未來(lái)能源研究所-木制品市場(chǎng)和森林政策對(duì)土地利用變化的影響

EffectsofWoodProducts

MarketsandForestPoliciesonLandUseChange

DavidN.Wear

WorkingPaper24-07

May2024

ResourcesfortheFuturei

AbouttheAuthor

DavidN.WearisanonresidentseniorfellowandthedirectoroftheLandUse,

Forestry,andAgricultureProgramatResourcesfortheFuture(RFF).Priortohis

arrivalatRFF,hespentmorethan30yearswithUSForestServiceResearchand

Development,mostrecentlyasaseniorresearchscientistandleaderofnationalandregionalresourceassessments.

HisworkhasbeenrecognizedwithnationalawardsfromtheUSSecretaryof

Agriculture,ChiefoftheUSForestService,andSocietyofAmericanForesters.Wear’sresearchfocusesonlinkingeconomicchoiceandbiophysicalelementsofnatural

resourcesystemstoprojectresourceconditionsandilluminatepolicyoptions.Hehaswrittenextensivelyonforesteconomics,forestcarbondynamicsatbroadscales,

timbersupply,theeffectsofriskonforestinvestments,andlanduse.Hehasledor

participatedininterdisciplinaryinitiativestounderstandtheinfluenceoflanduseandforestmanagementonavarietyofecosystemservices,includingbiologicaldiversityandwater.Asafederalscientist,heledorcontributedtoseveralnationalandregionalnaturalresourceassessmentsincludingtheSouthernForestResourceAssessment,SouthernForestFuturesProject,andNationalClimateAssessment.

Acknowledgments

ThisresearchwassponsoredbyaUSForestServiceWoodInnovationsGrant

(20DG11094200214)titled“TheImportanceofForestProductMarketsforForest

ConservationandRestoration”withResourcesfortheFuture.IbenefitedfromreviewsbyMattWibbenmeyerandJimBoyd.

EffectsofWoodProductsMarketsandForestPoliciesonLandUseChangeii

AboutRFF

ResourcesfortheFuture(RFF)isanindependent,nonprofitresearchinstitutionin

Washington,DC.Itsmissionistoimproveenvironmental,energy,andnaturalresourcedecisionsthroughimpartialeconomicresearchandpolicyengagement.RFFis

committedtobeingthemostwidelytrustedsourceofresearchinsightsandpolicysolutionsleadingtoahealthyenvironmentandathrivingeconomy.

Workingpapersareresearchmaterialscirculatedbytheirauthorsforpurposesof

informationanddiscussion.Theyhavenotnecessarilyundergoneformalpeerreview.Theviewsexpressedherearethoseoftheindividualauthorsandmaydifferfrom

thoseofotherRFFexperts,itsofficers,oritsdirectors.

AbouttheProject

IdeclarethatIhavenoknowncompetingfinancialinterestsorpersonalrelationshipsthatcouldappeartohaveinfluencedtheworkreportedinthispaper.

SharingOurWork

OurworkisavailableforsharingandadaptationunderanAttribution-NonCommercial-NoDerivatives4.0International(CCBY-NC-ND4.0)license.Youcancopyand

redistributeourmaterialinanymediumorformat;youmustgiveappropriatecredit,providealinktothelicense,andindicateifchangesweremade,andyoumaynot

applyadditionalrestrictions.Youmaydosoinanyreasonablemanner,butnotinanywaythatsuggeststhelicensorendorsesyouoryouruse.Youmaynotusethe

materialforcommercialpurposes.Ifyouremix,transform,orbuilduponthematerial,youmaynotdistributethemodifiedmaterial.Formoreinformation,visit

/licenses/by-nc-nd/4.0/.

ResourcesfortheFutureiii

Abstract

Woodproductsmarketsinfluencereturnstoforestlanduses,andpoliciestargeting

thesemarketscouldinfluencelanduseoutcomes,withimportantimplicationsfor

timberscarcity,aswellasconservationoutcomesfromwatershedprotectionto

carbonsequestration.Inthispaper,Imodelinteractionsbetweenwoodproduct

demands,timberprices,andlanduseswitchingintheUSSouthbetween1982and

2012andshowthattheamountandvolatilityoffinancialreturnstoforestsand

agricultureinfluencedbothafforestationanddeforestationdecisions.Ithensimulatetheeffectsofdemandexpansionandforestpoliciesonafforestation,deforestation,

andnetchangeinforestlandareausingcounterfactualscenariosappliedto2007–12.OnescenarioestimatestheeffectsofeconomicgrowthusingtheGreatRecessionasanaturalexperiment.Otherscenariosexaminepolicyapproachesthat(1)encourage

additionalwooduseinconstruction,(2)subsidizetreeplanting,and(3)change

agriculturalreturns.Theeconomicgrowth(recessioncounterfactual)scenarioresultsinlittlenetchangeinforestarea—increasedafforestationdrivenbyhighertimber

pricesisoffsetbydeforestationfromincreaseddemandforurbanlanduses.In

contrast,policiesthatincreasetimberdemandswithoutincreasingdemandsforurbanland(e.g.,usingmorewoodinthebuiltenvironment)aretwiceaseffectiveastree-

plantingsubsidiesatencouragingafforestationandexpandingforestlanduses.

Agriculturalreturnsmorestronglyaffectdeforestationdecisions(forest-to-agricultureswitching)thanafforestationdecisions(agriculture-to-forestswitching)andhavea

verysmallimpactonforestarea.Resultshighlighthowanypolicythatincreases

returnstoforestuses,suchasthroughcarbonoffsetmarkets,couldincreaseforestareainthisregionandothers.

EffectsofWoodProductsMarketsandForestPoliciesonLandUseChangeiv

Contents

Highlights

1

1.Introduction

2

2.Methods

3

2.1.LandUseChoice

3

2.2.WoodProductsMarkets

7

3.DataandEstimation

8

4.Results

11

5.SimulationAnalysis

12

5.1.BaseCase

13

5.2.EconomicGrowthCounterfactual

13

5.3.ExpandingWoodContentofNewConstruction

14

5.4.SubsidizedTreePlanting

14

5.5.AgriculturalPrices

14

6.Discussion

15

7.Conclusions

16

References

18

Tables

21

Figures

28

ResourcesfortheFuture

1

Highlights

.LinkedmodelsofwoodproductsmarketsandlandusechangeintheUSSouthhighlighttheinfluenceofwoodproductspricesoncomponentsoflanduse

change.

.CounterfactualsimulationindicatesthattheGreatRecessionledtobothreducedafforestationof0.728millionhectaresandreduceddeforestationofabout0.647millionhectaresbetween2007and2012.

.Policysimulationsindicatethatgrowingtimberdemandsismoreeffectivethanreducingtree-plantingcostsatexpandingoverallforestarea.

.Policiesthatexpandthewoodcontentofconstruction—thatis,growingwood

demandwithoutalsoexpandingdemandsforlandindevelopment—areespeciallyeffectiveatincreasingforestlanduses.

EffectsofWoodProductsMarketsandForestPoliciesonLandUseChange

2

1.Introduction

TherelationshipbetweenUStimberproductionandforestconditionshaschanged

substantiallysincethemid-twentiethcentury.Withthesubstantialcurtailmentof

harvestingfromfederallandintheearly1990s,old-growthharvestinginwestern

regionsshiftedrapidlytosecondgrowthandtotheSoutheast.Afforestationof

agriculturallandexpandedtheareaofactiveforestmanagementonprivatelands,

leadingtoincreasedproductionfromplantedforestsandstrongexpansioninthe

region’scarbonsink(WearandCoulston2015).Duringthisperiod,UStimberharvestsrosetopeaklevelsatabouttheturnofthecentury,whileforestareaandforest

inventoriesalsogrew(Oswaltetal.2019),suggestingthatlanduseandforest

managementdecisionsareresponsivetotimberpricesandthatpoliciesthataugmentforestreturnswouldfurtherincreasetheareaofforestsonprivateland.Thisstudy'sobjectiveistoestimatethepotentialforeconomicgrowthandpolicytoaffectprivateforestlandinresponsetoscarcitysignalsbasedonmodelsofhistoricalchanges.

Privatelandowners’anticipationofandresponsetofuturepricesisthekeyto

mitigatingfuturetimberscarcitybutisalsothemechanismforenhancingthe

provisionofseveralecosystemservices,fromwatershedprotectiontocarbon

sequestration.Modernlandconservationpolicyworkslargelythroughmarket

instrumentssuchasrentalpaymentstoinduceretirementofagriculturallands

throughUSDA’sConservationReserveProgram(Hellerstein2017)orcarbonpaymentstoinduceafforestationandothermanagementthroughvoluntaryorcomplianceoffsetmarkets(vanKootenandJohnston2016).Transferabledevelopmentprogramsrelyoneconomicallyrationaltradingoflanduseoptionstoprovideenvironmentalbenefits

(McConnellandWalls2009).Thenetcarbonbenefitsderivedfrombioenergywithcarboncaptureandstoragedependonpatternsofinducedlandusechange

(Daigneaultetal.2012;Zhaoetal.2024).Understandingthemechanismsandthe

potentialmagnitudeoflandusechangethereforeprovidesfoundationalinformationfortheeffectivedesignofconservationandclimatechangepolicies.

Inthispaper,IuseempiricalmodelsofhistoricalchangeandinsightsfromtheGreatRecessiontoexaminetheinteractionsamongwoodproductsmarkets,harvesting,andlandusechanges,aswellastheeffectsofvariousforestpolicies.ImodeldecisionsbylandownerstoconvertforeststootherusesortoestablishnewforestsintheUS

South,theregionwiththemostactiveforestinvestmentinthenation,withthelargestglobalproductionandconsumptionofwoodproducts(Wearetal.2016).Ilinkmylandusechoicemodeltoareduced-formnationaltimbermarketmodeltoestimatehow

marketfactorsaffectedlandusechoices,thensimulateoutcomesforalternative

scenarios.IlookfirstattheGreatRecessionasa“naturalexperiment”thatdisruptedwoodproductsmarketsbetween2007and2012andprovidesinsightsintohow

economicgrowthaffectsmarketsandlanduse.ThisrecessionstartedinDecember2007withthecollapseofsubprimemortgageinstrumentsandresultedin

unprecedentedcollapsesintheconstructionandsolidwoodproductssectors,with

ResourcesfortheFuture

3

substantialimpactontimberprices.Housingstartsfelltotheirlowestlevelssince

WorldWarII,andwhiletherecessionlastedsixquarters,housingstartsdidnotclimbaboveprerecessionlowsuntil2016(Prestemonetal.2018).

Isimulatehowthesectorandlandusechangeswouldhaveevolvedbetween2007

and2012absenttherecession,acounterfactualexperiment.Thisprovidesestimatesoftheimpactsoftherecessionand,moreimportantly,insightsintohoweconomic

growthandhousingdemandaffectbothafforestationstimulatedbyrisingtimber

pricesanddeforestationtosupporthousingdevelopment.Ithenusethesame

approachtomodeltheeffectsofthreepolicyscenariosthatwouldaltertimberreturns(1)ascenariothatincreasesthewoodcontentofnewconstruction,linked,for

example,toinitiativesfocusedonusingmasstimberintallcommercialandresidentialstructures,(2)ascenariothatappliestree-plantingsubsidies—amorefamiliarsupply-sidepolicytoolaimedatencouragingafforestation,and(3)ascenariothatreduces

cropreturns.Iconcludewithacomparisonoftheeffectsofthesepoliciesand

implicationsforabroadsuiteofecosystemserviceobjectives.

2.Methods

2.1.LandUseChoice

Myanalysisdevelopsamodeloflandusechoicesthatlinksforestproductspricestothechoicesthatdeterminechangesinforestuse.Theseincludetransitionsintoforestuses(afforestation)andoutofforestuses(deforestation).Becauselandusechangesarecostlytoreverseandarebasedonuncertainfuturereturns,optionvalueslikely

playanimportantroleinthechoicecalculus.Ithereforeusearealoptionsframeworkformodelinglandusechanges.Schatzki’s(2003)modelofafforestationinGeorgia

providesamethodforincorporatingrealoptionsconsiderationsinaneconometric

modelthatIextendtoallrelevantlandusechoicesusingalanduseshareformulation.Thisrealoptionsapproachextendspreviousstaticlandusechoicemodels(Lubowskietal.2008;Hardieetal.2000;Kimetal.2018)toaccountfordynamicsinawaythataccommodatesobservedvariabilityandhysteresisinlanduseoutcomes.Simplyput,considerationofrealoptionsindecisionmakingdampensshort-runintertemporal

changesbyplacingvalueondeferringthesechoicestotakeadvantageofadditionalinformation.Otherstudieshaveusedrealoptionssimulationapproachestoexplorelanduseswitchingtoanovellanduse(Songetal.2011forswitchgrass;Wearetal.

2015forgeneticallymodifiedeucalyptus)andtoexploreforestinvestmentresponsestobiophysicalandclimaterisk(Meietal.2019).

Iassumethatlandusechoicesreflecttheoutcomeofexpectedreturncomparisonsamongalternativeuses.Returnestimatesarebasedonthediscountedstreamof

expectedcostsandrevenuesforeachuse,conversioncosts,andoptionvaluesthatreflectopportunitiestoswitchinthefuture.Thestructureofcosts,returns,and

optionsaredissimilaramongusesandareinfluencedbythequalityofland.

EffectsofWoodProductsMarketsandForestPoliciesonLandUseChange

4

Agriculturalusesgenerateannualreturns,forestsgenerateperiodicandsometimesirregularreturns,andurbanusechoicesareessentiallyirreversibletransfers.(Theoptionvalueofreversiontoruralusesisconsiderednegligible.)

Iformulateabinarylandusecomparisonasfollows.Iassumeautility-maximizingrisk-neutrallandownerwhoallocateslandtotheusewiththehighestnetpresentvalueofreturns,netofconversioncostsandoptionvalues,basedonSchatzki(2003).Futurereturnsaredescribedbyastochasticdiffusionprocess,usingageometricBrownianmotion(GBM)processwithdriftμiandvarianceσi—parametersthatareinformedbyhistoricalmovementofreturns.Foragivenlandqualityq,thechoicebetweencurrentlanduseiandanalternativelandusejisbasedonavaluecomparison:

vq,i=max(Rq,i/r?μi+0vij,Rq,j/r?μj?ccij+0vji),(1)

whereRq,iistheaverageannualnetreturntolanduseinthecurrentperiod,risthediscountrate,μiisthedriftparameter,andccijisthecostofconvertingfromlanduseitolandusej.Forforests,Risanannualizedrentequivalentbasedonthepresentnetvalueofaninfiniteandpossiblyirregularstreamofcostsandreturns.Option

valuesaredefinedfortheoptionofconvertingfromthecurrentlanduseitotheotherlandusej(0vij),whichisforgonewithlanduseconversion,andtheoptionvalueofconvertingbacktolandusei(0vji),whichisobtainedwiththelanduseconversion.Optionvaluesimplythatchoicesarecostlytoreverseandsensitivetouncertainty

aboutfuturereturnstobothuses;thatis,OVisafunctionofreturnvolatility(a

functionofσiandσj)andthecorrelationbetweenreturns(pij;seeSchatzki2003fordetails).Thedenominatorofthereturnterminequation(1)defineshighervalues

whenreturnsaretrendingupward.(Notethatμimustbeboundedfromabovebyrtoavoidexplosivereturngrowth.)

Theincorporationofoptionvaluetermsdefinesthelandusechoiceasafunctionofexpectedreturns,returnvolatility,correlationofreturns,andconversioncosts.Asa

result,landusehysteresiscanarise,sobothlandusesiandjmaybeobservedwithinthesamequalityclass.Landusemayevenchangewithoutachangeinrentsbecauseallreturninformationinformstheexpectedvarianceofreturnsandthecorrelationofreturnsbetweenuses(i.e.,stablereturnsimplyareductioninuncertaintyaboutfuturereturnsovertime).

ResourcesfortheFuture

5

Myresearchquestionsfocusonhowlandusedynamicsaffecttheoverallareaof

forestwithinacounty,includingforestretention,afforestation,anddeforestation.

Usingthearea-frameinventoryoflanduseobservationsdefinedbytheNational

ResourceInventory(NRI),Isummarizeeachcounty’sobservedlandusechanges

betweensurveyyears(five-yeartimestepsbetween1982and2012)forfourbroadlandusecategories:agriculture(A),forest(F),urban(U),andother(O).The“other”classincludesfederalland,ConservationReserveProgram(CRP)land,andwaterbodies.Foreachcountyk,definealandusechangematrixas

Tk,t=

SAASFASUASOA

SAFSFFSUFSOF

SAOSFOSUOSOO

SAU

SFU

SUU

SOU

)

(2)

Tkdefinesthetransitionsharesforalllanduses,whereSijistheshareoflanduseiinperiodtthattransitionstolandusejinperiodt+1.Netlandusechangewithincountykisdefinedby

Ak,t+1=Tk,tAk,t,(3)

whereAisa1*4vectoroflandusesforeachcounty.NotethatTisindexedbytimetoreflectdifferentstates(i.e.,changesinexogenousvariablessuchasrentsthat

determinelanduseswitchingshares).

ToestimatethetransitionsharesthatconstituteTk,t,Idecomposethetransition

matrixintoasequenceofbinarychoices.Thefirststagedetermineschangesintoandoutofthe“other”category.Thismayrepresentclassificationnoiseor,inthecaseoftheCRP,anadministrativeprocess.WiththeexceptionoftheCRPbetween1982and1992,transitionsintoandoutofthe“other”categoryhavebeennegligible,andItreattheseasexogenous.

Thesecondstageislanddevelopment,ortheconversionofrurallandtourbanuses.Urbanuserentsareinfluencedbylocalconditionsthatshiftdemandsforlaborand

land.Iassumethatwhilethereturnstoanexistingruraluseinfluencewhenrurallandisdeveloped,asdescribedbyequation(1),returnstootherrurallandusesarenot

importanttothechoicetoconvertrurallandtoanurbanuse.Ialsoassumethaturbanlandusescannotbeconvertedbacktoruraluseswithinthetimeframeofthismodel,generallyconsistentwithobservedtransitionsintheNRIdata.Thisdefinesan

urbanizationsubmatrix:

Turbanization=l一一

]=l],(4)

EffectsofWoodProductsMarketsandForestPoliciesonLandUseChange

6

wherethetworowsofthematrixdefineurbanizationasbinarychoicemodelsfor

developingforestandagriculturallands,respectively:

SAU=Pr(U|A)=Pr(?0ⅤU>?0ⅤA?CCAU)=f(XAU,CCAU);SAA′=1?SAU;(5.a)

SFU=Pr(U|F)=Pr(?0ⅤU>?0ⅤF?CCFU)=f(XFU,CCFU);SFF′=1?SFU,(5.b)

whereXijisavectorofexplanatoryvariablesthatdeterminequasirentsandoptionvaluesforusesiandj,andCCijisconversioncostsbetweenusesiandj.

Thethirdstageofthedecisionstructureislanduseswitchingbetweenagriculturalandforestlanduses.Consistentwithequation(1),Iassumethatlanduseswitchingisinfluencedbyrents,conversioncosts,andthevolatilityofexpectedreturns.Theruralusedecisionsdefinethesubmatrix:

Trural=l

].

(6)

TherowsofTruraldefinetwobinarychoicemodelsforforestandagricultureland(afteradjustingsharestosumtoone):

SAF=Pr(F|A)=Pr(?0ⅤF>?0ⅤA?CCAF)=f(XAF,CCAF);SAA=1?SAF;

SFA=Pr(A|F)=Pr(?0ⅤA>?0ⅤF?CCFA)=f(XFA,CCFA);SFF=1?SFA,

(7.a)

(7.b)

whereXAF,XFAarevectorsofexplanatoryvariablesthatinfluencethedeterminationofrents,optionvalues,andconversioncostsforagriculturalandforestlanduses.Thefulltransitionmatrix,equation(2),canbeapproximatedwithequations(5a),(5b),(7a),and(7b),providingestimatesofthefirstthreecolumnsofrows1and2,andtheotherelements(Kij)determinedbyassumptions.Forhistoricalsimulation,Isetthekvaluestotheirobservedvalues.Forprojections,Idefineurbanandotherusesassinks,so

thatdiagonalelementsfortheseusesareequaltooneandoff-diagonalelementsareequaltozero.Thus

一

Tk,t=

(1?)

(1?)

KUA

KOA

(1?)

(1?)

KUF

KOF

SAU

一

SFU

KUU

KOU

),

(8)

wherehatsindicateestimatedvalues,andksareconstantssetbyassumption.

ResourcesfortheFuture

7

2.2.WoodProductsMarkets

ThelinkbetweenwoodproductsmarketsandlanduseisthroughtheXi,jvectors,

whichincludeforestrents,theirvariance,andcorrelationwithotherreturns.Forest

rentsaredefinedbylinkingtimberpricestoaforestryproductionfunctionsothat

changesinpricesaltertheright-handsidesofequations(5a),(7a),and(7b).Idevelopreduced-formtimberpriceequationstolinkdemanddeterminants(e.g.,housingstartsandGDP)tomarketclearingpricesfortimber,basedonastructuralmodelof

interactingwoodproductsandtimbersuppliesanddemands(WearandMurray2004;Prestemonetal.2018;Daigneaultetal.2016).Myreduced-formequationsderivefromastructuralmodelofwoodproductsandtimbersuppliesanddemands.Ipositan

aggregatetransformationanddualrestrictedprofitfunctionforwoodproducts:

Tr(X,Y)=0;(

9)

Rπ(Py,PNF,XF)=0,(

10)

whereXisavectorofinputs,includingtimber;Yisavectorofwoodproductoutputs;andprepresentspricevectorsforoutput(y)andnonfixedinputs(NF).XFreferstotheinputsthatarequasifixed.Thederiveddemandfortimberisdefinedby

?Rπ/?PT=TD=f(Py,PNF,XF).(11)

Thederivedsupplyforwoodproductsisdefinedas

?Rπ/?Pyi=Ys=f(Py,PNF,XF).(12)

Defineaggregatedemandforwoodproducts(Y)asafunctionofZ,avectorof

variablesreflectinggeneraleconomicactivity,constructionactivity,otherdemandshifters,andownprice:

YiD=f(Py,Z).

(13)

Basedonequations(11)and(12),areduced-formmodelforwoodproductspricesisdefinedas

Pyi=f(Px,XF,Z).(14)

Ispecifytimbersupplyasaninventoryadjustmentproblemonprivatelandsandan

administrativelydeterminedtimberharvestlevelonfederallands(TFED):

Ts=f(PT,PW,I)+TFED.(

15)

EffectsofWoodProductsMarketsandForestPoliciesonLandUseChange

8

Theprivatelandsupplyassumesthattimbersupplyisaffectedbycurrentprices

(includingotherinputprices)andfutureprices,proxiedbytheleveloftimber

inventories(I),abiophysicalscarcitymeasure.BecauseIisalteredbylanduse

outcomes,Ineedtoaccountforthisfeedbackinthesimulations.Thereduced-form

equationfortimberpricesisdefinedbyequations(11)and(15):

PT=f(PY,PW,XF,I,TFED).(

16)

Anestimatingequationfortimberpricessubstitutesequation(14)forPYiinequation(16):

PT=f(PW,XF,I,TFED,Z),(17)

withasetofexogenousvariablesontheright-handsiderelatedtothepricesofotherinputs(PW),fixedfactorsofproduction(XF),timberinventories(I),federaltimber

policy(TFED),andasetofwoodproductdemandshifters(Z).

3.DataandEstimation

Thefourbinarylandshareequations,(5a),(5b),(7a),and(7b),combinedwith

structuralconstraints,definethelandusetransitionmodel(equation(2)).Land

transitionsharesareconstructedfromplottransitionsrecordedbytheUSDA’sNRIandexpanded(usingaerialexpansionfactors)todefinetransitionsharesatthe

countylevelforsixfive-yearperiodsstartingin1982andendingin2012.Share

equations,wherethedependentvariableisboundedby[0,1],areestimatedas

fractionalresponseregressionmodels(PapkeandWoolridge1996;Woolridge2002)usinggeneralizedmethodsofmomentsestimation(implementedinRusingtheGLMprocedurewithaquasibinomialspecification;RCoreTeam2013).Usingthefractionalresponseapproachratherthanthemorecommonlog-oddstransformationobviatestheneedtomodifyextremevalues(0or1)forestimation.Toaddressthe

heteroscedasticitythatarisesfromtheboundednatureofthedependentvariable,

regressionsareweightedbythelandareaoftheinitialuseineachequation.Thefourshareequationsareformulateda