Error Consideration for Geocoding Police Reported Collision Data in California

Geographic Information Systems (GIS) are frequently used to analyze collision data. In order to utilize GIS, the data must be geocoded, or assigned a latitude and longitude coordinate by translating a descriptive location onto street network data. However, the ability for accurate spatial analysis can be limited by geocoding errors that may occur due to limitations in data collection technologies, incorrect data entry due to human error, or inaccurate street reference data. In the state of California there is an increased opportunity for data entry errors, given the long sequence of events and resulting paper trail that is required prior to finalizing each collision record. Data entry errors can occur during the initial traffic collision report completion, statewide database entry, state highway reference location input, or during a separate process to geocode fatal collisions. These data entry errors are incorporated into any geocoding process and frequently cause geocoding errors; but even in the absence of data entry errors, discrepancies in street network data can also result in geocoding inaccuracies. The objective of this paper is to summarize the sources of geocoding errors that occur before and after collision data is compiled into the California state database and the federal database of fatal collisions. Consideration is also given to the potential errors that can arise from the use of Global Positioning System coordinates as an alternative to geocoding. Finally, the impact of geocoding errors on traffic safety analysis is discussed in the context of specific applications currently available in California.