Does rising income increase or decrease damage ris

JournalofUrbanEconomics63(2008)788–802∗,AhmedMushfiqMobarakbaUniversityofMontana,EconomicsDepartment,32CampusDrive#5472,Missoula,MT59812-5472,USAbYaleUniversity,SchoolofManagement,Box208200,NewHaven,CT06520-8200,USAReceived30October2006;revised7May2007Availableonline22August2007AbstractRecentempiricalliteraturehasfoundanegativerelationshipbetweenincomepercapitaandmeasuresofriskfromnaturaldisaster,supportiveoflogicthathigherincomesallowcountriestomitigatedisasterrisk.Wearguethatbehavioralchangesatthemicrolevelinresponsetoincreasingincome(suchaslocationchoiceandextentofcostlyabatementactivity)mayleadtoanon-linearrelationshipbetweenaggregateincomesanddisasterdamages,wheretherisksincreasewithincomebeforetheydecrease.Inacountry-yearpaneldataset,weshowthatdisasterriskassociatedwithflooding,landslidesandwindstormsincreaseswithincomeuptoGDPpercapitalevelsof$5044,$3360,and$4688peryearrespectivelyanddecreasethereafter.Suchnon-linearimpactsareabsentforotherdisastertypessuchasextremetemperatureeventsandearthquakeswherethelinksbetweenhumanbehavioralchoicesandexposuretoriskarenotasstrong.Fromapolicyperspective,thissuggeststhatfortheleastdevelopedcountries,thedualgoalsofdisasterriskpreventionandeconomicdevelopmentcannotbeassumedtobecomplementaryforallformsofnaturaldisaster.Inadditiontoallocatingresourcestomanagedisasterrisk,thepoorestnationsmayhavetobemoreproactiveinenactingpoliciesthatalterthebehavioralchoicesofcitizensthatimpactacountry’sexposuretonaturaldisasterrisk.©2007ElsevierInc.Allrightsreserved.JELclassification:O1;O2;D1;Q54;R5Keywords:Economicdevelopment;Naturaldisaster;Kuznetscurve;Urbanization1.IntroductionThestringoflargenaturaldisastersworldwideinthelastfewyears,includingthetsunamiof2004thatdevas-tatedcoastalregionsofSoutheastAsia,earthquakesinSouthAsia,andhurricanesinthecoastalUnitedStatesandtheCaribbeanserveasshockingremindersofthe*Correspondingauthor.Fax:[phone].E-mailaddresses:[e-mail](D.K.Kellenberg),[e-mail](A.M.Mobarak).tremendouspowerofnaturetoeffectdeathsanddam-agesinbothdevelopedanddevelopingcountriesalike.Manyscholarshavearguedthatthesocialandeco-nomiccostsofnaturaldisastersaredisproportionatelybornebypoorpeopleindevelopingcountries.Kahn(2005)pointsout,forexample,thatbetween1980and1999,Indiaexperiencedfourteenearthquakesthatkilledatotalof12,137peoplewhiletheUnitedStatesexpe-riencednineearthquakesthatkilledonly137people.Worldwide,naturaldisasterskilledanestimated2.69millionpeopleandledtoUS$955billionineconomic0094-1190/$–seefrontmatter©2007ElsevierInc.Allrightsreserved.doi:10.1016/j.jue.2007.05.003D.K.Kellenberg,A.M.Mobarak/JournalofUrbanEconomics63(2008)788–802789damagesbetween1970and2001(Yang,2005),butonapercapitabasis,thelosseswere20timeslargerindevelopingcountriesthaninindustrializednations(Bendimerad,2000).Aseemingconsensushasemergedintheliteraturethattheriskofdeathanddamagesfromnaturaldisas-tersismonotonicallydecreasinginincome.Thelogicthateconomicdevelopmentallowsacountrytobet-termanageandmitigatetheriskfromdisastershasintuitiveappeal,andrecentempiricalresearchusingcross-sectionalorpanelmacrodata(e.g.UNDP,2004;Anbarcietal.,2005;Kahn,2005;ToyaandSkidmore,2007)reportresultsthatarebroadlysupportiveofthislogic.Adoptingthisviewalsolendsitselftoasimpleandattractivemacro-levelpolicyprescriptiontoman-agethehumanandeconomicriskfromnaturaldisas-ters:allowcountriestodevelop,andtheriskofdisasterdamagesshouldfall.Indeed,somepolicyanalystshavearguedthatthebestwaytoavoidlargedisasterdamagesisforpoorcountriestodevelopandgrowfaster(e.g.Okonski,2004;Hoke,2005).Thepurposeofthispaperistochallengethisemerg-ingconsensusinbothacademicandpolicycirclesonthemonotonicnegativerelationshipbetweendevelopmentanddisasterdamages.Wepresenttworelatedbitsofev-idencethatargueinfavorofanon-monotonicKuznetsinverted-Utyperelationshipbetweendevelopmentanddisasterdamages.First,wearguebasedonsimplebehavioralanalysisofrisk-averseindividualsthatitisreasonabletoexpectthatatlowincomelevels,peoplemaymakechoicessuchthatdisasterdamageswillincreasewithincome,andthattheoverallrelationshipbetweendisasterdam-agesandincomemaybenon-monotonic.Forexample,ifhigherlevelsofproductiongenerategreaterdisasterriskasaby-product,andifpeoplecanchoosetoforegoconsumptioninordertomitigatethatrisk,itisentirelypossiblethatthisrisk-returntrade-offmayswinginfa-vorofconsumptionatlowlevelsofincomewherethemarginalutilityofconsumptionishigh,butthenswingtowardsmitigationasincomeincreases.Second,weshowempiricalevidenceofsuchnon-linearitiesinapaneldatasetofdisasterevents,dam-ages,andeconomicdevelopmentcovering133devel-opedanddevelopingnations,spanning28years.Wefirstrunthelinearspecificationsakintothecurrentliter-atureandareabletoreplicatethenegativerelationshipbetweendevelopmentanddisasterdamagesthatisre-portedinthisliterature,butthenshowthatinrichermodels,thereisevidenceofmorecomplicatednon-linearimpactsofdevelopmentontheriskofd