1HealthInsuranceandConsumption:EvidencefromChina’sNewCooperativeMedicalSchemeChong-EnBaiBinzhenWu*TsinghuaUniversityAbstractTheprecautionarysavingmotiveisapopularexplanationfortheChineseSavingPuzzle,butfewempiricalstudiesquantifytheimportanceoftheprecautionarysavingsinChina.Weexploitthequasi-naturalexperimentprovidedbytheintroductionofahealthinsuranceprograminruralChinatoexaminehowtheinsurancecoverageaffectshouseholdconsumption.Theresultsshowthatthehealthinsurancecoverageonaveragestimulatesnon-healthcarerelatedconsumptionbymorethan5percent.Theeffectexistsevenforhouseholdsthathavenohealth-careexpenditure.Inaddition,theeffectisstrongerforpoorerhouseholdsandhouseholdswithworseself-reportedhealthstatus,whobothhavehigherriskofrelativelylargehealthexpenditure.Wealsofindthattheinsuranceeffectvarieswithhouseholds’experiencewiththeprogram.Particularly,theeffectisonlysignificantinvillagesinwhichhouseholdswitnesssomereimbursementresultingfromtheinsurancecoverage.Inaddition,theprogramstimulatesmoreconsumptionforexperiencedparticipantsthanfornewparticipantsoftheprograminthesevillages.Keywords:NewCooperativeMedicalScheme;Consumption;HealthInsurance;PrecautionarySavings;ChineseSavingPuzzleJELClassificationNos.:D12,E21,I18*Correspondingauthor.Email:wubzh@sem.tsinghua.edu.cn;Tel:86-10-62772371.Fortheirvaluablesuggestions,wewanttothankparticipantsintheworkshopsforSavingandInvestmentinChinaatTsinghuaUniversity,seminarparticipantsattheCentralUniversityofFinanceandEconomics,andparticipantsattheStanford-TsinghuaConferenceon“ChinesePolicyReform:TopicsforTroubledTimes.”21IntroductionChinesehighandrisingsavingratehasattractedalotinternationalattention.Householdsavingratehasrisenbyabouttenpercentagepointsbetween1995and2008,reaching28percentofthedisposableincomein2008,whichishigherthanmostofothercountriesincludingEastAsiancountries(Prasad,2009).TheliteraturehasproposedmanyexplanationsforthisChineseSavingPuzzle.Apopularoneisthatthedissolutionofthetraditionalsocialsafetynethascreatedmoreprecautionarysavings(ChamonandPrasad,2008;Meng,2003).1Chinesegovernmenthasmadealotefforttoimproveitssafetynet.Thesocialinsuranceprogramsdisbursed1.2trillionRMBin2009,withanannualgrowthrateof19.4%since2000.However,therehavebeenveryfewempiricalstudiesthatquantifythesizeofprecautionarysavingsinChina.GiventhatChinesesavingratehasimportantglobalimpacts,itiscrucialtolearnhowmuchthepublicinsuranceprogramsaffectconsumptionandsavingsinChina.Theexistingempiricalliteratureforthedevelopedcountriesdeliversquitemixedresultsontheroleoftheprecautionarysavings.Theresultsrangefrombeingverysmall(Dynan,1993;Guiso,Jappelli,andTerlizzese,1992;Hurst,etal.,2010;Starr-Mccluer,1996)ormodest(EngenandGruber,2001;Lusardi,1998)toquitelarge(Banksetal.,2001;CarrollandSamwick,1998;Fuchs-SchÜndelnandSchÜndeln,2005;Kazarosian,1997).Studiesinthedevelopingcountriesarestillintheirearlystages(LeeandSawada,2010;Meng,2003;ZhangandWan,2004).Mostofthestudiesfindasubstantialamountofprecautionarysavings.Recentstudiesexploittheexogenousvariationsoftheinsurancecoveragecausedbypolicychanges,includingGruberandYelowitz(1999),EngenandGruber(2001),andKantorandFishback(1996)fortheUS,Atella,Rosati,andRossi(2005)forItaly,WagstaffandPradhan(2005)forVietnan,andChou,LiuandHammitt(2003)forTaiwan.Mostofthesestudiesconfirmtheimportanceoftheprecautionarysavings,butitisnotclearwhethertheestimatescanbeappliedtoChina,nottomentionthatChinesecultureremainsapopularexplanationfortheChineseSavingPuzzle.1Otherexplanationsincludethelife-cyclemodelwithdemographicchanges(Kraay,2000;ModiglianiandCao,2004),highincomegrowthandhabit(HoriokaandWan,2007),changesinthereturnrateofinvestment(Wen,2009),theimbalanceinthesexratio(WeiandZhang,2009),financialunderdevelopment(Caballero,GourinchasandFarhi,2008),incomeinequality(Jin,Li,andWu,2010)andculturaldifference.3ThelaunchofpublichealthinsuranceprogramsinChinaprovidesnaturalexperimentstoinvestigatethesizeoftheprecautionarysavingsinChina.Thispaperexploitsoneofmostimportantpolicychangesintheruralareas:theintroductionoftheNewCooperativeMedicalScheme(NCMS)sinceJuly2003.Thispublichealthinsuranceprogramisheavilysubsidizedbythegovernment,andhasbeenintroducedsequentiallyindifferentcounties.Households’participationisvoluntary.Wefocusonthedouble-differencecomparisonbetweentheinsuredandthenon-participantsinthevillagesthathavelaunchedtheprogram.ThereasonisthathouseholdsinthesamevillagearemorecomparabletoeachotherthantohouseholdsinadifferentvillageanditcanreducethebiasresultedfromthecontemporarypolicychangesthatwereintroducedsimultaneouslywiththeNCMS.Thedifference-in-differenceframeworkhelpsremoveallthetime-invariantselectionbias.Selectionbiasontheobservablesisfurtherreducedbyallowingthetemporalchangeinconsumptiontovarywithincomeandhealthstatusorbyapplyingmatchingdifference-in-difference.ThedataweusecombinethelongitudinalRuralFixed-PointSurveybetween2003and2006andahouseholdsurveyontheNCMSforasubsampleofthe2006roundoftheRuralFixed-PointSurvey.Theresultsindicatethathouseholdconsumptionotherthanhealthe