Skip to main content
SearchLoginLogin or Signup

Geospatial technology and crime analysis

Andresen, M.A. (2009). Geospatial technology and crime analysis. In K. Hastings (Ed.), Conference proceedings of the GeoTec event 2009. Vancouver BC, Canada.

Published onJan 01, 2009
Geospatial technology and crime analysis


Geographic information analysis is a valuable tool in spatial crime analysis. The ability to integrate spatially-referenced crime data with other environmental attributes not only makes large scale analyses possible, but has advanced the field of spatial criminology. In this paper I discuss some of the issues that relate to the use of GIS in crime analysis regarding spatial data quality and the use of new geospatial data. Specifically, because of the widespread use of GIS in crime analysis spatial data quality and error propagation are serious concerns; and the availability of recently developed geospatial technologies in GIS increases the utility of GIS and the spatial analysis of crime.

Keywords: crime; spatial analysis; data quality; interoperability; ambient population

1. Introduction

The geographical analysis of crime has been undertaken since at least the 19th century in France and England and the early 20th century in the United States. This research, and the research that followed, clearly showed that there is a strong geographic bias in crime with that bias is different for different forms of criminal activity. Until the 1970s, the geography of crime research was dominantly performed by sociologists. After this time, the geography of crime increasingly became the interest of geographers, urban planners, and architects leading to theories of the geography of crime with a geographical imagination (Brantingham and Brantingham 1981). With the advent of geographical information systems (GIS) in recent years, geographical studies of crime have become easier in some ways (data integration and subsequent analysis) but more complicated in other ways (data quality and interoperability).

In this paper, I outline some of the issues that we now face when using GIS, and the corresponding geospatial technologies, to analyze crime. This is not to say that advances in GIS have hindered the geographical analysis of crime. Rather, the geographical analysis of crime has advanced because of the advances in GIS. Indeed, the geographical analysis of crime has flourished in recent years (Ratcliffe 2002), we simply now have a new set of challenges to deal with.

2. Data quality, geocoding, and crime analysis

Data quality issues are not unique to the geographic analysis of crime. There are, however, issues that are particularly important to the geographic analysis of crime, one of which is geocoding. There are many fields that must geocode discrete events for subsequent analysis in a GIS (traffic accidents, public health, and epidemiology, to name a few), and its importance in geographical crime analysis is paramount. There are two fundamental issues that have implications for data quality: geocoding success rates and the quality of the underlying spatial data, each clearly related to the other.

Geocoding success rates are limited to the quality of the data being geocoded and the quality of the data being geocoded to. The quality of the former may, of course, be improved by the user, but all too often these input data are counted in the tens of thousands. Therefore, even if the entire data set is “cleaned”, there may be a number of records that are incorrect. Ratcliffe (2004) identified a minimum acceptable success rate for geocoding that maintains the original geographical distribution of a phenomenon. This minimum acceptable success rate implies that a certain number of (criminal) incidents will not be geocoded because of improper addresses, street numbers, etc., but that is acceptable up to some limit because it is the geographical pattern that is under analysis not individual incidents. Ratcliffe (2004) identified this minimum acceptable success rate as 85 percent, a reasonable goal to set for any analysis.

The quality of the underlying spatial data (road networks, etc.) that incident data are geocoded to are also particularly important. Ratcliffe (2001), using TIGER-type road networks, found that more than 50 percent of geocoded (criminal) incidents are given geographic coordinates within the wrong land parcel. Clearly such incorrect geocoding is not acceptable in small-area analyses of crime, but this incorrect geocoding leads to (criminal) incidents being misallocated to the wrong census tract 5 to 7.5 percent of the time. Therefore, this lack of accuracy for TIGER-type geocoded address data may have significant implications for relatively aggregate forms of spatial analysis.

Figure 1. Data quality



The origin of this latter data quality issue is two-fold. First, for reasons of feasibility, assumptions must be made when calibrating geocoding algorithms. One of which is to assume as linear relationship along a 100-block; in other words, 125 Some Street is 25 percent along the 100-block of Some Street and 150 Some Street is 50 percent along the 100-block of Some Street. Though this may often be the case, it cannot always be assumed. Second, as outlined in Figure 1, roads in the road network may not be where they are expected to be. The census boundary in this figure is a census tract, so all census boundaries follow a road. It is easy to see that the road cuts through the interior of the census tract, Figure 1a, so all (criminal) incidents allocated to that road segment regardless of which side of the street they are on will be allocated to one census tract. In Figure 1b, the road segment literally weaves through six different census tracts, potentially misallocating most of the (criminal) incidents that are geocoded to this road segment. As such, the percentage of geocoded (criminal) incidents found to be allocated to the incorrect census tract by Ratcliffe (2001) may be considered low.

Within the social sciences, research can really only be performed that analyzes the accuracy of geocoding procedures. Only advances in geospatial technology, whether they are geocoding algorithms or better quality underlying spatial data, will have significant impacts here.

3. Interoperability/compatibility of data

Clearly related to data quality is the issue of data interoperability and comparability. A road network being properly aligned with census tracts is an interoperability issue, but interoperability extends far beyond a spatial mismatch. A further issue is that of interoperability between different data sets based on issues such as schematic and semantic heterogeneity; schematic heterogeneity may be having one municipality classifying each land use as its own variable (in a binary format) and another municipality classifying land use in one variable, whereas semantic heterogeneity may arise from using different terms for describing similar or identical objects through different language and meanings (Schuurman 2002, 2005).

Depending upon crime data availability, the use of data from multiple sources (municipalities and other levels of government) is common. As with data quality issues this is not unique to the geographical analysis of crime, and the increasing availability of spatial data from multiple source agencies is only making these issues more important. However, in the case of the geographical analysis of crime, not only is the spending of scarce public resources at stake, but potentially the safety of the populace when considering tactical or long-term crime analysis and prevention projects.

4. New geospatial data and crime analysis: the ambient population

Despite the challenges that the geographical analysis of crime faces because of data quality and interoperability issues, the availability of new geospatial data (because of relatively recent technologies) has allowed researchers and practitioners to attempt to answer questions that were posed decades ago. Specifically, the issue of properly measuring the population at risk of criminal victimization has been around for at least 40 years (see Boggs 1965). However, because of limitations in time, money, and feasibility, this issue has largely been unresolved. That is, until recently.
Figure 2. Resident and ambient populations

a) Resident population

b) Ambient population

Traditionally, the measurement of the population at risk has been through the use of census data. But the problem with such a procedure is that most people leave their home census unit (census tract, for example) during the day. As such, census population counts are not very useful for calculating crime rates that need a measure for the population at risk.

Oak Ridge National Laboratory has been producing estimates of the ambient population (LandScan Global Population Database, LGPD): the number of people present in a square kilometre cell any time of the day and all days of the year, incorporating diurnal and seasonal population changes. Thoroughly researched by Dobson and colleagues (2000, 2003, 2004), these data are freely available for non-commercial use and provide an excellent, though not perfect, measure of the number of persons at risk of criminal victimization. The LGPD calculates the relative attraction of each cell for the population based on land cover, road networks, slope, and nighttime lights and reallocates census populations accordingly at a scale relevant for the geographical analysis of crime. Andresen and colleagues (2006, 2007, 2008) have used these data in a criminological context.

As shown in Figure 2, the geographical distribution of the ambient population is quite different from the resident census population. As expected because of the goals of the census data collection the resident population is quite evenly distributed across the urban landscape (Vancouver, British Columbia), but the ambient population is clearly clustered in particular areas of the city. This changes the distribution of crime rates across the city quite substantially depending upon the spatial units under analysis such that there is little statistical relationship between resident- and ambient-based crime rates. Figure 3 highlights this change by taking the ratio of the ambient to the resident population. It is clearly evident that there are areas in Vancouver that gain and lose significant portions of their resident population throughout the day. As such, any inference for theory or policy based on resident-based crime rate calculations may be in error. Of course, the ambient population estimates are far from perfect, but they are a significant improvement over the census data alternative and Oak Ridge National Laboratory is currently improving their estimation algorithms.

5. Conclusion

As argued in this short paper, geospatial technology presents the geographical analysis of crime with a new set of challenges and a new set of opportunities. Social scientists must partner with scientists in other disciplines and industry in order to minimize these challenges in the future without compromising the new opportunities. Such a partnership will not only make the geography of crime research easier because of highly quality geospatial data, it will also make it more relevant and reliable for improving standards of living through a better understanding of crime.

Figure 3. Ambient-resident population ratio


Andresen, M.A. (2006). Crime measures and the spatial analysis of criminal activity. British Journal of Criminology 46: 258 – 285.

Andresen M.A. (2007). Location quotients, ambient populations, and the spatial analysis of crime in Vancouver, Canada. Environment and Planning A 39: 2423 – 2444.

Andresen, M.A. and G.W. Jenion (2008). Ambient populations and the calculation of crime rates and risk. Security Journal, forthcoming. DOI: 10.1057/sj.2008.1

Boggs, S.L. (1965). Urban crime patterns. American Sociological Review 30: 899 – 908.

Brantingham, P.L. and P.J. Brantingham (1981). Notes on the geometry of crime. In P.J. Brantingham, and P.L. Brantingham (eds) Environmental Criminology. Beverly Hills, CA: Sage Publications, 27 – 53.

Dobson, J.E. (2003). Estimating populations at risk. In S.L. Cutter, D.B. Richardson, and T.J. Wilbanks (eds) The Geographical Dimensions of Terrorism. New York and London: Routledge, 161 – 167

Dobson, J.E. (2004). The GIS revolution in science and society. In S.D. Brunn, S.L. Cutter, and J.W. Harrington, Jr. (eds) Geography and Technology. Dordrecht: Kluwer Academic Publishers, 573 – 587.

Dobson, J.E., E.A. Bright, P.R. Coleman, and B.L. Bhaduri (2003). LandScan: a global population database for estimating populations at risk. In V.Mesev (ed) Remotely Sensed Cities. London and New York: Taylor and Francis, 267 – 279.

Dobson, J.E., E.A. Bright, P.R. Coleman, R.C. Durfee, and B.A. Worley (2000). LandScan: A global population database for estimating populations at risk. Photogrammetric Engineering and Remote Sensing 66: 849 – 857.

Fotheringham, A.S., and D.W.S. Wong (1991). The modifiable areal unit problem in multivariate statistical analysis. Environment and Planning A 23: 1025 – 1044.

Gold, S.S. (2003). Watch us move: homeland security requires a new kind of population map. Popular Science January: 40.

Openshaw, S. (1984). The modifiable areal unit problem. CATMOG (Concepts and Techniques in Modern Geography) 38, Geo Books, Norwich.

Ratcliffe, J.H. (2001). On the accuracy of TIGER type geocoded address data in relation to cadastral and census areal units. International Journal of Geographical Information Science 15: 473 – 485.

Ratcliffe, JH (2002). Aoristic signatures and the temporal analysis of high volume crime patterns. Journal of Quantitative Criminology 18: 23 – 43.

Ratcliffe, J.H. (2004). Geocoding crime and a first estimate of a minimum acceptable hit rate. International Journal of Geographical Information Science 18: 61 – 72.

Schuurman, N. (2002). Flexible standardization: making interoperability accessible to agencies with limited resources. Cartography and Geographic Information Science 29: 343 – 353.

Schuurman, N. (2005). Social perspectives on semantic interoperability: constraints to geographical knowledge from a database perspective. Cartographica 40: 47 – 61.


Martin A. Andresen is Assistant Professor of Criminology and a member of the Institute for Canadian Urban Research Studies at Simon Fraser University. With a background in economics and (economic) geography, his research is concentrated in spatial crime analysis, co-offending, applied spatial statistics, and the geography of international trade. This research has been published in a number of academic publications in journals such as Annals of the Association of American Geographers, British Journal of Criminology, Canadian Geographer, Criminal Justice Policy Review, Environment and Planning A, Regional Studies, and Security Journal.

No comments here
Why not start the discussion?