评估可再生能源政策对发展中国家脱碳的影响（英）--世界银行VIP专享VIP免费

下载本文档

阅读 628
格式 pdf
大小 1.5 MB
约72页
2023-05-01
收藏
点赞(0)
海报
举报

P R W P 10329

Assessing the Impact of Renewable Energy Policies

on Decarbonization in Developing Countries

Clara Galeazzi

Jevgenijs Steinbuks

Laura Diaz Anadón

Infrastructure Chief Economist Oce

February 2023

Public Disclosure AuthorizedPublic Disclosure AuthorizedPublic Disclosure AuthorizedPublic Disclosure Authorized

Produced by the Research Support Team

Abstract

e Policy Research Working Paper Series disseminates the ndings of work in progress to encourage the exchange of ideas about development

issues. An objective of the series is to get the ndings out quickly, even if the presentations are less than fully polished. e papers carry the

names of the authors and should be cited accordingly. e ndings, interpretations, and conclusions expressed in this paper are entirely those

of the authors. ey do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and

its aliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.

P R W P 10329

is study oers the rst consistent attempt to identify

how energy sector decarbonization policies have aected

the energy mix over the past four decades across more than

100 developing countries. It applies systematic regression

analysis to ve energy sector decarbonization outcomes

and more than 75 policy instruments aggregated into seven

policy packages. Combining instrumental variables with

country interactions and country and time xed eects in

regional panels helps address potential endogeneity issues.

Only a handful of energy policy packages signicantly aect

the decarbonization of developing countries’ energy mix,

and the packages more often achieve a negligible or oppo-

site result than intended three years after implementation.

Policies that address counterparty risk have the highest

immediate eects. Eects of renewable policies on various

decarbonization outcomes improve slightly ve and seven

years after their implementation.

is paper is a product of the Infrastructure Chief Economist Oce. It is part of a larger eort by the World Bank to

provide open access to its research and make a contribution to development policy discussions around the world. Policy

Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. e author may be contacted

at cgaleazzi@worldbank.org and jsteinbuks@worldbank.org.

ASSESSING THE IMPACT OF RENEWABLE ENERGY POLICIES ON

DECARBONIZATION IN DEVELOPING COUNTRIES

Clara Galeazzi, Jevgenijs Steinbuks and Laura Diaz Anadón1

Keywords: developing countries, energy mix, renewable energy policies, RISE index

JEL: O13, Q48

1 Galeazzi: Centre for Environment, Energy and Natural Resource Governance, Department of Land

Economy, University of Cambridge; Belfer Center, John F. Kennedy School of Government, Harvard

University. Steinbuks: Senior Economist, Office of the Chief Economist, Infrastructure VP, the World

Bank. Diaz Anadón: Centre for Environment, Energy and Natural Resource Governance, Department

of Land Economy, University of Cambridge; Belfer Center, John F. Harvard Kennedy School,

Harvard University.

The authors would like to thank Juliette Besnard, Vivien Foster, Tigran Parvanyan, and Elisa Portale

for their helpful comments and suggestions and Eric Jardim for outstanding research assistance. They

appreciate the financial support from the World Bank Energy Sector Management Assistance

Program, the Department of Land Economy at the University of Cambridge, and the Stapley Trust.

They also appreciate the World Bank Energy Sector Management Assistance Program for sharing the

RISE dataset used in this paper.

/72

下载本文档

文本预览下载提示常见问题

PolicyResearchWorkingPaper10329AssessingtheImpactofRenewableEnergyPoliciesonDecarbonizationinDevelopingCountriesClaraGaleazziJevgenijsSteinbuksLauraDiazAnadónInfrastructureChiefEconomistOfficeFebruary2023PublicDisclosureAuthorizedPublicDisclosureAuthorizedPublicDisclosureAuthorizedPublicDisclosureAuthorizedProducedbytheResearchSupportTeamAbstractThePolicyResearchWorkingPaperSeriesdisseminatesthefindingsofworkinprogresstoencouragetheexchangeofideasaboutdevelopmentissues.Anobjectiveoftheseriesistogetthefindingsoutquickly,evenifthepresentationsarelessthanfullypolished.Thepaperscarrythenamesoftheauthorsandshouldbecitedaccordingly.Thefindings,interpretations,andconclusionsexpressedinthispaperareentirelythoseoftheauthors.TheydonotnecessarilyrepresenttheviewsoftheInternationalBankforReconstructionandDevelopment/WorldBankanditsaffiliatedorganizations,orthoseoftheExecutiveDirectorsoftheWorldBankorthegovernmentstheyrepresent.PolicyResearchWorkingPaper10329Thisstudyoffersthefirstconsistentattempttoidentifyhowenergysectordecarbonizationpolicieshaveaffectedtheenergymixoverthepastfourdecadesacrossmorethan100developingcountries.Itappliessystematicregressionanalysistofiveenergysectordecarbonizationoutcomesandmorethan75policyinstrumentsaggregatedintosevenpolicypackages.Combininginstrumentalvariableswithcountryinteractionsandcountryandtimefixedeffectsinregionalpanelshelpsaddresspotentialendogeneityissues.Onlyahandfulofenergypolicypackagessignificantlyaffectthedecarbonizationofdevelopingcountries’energymix,andthepackagesmoreoftenachieveanegligibleoroppo-siteresultthanintendedthreeyearsafterimplementation.Policiesthataddresscounterpartyriskhavethehighestimmediateeffects.Effectsofrenewablepoliciesonvariousdecarbonizationoutcomesimproveslightlyfiveandsevenyearsaftertheirimplementation.ThispaperisaproductoftheInfrastructureChiefEconomistOffice.ItispartofalargereffortbytheWorldBanktoprovideopenaccesstoitsresearchandmakeacontributiontodevelopmentpolicydiscussionsaroundtheworld.PolicyResearchWorkingPapersarealsopostedontheWebathttp://www.worldbank.org/prwp.Theauthormaybecontactedatcgaleazzi@worldbank.organdjsteinbuks@worldbank.org.ASSESSINGTHEIMPACTOFRENEWABLEENERGYPOLICIESONDECARBONIZATIONINDEVELOPINGCOUNTRIESClaraGaleazzi,JevgenijsSteinbuksandLauraDiazAnadón1Keywords:developingcountries,energymix,renewableenergypolicies,RISEindexJEL:O13,Q481Galeazzi:CentreforEnvironment,EnergyandNaturalResourceGovernance,DepartmentofLandEconomy,UniversityofCambridge;BelferCenter,JohnF.KennedySchoolofGovernment,HarvardUniversity.Steinbuks:SeniorEconomist,OfficeoftheChiefEconomist,InfrastructureVP,theWorldBank.DiazAnadón:CentreforEnvironment,EnergyandNaturalResourceGovernance,DepartmentofLandEconomy,UniversityofCambridge;BelferCenter,JohnF.HarvardKennedySchool,HarvardUniversity.TheauthorswouldliketothankJulietteBesnard,VivienFoster,TigranParvanyan,andElisaPortalefortheirhelpfulcommentsandsuggestionsandEricJardimforoutstandingresearchassistance.TheyappreciatethefinancialsupportfromtheWorldBankEnergySectorManagementAssistanceProgram,theDepartmentofLandEconomyattheUniversityofCambridge,andtheStapleyTrust.TheyalsoappreciatetheWorldBankEnergySectorManagementAssistanceProgramforsharingtheRISEdatasetusedinthispaper.21INTRODUCTIONSincetheendofthelastcentury,countriesaroundtheworldhaveenactedaplethoraofpowersectorandrenewableenergypoliciesatdifferentpointsintime,indifferentcombinations,toadvanceoneorseveralenvironmental,economic,security,orequitypolicygoals.Thesepoliciesencompassinstrumentsrangingfromfeed-intariffsforrenewableenergygenerationtorenewableenergytargetsandfrombiofuelblendmandatestocarbonpricingmechanisms.Morerecently,thesepolicieswerealsomotivatedbythegrowingchallengeofmitigatingclimatechangeonaglobalscale.Weofferthefirstcomprehensiveandsystematicassessmentofhowenergypolicies,largelyrelatedtothedeploymentofrenewableenergy,shapedthedecarbonizationoftheenergymix3to7yearsaftertheirimplementationacrossmorethan100developingcountriesandfourdecades.Indoingso,ourstudyfillsanimportantknowledgegap.Asystematicreviewofpublishedresearchthatassessedtheimpactsofawiderangeofdecarbonizationpolicyinstrumentsrelatedtorenewableenergyindicatedthatmostexistingstudiesfocusonadvancedindustrializedcountries(Peñascoetal.,2021).Weconsiderfiveinter-relateddecarbonizationoutcomeindicatorsoftheenergymix,allinrelativetermsasashareofthetotal:(1)renewableenergyconsumption;(2)renewableelectricityoutput;(3)fossilfuelenergyconsumption;(4)oil,gas,andcoalenergyelectricityproduction;and(5)oilelectricityproduction.OurpolicyvariablesaregroupedintosevenenergypolicypackagesleveragingtheWorldBankRegulatoryIndicatorsforSustainableEnergy(RISE)data.2Thedatasetincludesover75policyinstrumentsrelatedtorenewableenergy,power,andfueluseimplemented2Foradetaileddescriptionofthedatabase,pleaserefertoitsofficialwebsite:https://rise.esmap.org/.3overthelastfourdecades.Thesevenenergypolicypackagesare(1)Legalframework(LF);(2)Planningforexpansion(PE);(3)Incentivesandregulatorysupport(IR);(4)Attributesoffinancialandregulatoryincentives(AI);(5)Networkconnectionanduse(NC);(6)Counterpartyrisk(CR);and(7)Carbonpricingandmonitoring(CP).1.1CONTEXTANDMOTIVATIONOurstudyiscloselyrelatedtotwolargeempiricalliteraturestreams.First,wepositionouranalysiswithinthebroadsocialscienceliteraturethatanalyzestheeffectofindividualrenewableenergypoliciesandtheiroutcomes.Peñascoetal.,(2021)conductedasystematicreviewof211studiesthatevaluate"theeffectofaspecificpolicyinstrument[relatedtothecleanenergytransition]intoaspecificoutcome"acrossallthesocialsciences.Overall,theirreviewincludesstudiesof50countries.ArelevantfindingisthattheexistingliteratureisheavilybiasedtowardtheOECDcountries.So,instead,welookonlyatnon-OECDandcovermorethan100developingcountries.The211studiesondevelopedcountriessuggestasignificantvariationacrosstheeffectsofenergypolicyinstruments.Theyhelpsupporttheevidencethatmaterialendowmentand“integratedpolicyandeconomy-wideapproaches”coupledwith“enablingconditions(governance,institutions,behavior,innovation,policy,andfinance)”arekeytodecarbonization(Bangetal.,2015;Bednar-Friedletal.,2022;Boassonetal.,2020;Lamb&Minx,2020).Second,thisstudyalsoengageswithandcontributestoalargebodyofeconomicsliteratureonthereformofenergysectorgovernance(seeFoster&Rana(2020)andJamasbetal.(2005,2015)forexcellentsurveysofthisliterature)thatbeganinthe1980s.Inthespiritofthisliterature,ourworksubjectsthesupposedbenefitsofenergysectorreformto4econometricexaminationbasedonmulti-decadepaneldatasetsofdevelopedanddevelopingcountriesusingprogramevaluationtechniques.Asmallbodyofthisresearch(Cubbin&Stern,2006;Nagayama,2009;Senetal.,2016;Urpelainenetal.,2018),likethisstudy,attemptstoaddressendogeneityissuesusingtheinstrumentalvariables(IVs)approach.NoneofthestudiesusingIVapproacheshasexaminedclimate-relatedoutcomesofpowersectorreform.However,Mallawaarachchietal.(2021)discusshowstudyingtheclimate-relatedoutcomesofpowersectorreformscanhelpcoordinatepolicyagendaswithdifferingeconomicandenvironmentalobjectives.OurmostcloselyrelatedstudyisDoumbia(2021),whichanalyzestherelationshipbetweenthedegreeofapowermarket'scompetitivenessandpowersectoroutcomesindevelopingcountries,includingrenewableenergypenetration.However,unlikeourstudy,Doumbia(2021)doesnotaddressendogeneityissues,andtheiranalysisislimitedtoconditionalcorrelations.Severalwell-knownchallengesconfoundthequantitativeassessmentofthepowersectorandrenewableenergypolicyimpacts(Bacon,2018).Theseincludeendogeneityresultingfromomittedvariablesrelatedtothecountryandregionalcharacteristicsandsimultaneityandreversecausalityofpolicyenactmentandoutcomes.Estimatesarealsopronetomeasurementerrorsresultingfromalackofaccountingforthedepthofreformandcollinearityofpolicies.Wedesignthestudytoengagewitheachofthesechallengessystematically.Weincludetimeandcountryfixedeffectstoaccountforthecountryandtemporalomittedvariablesandestimateinstrumentalvariable(IV)regressionsacrosssixregionalpanels.OuridentificationstrategyforselectingappropriateIVsassumesthatdevelopingcountriesaremorelikelytoimplementrenewableenergypolicieswhentheyexhibit"closeness"tomajordonorsthatchampiontoday'sbestpracticesinenergymarkets.TheIVs5arerelevantbecausethepolicyinstrumentsrecordedintheRISEdatabasearemeanttoreflecttoday’sbestpracticesinenergymarkets(ESMAP,2022).OurmainIVisdevelopingcountries'foreignpolicyandpoliticalproximitytothemajorWorldBankdonors(France,Germany,Japan,theUnitedKingdom,andtheUnitedStates)asestimatedinBaileyetal.’s(2017)databaseonvotingintheUnitedNationsGeneralAssembly.Inaddition,weconsidertwootherIVsthatmeasureclosenessassociatedwithtradeforrobustnesspurposes,describedindetailinSection3.3.Whileachievingperfectidentificationincross-countrypanelregressionsisdifficult,ifnotimpossible,the"closeness"measuredthroughourIVsisrelevantbecauseweshowthatfinancefromdonorsoftenfacilitatesorpromotestheimplementationofthepoliciesinthefirstplaceandthattheyareplausiblyexogenoustodevelopingcountries'energymix.Wealsoaddressthedifficultiesofcodingvariablestoconsistentlyreflectthedepthofreformandissuesrelatedtothecollinearityofmanyrenewableenergypolicies.Finally,wedesigntwoalternativestothedefaultRISEpolicyindexdevelopedbytheWorldBanktooperationalizeeachofthesevenrenewableenergypolicypackagesintovariablesforregressionanalysis.Thefirstindexaddressesthedepthofthereformproblembecauseitweighsallpolicieswithineachpolicyreformtypeequally.ComparedtotheweightingbytheWorldBank,itiscomparativelymoresensitivetothenumberofpoliciesimplemented.Thesecondindexaddressesthecollinearityofpolicieswithineachpolicyreformtypebyweightinguncorrelatedpoliciesmorethanthosehighlycorrelatedtoothers.Theresultingcombinationsofindicesandmodelscreate18"base"regressionspecifications.Weestimateeachoftheseregressionspecificationsoverfiveindicatorsoftheenergymix,sevenpolicyreformtypes,andsixgeographicalregions,leadingto3,780regressionsintotal.6Additionally,following(Wooldridge,2001),weapplytwo-stageleastsquares(2SLS)countryinteractions,renderingtensofthousandsoffirst-stagecoefficients.WhenwerestrictthesampletotheIVregressionsthatplausiblymeettherelevancecriterionandaretheoreticallyconsistent,weobtain540first-stageregressionsthatconstitutethebasisforouranalysis.Overall,thispaperfillsagapbyfocusingonabroadsetofenergypolicypackages,awiderangeofdevelopingcountriesoveralongperiod,andimportantdecarbonizationoutcomesusingquasi-experimentaleconometrictechniques.Doingthisprovidesnewinsightsregardingtheextenttowhichenergypolicypackageslinkedtoemissionsreductions(andrenewabledeployment)inindustrializedcountriesarealsoassociatedwithsimilarresultsindevelopingcountriesovertheshort-andmedium-term(Bednar-Friedletal.,2022).1.2CONTRIBUTIONSThemainresultsofouranalysissuggestthatcontrollingfortime,country,andregionaldifferences,theeffectsofmostrenewableenergypolicypackagesontheenergymixindevelopingcountriesthreeyearsafterpassingthepoliciesarelargelyinsignificant.Only15.7percentoftheestimatedsecond-stageregressionsmeetthestatisticalsignificancethresholdofap-valuebelow10percent.Moreover,mostoftheestimatedstatisticallysignificantsecond-stageregressioncoefficientsarenegativeornegligible.Thatis,renewableenergypoliciescounterintuitivelyresultinthesameorahighershareoffossilfuelsourcesinthedevelopingcountriesenergymix.However,theperformanceofthesepoliciesgenerallyimproved–becomingnegligibleorslightlypositive-toachievetheirgoalsfiveand/orsevenyearsafterward.WeinterpretanddiscusstheseresultsinthecontextoftheSailingShipEffect(Gilfillan,1935;Ward,71967),whereinincumbentfossilfueltechnologiesdampentheshort-termeffectsofrenewableenergypolicies.Ourresultsmaydifferwhenconsideringtheimpactonmodern(non-hydro)renewablesinsteadofallrenewables.Toaddressthisconcern,weconductrobustnesschecksexcludinghydropowerandfindsimilarresults.Thelimitedimpactofthesepoliciescouldbedrivenbyahostofinterrelatedfactorsleadingtodifficultiesinsecuringfinanceinthesecountries(Eglietal.,2019;Moner-Gironaetal.,2021).Thisexplanationregardingdifficultiessecuringfinanceisinlinewiththeresultsinthispaper,indicatingthattheenergypolicypackagethataddressescounterpartyriskistheonethatismoreconsistentlyassociatedwithincreasesintherenewablesindevelopingcountries’energymixthreeyearsafterimplementation.Thisfindingoftheimportanceofaddressingcounterpartyrisksupportstherationaleforpoliciesthatmakeprojectsbankableforprivateinvestors,includinggovernmentguaranteesforelectricityauctions.Injuxtapositiontootherpolicies,theeffectsofthecounterpartyriskpackagetendtomoderateovertime,perhapsbecausetheyaddressmajorfinancinghurdlesintheshorterterm,makingroomforotherpoliciestohavepositiveeffectsovertime.Therefore,thefindingsofthispapercontributefurtherevidencetothenotionthatsignificantlyincreasingclimatefinancefordevelopingcountriesisanessentialelementcomplementingdomesticrenewableenergypolicies.Merelyrequestingadditionaldomesticdecarbonizationpoliciesindevelopingcountrieswillunlikelyyieldsignificantandtimelychangesintheenergymix,andthedevelopedworldshouldatleastfulfillitsclimatefinancecommitmentsundertheParisAgreement.Incombinationwiththerelativedearthofresearchontheimpactofdecarbonizationpoliciesindevelopingcountries,thispaper'sfindingsshowthatcountriesneedtoadopta8sustainedandcrediblepolicyeffort.Suchaneffortshouldholisticallyconsidertheenergysector,domesticinstitutionsandcapabilities,theinnovationsystem,socioeconomicimpacts,andbroaderSDGswhileadoptinganexperimentalattitudeandimplementingdatacollection,learningandadaptationmechanisms,andinternationalknowledgesharing.2EMPIRICALSPECIFICATIONWeexamineacontinuumoflinearrelationshipsbetweenarenewableenergypolicypackagexandanenergymixoutcome,yinacountrycandayeart(l).Asenergysectorpoliciestaketimetoimplement,weassumethatpolicyimplementationdoesnothaveanimmediateeffectontheoutcomevariableandconsiderthree-year,five-year,andseven-yearlags,asiscommoninpolicyevaluationliterature(Choi&Anadón,2014;Doblingeretal.,2019).Eachregressionequationcanbesummarizedasfollows:𝑦𝑐,𝑡=𝛼𝑐+𝛽𝑥𝑐,𝑡−𝑙+𝛾𝑡+𝜀𝑐,𝑡,(1)where𝛽isthecoefficientofinterest,𝛼and𝛾arethecountry-andtime-fixedeffects,and𝜀istheunobservederrorterm.Thetimefixed-effectestimateoftheeconometricmodel(1)willbebiasedandinconsistentifanunobservedtime-varyingcountryorregionalcharacteristics,suchaseconomicgrowthandothersocio-economicvariables,playapartinpolicyoutcomes,leadingtoomittedvariablesbias.Tomitigatetheseproblems,weexplicitlycontrolforregionalcharacteristicsbyestimatingcountryfixed-effectregressionmodelsseparatelywithintheWorldBankregionalgroups(AppendixTableA.1).Otherendogeneitybiasescouldresultfromthesimultaneityandreversecausalityofpolicyenactmentandoutcomes.Weattempttocorrectendogeneitybiasusingtheinstrumentalvariables(IV)approach.Intheory,IVssuccessfullyaddressthebiasesbyisolatingthe9exogenousportionoftherelationshipbetweentheindependentvariableofinterest,x,andthedependentvariable,y.Inourcontext,theIVapproachimpliesfindingavariable,z,whichaffectstheenactmentoftherenewableenergypolicy,butnottheenergymix,exceptthroughthispolicy.ThemainchallengeisfindingasuitableIV,atopicwediscussinthenextsection.Followingtheliterature(Stock&Watson,2011;Wooldridge,2001),weestimatethemodel(1)inthefollowingtwostages:𝑥𝑐,𝑡=𝛼𝑐+𝛽𝑧𝑐,𝑡+∑𝛿𝑐𝑧𝑐,𝑡𝐷𝑐𝑛𝑐=1+𝛾𝑡+𝑢𝑐,𝑡(2)𝑦𝑐,𝑡=𝜃𝑐+∑𝜗𝑐𝑥𝑐,𝑡−𝑙̂𝑛𝑐=1𝐷𝑐+∑𝜌𝑥𝑐,𝑡−𝑙̂𝑛𝑐=1+𝜇𝑡+𝑐,𝑡,(3)whereDisacountrydummyvariable,𝑥̂istheinstrumentedpolicyvariable,𝛼,𝜃,𝛾,and𝜇arethecountry-andtime-fixedeffects,and𝑢andvaretheunobservederrorterm.Thekeycoefficientsofinterestarethesecond-stage(S2)estimatesofpolicyvariablesinteractedbycountryfixedeffects,𝜗𝑐.Werestrictouranalysistothesub-setofthesecond-stageestimatesthatarelikelytosatisfytheIVrelevancecondition(i.e.,zmustbestronglycorrelatedwithx)andtheexclusionrestriction(i.e.,zonlyaffectsythroughitsimpactonx).AssessingtheIVrelevancecriterionisstraightforwardbycheckingtheF-statisticofthefirst-stageregression(2).Asthereisnovalidstatisticaltestfortheexclusionrestriction,wekeepthefirst-stageestimatesthatarestatisticallysignificantandhavetheoreticallyconsistentsigns.WediscussthestrengthofourIVsinSections4and5.103DATAANDCONSTRUCTIONOFVARIABLES3.1DEPENDENTVARIABLESWeusetheWorldBankWorldDevelopmentIndicators(WDI)asadatasourceforourdependentvariablesspanningthelastfourdecadesandmorethanahundreddevelopingcountries.AppendixTableA.1providesalistofallthecountriesincludedinthisstudyandtheirregionalclassification.Forrobustnesspurposes,weconsiderfiverelevantdependentvariables(Table1).WeexpectrenewableenergypolicypackagestonegativelyaffectthefirstthreeenergymixmeasuresinTable1.Incontrast,weexpectrenewableenergypolicypackagestopositivelyaffecttheremainingtwoenergymixmeasuresinTable1.Tostandardizeourestimationresultsacrossdifferentspecifications,wemultiplytheestimatedcoefficientsofinterestforthefirstthreeenergymixmeasuresbyminusonetoindicatethatapositivecoefficientleadstoimpactsalignedwithdecarbonization.Conversely,anegativecoefficientdenotesimpactsthatarenotalignedwithdecarbonization.AppendixFiguresA.1andA.2showthetimesseriesofeachdependentvariableovertime,aggregatedoverregionstopreservespace.Temporalpatternsindicatecleardifferencesacrossregionsinalloutcomevariables.Forexample,theshareoffossilfuelsintheenergymixisconsistentlyhigherinrelativelyoil-dominatedregions(e.g.,theMiddleEastandNorthAfrica)thanhydro-dominatedregions(LatinAmerica).Thesepatternsreinforceourrationaletoestimatetheregressionsseparatelyacrossregions.11Table1:DependentVariablesVariableAcronymUnitExpectedImpactofRenewableEnergyPoliciesFossilfuelenergy(oil,gas&coal)consumptionFCCPercentoftotalNegativeElectricityproductionfromfossilfuel(oil,gas&coal)sourcesEFFPercentoftotalelectricityoutputNegativeElectricityproductionfromoilsourcesEOSPercentoftotalelectricityoutputNegativeRenewableenergyconsumptionRECPercentoftotalfinalenergyconsumptionPositiveRenewableelectricityoutputREOPercentoftotalelectricityoutputPositiveSource:WorldBankWorldDevelopmentIndicators(WDI)database,2020.3.2EXPLANATORYVARIABLESOurexplanatoryvariablescomefromthe2018disaggregatedpolicyinstrumentdatabehindtheRegulatoryIndicatorsforSustainableEnergy(RISE)renewableenergy"trafficlight"indicatorsthattheEnergySectorManagementAssistanceProgram(ESMAP)attheWorldBankhaspublishedsince2010.Concretely,ourindependentvariablesareindexesthatrepresentsevenenergypolicypackages.EachenergypolicypackagecomprisesthedisaggregatedpolicyinstrumentsintheRISEdata.Inthissection,wedescribetheprimaryRISEdatasetandthethreealternativemethodswedesignedtoreducethedimensionalityofthedatasetandcreateourindependentvariablesindexes.IntheprimaryRISEdataset,eachpolicyiscodedintwovariables:oneindicatingwhetherthepolicyexiststhroughyes/noanswer,andthesecondspecifyingtheyearofthefirstinstanceofthepolicy,ifapplicable.3Whenrestructuringthedataset,wecombinethetwovariablesforeachpolicyinstrumentintoone.Itsvaluechangedfrom0to1whenthefirstpolicywasputintoplace,creatingthepaneldatasetneededfortherestoftheanalysis.3Forinstance,"Doesalegalframeworkforrenewableenergydevelopmentexist?"andtheyearforthefirstlegalframework.12Weaddressatleasttwofurtherchallengeswiththestructureoftheprimarydatasetforcreatingourindependentvariables.First,wefindseveraldiscrepanciesbetweenthesamepolicy'syes/noandyearsvariables.AppendixTableA.2describestheeighttypesofdiscrepanciesinthedatasetandoursolutiontoaddresseachtypeofdiscrepancy.Second,thereareoccasionalcontinuousvariablesinthedatasetandvariablesthatwecannottransformintopaneldata.Theseoccasionalcontinuousvariablesdisruptoureffortstoaddresscollinearityandreducethedimensionalityofthedataset,asdiscussedinthefollowingsection.PleaserefertoAppendixTableA.3formoredetails.Thefinaldatasethas76policiesoverapanelof133countriesbetween1980and2018,asmostpolicieswerenotimplementedbeforethen(seeAppendixTableA.4).ReducingdimensionalityWemustreducethehighlydimensionalpolicyinstrumentdataappropriatelyforregressionanalysisandinterpretation.ESMAPindicatorsinclude76renewableenergypoliciesgroupedintosevenhigh-levelenergypolicypackages,or“Headings”(Table2).Eachenergypolicypackagecontainsadifferentnumberofpoliciesandsometimescontainsclusters(orgroups)ofpolicies.ThedendrograminAppendixBshowsthatsimilarpolicyinstrumentsusuallyresidewithinthesameHeading.We,therefore,usetheseHeadings,orenergypolicypackages,asourindependentvariables.Table2:HeadingsintheRISERenewableEnergyPillarHeading/EnergypolicypackageAcronyms1LegalframeworkforrenewableenergyLF2PlanningforrenewableenergyexpansionPE3IncentivesandregulatorysupportforrenewableenergyIR4AttributesoffinancialandregulatoryincentivesAI5NetworkconnectionanduseNC6CounterpartyriskCR7CarbonpricingandmonitoringCPSource:RISEwebsite,https://rise.worldbank.org.13Itisnecessarytodesignindexesthatrepresenttheenergypolicypackagessothattheoreticalandstatisticalcomparisonsbetweenthemmakeeconomicsenseandreflectcharacteristicsliketheextentofrenewableenergypolicyreformortheextenttowhichpolicyinstrumentsareadoptedtogether.Weusethreedifferentmethodstotallypoliciesor"operationalize"eachenergypolicypackage,potentiallyaffectingtheregressionresults.Thefirstalternativeisthe“RISEindex,”basedonthepracticesESMAPusestocreatetheaforementionedaggregatetrafficlightindicatorsininstitutionalpublications(Fosteretal.,2018).TheRISEindexweighspolicygroupsequally,butasshowninAppendixA.3,groupsandsub-nestedgroupscancontaindifferentquantitiesofpolicies.Therefore,intheRISEindex,theweightgiventoeachpolicyisaffectedbyhowmanypoliciesareinthegroupthispolicybelongsto.Forinstance,whentherearefourpoliciesinagroup,eachisworth25percent,andwhentherearetwopolicies,eachisworth50percent.Thearbitraryaspectofweighingthepolicieswithingroupsispartofourmotivationtoseekalternativemethodstocreateotherindicesthatwecanuseasindependentvariables.Analternativeapproachistocreateanindexthatsumsuptheenactedpoliciesateachtime.Suchanindex,whichwewillcallthe"Summationindex,"hasbeentriedbefore(Cubbin&Stern,2006)andisthesimplestalternativetotheRISEindexforeachofthesevenenergypolicypackagesshowninTable2.WhenappliedattheHeadinglevel,theSummationindexweighsallpoliciesequallyandproducesamanageablenumberofindependentvariables.Arguably,becauseallpoliciesareweightedequallywithinHeadings,itisalsoaproxyforthedepthofreformacrosscountries.WeusetheSummationindexasoursecondoption.Weproposeathird,"CompositeIndex,"basedoncorrelationanalysisforrobustness.TheCompositeindexfirstreducesdimensionalitybydroppinghighlycorrelatedvariablessothathighlycollinearpoliciesarenotcountedseveraltimesover.Itthensumstheremaining14variablesbyHeading.WedescribedetailsofconstructingtheCompositeindexinAppendixC.AppendixTableA.5comparestheweightsbetweenindexes.Inadditiontothethreeindexes,werunanddiscardPrincipalComponentAnalysis(PCA)asalastalternativetoreducethedimensionalityofthedataset.PCAmakesuseofhowvariablesrelatetoeachotherintheircorrelationmatrix,summarizingthedirectionsinwhichthedataisdispersed(Eigenvectors)andtherelativeimportanceofthedirections(Eigenvalues).Basedontheinput,PCAcreatesthesamenumberofnewvariables(“components”)butordersthemtodecreasetheamountofinformationcontained.Usingonlythefirstfewcomputedcomponents,itispossibletoreducethedataset'sdimensionalityandretainitsinformation.Nevertheless,itishardtojustifyusingPCAbecauseitscomponentslackmeaning,whichisessentialtoourresearchquestion.ThisisexactlytheproblemthatCubbinandStern(2006)runintowhenrunningregressionswithoutputfromPCAanalysis.Inthecaseofourdataset,wefindthatPCAretainstoomanycomponentsandlosestoomuchinformationcomparedtotheRISEandSummationindices.Overall,eachindependentvariablerepresentsanaggregateofrenewableenergypoliciesorenergypolicypackages.However,toconstructtherepresentationofrenewableenergypolicypackagesusefulforregressionanalysis,weemploythreeweightingmethods:theRISE,theSummation,andtheCompositeindexes.Becauseweareworkinginapanelformat,eachcountryhasthreeversionsofeachofthesevenindependentvariablesoverseveraldecades.AppendixFigureA.3compareseachindexbyHeading,aggregatedbyregion,for2015.Thatfigureshows,asexpected,thatthedifferencebetweenindicesisthelargestacrossHeadingsthatcontainmanydifferentpolicies.153.3INSTRUMENTALVARIABLES(IVS)TofindsuitableIVs,weexploreinternationalpoliticaleconomyaspectshighlightedintheliteratureonenergysectorgovernancereformthatspansseveraldecadesandregions.Wepositthatcountriesaremorelikelytoimplementregulatoryenergypolicieswhentheydisplayarelativelyhigherlevelofaffinity,orcloseness,withdevelopedcountriesthatchampionincreasedprivatesectorparticipationandotherrelatedchangesintheirloansrelatedtopowermarkets.Thisassumptionbroadlysatisfiestherelevancecondition.Theimportanceofconditionalityofreformforfinancialaidinenergysectorsofdevelopingcountriesiswellrecognizedintheliterature.(Heniszetal.,2005)(Heniszetal.,2005),forexample,arguethat"internationalpressuresofcoercionandemulationstronglyinfluencethedomesticadoptionofmarket-orientedreforms.”4Theliteraturehasalsoestablishedthatthepoliticalideologyofdevelopingcountrygovernmentsmattersforacceptingandcomplyingwiththeconditionalitytermsofeconomicreform(Smetsetal.,2013),includingpowersectorreform(Rufín,2003).Forexample,Imam,Jamasb,andLlorca(Imametal.,2019)findthatleft-winggovernmentsintheSub-SaharanAfricaregionareconsistentlylesslikelytosuccessfullyimplementpowersectorreformandimprovepowersectoroutcomes(i.e.,installedcapacityandelectricityaccess).Basedontheaboveconsideration,wearguethatdevelopingcountries'governmentswhosepoliticalideologyisclosertomajorWesterndonorsaremorelikelytoimplementrenewableenergyreforms.WeconsidertwochannelsthatmayrepresentaffinitytothemainG-7donors(France,Germany,Japan,theUnitedKingdom,andtheUnitedStates)whileotherwisenotcorrelatedtoenergymixoutcomes:(1)similarityinforeignpolicy;and(2)connectionthroughtrade.4Theyfindthatcoercionoccurs“[…in]asmanyas205countriesandterritoriesbetween1977and1999[with]thecoerciveeffectofmultilaterallendingfromtheIMF,theWorldBankorRegionalDevelopmentBanks[…]increasingovertime."16Wepinpointthreemeasurablewaysthisrapportmaybemeasuredandevidencedovertimeinthedata.Table3summarizesthechosenIVs,whichwedescribeandsupportintheparagraphsbelow.AppendixFiguresA.4andA.5illustratethetemporalpathofthethreeIVsaggregatedoverregionsfrom1980to2018.Table3:SummaryofInstrumentalVariablesClosenesstodonorsIVSupportedinForeignpolicyUNGeneralAssemblyvotingBaileyetal.(2017)TradeRelativetradevalueaggregates(Rufín,2003)TradeTradeagreementsinplaceSources:Baileyetal.(2017);UnitedNationsComtradedatasetviatheDatabaseforInternationalTradeAnalysis(BACI),bytheCenterforProspectiveStudiesandInternationalInformation(CEPII);EuropeanCommission.WeconsiderthatitmaytaketimetoimplementpoliciesfollowinganincreaseinclosenessmeasuredthroughourIV.Therefore,weuseamovingaverageoverfiveyearsoftheIVs.Althoughbilateralrelationshipscanberelativelyslowtochange,administrationchangesindemocraciesmayresultinmoreabruptchanges,sowealsoconsiderthemovingaverageoverthreeyearsoftheIVs.ForeignPolicyClosenessIVTorepresentchangesinforeignpolicypreferences,weuseadyadicdatasetbehindtheAffinityofNationsindexbyBaileyetal.(Baileyetal.,2017).TheyuseadynamicordinalspatialmodelonasingledimensiontoestimatestatepreferencestowardtheUS-ledliberalorder,asreflectedthroughtheUnitedNationsGeneralAssemblyvoting.ThemeasureistheIdealPointIndex.Ontheotherhand,theIdealPointDistanceisthedifferencebetweentheIdealPointsforallcountrydyadsthatparticipateintheUnitedNationsGeneralAssembly(e.g.,France17andGabon).Therefore,theIdealPointDistancesuggeststhedifferencebetweenthepreferencefortheUS-ledliberalorderforanytwocountriesinanygivenyear.Observethatthevotingdatasetdoesnotsimplymeasuresimilarityacrossallvotes.Rather,itestimatesthedistanceofvotingtowardaspecifictopic.AnchoringthecontentoftheestimatesinonetopichelpsaddresstheissuethattheG-7donorsdonotalwaysvotethesameway.Tounderstandhowpreferencesinoursamplechangedcomparedtothefivedonorsovertime,wesumtheyearlyIdealPointDistancebetweeneachcountryandthedonors.Sincewewantaclosenessindicatorratherthanadistanceindicator(i.e.,wewantourIVtobepositivelyassociatedwithrelativeclosenesstothedonors),wemultiplythesummedIdealPointDistancebythenegativeone.InternationalTradeClosenessIVHenisz,Zelner,andGuillén(Heniszetal.,2005)observethatcountriesimplementingeconomicreformsoftenimitatetheirtrade-relatedpeers.Theyarguethat"theintensityoftradetransactionsreflectsthedensityofthesocialnetworkinwhichagivencountryisembedded[…]andthereforethelevelofformalizedconformitywithinthenetwork."5Tocapturechangesinclosenessthroughtrade,wecomputethepercentageoftradethatcorrespondstoexchangewiththedonorsforeachrecipientcountryeachyear.TheexchangedatacomesfromtheUnitedNationsComtradedatasetviatheDatabaseforInternationalTradeAnalysis(BACI)(Gaulier&Zignago,2011).WealsogenerateapaneldatasetontheexistenceofTradeAgreementswiththeEuropeanUnion.ThesourceofthedataistheEUCommission's"currentstateofplay"agreementsinplace(seeAppendixTableA.6).This5Furthermore,theyarguethat"…policiesdirectlyreflectthelevelofformalizedconformitywithinatradenetwork.Inaworldcharacterizedbyuncertaincause-effectrelationships,thepolicyinitiativesundertakenby"relevantothers"suchastradepartnersrepresentanormativemodelthatlendscredencetoanalogousdomesticpolicyinnovationsandmaytriggeracross-nationaldiffusionprocess."18variableisbinaryandtakesavalueof1sincethetimethetradeagreementcameintoforce.Ifthereisnotradeagreement,thevariableis0fortheentiretyofthetimeseries.4RESULTSWepresentestimationresultsandspecificationtestsoftheempiricalmodel(equations2and3)estimatedoversixregions,fivedependentvariables,sevenpolicy-typevariables,threeIVs,twoIVmovingaverages,andthreeaggregationindices.Altogetherweestimate3,780regressions,towhichwealsoapplycountryinteractions.Theregressionsrenderthousandsofcoefficientsthatconstitutethebaseforouranalysis(Table4).Table4:EstimatedempiricalspecificationsSpecificationOptionsIVtype1.AffinitythroughUnitedNationsGeneralAssemblyvoting2.Affinitythroughbilateraltrade3.AffinitythroughEUtradeagreementsIVmovingaverage1.Fiveyearsmovingaverage2.ThreeyearsmovingaverageIndices1.RISE2.Composite3.SummationRegions1.EastAsia&thePacific(EAP)2.Europe&CentralAsia(ECA)3.LatinAmerica&Caribbean(LAC)4.TheMiddleEast&NorthAfrica(MENA)5.SouthAsia(SAS)6.Sub-SaharanAfrica(SSA)Dependentvariables1.Fossilfuelenergyconsumption,%ofthetotal(FFC)2.Electricityproductionfromfossilfuels(oil,gas&coalsources),%ofthetotal(EFF)3.Electricityproductionfromoilsources,%ofthetotal(EOS)4.Renewableenergyconsumption,%oftotalfinalenergyconsumption(REC)19SpecificationOptions5.Renewableelectricityoutput,%oftotalelectricityoutput(REO)RenewableEnergyPolicies1.Legalframeworkforrenewableenergy(LF)2.Planningforrenewableenergyexpansion(PE)3.Incentivesandregulatorysupportforrenewableenergy(IR)4.Attributesoffinancialandregulatoryincentives(AI)5.Networkconnectionanduse(NC)6.Counterpartyrisk(CR)7.Carbonpricingandmonitoring(CP)Followingtheliterature,wedefinethefirststageestimatefortheIVregressionsasstatisticallysignificantwhenithasap-valueatorbelow5%andanF-statisticabove10(Stock&Watson,2011).Additionally,weimposeatheoreticalrestrictionthattheestimatedrelationshipbetweentheinstrumentalandtheendogenousvariableinthefirststageispositive(i.e.,closenesstodonorsincreasesthelikelihoodofadoptingarenewableenergypolicy).Withthetheoreticalrestrictionabove,thesecond-stagecoefficientsarguablyrepresentthecausaleffectofthesevenenergypolicypackagesonthefiveenergymixoutcomes.Whiletheexogeneityofthesecond-stageestimatescannotbeestablishedwithcertainty,weperformadditionalrobustnessteststoestablishthevalidityofchoseninstrumentalvariables.Oneestablishedfindingintheeconomicsliteratureisthat"gettingsimilarresultsfromalternativeinstrumentsenhancesthecredibilityofinstrumentalvariableestimates"(Murray,2006)p.118.Theuniquenessofourempiricalapproachallowsustoconductaformaltestfordifferencesinthesecond-stageestimatesresultingfromdifferentinstrumentalvariables.Weregressavectorofestimatedsecond-stagecoefficientsforeachofthefiveenergymixoutcomesonthedummiesfortheinstrumentalvariableusedtoobtainthesecond-stageestimate,theenergypolicypackage,thetypeofindexused,andthecountryandfixed-effects.20Table5showstheestimatedregressionresults.Weonlyobservestatisticallysignificantdifferencesbetweenthesecond-stageestimatesobtainedfromaffinitythroughtheUNGAvotingandtheothertwoinstrumentalvariablesforoneofthefiveenergymixoutcomes(electricityproductionfromoilsources).Thereareeithernostatisticallysignificantdifferencesacrossestimatedsecond-stagecoefficients,andthethreechoseninstruments,orthedifferencesareonlymarginallysignificantforallotherenergymixoutcomes.Theseresultsgiveusgreaterconfidencethatoursecond-stageestimatesreflectthecausaloutcomeofrenewableenergypolicyreforms.Table5:SpecificationRegressionTestforIVExogeneity(1)(2)(3)(4)(5)FCCEFFEOSRECREOIV:EUAgreements0.270.781.85-0.43-0.22(0.16)(0.34)(1.44)(0.30)(0.24)IV:Affinitythroughbilateraltrade0.796.344.23-0.15-1.27(0.67)(3.27)(0.72)(0.42)(0.67)Policy:Planningforrenewableenergyexpansion-1.11-2.95-2.51.83.06(0.57)(0.93)(1.48)(0.68)(1.33)Policy:IncentivesandregulatorysupportforRE-0.33-1.02-1.020.211.16(0.69)(1.06)(1.56)(0.44)(0.94)Policy:Attributesoffinancialandregulatoryincentives-0.47-0.9-0.950.732.21(0.33)(0.52)(0.87)(0.73)(1.13)Policy:Networkconnectionanduse-0.14-0.391.380.742.19(0.33)(0.39)(1.00)(0.35)(0.89)Policy:Counterpartyrisk0.150.470.73-0.260.51(0.36)(0.96)(1.86)(0.44)(0.84)Policy:Carbonpricingandmonitoring-0.450.811.68(0.84)(0.29)(0.80)Index:Composite-0.22-0.1-0.980.67.99(0.08)(0.30)(0.46)(0.32)(0.38)Index:Summation-0.050.14-0.28-0.090.02(0.07)(0.23)(0.45)(0.12)(0.14)Constant-0.62-1.65-3.610.24-0.2(0.33)(0.50)(0.84)(0.36)(0.88)Observations707717726947923R-squared0.0710.1460.0750.010.055CountryFixedEffectsYESYESYESYESYES21WealsoperformKruskal-WallisHspecificationteststodeterminewhetherthereisastatisticallysignificantdifferencebetweenthenumberofeligiblefirst-stageresultsobtainedusinganyindices,IVs,andIVmovingaveragessummarizedinTable4.Wechosethistestbecauseitisimpervioustothenormalityassumption,likelytobeviolatedacrossdistributionsofestimatedfirst-stagecoefficients.TheresultsinthefirstcolumnofTable6rejectthehypothesisthatdifferentwaysoftallyingpolicieswithinenergypolicypackagesmakeadifferenceintheregressionestimates.Inotherwords,thereisnosignificantdifferencebetweenthethreeindicesacrossthenumberofeligiblefirst-stageregressions.ThechoiceofanIVvariabledoesmakeasignificantdifferenceacrossthenumberofeligiblefirst-stageregressions(Table6,secondcolumn).However,asshowninTable5,thesefirst-stagedifferencesdonotaffectthesecond-stageestimatesconsistently.Finally,thereisnosignificantdifferenceacrossthenumberofeligiblefirst-stageregressionswhencomparingmovingaveragesof3and5periodsforourIVs(Table6,thirdcolumn).Table6:Kruskal-WallisHtestssummaryIndicesIVIVmovingaverageschi-squaredp-valuechi-squaredp-valuechi-squaredp-value0.880.6415.160.00050.1950.66Note.Statisticaltiesarenotaccountedfor.ResultswithstatisticaltiesareindistinguishablefromtheonesreportedintheTable.Basedontheoutcomesofthetwotestsabove,wecontinueouranalysiswiththeempiricalspecificationthatyieldsthehighestnumberofpositiveandstatisticallysignificantfirst-stagecoefficients:(i)theRISEindex;(ii)theIVbasedonaffinitythroughtheUnitedNationsGeneralAssemblyvoting,and(iii)theIVmovingaverageof5years.22Ofthe1,902first-stageregressionsinthechosenempiricalspecification,28percent(or540coefficients)areeligibleforthesecond-stageestimation(AppendixTableA.7).ThetotalrowofTable7showstheshareofsignificantsecond-stagecoefficientsasapercentageoftotaleligiblefirst-stageregressions.Only15.7percent(or85coefficients)oftheestimatedsecond-stageregressionsmeetthestatisticalsignificancethresholdofap-valuebelow10percent.Theseresultsindicatethat,onaverage,roughlyoneoutofsixrenewableenergypolicypackageshavehadanyimpactonthedevelopingcountriesenergymixsuchthatlimitingthisstudytoamorerecenttimeframeduringwhichclimateconcernsgainedimportanceinenergypolicydecisionswouldhaveseverelyrestrictedouranalysis.Inaddition,amere4.8percent(or26coefficients)oftheestimatedsecond-stageregressionsarepositiveandstatisticallysignificant.Morethantwooutofthreesignificantenergypolicypackageshavehadtheoppositeeffectthanintended.Astheinputdataarestandardized,therelationshipismeasuredinunitsofstandarddeviationdistancefromthemeanandcanbecomparedacrossenergypolicypackagesandenergymixoutcomes.Thesignificanceandeffectivenessofthereformseemtovaryacrossgeographicregions.WiththeexceptionsoftheEuropeandCentralAsia,andtheSub-SaharanAfricaregions,allotherregionshadalessthan5percentshareoftherenewableenergypolicieswithasignificanteffectontheenergymix(Table7,top).Moreover,contrarytowhatonewouldexpect,boththesignificanceandeffectivenessofrenewableenergypoliciesdeclinewithincomelevel.(Table7,bottom).AppendixTableA.8displaysthissameinformationbyenergypolicypackageandenergymixoutcome.23Table7:Significantsecond-stagecoefficientsasapercentageofeligiblefirst-stageregressions,combiningtheresultsforall7energypolicypackages.GroupTotaleligiblefirst-stageregressionsSharewithsignificantsecond-stagecoefficientsSharewithpositiveandsignificantsecond-stagecoefficientsRegionSSA22422.8%6.7%EAP864.7%1.2%ECA5044.0%18.0%LAC1166.0%0.0%MENA362.8%2.8%SAS280.0%0.0%Total54015.7%4.8%IncomeHigh490.0%0.0%UpperMiddle11315.0%0.9%LowerMiddle24615.4%5.3%Low13222.7%9.1%Total54015.7%4.8%Notes:Regressionspecification:RISEindex,UNGAaffinityIVwithfiveyearsmovingaverages.SSA:Sub-SaharanAfrica.EAP:EastAsia&thePacific.ECA:Europe&CentralAsia.LAC:LatinAmerica&Caribbean.MENA:theMiddleEastandNorthAfrica.SAS:SouthAsia.Whiletheaggregateresultssuggestthatrenewableenergypolicieshadamodestsignificanteffectontheenergymix,theyfallshortofexplainingtheeffectofeachofthesepoliciesseparatelyandovertime.Figure1illustratestheresultsofourmainresearchquestion,whichshowsdistributionalboxplotsoftheeffectsofeachenergypolicypackageaggregatedacrossallregions.Inadditiontothedefaultlagof3years,weconsiderthepossibilitythattheeffectofeachenergypolicypackagechangeswithtimeandanalyzelagsof5and7years.Whenbreakingdownbyenergypolicypackage,itbecomesimportantthatcoefficientsrelatedtodifferentenergymixoutcomesforthesamecountryclustertogether(e.g.,Kenya,Eritrea,andAngolainAppendixFigureA.6)duetotheinherentsimilarityoftheoutcomevariables.24Toavoidbiasingtheresultsbyenergypolicypackagetowardsthecountriesforwhichtherearemoreavailableoutcomes,wekeeponlyoneoutcomecoefficientatrandombyenergypolicypackageandcountryinsubsequentanalysis.Becausethereisnotheoreticalreasontopreferanyoutcomevariableoveranother,andbecausewehaveshownthattheytendtocluster,wesimplychoosethefirstavailablecoefficientfromtheoutcomesfromalistorderedinthesamewayasthelefttorightcolumnsinTable5.Here,wetakeamomenttoconsiderthatofourthreefossilfuelvariables,onlyoneofthem(electricityfromoilsources)excludesnaturalgas,whichmaybeconsideredatransitionfuelinsomecountries.AppendixFigureA.6helpsusconsiderthepossibilitythatlumpinggasinwithotherfossilfuelscouldtheoreticallybedrivingourpessimisticresultsregardingtheimpactofenergypolicypackagesonthedecarbonizationoftheenergymix.Becausetheeffectsoftheenergypolicypackagesonfossilfuels(red)andrenewables(blues)tendtocluster,potentialmisclassificationofnaturalgasisunlikelytodriveourresults.PatternsfromFigure1(andAppendixTablesA.9-A.10,whichsummarizethemeansofestimatedsecond-stagecoefficientsacrossregionsandincomecategories)showthatallenergypolicypackagesexceptcounterpartyriskhadconsistentlyhigheraverageeffectsovertime.Moreover,planningforexpansion,incentives,andregulatorysupport,attributesoffinancialandregulatoryincentives,andnetworkconnectionanduseenergypolicypackagesovercomenegativemedianssevenyearsaftertheirimplementation(Figure1).25Figure1:Boxplotofnormalizedsecond-stagecoefficientsbyanenergypolicypackageNotes:Theunitoflagsareyears;differentlagsaredenotedbythecolorsinthelegend.LF=Legalframework;PE=Planningforexpansion;IR=Incentivesandregulatorysupport;AI=Attributesoffinancialandregulatoryincentives;NC=Networkconnectionanduse;CR=Counterpartyrisk.Regressionspecification:RISEindex,UNGAaffinityIVwithfiveyearsmovingaverages.Intheboxplot,thereisaboxfromthefirstquartiletothethirdquartile,withthe2ndquarter(50%percentile)markedbytheinternallineofthebox.Thewhiskersextendingfromtheboxesgofromeachquartiletotheminimumormaximum,excludingoutliers.Table8:Averageofnormalizedsecond-stagecoefficientsbyenergypolicypackagesandyearlylagsLagsinyearsEnergypolicypackage357LegalFramework-0.66-0.250.90Planningforexpansion-2.200.483.02Incentivesandregulatorysupport-0.30-0.922.04Attributesoffinancialandregulatoryincentives-5.355.4819.17Networkconnectionanduse-0.08-0.431.48CounterpartyRisk0.38-2.24-0.67Notes:Regressionspecification:RISEindex,UNGAaffinityIVwithfiveyearsmovingaverages.Normalizedsecond-stagecoefficient26Unlikeotherenergypolicypackages,thecounterpartyrisktypehasthehighestandonlypositivemedianandmeanclosesttoimplementation(lagof3years).Itsmedianishigherwhenusingatimelagof5years,too,exceptwhencomparedtofinancialandregulatoryincentivesattributes.AppendixTableA.3describesthecontentoftheenergypolicypackages.Thecounterpartyriskpackageincludesgovernmentguaranteesorothermeanstoensurethecreditworthinessofprojectsprocuredthroughauctionsorotherwise.Oneinterpretationisthatmitigatingcounterpartyriskcouldhavecomparativelyimmediateeffects.Anotherinterpretationoftheresultisthataddressingcounterpartyriskinfluencesoutcomespositivelywhilesupportingotherenergypolicypackagesthattakemoretimetohavetheintendedoutcomes.Eitherway,thisresultgivescredencetotheideathatpoliciesthataddressthebankabilityofprivateinvestmentinrenewableenergyarecrucialforenergydecarbonization.Inthefollowingsection,wechallengeourresultsinthreewaysandarriveatsimilarconclusions.5ROBUSTNESSCHECKSANDDISCUSSIONBelowweexploretherobustnessofourfindingstoassumptionsusedinconstructingthedatasampleandvariables.Wethendiscussthecaveatsandpolicyimplicationsofourfindings.275.1ROBUSTNESSCHECKSLimitingthecountrysampleOuranalysisofthesecond-stageestimates'variationacrosstimeassumesthateconometricmethodssuchascountry-fixedeffectsandregionalpanelscanfullyeliminateallconfoundingcountry-specificcharacteristicsfromthedataandmakecross-countrycomparisonspossible.Ifthiswerenotthecase,onecouldnottakeourresultsatfacevaluebecausethecountrysampleisnothomogenousacrossalllagsandenergypolicypackages.Forinstance,estimatedcoefficientsofthelegalframeworkfortherenewableenergypolicypackageexistforGhanaonlyinlag3andPeruonlyinlag5.Forrobustnesspurposes,wechallengeourresultsbynarrowingthecomparisononlytocountrieswithsignificantcoefficientsacrossallthreetimelags.Thereareonlyninecountriesthatfulfillthiscriterion.Theresultsinthisrestrictedsamplestillpointtoatemporaldimensionofeffectsacrossenergypolicypackages.ExamplesofconsistentimprovementsinenergymixmetricsbyenergypolicypackageareevidencedinattributesoffinancialandregulatoryincentivesandincentivesandregulatorysupportforrenewableenergypolicypackagesforKenyaandUkraine,respectively.Unfortunately,thesizeofthedatasampleavailableforthisrobustnesscheckrestrictsusfromreasonablyaveragingoverenergypolicypackagesseparately.However,averagesoftheeffectsofallenergypolicypackagestogetherstillyieldincreasesovertime(0.73,1.72,3.83inlags3,5,and7,respectively).Moreover,theresultholdsevenwhenweexcludeoutliers(inthissample,averagesare0.14,0.24,and0.29inlags3,5,and7,respectively).KeepingonlymodernrenewablesbyremovinglargehydropowerWeacknowledgethattheintroductionoflargehydropowerinsteadofmodernrenewablesmayobscuretheeffectsoftheenergypolicypackageswestudy.Therefore,for28robustness,were-estimatethefinalspecification(i.e.,using(1)theRISEindex;(2)theIVbasedonaffinitythroughUnitedNationsGeneralAssemblyvoting;and(3)theIVmovingaverageof5years)withoutlargehydropower.WeremovehydropowerbyreplacingRECandREO(“Renewableenergyconsumption”and“Renewableelectricityoutput”)variableswiththesharesof“ModernEnergyGenerationoverTotalGeneration”and“ModernEnergyCapacityoverTotalInstalledCapacity”asourdependentvariables.Werefertothenewfivedependentvariablesasthe“altereddependentvariables.”Resultsarelittlechanged.AlthoughsomeregionsexperiencedimprovementscomparedtotheresultsofTable7(AppendixTableA.11),theenergypolicypackagesstillhadaminimalornegativeimpactontheenergymix,whilesecondstageresultsstillclusteredtogether(AppendixFigureA.7).FigureA.8andTableA.12intheAppendixreplicatethemainresultsusingthealtereddependentvariables.Themedianandmeanresultsareslightlydifferent.Themeansoftypesrelatedtothelegalframework,attributesoffinancialandregulatoryincentives,andnetworkconnectionandusehoverclosetozeroinsteadofhavingimmediatenegativeeffects.Themeansofplanningforexpansionandattributesoffinancialandregulatoryincentivesstillincreaseconsistentlyovertime,thoughthemeansofallremainingthreeenergypolicypackagespeakinthesecondperiod.Theenergypolicypackagethatdealswithcounterpartyriskremainstheonewiththehighestmeanclosesttoimplementation,anditgoesdownovertime.Mediansarenegativeorclosetozeroforallenergypolicypackages,exceptforattributesoffinancialandregulatoryincentives,whichincreasesovertime,andforcounterpartyrisk,whichpeaksafterfiveyearsandbecomesnegativeaftersevenyears.29ConsideringabsoluteinsteadofrelativechangesinmodernrenewablesWeconsiderthattheabsoluteeffectsofthetwonewmodernrenewableoutcomes(“ModernEnergyCapacity”and“ModernEnergyGeneration”)areobscuredbysimilarordisproportionategrowthintheoverallenergymix(thedenominator),whichincludesincumbenttechnologieslikefossilfuelsandlargehydropower.We,therefore,re-estimatethefinalspecification(i.e.,using(1)theRISEindex;(2)theIVbasedonaffinitythroughUnitedNationsGeneralAssemblyvoting;and(3)theIVmovingaverageof5years)consideringtheabsolute,notrelative,changesinmodernrenewables.Figure2:Boxplotofnormalizedsecond-stagecoefficientsforonlytheabsolutemodernrenewablecapacityoutcomeNotes:Theunitoflagsareyears;differentlagsaredenotedbythecolorsinthelegend.LF=Legalframework;PE=Planningforexpansion;IR=Incentivesandregulatorysupport;AI=Attributesoffinancialandregulatoryincentives;NC=Networkconnectionanduse;CR=Counterpartyrisk.Regressionspecification:RISEindex,UNGAaffinityIVwithfiveyearsmovingaverage.Intheboxplot,thereisaboxfromthefirstquartiletothethirdquartile,withthe2ndquarter(50%percentile)markedbytheinternallineofthebox.Thewhiskersextendingfromtheboxesgofromeachquartiletotheminimumormaximum,excludingoutliers.Normalizedsecond-stagecoefficient30Figure3:Boxplotofnormalizedsecond-stagecoefficientsforonlytheabsolutemodernrenewablegenerationoutcomeNotes:Excludesoutliers.Theunitoflagsareyears;differentlagsaredenotedbythecolorsinthelegend.LF=Legalframework;PE=Planningforexpansion;IR=Incentivesandregulatorysupport;AI=Attributesoffinancialandregulatoryincentives;NC=Networkconnectionanduse;CR=Counterpartyrisk.Regressionspecification:RISEindex,UNGAaffinityIVwithfiveyearsmovingaverage.Intheboxplot,thereisaboxfromthefirstquartiletothethirdquartile,withthe2ndquarter(50%percentile)markedbytheinternallineofthebox.Thewhiskersextendingfromtheboxesgofromeachquartiletotheminimumormaximum,excludingoutliers.Thenumberofcoefficientsforthesetwooutcomevariablesindividuallyistoolowtovisualizebytheenergypolicypackage.We,therefore,aggregatetheresultsintoboxplots(seeFigures2and3)thatrepresentthethreetimelagswestudy.AppendixTableA.13showstheaverageeffectofthepolicypackagesonthesetwoabsoluteoutcomes,bytimelag.Overall,theimmediateeffectsofpoliciesaresmallandclosetozero;temporalresultsalsoechotheresultsofpreviousanalyses.Normalizedsecond-stagecoefficient315.2DISCUSSIONInthispaper,weaddressthefollowingquestions:Howdotheeffectsofsevenmajorenergypolicypackagesonthedeploymentofcleanenergytechnologiescompareindevelopingcountries?Andhowdosucheffectschangefromtheshorttomediumtermafterimplementation,bypolicycategory?Peñascoetal.'s(2021)relevantsystematicreviewofdevelopedcountriescanhelpguideexpectationsonourcountrysample.Thereviewshowsthatabout50percentofevaluationsfindthatregulatorypoliciesthatestablishrenewableenergyobligations,includinglegalframeworksforrenewableenergy,whichisoneofourenergypolicypackages,donotaffectrenewabledeployment.ResultsofstudiesonGHGtradingschemesinourenergypolicypackagesoncarbonpricingandmonitoringarealsomixed.Fifty-threepercentreportnoimpact,and8percentreportanegativeimpact.Ontheotherhand,75percentofstudiesonthepolicygroupsthatincludetaxesandgrantsfindpositiveeffects.Theliteraturebroadlysupportstheeffectsoffeed-intariffsandfeed-inpremiums(broadly,subsidiesforrenewableenergyandpartofourincentivesregulatorysupportenergypolicypackage).About86percentofevaluationsfoundthatfeed-intariffspositivelyaffectedrenewableenergydeployment.Today,feed-intariffsarelosinggroundtoauctions(competitivebiddingprocessesforprivatesectorinvestmentinrenewableenergydeploymentthatareincludedinourenergypolicypackageonattributesoffinancialandregulatoryincentives).Fifty-ninepercentofstudiesonauctionsreportapositiveimpact,and41percentreportanegativeornegligibleimpactondeployment.Whiletherearerelativelylessdatatoassessauctions,itisunderstoodthatdesignelementsarecrucialtosuccess.Consistentwiththeliteratureontheenergysectorreformindevelopingcountries(Foster&Rana,2020;Jamasbetal.,2005,2015),ourmainresultsandrobustnesschecks32pointtoveryloweffectivenessacrossenergypolicypackagestonearthedecarbonizationoftheenergymix.Inourresults,onlyone-sixthofthecoefficientsrepresentingenergypolicypackageshaveevenmodeststatisticalsignificance.Moreover,mostofthemarethesignoppositetowhatonewouldexpect.Astheystand,theseresultsseemtosuggestthat,atleastwithin3-7yearsstudied,decarbonizationpoliciesinourcountrysamplesfailtodeliverontheirgoalsofreducingtheshareoffossilfuelsintheirenergymix.Theresultsmaybedrivenbyahostofinterrelatedissuesindevelopingcountries,andtheyalllikelyplayarole.Theinstitutionalcapacityinthecountriesweexploreisweakerascomparedtothecountriescoveredbythepolicyevaluationliteratureindevelopedcountries(Foster&Rana,2020),wheretheimpactonenergymixesandrenewableenergytechnologydeploymentsseemtobemorepositive.Arelativelackofinstitutionalcapacitycanalsonegativelyaffecttheabilitytosecurefinanceindevelopingcountries.Thisproblemisitselfconnectedtofactorssuchasmacroeconomicconditionsandalackofinfrastructure(Eglietal.,2019;Moner-Gironaetal.,2021),althoughthecrucialrolefinanceplaysindecarbonization(Buchneretal.,2019;IRENAandClimatePolicyInitiative(CPI),2020;Macquarieetal.,2019;Steckeletal.,2017).Theimportanceofsecuringfinanceisinlinewithourresultssurroundingthecounterpartyriskpackage.Indeed,itistheonlyenergypolicypackagethatyieldsanincreaseinrenewablesindevelopingcountriesenergymixthreeyearsafterimplementation.Thisresultagaintiesinwithexistingresearch.AccordingtoPeñascoetal.,(2021),“duediligenceofprojectsfromcommercialorinvestmentbanks”iscrucialforthesuccessofauctionsindevelopedcountries.33Inadditiontotherelativelypositiveeffectsofpoliciesthataddresscounterpartyrisk,thereissomefurtherbasisforoptimism,astheeffectivenessoftheenergypolicypackagesimprovesovertimeoverall.WepositthattheSailingShipEffect(Gilfillan,1935;Ward,1967),whereincumbenttechnologiestemporarilyimprovetheirproductivityinresponsetocompetitivethreatsbynewtechnologies,couldbeapotentialdriverforthesedynamics.Interactionbetweenenergypolicypackagescouldalsohelpexplaingreaterpositiveeffectsovertime.Addressingcounterpartyriskfirstmightbuttressandsupportotherenergypolicypackagesovertime.Despiteoureffortstoidentifythecausalrelationshipbetweenenergypolicypackagesanddecarbonization,ouranalysisislimitedbytheextenttowhichourmethods,throughIvsandcontrols,canaddressotherpatternsshapingtheenergysectorinourcountrysample.Theseinclude,forexample,changesinenforcementcapabilitiesovertime,whichstaticcountryfixedeffectscannotcontrol.Whileweestablishthelimitedeffectivenessoftherenewableenergypoliciesinachievingdecarbonizationinoursample,thevalidityoftheexclusionrestrictionoftheIvscouldbealimitationpotentiallyaffectingtheresults.Inaddition,asmeasuredthroughforeignpolicyandtradeIvs,closenesstomajordonorscouldaffectpoliciesoutsidethepowersector.Thosenon-powersectorpolicies,inturn,mayhavehadsomeeffectsontheenergymix.Nevertheless,wecouldnotfindalternativeinstrumentsanddatathatcoveredthebreadthofgeography,powersectorpolicy,andoutcomesthatourresearchquestionsentailed.Furtherresearchmaybeabletoconsiderotherinstruments,especiallyiftheanalysisisnarrowerinscope.Last,thisdiscussionconsideredthattheinterrelationbetweenenergypolicypackagescouldhelpexplainourfindingofmorepositiveeffectsovertime.TheevolvinginterdisciplinaryanalyticalframeworkofpolicymixesspearheadedbyRogge&Reichardt34(2016)isadescriptiveconceptualframeworkforthepolicy-makingprocessandisappropriateforsocialscienceresearchquestionsinmultiplefields.ThelimitedrelevantempiricalworkinenergyincludesSchmidt&Sewerin,(2019),whoanalyzepolicymixesinninedevelopedcountries,althoughthatstudydoesnotconsiderpolicyinteractions.6CONCLUSIONSANDPOLICYIMPLICATIONSAchievingdecarbonizationworldwiderequiresarobustunderstandingofhoweffectivedifferentrenewableenergyandclimatepoliciesareinabroadrangeofcountriesandtheirdynamicsovertime.Researchonthetopicisespeciallyimportantindevelopingcountries,wheretheliteratureisrelativelyscarce.Thisstudyshedslightontheseimportantissuesbyconductingafirstsystematicassessmentofhowsevenrenewableenergypolicypackagesaffecttheenergymixindevelopingcountriesovertime.WerelyonthebackgrounddatabehindtheRISEindicatorspublishedbytheESMAPattheWorldBank.Weaddressseveralwell-knowneconometricissuesintheexistingliteraturethatusessimilardatasets,includingomittedvariablesandsimultaneityandreversecausalitybetweenenergypolicypackagesandoutcomes.Weestimatethousandsofindicator-instrument-outcome-levelcountryandtimefixed-effectsregressionsoverregionalpanelscoveringmorethan100developingcountriesandfourdecadesofenergysectorpolicies.Wecrediblyevaluatetherobustnessofindicators’measurement,qualityofinstrumentalvariablesused,andsignificanceanddirectionofestimatedenergypolicypackagecoefficients.Notably,wefindnomajormeasurementdifferenceswhenwetrydifferentpolicyinstrumentaggregationoutcomes,allowingustoconcludethattheaggregationmethodusedbytheWorldBankisrobusttopotentialunder-andoverweightingproblems.Wealsofindnomajordifferencesinthesecond-stageestimatesobtainedby35differentIVs,whichexploitdifferentsourcesofarguablyexogenousvariationinrenewableenergypolicies.Thisresultaddsrobustnesstoouridentificationapproach.Ourfindingsoftheeffectsofrenewableenergypoliciesondecarbonizationoutcomesindevelopingcountriesarequitepessimistic.Onlyone-sixthoftheestimatedpolicycoefficientshaveevenmodeststatisticalsignificance.Moreover,manyofthesepolicieshaveasignoppositetowhatonewouldexpect(theyareassociatedwithnegativeimpactsonenergymixesintermsofpromotingdecarbonization)orminimaleffects.Theseresultssuggestthatatleastshort-tomedium-termrenewableenergypoliciesindevelopingcountriesmayfailtodeliverontheirgoalsofreducingtheshareoffossilfuelsintheirenergymixwithoutotherconcurrentchangesbothattheinternationalanddomesticlevels.Theresultspointtoimportantavenuesforpolicyandfutureresearch.Wesuggestthepossibledriversofthisresult,suchasweakinstitutionalcapacity,thattranslateintodifficultiessecuringfinance.Theresultssuggestthatwithoutadditionalinternationalclimatefinanceandinvestmentsininstitutionalcapacity,effortstocreateadditionaldecarbonizationpoliciesmaynotsignificantlyimpacttheenergymixinthesecountriesintheshorttomediumterm.Fulfilling(andlikelyexceeding)commitmentsmadeduringtheParisAgreementandtheGlasgowClimateActforclimatefinanceareessentialtoenablelowercarbonenergymixes.OurresultsalsopointtoapossibleroleforadditionalSouth-Southinteractionstobuildonexperiencesrelatedtopolicydesign,enforcement,andmonitoring.Thereare,however,somefindingsthatlendabasisforoptimism.WeseethattheeffectivenessofrenewableenergypoliciesimprovesovertimeanddiscussevidencefortheSailingShipEffect.Additionally,policiesthataddresscounterpartyriskhavethegreatestimmediateimpact,underscoringtheimportanceofaccesstofinanceonincorporatingrenewablesintotheenergymix.36Ourresultsarerobusttodifferentwaysofcodingandaggregatingpoliciesandtovariousrobustnesschecks,whichincludeanalyzingamorestringentcountrysample,removinglargehydropowerfromoutcomevariables,andassessingabsoluteinsteadofrelativeoutcomevariables,amongothers.Weseeseveralvenuesforfutureresearch.Understandingofthecausalmechanismexplainingourresultscanbefurtherimproved.Studiesthatutilizemoregranulardataattheindustrylevelarenecessarytoelucidatetheeffectsofunobservedfactorsindevelopingcountries,suchastheextenttowhichrenewablepoliciesareenforced.Anotherimportantdirectionforfutureresearchistostudyfirm-levelresponsestorenewableenergypolicies,includingimplicationsforproductivity,entryandexit,theturnaroundofcapitalstock,andotherconstraintstorenewableenergytechnologyadoption.REFERENCESBacon,R.(2018).TakingStockoftheImpactofPowerUtilityReforminDevelopingCountriesALiteratureReviewWorldBankPolicyResearchWorkingPaperNo.8460.Bailey,M.A.,Strezhnev,A.,Voeten,E.,&Walsh,E.A.(2017).EstimatingDynamicStatePreferencesfromUnitedNationsVotingData.JournalofConflictResolution,61(2),430–456.Bang,G.,Underdal,A.,&Andresen,S.(2015).Thedomesticpoliticsofglobalclimatechange:Keyactorsininternationalclimatecooperation.TheDomesticPoliticsofGlobalClimateChange:KeyActorsinInternationalClimateCooperation,1–216.https://doi.org/10.4337/978178471493237Bednar-Friedl,B.,Biesbroek,R.,Schmidt,D.N.,Alexander,P.,Børsheim,K.Y.,Carnicer,J.,Georgopoulou,E.,Haasnoot,M.,LeCozannet,G.,Lionello,P.,Lipka,O.,Möllmann,C.,Muccione,V.,Mustonen,T.,Piepenburg,D.,&Whitmarsh,L.(2022).Europe:ClimateChange2022:Impacts,Adaptation,andVulnerability.ContributionofWorkingGroupIItotheSixthAssessmentReportoftheIntergovernmentalPanelonClimateChange.Boasson,E.,Wettestad,J.,Leiren,M.,Szulecki,K.,Inderberg,T.H.,Bäckstrand,K.,Faber,H.,Reimer,I.,&Banet,C.(2020).ComparativeRenewablesPolicy:Political,OrganizationalandEuropeanFields.Buchner,B.,Clark,A.,Falconer,A.,Macquarie,R.,Meattle,C.,Tolentino,R.,&Wetherbee,C.(2019).GlobalLandscapeofClimateFinance2019.Choi,H.,&Anadón,L.D.(2014).Theroleofthecomplementarysectoranditsrelationshipwithnetworkformationandgovernmentpoliciesinemergingsectors:Thecaseofsolarphotovoltaicsbetween2001and2009.TechnologicalForecastingandSocialChange,82,80–94.Cubbin,J.,&Stern,J.(2006).Theimpactofregulatorygovernanceandprivatizationonelectricityindustrygenerationcapacityindevelopingeconomies.WorldBankEconomicReview,20(1),115–141.Doblinger,C.,Surana,K.,&Anadón,L.D.(2019).Governmentsaspartners:TheroleofalliancesinU.S.cleantechstartupinnovation.ResearchPolicy,48(6),1458–1475.Doumbia,D.(2021).PowerMarketSophisticationandSectorOutcomes:AFocusonSocialPerformance,ElectricityReliability,andRenewableEnergyPenetration.WorldBankPolicyResearchWorkingPaperNo.9585.Egli,F.,Steffen,B.,&Schmidt,T.S.(2019).Biasinenergysystemmodelswithuniformcostofcapitalassumption.NatureCommunications201910:1,10(1),1–3.38ESMAP.(2022).BuildingResilience:RegulatoryIndicatorsforSustainableEnergy.TheWorldBank.Foster,V.,&Rana,A.(2020).RethinkingPowerSectorReformintheDevelopingWorld.RethinkingPowerSectorReformintheDevelopingWorld.Gaulier,G.,&Zignago,S.(2011).BACI:InternationalTradeDatabaseattheProduct-level.InMpra(Vol.31398,p.33).Gilfillan,S.C.(1935).InventingtheShip.Follett.Greenacre,M.,&Primicerio,R.(2013).Measuresofdistancebetweensamples:Euclidean.InMultivariateAnalysisofEcologicalData.FBBVA.Henisz,W.J.,Zelner,B.A.,&Guillén,M.F.(2005).TheWorldwideDiffusionofMarket-OrientedInfrastructureReform,1977–1999.AmericanSociologicalReview,70(6),871–897.Imam,M.I.,Jamasb,T.,&Llorca,M.(2019).Politicaleconomyofreformandregulationintheelectricitysectorofsub-SaharanAfrica.WorkingPaper.UniversityofCambridge,FacultyofEconomics.IRENAandClimatePolicyInitiative(CPI).(2020).GlobalLandscapeofRenewableEnergyFinance.Jaccard,P.(1908).DistributiondelaflorealpinedansleBassindesDransesetdansquelquesregionsvoisines.BulletindeLaSocietéVaudoiseDesSciencesNaturelles,37,241–272.Jamasb,T.,Mota,R.,Newbery,D.,&Pollitt,M.(2005).ElectricitySectorReforminDevelopingCountries:ASurveyofEmpiricalEvidenceonDeterminantsandPerformance.WorldBankPolicyResearchWorkingPaperNo.3549.39Jamasb,T.,Nepal,R.,&Timilsina,G.R.(2015).AQuarterCenturyEffortYettoComeofAge:ASurveyofPowerSectorReformsinDevelopingCountries.WorldBankPolicyResearchWorkingPaperNo.7330.Lamb,W.F.,&Minx,J.C.(2020).Thepoliticaleconomyofnationalclimatepolicy:Architecturesofconstraintandatypologyofcountries.EnergyResearchandSocialScience,64.Macquarie,R.,Naran,B.,Rosane,P.,Solomon,M.,Wetherbee,C.,Buchner,B.,Choi,J.,Davila,E.,Dreyer,C.,&Falconer,A.(2019).UpdatedviewontheGlobalLandscapeofClimateFinance2019.Mallawaarachchi,T.,Nepal,R.,&Ross,H.(2021).MeetingenvironmentalobjectivesthroughenergysectorreformsinAsiaandthePacific.https://doi.org/10.1080/14486563.2021.1989009,28(4),305–313.Moner-Girona,M.,Bender,A.,Becker,W.,Bódis,K.,Szabó,S.,Kararach,A.G.,&Anadón,L.D.(2021).Amultidimensionalhigh-resolutionassessmentapproachtoboostdecentralisedenergyinvestmentsinSub-SaharanAfrica.RenewableandSustainableEnergyReviews,148,111282.Murray,M.P.(2006).AvoidingInvalidInstrumentsandCopingwithWeakInstruments.TheJournalofEconomicPerspectives,20(4),111–132.Nagayama,H.(2009).Electricpowersectorreformliberalizationmodelsandelectricpowerpricesindevelopingcountries:Anempiricalanalysisusinginternationalpaneldata.EnergyEconomics,31(3),463–472.Peñasco,C.,Anadón,L.D.,&Verdolini,E.(2021).Systematicreviewoftheoutcomesandtrade-offsoftentypesofdecarbonizationpolicyinstruments.InNatureClimateChange(Vol.11,Issue3,pp.257–265).NatureResearch.40Rogge,K.S.,&Reichardt,K.(2016).Policymixesforsustainabilitytransitions:Anextendedconceptandframeworkforanalysis.ResearchPolicy,45(8),1620–1635.Rufín,C.(2003).Thepoliticaleconomyofinstitutionalchangeintheelectricitysupplyindustry:Shiftingcurrents.EdwardElgar.Schmidt,T.S.,&Sewerin,S.(2019).Measuringthetemporaldynamicsofpolicymixes–Anempiricalanalysisofrenewableenergypolicymixes’balanceanddesignfeaturesinninecountries.ResearchPolicy,48(10),103557.Sen,A.,Nepal,R.,&Jamasb,T.(2016).ReformingElectricityReforms?EmpiricalEvidencefromAsianEconomies.InTheOxfordInstituteforEnergyStudies(OIES).Smets,L.,Knack,S.,&Molenaers,N.(2013).Politicalideology,qualityatentryandthesuccessofeconomicreformprograms.TheReviewofInternationalOrganizations,8(4),447–476.Steckel,J.C.,Jakob,M.,Flachsland,C.,Kornek,U.,Lessmann,K.,&Edenhofer,O.(2017).Fromclimatefinancetowardsustainabledevelopmentfinance.WileyInterdisciplinaryReviews:ClimateChange,8(1),e437.Stock,J.H.,&Watson,M.W.(2011).IntroductiontoEconometrics,3rdEdition.Pearson.Urpelainen,J.,Yang,J.,&Liu,D.(2018).PowerSectorReformsandTechnicalPerformance:GoodNewsfromanInstrumentalVariableAnalysis.ReviewofPolicyResearch,35(1),120–152.https://doi.org/10.1111/ropr.12275Ward,W.H.(1967).TheSailingShipEffect.InstituteofPhysicsandThePhysicalSociety.Warner,R.(2007).AppliedStatistics:FromBivariateThroughMultivariateTechniques.Components.https://doi.org/10.1017/CBO9781107415324.004Wooldridge,J.M.(2001).EconometricAnalysisofCrossSectionandPanelData.TheMITPress.41Yim,O.,&Ramdeen,K.T.(2015).HierarchicalClusterAnalysis:ComparisonofThreeLinkageMeasuresandApplicationtoPsychologicalData(Vol.11,Issue1).42APPENDIXA:SUPPORTINGTABLES&FIGURESTableA.1:CountriesinthedatasetCountCountryReg.KeptCountCountryReg.KeptCountCountryReg.Kept1CambodiaEastAsia&Pacific(EAP)149AustraliaOECD092AfghanistanSouthAsia(SAS)12China150Austria093Bangladesh13Indonesia151Belgium094India14LaoPDR152Canada095Maldives0,missingdata5Malaysia153Chile1,movedtoLAC96Nepal16Mongolia154CzechRepublic097Pakistan17Myanmar155Denmark098SriLanka18PapuaNewGuinea156Finland099AngolaSub-SaharanAfrica(SSA)19Philippines157France0100Benin110Singapore158Germany0101BurkinaFaso111SolomonIslands159Greece0102Burundi112Thailand160Hungary0103Cameroon113Vanuatu161Ireland0104CentralAfricanRepublic114Vietnam162Israel0105Chad115ArmeniaEurope&CentralAsia(ECA)163Italy0106Congo,Dem.Rep.116Azerbaijan164Japan0107Congo,Rep.117Belarus165Korea,Rep.0108Côted’Ivoire118Bulgaria0,EUmember66Netherlands0109Eritrea119Croatia0,EUmember67NewZealand0110Ethiopia120Kazakhstan168Norway0111Ghana121KyrgyzRepublic169Poland0112Guinea14322Romania0,EUmember70Portugal0113Kenya123RussianFederation171SlovakRepublic0114Liberia124Serbia172Spain0115Madagascar125Tajikistan173Sweden0116Malawi126Türkiye174Switzerland0117Mali127Turkmenistan175UnitedKingdom0118Mauritania128Ukraine176UnitedStates0119Mozambique129Uzbekistan177AlgeriaTheMiddleEast&NorthAfrica(MENA)1120Niger130ArgentinaLatinAmerica&Caribbean(LAC)178Bahrain1121Nigeria131Bolivia179Egypt,ArabRep.1122Rwanda132Brazil180Iran,IslamicRep.1123Senegal133Colombia181Jordan1124SierraLeone134CostaRica182Kuwait1125Somalia135DominicanRepublic183Lebanon1126SouthAfrica136Ecuador184Morocco1127SouthSudan0,missingdata37ElSalvador185Oman1128Sudan138Guatemala186Qatar1129Tanzania139Haiti187SaudiArabia1130Togo140Honduras188Tunisia1131Uganda141Jamaica189UnitedArabEmirates1132Zambia142Mexico190WestBankandGaza1133Zimbabwe143Nicaragua191Yemen,Rep.144Panama145Paraguay146Peru147Uruguay148Venezuela,RB1Source:ESMAPRISEdataset.Notes:OurcountrysampleexcludesEUmembersandAustralia,Norway,GreatBritain,Japan,Korea,andSwitzerland.4445FigureA.1:Dependentvariablesnegativelyaffectedbyrenewableenergypolicies(distributionbyregions)Source:WDIandauthors’elaborationbasedonthemethodsdescribedinthispaper.Notes:SSA=Sub-SaharanAfrica;EAP=EastAsia&thePacificECA=Europe&CentralAsia;LAC=LatinAmerica&Caribbean;MENA=theMiddleEastandNorthAfrica;SAS=SouthAsia.46FigureA.2:Dependentvariablespositivelyaffectedbyrenewableenergypolicies(distributionbyregions)Source:WDIandauthors’elaborationbasedonthemethodsdescribedinthispaper.Notes:SSA=Sub-SaharanAfrica;EAP=EastAsia&thePacific;ECA=Europe&CentralAsia;LAC=LatinAmerica&Caribbean;MENA=theMiddleEastandNorthAfrica;SAS=SouthAsia.47TableA.2:RISEdiscrepanciesandvariablesremovedfromtheanalysisTypeofdiscrepancyDummyYearDecisionRationale1Dummyandyeardiscrepancy0YearshouldnotbespecifiedbutisspecifiedFavoredtheyearcolumnTheyearcolumnismorespecificinformationthanthedummycolumn.Ifthereisinputforthemorespecificcolumn,thenweassumethatithasbeenverifiedandiscorrect.2Potentialdummyandyeardiscrepancy0Yearshouldbe0,butitisNA,N/A,notapplicable,ormissingFavoredthedummycolumn,treatedyearas“0”Wecannotuseayearifwedonothaveit.3Dummyandyeardiscrepancy1Yearshouldbespecified,butis0TreatedyearasNA(“.”)Treatingyearsas“no”(with0)wouldbeincorrectbecausethereformseemstohavebeenmade.However,withoutayear,wecannotcounttheminapanel.4Dummyandyeardiscrepancy1Yearshouldbespecified,butismissingTreatedyearasNA(“.”)Treatingyearsas“no”wouldbeincorrectbecausethereformwasmadeaccordingtothedummycolumn.However,withoutayear,wecannotcounttheminapanel.5Yearlookssuspicious1YearseemstooearlyNoactionSomeyearsareveryearly,examplesre.2.1.6.yr(1895)orre.6.3.1.3.yr(1923).Wegivethedatasetthebenefitofthedoubt.6DummyandyeardiscrepancyNAYearshouldbeNA,butisspecifiedFavoredtheyearcolumnTheyearcolumngivesmoreinformationthanthedummycolumn.Ifthereisinputforthemorespecificcolumn,thenweassumethatithasbeenverifiedandiscorrect.7PotentialdummyandyeardiscrepancyNAYearshouldbeNA,butis0Favoredthedummycolumn,treatedyearasNA(“.”)Seemsliketheyearcolumnwasgivena“0”becauseitwas“NA”inthedummycolumn.Butwetreatmissinginthedummycolumnas“NA”.So,wefavoredthedummycolumn.Source:ESMAPRISEdatasetandauthors’elaborationbasedondatasetdescribedinthispaper.48TableA.3:RenewableenergypoliciescoveredintheRISEdatasetHeadingsRISEIDOurIDQuestionLegalframeworkforrenewableenergy(LF)1.1.1re_1_1Doesalegalframeworkforrenewableenergydevelopmentexist?1.2.1re_1_2Doesthelegalframeworkallowprivatesectorownershipofrenewableenergygeneration?Planningforrenewableenergyexpansion(PE)2.1.1re_2_1_1Doesanofficialrenewableenergytargetexist?2.1.2re_2_1_2Isthetargetlegallybinding?2.1.3re_2_1_3IstheREtargetlinkedtointernationalcommitments(e.g.,NDCorregionalcommitment)?2.1.4re_2_1_4Isthetargetbasedonatransparentmethodology?2.1.5re_2_1_5Istherearenewableenergyactionplanorstrategytoattainthetarget?2.1.6re_2_1_6Isthereanyprovisionforconsultationwiththepublicontherenewableplan?2.2.1re_2_2_1Isthereanassessmentoftheroleofrenewablesintheelectricitysupply?2.2.2re_2_2_2Isthereatargetforrenewablesinelectricity?2.3.1re_2_3_1Isthereanassessmentoftheneedsforheatingandcoolinginbuildingsandindustryinthecountryandofhowrenewablescancontribute?2.3.2re_2_3_2Isthereaspecifictargetforrenewablesforheatingandcooling?2.4.1re_2_4_1Isthereanassessmentofthepotentialroleforrenewablesintransportincludingsandelectrification?2.4.2re_2_4_2Isthereaspecifictargetforrenewablesintransport?2.5.1re_2_5_1DoestherenewableplanorstrategyestimatetheamountofinvestmentnecessarytomeettheREtarget?2.5.2re_2_5_2Isthereaninstitutionresponsiblefortrackingprogressinrenewableenergydevelopment?2.5.3re_2_5_3Isthereanyperiodicreportingmechanismforrenewableenergyprogress?2.5.4re_2_5_4Isthereamechanismforadjustingtheplanbasedonreportingofrenewableenergydeployment?2.5.5re_2_5_5Iscurrentpolicyenvironmentconducivetorenewableenergydeployment?2.6.1re_2_6_1Isgenerationandtransmissionplanningintegrated?2.6.2re_2_6_2Isplanningfordispatchincludedinthegenerationandtransmissionplan?2.6.3re_2_6_3Isthegenerationplanbasedonaprobabilisticapproach?2.6.4re_2_6_4Doesthecurrenttransmissionplanningconsiderrenewableenergyscale-up?2.7.1re_2_7_1Doesthegovernmentendorseandusethesolar/windresourcemapsanddataapplicabletotheircountrythatareavailablethroughtheGlobalSolarAtlas/GlobalWindAtlas,orhavetheypublishedsomeothersolar/windresourcemapthatconformstobestpracticeinthelastfiveyears?2.7.2re_2_7_2HasthecountrycarriedoutgeospatialplanningorproducedzoningguidancetoinformthecommercialdevelopmentoftheREresource?2.7.3re_2_7_3Hasthegeospatialplanningorzoningguidancebeencarriedoutaccordingtobestpracticebyi)beingundertakenaspartofastrategicenvironmentalandsocialassessmentorequivalentprocess;andii)bymakingtheoutputspublicallyavailable?Incentivesandregulatorysupportforrenewableenergy(IR)3.1.1re_3_1_1DoesthecountryofferlongtermPPA’sforrenewableelectricityproductionforlargescaleproducers(e.g.via.Feed-in-tariffs,PPA’sawardedthroughauctionsetc.)3.1.2re_3_1_2DoesthecountryofferlongtermPPA’sforrenewableelectricityproductionforsmallscaleproducers(e.g.via.Feed-in-tariffs,PPA’sawardedthroughauctionsetc.)3.1.3re_3_1_3DoesthegovernmentpublishclearandpracticalguidanceonwhatpermissionsarerequiredtodevelopaREelectricityproject?3.1.4re_3_1_4Doesthegovernmentofferotherdirectfiscalincentivesforrenewableelectricity(e.g.capitalsubsidies,grantsorrebates,investmenttaxcredits,taxreductions,productiontaxcredits,FITsforlargeproducers?)3.2.1re_3_2_1DoesthecountryprovideprioritizedaccesstothegridforRE?3.2.2re_3_2_2DoREprojectsreceivepriorityindispatch?49HeadingsRISEIDOurIDQuestion3.2.3re_3_2_3Arethereprovisionstocompensatesellerifofftakeinfrastructureisnotbuiltintime?3.2.4re_3_2_4AretheremechanismstocompensateREprojectsforlostgenerationduetocertaincurtailmentsafterprojectcommissioning?3.2.5re_3_2_5Isthecompensationduebecauseofcurtailmentactuallygivenout.3.3.1re_3_3_1Isthereabiofuelsblendingmandateorotherobligationtousebiofuels?3.3.2re_3_3_2Aretheresustainabilitycriteriawhichbiofuelswhichcontributetothemandatemustmeet?3.3.3re_3_3_3Ifthereisaplanforproducingbiofuelsinthecountry,hasthisincludedanassessmentofsustainabilityimpacts(e.g.againsttheGBEPSustainabilityindicators)includinganassessmentofimpactsonfoodsecurity.3.3.4re_3_3_4Isthereatleastoneschemetoencourageuseofelectric/hybridvehicles?(e.g.Taxbenefittoconsumersandmanufacturers,etc.)3.4.1re_3_4_1Arethereanypoliciestoencouragedeploymentofanyrenewableenergyheatingandcoolingtechnologies?3.4.2re_3_4_2Aretherespecificmeasures(financialsupportorpromotion)designedtoencouragetheuseofrenewablesintheheatingandcoolingsectors?3.4.3re_3_4_3Areopportunitiesforrenewableheatpromotedalongsideenergyefficiencymeasuresinbuildingsand/orindustry?Attributesoffinancialandregulatoryincentives(AI)4.1.1re_4_1_1IscompetitionusedtoensurelargescaleREgeneration(projects>10MW)iscostcompetitive(e.g.throughauctionsforPPA’s)?4.1.1.1re_4_1_2_1Isthereascheduleforfuturebids/auctionsavailableforinvestors?4.1.1.2re_4_1_2_2Isthereapre-qualificationprocesstoselectbidders?4.1.2.3re_4_1_2_3Aretariffsindexed(inpartorinwhole)toaninternationalcurrencyortoinflation?4.1.1.4re_4_1_2_4Arethereprovisionstoensurefullandtimelyprojectcompletion(e.g.bid-bonds,projectmilestones)4.1.1.5re_4_1_2_5Areprojectsawardedthroughauctions/bidsonline/ontracktobeonlineonstateddate?4.1.1.6re_4_1_2_6Haveauctions/bidsmetstatedtargetforinstallations?4.2.1re_4_2_1Cansmallproducers(residential,commercialrooftopPV,etc)connecttothegrid?4.2.2re_4_2_2Arecontractswithfixedtariffsavailableforsuchproducers?4.2.3re_4_2_3Isthereascheduleorclearrules(e.g.capacitybasedlimits)foradjustingthetarifflevelovertime?4.2.4re_4_2_4Aredifferenttariffsavailablefordifferenttechnologiesandsizesofthegenerationplant?4.2.5re_4_2_5Isthereamechanismtocontrolthecapacitybuiltundereachtariff?4.2.6re_4_2_6Aretariffsindexed(inpartorinwhole)toaninternationalcurrencyortoinflation?Networkconnectionanduse(NC)5.1.1re_5_1_1Doesthecountryhaveagridcodethatclearlyspecifiesconnectionprocedures?5.1.2re_5_1_2Dotheconnectionproceduresmeetinternationalbestpractices?5.1.3re_5_1_3Doesthegridcodeincludemeasuresorstandardsaddressingvariablerenewableenergy?5.1.4re_5_1_4Arethererulesdefiningtheallocationofconnectioncosts?5.1.5re_5_1_5Isthetypeoftheconnectioncostallocationpolicyconsideredshallow(gridoperatorpaysforconnectioncosts)?5.2.1re_5_2_1Arethererulesthatallowelectricitycustomerstopurchasepowerdirectlyfromathirdparty(i.e.anentityotherthanthedesignatedutilityinaservicearea)?5.2.2re_5_2_2Dotherulesdefinethesizeandallocationofcostsforuseofthetransmissionanddistributionsystem(e.g.wheelingcharges,locationalpricing?)5.3.1re_5_3_1Doesthecountrycarryoutregularassessmentsoftheflexibilityoftheelectricitygridandtheissuesrelatingtorenewablesintegration?5.3.2re_5_3_2Canrenewableenergyprojectssellintobalancing/ancillaryservices?5.3.3re_5_3_3Arethererulesforexchangingpowerbetweenbalancingareasthatpenalizevariablerenewableenergy,e.g.throughimbalancepenalties?(onlyscoredincountrieswithmultiplebalancingareas)5.3.4re_5_3_4Arethereprovisionsinthepowerexchangerulesthatallowforplantforecasting?(onlyscoredincountrieswithmultiplebalancingareas)5.3.5re_5_3_5DoesthecountryintegratehighqualityforecastingforanyvariableREresources(eitherthroughsubscriptionserviceorprovidedbynationalagencies)intotheirdispatchoperations?50HeadingsRISEIDOurIDQuestion5.3.6re_5_3_6Aredispatchoperationsbeingcarriedoutinrealtime?Counterpartyrisk(CR)6.1.1Arethefollowingfinancialratiosofthecounterpartydeemedcreditworthy?6.1.1.1Currentratio,<1–0inbetween–scale>=1.2–256.1.1.2EBITDAmargin;<0–0inbetween–scale>=15%--256.1.1.3Debtservicecoverageratio;<1–0inbetween–scale>=1.2–256.1.1.4Dayspayableoutstanding;>180–0inbetween–scale<=90–256.2.1re_6_2_1Isthecounterpartyunderwrittenbyagovernmentguaranteeorarethereothermechanismstoensurecreditworthiness(e.g.throughaletterofcredit,escrowaccount,paymentguarantee,orother)?6.2.2re_6_2_2ArestandardPPAsbankable?6.3.1.1re_6_3_1Generation,Arethefinancialstatementsofthelargestutilitypubliclyavailableinthefollowingcategories?6.3.1.2Transmission,Arethefinancialstatementsofthelargestutilitypubliclyavailableinthefollowingcategories?6.3.1.3Distribution,Arethefinancialstatementsofthelargestutilitypubliclyavailableinthefollowingcategories?6.3.1.4Retailsales,Arethefinancialstatementsofthelargestutilitypubliclyavailableinthefollowingcategories?6.3.2.1re_6_3_2Generation,Ifyes,aretheyauditedbyanindependentauditorforthefollowingcategoriesofutilities?6.3.2.2Transmission,Ifyes,aretheyauditedbyanindependentauditorforthefollowingcategoriesofutilities?6.3.2.3Distribution,Ifyes,aretheyauditedbyanindependentauditorforthefollowingcategoriesofutilities?6.3.2.4Retailsales,Ifyes,aretheyauditedbyanindependentauditorforthefollowingcategoriesofutilities?6.3.3.1re_6_3_3Generation–Electricityavailableforsaletoend-users,Arethefollowingmetricspublishedinaprimaryofficialdocument(bytheutility,regulatororministryand/orgovernment)?6.3.3.2Transmission–Transmissionlossrate,Arethefollowingmetricspublishedinaprimaryofficialdocument(bytheutility,regulatororministryand/orgovernment)?6.3.3.3Distribution–Distributionlossrate,Arethefollowingmetricspublishedinaprimaryofficialdocument(bytheutility,regulatororministryand/orgovernment)?6.3.3.4RetailSales–Billcollectionrate,Arethefollowingmetricspublishedinaprimaryofficialdocument(bytheutility,regulatororministryand/orgovernment)?6.3.4re_6_3_4Istheutilityoperatinganincidence/outagerecordingsystem(orSCADA/EMSwithsuchfunctionality)?6.3.5IstheutilitymeasuringtheSAIDIandSAIFIoranyothermeasurementsforservicereliability?6.3.5.1Arethemeasurementsreportedtotheregulatorybody?6.3.5.2Arethemeasurementsavailabletopublic?Carbonpricingandmonitoring(CP)7.1re_7_1Isthereacarbonpricingmechanism(egcarbontax,emissionstradingscheme)implementedinthecountry,coveringpartorallofthecountry’sgreenhousegasemissions?)7.2re_7_2Isthereamonitoring,reportingandverificationsystemforgreenhousegasemissionsinplace?Source:ESMAPRISEdatasetandauthors’elaborationbasedondatasetdescribedinthispaper.Notes:Donotcontaintheyear,cannotbeusedinapanelformat.51TableA.4:AttributesofrawdataCountries133Firstyear1875,SwitzerlandVariablesprimary168Variablescleaned76Policiesdirectlyinheadings/PICs4Policiesingroups,nestedonce66Policiesingroups,nestedtwice6Source:ESMAPRISEdatasetandauthors’elaborationbasedondatasetdescribedinthispaper.52FigureA.3:ComparisonofRISE,Summation,andCompositeIndices,byheading,overtheregion,for2015Source:ESMAPRISEdatasetandauthors’elaborationbasedondatasetdescribedinthispaper.53TableA.5:RISEversusCompositeweightsusedtocreatetheexplanatoryvariablesinthe“type”columnTypeOurIDCompositeindexweightRISEindexweightTypeOurIDCompositeindexweightRISEindexweightLegalframeworkforrenewableenergyre_1_111Networkconnectionandusere_5_1_10.50.2re_1_211re_5_1_20.50.2Planningforrenewableenergyexpansionre_2_1_10.3330.167re_5_1_310.2re_2_1_210.167re_5_1_40.50.2re_2_1_310.167re_5_1_50.50.2re_2_1_410.167re_5_2_110.5re_2_1_50.3330.167re_5_2_210.5re_2_1_610.167re_5_3_110.167re_2_2_110.5re_5_3_210.167re_2_2_20.3330.5re_5_3_310.167re_2_3_110.5re_5_3_410.167re_2_3_210.5re_5_3_510.167re_2_4_10.50.5re_5_3_610.167re_2_4_20.50.5Counterpartyrisk6.1.1re_2_5_110.26.1.1.1re_2_5_210.26.1.1.2re_2_5_310.26.1.1.3re_2_5_410.26.1.1.4re_2_5_510.26.2.110.5re_2_6_110.256.2.210.5re_2_6_210.256.3.1.10.250.03125re_2_6_310.256.3.1.2re_2_6_410.256.3.1.3re_2_7_110.3336.3.1.4re_2_7_20.50.3336.3.2.10.250.03125re_2_7_30.50.3336.3.2.2Incentivesandregulatorysupportforrenewableenergyre_3_1_10.50.256.3.2.3re_3_1_20.50.256.3.2.4re_3_1_310.256.3.3.10.250.03125re_3_1_410.256.3.3.2re_3_2_10.50.26.3.3.3re_3_2_20.50.26.3.3.4re_3_2_310.26.3.40.250.03125re_3_2_40.50.26.3.5re_3_2_50.50.26.3.5.1re_3_3_10.50.256.3.5.2re_3_3_20.50.25re_7_10.50.554TypeOurIDCompositeindexweightRISEindexweightTypeOurIDCompositeindexweightRISEindexweightre_3_3_310.25Carbonpricingandmonitoringre_7_20.50.5re_3_3_410.25re_3_4_10.50.333re_3_4_20.50.333re_3_4_310.333Attributesoffinancialandregulatoryincentivesre_4_1_11Notscoredre_4_1_2_110.167re_4_1_2_210.167re_4_1_2_310.167re_4_1_2_40.3330.167re_4_1_2_50.3330.167re_4_1_2_60.333re_4_2_110.167re_4_2_210.167re_4_2_310.167re_4_2_410.167re_4_2_510.167re_4_2_610.167Source:RISEdatasetandauthors’elaborationbasedonmethodsdescribedinthispaper.Notes:Notscored;Donotcontaintheyear,cannotbeusedinapanelformat.55FigureA.4:ClosenesstomajordonorsthroughUNGAvoting(distributionbyregions)Source:Author’selaborationbasedon(Baileyetal.,2017).56FigureA.5:Closenesswithmajordonorsthroughtrade,1995-2015,%oftotal(distributionbyregions)Source:UNComtradeviatheCEPIIBACIdataset.57TableA.6:TradeagreementsinplacewithEUCommissionCountryYearinplaceCountrykeptinsample(1=yes;0=no)Armenia19991Azerbaijan19991Canada20170Switzerland19800Chile20031Côted’Ivoire20160Comoros20140Colombia20131CostaRica20131DominicanRepublic20081Algeria20051Ecuador20131Egypt,ArabRep.20040Ghana20161Guatemala20131Honduras20131Israel20001Jamaica20081Jordan20021Japan20190Kazakhstan20161Korea,Rep.20150Lebanon20061Morocco20001Madagascar20121Mexico20001Mozambique20161Nicaragua20131Norway19940Peru20131Singapore20191SolomonIslands20201ElSalvador20131Serbia20131Tunisia19981Türkiye19951Ukraine2016158CountryYearinplaceCountrykeptinsample(1=yes;0=no)SouthAfrica20161Zimbabwe20121Source:EUCommissionwebsite.59TableA.7:EligibleS1coefficientsforbasespecificationsIndexIvsMovingaverageSignificant,f>10Significant,f>10,positiveRiseUNGAaff.5784540RiseUNGAaff.3698516CompositeUNGAaff.5713479CompositeUNGAaff.3656471SummationUNGAaff.5703471SummationUNGAaff.3640449SummationEUagreements5262194RiseEUagreements5248180RiseEUagreements3245177CompositeEUagreements5243175SummationEUagreements3242174CompositeEUagreements3240172RiseTradew.donors395331RiseTradew.donors593428CompositeTradew.donors391425SummationTradew.donors590724CompositeTradew.donors588523SummationTradew.donors389721Notes:Tableisorderedbybasedspecificationsthat,inadditiontobeingsignificantatap-valueof0.05withanF-statisticofatleast10,hadthesignwetheorizedfortheIVs.Columnfourindicatesthenumberofcoefficientswithap-valuebelow5%andanf-statisticabove10.Columnfivecontainsthesameinformationfilteredforcoefficientswithapositivesign.60TableA.8:EligibleS2coefficients,bypolicyandoutcomes,lag3.FossilfuelenergyconsumptionElectricityproductionfromoil,gas&coalElectricityproductionfromoilsourcesRenewableenergyconsumptionRenewableelectricityoutputTotalLegalframework1326Planning215210Inc/reg.support511512Attributesoffin/reginc5614319Networkconn.&use6412215Counterpartyrisk5633623Co2price&mon.0Total21206182085Note:Regressionspecification:RISEindex,UNGAaffinityIVwithfiveyearsmovingaverage.61FigureA.6:ScatterplotofS2coefficientsfortheSSAregionshowingthatoutcomesforthesamecountrytendtoclustertogether,lagof5Notes:LF=Legalframework;PE=Planningforexpansion;IR=Incentivesandregulatorysupport;AI=Attributesoffinancialandregulatoryincentives;NC=Networkconnectionanduse;CR=Counterpartyrisk.Toavoidanoverpopulatedgraph,weshowoneregiononly.Regressionspecification:RISEindex,UNGAaffinityIVwithfiveyearsmovingaverage.X-axis:Energypolicypackage.Y-axis:outcomes(colors).Clusteredoutcomesareshownincircles.62TableA.9:Averageeffectofpoliciesacrossregions,bythesecond-stagelagRegion/PolicyLag357Sub-SaharanAfrica-0.071.632.50EuropeandCentralAsia-1.97-0.188.28AllOtherRegions-0.14-0.200.69Notes:DataforregionsexcludingtheSub-SaharanAfricaregionandtheEuropeandCentralAsiaregionwereaggregatedintoAllOtherRegionsduetoasmallnumberofobservationsfortheseregions.Regressionspecification:RISEindex,UNGAaffinityIVwithfiveyearsmovingaverage.63TableA.10:Averageeffectofpoliciesacrossincomecategories,bythesecond-stagelagIncomeCategory/PolicyLag357Low-incomeeconomies0.550.190.24Lower-middle-incomeeconomies-2.671.148.09Upper-middle-incomeeconomies-0.98-2.19-0.70Notes:Regressionspecification:RISEindex,UNGAaffinityIVwithfiveyearsmovingaverage.TableA.11.Shareofsignificantsecond-stagecoefficientsasapercentageoftotaleligiblefirst-stageregressionsbyregions(top)andincomelevels(bottom);replacingsharesof“Renewableenergyconsumption”and“Renewableelectricityoutput”withsharesof“ModernEnergyGenerationoverTotalGeneration”and“ModernEnergyCapacityoverTotalInstalledCapacity”toexcludelargehydropowerGroupTotaleligiblefirst-stageregressionsSharewithsignificantsecond-stagecoefficientsSharewithpositiveandsignificantsecond-stagecoefficientsRegionSSA22718%13%EAP9312%0%ECA4053%35%LAC966%6%MENA4924%22%SAS3238%31%Total53719%13%IncomeHigh360%0UpperMiddle14217%11%LowerMiddle25225%16%Low10715%14%Total53719%13%Notes:Regressionspecification:RISEindex,UNGAaffinityIVwithfiveyearsmovingaverage.64FigureA.7:ScatterplotofS2coefficientsfortheSSAregion,showingthatoutcomesforthesamecountrytendtoclustertogether;replacingsharesof“Renewableenergyconsumption”and“Renewableelectricityoutput”withsharesof“ModernEnergyGenerationoverTotalGeneration”and“ModernEnergyCapacityoverTotalInstalledCapacity”toexcludelargehydropower,lagof5Notes:LF=Legalframework;PE=Planningforexpansion;IR=Incentivesandregulatorysupport;AI=Attributesoffinancialandregulatoryincentives;NC=Networkconnectionanduse;CR=Counterpartyrisk.Toavoidanoverpopulatedgraph,weshowoneregiononly.Regressionspecification:RISEindex,UNGAaffinityIVwithfiveyearsmovingaverage.X-axis:energypolicypackage.Y-axis:outcomes(colors).Clusteredoutcomesareshownincircles.65FigureA.8.Boxplotsecond-stagecoefficientsbyenergypolicypackages,bythesecondstagelag;replacingsharesof“Renewableenergyconsumption”and“Renewableelectricityoutput”withsharesof“ModernEnergyGenerationoverTotalGeneration”and“ModernEnergyCapacityoverTotalInstalledCapacity”toexcludelargehydropowerNotes:LF=Legalframework;PE=Planningforexpansion;IR=Incentivesandregulatorysupport;AI=Attributesoffinancialandregulatoryincentives;NC=Networkconnectionanduse;CR=Counterpartyrisk.Toavoidanoverpopulatedgraph,weshowoneregiononly.Regressionspecification:RISEindex,UNGAaffinityIVwithfiveyearsmovingaverage.X-axis:energypolicypackage.Y-axis:outcomes(colors).Clusteredoutcomesareshownincircles.Tocorrectlyviewthegraph,theaxeswerecutat-40,leavingoutonedatapointat-60forPElagof5.Intheboxplot,thereisaboxfromthefirstquartiletothethirdquartile,withthe2ndquarter(50%percentile)markedbytheinternallineofthebox.Thewhiskersextendingfromtheboxesgofromeachquartiletotheminimumormaximum,outliersincluded.66TableA.12.Averageeffectofenergypolicypackages,bythesecond-stagelag;replacingsharesof“Renewableenergyconsumption”and“Renewableelectricityoutput”withsharesof“ModernEnergyGenerationoverTotalGeneration”and“ModernEnergyCapacityoverTotalInstalledCapacity”toexcludelargehydropowerEnergypolicypackage357LegalFramework0.860.930.43Planningforexpansion-2.13-1.492.88Incentivesandregulatorysupport-0.180.84-1.51Attributesoffinancialandregulatoryincentives0.794.079.18Networkconnectionanduse1.702.191.35CounterpartyRisk2.531.661.11Carbonpricingandmonitoring0.09Notes:Regressionspecification:RISEindex,UNGAaffinityIVwithfiveyearsmovingaverage.TableA.13.Averageeffectofallenergypolicypackagestogetherontwoabsoluteenergyoutcomes,bythesecond-stagelagEnergyoutcome357Absolutemodernrenewablecapacity0.210.340.49Absolutemodernrenewablegeneration0.551.071.48Notes:Regressionspecification:RISEindex,UNGAaffinityIVwithfiveyearsmovingaverage.APPENDIXB:DENDROGRAMOFPOLICYDATASETDendrogramsarewidelyusedtofindhomogeneousgroupsinobservations,orinourcase,policyinstruments,thatdifferfromeachother.Theycanhelptheresearchertoidentifythestructureofthedataortogroupvariablesbasedontheirsimilarity.TheresultsofthedendrogramleadustogrouppolicieswithintheHeadingsthatwereprovidedtousbyESMAP.Todepicthowthepoliciesrelatetooneanother,wefirstcreateadissimilarity(1-similarity)matrixbasedontheJaccardcoefficient.Inbrief,theJaccardcoefficient(Eq.B.1)is67theproportionofoccurrencesinwhichbothvariables(policies)takeavalueofoneinthepaneldataset,overtheoccurrenceofallothercombinations,exceptbothvariablestakingavalueofzero.Jaccardcoefficient=a/(a+b+c)Jaccardcoefficient=a/(a+b+c)Eq.B.1TableAppendixB.1.VariablesintheJaccardcoefficient.Var1,1Var1,0Var2,1abVar2,0cdSource:(Jaccard,1908).Wethensystematicallymergesimilarpoliciesintogroups,creatinganagglomerativehierarchicalclusteredvisualization.Intheresultingdendrogram,eachpolicyisplacedalongthey-axisandisconnectedtootherpoliciesviaahorizontallinethatendsattheircorrespondingsimilarityvalue.Theshorterthelines,themoresimilarthepolicies.Theshapeofdendrogramschangesaccordingtothemethodoflinkinggroups.Themethodspertinenttobinarydataaresingle,complete,andaveragelinkages.Eachmethodhasitslimitations.Thesinglelinkagemayproduce“chaining”,inwhichseveralclustersarejoinedbecauseoneoftheircasesiswithinproximityofacasefromaseparatecluster.However,incompletelinkage,outlyingcasespreventcloseclustersfrommerging.Wechoosethethirdmethod,averagelinkages,as,intheory,itprovidesacompromisebetweensingleandcompletelinkage(Greenacre&Primicerio,2013;Yim&Ramdeen,2015).InFigureB.1,thethreeverticaldashedredlinesallowthereadertocomparedissimilaritiesvisually.Inourcase,thedendrogramhelpsfurthersupporttheideathatpolicieswithinHeadingsaremostsimilartoeachotherandnotmostsimilartothoseinotherHeadings.68Overall,itsupportsourtheoreticalrationaletooperationalizetheindependentvariablesusingthestructurethatwaspre-determinedbythedatasetweacquired.FigureB.1:DendrogramclustervisualizationofpoliciesintheRISEdataset.Source:RISEdatasetandauthors’elaborationbasedonmethodsdescribedinthischapter.69APPENDIXC:CORRELATIONANALYSISOurfirstoptionincorrelationanalysisistoapplya"survival"method,whichkeepsonlyuncorrelatedvariables.Westarttheexerciseatthemostdisaggregatedgrouplevelandthenpitany'surviving'variablesagainstremainingvariablesathigheraggregationlevels,firstthemoreaggregategroup,stoppingthentheheadinglevel.Weusethe𝜙statistic,whichissuitablefortrulydichotomousvariables.ItisequivalenttoPearson's𝜌=√𝜒2/𝑁andina2x2contingencytable,theoutputisequivalenttoCramer'sV,Spearman's𝜌,andPearson'scorrelation(Warner,2007).Inthe"survival"method,weassumethatkeepingoneoftwohighlycorrelatedvariablesretainsenoughinformationtorepresentboth.Wealsoassumetransitivity.Inotherwords,weassumethatpittingsurvivingsub-groupvariablesagainstvariablesinhigherlevelsofaggregationwouldgivesimilaroutcomesthandoingthesameanalysiswiththeonesthatwereremovedfromthepool.Thisallowsustocomparepairsandkeeponlyonevariableofthepairwhenthecorrelationisaboveapredeterminedthreshold.Analternativeistousegroupaverages.Wefirstobtain𝜙statisticsforallpairswithinaheadingandcomputetheaverage𝜙statisticforthatheadingifthepvalue<0.05.Whentheaverage𝜙statisticishigherthanapredeterminedcutoff,thenwekeeptheaverageoftheheadingasanewvariablethatsummarizestheheadingandmoveontothenextheading.However,ifitisnot,werepeattheexercisebygroupandsub-group.Additionally,wemustchoosethe𝜙statisticcut-offpointforwhatshouldbeconsidered"high"enoughcorrelations,buttheparametertouseisnotimmediatelyevident.Thetrade-offhereisbetweenretaininginformationorproducingamanageablenumberofindependentvariables,eachneedingseparateregressions.70Weruntheanalysisatseveralcut-offlevelsforbothmethodstodiscoverwhetherthereisanembeddedtippingpointinthedatathatminimizescuttingdataandmaximizesinformation(FigureC.1).FigureC.1suggeststhehigherthecut-off,themorevariableswekeep,whichisexpected.However,theredoesnotseemtobeaunilaterallyoptimalcut-offpointattheaggregatelevel.Thiswouldhaveoccurredifeachincreasingthecut-offpointdidnotsignificantlyaltertheaggregatenumberofvariablesretainedbytheanalysis.Cut-offSurvivalmethodAveragemethod0.528230.646290.755290.863300.96630FigureC.1:Numberofvariables/groups/headingsremaining(y-axis),bycut-offpoint(x-axis),survivalmethod(dashed),averagemethod(solid)Source:ESMAPRISEdatasetandauthors'elaborationbasedondatasetdescribedinthispaper.Thesurvivalmethodretainsfewerindependentvariablesthantheaveragemethod.Whileaveragingretainsmoreinformationthanthesurvivalmethod,itmayobscureunderlyingdisparitiesintheaveragedvariables.Specifically,usinggroupaveragesfailstoidentifypoliciesthataredifferentfromothersintheirgroup.Bothmethodsmaymaskandcompounderrors.Atthep=0.05level,fiveoutofevery100correlationsfailtorejectthenullhypothesis.Ifthesevariablesarecarriedintothesubsequentrounds,thenthepossibilityofrejectingthenullhypothesisiscarriedwiththem.Themostglaringshortcomingtousingonlycorrelationmethodsisthatthedimensionalityisnotconsiderablyreduced.Usingtheminimalcut-offof0.5,weareleftwithatleast23variables.Therefore,thesecondstep,summation,createsourfinal"Compositeindex"thatiscomparabletothedefaultRISEandtheSummationindexdescribedinthetext.0102030405060700.450.50.550.60.650.70.750.80.850.9

1、当您付费下载文档后，您只拥有了使用权限，并不意味着购买了版权，文档只能用于自身使用，不得用于其他商业用途（如 [转卖]进行直接盈利或[编辑后售卖]进行间接盈利）。
2、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。
3、如文档内容存在违规，或者侵犯商业秘密、侵犯著作权等，请点击“违规举报”。

碎片内容

碳中和的最新文档