vehiclelockdown

IZADPNo.
13278UmutOguzogluCOVID-19LockdownsandDeclineinTrafficRelatedDeathsandInjuriesMAY2020IZADPNo.
13278COVID-19LockdownsandDeclineinTrafficRelatedDeathsandInjuriesMAY2020UmutOguzogluUniversityofManitobaandIZACOVID-19LockdownsandDeclineinTrafficRelatedDeathsandInjuries*Thispaperinvestigatesthedeclineintrafficaccidents,fatalitiesandinjuriesduringthemonthsthatCOVID-19stay-at-homeordersimplementedinTurkey.
TakingintoaccountthedeclineinaccidentsinMarchandApriltogether,theseratesroughlytranslateto200trafficrelateddeathsand17,600injuriesavoidedduringthemonthsthatstay-at-homeorderswereinplace.
TheDifferenceinDifferenceestimatesthatexploitvariationinquarantineordersamongsmallcities,IalsoshowthatstricterrulesinAprilareresponsibleforthedeclineofaccidentswithdeathorinjuryby35percent,deathby72percentandinjuriesby19percent.
JELClassification:P48,Q53,Q58Keywords:COVID-19,safer-at-home,lockdowns,trafficaccidents,fatality,injury,TurkeyCorrespondingauthor:UmutOguzogluDepartmentofEconomicsUniversityofManitoba647FletcherArgueBld.
WinnipegManitobaCanadaE-mail:umut.
oguzoglu@umanitoba.
ca*IthankJuliaWitt,SerkanOzbeklik,ResulCesur,SoodehSaberianandRyanGodwinforthevaluablecomments.
IamgratefulforSidraHameedforexcellentresearchassistantship.
Allmistakesaremyown.
2IntroductionThedeclineintrafficaccidentsduringCOVID-19lockdownshavebeennotedacrosstheworldinacademicarticles(seeamongothersforSpain(Aloietal2020),Brazil(Statista,2020),India(Sarla,2020),theUS(Brodeuretal,2020;ShillingandWatjen,2020)),andinthemedia.
Inadditiontodecreasingthenumberoffatalitiesandinjuriesdirectly,thedeclineintrafficaccidentsalsoindirectlyhelpedhospitalscopewiththepandemicmoreeffectivelybyfreeingupresources.
Chiletetal(2020)foundthatinChile,thedeclineinmobilityhadasignificantimpactonnon-respiratoryemergencyroomvisits,especiallyforvisitsrelatedtotraumathatcouldbetrafficrelated.
Moreover,giventhatupto3%ofdevelopingcountries'GDPisspentonemergencyresponseandhealthcareduetotrafficaccidents(WHO,2018),inadditiontoproductivitygainduetomissingdisabilitiesandlossoflife,thedeclineintrafficaccidentswillmakeavailablemuchneededfinancialresourcesthatcanbeallocatedtofightpost-pandemicrecessions1.
Inthispaper,Iquantifythedeclineintrafficaccidents,fatalitiesandinjuriesinTurkeyusingcity-levelaccidentreportsfromDecember2018toApril2020,andmonthlycountry-levelreportssinceJanuary2013.
InresponsetoasharpincreaseinCOVID-19cases,Turkeyimposedverystrictstay-at-homeorderssincethemiddleofMarchwhich,amongothermeasures,includetravelrestrictions,schoolandnon-essentialserviceclosures,permanentquarantineforunder20andover65yearoldsandindividualswithco-morbidities,andcurfewsonweekendsthatapplytoeveryonein31cities.
ThedeclineinmobilityinTurkey,especiallyinIstanbul,oneofthemosttraffic-congestedcitiesintheworld(ranked9outof4162),hasbeencomparabletoregionswithsevereoutbreaks.
Accordingtoresultsusingtheentiresampleofcities,theincidencerateofaccidentsthatinvolvedeathorinjuries,andaccidentswithmateriallossareestimatedtohavedeclinedby60and46percentrespectivelyinApril.
TheaverageestimateddecreaseinincidenceratesofinjuriesanddeathsinAprilwas43and65percentrespectively.
TakingintoaccountthedeclineinaccidentsinMarchandApriltogether,theseratesroughlytranslateto200trafficrelateddeathsand17,600injuriesavoidedduringthemonthsthatstay-at-homeorderswereinplace.
TheDifferenceinDifferenceestimatesthatexploitvariationinquarantineordersamongsmallcities,IalsoshowthatstricterrulesinAprilareresponsibleforthedeclineofaccidentswithdeathorinjuryby35percent,deathby72percentandinjuriesby19percent.
BriefChronologyofCOVID-19measuresinTurkeyAfterthefirstcaseofCOVID-19infectionwasannouncedonMarch11,theTurkishgovernmentimplementedaseriesofmeasuresrestrictingmobility.
OnMarch12,allschoolsincludinguniversitiesseizedoperation.
Shortlyafter,bars,nightclubs,museumsandpicnicareascloseddown,varioussportsleaguesweresuspended.
OnMarch21,acurfewwasimposedonpersons1For,theUS,Brodeuretalestimatethatbenefitsfromavoidedtrafficaccidentsrangefrom$7billionto$24billion.
2OtherTurkishcitiesinglobalcongestionrankingsincludeAnkara(ranked100),Izmir(133),Antalya(144),Adana(181),Bursa(208),Gaziantep(236),Mersin(246),Konya(329).
Forcongestionrankingssee:https://www.
tomtom.
com/en_gb/traffic-index/ranking/3age65andolderandpersonswithchronicillnesses.
Also,non-essentialservicessuchashairdressers,barbershopsandbeautysalonswereclosed.
SinceMarch27,individualsrequiretravelpermissionforintercitybustravels.
OnMarch30,thenumberoftaxisallowedintrafficwaslimited.
OnApril3,theinteriorministryquarantined30metropolitancitiesandthecityofZonguldak.
Acurfewimposedonpersonsyoungerthan20inthequarantinedcitiesinadditiontothecurfewonpersons65andolderthatwasalreadyinplace3.
OnApril10,a48hourscurfewimposedoneveryone.
Curfewswereperiodicallyre-introducedsincethen.
Theseruleshavehadaprofoundimpactonthemobilitypatternsthroughoutthecountry.
TherateofdrivingandwalkingdroppedsharplyinIstanbul,almostatacomparablemagnitudetointernationalregionswithsevereoutbreakssuchasLombardyRegioninItaly,NewYorkCityandMadrid(seeFigure1).
3ForamoreextensivecoverageofCOVID-19measuressee:https://www.
duvarenglish.
com/health-2/coronavirus/2020/04/09/ankara-vs-covid-19-chronology-of-turkeys-coronavirus-battle/4DataTwodatasourceswereusedfortheanalysisthatispresentedinthenextsection.
ThefirstisthesummaryreportsfromtheMinistryofTransportandInfrastructure(MTI)whichprovidemonthlydataonallaccidents,fatalitiesandinjuriesstartingfrom2013forthewholecountry4.
Theseconddatasourceisthemonthlycity-leveltrafficaccidentreportspublishedontheTurkishGeneralDirectorateofSecurity's(TGS)websitefromDecember2018toApril20205.
Theinformationincludedatthecitylevelarecityandmonthoftheaccidents,numberofaccidentsbrokendownbytype(accidentwithdeathorinjury,accidentwithmaterialloss),numberofdeathsandnumberofinjured.
Furtherdetailssuchaswhowasatfault(driver,pedestrian,vehicle,road),numberofvehiclesinvolvedintheaccident,typeofcrash…etcarealsoavailable,however,thisinformationisprovidedforthewholecountry,notforindividualcities.
Reportsdonotincludedemographicinformation.
Notethatcity-levelreportsonlyrecorddeathsthatoccurredatthesceneoftheaccident,whileMTIreportsincludefatalitiesbothatandaftercrashsince2015.
In2019,174,590accidentswithdeathsorinjuriesoccurredinTurkeywhere285,638peoplewereinjured,and2524peoplediedatthecrashscenewhichreflectsonlyhalfofthefatalitiesthatusuallyoccurwithin30daysofacrash.
Theper-cityaverageforeachmonthbetweenDecember2018andFebruary2020was2.
3deathsatthecrashsceneand260injuries.
Therehavebeenonaverage150accidentswithdeathsand/orinjuriespercityinamonthand232accidentswithmaterialloss(seeTable1).
Thesenumbersdramaticallydeclinedduringthelockdowns.
PercentagechangeinalltrafficaccidentsandaccidentsthatinvolveinjuryordeathcomparedtothepreviousyearisreportedinFigure2forthe10mostpopulouscitiesinTurkey.
Thereisasharpcontrastbetweenmonthswherestay-at-homeorderswereineffectandmonthsprecedingthem.
Especially,thesuddendropinApril,thefirstfullmonththatthelockdownmeasureswereineffect,isstriking.
Also,theeffecthasbeenmorepronouncedthemorethecitywascongested(seeFigure3).
4Reportsaredownloadedfrom:TurkishStatisticalInstitute'sOfficialStatisticsPortalhttp://www.
resmiistatistik.
gov.
tr/.
Reportfor2018canbedownloadedfromMTI'swebsitehttps://www.
kgm.
gov.
tr/Sayfalar/KGM/SiteTr/Trafik/TrafikKazalariOzeti.
aspx5ReportsinPDFformatfor2020canbedownloadedfromfollowinglinkhttp://trafik.
gov.
tr/istatistikler37.
AllpastreportsandfulldatausedinthepaperincsvformatareavailablefromtheAuthor.
56MethodologyThesimplemodelthatdescribesthecitylevelaccidentoutcomesforcityiatmonthmcanbecharacterisedasfollows:where!
!
"isoneofthefollowingcityleveloutcomes:numberofaccidentswithdeathorinjury,numberofaccidentswithmaterialloss,numberofdeaths,andnumberofinjuries.
identifiesmonthsthatsocialdistancingruleswereineffect(individualdummiesforMarchandApril2020)arecity,monthandyearfixedeffectsrespectively.
Notethattherehasbeennocitylevelvariationinthetimingofinitialstay-at-homerules6.
However,aquarantinewasimposedononly31citiesinApril.
AccordingtoAppleMobilitydata,declinesindrivingandwalkinginIstanbul(oneofthequarantinedcities)wasmorethanthedeclineforthewholecountry(seeFigure3).
OnecanexploitthisvariationinpoliciestoidentifytheimpactofstrictermeasuresbyestimatingamodelwithinteractionforquarantinedcitiesasinBrodeuretal(2020).
However,30outof31citiesthatwerequarantinedaremetropolitanmunicipalities.
Sincemetropolitancitydesignationisprimarilyafunctionofthepopulation(morethan750,000)andgiventhatmetropolitan6Citylevelvariationinimplementationofthemeasuresislikelybutisnotobservable,thisvariationispartlycapturedbycityandmonthfixedeffects.
7municipalitiesareallocatedmorefinancialandadministrativeresources,theymaynotbedirectlycomparabletoothercitiesintermsofpre-pandemictrafficcongestionorintheirabilitytoenforcestay-at-homeordersduringthepandemic.
TheonlycitythatwasinquarantineduringAprilthatisnotametropolitanmunicipalityisZonguldak.
ThereasonbehindZonguldak'squarantinewasnotthepopulationsizebutitshighprevalenceofrespiratoryillnesses(Zonguldakisamajorcoalproducer).
Zonguldakhasapopulationofaround500,000andismorelikelytobearepresentativeofnon-metropolitancitiesinthesample.
Usingthisvariation,itispossibletoobtainamoreaccuratepictureoftheeffectofthequarantine.
Irestrictthesampletoonlynon-MetropolitancitiesandestimateaDiff-in-DiffmodelwiththeinteractionforZonguldakforthemonthofApril.
ResultsHere,IfocusonlyonthecitylevelestimatesthatusedatafromDecember2018onwards.
Thecountry-levelestimatesthatusealongertimeperiodarereportedintheAppendix.
AllmodelsareestimatedusingPoissonmodels,coefficientspresentedinthetablesareincidencerate-ratios(avaluebelowonemeansadecreaseintheincidenceoftheoutcomevariable).
AccordingtoTable2,inMarchandApril2020,trafficaccidentswithdeathorinjurydeclinedapproximatelyby20percentand60percentrespectively,comparedtothesamemonthslastyear.
Thedeclineinaccidentswithmateriallossare13and46percent.
Thisresultedinadeclineinreportedinjuriesby13and75percent.
ThedeclineindeathsatcrashscenesareonlyinApril,by15percent,anditisimpreciselyestimated.
Next,IfocusontheDiff-in-DiffestimatesthatusethecityofZonguldakasthetreatedgroupandtherestofthenon-metropolitancitiesasthecontrolgroup.
AccordingtotheresultsinTable3,thedecreasethatcanbeattributedtothestrictermeasuresinAprilareapproximately35percentinaccidentswithdeathorinjury,32percentinaccidentswithmateriallossand19percentininjuries.
Therehasnotbeenasignificantdeclineindeathsatthecrashsceneduetostrictermeasures.
AdjustedDeathsandInjuriesTheestimateddeclineintrafficfatalitiesreportedabovemostlikelytounderrepresenttheactualdeclineinliveslostduetotrafficaccidents.
Sincecity-levelreportsonlyrecorddeathsatthecrashscene,thefatalitiesshouldbeadjustedtoreflecttheactualtrafficaccidentdeaths.
AccordingtothelatestavailablestatisticsinMTIreports,in2018,3368peoplediedatthesceneoftheaccidentwhile3307diedwithin30daysduetoinjuriescausedbytheaccidents.
Therefore,theactualnumberofdeathsislikelytobetwicethenumbersthatareavailableinthemonthlyreports.
Onepossibilitytoadjustthefatalitiesistore-classifyaportionofthereportedinjuriesasdeath.
Usinghospital-leveladministrativedatafromTurkey,Demiraletal.
(2010)estimatethat3.
9%ofindividualswhowereadmittedtohospitalswithtrafficinjuriesdiewithin30days.
This,howevermostlikelynottoreflectthefatalityrateforallinjuredsinceonlyseverecaseswouldbeadmittedtothehospital.
Anotherpossibility,whichIimplementforthecountry-levelregressions,istouse8theratioofallfatalitiestofatalitiesatthecrashscenetocalculateanadjustmentratio.
ThesestaticsareavailableinMTIreportssince2015whichgivesamedianadjustmentrateof2.
027.
Forthecity-levelanalysis,Iadjustedbothfatalitiesandinjuries.
Comparingthetotalofmonthlyinjuriesfor2018fromthecitylevelreports(310,109)whichdoesnottakeintoaccountpost-crashdeathsandonethatdoesfromMTIforthesameyear(307,071),Ichoseapost-crashfatalityrateforinjuredas1%andincreased(decreased)monthlydeaths(injuries)by1%8.
TheestimatedresultsforadjusteddeathsandinjuriesforthewholesampleandthesmallerDiff-in-DiffsamplearereportedinTable4.
Theestimateddeclineintrafficaccident-relateddeathsis43percentinApril,andtheestimateddeclineininjuriesisaround65percent.
AccordingtoDiff-in-Diffresults,deathsdeclineby72percentduetostrictermeasures,andinjuriesby19percent.
CheckforPre-TrendIre-estimatethemodelsbyincludingindicatorsformonthsfromDecember2019toApril2020,theresultsarepresentedinTable5forthewholesampleandinTable6fortheDiff-in-Diffsample.
Accordingtoresults,thetrafficaccidentoutcomeswereslightlyhighercomparepreviousyearuntilthestay-at-homeruleswereimplemented.
DiscussionEveryyear1.
35millionpeopledieandupto50millionareinjuredintrafficaccidents.
90%oftheselossesoccurinlow-andmiddle-incomecountries(WHO,2018;Puvanachandraetal.
2012).
Highpopulationdensityandroadcongestionaremajorfactors(Ashrafetal.
(2019);GolobandRecker(2003);Jonesetal.
(2003)).
In2019,174.
590accidentswithdeathsorinjurieshappenedinTurkeywhere285,638peoplewereinjured,and2524peoplediedatthecrashscenewhichreflectsonlyhalfofthefatalitiesthatoccurwithin30daysofacrash.
Inadditiontotheunmeasurablepersonallosstotheindividualsandtheirfamilies,thesetrafficaccidentstakeanimmensetollontheeconomyintermsofcostofemergencyresponse,healthcareandlossofproductivity.
Thispaperexaminesthedeclineintrafficaccidentsandrelatedinjuriesandfatalitiesasaresultofthestay-at-homerulesthatwereintroducedtofightrapidlyincreasingCOVID-19casesinTurkey.
TheresultssuggestthatduringAprilwhenthestrictestmeasureswereinplacefortheentiremonth,trafficaccidentsdroppedapproximatelyby60percent,deathsdeclinedby43percentandinjuriesdeclinedby64percentcomparedtoApriloflastyear.
ThisisinadditiontodeclinesinMarchwhenrelativelymilderruleswereineffectduringhalfofthemonth.
ApplyingtheseestimatestostatisticsfromMarchandApril2019,theestimatesimplythatapproximately21,000accidents,17,6000injuriesand200deathsavoidedinonlyoneandahalfmonthsthatthestay-at-home7Isimplycomputeratiooftotalfatalitiestofatalitiesatthecrashsceneforeachmonth.
Themedianis2.
02.
8Also,thenumberofindividualswhodiedafterthecrashisroughlyonepercentofthereportedinjuries.
Notethatthisisroughlyequivalenttousingtheadjustmentrate.
9policieswereinplace9,morethanhalfoftheseindividualswhohadotherwisediedorinjuredarelikelytohavebeenbetween24and64yearold(seeTable8inAppendix).
Incontrast,therewere3174officialdeathsattributedtoCOVID-19betweenMarch10andApril30inTurkey10.
9TotalaccidentsinMarchandApril2019were32,401and31,761respectively.
Adjustednumberofdeathswere319and393,injurieswere20,102and20,552.
ImultipliedthesenumberswiththeestimatesfromTable2andTable4.
10See:https://covid19.
saglik.
gov.
tr/10References:AshrafI,HurS,ShafiqM,ParkY(2019)Catastrophicfactorsinvolvedinroadaccidents:Underlyingcausesanddescriptiveanalysis.
PLoSONE14(10)Brodeur,Abel,NikolaiCook,andTaylorWright.
OntheEffectsofCOVID-19Safer-At-HomePoliciesonSocialDistancing,CarCrashesandPollution.
No.
13255.
InstituteofLaborEconomics(IZA),2020.
Demirel,B.
,Demircan,A.
,Akar,T.
,Kele,A.
,&Bildik,F.
(2010).
lkemizdeTrafikKazalarndalenlerinGerekSaysNedir.
PamukkaleTpDergisi,(2),70-76.
FraserShillingandDavidWaetjen.
SpecialReport(Update):ImpactofCOVID19MitigationonNumbersandCostsofCaliforniaTrafficCrashes.
Availableonline:https://roadecology.
ucdavis.
edu/files/content/projects/COVID_CHIPs_Impacts_updated_415.
pdfGolob,ThomasF.
,andWilfredW.
Recker.
"Relationshipsamongurbanfreewayaccidents,trafficflow,weather,andlightingconditions.
"Journaloftransportationengineering129.
4(2003):342-353.
Jones,A.
P.
,Haynes,R.
,Kennedy,V.
,Harvey,I.
M.
,Jewell,T.
,&Lea,D.
(2008).
GeographicalvariationsinmortalityandmorbidityfromroadtrafficaccidentsinEnglandandWales.
Health&place,14(3),519-535.
Naci,Huseyin,andTimothyD.
Baker.
"ProductivitylossesfromroadtrafficdeathsinTurkey.
"InternationalJournalofInjuryControl&SafetyPromotion15.
1(2008):19-24.
Puvanachandra,P.
,Hoe,C.
,zkan,T.
,&Lajunen,T.
(2012).
BurdenofroadtrafficinjuriesinTurkey.
Trafficinjuryprevention,13(sup1),64-75.
Sarla.
G.
COVIDDairies:AnIndianPerspective.
JMedicalResearchandSurgery2020.
1(3):1-4.
Wesson,H.
K.
,Boikhutso,N.
,Bachani,A.
M.
,Hofman,K.
J.
,&Hyder,A.
A.
(2014).
Thecostofinjuryandtraumacareinlow-andmiddle-incomecountries:areviewofeconomicevidence.
Healthpolicyandplanning,29(6),795-808.
WorldHealthOrganization.
"RoadTrafficInjuries".
https://www.
who.
int/health-topics/road-safety#tab=tab_1WorldHealthOrganization.
Globalstatusreportonroadsafety2018.
WorldHealthOrganization,2018.
11Table1:SummaryStatisticsofAccidentOutcomes:FromDec2018toFebruary2020meansdminmaxAccidentsw.
Death/Injury159.
8562208.
211531535Accidentsw.
MaterialLoss232.
358515.
216354367Deaths2.
3420482.
93979019Injuries259.
2876304.
986152256March2020meansdminmaxAccidentsw.
Death/Injury131.
0494179.
289771165Accidentsw.
MaterialLoss202.
7901449.
737963128Deaths1.
8024692.
28265109Injuries202.
9259250.
1352101593April2020meansdminmaxAccidentsw.
Death/Injury68.
8148188.
587124549Accidentsw.
MaterialLoss118.
9877259.
753471775Deaths1.
7530862.
244609010Injuries95.
64198116.
8268572712Table2:PanelPoissonModelswithCityFixedEffects(1)(2)(3)(4)Accidentsw.
Death/InjuryAccidentsw.
MaterialLossDeathsInjuriesMar-20200.
786***0.
874***1.
0860.
778***(0.
0128)(0.
0112)(0.
158)(0.
0101)Apr-20200.
407***0.
534***0.
8560.
359***(0.
00763)(0.
00771)(0.
122)(0.
00552)Observations1377137713771377Exponentiatedcoefficients;StandarderrorsinparenthesesModelsincludemonth,yearandcityfixedeffects*p<0.
10,**p<0.
05,***p<0.
01Table3:PoissonDiff-in-DiffEstimates(1)(2)(3)(4)Accidentsw.
Death/InjuryAccidentsw.
MaterialLossDeathsInjuriesApr-20200.
469***0.
605***0.
7600.
396***(0.
0181)(0.
0217)(0.
183)(0.
0179)Zonguldak1.
292***1.
359***1.
1001.
220***(0.
0957)(0.
114)(0.
0968)(0.
0891)Apr-2020#Zonguldak0.
754***0.
684***0.
00000306***0.
812***(0.
0216)(0.
0200)(0.
00000315)(0.
0303)Constant48.
12***78.
24***0.
448***90.
19***(4.
014)(7.
082)(0.
112)(7.
757)Observations867867867867Exponentiatedcoefficients;StandarderrorsinparenthesesSample:non-metropolitancitiesModelsincludemonth,yearfixedeffects,StdErrorsareclusteredatcitylevel*p<0.
10,**p<0.
05,***p<0.
0113Table4:PanelPoissonModelswithCityFixedEffects-AdjustedDeathandInjury(1)(2)(3)(4)DeathsInjuredDeathsInjured0.
7960.
790***Mar-20200.
9020.
778***(0.
117)(0.
0366)(0.
0875)(0.
0102)Apr-20200.
574***0.
359***0.
546***0.
365***(0.
0587)(0.
00554)(0.
0854)(0.
0183)Zonguldak1.
1251.
219***(0.
0856)(0.
0886)Mar-2020#Zonguldak1.
742***1.
019(0.
136)(0.
0326)Apr-2020#Zonguldak0.
286***0.
813***(0.
0359)(0.
0307)Constant1.
22885.
74***(0.
170)(7.
409)Observations13771377867867Exponentiatedcoefficients;StandarderrorsinparenthesesModelsincolumns1and2areestimatedwithPanelPoissonwithcityfixedeffectsandincludemonth,yearandcityfixedeffects.
Modelsin(3)and(4)includemonthandyearfixedeffectsandclusteredstandarderrorsatcitylevel.
*p<0.
10,**p<0.
05,***p<0.
0114Table5:Pre-TrendPoissonwithFixedEffects(1)(2)(3)(4)Accidentsw.
Death/InjuryAccidentsw.
MaterialLossDeaths(Adjusted)Injuries(Adjusted)Dec-20191.
100***1.
0101.
1111.
058***(0.
0142)(0.
00988)(0.
0828)(0.
0109)Jan-20201.
070***0.
898***0.
9011.
067***(0.
0143)(0.
00875)(0.
0715)(0.
0113)Feb-20201.
056***1.
025**1.
1111.
033***(0.
0148)(0.
0109)(0.
0927)(0.
0116)Mar-20200.
836***0.
834***0.
8970.
818***(0.
0110)(0.
00881)(0.
0701)(0.
00864)Apr-20200.
433***0.
510***0.
571***0.
377***(0.
00695)(0.
00638)(0.
0483)(0.
00505)Observations1377137713771377Exponentiatedcoefficients;StandarderrorsinparenthesesModelsincludemonth,yearandcityfixedeffects*p<0.
10,**p<0.
05,***p<0.
0115Table6:Pre-TrendPoissonDiff-in-DiffEstimates(1)(2)(3)(4)Accidentsw.
Death/InjuryAccidentsw.
MaterialLossDeaths(Adjusted)Injuries(Adjusted)Dec-20191.
056*0.
9541.
305**1.
017(0.
0347)(0.
0359)(0.
140)(0.
0450)Zonguldak1.
259***1.
340***1.
1271.
160**(0.
0933)(0.
113)(0.
0852)(0.
0847)Dec-2019#Zonguldak1.
061**1.
163***0.
9001.
195***(0.
0274)(0.
0264)(0.
0801)(0.
0392)Feb-20200.
9731.
146***1.
0470.
946(0.
0284)(0.
0517)(0.
129)(0.
0385)Feb-2020#Zonguldak1.
178***1.
067*0.
693***1.
285***(0.
0345)(0.
0353)(0.
0616)(0.
0486)Jan-20201.
104***0.
898***1.
1081.
135***(0.
0316)(0.
0288)(0.
153)(0.
0366)Jan-2020#Zonguldak1.
214***1.
073***1.
372***1.
518***(0.
0320)(0.
0234)(0.
128)(0.
0390)Mar-20200.
831***0.
876***0.
8600.
829***(0.
0255)(0.
0412)(0.
0950)(0.
0348)Mar-2020#Zonguldak1.
094***0.
900***1.
739***1.
071**(0.
0275)(0.
0237)(0.
142)(0.
0352)Apr-20200.
451***0.
583***0.
590***0.
384***(0.
0155)(0.
0212)(0.
0872)(0.
0170)Apr-2020#Zonguldak0.
774***0.
694***0.
285***0.
854***(0.
0219)(0.
0213)(0.
0360)(0.
0328)Constant47.
41***77.
50***1.
570***83.
73***(3.
534)(5.
931)(0.
160)(6.
257)Observations867867867867Exponentiatedcoefficients;StandarderrorsinparenthesesSample:non-metropolitancitiesModelsincludemonth,yearfixedeffects,StdErrorsareclusteredatcitylevel*p<0.
10,**p<0.
05,***p<0.
0116APPENDIX:Table7:PanelPoissonModels–CountryLevelResults(1)(2)(3)(4)AllAccidentsInjuredDeathsDeaths(Adjusted)Mar-20200.
784***0.
817***1.
0181.
005(0.
00647)(0.
00820)(0.
110)(0.
0788)Apr-20200.
404***0.
343***0.
8960.
831**(0.
00396)(0.
00444)(0.
0978)(0.
0653)Constant10830.
3***16986.
8***219.
5***428.
1***(41.
20)(55.
18)(6.
590)(11.
04)Observations88888852Exponentiatedcoefficients;Standarderrorsinparentheses.
Adjusteddeathsestimatesusedatafrom2015onwards.
Adjustmentwasappliedtoyears2019and2020wherenumberofdeathsafterthecrashisnotavailable.
Modelsincludemonth,yearfixedeffects*p<0.
10,**p<0.
05,***p<0.
0117Table8:TrafficAccidentDeathsandInjuriesbyAgeGroupsAgegroups0-910-1415-1718-20YearKilledpersonsInjuredpersonsKilledpersonsInjuredpersonsKilledpersonsInjuredpersonsKilledpersonsInjuredpersons2013202193039513405116155991872354020141812092097140831131684719525253201538722738186145852911787543627191201633922530159143202621770942426615201733122342175146342811689438825247201830923765156155242001632236925995Agegroups21-2425-6465+UnknownYearKilledpersonsInjuredpersonsKilledpersonsInjuredpersonsKilledpersonsInjuredpersonsKilledpersonsInjuredpersons20132242903123541573014881506019159020142743119321891591774521658523100120155223405942051694981474176562981920164923308142121703001365184114784620175333226442661693751272178981811729201849032515377117207713492003131842Source:GeneralDirectorateofPublicSecurityandGeneralCommandofGendarmerieNote:TablewasretrievedfromTurkishStatisticalInstituteWebsite