• No results found

Green investment under time-dependent subsidy retraction risk

N/A
N/A
Protected

Academic year: 2022

Share "Green investment under time-dependent subsidy retraction risk"

Copied!
30
0
0

Laster.... (Se fulltekst nå)

Fulltekst

(1)

ContentslistsavailableatScienceDirect

Journal of Economic Dynamics & Control

journalhomepage:www.elsevier.com/locate/jedc

Green investment under time-dependent subsidy retraction risk

Verena Hagspiel

a,

, Cláudia Nunes

b

, Carlos Oliveira

c,d

, Manuel Portela

b

aDepartment of Industrial Economics and Technology Management, Norwegian University of Science and Technology, Trondheim 7491, Norway

bDepartment of Mathematics and CEMAT, Instituto Superior Té cnico, Av. Rovisco Pais, Lisboa 1049-001, Portugal

cISEG-School of Economics and Management, Universidade de Lisboa Rua do Quelhas 6, Lisboa 1200-781, Portugal

dREM-Research in Economics and Mathematics, CEMAPRE Portugal

a r t i c l e i n f o

Article history:

Received 31 October 2019 Revised 13 May 2020 Accepted 16 May 2020 Available online 28 May 2020 Keywords:

Real options Renewable energy Policy risk

Non-homogeneous poisson process Renewable energy support schemes Feed-in-tariffs

a b s t r a c t

Drivenbyambitioustargetstoreducegreenhousegasemissionsmanycountrieshavein- troducedsupport schemes to accelerateinvestments in renewableenergy. However, in recentyearsexperience showedthat,overtime,retractionofsupportschemes becomes morelikely.Thishasasevereeffectoninvestmentbehaviour.Inthispaperwestudythe effectofapotentialsubsidyretractionofafeed-intariff (FIT)oninvestmentinrenewable energycapacity,whereweexplicitlyaccountforthefactthatthelikelihoodofpolicyre- tractionmaychangeovertime.We showthatthe rangeofFITs,forwhichitisoptimal toinvestimmediately,decreasesthelongerasubsidyhasbeeninplace.IftheFIToffered istoosmalland/orthesubsidyhasbeeninstalledtoolongago,itisoptimalforthefirm towaitwithinvestmentuntil thesubsidyiseventuallyretractedand freemarketcondi- tionsprovetobeprofitableenough.Furthermore,weshowthatwhetherapolicymaker aimingatacceleratinginvestmentprefersinvestorstoconsiderretractionrisktobetime- dependent ornot,depends onhowmuch time haspassed since thesubsidy hasbeen introducedatthemomenttheinvestorconsidersinvestmentforthefirsttime.

© 2020TheAuthor(s).PublishedbyElsevierB.V.

ThisisanopenaccessarticleundertheCCBYlicense.

(http://creativecommons.org/licenses/by/4.0/)

1. Introduction

Increasingtheshareofrenewableenergy(RE)productiontotheoverallenergymixisrecognizedascriticalinreaching ambitioustargetstoreducegreenhousegasemissions(EuropeanCommission,2015).Duetothederegulationofthemajor- ityofelectricity marketsworldwide, itisprivate investorswithan objectiveofmaximizingprofitthat decidewhetherRE projectsarebuiltornot(AbadieandChamorro,2014).Manycountrieshaveintroduceddifferenttypesofsupportschemes aimed at acceleratinginvestments inRE. Governments therewith, want to ensure competitivenessof RE production and encourage investments. By 2019 nearly all countries worldwide have employed RE support policies and targets (REN21, 2019). Feed-inpricingpolicies,i.e.feed-in tariffs(FITs) andfeed-in premiums (FIPs),havebeeninstrumental inencourag-

Corresponding author.

E-mail address: verena.hagspiel@ntnu.no (V. Hagspiel).

https://doi.org/10.1016/j.jedc.2020.103936

0165-1889/© 2020 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license.

( http://creativecommons.org/licenses/by/4.0/ )

(2)

ing REprojectsworldwide, sincethey provideastableincometogenerators andhelpincrease thebankabilityofprojects (IRENAetal.,2018).By2019atleast111countrieshaveemployedsometypeoffeed-inpolicies(REN21,2019).

Policyuncertaintyintheformofsuddenrevisionsorretractionsofrenewablesupportschemes,however,hadadramatic impactoninvestmentsinREprojectsduringrecentyears.Thistypeofpolicyriskrepresentsasignificantchallengeforactors intheREsectorcurrently(Whiteetal.,2013)andhindersneededinvestments.Policyuncertaintyisfrequentlymentioned as(oneof)thekeyreasonsforthelackofinvestmentsinRE(Whiteetal.,2013,ClimateFinanceLeadershipInitiative,2019)1 OnecountrywheretheeffectofpolicyriskrecentlyhaddramaticimpactonREinvestmentsisAustralia.In2019investment inREprojectsacrossthecountry“hassloweddramatically” afterarecord-breakingtwoyears” accordingtotheCleanEnergy Council,Australia’sREassociation.Investorswerecitingnationalpolicyuncertaintyasoneofthetwokeyfactorsforpulling backfrominvestments(TheSydneyMorningHerald,2019).AnotherrecentexamplecomesfromthesolarindustryinChina, whereonJune1,2018authoritiessuddenlyandveryunexpectedly,strictlylimitednewsolarinstallationsthatwouldqualify forFITs.Analystsestimated,thatthissuddenpolicychangesetastoptoatleast20GWofsolarprojectsthatwereexpected tobebuiltinChinain2018(TheEconomist,2018).

Fromaninvestor’sperspectiveexperienceshowsthatovertime,retractionorrevisionofsupportschemesbecomemore likely.Oneofthemainreasonsforthisistechnological development.Themainmotivationforpolicymakerstograntsup- port to renewable electricityprojects inthe formof subsidiesisto ensure their competitivenesscompared totraditional powersources,likecoalorgas.Asrenewableelectricitytechnologiesmature,investmentintheseresourcesbecomesprof- itableevenwithoutgeneroussupportfromgovernments.Therefore,firmsmayexpectthatgovernmentseventually wantto terminate thesesupportschemes orrevisethem inwaysthat make themlessgenerous. Arecentexample isChina, that significantlycutthetotalsizeofitsrenewablepowersubsidiesfor2020comparedtoprioryears2.China’sgovernmentalso announcedto stopsubsidisingonshorewind capacityby 2021andisreportedlyconsidering toscaleback oreven abolish subsidiesfortheoffshorewindpowersectorin2022“asthecostsofrenewablescontinuetofallandtheindustrybecomes moreeconomicallycompetitive” (CXTech(2020)).Other reasonsforincreasinglikelihoodofpolicyretractionovertime are diminishingpublic funds,changesofgovernmentsorthefactthat REcapacitygoalshavebeenreached. Thereare several cases wherebudget constraints ledregulators to retract orsignificantly reduce provided FITs. Oneprominent exampleis Spain,wherea newgovernmentannounceda drasticretroactivelyimplementedcut ofsubsidiestoelectricity suppliersin 2013as a consequenceof a EUR26bn tariff deficit, which wasbesides the effects ofthe financial crisis alsoa result of a toogeneroussystemofsubsidiesprovidedbythe previousgovernment(FinancialTimes,2013).Similarly, Italyprovided generoussubsidiesfundedbypowerconsumersleadingtosignificantinvestmentinREfrom2005to2012buthadtocurtail theirrenewable supportschemessubstantiallyin2015 inordertorelieveconsumers,whowerestrugglingwithhighelec- tricitybillsintheaftermathoftheeconomiccrisis(TheWallStreetJournal,2014).AlsoregulatorsintheUKwereforcedto drasticallycuttoogeneroussubsidiesin2015(DepartmentofEnergy&ClimateChange,2015).

While FITs have long been considered to be the most effective scheme for accelerating development of RE sources (Coutureand Gagnon, 2010, del Rioand Mir-Artigues, 2012, Ritzenhofen andSpinler, 2016), we currently can observe a shiftfromtariff-basedinstrumentsto competitiveauctions (REN21,2019).Thismaylead investorstoperceivean increase intheriskofretractionofcurrentlyinstalledFIT schemes.Thereforeweproposetostudytheeffectofapotentialsubsidy retractiononinvestmentinREcapacityassumingthatthelikelihoodofpolicyretractioninnotconstantovertime.

WedevelopamodelinwhichaprofitmaximisinginvestorhastheoptiontoinvestinanREproject.Thecurrentsubsidy schemeprovides investorsin REprojects witha fixedFIT foreach unit produced, therewithshielding them frommarket uncertainty.However,investmentsarethreatenedbyapotentialsubsidyretraction,whichisassumedtobecomemorelikely thelongerthesubsidyschemehasbeenprovidedtoinvestors.Aftersubsidyretractionelectricityproducedmustbesoldon thespotorfuturesmarketatapricethatvariesovertime.Someoftheexamplesofsuchretroactivesubsidyrevisionshave occurredinseveralcountriesinthelast decade(REN21, 2015). ExamplesareSpain, Belgium, theCzechRepublic, Bulgaria andGreece(BoomsmaandLinnerud,2015).Inordertorelateourresultstoarealisticcaseofgreeninvestmentwepresent acasestudyconsideringinvestmentinanonshorewindprojectusingthemostrecentavailabledataforthecaseofEurope.

FITsandelectricitypricedataistakenfromGermany,wheredifferenttypesofFITsupportschemeshavebeenappliedsince 19913Finally,we present extensionsto our model,that account for differenttypes of limits forthe subsidy period.We first considerthecasethattheregulatorannouncesthesubsidytobeprovidedforalimitedtimeperiod.Second,weanalysethe case,wheresubsidyisannouncedtoberetractedincaseelectricitymarketpriceraisestoacertainlevel.

Applying arealoptions approachallows usto accountforimportantcharacteristics ofREinvestments.RE investments entaillargesunkinvestmentcosts,whichareoftenspecifictotheconsideredproject.Second,theprojectvaluedependson uncertainfutureframeworkconditions,likefluctuatingelectricitypricesorchangingsupportschemes.Third,investorshave theoptiontopostponeinvestmentsifcurrentframeworkconditionsdonotjustifyimmediateinvestment.

OurresultsshowthattherangeofFITs,forwhichitisoptimaltoinvestimmediately,decreasesthelongerasubsidyhas beeninplace.IftheFITofferedistoosmalland/orthesubsidyhasbeeninstalledfortoolong,itisoptimalforthefirmto

1Recent examples are, for instance, The Climate Finance Leadership Initiative, a group of banks and asset managers assembled by Michael Bloomberg, that released a report outlining that one of the problems of securing investment for low-carbon energy in emerging-market countries is the unpredictability of government policies ( Climate Finance Leadership Initiative, 2019 ).

2https://oilprice.com/Latest- Energy- News/World- News/China- Slashes- Renewable- Subsidies.html .

3The German Electricity Feed-in Act (1991) was, in fact, the first green electricity feed-in tariff scheme in the world.

(3)

waitwithinvestmentuntilthesubsidyhasbeenretractedandfreemarketconditionsareprofitableenough.Specifically,we derivetheoptimalelectricitymarket pricethresholdthattriggers investmentinthiscase.Wefindthat increasing market pricevolatility(drift)discourages(encourages)investment,alsoforthecasesthatinvestmentismadewhenthesubsidyis stillprovided.

Furthermore,weshowthatconsideringsubsidyretractionrisktobetime-dependentcansignificantlyaffecttheoptimal investment strategy compared to the casewhen subsidy retraction risk isconsidered constant over time. While we can confirm earlierresearch inthat policy risk discourages investment (see, e.g., Dalby etal., 2018, Boomsma andLinnerud, 2015 orFuss etal., 2008 forsimilar results), itis not straightforward to concludewhethera policy maker, who aims to accelerateinvestment,wouldprefer investorstoconsidersubsidyretraction riskto betime-dependentornot.In factthis dependsontheFITprovidedandthetimepassedsincethesubsidyhasbeenintroduced.

Finally,weconcludethat settingalimit forthe periodduringwhichthesubsidyisprovided bothinthe formoftime oraspecificmarketprice,discouragesinvestment.Thebestfromapolicymakerspointofviewwiththemainobjectiveto accelerateinvestmentistoannounceasupportschemeforanunlimitedperiodoftime.

Realoptionstheoryhasbeenincreasinglyappliedtoinvestmentproblemsintheenergysectorundermarketuncertainty andpolicy changein recentyears.Early contributions towork on policy andinvestmentfrom areal options perspective havefocused ontaxpolicyuncertainty((Dixit andPindyck,1994,Chapter9),HassettandMetcalf,1999,PawlinaandKort, 2005).Anothergroup ofpapersstudied theeffectclimatechangepolicy uncertaintyrepresentedasexogenous eventthat createsuncertaintyinthecarbon priceoninvestmentinRE sources(Yang etal.,2008, Fussetal.,2008). Aratherrecent strandofliteraturehasanalysedtheeffectofpolicyuncertaintyintheformofrandomprovision,revisionorretractionofa subsidy(see,e.g.,Boomsmaetal.,2012,BoomsmaandLinnerud,2015,RitzenhofenandSpinler,2016,EryilmazandHomans, 2016andChronopoulosetal.,2016).Similartoourcase(RitzenhofenandSpinler,2016)and(BoomsmaandLinnerud,2015) studytheeffectofapotentialfutureretroactivelyappliedsubsidyretraction.RitzenhofenandSpinler(2016)consideracase whereregulatorsmaydecidetoswitchfromafixed-priceFITtoafree-marketregime.Theyfindthatuncertaintyregarding futureregulatoryregimesdelaysorevenreducesinvestmentactivityforFIT levelsnearelectricitymarketpricesandhigh probabilitiesofanimminentregimeswitch.Similarly,BoomsmaandLinnerud(2015)concludethattheprospectsofsubsidy terminationslowsdowninvestmentsifsubsidyitisretroactivelyapplied,butspeedsupinvestmentsifitisnot.Weconfirm theirresultthatpotentialretroactivesubsidyretractionreducesinvestmentactivity.

However,all oftheaforementionedpapersanalysingpolicy uncertaintyintheformofsubsidyretractionconsiderthat thelikelihoodofsubsidyrevisionsisconstantovertime.Wecontributetothisstrandofliteraturebyallowingtheprobability ofsubsidyretractiontodependonthetimesincethesupportschemewasoriginallyintroduced.Specifically,welookatthe casethatinvestorsperceivesubsidyretractiontobecomemorelikelythelongersubsidyhasbeenprovided.

There is also a growing body of empirical work studying policy risk in the power sector. The majority of this work focuses on analyzing how policy changes and/or policy uncertainty impact investments in renewables (see, for exam- ple,Eyraud et al.(2011), LüthiandWüstenhagen (2012), Linnerudet al.(2014), Karneyeva andWüstenhagen (2017)and Sendstadetal.(2020)).Here,Sendstadetal.(2020)specifically,investigatestheeffectofretroactive policychangesonthe investmentdecisions inrenewable energyandfinds that suddenunexpected policy changes deterfurtherinvestment ac- tivityinrenewables.Weare,however,not awareofanypaperthat attemptstomeasureinvestors’perceivedpolicy riskin thepower sector nor attempts to measure policy risk forthe power sector in general. A recentstrand of literature,ini- tiated by the work of(Baker et al., 2016), attempts to measure policy uncertainty in a more general sense referring to the probability of political decisions, events, or conditions significantly affecting current and future business conditions.

Bakeretal.(2016)developedanewindexofeconomicpolicyuncertaintybasedonnewspapercoverage frequencyforthe UnitedStates. Using firm-level data they findthat policy uncertainty isassociated withgreater stockpricevolatility and reducedinvestmentandemployment inpolicy-sensitive sectorslike defense,health care,finance, andinfrastructurecon- struction. We are however, not aware ofany work usingthis methodology to measure policy uncertainty forthe power sector.

Fromamoretechnicalpointofview,policyuncertaintyisusuallymodelledintheliteraturebyahomogeneousstochastic process withdifferent states,each one representing a differentlevel of subsidy (or the retraction ofit). Technically, our contributiontothestateoftheartistoconsiderannon-homogeneousstochasticprocesstomodelpolicyuncertainty.Inthis casetheinvestmentproblemreliesontheresolutionofatime-dependentoptimalstoppingproblemwheredifferentregimes areconsidered.Fromapuremathematicalpointofview,thistypeofoptimalstoppingproblemsarestudiedbyOliveiraand Perkowski(2019).Inthistheoreticalpaper,thesystemofHamilton-Jacobi-Bellman(HJB)equationsisderived andtechnical conditionssatisfiedbyvaluefunctionsarepresented.Sinceoursetup fitsintheonepresentedintheprevious paper,we take intoaccount some ofthe technicalresults ofOliveira andPerkowski (2019). Ourpaper contributesto theliterature bypresentingthespecificsolution ofsuch astoppingproblemandderivesexplicitexpressions forthevalue functionand thresholdcurvesfortheinvestmentproblemconsidered.Thishas,tothebestofourknowledge,notbeendonebefore.

Theremainderofthispaperisorganisedasfollows.Section2introducesabenchmarkmodel,forwhichwepresentthe solutionprocedureinSection3.Comparativestaticsresultsforthebenchmarkmodelare presentedinSection 4.We then checktherobustnessofourmodelinSection5,relaxingakeyassumptionwemakeinthebenchmarkmodel.Thesolution forthe robustnessmodel ispresented in Section 5.1, andthe results are analysed inSection 5.3, basedon a casestudy introducedinSection5.2.WefinallypresentrelevantextensionsinSection6,andconcludeinSection7.

(4)

2. Model

WestudyaprofitmaximisingfirmthatconsiderstheoptiontoinvestinaREproject.Inordertoaccelerateinvestment innewREprojectsregulatorsarecurrentlyprovidingasubsidyintheformofafixedfeed-in-tariff (FIT)thatoffersasubsidy paymentof(p+F)perMWhofelectricityproducedduringthecontract.Investorsareawareofthepossibilitythat,atsome random pointin time,

ν

,regulatorsmayretract the currentsubsidyscheme,which would,asaconsequence, exposethe firm tomarketrisk. SimilartoBoomsma etal.(2012),BoomsmaandLinnerud(2015),Ritzenhofen andSpinler(2016) and Dalbyetal.(2018),weareconsideringasinglesubsidyrevision.Afterthesubsidyhasbeenretractedthefirmwillsellthe electricityatmarketpriceperMWhofelectricityproduced.

Weconsiderthatretractionofthesubsidysupportschemewilloccuratan exponentiallydistributedrandomtime,de- noted by

ν

, withintensity

λ

(t), that is assumedto be increasing withtime. Therefore, retractionof the subsidysupport

schemebecomesmorelikelyastimegoesby.Forillustrationpurposeswe assumethefollowingtime-dependentintensity

λ

(s)=

λ

0s.Then

Fν

(

u

)

=1euλ0sds=1eλ0u2/2.

Mostofthequalitativeresultsthat weshowinthepapercarryoverforotherintensityfunctions,aslongastheyarenon- decreasingint.

In what follows, the stochastic process

θ

provides usthe information regarding the state of the subsidy. Specifically,

θ

=1whilethesubsidyisstillactiveand

θ

=0,otherwise.Therefore,

θ

s=

1, s<

ν

0, s

ν

Duetothestructureof

ν

,

θ

isnot,apriori,aMarkovprocess,unlessonetakesintoaccount theinformationregardingthe momentwhen thesupport schemewasimplementedandthecurrenttime. Therefore,fromnowon,we considerthe bi- dimensionalpair(t,

θ

),wheretrepresentsthetimeelapsedsincethemomentwhenthesupportschemewasimplemented.

Thisnewprocessis,infactaMarkovprocess.

Inlightoftheinformationintroducedinthepreviousparagraph,onecanupdatethedistributionof

ν

takingintoaccount

that

ν

> t. We therefore, introduce a new random variable, represented by

ν

t. The distribution function for

ν

t can be

computedastheconditionaldistributionoftherandomvariable

ν

,asweshowinthenextequation:

P

( ν

t>t+s

)

=P

( ν

>s+t

| ν

>t

)

=ett+sλ0udu.

Therandomvariable

ν

t denotesthe(random)timeremaining untilretractionofthesubsidy,giventhatithasbeenactive

inthelasttunitsoftime.Additionally,onecannoticethattheexpectedtimeof

ν

t isequalto

E[

ν

t]=

√2

π

1−

t

λ

0

eλ0t2/2

λ

0 , (1)

wheredenotes thedistribution functionofa normaldistribution,withmean0andvariance 1.Wenote that E[

ν

t] isa

decreasingfunctionoftandverifiesE[

ν

t]→0.Thishighlightsthatthesubsidyretractionislikelytooccursoonerwhent increases.

We furthermore assume that the electricity market price, hereafter denoted by P=

{

Ps:s≥0

}

, follows the following stochasticdifferentialequation:

dPs=

μ ( θ

s

)

Psds+

σ ( θ

s

)

PsdWs, P0=p (2)

where

μ ( θ )

=

0,

θ

=1

μ

,

θ

=0,

σ ( θ )

=

0,

θ

=1

σ

,

θ

=0 (3)

andW=

{

Ws:s≥0

}

isa Brownianmotion.Byconstruction,theprocess P incorporatestheinformationregardingthe ac- tivenessofthesubsidy.Therefore,aslongastheFITisprovided,theunitpriceisfixedandequaltop.Inthisperiodthefirm earns p+F foreachunit producedby unitoftime.We interpretthisFIT asa premiumFofferedontopoftheelectricity priceatthebeginningoftheplanninghorizon,i.e.p+F.Uponretractionofthesubsidy,thefirmearnstheelectricitymar- ketprice,whichisassumedtofollowageometricBrownianmotion(asin,forexample,Fletenetal.(2007),Ritzenhofenand Spinler (2016),Boomsma andLinnerud(2015) andChronopoulos etal.(2016)), withdrift

μ

andvolatility

σ

.We further-

more,assumethattheprojectissufficientlysmallnottoaffectlong-termelectricityprices,i.e.theproducerisaprice-taker, followingamongothers(Boomsmaetal.,2012),(RitzenhofenandSpinler,2016)and(BoomsmaandLinnerud,2015).

Giventhedefinitionof

μ

(

θ

) and

σ

(

θ

), wemake thesimplifyingassumptionthat atthemomentofsubsidyretraction, themarketpriceisequaltop,independentlyofthe timeelapsedsincethebeginning oftheplanninghorizon.Inpractice however,once the subsidy isretracted,the company will sellthe electricity atthemarket price, which is not knownto thefirm beforehand(asthe marketpriceevolvesrandomlyover time).Therefore,inthat case, thepriceoftheelectricity

(5)

atthe momentoftheinvestmentwillbe arandomvariable,withadensityfunctionthat we,ingeneral, cannotcompute explicitly.Thismeansthatwecannotfindclosed-formexpressionsforthewaitingregionandthevaluefunction.Makingthis simplifyingassumptionallowsustoderiveanalyticalresultsfortheproblem,whichwillserveasanimportantbenchmark case.

InSection5werelaxthisassumption,allowingtheelectricitymarketpricetoevolveaccordingtoageometricBrownian motionfromthebeginningoftheconsideredplanninghorizonon.Inthiscase,weareonlyabletofindthewaitingregion andthevaluefunctionconditionalonthemarketpriceattheretractiontime.Then,tosolvethe(unconditionaloptimisation) problem,one needs to compute theexpectedvalue ofthe value function,whichis ahighlynon-linear function.Forthat reason,thesecomputationsare,ingeneral,onlypossibletodofromanumericalpointofview.However,wewillshowthat thequalitativeresultsofthebenchmarkmodelarerobusttothisassumption.

Furthermore,inthespecial casethat the intensityofretractionis constant,the involvedexpressions andexpectations turnouttobesimpler,andweareabletoprovideananalyticalresultforthevaluefunctionevenwhenthepriceisevolving fromthebeginning(seeSection5.1forthederivationsofthiscase).

Theprofitfunctionofthefirm,herebydenotedθ(.),isgivenby

θ

(

Ps

)

=

K

(

p+F

)

C

(

K

)

,

θ

=1

KPsC

(

K

)

,

θ

=0 (4)

whereK>0representstheannualproduction,C(.)denotestheproductioncosts,andp+F representstheFIT.Uponinvest- mentthefirmneedstopayasunkcostofI.

Theinvestmentproblemofthefirmcanthenbeformulatedasthefollowingoptimalstoppingproblem:

Vθ

(

p,t

)

=sup τ Eθ,p,t

τ e−rs

θ

(

Ps

)

dse−rτI

(

K

)

, (5)

whereweusetheindex

θ

∈{0,1}todenotethecurrentstateofthesubsidy(non-activeoractive,respectively),andristhe exogenouslygivendiscountrate.Furthermore,duetothetime-dependenceoftheintensityrate

λ

0t,atwhichthesubsidyis retracted,weneedtotaketimeintoaccountintheoptimisationproblem(5).Forthatreason,Vθ hasanexplicitdependency onthevariablet,whichweincludenotation-wise.

Fromthemathematicalpointofview,thisisanoptimalstoppingproblem,whereboththeprofitfunctionandtheprice dynamicsareaffectedbyaswitchingregime.In(OliveiraandPerkowski,2019),theauthorsconsideramoregeneralset-up, wherethey deduce thesystemof Hamilton-Jacobi-Bellman(HJB)equations that characterise thevalue function andthey proveexistenceanduniqueness ofsolutiontotheequations.Sinceinthatcasetheauthorsdealwithan exitproblem,we re-writeourproblem(5)inamoreconvenientwayasfollows

Vθ

(

p,t

)

=sup τ Eθ,p,t

0

e−rs

θ

(

Ps

)

dsτ

0

e−rs

θ

(

Ps

)

dse−rτI

(

K

)

(6)

=Eθ,p,t

0

ers

θ

(

Ps

)

ds +sup

τ Eθ,p,t

τ

0

ers

(

θ

(

Ps

))

dserτI

(

K

)

.

Onemaynoticethat thecomputations regardingthefirst expectedvaluein thesecond lineof(6)are straightforward sinceitdoesnotdependonthestoppingtime

τ

.Thereforewearelefttofindthesolutionofthesecondterminthesecond

lineof(6),whichwedenoteby

v

θ

(

p,t

)

=sup τ Eθ,p,t

τ

0

ers

(

(

Ps

))

dserτI

(

K

)

. (7)

WenextpresenttheHamilton-Jacobi-Bellman(HJB)equationsthatcharacterisevθ(p,t).Inthiscase,aswehavetworegimes (thatcorrespondtotwopossiblestatesofthesubsidy:activeorinactive)thatinfluencethedynamicsofthepriceandprofit functionofthefirm,weneedtosolvetwosetsofHJBequations:

min

(

r

v

0

(

p,t

)

L0

v

0

(

p,t

)

+

0

(

p

)

,

v

0

(

p,t

)

+I

(

K

) )

=0, (8)

min

(

r

v

1

(

p,t

)

L1

v

1

(

p,t

)

+

1

(

p

)

,

v

1

(

p,t

)

+I

(

K

) )

=0, (9)

whereLi(fori=0,1)istheinfinitesimalgeneratorofthebi-variateprocess(

θ

,P)=

{

(

θ

s,Ps),s>0

}

.Inviewoftheassump-

tionsabout

θ

andP,thefollowingholds:

L0

v

0

(

p,t

)

=

μ

p

v

0

(

p,t

)

p +

1

2

σ

2p2

2

v

0

(

p,t

)

p2 ,

L1

v

1

(

p,t

)

=

v

1

(

p,t

)

t +

λ (

t

) ( v

0

(

p,t

)

v

1

(

p,t

) )

.

Finally,wenotethatinOliveiraandPerkowski(2019)theauthorsdonotconsideraspecificprofitfunctionnorspecificprice dynamics.Therefore,allthecomputationsandtheanalysisnecessarytoobtainthevaluefunctionswillbepresentedinthe nextsections.

(6)

3. Modelsolution

Inthissectionwe presentthevalue functionofthefirm dependingonwhetherthesubsidyisstill presentornot.We firstderivethevaluefunctionforthecasethatthesubsidyhasbeenretractedalready(seeSection3.1)andthencontinue withderivingthevaluefunctionforthecasethatthesubsidyisstillactive(seeSection 3.2).Wethen presenttheoptimal investmentstrategyofthefirm.

3.1. Solutionwhenthesubsidyhasbeenretracted

Incasethesubsidyhasbeenretracted,wehaveastandardinvestmentproblem,forwhichthesolutionisalreadyknown;

see,forinstance,DixitandPindyck(1994).Moreover,we knowthatforsuchacasetheinvestmentdecisionisathreshold decision,meaningthatinvestmentisoptimalforlargevaluesofprice(pp0)while,forsmallvaluesofprice(p<p0),itis optimaltowait.Uponinvestment,thefirmsearnsaperpetualvaluegivenby:

Eθ,p,t

0

e−rs

(

KPsC

(

K

) )

ds I

(

K

)

= K r

μ

p

C

(

K

)

rI

(

K

)

.

Before investment (ie., p<p0), thefirm does not earnanyprofits.Therefore, thevalue of thefirm isequal tothe value of the optionto invest. Furthermore,asthe subsidy is not active, there isno longer the need forthe time dependency.

Therefore,wemaydroptheexplicitdependencyontofthevaluefunctionV0.

In view oftheseconsiderations, andapplicationofresults providedby Dixitand Pindyck(1994),we endup withthe resultpresentedinthefollowingproposition.Theproofsallpropositionscanbefoundintheappendix.

Proposition1. Thevaluefunctionwhenthesubsidyisnolongeravailable,i.e.V0(p),isgivenby:

V0

(

p,t

)

V0

(

p

)

=

Apd1, pp0

K

rμpC(rK)I

(

K

)

, p>p0 where

p0= d1

d1−1

(

I

(

K

)

+C

(

K

)

r

)

×r

μ

K , (10)

A= K r

μ

d11

p01d1, (11)

withd1=σ22μ+

4μ24μσ2+σ4+8rσ2

2σ2 >1.

3.2. Solutionwhenthesubsidyisstillactive

When thesubsidy isstill active,the firm’srevenueswouldbe constantupon investment.Giventhe electricityprice p atthe momentthe firm decidesto invest,the firm willreceiveby unit ofelectricity produced pwitha top-up payment ofanamountFduringthelife ofthesubsidy.Oncethesubsidywillberetracted,thefirm willbe exposedtooutputprice uncertainty.Thus,weneedtotakeintoaccountthepossiblechangesinthestateprocess

θ

.

Asdiscussedabove,inordertofindV1onehastocomputethefunctionv1,whichsatisfies r

v

1

(

p,t

)

v

1

(

p,t

)

t

λ

0t

( v

0

(

p

)

v

1

(

p,t

) )

+K

(

p+F

)

C

(

K

)

=0,

inthecontinuationregion.Hereweomitthetimedependencyofv0ont,astheproblemistime-homogeneousinthiscase (seeSection3.1).

Thesolutionofthisproblemisnotstraightforward.Wepresentthesolutioninthefollowingproposition.

Proposition2. AssumethatFissuchthatitsatisfiesthefollowingcondition F≡ −

σ

2

2 1 K

C

(

K

)

+rI

(

K

)

r d1<F< 1 K

rI

(

K

)

+C

(

K

)

1−E[erν0] ≡F.

Thenthevaluefunctionfortheoptimalstoppingproblempresentedin(6)isgivenby:

V1

(

p,t

)

=

⎧ ⎪

⎪ ⎩

V1,1

(

p,t

)

, p<c1

V1,2

(

p,t

)

,

(

p,t

)

[c1,p0

)

×[t1

(

p

)

,

)

V1,3

(

p,t

)

,

(

p,t

)

∈[c1,p0

)

×[0,t1

(

p

))

V1,4

(

p,t

)

, p>p0

where,

V1,1

(

p,t

)

=V1,2

(

p,t

)

=Apd1E

e−rνt

,

(7)

Fig. 1. Investment and continuation regions in the ( p, t )-plane.

V1,3

(

p,t

)

=V1,4

(

p,t

)

=

K

(

p+F

)

rC

(

K

)

rI

(

K

)

+

K p r

μ

K

(

p+F

)

r

E[e−rνt], andp0isgivenby(11).Finallyt1(p)isuniquelydefinedbytheequation:

Apd1E

erνt1(p)

=

K

(

p+F

)

rC

(

K

)

rI

(

K

)

+

K p r

μ

K

(

p+F

)

r

E[erνt1(p)]. (12)

Additionally,c1c1(F)verifiesthefollowingconditions:

(

i

)

c1

(

F

)

(

0,p0

)

and t1

(

c1

)

=0, if F<F<F

(

ii

)

lim

FF

c1

(

F

)

=0, and lim

FFc1

(

F

)

=p0.

TheresultspresentedinProposition2showthatwehavetodistinguishbetweenfourregionsforthevaluefunctionin casethesubsidy isstill active.Fig. 1illustrates the shapesofthe differentregions, assuming that the intensityfunction,

λ

(t), atwhich the subsidyis retrievedis non-decreasing int. Ifthe initial value ofthe process P isstrictly smaller than athresholdc1 thefirm isin thecontinuation region.We notethat thethreshold c1 isdefinedimplicitlyby theequation t1(c1)=0,asstatedinProposition2.Weindicatethiscontinuationregionby“Wait(1)” inFig.1.Forsuchlowvaluesofpit isinfactneveroptimalforthefirmtoinvest,neitherbeforenorafterthesubsidyisretracted,independentlyofhowmuch timehaspassedsincethesubsidyhasbeeninstalled.Oncethesubsidyhasbeenretractedthefirmwill investassoonas thepricehitsthelevelp0.Thevaluefunctioninthisregion,V1,1(.,.),isthereforeequaltothevalueoftheoptiontoinvest atp0,scaledbythediscountfactorE[erνt],whichaccountsfortheexpectedtimeittakesuntilthesubsidyisretracted,for agivent.

IfthecurrentvalueoftheprocessPislargeenough,specifically pp0,itisoptimalforthefirmtoinvestimmediately independenton whetherthe subsidy isstill provided ornot. We indicate thisregion by “Invest (4)” in Fig.1. Thevalue functioninthisregion,V1,4(.,.)consistsoftwo terms.The firsttermrepresentsthediscountedexpectedfuturecashflows oftheprojectincasethefirmwouldreceivetheFITforeverminustheinvestmentcost.Thesecondtermaccountsforthe additionalrevenuesearned inthe open marketassoon asthe subsidyis retracted,scaled by afactor accountingforthe expectedtimeuntilthesubsidyisretracted.

Forintermediatevaluesofpbetweenc1andp0,whetherthefirmisinthecontinuationregionorinthestoppingregion dependsonhowmuchtime haspassedsincethesubsidywasinstalled(represented byt).Here wedistinguishtwocases.

Ifp is such that [c1,p0] and sufficientlymuch time has passedsince the introduction ofthe subsidy, i.e.t>t1(p),it is neveroptimalforthefirmtoinvestbeforethesubsidyhasbeenretracted.Oncethesubsidyhasbeenretracteditisoptimal forthefirmtoinvestatthemomentthepricehitsthelevelp0.Thisissimilartothecaseinthefirstcontinuationregion (Wait(1)).Thevaluefunctioninthiscontinuationregion,indicatedby“Wait(2)” inFig.1,denotedbyV1,2(.,.),istherefore equaltothevalueoftheoptiontoinvestoncethesubsidyhasbeenretractedscaledbyadiscountfactoraccountingforthe expectedtime untilsubsidyretraction.If, however,relatively littletime haspassedsincethesubsidyhasbeenintroduced, specificallytt1(p),itisoptimalforthefirmtoinvestimmediately.Thevaluefunctionforthiscase,V1,3(.,.),isequaltothe valueintheother investmentregion.Thisinvestmentregionisindicatedby “Invest(3)” inFig.1.Notethat theboundary curveseparatingtheinvestmentandcontinuationregionsinthiscaseincreaseswiththevalueofp,aswellasthetimet,for whichthesubsidyhasbeenactivealready.Theparticularshapeoftheboundarycurve dependsontheparticularintensity

(8)

Fig. 2. Investment and continuation regions for the case that the subsidy is active (left graph) and that the subsidy has been retracted already (right graph).

rateatwhichthesubsidyisexpectedtoberetrieved.However,aslongas

λ

(.)isanon-decreasingfunction,theretractionof thesubsidybecomesmorelikelyastimepasses, andtherefore,one mayexpectthat,alsointhiscase,theboundarycurve isupwardslopinginthe(p,t)-plane4

Notethatthe investmentstrategy(giventhesubsidyisstill present)dependson howmuchtimehaspassedsincethe subsidyintroductionatthemomentthefirmconsiderstheinvestmentproblemforthefirsttime.Incasethefirmconsiders investmentatthemomentthesubsidyhasbeenintroduced,i.e.t=0,thenthehorizontallineatt=0inFig.1represents thedecisionruleasafunctionofp.Fort=0thefirmwillinvestifp>c1.Ifthefirmconsiderstheinvestmentproblemfor thefirsttimewhent>0timehasalreadypassedsincethesubsidyintroduction,thenthethresholdcurvet1(p)determines theinvestmentstrategy. Fort=t,forexample,thefirminvestsimmediatelyifp>p,asindicatedinFig.1.Otherwise,it isoptimalforthefirmtowaitwithinvestmentuntilthesubsidyhasbeenretractedandthepricehasreachedthethreshold p0.

Fig.2illustratestheinvestmentandwaitingregionforbothcases.Theleftgraphshowsthedifferentregionsforthecase thatsubsidyisactiveandtherightgraphillustratestheinvestmentandwaitingregionsforthecasethatsubsidyhasbeen alreadyretracted.Considerthepoint(tˆ,pˆ)indicatedintheleftgraphofFig.2.Ifthefirmconsidersinvestmentforthefirst timeattˆforavalueofp=pˆ,thenitisoptimalforthefirmtoinvestimmediately.Thetimeperiodittakesuntilsubsidyis eventuallyretractedthenonlyaffectstheprojectrevenuesearnedbutnottheinvestmentstrategy.Now considerthepoint (tˆˆ,p)ˆˆ insidethecontinuationregion.Ifthefirmconsidersinvestmentforthefirsttimeaftertˆˆtimeunitshavepassedsince theintroductionofthesubsidy,theoptimaldecisionofthefirmistowaitwithinvestment.Infact,thefirm willfirstwait untilthesubsidyhasbeenretracted.Thismovementintimeisindicatedbytheverticalarrowtowardstheretractiontime v.Oncethesubsidyhasbeenretractedtheoptimalinvestmentstrategy isillustratedintherightgraphofFig.2.Thefirm willthencontinuewaitinguntilthepriceprocessP(s)hitstheinvestmentthresholdp0,atwhichitisoptimaltoinvest.

ThefollowingremarkcommentsontheinvestmentstrategythatresultswhenatoosmallortoolargeFITisofferedto thefirm.

Remark1.

(a) Ifthedecision-makeroffers aFIT equalto p+F foreach unit produced,then itholdsthat c1=0.This meansthat itisoptimalforthefirmtoinvestforanyinitialpricepgiveninvestmentisconsidered att=0.As aconsequence, theoptimalstrategyillustratedinFig.1changesinthesensethatthewaitingregion(1)disappears;Moreover,when F>F, the waiting region (1) will not exist eitherand the investment region (regions (3) plus (4)) will be larger thantheoneobtainedwhenF=F asforthiscaseitholdsthatt1(p)>0forevery p∈[0,p0).Thismeans,thatitis optimalfora firm toinvestforevery pricepaslong aslessthant1(p=0)time haspassedsince thesubsidy has beenprovided.

(b) IfF=F,then both the waitingregion(2) andthe investmentregion (3)depicted inFig. 1will disappearbecause limFFc(F)=p0,whichimpliesthattheboundaryt1turnsintoaverticallinein p0.Inthiscase,thefirmdoesnot haveany additionalincentiveto makethe decisionearlier thanin thecasethat subsidy is not provided.It isalso importanttonotethatwhenF<F,theinvestmentdecisionwillbeoptimalonlyforpricesgreaterthan p0.

Remark1showsthatiftheFITofferedissufficientlysmallerthanthecurrentmarketprice,specifically,iftheFIToffered issmallerorequaltop+F (notethatF<0),thenitismoreprofitableforthefirmtoinvestinthefreeelectricitymarket

4Indeed, one may confirm that the proof of Proposition 2 in Appendix A.2 only rely on the monotony of the function λ( t ) and not on its specific expression.

(9)

wherenosubsidyisprovided.ThismeansthatpolicymakerscannotspeedupinvestmentbyofferingaFITbelowthislevel.

InthefollowingwewillfocusonanalysingthecasesthatFissetsuchthattheconditionpresentedinProposition2holds.

Inthenext proposition,we providetheexpectedtime toundertake theinvestmentdecisiongivenitisnot optimalto investimmediately.

Proposition3. Incasethecurrentpriceisp,thetimeelapsedsincetheintroductionofthesubsidyisequaltot,thesubsidy is stillactiveanditisnotoptimalforthefirmtoinvestimmediately,theexpectedtimetoundertaketheinvestmentdecision,which wedenotebyE1,t[

τ

I],isgivenby

E1,t[

τ

I]=

E[

ν

t]+ lnpp0

μ

12

σ

2

1{t>t+

1(p),μ>12σ2}, (13)

whereE[

ν

t]isgivenbyEquation(1),and1A denotestheindicatorfunctionofpropositionA5

Theexpression(15)fortheexpectedinvestmenttimestatedinProposition3shouldbeunderstoodasfollows.Ifthetime twhen thefirm considers investmentissuch thatt<t1(p), itisoptimalforthe firm toinvestimmediately. Thismeans thatinthiscasetheexpectedtimetoinvestmentisequaltozero.Incaset>t1(p),however,itisnotoptimalforthefirm toinvest immediately. Then thefirm first waits until the subsidyis retracted(which willtake E1,t[

ν

t] time, on average).

Oncethesubsidyisretractedthefirm willwait untilthemarketpriceincreasesup top0 (whichwilltake

lnpp0 μ12σ2

time, onaverage),andthentaketheinvestmentdecision.Notethattheexpectedtimeforthepricetohit p0isfiniteifandonly if

μ

>12

σ

2(seeWillmottetal.,1995).

4. Comparativestaticsresults

Inthissectionwe willfirstpresentsome analyticalresultsrelatedtothebehaviouroftheinvestmentboundaryinkey parameters.Wethencomparehowtheoptimalinvestmentstrategyandtheexpectedtimeofinvestmentdifferswhenthe firmconsiderstheretractionrisktobeconstantcomparedtotime-dependent.Inordertoallowcomparisonwealsopresent theanalyticalresultsforthecaseofconstantretractionrisk,i.e, weconsiderthecase

λ

(t)=

λ

ˆ.Notethatweassumehere that the intensityrate of retraction increaseslinearly withtime, i.e.,

λ

(t)=

λ

0t, order to derive the expressions for the relevantquantitiesexplicitly.

We note againthat for thecase that the subsidyis no longer active, theproblem isa standard one. Detailed results regardingthecomparativestaticsofthiscasecanbefoundintheliterature(see,forexample,DixitandPindyck(1994))and canbe easily derived analytically.Indeed,one can provethat p0 increaseswith

σ

anddecreases with

μ

.Thismeans the

higherthemarketpriceuncertaintythelargertheinvestmentthresholdasthevalue oftheoptiontoinvestincreases.The higherthetrend ofthepriceprocess, thesmaller theinvestmentthreshold asthevalue oftheproject increasesandthe firmismoreeagertoundergoinvestmentandtherewith,receivethe projectvalue. Inthefollowingwenow focusonthe resultsrelatedtotheinvestmentstrategywhenthesubsidyisstillactive.Proposition4presentsresultsonthebehaviorof theinvestmentthresholdcurveinkeyparameters.

Proposition4. Theinvestmentthresholdcurve,t1(p),increaseswith

μ

andFanddecreaseswith

σ

and

λ

0,i.e.

t1

(

p

)

∂μ

>0,

t1

(

p

)

∂σ

<0,

t1

(

p

)

∂λ

0 <0,

t1

(

p

)

F >0.

Theboundaryc1decreaseswith

μ

andFandincreaseswith

σ

and

λ

0,i.e.

c1

∂μ

<0,

c1

∂σ

>0,

c1

∂λ

0 >0,

c1

F <0.

TheresultspresentedinProposition4show thatkeepingallother parameters constant,increasingthevolatility (drift) decreases(increases)theoptimalinvestmentboundary,therewithdecreasing(increasing)theinvestmentregionforthecase thatthesubsidyisactive.Thisisnotsurprisingashigherelectricitypriceuncertaintyincreasesthevalue ofthefirm’sin- vestmentopportunityasitwillreceivetheelectricitypriceoncethesubsidyhasbeenretracted.Therefore,thefirmdemands ahigherFIT(i.e.largerp,aswekeepFconstantforthiscase)toundergoinvestmentimmediately.Ahigherelectricityprice drift,however,increases thevalue of theproject.Therefore, forhigherdrift thefirm has alarger incentiveto investim- mediately,whichmeansthat therangeofpricespforwhichitisoptimaltoinvestimmediatelyatthefirsttimethefirm considersinvestment,islarger.IncreasingFandtherewith,theFIToffered,hasthesameeffect.Proposition4alsoshowsthat increasingthe likelihoodofpolicy retraction,i.e.increasing

λ

0,decreases theinvestment region,andtherewith therange ofFITsforwhichthefirmis willingtoinvestimmediately.The reasonforthat isthatincreasing

λ

0 decreases theproject value,asthefirm expectsto receivetheFIT over ashorterperiodoftime. Therefore,it demandshigherlevels oftheFIT

5If μ< 12σ2, then this expected time is infinite, as follows from standard results of stochastic processes; see, for instance, Kulkarni (2016) .

(10)

tojustifyimmediateinvestmentwhensubsidyis(still)provided.Fromapolicymakersperspective thataimstoaccelerate investment,policyriskisdamagingasitreducesthefirm’swillingnesstoinvest.Ifinvestorsperceivepolicyrisktobelarger, they willbemorehesitanttoinvest.Inthat caseregulatorswouldneedtoofferahigherFIT inorderto triggerthesame amountofinvestment.

WealsonotethattheresultofProposition4showsthatincasethepricepliesin(c1,)and

μ

orFincrease,thefirm

maystilldecidetoinvestevenifmoretimeelapsedsincetheintroductionofthesubsidies.Onthecontrary,if

σ

increases,

itmaybetoolateforafirmtoinvestinthatrangeofprices.Inthatcasetheinvestmentwillonlyoccurafterthesubsidyis removedandthemarketpricereachesthevalue p0.Thismeansthatourresultsarecoherentwiththestandardrealoptions resultthatincreasingthevolatilitypostponestheinvestmentdecision.

We now analyze how the investment strategy differs when subsidy retraction risk is assumedto be time-dependent compared to the case that the firm assumes it to stay constant over time. Therefore, we provide the equivalent of Proposition 2forthecase

λ

(t)=

λ

ˆ inthe followingproposition. Thismeansthat theprobability that thesubsidy willbe retractedis assumedto beconstant witha rate

λ

ˆ.In thiscasethetime until retraction isexponentially distributed,and we aredealingwithahomogeneous process,asoppositionofthetime-dependentcase,that leadstoa non-homogeneous process.

Proposition5. Assumethat

λ

(t)=

λ

ˆ,

t,andFissuchthat

FC

(

K

)

+rI

(

K

)

Kp0<F <r+

λ

ˆ K

I

(

K

)

+C

(

K

)

r

F˜.

Then,thevaluefunctionfortheoptimalstoppingproblempresentedin(5)isgivenby:

V1

(

p

)

=

Apd1ˆλ

r+λˆ p<p

K(p+F)

r+ˆλC(rK)+r−K pμrˆλ

+λˆI

(

K

)

pp, wherepisthesmallestsolutionofthefollowingequation:

A

(

p

)

d1ˆ

λ

r

λ

=

K

(

p+F

)

r+

λ

ˆ

C

(

K

)

r + K p

r

μ λ

ˆ

r

λ

I

(

K

)

. (14)

Theexpectedtimetotakethedecisiontoinvest,incasep<p,isgivenby E1[

τ

I]=

1

λ

ˆ + lnpp

μ

12

σ

2

1{μ>1 2σ2}.

ForthecasethatsubsidyhasbeenretractedalreadytheresultsarethesameaspresentedinProposition1.

Incasetheretractionprobabilityisconsideredtostayconstantovertime,itisoptimalforthefirmtoundergoinvestment immediatelythefirsttime itconsidersinvestment,ifpislargerorequalthanthethresholdp.Otherwise,itisoptimalto waituntilthesubsidyisretractedandtheelectricitymarketpricehitsthethresholdp0(givenbyequation(11)).Theoptimal decisioninthiscaseisindependentofhowmuchtimehaspassedsincethesubsidyhasbeenintroduced.

Wenowcomparethecasesofconstantversustime-dependentretractionprobability.Notethatifthefollowingrelation betweentheintensityparameters

λ

0and

λ

ˆ holds:

λ

ˆ2

λ

0 =

2

π

,

then the expectedtime to retraction attime t=0is the sameforthe two cases,i.e. E[

ν

0]= 1ˆλ.As E[

ν

t] is a decreasing functionoft,itfollowsthat

E[

ν

t]< 1

λ

ˆ, fort>0,

which means that as time since the introduction of the subsidy passes, the expected time until retraction in the non- homogeneous case(i.e.,when

λ

(t)=

λ

0t) decreases andthedifference betweenthissituation andthehomogeneous one increases.Forthefollowinganalysiswe willassumethattheexpectedtimetoretractionattimet=0isthesameforthe twocases.

Fig. 3illustrates how theinvestment strategies ofthe casesofconstant andtime varying retraction probability differ.

Thebluesolidlineillustratestheinvestmentboundaryt1(p)forthecaseoftime-dependentretractionrisk,whiletheblue dashedlinerepresentstheboundarybetweenthewaitingandinvestmentregionfortheconstantcase.Notethatournumer- icalresultsindicatethat c1<p alwaysholds.AsdepictedinFig.3,one maydefine tt1(p),whichrepresentsthelast momentsincesubsidyintroduction,atwhichitisstilloptimaltoinvestforaFITschemepaying p+F perunitproduced ifthefirmconsidersretractionrisktobetime-dependent.

WefindthatiftttherangeofFITs,p+F,forwhichthefirmwillinvestimmediatelyislargerifitconsidersretraction probability tobetime-dependent. Ifhowever,t> t,theopposite holds.Afirmthat considersretractionprobability tobe

Referanser

RELATERTE DOKUMENTER

For applications to communicate over a computer network, they have to use a transport protocol. The dominant ones today are TCP and UDP, with TCP being preferred by the network

The impact of uncertainty and irreversibility depends on how the system evolves over time, which is determined by the intertemporal conditions: The investment decision at t; given

In order to better classify the corresponding forecasting results, we also performed VaR forecasts based on five alternative VaR models, namely the Heston, Heston-Nandi,

Additionally, we find that, although an embedded option to invest in a more efficient technology may mitigate the impact of policy uncertainty in the case of sudden subsidy

It is found how the anaerobic portion of total energy depends on time, and a sprint cross-country skiing scenario is used to study how racing time and recovery depends on

IGHT_METHODMethod used to determine whether the net was hauled in daylight or at night time, which depends on the time information available: 1 - DAY_NIGHT is based on calculated

We consider this long-term return to be dependent on sustainable development in economic, environmental and social terms, and integrate our responsible investment efforts into the

investment promotion efforts in the automotive sector in year t and at the same time simplified registration procedures for foreign investors, to the extent that the latter