PCI’s latest release explores crime risk in Australia and provides the first comprehensive depiction of neighborhood crime risk in the country. Our data science team reconciled a wide range of sources, with different publication timelines and definitions of crime types to produce consistent measures of the seven major crime types.
An accurate crime score
National crime statistics are published by the Australian government, reporting annual crime counts for most areas. These reports lag reality by as much as a year and a half, and crime counts are presented at the territory or state level. Some local municipalities publish richer data sets at more regular intervals, but each reporting agency uses their own breakdown of major crime types and subtypes and their own publication schedule.
These differences of how crimes are recorded and the frequency with which they are released can conflate understandings of risk. For instance, some reporting agencies might distinguish between assault and battery, as well as major and minor assaults, where other agencies could collapse this data merely into “assaults”. This results in an apples-to-oranges type problem, where you might be comparing two different definitions of assault across different years of data.
Thus, crime counts might vary from one another because of differences in definition, as opposed to actual differences in the number of offenses. Though these issues are reconciled at the national level, the national data is outdated, highly aggregated, and only provided annually.
Reconciling these publication differences is a crucial step to produce a measure of crime risk at the neighborhood and monthly scale. Our data science team employed a variety of statistical machinery to build translations between the local municipality data and the national crime statistics to reconcile these differences. We then forecast offenses by location into the near future, producing a consistent measure of risk in both time and space.
Australian neighborhoods, severity weighting and floating populations
The result of this effort is, for the first time, the ability to compare crime risk in a neighborhood in Sydney relative to a neighborhood in Adelaide. We are able to produce crime scores for 2,214 of the 2,310 neighborhoods in Australia (“neighborhoods” are defined using Australia Statistical Area level 2 geography), covering 99.5% of the country’s population. Sparsely populated areas, with below 1 person per 30 square kilometers, were not assigned a risk color code and not ranked in nationwide comparison. This is the result of either a lack of granular data or a distrust in the information being provided.
Similarly to PCI in other countries, we also correct for “floating populations” — differences in residential population versus the actual number of people inhabiting each district — and employ severity weighting when aggregating crime types to an overall index.
Make informed decisions and mitigate risk
As with prior PCI releases, Pinkerton utilizes cutting-edge data science to enrich our more than 170 years of real-world experience and expertise in helping organizations manage their risk. The Australian crime index provides an unrivaled understanding of granular crime risk, enabling you to make informed decisions across space and time.
