Bicycle-motor-vehicles crashes and bikeway network in Berlin (2011-2015)

AbstractThis dataset compiles georeferenced crash records, bikeway infrastructure data, and mobility exposure indicators to analyze the relationship between bikeway design and cyclist safety in Berlin. The data were collected and processed for the study “Bikeway Design as Determinant of Bicycle–Motor-Vehicle Crashes in Berlin” (2025). The dataset covers 25,688 bicycle–motor-vehicle crashes reported by the Berlin Police Department between January 2011 and December 2015, matched with six bikeway typologies across the city road network, and aggregated by 15 spatial observation units. Supplementary data on bicycle and motor vehicle trip volumes and population counts were derived from the Berlin Senate and the Berlin-Brandenburg Statistical Office.Spatial and Temporal CoverageGeographic coverage: Berlin, GermanySpatial units: 23 administrative districts aggregated into 15 observation groupsTemporal coverage: January 1, 2011 – December 31, 2015Data SourcesBerlin Police Department: Geocoded, anonymized crash records (38,917 total; filtered to 25,688 bicycle–motor-vehicle collisions).Berlin Senate Department for Urban Development and the Environment (2008, 2013): Shapefiles of bikeway networks and typologies.Berlin-Brandenburg Statistical Office (BBSO): Population and mobility survey data.Berlin Travel Survey (SrV 2008; 2013): Modal split and trip volumes.VariablesCrash data (dependent variable):Number of bicycle–motor-vehicle crashes (N), per bikeway typology and spatial unitInjury severity levels: fatalities, severe, moderate injuriesBikeway infrastructure (independent variables):Bikeway length (km) per typology and observation unitSix typologies:PBP – Protected bikeway paths (natural barrier, >5 m from motor traffic)DBL1 – Dedicated bicycle lanes, kerb-adjacent, continuous lineDBL2 – Dedicated bicycle lanes adjacent to parking, dashed lineSUBBB – Shared bus–bike lanesSUBBP – Side-path bicycle lanes adjacent to sidewalksNBI – No bicycle infrastructure (mixed traffic)Mobility exposure data:Mtrip – Estimated annual motorized vehicle trips per spatial unitBtrip – Estimated annual bicycle trips per spatial unitPopulation:Number of inhabitants per spatial unit (BBSO 2013).Data StructureFormat: CSV and shapefile (GIS) tablesObservations:25,688 bicycle–motor-vehicle crashes (row-level crash dataset, anonymized)90 records in long format (15 observation units × 6 bikeway typologies) for regression analysisSpatial resolution: 3 m buffer around bikeway axis used to assign crashes to typologiesMethodsCrash records were filtered to include only collisions between one bicycle and one motorized vehicle (excluding public transport). Crashes were spatially matched to bikeway typologies using GIS shapefiles. Bikeway lengths were calculated per typology and spatial unit. Modal split and population data were used to estimate bicycle and motor vehicle trips. These data were aggregated into a panel suitable for multivariate linear regression and negative binomial generalized linear models.Use and ApplicationsThe dataset enables:Comparative safety analysis of bikeway typologiesReplication of regression models linking bikeway length, trip volumes, and crash riskExposure-adjusted crash risk estimation for urban road safety researchPolicy analysis of infrastructure retrofit scenarios (e.g., converting mixed-traffic lanes into DBL1 or SUBBB).LimitationsAggregated to 15 observation groups, which may obscure segment-level variation.Modal split trip estimates rely on survey data rather than continuous traffic counters.SUBBP and DBL2 models were less robust due to heterogeneity of design features.LicensingThe dataset is released under a CC BY 4.0 license, allowing reuse with attribution.