ.R.McKean,"NonmarketValuesfromTwoDecadesofResearchonRecreationDemand,"inV.K.Smith,ed.,AdvancesinAppliedMicro P Economics,Volume5.JAIPressInc.,1990.@(#(# 0   Welsh,M.,"ExploringtheAccuracyoftheContingentValuationMethod:ComparisonswithSimulatedMarkets,"Ph.Dthesis,DepartmentofAgriculturalEconomics,TheUniversityofWisconsinMadison(1986). (#(# #)XQhXXIXM8#