Poverty-Growth-Inequality Triangle under Globalizaon: Time ...

Poverty-Growth-InequalityTriangleunderGlobaliza9on: TimeDimensionsandtheControlFactors

oftheImpactsofIntegra9on

YumekaHIRANOandShigeruOTSUBOGraduateSchoolofInterna>onalDevelopment

NagoyaUniversity

The13thInterna>onalConven>onoftheEastAsianEconomicAssocia>oninSingapore

20October,2012

1

ThisresearchisfundedbytheJapanSocietyforthePromo>onofScience(JSPS)ResearchFunds(KibanA-22252005andChallengingExploratoryResearch23653066).YumekaHiranoisaJSPSDoctoralResearchFellow.ShigeruOtsuboisaprofessorandthedirectorfortheEconomicDevelopmentPolicy&ManagementProgramattheGSID,NagoyaUniversity.

Contents

1.  Introduc>on:Ques>onstobeaddressed2.  LiteratureSurvey3.  MethodsandEmpiricalModels4.  Data:Whatdoesitshow?5.  RegressionResults6.  ConcludingRemarks

2

1.Introduc>on:P-G-IRela>onshipunderGlobaliza>on

Trade-off?

Development GovernanceInstitutions

ElasticityofPovertyReduction 'Pro-Poor' ElasticityofPovertyReduction

w.r.t.G rowth w.r.t.Distribution

GrowthIncreasesin

M eanIncom eLevel

InequalityD istributionofIncom e

D istributionofAssets

PovertyReductionin

AbsolutePoverty

GlobalizationTradeIntegration

FinancialIntegration

H RIntegration

Globalization

Source:Otsubo,Shigeru(2009),LeadingIssuesinDevelopmentwithGlobaliza8on,Fig.1-2,p58.3

1.Introduc>on:Ques>onstobeaddressed

•  Growthisgoodforthepoor(Kraay,2002)•  Inequalityisbadforthepoor(Ravallion,2005)•  Thesocio-economicstructure,ins>tu>ons,culture,policies,and

othercountryspecificfactorsarethedeterminingfactorsoftherela>onshipsamonggrowth,inequality,andpovertyreduc>on.

•  Theelas>cityofpovertyreduc>onwithrespecttoeconomicgrowth(growtheffect)isonlylessthanhalfoftheelas>cityofpovertyreduc>onwithrespecttochangesinincomedistribu>on(distribu>oneffect)(Bourguignon,2003;Ravallion,2005).

4

Whethergrowthstrategyunderglobaliza9on(trade,financeandinvestment,andlabormigra9on),whichisconsideredtoexpanddomes9cdisparity,indeedachievespovertyreduc9onornot?

StocktakingofCross-CountryEmpiricalAnalysis

1.1Integra>onisgoodforgrowth?

5

•  (lnyct–lnyc0)/T=α+βlnyc0+λt’+εct’[1]•  (lnyct–lnyc0)/T=α+βlnyc0+γlnZct+λt’+εct’[2]

yc0:theini>allevelofpercapitaincomeforcountrycyct:thelevelofincomeatyeartforthesamecountrycZct:condi>ons/factorsforincomeconvergenceatyeartforthesamecountrycλt’:>me-specificfixedeffects.εct’:arandomerrorterm.

Globaliza>on(economicintegra>on)asacrucialfactor,togetherwithpolicystancesandins>tu>onalquali>es,forthecondi>onalβ-convergence

6

Source:Authors’compila>on

1.1Integra>onisgoodforgrowth?

1.2IsTradeIntegra>onDistribu>onNeutral?

7Source:Authors’compila>onusinggrowthspellsof5-9years.

1.2Isintegra>onneutraltoincomedistribu>on?

l  DollarandKraay(2004):² Nosignificantcorrela>oncanbeobservedbetweenthechangesinthe

trade/GDPra>os(orchangesintariffrateandcapitalregula>ons)andthechangesintheGini,andthatnotonlytheeconomicgrowthbutalsotradeintegra>onis“onaverage”neutraltoincomedistribu>on.

l Milanovic(2005):² Tradeintegra>ongivesanega>veimpactespeciallyamongthemiddle

andlowestclasshouseholdsinpoorna>on(althoughposi>veeffectwasobservedinhighincomefamiliesofhigh-incomecountries).

²  Investmentintegra>onislikelytobenefitonlythetopincomeclass,althoughitisnotyetsta>s>callysignificant.

8

1.3Whataretheappropriate>medimensionsinanalyzingtheroleofins>tu>ons?

•  Previousstudiesfocusedonthequalityofeconomicpolicies(trade,fiscalandmonetarypolicies,etc.)

•  Thispaperpaysmoreaven>ontoins>tu>onalquality/capacityasthecontrolfactorsoftheimpactsofintegra>on.

•  Ins>tu>onhasdifferentnature,whichtakeslonger>metogiveitseffectsthanpolicychanges(Williamson,2000).

•  Qualityofins>tu>onsshouldbeoneofthekeydeterminantsforeconomicgrowthandpovertyreduc>oninthelongerrun.

9

3.MethodsandEmpiricalModels

1)Globaliza>onandGrowth:Doesintegra>onpromotegrowthandincomeconvergence?(lnyct–lnyc0)/T=α+βlnyc0+γlnGinic0+δlnZct+θlnGct+λt’+εct’[3]　　　

2)Globaliza>onandInequality:Isintegra>onneutraltoincomedistribu>on?a)Es>mate1)forthegrowthratesofaverageincomeofthepoorestquin>le(andtestΔθ)(lnypct–lnypc0)/T=α+βlnypc0+γ(lnyct–lnyc0)/T+δlnGinic0+η lnZct+θlnGct+λt’+εct’[4]b)Es>matelevelrela>onshipalaDollarandKraay(2004,Eq.3).lnypct=α+βlnyct+δlnZct+θlnGct+μc+λt+εct[5]Countryfixed-effectses>ma>onsareconducted(i.e.es>ma>ngμc).c)Percapitaincomein1)isreplacedbymeasurementsofinequality(Gini,5thquin>leincomeshare)

(lnGinict–lnGinic0)/T=α+βlnGinic0+γ[(lnyct–lnyc0)/T]+δlnZct+θlnGct+λt’+εct’　[6]3)PoveryandGrowth/Inequality:Changesinpovertyindicatorsareregressedongrowthand(levelsof/changesin)inequalitymeasurementswithtwosetsofcontrolvariablesZandG.(lnPovct–lnPovc0)/T=α+βlnPovc0+γ[(lnyct–lnyc0)/T]+δlnGinic0+ε[(lnGinict–lnGinic0)/T]+ηlnZct+θlnGct+λt’+εct’[7]

10

AseriesofpanelregressionanalysesingrowthspellsformajorsegmentsoftheP-G-Itriangleunderglobaliza>on

11

Lessons from the Cross-country Estimates of the Kuznets Curve

Source: Bourguignon 2004, Figure 5

Ifthemodelstartsfromagrowthspellequa>onwithanassump>onofcountry-specificeffectsingrowths/changes,howwouldthees>matedcoefficients(sensi>vi>es)behave?

3.1FixedEffects

3.2Instruments

•  Thecurrentstudyusessimpleandstandardizedinstrumentsoflaggedvaluesforthelevels(includinglaggedperiodaveragesforperiodaverages),andini>alvalueandlaggedrateofchangesforgrowthspells.

•  Explanatoryvariablesareinstrumentedonlywhenitseemslogicaltomobilizeinstruments.Ingrowthregressions(Eq.[3]),mostvariables,exceptforini>alvalues,areinstrumented.

•  Inlevelregressionsforthepoorestquin>leincome(Eq.[5]),onlyna>onalaveragepercapitaincomeisinstrumented.

•  Similarly,onlyincomegrowthspellsareinstrumentedintheGiniconvergenceregressions(Eq.[6]).Inthepovertychange(P-G-I)regressions(Eq.[7]),onlyGinichangesareinstrumented.

12

3.3TimeDimensions

Impactsindifferent>medimensions1)  Short-term(1to4years)2)  Medium-term(5to9years)3)  Long-term(overtenyears)Inordertoavoidmixingupkinksandshi}singrowthchangesineachspell,astheproblemspointedoutbyPritchev(2000)andetal.

13

Figure4:KinksinDecadalGrowthPerformance

Source:Authors’compila>onbasedonWDI

4.Data:Whatdoesitshow?•  Thegrowthspellsoftheaforemen>onedthreedifferent>medura>ons

arecompiledfromtheavailabledatapointsofbovom-quin>leincomesharesduringtheperiodfrom1978to2010.

•  MostofthedatasetcomesfromWDI,exceptHDIindexfromUNDP,Ins>tu>onIndexfromICRGandFreedomHouse.

14

AverageYear No.ofCountries

No.ofGrowthSpells

Period

ShortTerm 1.75 79 561 1981-2010

MediumTerm 5.72 99 245 1978-2010

LongTerm 11.17 86 116 1979-2010

•  Highergrowthperformanceinthedecadaldummyforthe2000s.(Eq.[2]).•  Incomeconvergence,bothamongeachregionandamongeachincomegroup,is

detectedacrossgrowthspellsofdifferent>medura>ons,withhighersignificancefoundinthelonger-termdatabase.(Eqs.[3]&[4])

•  Crisesdummiesfoundanega>veandsignificantcoefficientintheshortspells,butfoundposi>veandsignificantcoefficientsinthelongerspells.Countriesexperiencingcrises(DebtCrisis:1982-83,AsianFinancialCrisis:1997-98,andtheWorldFinancialCrsis:2008-09)tendtoovercometheshort-termnega>veimpactwithreformsandcounterpoliciesandenjoyevenhigherrateofgrowthsasendresults.(Eq.[6])

•  Surprisingly,landlocked-countrydummiesfoundposi>ve(andsignificantinshortspells)coefficients.(Eq.[7])

•  Posi>veandsignificantcoefficientswerefoundfortransi>onaleconomiesandtheFormerSovietUnion(FSU)dummies,acrossgrowthspellsofthethree>medura>ons.(Eqs.[8]&[9]).

•  Ini>alinequalitytendstounderminesubsequentgrowthperformance(astotaleffects).(Eq.[10]-[11])

15

4.Data:Whatdoesitshow?(Table1-3)

16

FromTable4Cross-correla>ons•  thecorrela>onbetweenrateofchangesininequality(GINI)andthatin

povertyheadcountra>o(HCR)istwiceashighinthelongerspells(0.46)ascomparedtothatinthemedium-termspells(0.23).

•  Whilethegrowthintheaverageincomeofthebovomquin>leishighlycorrelatedwithoverallGINI,therateofchangesinthena>onalaverageincomeislargelyneutraltoincomedistribu>oninthesegrowthspells.

17

4.Data:Whatdoesitshow?

5.RegressionResults

PleaseRefertotheTablesofRegressionResults.

18

5.1GrowthRegressions(Block1)Table5:•  Asignificantposi>veimpactongrowthofFDIs(theperiodaverage

ra>osofFDItoGDP)isdetected.•  Remivances(thatrepresentsmigrantworkers)havenega>ve

rela>onswith(par>aleffectson)growthperformancewithlesstomoderatesignificance.

•  Theresultsforthetradeintegra>onaremixed.Asasingleexplanatoryvariable,theini>alvalueoftradetoGDPra>ohasalesstomoderatelysignificantposi>vecoefficient.Inasetofpolicy,ins>tu>onvariablesandtheFDI,theaveragerateofchangeintradetoGDPra>ohasamoderatelysignificantposi>vecoefficient.

•  Otherresultsarelargelyinsignificantandsome>mescounterintui>ve.

19

20

5.1GrowthRegressions(Block1)Table6-7:•  Althoughtheresultsobtainedfromthevola>leshort-termgrowth

spellsaresomewhatunstable,itisgenerallytruethattheimpactofmacroeconomicpolicymanagementismoresignificantlyvisibleinshorterspellsascomparedtothoseinthelongerspells.

•  Whilethesizeofthegovernmentisgenerallynega>velyassociatedwithgrowthperformanceintheshortterm,thateffectbecomesnegligibleinthelongerterm.

•  Theimportanceofinfla>oncontrol,i.e.,thepreserva>onofthevalueofcurrency,persiststhroughoutthe>medura>ons.

•  Thesignificanceoftheimpactofins>tu>onalqualitytendstoincreaseinthelonger>mespans.Italsoincreasesrela>vetothoseofmacroeconomicpolicymanagement/stancevariables.Ittakeslonger>metoobservetheposi>veimpactofhigherins>tu>onalquality,justlikeittakeslonger>metobuildit.

21

22

Table8•  Tradetendstobepro-poorwithlesssignificance(Column5),while

FDItendstobean>-poorwithlesstomoderatesignificance.•  Ins>tu>onalqualityasmeasuredbyICRGcompositeandpoli>cal

situa>onindicatorstunedouttobesignificantlyan>-poorinarela>vesense.Thoseindicatorsofins>tu>onalquality(orgovernance)aredevelopedwithins>tu>onalinvestorsinmind,andthuslargelypro-investment,pro-growth,andsupply-sideoriented.

•  Whiledemocracyandsecurityhaveposi>vees>matedcoefficientswithlesssignificance,nega>veandsignificantcoefficientisavachedtoacapacityindicatorthatmeasuresacountry’soverallmanagementcapacity.

•  Thisresultalsoindicatestheneedofdevelopingpro-equality(thus,pro-poor)ins>tu>onalqualityindicatorsfortheanalysesaimedatiden>fyingfactorsformoreequitable,pro-poorgrowth.

23

5.2InequalityRegressions(Block2)

24

5.2InequalityRegressions(Block2)

Table9:es>matesofthecross-countryfixedeffects(thatshouldadduptozero)areobtainedusingthefivedequa>oninthecolumn(12)inTable8.•  Ethiopia,Moldova,Zambia,Ukraine,Pakistan,Belarus,

Vietnam,KyrgyzRepublic,BurkinaFaso,andIndiaareiden>fiedasthetoptencountriesofpro-poorcountryfixedeffects.

•  Brazil,Columbia,Bolivia,Panama,Argen>ne,SouthAfrica,Venezuela,Guatemala,Honduras,andParaguayareiden>fiedasthecountrieswithleastpro-poorfixedeffects.

•  Thisiden>fica>onofpro-poorfixedeffects,however,dependsheavilyontheselec>onofperiodsinthesetofobserva>ons.

25

26

5.2InequalityRegressions(Block2)Table10•  ThereisasignificantabsoluteconvergenceinGiniacrosscountries(Column1).As

such,nopar>cularrela>onshipbetweenincomeandinequalityisnecessitated.GiventhecountryspecificeffectspostulatedinEq.[5],thesefixedeffectsdropoutintheregressionswithgrowthspells(Eq.[6]).

•  AninsignificantcoefficientfoundinColumn(2)andmixed(sign)andinsignificantcoefficientsavachedtoincomegrowthterminthecolumnsofTable9signifythisfact.

•  Moderatelysignificantandnega>vecoefficientsavachedtoincomegrowthimplythat,whenthesegmentofincomegrowthcausedbyfacesofintegra>onisseparatelyaccountedfor,growthpropelledbydomes>creasonstendtoreduceinequality.Inotherwords,thisinturnimpliesthatgrowthpropelledbyintegra>ontendstowideninequality.

•  Amongthethreemodali>esofintegra>on,remivances—thatsignifytheexistenceofinterna>onalmigrantworkers—areslightlysignificantandhaveverystablenega>veimpact(pro-equalityimpact)onGini.

•  Higherins>tu>onalqualityasmeasuredinICRGcompositeandpoli>calsitua>onindicatorsexpandsinequality,thoughitpromotesgrowthasseeninBlock1.

27

28

5.3PovertyRegressions(Block3)Table11•  Thereismoderatelysignificantabsoluteconvergenceinpoverty

headcountra>os(HCR)acrosscountries(Column1).•  Thisstudy,withmanygrowthspellstakenfromcountriesin

transi>onincludingtheFSUstates,findshigherelas>cityofpovertyreduc>onforeconomicgrowth(Columns2,3and4).Theresultsofregressionswithpolicyvariables(Column5)indicatethatalargergovernmenttendstobepro-poor,andthatinfla>onisan>-poor.

•  Ins>tu>onalqualityasmeasuredintheinvestor-orientedICRGindicators,whilepromo>ngoveralleconomicgrowth,affectsequality(GINIandbovomquin>leincomerela>vetothecountryaverage)nega>vely.

•  higherins>tu>onalquality(ICRGindicators)reducespoverty(HCR)(Columns6and7).

•  Noneofthevariablesofintegra>onturnsouttobe(strongly)significantintheequa>onspresentedinColumns(8)and(9).

29

5.3PovertyRegressions(Block3)•  Asatendency,ahigherini>alleveloftradeintegra>on

(tradetoGDPra>o)tendstoreducepovertyinasubsequentperiod,withlesstomoderatesignificance.Ahigheraveragerateofchangeinintegra>onthroughtrade,however,tendstoincreasepoverty.

•  AlthoughFDIspromotegrowthrathersignificantly,thesameFDIstendtoexpandinequality(Gini),asreportedearlier.

•  Thecombinedimpactonpovertyreduc>onisthusmixedandsta>s>callyinsignificantduringthisexercise.Althoughtheimpactofremivancesisstablyconducivetopovertyreduc>on,itislargelyinsignificantsta>s>callyamongthesamplecollectedforthisstudy.

30

31

6.ConcludingRemarks•  1)theabsoluteincomeconvergenceineachincomegroup

(withdivergencebetweenincomegroups);•  2)thecondi>onalincomeconvergenceamongcountries

acrossincomegroups;and•  3)theabsolute(and,ofcourse,condi>onal)convergencein

inequality(GINI)andpovertyheadcountra>os(HCR)acrosscountries.

•  However,unlikeearlierstudiesthatfoundhigherinequalityelas>cityofpovertyreduc>on,thisstudy,withmanygrowthspellstakenfromcountriesintransi>onincludingtheFSUstates,findshigherelas>cityofpovertyreduc>onforeconomicgrowth.

32

6.ConcludingRemarks•  rela>velyhighersignificanceof‘ins>tu>ons’withlowersignificanceof

‘policystances’inthelongergrowthspells.•  Amongthethreedimensionsofintegra>ontestedinthisstudy

(interna>onaltrade,FDI,andremivances),highersignificanceexistsin:1)FDI(posi>ve)ingrowth;2)FDI(nega>ve)andtrade(posi>ve)intheincomeofbovomquin>lerela>vetona>onalaverage;and3)remivances(nega>ve)inpoverty(i.e.,reducepovertyheadcountra>o),butnotwithsufficientsta>s>calsignificance(atleastnotyet).

•  Tradeismostlyneutraltoincomedistribu>on(GINI),whileFDIandremivancestendtoexertnega>veandposi>veimpacts,respec>vely.

•  Ins>tu>onalqualityasmeasuredintheinvestor-orientedICRGindicators,whilepromo>ngoveralleconomicgrowth,affectsequality(GINIandbovomquin>leincomerela>vetothecountryaverage)nega>vely.AsthecombinedresultoftheP-G-Itriangle,however,higherins>tu>onalquality(ICRGindicators)reducespoverty(HCR).

33

Thankyouforyouraven>on.

34