Critical18jjj.com

FirearmAvailabilityandHomicideRatesacross26High-IncomeCountriesDavidHemenway,PhD,andMatthewMiller,MD,MPH,ScDBackground:Amongdevelopedna-tions,theUnitedStateshasthehighestrateofciviliangunownership,andthehighesthomiciderate.
WeexaminewhethertheUnitedStatesismerelyanexception,orifarelationshipbetweengunavailabilityandhomicideexistsacrossalldevelopednations.
Methods:Homicideratesfortheearly1990scomefrom26of27ofthehighlyindustrializedorhigh-incomecountrieswithgreaterthan1millionpop-ulationasclassifiedbytheWorldBank.
Twocommonproxiesforgunavailabilityareused,thepercentageofsuicideswithafirearm,andthe"Cookindex,"theaver-ageofthepercentageofsuicideswithafirearmandthepercentageofhomicideswithafirearm.
Results:Insimpleregressions(nocontrolvariables)across26high-incomenations,thereisastrongandstatisticallysignificantassociationbetweengunavail-abilityandhomiciderates.
Conclusion:Acrossdevelopedcoun-tries,wheregunsaremoreavailable,therearemorehomicides.
Keywords:Firearms,Guns,Homi-cide,International.
JTrauma.
2000;49:985–988.
Ofalldevelopednations,theUnitedStateshasthemostprivatelyownedguns,andthehighestratesofhomi-cide.
ApproximatelytwothirdsofmurdervictimsintheUnitedStatesarekilledwithagun.
Isgunavailabilityassociatedwithhomicidesacrossalldevelopednations,oristheUnitedStatesjustanexceptionalcaseAcrossdevelopednations,themostwidelycitedinter-nationalstudyexamined14countries.
1Gunownershipinfor-mationwasobtainedfromcomparabletelephoneinterviews.
Therateofgunhomicide,andthetotalhomicideratewassignificantlycorrelatedwithlevelsofgunownership;therewasnosignificantcorrelationbetweennongunhomicideandgunownership.
Astheauthorsofthatstudynote,atleastthreemethodologicconsiderationslimittheinferencesthatcanbedrawnfromtheresults:onlyasmallnumberofcountrieswassampled,thecountriesselectedforinclusionwerenotarandomsampleofdevelopednations,andhomi-cidedatawereforyearsearlierthanthedataongunowner-shiprates.
Ourstudyextendsthispreviousworkbyanalyzingcon-temporaneousdataonfirearmavailabilityandhomicideratesamong96%(26of27)ofallthecountrieswithmorethan1millioninhabitantsdefinedbytheWorldBankashighin-comeorhighlyindustrialized.
MATERIALSANDMETHODSUsingdatareportedbyresearchersattheCentersforDiseaseControlandPrevention(CDC),2countrieswerese-lectedinaccordancewiththeWorldBank'sdefinitionofnationsthatarehighlyindustrializedorhavehigh-incomeeconomies(1992grossnationalproductpercapita$8,356).
EachyeartheWorldBankclassifiesallnationsbyincomebasedontheirgrossnationalproduct.
Ofthenationswithpopulationsgreaterthan1millionincludedinthe1994WorldDevelopmentReport,327wereclassifiedashighlyindustri-alizedorashavinghigh-incomeeconomies.
Twenty-six(96%)ofthese27countriesprovideddataforanalysis.
Over76%provideddatafor1993or1994;therestgavedatafor1990,1992,or1995.
2Thecountries,theyearforwhichthedatawereprovided,populations,andoverallhomicidedeathratesaregiveninTable1.
Weanalyzetheassociationbetweenfirearmavailabilityandtotalhomicideratesinboththe26high-incomeorhighlyindustrializednationsdescribedabove,aswellasinthesampleof36countriescompiledbytheCDCresearchers,2whichincludesanadditional10"uppermiddleincomena-tions"thatprovideddata(53%ofthe19uppermiddleincomecountriesthatwereaskedprovidedtherequesteddata).
Asaproxyforgunavailability,weusethepercentageofsuicideswithafirearmfortheindependentvariable;wealsouseasecondproxy,theaverageofthepercentageofhomi-cidesandsuicidesattributabletofirearms(Cookindex).
Thefirstmeasureishighlycorrelatedwiththepercentageofhouseholdsreportinggunownershipinstudiesbothof170UScities(r0.
86)4andacross16developednations(r0.
91).
5Thismeasurehasoftenbeenusedasaproxyforgunavailability.
4,6–8TheCookindexhasbeenvalidatedforur-banpatternsofregionalgunownershipwithdatafromsur-veysofresidents(r0.
91).
9TheCookindexhasalsobeenvalidatedattheregionallevelintheUnitedStateswithSubmittedforpublicationFebruary14,2000.
AcceptedforpublicationAugust9,2000.
Copyright2000byLippincottWilliams&Wilkins,Inc.
FromtheHarvardInjuryControlResearchCenter,HarvardSchoolofPublicHealth,Boston,Massachusetts.
ThisresearchwassupportedinpartbygrantsfromtheCentersforDiseaseControlandPrevention,theJoyceFoundation,theRobertWoodJohnsonFoundation,andtheCenteronCrimes,CommunitiesandCultureoftheOpenSocietyInstitute.
Addressforreprints:DavidHemenway,PhD,Director,HarvardInjuryControlResearchCenter,HarvardSchoolofPublicHealth,677HuntingtonAvenue,Boston,MA02115;email:hemenway@hsph.
harvard.
edu.
TheJournalofTRAUMAInjury,Infection,andCriticalCareVolume49Number6985GeneralSocialSurvey–basedestimatesofhandgunowner-ship(r0.
96)andfor21statesthatreportedhouseholdfirearmownershiplevelsonBehavioralRiskFactorSurveil-lancesurveys(r0.
83).
10Internationaldataforthe26developedcountrieswerenotavailableonratesofhandgunownershiporhandgunfatalities.
Analysesareconductedus-ingboththehomiciderateandthenaturallogofthehomiciderate,thelatterbetterapproximatinganormaldistribution.
RESULTSAcrossthe26developedcountries,thesimplecorrelationcoefficientbetweenthegunavailabilityproxyandthetotalhomiciderateis0.
69;therelationshipissignificantatp0.
000.
UsingtheCookindex,thecorrelationcoefficientis0.
74andtherelationshipissignificantatp0.
000(Table2).
Therelationshipbetweengunavailabilityandhomicideissimilarforbothformsofthedependentvariable—thetotalhomiciderateorthenaturallogofthehomiciderate.
DroppingtheUnitedStatesfromtheanalysis,theresultsremainsignificantwhentheCookindexisused(p0.
01)butnotwhenthepercentageofsuicideswithafirearmisusedasthegunproxy(Table2).
ThefirearmhomiciderateisverystronglyassociatedwiththeCookindex;thenonfirearmho-miciderateshowsnostatisticallysignificantcorrelation(notshown).
TheonecountrydefinedbytheWorldBankashighincomeorhighlyindustrializedthatmightnotbeconsideredas"developed"isKuwait.
EliminatingKuwaitfromtheanal-ysisdoesnotalterthefindings(notshown).
NorthernIrelandTable1HomicideRatesfor26High-IncomeCountriesCountryYearPopulation(inthousands)TotalHomicideRateper100,000SuicideswithaFirearm(%)UnitedStates1993257,7839.
9360.
9NorthernIreland19941,6426.
8215.
9Finland19945,0883.
2421.
2Israel19935,2612.
3226.
1Italy199256,7642.
2513.
9Scotland19945,1322.
242.
7Canada199228,1202.
1628.
2Australia199417,8381.
8618.
6Taiwan199421,0861.
781.
7Singapore19942,9301.
711.
2NewZealand19933,4581.
4716.
7Belgium19909,9671.
4113.
4Switzerland19947,0211.
3226.
4Sweden19938,7181.
3013.
3HongKong19935,9191.
230.
7Denmark19935,1891.
2110.
1Germany199481,3381.
177.
5Austria19948,0291.
1718.
4France199457,9151.
1224.
7Netherlands199415,3821.
113.
0Kuwait19951,6841.
013.
6Norway19934,3240.
9729.
0Spain199339,0860.
955.
5Ireland19913,5250.
629.
6Japan1994124,0690.
620.
2England/Wales199251,4290.
554.
3DatafromKrugetal.
,1998.
2Table2Coefficients(withSignificanceLevels)andCorrelationCoefficientsinRegressionsofGunAvailabilityandHomicidefor26High-IncomeCountries,1990saDependentVariable:HomicideRatefor26DevelopedNationsGunAvailabilityProxy1(%SuicidesUsingGun)GunAvailabilityProxy2(CookIndex)CorCorCruderate10.
1(0.
000)0.
699.
2(0.
000)0.
74ExcludingUS2.
9(0.
236)0.
254.
9(0.
003)0.
57Naturallogofrate2.
9(0.
001)0.
612.
9(0.
000)0.
70ExcludingUS1.
8(0.
112)0.
322.
2(0.
005)0.
55,coefficients;Cor,correlationcoefficients.
aRawdatafromKrugetal.
,1998.
2TheJournalofTRAUMAInjury,Infection,andCriticalCare986December2000wasinthemidstof"thetroubles"orcivilwarin1994.
IfNorthernIrelandiseliminatedfromtheanalysis,theresultsremainstatisticallysignificant(notshown).
Weightingthe26observationsbythepopulationofthecountryincreasestheassociationbetweenfirearmavailabilityandhomicide(notshown),sincesomeofthelargestcountriesareattheextremes.
TheUnitedStateshasmanygunsandmanyhomicides,whereasJapanandEnglandhavefewgunsandfewhomicides.
Whenall36countriesforwhichKrugetal.
2obtaineddataareconsidered(the26high-incomeorhighlyindustrializedand10uppermiddleincomenations),theresultsoftenremainsig-nificant(Table3).
Forexample,thenaturallogofthehomiciderateispositivelyassociatedwithbothmeasuresoffirearmavail-ability(p0.
05),andthecrudehomiciderateissignificantlyassociatedwiththeCookindex(p0.
05)(Table3).
DISCUSSIONResultsfromoursimpleregressionsof26developednationsshowahighlysignificantpositivecorrelationbetweentotalhomicideratesandbothproxiesforgunavailability.
WhenweusetheCookindex,thecorrelationisparticularlystrong,andresultsremainstatisticallysignificantwhentheUnitedStatesisexcluded.
Itis,ofcourse,possiblethatthefindingscouldbe"ex-plained"byothervariables;ourregressionscontainonlyoneindependentvariable,ameasureofgunavailability.
How-ever,bylookingexclusivelyatindustrializedandhigh-in-comenations,wedocontrol,inpart,forsomesocialandeconomicvariables.
Cross-sectionalstudieslikeoursdonotprovideinforma-tionaboutcausality.
Forexample,arelationshipbetweenvariablesxandymaybeattributabletoxcausingy,ycausingx,orathirdvariableaffectingboth.
OurresultsareconsistentwiththeKilliasresults.
1Com-paredwithourstudy,Killiasusedameasureofratesofhouseholdgunownership—comparablehouseholdsurveyre-sults—butincludedfewernations.
Ourresultsarealsocon-sistentwithmanyecologicalstudiesoftheUnitedStates,whichfindthatinareaswithmoregunstherearemorehomicides,6,8,11–13andwithcase-controlstudiesthatfindaguninthehometobeariskfactorforhomicidevictimizationandtheperpetrationofmurder.
14–17Currently,thereisno"goldstandard"forthemeasureofgunavailability.
Likeourmeasure,thepercentageofhouse-holdswithafirearmisonlyacrudeproxyofgunavailabilityforuseinhomicide,forittellsnothingaboutthenumberofgunsperhousehold,thetypes(e.
g.
,handguns)orcalibersoftheguns,thestorageoftheguns,thecarryingoftheguns,ortheeasewithwhichurbanadolescentscanobtainhandguns.
Inaddition,surveystypicallyunderrepresentpoorpeople(e.
g.
,householdswithouttelephones),andindividualswhoowngunsillegallymaynotreportthem.
Italsoappearsthat,atleastintheUnitedStates,womenlivingintwo-adulthouseholdswithgunsoftendonotknowthatthereisagunintheirhome.
18,19Fortunately,householdgunownershipandbothCookindexandthepercentageofsuicideswithagunarehighlycorrelated,suggestingthattheyareprovidinginfor-mationaboutsimilarconstructs.
Anotherlimitationofourstudyisthatthemeasuresofhomicideandfirearmavailabilitymayvaryfromcountrytocountrybecauseofdifferencesinthespecificityandsensi-tivityofthesurveillancesystems.
Forexample,differentcountriesmaynothavecomparabledefinitionsofhomicide(e.
g.
,whetherguerillaorterroristactivitiesshouldbeinclud-ed).
Problemswithdatacomparabilityandaccuracyarepar-ticularlyacutefornonindustrializednations.
Onthisbasisalone,itisexpectedthatincludingindustrializedandnonin-dustrializednationsinasingleanalysis,despitethelargernumberofobservations,couldreducethepowertofindastatisticallysignificantassociationbetweenthevariables.
Socialscientistsaretaughtthatitisgenerallyinappro-priatetoincludeindustrializedandnonindustrializedcoun-triesinthesameanalysis,becausetheirsocial,political,andeconomicstructuresareusuallyquitedifferent;confoundingbyculturalfactorsislikelytobemoresubstantialwhenbothindustrializedandnonindustrializednationsareincludedinthestudy.
Forexample,inthesampleof36nations,71%ofthefirearmdeathsinthehigh-incomecountriesweresui-Table3Coefficients(withSignificanceLevels)andCorrelationCoefficientsinRegressionsofGunAvailabilityandHomicide,for36UpperMiddleIncomeandHigh-IncomeCountries,1990saDependentVariable:HomicideRatefor36DevelopedandLesserDevelopedNationsGunAvailabilityProxy1(%SuicidesUsingGun)GunAvailabilityProxy2(CookIndex)CorCorCruderate10.
5(0.
154)0.
2412.
3(0.
045)0.
34ExcludingUS9.
0(0.
324)0.
1711.
8(0.
092)0.
29Naturallogofrate2.
7(0.
018)0.
392.
8(0.
002)0.
49ExcludingUS2.
3(0.
095)0.
292.
6(0.
012)0.
42,coefficient;Cor,correlationcoefficients.
aRawdatafromKrugetal.
,1998.
2The10additionalcountriesthatareaddedtothe26high-incomecountriesareBrazil,Mexico,Estonia,Argentina,Portugal,Slovenia,Greece,Hungary,Mauritius,andSouthKorea.
Brazil,Mexico,andEstoniahaveabouttwotothreetimesthemurderrateoftheUnitedStates.
FirearmAvailabilityandHomicideVolume49Number6987cides,whereas72%ofthefirearmdeathsintheuppermiddleincomecountrieswerehomicides.
2Onestudymixedtogetherthe10nonindustrializedna-tions(includingEstonia,Mexico,Brazil,Mauritius,andSlo-venia)withthe26high-incomenationsinananalysisoffirearmsandhomicide.
4Apriori,theresultswouldbeex-pectedtobebiasedtowardthenull.
Yetevenincludingbothdevelopedandlessdevelopednationsintheanalysis,theassociationbetweenthehomiciderateandthegunavailabil-itymeasuresarepositiveandusuallystatisticallysignificant.
Inouranalysis,wefind,forindustrializedcountries,averystrongandhighlysignificantassociationbetweengunavailabilitylevelsandtotalhomiciderates.
Therelationshipholdseventhoughthenumberofobservationsisfairlysmall(n26)andthemeasuresofgunavailabilityareonlyproxies.
TherelationshipdoesnotseemtobeattributableentirelytotheUnitedStates—whichhasmoregunsandmorehomicidesthanotherdevelopednations—becausetheresultsoftenholdwhentheUnitedStatesisexcludedfromtheanalysis.
Moregunsareassociatedwithmorehomicidesacrossindustrializedcountries.
REFERENCES1.
KilliasM.
Internationalcorrelationsbetweengunownershipandratesofhomicideandsuicide.
CanMedAssocJ.
1993;148:1721–1725.
2.
KrugEG,PowellKE,DahlbergLL.
Firearm-relateddeathsintheUnitedStatesand35otherhigh-andupper-middle-incomecountries.
IntJEpidemiol.
1998;27:214–221.
3.
WorldBank.
WorldDevelopmentReport.
NewYork:OxfordUniversityPress;1994:251–252.
4.
KleckG.
TargetingGuns:FirearmsandtheirControl.
NewYork:AldinedeGruyter;1997.
5.
KilliasM.
Gunownership,suicide,andhomicide:aninternationalperspective.
In:delFrateAA,ZvekicU,vanDijkJJJ,eds.
UnderstandingCrime:ExperiencesofCrimeandCrimeControl.
Rome:UNICRI;1993:289–303.
6.
LesterD.
Firearmavailabilityandtheincidenceofsuicideandhomicide.
ActaPsychiatrBelg.
1988;88:387–393.
7.
LesterD.
Guncontrol,gunownershipandsuicideprevention.
SuicideLifeThreatBehav.
1988;18:176–180.
8.
LesterD.
Relationshipbetweenfirearmavailabilityandprimaryandsecondarymurder.
PsycholRep.
1990;67:490.
9.
CookPJ.
Thetechnologyofpersonalviolence.
In:TonryM,ed.
CrimeandJustice.
Vol14.
Chicago:UniversityofChicagoPress;1991.
10.
MillerM,AzraelD,HemenwayD.
Firearmavailabilityandunintentionalfirearmdeaths.
AccidAnalPrev.
Inpress.
11.
BrearlyHC.
HomicideintheUnitedStates.
ChapelHill,NC:UniversityofNorthCarolinaPress;1932.
12.
SeitzST.
Firearms,homicide,andguncontroleffectiveness.
LawSocRev.
1972;6:595–614.
13.
BrillS.
FirearmAbuse:AResearchandPolicyReport.
Washington,DC:PoliceFoundation;1977.
14.
KellermannAL,RivaraFP,RushforthNB,etal.
Gunownershipasariskfactorforhomicideinthehome.
NEnglJMed.
1993;329:1084–1091.
15.
CummingsP,KoepsellTD,GrossmanDC,etal.
Theassociationbetweenthepurchaseofahandgunandhomicideorsuicide.
AmJPublicHealth.
1997;87:974–978.
16.
RowlandJ,HoltzhauerF.
Homicideinvolvingfirearmsbetweenfamily,relativesandfriendsinOhio:anoffender-basedcase-controlstudy.
AmJEpidemiol.
1989;130:825–825.
17.
KleckG,HoganM.
Anationalcasecontrolstudyofhomicideoffendingandgunownership.
SocProbl.
1999;46:275–293.
18.
LudwigJ,CookPJ,SmithTW.
Thegendergapinreportinghouseholdgunownership.
AmJPublicHealth.
1998;88:1715–1718.
19.
AzraelD,MillerM,HemenwayD.
ArefirearmsstoredsafelyinhouseholdswithchildrenItdependsonwhomyouask.
Pediatrics.
2000;106:E31.
TheJournalofTRAUMAInjury,Infection,andCriticalCare988December2000