The impact of geocoding method on the positional accuracy of residential burglaries reported to police

Abstract

Purpose The purpose of this paper is to better understand the variability in burglary geocoding positional accuracy between United States Census Topologically Integrated Geographic Encoding and Referencing (TIGER)-based street geocoding and results produced using reference data made publicly available by Google. Design/methodology/approach This research compares the Euclidian distance between ground-truthed burglaries and results produced using two different geocoding reference data sets: TIGER-based street geocoding and publicly available data within Google Earth. T-tests and z-tests are used to discern whether positional errors are statistically significant. Findings Both within suburban and urban jurisdictions, Google outperformed street geocoding in terms of positional accuracy. Positional errors on average were 1/4th as large for Google in a suburban setting and 1/5th as large in an urban setting compared to street geocoding. Practical implications Police departments that are relying on street geocoding techniques may achieve improved spatial precision by using Google’s reference data if they contain parcel-level information. Moreover, relying on less precise spatial referencing methods may place burglaries in locations where the events do not actually occur or cluster. Originality/value This is the first analysis of law enforcement data to examine the positional accuracy of geocoded offense data using Google Earth compared to the commonly used street geocoding method of interpolation.

Testing the Spatial Accuracy of Address-Based Geocoding for Gunshot Locations

“…If this is the case, this could result in a larger proportion of cases that were coded as having some transcription errors, and could result in larger distance errors. We cannot say whether the resulting 10% of cases that had transcription errors is the norm, but prior analysis of geocoding accuracy would not have uncovered this type of error (with perhaps the exception of Mazeika & Summerton (2017) if they were unable to find a building at a listed address).…”

“…While this threshold appears to be quite reachable, a study by Cohen (2006) noted large yearly differences, with a low 72% for one year and a high 91% for another year. The latter number was echoed when Mazeika and Summerton (2017) successfully geocoded about 93% of burglaries in their study. Using google instead of the street geocoding they achieved a much higher match rate, over 99%.…”

“…There are a few different methods for geocoding currently employed. It has traditionally been done through the use of street geocoding, where a database matches street segments to locations and then interpolates the exact location based on the range of addresses on the street segment (Mazeika & Summerton, 2017). There are multiple databases that can be used for street geocoding, in the US the TIGER database covers the whole country and is the most common (Hart & Zandbergen 2013).…”

“…Hart and Zandbergen (2013) used address point data compared with street geocoding to determine the positional accuracy of different street geocoding and noted that local street line reference data outperformed the national data such as TIGER. With the fast paced technology development there are now multiple commercial services that can provide geocoding, and although Mazeika and Summerton (2017) note that this likely was the least common method within the criminal justice system this is subject to rapid change, and currently the extent of its usage is unknown.…”

“…The true error associated with crimes that are more difficult to geocode, such as shootings, remain to be showed. The finding that the positional accuracy of geocoding was increased through the use of new technology, google (Mazeika & Summerton 2017), also points to a need for further study on how technological innovation, such as GDT, may be of use in accurate geocoding.…”