CorruptionKills:GlobalEvidencefromNaturalDisastersSerhanCevikandJoãoTovarJallesWP/23/2202023OCT©2023InternationalMonetaryFundWP/23/220IMFWorkingPaperEuropeanDepartmentCorruptionKills:GlobalEvidencefromNaturalDisastersPreparedbySerhanCevikandJoãoTovarJalles1AuthorizedfordistributionbyBernardinAkitobyOctober2023IMFWorkingPapersdescriberesearchinprogressbytheauthor(s)andarepublishedtoelicitcommentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseoftheauthor(s)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.AbstractNaturaldisastersareinevitable,buthumanitarianandeconomiclossesaredeterminedlargelybypolicypreferencesandinstitutionalunderpinningsthatshapethequalityofpublicinfrastructure(includingemergencyresponsesandhealthcareservices)andgovernbusinesspracticesandtheadherencetobuildingcodes.Inthispaper,weempiricallyinvestigatewhethercorruptionincreasesthelossofhumanlivescausedbynaturaldisasters,usingalargepanelof135countriesduringtheperiod1980–2020.Theeconometricanalysisprovidesconvincingevidencethatcorruptionincreasesthenumberofdisaster-relateddeaths,aftercontrollingforeconomic,demographic,healthcareandinstitutionalfactors.Thatis,thehigherthelevelofcorruptioninagivencountry,thegreaterthenumberoffatalitiesasashareofpopulationduetonaturaldisasters.Ourresultsshowthatthedevastatingimpactofcorruptiononlossofhumanlivescausedbynaturaldisastersissignificantlygreaterindevelopingcountries,whichareevenmorevulnerabletononlineareffectsofcorruption.JELClassificationNumbers:D31;D73;H41;P16;Q54Keywords:Corruption;institutions;naturaldisasters;fatalitiesAuthor’sE-MailAddress:scevik@imf.org;joaojalles@gmail.com1TheauthorswouldliketothankAzarSultanovforhelpfulcommentsandsuggestions.I.INTRODUCTIONNaturaldisastersareinevitable,resultinginsignificanteconomiclossesandtensofthousandsofdeathsinmostyearsacrosstheworld.Inhighfatalityyears,whichtendtobethosewithmajorearthquakesorcyclones,thenumberofdeathscausedbynaturaldisastersmayreachhundredsofthousands(Figure1).2Overthecourseofmodernhistory,therehasbeenacontinuousreductioninthenumberoffatalitiescausedbynaturaldisastersowingtobetterlivingstandards,moreresilientphysicalinfrastructure,betterearlywarningindicatorsandstrongeremergencyresponsesystems(Figure2).However,therearestillimportantdisparitiesacrosscountriesinhumanitarianandeconomiclosses.Forexample,anearthquakemeasuring7ontheRichterscaledevastatedHaitiandkilledmorethan200,000peoplein2010,whileearthquakesofsimilarmagnitude(7.2ontheRichterscale)causedonlyminorfracturesandinjuriesinMexicoandNewZealand.Couldgeographicandsocioeconomicfactorsaloneexplainsuchastrikingdifferenceindisasteroutcomes?Wethinknot.Theimpactofnaturaldisasters,inourview,isalsoattributabletopolicypreferencesandinstitutionalunderpinningsthatdeterminethequalityofpublicinfrastructure,theeffectivenessofemergencyresponsesandhealthcareservicesandgovernbusinesspracticesandtheadherencetobuildingcodes.Thisisnotthefirstattemptintheliteraturetoanalyzeeconomic,institutionalandsocialfactorsindetermininglossesassociatedwithnaturaldisasters(Albala-Bertrand,1993;TolandLeek,1999;Haque,2003;Anbarcietal.,2005;Kahn,2005;SkidmoreandToya,2007;KellenbergandMobarak,2008;Raschky,2008;Noy,2009;PadliandHabibullah,2009;SchumacherandStrobl,2011;Loayzaetal.,2012;Cavalloetal.,2013;Klomp,2016;Taghizadeh-Hesaryetal.,2019).Corruption—commonlydefinedastheabuseofentrustedpowerforprivategain—isshowntohavedetrimentaleffectsoneconomicdevelopment,socialcohesionandtrust,andpoliticalstabilityandeffectivegovernance(Mauro,1995;Tanzi,1998;Mo,2001;AlesinaandWeder,2002;HabibandZurawicki,2002;PellegriniandGerlagh,2004;MeonandSekkat,2005;Rose-Ackerman,2006;Aidtetal.,2008;Hodgeetal.,2011;D’Agostinoetal.,2016;Huang,2016;ChangandHao,2017;FarzaneganandWitthuhn,2017;CieślikandGoczek,2018;GründlerandPotrafke,2019;Uberti,2022).Mostcloselyrelatedtothispaper,Escalerasetal.(2007)findthatcorruptionispositivelyrelatedtoearthquake-relateddeathsin75countriesovertheperiod1975–2003.ThisreflectsamultitudeofchannelsthroughwhichcorruptiondetermineslossesassociatedwithFigure1.NaturalDisaster-RelatedDeathsAcrosstheWorldSource:OurWorldinDatabasedontheEM-DATdatabase.2OurWorldinDataprovidesaconcisepresentationofdisastersbasedontheEM-DATdatabase,whichisusedinthispaper:https://ourworldindata.org/century-disaster-deaths.4naturaldisasters:(i)inadequateinfrastructure,weakbuildingcodesandunsafeconstruction(Brinkerhoff,2008;AlamandVennemo,2014;Iqbal,2018);(ii)slowandinefficientemergencyresponses,reliefdistributionandhealthcare(Akhtaruzzaman,2011;KlompanddeHaan,2013);(iii)disproportionateimpactonvulnerablegroupsofthesocietyduetoinequalitiesindisasterpreparednessandresponseandaccesstoinformation,resources,andsupport(Guptaetal.,2002;LehoucqandMolinas,2002);and(iv)lackofaccountabilityandtransparencyingovernance(Shah,2006;Heywood,2007).Inthispaper,weusealargepanelof135countriesduringtheperiod1980–2020andfocusontheroleofcorruptionindeterminingthelossofhumanlivescausedbynaturaldisasters.Corruptionisacomplexphenomenonthataffectsallcountries,butitseconomic,institutional,politicalandsocialcausesandconsequencesshowgreatvariationacrosscountries.Theeconometricanalysisprovidesconvincingevidencethatcorruptionincreasesthenumberofdisaster-relateddeaths,aftercontrollingforeconomic,demographic,healthcareandinstitutionalfactors.Hence,wecaninferthatthehigherthelevelofcorruption,thegreaterthenumberoffatalitiesperpopulationinnaturaldisasters.Toputthisempiricalfindingintoperspective,thedifferencebetweentheleastandmostcorruptcountriesinoursampleimpliesasixfoldincreaseinthenumberofdeathsperpopulationcausedbynaturaldisasterinagivenyear.Ourresultsalsoshowthatthedevastatingimpactofcorruptionisgreaterindevelopingcountriesthaninadvancedeconomiesandtherearenonlineareffectswithhigherlevelsofcorruptionresultinginanevenlargernumberofdeathsfromnaturaldisasters,especiallyindevelopingcountries.Inourview,thisreflectsthelowqualityofbuildingsandinfrastructureandtheweaknessofhealthandriskmanagementsystemsduetowidespreadcorruption.Theseempiricalfindingsarerobusttovariousregressionspecificationsandsampleheterogeneity,whichweusetoobtainagranularanalysisoftheimpactofcorruptiononlossofhumanlivesFigure2.AnnualAverageofNaturalDisaster-RelatedDeathsAcrosstheWorldSource:OurWorldinDatabasedontheEM-DATdatabase.5causedbynaturaldisasters.Allinall,theempiricalresultspresentedinthispaperhighlightthecriticalrelationshipbetweeneconomicdevelopmentandinstitutionalcapacityinstrengtheninggoodgovernance.Promotinganti-corruptionmeasurestostrengtheninstitutionsandcreateaconduciveenvironmentforgreatertransparencyingovernanceandappropriateuseofpublicresourcesis,therefore,paramountinthisregard.Theremainderofthispaperisorganizedasfollows.SectionIIdescribesthedatausedintheempiricalanalysis.SectionIIIintroducesthesalientfeaturesofoureconometricstrategy.SectionIVpresentsanddiscussestheempiricalresults,includingaseriesofrobustnesschecks.Finally,SectionVoffersconcludingremarksincludingpolicyimplications.II.DATAOVERVIEWWeputtogetherapaneldatasetofannualobservationscovering135countriesovertheperiod1980–2020.Thedependentvariableisthenumberofdeathsperpopulationduetonaturaldisastersinagivenyear,whichisobtainedfromtheEmergencyEventsDatabase(EM-DAT)compiledbytheCentreforResearchontheEpidemiologyofDisasters(CRED)attheUniversitéCatholiquedeLouvaininBelgium.TheEM-DATprovidesdataontheoccurrenceandeffectsofover22,000large-scalenaturaldisastersacrosstheworldfrom1900tothepresentdayandoffersinformationondifferentcategoriesofnaturaldisastersincludinggeophysical(earthquake,massmovementandvolcanicactivity),meteorological(extremetemperature,fogandstorm),hydrological(flood,landslideandwaveaction),climatological(drought,glaciallargeoutburst,wildfire),andbiological(epidemic,insectinfestationandanimalaccident).3▪Geophysical:Ahazardoriginatingfromsolidearth.Thistermisusedinterchangeablywiththetermgeologicalhazard.▪Meteorological:Ahazardcausedbyshort-lived,micro-tomeso-scaleextremeweatherandatmosphericconditionsthatlastfromminutestodays.▪Hydrological:Ahazardcausedbytheoccurrence,movement,anddistributionofsurfaceandsubsurfacefreshwaterandsaltwater.▪Climatological:Ahazardcausedbylong-lived,meso-tomacro-scaleatmosphericprocessesrangingfromintra-seasonaltomulti-decadalclimatevariability.▪Biological:Ahazardcausedbytheexposuretolivingorganismsandtoxicsubstances(e.g.venom,mold)orvector-bornediseases.Thenumberofdeathsperpopulationandpereventvariesonaverageaccordingtothetypeofnaturaldisaster(Figure3).Weather-relateddisasterssuchasdroughtsandfloodscausedextremelylargenumberofdeathsinthefirsthalfofthe20thcenturybuthavebecomelessimpactfulintermsofhumanitarianlossesovertime,owingtoimprovementsininfrastructureand3Weexcludeextraterrestrialdisasterscausedbyasteroids,meteoroids,andcometsastheypassnear-earth,entertheearth’satmosphere,and/orstriketheearth,andbychangesininterplanetaryconditionsthateffecttheearth’smagnetosphere,ionosphere,andthermosphere.6emergencymanagementincludinginternationalaid.Naturaldisasterscausedbyclimatechangesuchasextremetemperatureandstorms,ontheotherhand,havebecomemorepronouncedoverthepasthalfcentury.Large-scaleearthquakeshavealwaysresultedinsignificanteconomicandhumanitarianlosses,especiallyincountrieswithweakinstitutionalandphysicalinfrastructure.Themainexplanatoryvariableofinterestiscorruption,whichismeasuredbythecorruptionindexconstructedbytheInternationalCountryRiskGuide(ICRG).Corruptionisdefinedas“theextenttowhichpublicpowerisexercisedforprivategain,includingpettyandgrandformsofcorruption,aswellascaptureofthestatebyelitesandprivateinterests.”Thesurvey-basedcorruptionindexrangesfrom0(highestpotentialrisk)to6(lowestpotentialrisk).4Weinverttheindexsothatthenewvariableisincreasinginthedegreeofcorruption.TheICRGdatasetprovidesthemostcomprehensivecoverageacrosscountriesandovertime,startingin1984.Tomaximizethesamplesizeofnaturaldisasters,weusethe1984valueofthecorruptionindex(aswellasotherinstitutionalvariablesdescribedbelow)fortheperiod1980–1983.Althoughthisisastrongassumption,thecorruptionindexandotherinstitutionalfactorsusedintheanalysisareslow-movingvariables.5Followingtheliterature,weintroduceseveralcontrolvariables,includingrealGDPpercapita,tradeopennessasmeasuredbytheshareofexportsandimportsinGDP,urbanizationasmeasuredbytheshareofurbanpopulationintotal,andthenumberofhospitalbedsperFigure3.GlobalDeathsfromNaturalDisastersAcrosstheWorld,bytypeSource:OurWorldinDatabasedontheEM-DATdatabase.4TheICRGdatabaseisavailableathttps://www.prsgroup.com/explore-our-products/icrg/.5Ourbaselineestimationresultsremainunchangedwhenwealternativelyusetheperiod1984–2020.7population,whicharedrawnfromtheWorldBank´sWorldDevelopmentIndicatorsdatabase.Wealsoincludeadditionalvariablestocontrolforbroaderinstitutionalcharacteristics,whichcouldinfluenceboththelevelofcorruptionandthehumanitariancostofnaturaldisasters.6Specifically,weusecompositeindicesofbureaucraticquality,democraticaccountabilityandgovernmentstability,whichareobtainedfromtheICRGdatabase.Bureaucraticqualitymeasurestheinstitutionalstrengthandthelevelofexpertisetogovernwithoutdrasticchangesinpolicyorinterruptionsingovernmentservices.Countriesthatlackthecushioningeffectofastrongbureaucracyreceivelowpointsbecauseachangeingovernmenttendstobetraumaticintermsofpolicyformulationandday-to-dayadministrativefunctions.Democraticaccountabilityisameasureofhowresponsivegovernmentistoitspeople,onthebasisthatthelessresponsiveitis,themorelikelyitisthatthegovernmentwouldfall,peacefullyinademocraticsociety,butpossiblyviolentlyinanon-democraticone.Ingeneral,thelowestriskpointisassignedtoalternatingdemocracies,whilethehighestriskpointisassignedtoautarchies.Governmentstabilityisanassessmentbothofthegovernment’sabilitytocarryoutitsdeclaredprogramsaswellasitsabilitytostayinoffice.DescriptivestatisticsforthevariablesusedintheempiricalanalysisarepresentedinTable1.Thereisasignificantdegreeofdispersionacrosscountriesinthenumberofdisaster-relatedfatalitiesandconsiderableheterogeneityineconomicandinstitutionalfactorsbetweencountriesandwithinandbetweenincomegroups.Forexample,advancedeconomies,onaverage,havealowerlevelofcorruptionandexperiencefewernaturaldisastersthandevelopingcountries.Correspondingly,thenumberofdeathscausedbynaturaldisastersinadvancedeconomiesissignificantlylowerthanindevelopingcountries.Table1.SummaryStatisticsVariableObservationsMeanStd.dev.MinimumMaximumNaturaldisaster-relateddeathsperpopulation7,9441.331.30.02,331.5Corruption4,7373.01.40.06.0RealGDPpercapita7,12611,02316,130166Tradeopenness6,51283.852.60.0114,048Urbanization7,91353.323.94.3442.6Healthcare(hospitalbeds)3,4154.53.40.1100.0Bureaucraticquality4,8002.21.20.019.9Democraticaccountability4,8003.91.60.04.0Governmentstability4,8007.42.00.06.012.0Source:EM-DAT;ICRG;WorldBank;author'scalculations.III.ECONOMETRICSTRATEGYANDEMPIRICALRESULTSTheobjectiveofthispaperistoempiricallyinvestigatetheroleofcorruptionindisaster-relatedfatalitiesin135countriesovertheperiod1980–2020.Takingadvantageofthepanelstructureinthedata,weestimatethefollowingbaselinereduced-formempiricalspecification:6Theexclusionoftheseinstitutionalvariablesmayleadtoapotentialomittedvariablebias.8𝐷𝑒𝑎𝑡ℎ𝑠𝑖𝑡=𝛽1+𝛽2𝐶𝑜𝑟𝑟𝑢𝑝𝑡𝑖𝑡−1+𝛽3𝑋𝑖𝑡−1+𝜂𝑖+𝜇𝑡+𝜀𝑖𝑡(1)where𝐷𝑒𝑎𝑡ℎ𝑠𝑖𝑡isthelogarithmofthenumberofdeathsperpopulationduetonaturaldisastersincountryiandtimet,whicharewinsorizedat5thand95thpercentilestomitigatetheeffectsofextremeoutliers;𝐶𝑜𝑟𝑟𝑢𝑝𝑡𝑖𝑡isanindicatorofcorruption—themainvariableofinterestinthisanalysismeasuredfromlesstomorecorruptionbyinvertingtheoriginalscale;𝑋𝑖𝑡denotesavectorofcontrolvariablesincludingthelogarithmofrealGDPpercapita,tradeopenness,urbanization,thelogarithmofhospitalbeds,andmeasuresofbureaucraticquality,democraticaccountabilityandgovernmentstability(whicharerescaledforhighervaluestoindicatehigherrisk).Allexplanatoryvariablesincludingcorruptionarelaggedbyoneperiodtoreducepotentialreversecausalityconcerns.The𝜂𝑖arecountry-fixedeffectstocaptureunobservedheterogeneityacrosscountries,andtime-unvaryingfactorssuchageographicalvariableand𝜇𝑡aretime-fixedeffectstocontrolforglobalshocks(suchastheglobalbusinesscycleorcommoditypriceshocks).𝜀𝑖𝑡isanidiosyncraticerrortermsatisfyingusualassumptionsofzeromeanandconstantvariance.WeusetheDriscoll-Kraay(1998)robuststandarderrors,whichassumetheerrorstructuretobeheteroskedastic,autocorrelateduptosomelagandpossiblycorrelatedbetweenthegroups.Theempiricalanalysis—robusttovarioussensitivitychecks—providesconsistentevidencethatcorruptionhasastatisticallysignificanteffectonfatalitiescausedbynaturaldisastersacrosstheworld.First,weestimatethespatiallycorrelatedconsistentmodelforthenumberofdeathsperpopulationduetonaturaldisastersinagivenyear.Theseresults,presentedinTable2,demonstrateaconsistentpicturewiththesignsofallestimatedparameterscorrespondingtotheirexpectedvaluesacrossdifferentspecifications.Corruption—themainexplanatoryvariableofinterestinthisanalysis—issignificantlyandpositivelyassociatedwithahighernumberofnaturaldisaster-relateddeathsinoursampleof135countriesduringtheperiod1980–2020.Theestimatedcoefficientoncorruptionisstatisticallysignificantacrossallspecification,therebyimplyingthata1percentincreaseincorruptionleadstoanincreaseofabout2.1percentinthenumberofdeathsperpopulationcausedbynaturaldisaster,aftercontrollingforothereconomic,demographic,healthcareandinstitutionalfactors.Hence,wecaninferthatthehigherthecorruptionindexinagivencountry,thegreaterthenumberoffatalitiesasashareofpopulationduetonaturaldisasters.Toputthisfindingintoperspective,thedifferencebetweentheleastandmostcorruptcountriesinoursampleimpliesasixfoldincreaseinthenumberofdeathsperpopulationcausedbynaturaldisasterinagivenyear.Withregardstocontrolvariables,weobtainconsistentandintuitiveestimationresults.ThelevelofrealGDPpercapitaisinverselycorrelatedtonaturaldisaster-relateddeaths,suggestingthatdisasterstendtoresultinfewerfatalitiesincountrieswithhigherlevelsofincome.ThecoefficientonrealGDPpercapitaislargerinmagnitude(thanthatoncorruption)butstatisticallyinsignificantacrossallspecifications.Likewise,wefindthattradeopenness—ameasureofinternationaleconomicintegrationanddevelopment—doesnotappeartohavestatisticallysignificanteffectonnaturaldisasterdeaths.Bothurbanizationandhealthcareconditionsarecrucialfactorsindeterminingcross-countrydifferencesinthenumberofdeathsperpopulation9causedbynaturaldisasters.Thecoefficientsonurbanizationandhealthcareindicateastrongandstatisticallysignificantnegativerelationshipbetweentheshareofpopulationlivinginurbanareasandthestrengthofthehealthcaresystemandnaturaldisaster-relateddeathsperpopulation.Finally,weintroduceaseriesofinstitutionalandpoliticalvariables,whichdonotaltertheresults,butprovidemoreinformationonfactorsaffectingthehumanitarianimpactofnaturaldisasters.Allthreemeasures—bureaucraticquality,democraticaccountability,andgovernmentstability—contributetoadeclineinthenumberofdeathsperpopulationcausedbynaturaldisasters,butwithvaryingdegreesofstatisticalsignificance.Inotherwords,countrieswithhigherbureaucraticquality,greaterdemocraticaccountabilityandmorestablegovernmentstendtohavelowermortalityfromnaturaldisasters.Forrobustnessandtoobtainabetterunderstandingofhowthelevelofeconomicdevelopmentshapestheimpactofcorruptiononnaturaldisasterdeaths,weestimatethemodelseparatelyfordifferentincomegroups—advancedeconomiesanddevelopingcountries—andpresenttheseresultsinTable3.ThisdisaggregationrevealsastrikingcontrastintheimpactofcorruptiononTable2.DeterminantsofNaturalDisaster-RelatedDeaths(Dependentvariable:Annualnumberofdeathsperpopulation)Specification(1)(2)(3)(4)SampleofcountriesALLALLALLALLCorruption0.02030.02070.02040.0205Income(0.007)(0.008)(0.008)(0.008)Openness-0.0675-0.0718-0.0988-0.0939Urbanization(0.046)(0.055)(0.065)(0.068)Healthcare0.03230.01580.01350.0039Bureaucraticquality(0.028)(0.024)(0.023)(0.024)Democraticaccountability-0.3405-0.6330-0.6978-0.6220Governmentstability(0.311)(0.351)(0.370)(0.347)-0.0695-0.0877-0.0897-0.0877(0.033)(0.049)(0.049)(0.051)-0.0714(0.037)-0.0080-0.0108(0.012)(0.008)Observations1,8121,5091,5091,509Countries117888888R2_weighted0.0412CountryEffectsYes0.06630.05950.0614TimeEffectsYesYesYesYesYesYesYesNote:Driscoll-Kraayestimation.Standarderrorsinparenthesis.,,denotestatisticalsignificanceatthe10,5and1percentlevels,respectively.Aconstanttermisincludedbutomittedinthetable.Countryandtimeeffectsareincludedbutnotshownforreasonsofparsimony.10naturaldisaster-relatedfatalitiesacrosscountrieswithvaryinglevelsofeconomicdevelopment.Whilecorruptionhasnosignificanteffectinadvancedeconomies,ithasastatisticallyhighlysignificanteffectindevelopingcountries.Theestimatedcoefficientoncorruptionisstatisticallysignificantacrossallspecification,therebyimplyingthata1percentincreaseincorruptionleadstoanincreaseofalmost2.5percentinthenumberofnaturaldisaster-relateddeathsperpopulationindevelopingcountriesofoursample.Thisfinding,inourview,confirmsthecriticalrelationshipbetweeneconomicdevelopmentandinstitutionalcapacityinstrengtheninggoodgovernanceandcombatingcorruption,whichunderminethequalityofphysicalandinstitutionalinfrastructureandtherebyleadtoanincreaseinthenumberofdeathsfromnaturaldisasters.Finally,wesplitthesamplebythemedianlevelofcorruptionandestimatethemodelseparatelyforcountrieswithhighandlowlevelsofcorruption.Theseresults,presentedinTable4,validatethedeleteriouseffectsofwidespreadcorruption,especiallyindevelopingcountries.Wefindthattheimpactofcorruptiononnaturaldisaster-relateddeathsisnonlinear—increasingwiththelevelofcorruption.Table3.DeterminantsofNaturalDisaster-RelatedDeaths(Dependentvariable:Annualnumberofdeathsperpopulation)Specification(1)(2)(3)(4)SampleofcountriesALLAEEMLICCorruption0.02070.01900.02670.9558Income(0.008)(0.014)(0.011)(0.000)Openness-0.07180.0031-0.0858Urbanization(0.055)(0.081)(0.043)-1.0629Healthcare0.01580.0065-0.0377(0.000)Bureaucraticquality(0.024)(0.019)(0.058)-0.6330-0.4655-1.50456.7233(0.351)(0.348)(0.507)(0.000)-0.0877-0.0689-0.10540.0000(0.049)(0.059)(0.088)(0.000)-0.0714-0.0966-0.0687(0.037)(0.039)(0.044)-0.0551(0.000)-0.4330(0.000)Observations1,50974976025Numberofgroups8829597R2_weighted0.12660.14051.0000CountryEffects0.0663YesYesYesTimeEffectsYesYesYesYesYesNote:Driscoll-Kraayestimation.Standarderrorsinparenthesis.,,denotestatisticalsignificanceatthe10,5and1percentlevels,respectively.Aconstanttermisincludedbutomittedinthetable.Countryandtimeeffectsareincludedbutnotshownforreasonsofparsimony.11Table4.DeterminantsofNaturalDisaster-RelatedDeaths(Dependentvariable:Annualnumberofdeathsperpopulation)Specification(1)(2)(5)(6)LevelofCorruptionHighLowHighLowSampleofcountriesAllAllEMEMCorruption0.01470.02090.02250.0001Income(0.008)(0.015)(0.010)(0.018)Openness-0.1458-0.1223-0.1296-0.0658Urbanization(0.086)(0.104)(0.089)(0.128)Healthcare-0.03830.0519-0.0471-0.0286Bureaucraticquality(0.043)(0.039)(0.040)(0.328)-0.4721-0.1351-3.0012-0.6349(0.423)(0.352)(0.442)(1.596)-0.2220-0.0672-0.24250.0647(0.086)(0.037)(0.093)(0.095)-0.0900-0.0675-0.0928-0.0460(0.041)(0.033)(0.042)(0.042)Observations492819427147Numberofgroups37343310R2weighted0.18810.10300.19900.4173CountryEffectsYesYesYesYesTimeEffectsYesYesYesYesNote:Driscoll-Kraayestimation.Standarderrorsinparenthesis.,,denotestatisticalsignificanceatthe10,5and1percentlevels,respectively.Aconstanttermisincludedbutomittedinthetable.Countryandtimeeffectsareincludedbutnotshownforreasonsofparsimony.IV.CONCLUSIONNaturaldisastersareinevitable,resultinginsignificanteconomiclossesandtensofthousandsofdeathsinmostyearsacrosstheworld.Whiletherehasbeenacontinuousreductioninthenumberoffatalitiescausedbynaturaldisastersoverthepastcenturyowingtobetterlivingstandards,moreresilientphysicalinfrastructure,betterearlywarningindicatorsandstrongeremergencyresponsesystems,therearestillsignificantdisparitiesacrosscountriesinhumanitarianandeconomiclossescausedbynaturaldisasters.Thisisnotthefirstattemptintheliteraturetoanalyzeeconomic,institutionalandsocialfactorsindetermininglossesassociatedwithnaturaldisastersthatcontributetolossesassociatedwithnaturaldisasters,butweusealargepanelof135countriesoveralongperiodspanningfrom1980to2020andparticularlyfocusontheroleofcorruption.Theempiricalanalysisprovidesconvincingevidencethatwidespreadcorruptionincreasesthenumberofdisaster-relateddeaths,aftercontrollingforeconomic,demographic,healthcareandinstitutionalfactors.Hence,wecaninferthatthehigherthelevelofcorruption,thegreaterthenumberoffatalitiesperpopulationinnaturaldisasters.Toputthisempiricalfindingintoperspective,thedifferencebetweentheleast12andmostcorruptcountriesinoursampleimpliesasixfoldincreaseinthenumberofdeathsperpopulationcausedbynaturaldisasterinagivenyear.Ourresultsshowthatthisimpactisstrongerindevelopingcountriesthaninadvancedeconomies,highlightingthecriticalrelationshipbetweeneconomicdevelopmentandinstitutionalcapacityinstrengtheninggoodgovernanceandcombatingcorruption.Wealsofindevidenceofnonlineareffectswithhigherlevelsofcorruptionresultinginanevenlargernumberofdeathsfromnaturaldisasters,especiallyindevelopingcountries.Thesefindingsarerobusttovariouseconometricspecificationsandsampleheterogeneity,whichweusetoobtainagranularanalysisoftheimpactofcorruptiononlossofhumanlivescausedbynaturaldisasters.Corruptionisacomplexphenomenonthataffectsallcountries,butitseconomic,institutional,politicalandsocialcausesandconsequencesshowgreatvariationacrosscountries.Empiricalfindingspresentedinthisstudyshowthatdevelopingcountriestendtobemorevulnerabletothedeleteriousimpactofcorruptioninnaturaldisasters.Inourview,thisreflectsthelowqualityofbuildingsandinfrastructureandtheweaknessofhealthandriskmanagementsystemsduetowidespreadcorruption.Ourresultsthereforehighlighttheimportanceofpromotinganti-corruptionmeasurestostrengtheninstitutionsandcreateaconduciveenvironmentforgreatertransparencyingovernanceandappropriateuseofpublicresources.13REFERENCESAkhtaruzzaman,M.(2011).“Post-DisasterReliefOperationsinBangladesh:ContextandChallenges,”Disasters,Vol.35,pp.803–818.Alam,M.,andH.Vennemo(2014).“CorruptionandCatastrophicDisaster:AReview,”EcologicalEconomics,Vol.108,pp.49–58.Albala-Bertrand,J.(1993).PoliticalEconomyofLargeNaturalDisasters(NewYork:OxfordUniveristyPress).Aidt,T.,J.Dutta,andV.Sena(2008).“GovernanceRegimes,CorruptionandGrowth:TheoryandEvidence,”JournalofComparativeEconomics,Vol.36,pp.195–220.Alesina,A.,andB.Weder(2002).“DoCorruptGovernmentsReceiveLessForeignAid?”AmericanEconomicReview,Vol.92,pp.1126–1137.Anbarci,N.,M.Escaleras,andC.Register(2005).“EarthquakeFatalities:TheInteractionofNatureandPoliticalEconomy,”JournalofPublicEconomics,Vol.89,pp.1907–1933.Brinkerhoff,D.(2008).“BuildingGovernanceandAnticorruptioninDisaster-ProneCountries:PolicyLessonsfromNepal.”PublicAdministrationandDevelopment,Vol.28,pp.327–341.Cavallo,E.,S.Galiani,I.Noy,andJ.Pantano(2013).“CatastrophicNaturalDisastersandEconomicGrowth,”ReviewofEconomicsandStatistics,Vol.95,pp.1549–1561.Chang,C.,andY.Hao(2018).“EnvironmentalPerformance,CorruptionandEconomicGrowth:GlobalEvidenceUsingaNewDataset,”AppliedEconomics,Vol.49,pp.498–514.Cieślik,A.,andL.Gozcek(2018).“ControlofCorruption,InternationalInvestment,andEconomicGrowth:EvidencefromPanelData,”WorldDevelopment,Vol.103,pp.323–335.D’Agostino,G.,J.Dunne,andL.Pieroni(2016).“GovernmentSpending,Corruption,andEconomicGrowth,”WorldDevelopment,Vol.84,pp.190–205.Escaleras,M.,N.Anbarci,andC.Register(2007).“PublicSectorCorruptionandMajorEarthquakes:APotentiallyDeadlyInteraction,”PublicChoice,Vol.132,pp.209–230.Farzanegan,M.,andS.Witthuhn(2017).“CorruptionandPoliticalStability:DoestheYouthBulgeMatter?”EuropeanJournalofPoliticalEconomy,Vol.49,pp.47–70.Gründler,K.,andN.Potrafke(2019).“CorruptionandEconomicGrowth:NewEmpiricalEvidence,”ifoWorkingPapersNo.309(Munich:LeibnizInstituteforEconomicresearchattheUniversityofMunich).Gupta,S.,H.Davoodi,andE.Tiongson(2002).“CorruptionandtheProvisionofHealthcareandEducationServices,”InG.T.AbedandS.Gupta(Eds.),Governance,Corruption,andEconomicPerformance(Washington,DC:InternationalMonetaryFund).Habib,M.,andL.Zurawicki(2002).“CorruptionandForeignDirectInvestment,”JournalofInternationalBusinessStudies,Vol.33,pp.291–307.Haque,C.(2003).“PerspectivesofNaturalDisasterinEastandSouthAsia,andthePacificIslandStates:SocioeconomicCorrelatesandNeedAssessment,”NaturalHazards,Vol.29,pp.465–483.14Heywood,P.(2007).“TheEthicalFoundationsofPublicAdministration,”PublicAdministrationReview,Vol.67,pp.96–102.Hodge,A.,S.Shankar,P.Rao,andA.Duhs(2011).“ExploringtheLinksBetweenCorruptionandGrowth,”ReviewofDevelopmentEconomics,Vol.15,pp.474–490.Huang,C.(2016).“IsCorruptionBadforEconomicGrowth?EvidencefromAsia-PacificCountries,”NorthAmericanJournalofEconomicsandFinance,Vol.35,pp.247–256.Im,K.,M.Pesaran,andY.Shin(2003).“TestingforUnitRootsinHeterogeneousPanels,”JournalofEconometrics,Vol.115,pp.53–74.Iqbal,M.(2018).“PoliticalEconomyofBuildingRegulationsEnforcement:EmpiricalInsightsfromPakistan,”LandUsePolicy,Vol.73,pp.435–444.Kahn,M.(2005).“TheDeathTollfromNaturalDisasters:TheRoleofIncome,Geography,andInstitutions,”ReviewofEconomicsandStatistics,Vol.87,pp.271–284.Kellenberg,D.,andA.Mobarak(2008).“DoesRisingIncomeIncreaseorDecreaseDamageRiskfromNaturalDisasters?”JournalofUrbanEconomics,Vol.63,pp.799–802.Klomp,J.(2016).“EconomicDevelopmentandNaturalDisasters:ASatelliteDataAnalysis,”GlobalEnvironmentalChange,Vol.36,pp.67–88.Klomp,J.,andJ.deHaan(2013).“CorruptionandtheEffectivenessofForeignAid,”EuropeanJournalofPoliticalEconomy,Vol.29,pp.84–101.Koenker,R.,andG.Bassett(1978).“RegressionQuantiles,”Econometrica,Vol.46,pp.33–50.Lehoucq,F.,andJ.Molinas(2002).“StuffingtheBallotBox:Fraud,ElectoralReform,andDemocratizationinCostaRica,”ComparativePoliticalStudies,Vol.35,pp.749–775.Loayza,N.,E.Olaberria,J.Rigolini,andL.Christiaensen(2012).“NaturalDisastersandGrowth:GoingBeyondtheAverages,”WorldDevelopment,Vol.40,pp.1317–1336.Machado,J.,andJ.SantosSilva(2019).“QuantilesviaMoments,“JournalofEconometrics,Vol.213,pp.145–173.Mauro,P.(1995).“CorruptionandGrowth,”QuarterlyJournalofEconomics,Vol.110,pp.681–712.Mo,P.(2001).“CorruptionandEconomicGrowth,”JournalofComparativeEconomics,Vol.29,pp.66–79.Noy,I.(2009).“TheMacroeconomicConsequencesofDisasters,”JournalofDevelopmentEconomics,Vol.88,pp.221–231.Padli,J.,andM.Habibullah(2009).“NaturalDisasterandSocioeconomicFactorsinSelectedAsianCountries:APanelAnalysis,”AsianSocialScience,Vol.5,pp.65–71.Raschky,P.(2008).“InstitutionsandtheLossesfromNaturalDisasters,”NaturalHazardsandEarthSystemSciences,Vol.8,pp.627–634.Rose-Ackerman,S.(2006).InternationalHandbookontheEconomicsofCorruption(Cheltenham,UK:EdwardElgar).Schumacher,I.,andE.Strobl(2011).“EconomicDevelopmentandLossesduetoNaturalDisasters,”EcologicalEconomics,Vol.72,pp.97–105.15Shah,A.(2006).“CorruptionandDecentralizedPublicGovernance,”WorldBankPolicyResearchWorkingPaperNo.3824(Washington,DC:WorldBank).Skidmore,M.,andH.Toya(2007).“EconomicDevelopmentandtheImpactofNaturalDisasters,”EconomicLetters,Vol.94,pp.20–25.Taghizadeh-Hesary,F.,N.Yoshino,A.Mortha,andT.Sarker(2019).“QualityInfrastructureandNaturalDisasterResilience,”ADBIWorkingPaperNo.991(Manila:AsianDevelopmentBankInstitute).Tanzi,V.(1998).“CorruptionAroundtheWorld:Causes,Consequences,Scopes,andCures,”IMFWorkingPaperNo.98/63(Washington,DC:InternationalMonetaryFund).Tol,R.,andF.Leek(1999).“EconomicAnalysisofNaturalDisasters,”InT.Downing,A.Olsthoorn,andR.Tol(Eds.),ClimateChangeandRisk(London:Routledge).Uberti,L.(2022).“CorruptionandGrowth:HistoricalEvidence,1790–2010,”JournalofComparativeEconomics,Vol.50,pp.321–349.
VIP
