Why Bike-Friendly Cities are Safer for all Road Users
Wesley Marshall
ABSTRACT Giventhegrowingevidencesuggestingthatcitieswithhigherbicyclingratesfindlowerfatalityrates,weexamineroadsafetydatafrom24Californiacities.Thisanalysisincludedaccountingforcrashesacrossallseveritylevelsbutalsoforthreedifferentclassesofroadusers:vehicleoccupants,pedestrians,andbicyclists.Additionally,welookedatissuesofstreetandstreetnetworkdesigntoseewhatrolethesecharacteristicsmightplayinaffectingroadsafetyoutcomes.Overall,highbicyclingcitiesgenerally showamuch lower riskof fatalcrashes forall roaduserswhencompared tomostof theothercities inourdatabase. The fact that thispatternof low fatality risk isconstant for all classes of road users strongly suggests that the crashes are taking place at lowerspeeds.Themost notable difference foundbetween the safer and less safe citieswas thedensity ofstreet intersections.Whilewe do not yet have the data to fully disentangle the various contributingfactors, our results strongly suggest that safety benefits for all road users can be derived from acombination of the same steps that tend to attract more bicyclists. In other words, improving thestreetstobetteraccommodatebicyclesmayinfactleadtoaself‐reinforcingcyclethatcanhelpenhanceoverallsafetyforallroadusers.
INTRODUCTION
Davis,California,oftenreferredtoasthebicyclecapitalofAmericasincebecomingthefirstcitytogain"platinum"statusfromtheLeagueofAmericanBicyclists,shouldalsoberenownedforanotherreason:roadsafety.From1996through2007,theyearsexaminedforthisstudy,Davisendured only nine fatal road crashes, of which only three occurred on regular, non‐limitedaccessstreets.AnddespiteagreaterpercentageofpeoplebikingtoworkthananyothercityintheUnitedStates,notasingleoneofthesefatalcrashesinvolvedabicyclist.WithafatalcrashrateinDavisof lessthan1.5per100,000residents,farfewerpeoplearekilledontheirroadsthanintheU.S.asawhole,whichaverage14.5fatalitiesper100,000residents.AnotherAmerican city recognized as a “platinum”bicycling city, Portland,Oregon, increasedbicyclemodesharefrom1.2%in1990to5.8%in2000.Atthesametime,thetotalnumberofroad fatalitieswent fromaveragingover60peryeararound1990 to fewer than35peryearsince2000(1).Moreover,therewereonly20totalroadfatalitiesinPortlandin2008,whichisaremarkablesafetyrecord(3.6fatalitiesper100,000residents)foracityofover550,000people.Such fatal crash rates compare extremely favorably with the countries reporting the lowestcrashratesintheworldsuchastheNetherlandsat4.9per100,000residents(2).These outcomes are not uncommon; other researchers have reported notable decreases infatality rates in cities that have successfully increased their bicycle mode share (3, 4).Conventionalthinkingaboutroadsafetywouldsuggestthattheoutcomeoflowerroadfatalityrateswithmorebicycle riderswouldbeunlikely since, ingeneral,bicycle ridersexperienceamuchhigherfatalityratepermiletraveled.Butgiventhegrowingevidencesuggestingthatthisisnot thecase,weexamineroadsafetydata from24Californiacities in thispaper togarnerevidenceastowhycitieswithhighratesofbicycleusetypicallyseelowerratesofroadfatalityforallroadusers.Inothertobetterunderstandthetrendsinthesecities,wenotonlyexaminethe number of crashes of different severity levels but also the relative risk of a fatality or asevere injurygiven the fact thata crashoccurred. Theseanalyseswereconducted for threeclassesof roadusers ‐pedestrians,bicycle ridersandvehicleoccupants ‐ inorder tohelpusunderstandiftheunderlyingpatternsweresimilarforallroadusertypes.Wealsousedcensusdataasa roughestimateof thenumberofpeoplewalking,bikinganddriving ineach city inorder to gain a better understanding of the relative exposure rates in these cities for thedifferentclassesofroadusers.Finally,welookedatissuesofstreetandstreetnetworkdesigntoseewhatrolethesecharacteristicsmightplayinaffectingroadsafetyoutcomes.
STUDYBACKGROUND
This researchwas based on an initial database of over 150 California cities. We focused onCaliforniacitiesinordertohelpmaintainconsistencyinthedata,especiallyincomparinginjuryseverity outcomes. The earlier papers based on this dataset concentrated on the streetnetworkscharacteristicsof24oftheseCaliforniacitiesrepresentingtwelvemedium‐sizedcitieswithgoodsafetyrecordsandtwelvewithpoorsafetyrecords(5,6). Inthisstudy,wefurthersub‐dividethegroupoftwelvesafercities intothefollowingthreegroupsof fourcitesbaseduponbicyclemodeshare:highbicyclingcities,mediumbicyclingcities,andlowbicyclingcities.Thecitiesincludedwere:
Group1:HighestBicyclingSaferCities
• Berkeley •Davis • Chico •PaloAlto
Group2:MediumBicyclingSaferCities
• Alameda •SantaBarbara• SanLuisObispo •SantaCruz
Group3:LowBicyclingSaferCities
• Cupertino •Cupertino• Danville •SanMateo
Group4:LessSafeCities
• Antioch •Redding• AppleValley •Rialto• Carlsbad •Temecula• Madera •Turlock• MorganHill •Victorville• Perris •WestSacramento
Journeytoworkdatawascollectedalongwithstreetnetworkmeasures,streetcharacteristics,socioeconomicdata, traffic flow information, andover230,000 individual crash records fromelevenyearsofcrashdata. Allof this informationwasgeo‐coded inaGISdatabasewith theintentionoffacilitatingamorecomprehensivespatialanalysis.
LITERATUREREVIEW
Fewstudieshavespecificallylookedathowsafetyvariesforallroadusersdependingupontheamountofwakingorbikingthat isoccurring. Transitusagehowever isonemodethathas infactbeenevaluatedintermsofoverallroadsafety.Inaninternationalstudy,KentworthyandLaube concluded that citieswithhigher transitusealso tended tohave loweroverall fatalityrates(7).Litman,inaseparatestudy,foundthatthepercapitafatalityratesofU.S.citieswerelowerwithincreasedtransituse(8).Onereasonbehindtheseresults,astheauthorspointout,isthatmoretransitusetendstolowertheoverallamountofvehicleuse.
Ifreducingvehicleusethroughmoretransitusagecanhelpintermsofoverallroadsafety,thentheideathatincreasesinbikingandwalkingcanhaveasimilareffectispromising.However,itisimportanttounderstandthatthefatalityrateintermsofmilestraveledforvehicleoccupantsisapproximatelytentimesthatoftransituserswhilemoststudieshaveshownthatthefatalityrates in terms ofmiles traveled for biking andwalking are higher than for driving (8). Onepotentially confounding factor is that calculating safety on a per‐mile basis might not beappropriate given that most biking and walking trips are generally shorter in distance thandrivingtrips.Anotherpointtoconsideristhehandfulofstudiesfindinganincreaseinoverallbicyclistandpedestriansafetyemergingwith increasingnumbersofbikersandwalkers. Thethinkingisthatadriverchangeshisorherexpectationsbasedupontheperceivedpossibilityofencounteringabicyclistorpedestrian.Sowhenthenumberofbikersandwalkersincreasestothepointwheredriversbegintoexpectconflicts,thedriver'sbehaviorbeginstochangeforthebetter.
For example, a 1996 study by Lars Ekman found no linear association between bicyclistexposure and conflict rate in a comprehensive study conducted in Sweden (9). To bemorespecific,Ekmandeterminedthattheconflictrateforanindividualbicyclistwashigherwhenthenumberofbicyclistswaslow,withthisconflictratesubsidingastheflowofbicyclistsincreased.Intermsofconflictratesforabicyclist,thenumberofbicyclistswasmoresignificantthanthenumberofvehiclesontheroad.Ekmanalsofoundthattherisktopedestrianswasnotaffectedbythenumberofpedestrians.
AnotherexamplesistakenfromthecityofCopenhagen,whereitwasfoundthatbetween1990and 2000, a 40% increase in bicycle kilometers traveled corresponded to a 50% decrease inseriously injuredbicyclists (4). And ina2003studyofCalifornia cities,Peter Jacobsen foundresultssubstantiatingthisideaofsafetyinnumbers.Basedon68Californiacities,butonlyoneyearof crashdata, the results showed that the individual chanceof abicyclist or pedestrianbeingstruckbyacardropswithmorepeoplebikingandwalking(3).
Theseresultsare interestingbecauseconventionalwisdomlinksan increase inexposurewithan increase in risk. Although not easily transferable to overall road safety, these studies dobegintosuggestsomeexplanationastowhyplaces likeDavis,Portland,andtheNetherlandsmightbesafety thanplaceswith lowerbikeuse. While switching fromdriving to transithasbeenshowntodecrease individualrisk,switchingfromdrivingtobikingorwalkingwould,onaverage, increase individual risk. However, that average risk number does not explicitlyconsidersituationswerethereisacriticalmassofbikersandwalkersthatmaybeabletofindbettersafetyinlargernumbers.Inthosecases,theideathatswitchingfromdrivingtobikingorwalkingcanactuallyreduceone’sindividualriskisapossibility.Intermsofoverallroadsafety,strategiesknowntoincreasebikingandwalkingsuchastrafficcalminganddecreasingvehiclespeedshavealsobeenshowntoleadtobetterroadsafetyoutcomes(10,11).Together,suchstrategiescouldhelptoreduceoverallvehiclemilestraveled(VMT),whichcouldalsoplayaroleinimprovingroadsafety(12).
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
6.0%
7.0%
8.0%
9.0%
High Bicycling Medium Bicycling Low Bicycling
SAFER CITIES LESS SAFE CITIES U.S. Average
Mod
e Sh
are
(200
0 C
ensu
s)
Biking Mode Share Pedestrian Mode Share Transit Mode Share
RESULTS
For the purposes of this study, the crashes analyzed only include those that did occur onsurface streets and not those on limited access highways. This was done in order to fairlycomparecrashesonroadswherewalkingandbikingwouldbereasonablyexpected. Tables1and2summarizethedataforthisresultssection. ModeSharesBasedon2000Census journey‐to‐workdata,Figure1depictsbiking,walking,and transituseforeachsetofcities.AlsoshownistheU.S.averageforbiking,walking,andtransituseat0.4%,2.9%, and 4.6%, respectively (13). The high‐bicycling cities in our study havemore than 20timesmore biking than the U.S. average,more than 2.5 timesmorewalking, and 1.5 timesmoretransituse.Thelow‐bicyclingcitiesandlesssafecitiesmatchtheU.S.averageforbikingandfallbelowtheU.S.averageforwalkingandtransituse.Overallvehiclemodeshareiswellunder80%forthehigh‐bicyclingcities,82%forthemedium‐bicyclingcities,andover94%forthelow‐bicyclingcitiesandthelesssafecities.Figure1 Bicycle,Pedestrian,andTransitModeShares(2000Census)
RoadSafetyIn terms of road safety, the differences are not always found in terms of the overall crashnumbers.Infact,thecitieswiththelowerfatalityrateswouldseemtobelesssafeifweonlylookedatoverallcrashfrequency.Thisisanimportantdistinctionbecausemanysafetystudiesoftenfocusontheoverallnumberofcrashesandignorecrashseverity.Inourresults,anotherimportant difference seems to be in what is happening after the crashes occur. The crashseverity risk outcomes ‐ based upon the percentage of crashes for each road user type thatresult ina fatality ‐showthat ifyouare inacrash inoneof theGroup4cities, thenyouaremuchmore likely to die than if the crash tookplace in a city fromoneof theother groups.Overall,theriskofafatalityshouldacrashoccurissimilarforthethreegroupsofsafercitiesforeach roaduser type. For the less safecities, thechanceofavehicleoccupantorpedestriancrashresultinginafatalityisoverfourtimesgreaterthanwhatwefoundineachofthesafergroupsofcities.Moreover,thechanceofabicyclecrashresultinginafatalityisover11timesgreaterinthelesssafecitiesthaninthesafercities.Anotherkeyconsiderationinbetterassessingsafetyisconsideringrelativeexposure.Withtheintention of getting a better handle on the relative amounts of driving, biking, walking, andtransit use in these sets of cities, we used a simple road user exposuremetric inwhichwemultipliedcitypopulationbymodeshare to finda roughnumberof travelersbyeachmode.This is similar to themethod used by Jacobsen; in his study, he assumed that even thoughjourney‐to‐worktripsrepresentasmallpercentageoftotaltrips,thepercentageofeachmodefoundforcommutersisproportionaltoalltrips(3).Thoughthisexposuremetricisadmittedlyimprecise and might be inaccurate if we were interested in absolute rates for vehicle,pedestrian, and bicycle safety, it should function adequately as a proxy toward finding therelativesafetyratesforthese24cities.Toput thisapproach intocontext,Figure2depicts the fatalcrashesnotoccurringonsurfacestreetsovertheelevenyearstudyperiodforonecityfromthehighestbicyclinggroup,SantaBarbara,andonefromthelesssafegroupsofcities,Rialto. Thesetwocitieshavealmostthesamelevelofpopulation(~92,000)livingatalmostthesamepopulationdensity(~5,000peoplepersquaremile).Despitethesesimilarities,bicyclingmodeshareinSantaBarbaraisover3.6%(onthelowendofoureighthigherbicyclingcities)whilebicyclingmodeshareinRialtoisnearlynegligibleat0.2%. Walkingmodeshare isover6.5% inSantaBarbaraand1.3% inRialto. Interms of fatality rates, Santa Barbara had 19 vehicular deaths with over 78,000 estimatedvehicleusersforarateof2.2vehicledeathsperyearper100,000driverswhileRialtohad68vehiculardeathsandover88,000estimatedvehicleusers fora rateofover7.0driverdeathsperyearper100,000drivers. Forwalking,SantaBarbaraexperienced16deathsoverelevenyears with over 6,000 estimated walkers for a rate of 24.2 pedestrian deaths per year per100,000pedestrians.Rialtohad39deathswithlessthan1,200estimatedwalkersforarateofalmost 300 pedestrian deaths per year per 100,000 pedestrians. Santa Barbara also had anestimated3,356estimatedbicyclistswithonly twodeathsoverelevenyears fora rateof5.4bicyclistdeathsperyearper100,000bicyclists.ForRialto,wefindonefewerbicyclistdeathbutonly165estimatedbicyclistsforarateof55.1bicyclistdeathsperyearper100,000bicyclists.
HighBicycling MediumBicycling LowBicycling
Population (2000averagepercity) 70,328 65,742 61,087 59,845PopulationDensity (peoplepersq.mi.) 6,037 5,364 5,808 2,673Income (2000average) 51,669 46,579 81,721 46,408
VehicleModeShare 76.3% 82.0% 94.0% 95.8%
BikingModeShare 8.1% 3.4% 0.5% 0.6%PedestrianModeShare 7.5% 6.6% 1.8% 1.8%TransitModeShare 8.2% 8.0% 3.7% 2.0%
EstimatedNo.ofBicyclists 5,697 2,227 299 345EstimatedNo.ofPedestrians 5,268 4,352 1,082 1,060EstimatedNo.ofDrivers 53,625 53,908 57,422 57,302(estimatesbaseduponmodeshare&population)
VehicleFatalities 10.3 11.3 6.5 37.8VehicleSevereInjuries 61.5 52.3 52.5 83.1VehicleOtherInjuries 2,315.5 1,878.5 1,861.3 1,673.0VehicleTotalInjuries 2,387.3 1,942.0 1,920.3 1,793.8VehiclePropertyDamageOnly 5,471.8 5,519.8 3,648.8 3,769.5(crashcountsaveragedpercityfor19962007)
VehicleFatalityRisk 0.19% 0.15% 0.14% 0.76%(%chanceofcrashresultinginfatality)
VehicleFatalityRate 1.0 1.1 0.6 10.3VehicleSevereInjuryRate 6.0 5.0 5.1 22.6VehicleOtherInjuryRate 224.3 181.0 168.4 455.0(avg.peryearper100,000estimateddrivers)
PedestrianFatalities 7.8 8.5 4.3 16.8PedestrianSevereInjuries 26.8 33.5 20.0 21.3PedestrianOtherInjuries 292.0 244.3 142.0 102.3PedestrianTotalInjuries 326.5 286.3 166.3 140.4(crashcountsaveragedpercityfor19962007)
PedestrianFatalityRisk 3.07% 3.01% 3.01% 12.67%(%chanceofcrashresultinginfatality)
PedestrianFatalityRate 7.6 10.1 20.4 246.2PedestrianSevereInjuryRate 26.4 40.0 96.0 313.5PedestrianOtherInjuryRate 288.0 291.5 681.5 1,503.9(avg.peryearper100,000estimatedpedestrians)
BicycleFatalities 0.8 1.0 0.0 1.8BicycleSevereInjuries 24.5 32.8 11.8 48.3BicycleOtherInjuries 539.0 398.0 202.3 111.1BicycleTotalInjuries 564.3 431.8 214.0 161.3(crashcountsaveragedpercityfor19962007)
BicylistFatalityRisk 0.14% 0.22% 0.00% 1.36%(%chanceofcrashresultinginfatality)
BicycleFatalityRate 0.7 2.3 0.0 82.9BicycleSevereInjuryRate 22.3 76.4 203.9 2,185.2BicycleOtherInjuryRate 491.5 928.5 3,510.0 5,022.1(avg.peryearper100,000estimatedbicylists)
PedestrianSafety
VehicleSafety
BicycleSafety
GeneralInform
ation
SAFERCITIES LESSSAFECITIES
Table1 SummaryofResultsforCrashesNotonLimitedAccessHighways
Nowifwetakethisanalysistothecitygroups,wediscoverthateventhoughthelesssafecitieshavethelowestnumberofcrashesoccurring,Table1showsthatthesecitiesalsofoundhighervehicleoccupantcrashratesacrossallseveritylevels.Anotherkeyconsiderationisthefactthateventhoughthelesssafecitieshadverylowratesofbikingandwalking,theyalsoexperiencedfarmorebicyclistandpedestrianfatalitiesthantheothergroupsofcities.Forapedestrian,thefatalityrateismorethan24timesgreaterinthelesssafecitiesthanineitherofthecitygroupswithsignificantbiking,almosttentimesgreaterforasevereinjury,andoverfivetimesgreaterforallotherpedestrianinjuries. Forthesafercitieswithlowbicycling,thepedestrianfatalityrateisapproximatelytwicethatfoundinthehigherbikingcities.
Figure2 Bicycle, Pedestrian, and Vehicle Fatalities for Santa Barbara & Rialto
(1996‐2007)
HighBicycling MediumBicycling LowBicycling
MeasureforStreetNetworkDensityIntersectionDensity 114.2 103.2 101.2 62.7(intersectionspersq.mi.)
MeasureforStreetConnectivityLinktoNodeRatio 1.39 1.38 1.25 1.29(#links/#nodesincludingdeadends)
CenterlineMilesofMajorRoads 49.5 45.9 26.9 65.2CenterlineMilesofMinorRoads 144.8 119.2 113.6 210.8TotalCenterlineMiles 199.0 169.6 146.0 281.8(averagepercity)
Sidewalks 50.3% 38.3% 85.6% 48.4%BikeLanes 24.9% 23.6% 38.4% 15.6%OnStreetParking 41.1% 28.4% 42.8% 23.0%(%lengthofarterial/collectortypestreets)
Avg.No.ofLanes 2.7 2.4 3.7 3.1Avg.WidthofRoadwayCrossSection 50.9' 46.9' 59.7' 54.4'(averageonarterial/collectortypestreets)
SAFERCITIES LESSSAFECITIES
StreetNetwork&StreetDesign
Forabicyclist,thefatalityrateismorethan75timesgreaterinthecitieswiththepoorsafetyrecordscomparedtothosewithsignificantbiking,oversixtytimesgreaterforasevereinjury,andover seven timesgreater for all otherbicyclist injuries. The safer citieswith lowbicyclemodeshareshadzerobicyclefatalities.However,intermsofallotherinjuries(includingsevereinjuries)toabicyclist,thecrashrateswereover4timesgreaterinthelowbicyclingsafecitiesthaninthehigherbikingcities.StreetNetworkCharacteristics&StreetDesignOverall,thevariationinrelativefatalityrates,aswellasthefactthatacrashoccurringinoneofthelesssafecitieshasamuchhigherchanceofresultinginafatality,suggestsdifferencesinthestreet network and in the design of the street. The data shown in Table 2 supports thesefindings.Theconstantfactorforallthreegroupsofsafercitieswhencomparedtothelesssafecitieswasintersectiondensity.Forthetwogroupsofhigherbicyclingcities,theyalsotendedtobeslightlymoreconnectedwithfewerlanesandanarrowercross‐sectiononthemajorstreetsthanbothgroupsoflow‐bicyclingcities.Table2 Street&StreetNetworkCharacteristicsInordertotakingacloserlookatintersectiondensitywithrespecttosafety,wedevelopedthegraphsshowninFigures3and4.Forallroadusers,thechancethatacrashwouldresultinafatalitytendedtobelowerforthecitieswithlowerdensitystreetnetworks. Thissametrendwasfoundforvehiclecrashes,pedestriancrashes,aswellasbicyclecrashes.Asfortheotherstreetdesignconsiderations,welookagainatSantaBarbaraandRialtoshownin Figure 2. Overall, Santa Barbara had the fewest average number of lanes on thearterial/collector roads of any city in the database while Rialto averaged almost a full lane
11
1501251007550
0.014
0.012
0.010
0.008
0.006
0.004
0.002
0.000
Intersection Density (intersections per sq. mi)
Ch
an
ce o
f a
Ve
hic
le C
rash
Re
sult
i ng
in a
Fa
tal it
y
High Bik ingMedium Bik ingLow Bik ingLess Safe Cities
City Group
Vehicle Occupant Crash Fatality Risk vs. Intersection Density
more. SantaBarbaraalsohasmorethanthreetimesthe lengthofbike lanesonthesesameroads and about 30%more on‐street parking – all ofwhich seem to play a role in the roadsafetyandbiking/walkingoutcomesforSantaBarbara.Another interesting example is Carlsbad ‐ oneof the less safe cities ‐which also happens tohavethehighestpercentageofbikelanesonthearterial/collectorroadsofallthecitiesinthedatabase with nearly 70% of the total length of these roads having a bike lane present.However, Carlsbad is on the low end of the street connectivity and street network densityrangeandalsohas thehighestaveragenumberof lanespresentonthesemajorroads in thedatabase. Soevenwithahighdegreeofbike lanespresent,Carlsbad’sbicyclemodeshare isonly0.3%.Ontheotherhand,Berkeley‐oneofthehighestbikingcities‐hasoneofthelowestpercentagesofbikelanespresentonthemajorroads.Inthiscase,thedifferencemightbeinthefactthatBerkeleyhasthehigheststreetconnectivityandstreetnetworkdensityofallthecitiesaswellasotherstrategiesforaccommodatingbicyclistssuchasbikeboulevards.Thisiscertainlynottosaythatbikelanesarehazardousbecausethesafer,highbicyclingcitiesdidinfacttendtohavemorebikelanes.Forinstance,DavisandAntiochfindverysimilarpopulationdensities,streetconnectivities,andstreetnetworkdensities,butDavishassignificantlybettersafetyoutcomesandalsohappens to coveralmost2.5 timesmoreof theirmajor roadswithbikelanes.Overall,theresultssuggestthatmanyofthesestreetdesignfactors,alongwiththestreet network measures such as intersection density, seem to work in coincidence towardhelpingcreateanenvironmentwithahigherdegreeofbikingandwalkingaswellasimprovedroadsafety. Figure3 ChanceofVehicleCrashResultinginaFatalityvs.IntersectionDensity
12
1501251007550
0.04
0.03
0.02
0.01
0.00
Intersection Density (intersections per sq. mi.)
Ch
an
ce o
f a
Bik
e C
rash
Re
s ult
i ng
i n a
Fa
tal it
y
High Bik ingMedium Bik ingLow Bik ingLess Safe Cities
City Group
Bike Crash Fatality Risk vs. Intersection Density
1501251007550
0.30
0.25
0.20
0.15
0.10
0.05
0.00
Intersection Density (intersections per sq. mi.)
Ch
an
ce o
f a
Pe
de
stri
an
Cr a
sh R
es u
ltin
g in
a F
ata
lity
High Bik ingMedium Bik ingLow Bik ingLess Safe Cities
City Group
Pedestrian Crash Fatality Risk vs. Intersection Density Figure4 ChanceofaPedestrianorBicyclistCrashResultinginaFatalityvs.Intersection
Density
13
CONCLUSION
Highbicyclingcitiesgenerallyshowamuchlowerriskoffatalityorseverecrashesforallroadusers when compared to many of the cities in our data base. The fact that this pattern isconsistent for all classes of road users strongly suggests that the crashes are taking place atlower speeds in these high bicycling cities. The reason for lower speeds might be due tofeaturessuchastrafficcalmingandotherdesignelementsthatcanhelpattractlargenumbersofbicyclists. Ourstreetdatabasecontainssomehintsofthesetrends‐forexample,thehighbikingcitiestendtohavemorebikelanes,fewertrafficlanes,andmoreon‐streetparking.Atthe same time, large numbers of bicycle users might also help lower vehicle speeds. It isimportanttonotethatthehighbikingcitiesdonotnecessarilyhaveloweroverallcrashrates;rather,theyhavemuchlowerseveritylevelsforthosecrashesthatdooccur.Ourresultsalsoshowthatthereisagroupoffourcitiesthathavebothlowseveritylevelsandlow bike use. These cities represent an interesting hybrid exhibiting some characteristics incommonwithboththehigh‐bicycling/lowfatalitycitiesaswellasthelow‐bicycling/highfatalitycities.Thesefourcitiestendedtohavehighintersectiondensitiessimilartothevaluesfoundinthe high‐bicycling cities; they also tended to have low levels of street network connectivity,moreakintothelow‐bicycling/highfatalitycities.Inotherwords,thissubsetofcitiesfeaturedlocalstreetshigh incul‐de‐sacsbutatarelativelyhighdensity. Thesecitiesalsorevealsomeother unique features thatmight contribute to their lower fatality rates, including far fewermajorroadsthanfoundintheothercitygroups.Overall,thebiggestdifferencefoundbetweenthethreegroupsoflowerfatalitycitiesandthehigh fatality cities was intersection density. The graphs depict the relationships betweenfatality risks and intersection densities for vehicle occupants, pedestrians, and bicyclists,respectively. Our results consistently show that high intersection density appears to be thesinglemostimportantstreetdesignfactoraffectingcrashseverity.However,thereappearstobeotherfactorsatworkinleadingtotheselowerfatalityratesforboththehigh‐bicyclingcitiesand the low‐bicycling/low fatality cities. In the caseof thehigh‐bicycling cities, these factorsmight include theworkdone tomake the streetsattractive tobicyclists aswell as the sheerpresenceofmanybikeriders. Wedonotyethavethedata todisentangle theseeffects,butour results strongly suggest that safety benefits for all road users can be derived from anamalgamationofthestepstakentoattractmorebicyclists;thatis,aslongaswedefinesafetyintermsofreducingfatalityandseverecrashesandnotjustintermsofreducingoverallratesofcrashes.Improvingthestreetstoaccommodatebicyclesmayinfactleadtoaselfreinforcingcyclethatcanhelpenhanceoverallsafetyforallroadusers.Thiscombinationoffactorsseemstogoalongwaytowardoverallsaferandmoresustainablecities.
14
REFERENCES
1. CityofPortlandBureauofTransportation,AnnualReport.2008‐09:Portland,OR.2. OrganisationforEconomicCo‐operationandDevelopment(OECD),QualityofLife:
Transport,inOECDFactbook.2006:Paris.3. Jacobsen,P.L.,Safetyinnumbers:morewalkersandbicyclists,safetywalkingand
bicycling.InjuryPrevention,2003(9):p.205‐209.4. Jensen,N.,CyclePolicy2002‐2012.2002,CityofCopenhagenRoadsandParks
Department:Copenhagen.5. Marshall,W.andN.Garrick,ConsideringtheRoleoftheStreetNetworkinRoad
Safety:ACaseStudyof24CaliforniaCities.UrbanDesignInternationalJournal(accepted),2009.
6. Marshall,W.andN.Garrick,TheEffectofStreetNetworkDesignonWalkingand
Biking(recommendedforpublication).TransportationResearchRecord:JournaloftheTransportationResearchBoard,2009.
7. Kenworthy,J.andF.Laube,MillenniumCitiesDatabaseforSustainableTransport.
2000,InstituteforSustainabilityandTechnologyPolicy,distributedbytheInternationalUnionofPublicTransport.
8. Litman,T.,EvaluatingPublicTransitBenefitsandCosts.2009,VictoriaTransport
PolicyInstitute:Victoria,B.C.9. Ekman,L.,OntheTreatmentofFlowinTrafficSafetyAnalysis‐anonparametric
approachappliedonvulnerableroadusers.2006,LundInstituteofTechnology:Lund,Sweden.
10. Pucher,J.andL.Dijkstra,MakingWalkingandCyclingSafer:LessonsfromEurope.
TransportationQuarterly,2000.54(3).11. Retting,R.,S.Ferguson,andA.McCartt,AReviewofEvidence‐BasedTraffic
EngineeringMeasuresDesignedtoReducePedestrian‐MotorVehicleCrashes.AmericanJournalofPublicHealth,2003.93(9):p.1456‐1463.
12. Litman,T.SmartGrowthSafetyBenefits.2007[cited;Availablefrom:
http://www.planetizen.com/node/28523.13. Pisarski,A.,CommutinginAmericaIII:TheThirdNationalReportonCommuting
PatternsandTrends,NationalCooperativeHighwayResearchProgram(NCHRP)Report550,TransitCooperativeResearchProgram(TCRP)Report110.2006,TransportationResearchBoard:Washington,D.C.
