Crime Data

Tabled: 5 September 2018

4 Analysing and releasing crime data

CSA checks and sorts crime data—a process known as transformation—before it is considered reliable and meaningful enough to publish. Transformation can involve addressing errors, missing data and inconsistencies, and deciding whether to include or exclude very small datasets or outliers. It can also include grouping certain categories of data to make it easier to understand. Victoria Police provides CSA with raw recorded crime data, and CSA publishes it in accordance with a number of classifications, including its offence classification, based on ANZSOC.

4.1 Conclusion

CSA's crime statistics are complete and accurate, and the public can be confident that its published data tables contain reliable information on crime in Victoria since 2008. We recalculated the data in CSA's Recorded Offences reports for the 10 years from 2008 to 2017 and obtained an almost exact match between our results and those published by CSA. We also used LEAP to recalculate CSA's Criminal Incidents dataset from 2009 to 2017. Over the course of the years examined, we found differences of less than 2 per cent across each year in both calculations. From this, we conclude that CSA's reporting of crime rates over time closely reflects Victoria Police's actual performance.

CSA's methodology is transparent—it publishes information on its website about its processes and its counting and classification rules so that the public can understand how it uses Victoria Police data and how the public can interpret the statistics.

CSA developed an offence ranking system that closely aligns with the ABS NOI and adapted it to reflect Victorian legislation on crime. This ranking system is used to classify the offence type of incidents within the Criminal Incidents table. Again, the methodology is sound and allows the public to understand the level of crime that affects victims. However, the Criminal Incidents dataset should be used in conjunction with the Recorded Offences table rather than as a standalone measure of crime.

Victoria Police is using crime data to inform its decision-making. It uses data from LEAP to track and compare the level of crime in the community and to understand its progress against its publicly reported key performance indicators. Crime data is also helping PSAs to understand crime trends and where they can best focus their efforts.

4.2 Processes for transforming data into crime statistics

CSA publishes crime statistics in quarterly and annual reports. Its website contains information about how it transforms Victoria Police's LEAP data into published crime statistics.

CSA receives a monthly subset of LEAP data based on an extraction script written when CSA was first established. After each quarter, CSA conducts a series of high-level quality checks, compares the data to previous extracts to see if the number of records is within expectations, and tests for any missing data. When there are unexpected results, CSA consults with Victoria Police. CSA then classifies the data and applies a series of counting rules. CSA analyses notable movements and trends for the media release to help readers understand the information. Once the clearance processes and analysis are complete, it releases aggregated, de-identified data reports to the public on its website. CSA's documentation of its processes shows that it implements a clear and consistent approach for calculating crime statistics.

Individuals and researchers can easily access data on the CSA website, which includes an interactive ' crime by location ' tool. CSA also provides crime statistics in response to information requests—it receives about 1 000 requests each year from Victoria Police, government agencies (especially DJR), academics and the public. CSA has a process for managing and responding to data requests, which takes into account the nature of the request, its purpose and any potential for the data to identify individuals. Councils and ministers use data from CSA to explore potential crime risks for particular areas of a community. Government departments make data requests to inform policy and project work.

The monthly extract provided by Victoria Police does not contain the free-text 'narrative' field, which records specific details of incidents and actions taken to solve crime. This is an important part of the crime record, as members will usually only amend or add information to this field. However, it is possible the information recorded in the 'narrative' field may contradict data in the other fields and vice versa.

Victoria Police provided CSA with its first extract of 'narrative' data from LEAP in January 2018. This was a one-off provision of data. CSA considers 'narrative' information valuable as an additional source of information for validation. Victoria Police and CSA have yet to determine guidelines for the use of 'narrative' information in CSA's transformation process.

CSA publishes a number of data tables, as shown in Figure 4A. When the media reports on a rise or fall in the crime rate, they often use the Recorded Offences table to derive it while, in theory, other tables can also be used.

Figure 4A
Data tables published by CSA

Table

Description

Publishing frequency

Recorded Offences

The number of offences recorded by Victoria Police in LEAP during the reference period. This includes offences where no offender is identified or charged.

Quarterly

Alleged Offender Incidents

The number of offenders given an outcome by Victoria Police within the reference period. An individual may be an alleged offender multiple times within the period. If an incident includes multiple offences, it is assigned the offence category of only the most serious offence in the incident, known as the principal offence.

Quarterly

Victim Reports

The number of victim reports made to Victoria Police within the reference period. An individual may have reported multiple incidents of victimisation within the period.

Quarterly

Family Incidents

The number of family incident reports submitted by Victoria Police and recorded in LEAP within the reference period.

Quarterly

Unique Offenders

A snapshot of unique offenders within the reference period. An offender who has had multiple incidents is counted once.

Annually

Unique Victims

A snapshot of unique victims within the reference period. An individual who has reported multiple incidents of victimisation is counted once.

Annually

Criminal Incidents

A measure of criminal events. They may include multiple offences, offenders and victims recorded against one date and location.

Quarterly

Source: VAGO based on information provided by CSA.

Criminal Incidents table

CSA started reporting on incident-based crime in December 2017. Its intention in developing this additional table is to create a measure that is more aligned with how the community experiences criminal incidents—that is, as the criminal event that occurred rather than the unique offences that were involved. The introduction of the new measure is intended to complement the Recorded Offences table, and CSA publishes the incident crime rate alongside the Recorded Offences crime rate.

An incident is a distinct event that can be made up of one or many sub-incidents or offences occurring at the same time and location. It can also include multiple victims and alleged offenders. As a criminal incident can involve multiple offences and, therefore, multiple offence types or charges, CSA has developed a method for determining the 'principal offence', which becomes the offence type used for grouping incidents in the report. Figure 4B shows the method CSA uses.

Figure 4B
CSA's method for assigning principal offences or charges for incidents

Figure4B shows CSA's method for assigning principal offences or charges for incidents

Note: The fourth example shows the impact of charges when calculating the principal offence/charge for an incident—offences that have charges attached are prioritised over those that do not. In this example, a charge has been laid for assault but not for drug trafficking, thus assault becomes the principal charge even through drug trafficking has a higher ranking on CSA's Offence Index.
Source: VAGO based on CSA's explanatory notes.

The principal offence represents the most serious charge associated with the incident or the most serious offence committed (if no charges are yet laid) within an incident. The CSA Offence Index is largely based on the NOI, which ranks offence codes based on seriousness. CSA uses the police outcome associated with the most serious offence or charge to represent the overall investigation status of the incident.

We compared CSA's offence ranking against the NOI to test whether the rankings align and whether the assigned order of seriousness is reasonable. A direct alignment of the ranking systems is not possible because crimes are not named and defined in exactly the same way across the criminal legislation of different states. However, we observed that CSA has broadly applied the same order of priority for offences as the NOI, so we consider CSA's determination of principal offence for the Criminal Incidents table to be appropriate. The NOI measure of a crime's seriousness appears to reflect the harm imposed on victims, and we assessed this as a reasonable approach for assigning the principal offence of criminal incidents. We used CSA's offence ranking in our analysis of the Criminal Incidents table, which is discussed in Section 4.4.

Reporting based on incidents rather than just offences allows the public to understand the number of criminal events rather than the number of offences—CSA states that this is closer to how the community experiences a criminal event. However, the Criminal Incidents table may not highlight the increases and decreases in offences that do not amount to the principal offence. Although these offences are not the most serious part of an incident, they are still significant and relevant pieces of information about reported criminal activity. The Criminal Incidents table is most useful when considered alongside the Recorded Offences table.

4.3 Applying counting rules and classifications

CSA has a transparent process for the transformation of Victoria Police crime data. These transformations categorise the data in accordance with national recording standards such as ANZSOC. CSA has modified ANZSOC where required to match Victorian criminal law. This is an appropriate adaptation to make the classifications relevant to Victoria.

We reviewed the statistical transformations of the data and found that they comply with CSA's own policies and standards. The crime data that CSA receives from Victoria Police contains all sub-incidents that include an offence and selected other events. These events include family violence reports and traffic cautions issued to juveniles.

Adjustments

CSA applies two key counting rules to Victoria Police data:

  • The first includes data on juvenile traffic cautions that Victoria Police does not view as offences but collects information on.
  • The second excludes offences not managed by Victoria Police such as some animal cruelty offences, interstate motor vehicle theft and traffic infringements—which Victoria Police includes in its internal reporting.

These are reasonable adjustments.

Indigenous status data

Victoria Police records a person's response regarding his or her Indigenous status on each occasion that an incident is recorded against the person. Victoria Police suggests that members may avoid asking whether a person identifies as Indigenous because they perceive it to be sensitive and, instead, either guess the person's Indigenous status or record 'unknown'.

To manage the high number of Indigenous status fields marked 'unknown' in the data provided by Victoria Police, CSA applies a counting rule to the Indigenous status data to improve its quality and make it suitable for release. CSA uses the 'most frequent' rule to allocate an Indigenous status in its reports. According to the 'most frequent' rule:

  • if a person has made only one meaningful response ('yes' or 'no'), CSA uses that response for all records
  • if a person has made both 'yes' and 'no' responses, CSA allocates the most frequently appearing response.

CSA's method may reduce the impact of any data entry error on resulting crime statistics.

We accept there will be legitimate cases where 'unknown' is an appropriate option for members to record in response to the Standard Indigenous Question, but frequent recording of this response may suggest that reporting is burdensome or that members do not consider it important enough to do properly. This limits the usefulness of the data for analysis.

Victoria Police has an opportunity to strengthen this process by training members on the importance of collecting information accurately and having supervising sergeants monitor members' efforts to identify the correct response before they record 'unknown'. Further, Victoria Police should centrally monitor use of the 'unknown' response for Indigenous status.

Categorisation

Police need crime records to contain specific information, including geographical locations and location types, information about relationships between the people involved, and descriptions of people, places, injuries or weapons. LEAP records should capture very detailed information to allow police and intelligence teams to identify trends in criminal behaviour and develop advice for how to best respond to crime.

Published crime statistics, on the other hand, are easier to interpret at a less granular level. For example, the public may be more interested in knowing about a rise or fall in assaults as a category of offence rather than separating this out by the individual crime codes that make up the assault category.

Within CSA's public reporting, details of individual offences are grouped into categories to help the public more meaningfully measure the impact of crimes on the community. For example:

  • the crime division 'property and deception offences' contains a crime subdivision for 'theft'
  • the 'theft' subdivision contains seven groups that identify types of theft such as 'motor vehicle theft' and 'receiving or handling stolen goods'—these groups contain the individual criminal codes assigned to each offence.

The 'theft' subdivision contains more than 170 different offence codes that distinguish specific crimes, some of which are no longer used due to repealed legislation. This level of detail is unlikely to be of use to the general public seeking to understand crime trends.

CSA categorises geographic location in multiple ways—statewide level, the four police regions, the 54 PSAs, local government areas, and suburbs and postcodes that broadly fit within a PSA. Victoria Police also captures data on location types, such as car parks or residential streets, and CSA groups these into categories. Location types identify characteristics of certain locations so that where certain types of locations are repeatedly targeted for particular crimes, Victoria Police can identify the trend. This categorisation is reasonable, as it preserves the originating police region and allows the public to explore the statistics by local government areas or suburbs.

4.4 Reconciling Victoria Police data and published crime statistics

Offence count reconciliation

We compared the data extract that Victoria Police provided to us with published crime statistics in CSA's Recorded Offences table . To do this, we applied CSA's processes and counting rules to the data extract provided by Victoria Police. The result was an almost exact match between our analysis and the publicly reported figures across nine years (2008 to 2016). In 2017, our calculations differed by 1 per cent—this variance is most likely due to the different extract dates used for our data compared with that supplied to CSA, resulting in 29 additional offences being included in our data. We therefore concluded that the number of offences publicly reported by CSA reflects what is recorded in LEAP.

Incident count reconciliation

We also compared the extract that Victoria Police provided to us with the Criminal Incident table by applying CSA's process and counting rules for this data. Our analysis varied by 5 per cent overall across the nine-year period (2009 to 2017 ) and varied by 5 per cent or less in all but one of the offence divisions (the 'other' category).

As with the offence count reconciliation, the difference in the results may be due to the different dates on which our data and CSA's were extracted. Figure 4C shows the two sets of results.

Figure 4C
Comparison of VAGO and CSA calculation of crime incidents by year ending September, 2009 to 2017

Figure4C shows a comparison of VAGO and CSA calculation of crime incidents by year ending September, 2009 to 2017

Note: The difference in the 2017 results can be attributed to different data extraction dates in September. Years ending in September have been used for analysis as the data Victoria Police provided to VAGO was extracted in September.
Source: VAGO based on Victoria Police and CSA data.

Figure 4C indicates that despite the differences in raw figures, similar conclusions can be drawn from both calculations of the data—that criminal incidents reduced by roughly the same rate between September 2016 and September 2017—approximately 5 per cent.

4.5 Analysing and using crime data

We looked at whether Victoria Police uses crime data for decision-making. We reviewed the areas responsible for internal data analysis and the ways the data is used.

Within Victoria Police, CSU provides data for internal information requests, for thematic briefings to the chief commissioner, and for CSA's analysis and public reporting. Its responsibilities include tracking Victoria Police progress against BP3 measures and forecasting crime trends for the forthcoming quarter.

Victoria Police uses crime data at its regular meetings to discuss progress and inform strategic and resourcing decisions. Senior staff can access data analysis tools on the intranet to complete analysis at any time. This makes it easy for members to check data and crime trends for a range of common queries.

CompStat

Compare Statistics (CompStat)—a forum for discussing accountability, focusing on elements of police performance management including crime reduction, intelligence, strategy and resource management. The concept was originally developed by the New York City Police Department and has been widely adopted by other police departments worldwide.

Assistant commissioners participate in CompStat regularly, a forum for discussing, analysing and comparing the performance of regions. Through this forum, participants can use data to:

  • evaluate performance against agreed corporate goals
  • encourage effective use of resources to achieve agreed goals
  • provide feedback to an area under review
  • discuss meeting agreed performance measures.

Victoria Police's CSU provides a range of relevant data and analyses to support CompStat to fulfil its responsibilities. CompStat briefing packs include scatter plots to show how divisions compare on clearance rates and number of offences. For example, the briefing packs may compare clearance rates for the offence 'theft of motor vehicles' to the state average and PSAs that are outliers with very high or low clearance rates. Data in the CompStat briefing packs may also present performance as percentage increases in crime. However, if this involves small numbers, it is not always meaningful.

Completing and sharing this analysis at CompStat meetings may act to keep divisions accountable and keep senior management aware of outliers and trends that need a response.

CompStat provides a forum for police to debate the drivers of crime, preventative strategies, and the balance between achieving reductions in total crime and 'harm crimes'—crimes that cause greater damage to people and their perceptions of safety.

In our interviews, four senior members stated that reducing harm crimes such as aggravated burglaries does not seem to help with reducing total crime. Reducing total crime means devoting resources to solving high‑volume lower‑harm crimes, which may divert attention and resources from lower‑volume but higher-harm crimes. They expressed frustration that reducing total crime is typically the focus of crime statistics and media coverage, and then becomes an internal priority. They argue that the priority should be to reduce harm crimes rather than total crime, which includes a lot of petty crime.

Tasking and coordination

PSAs also conduct monthly tasking and coordination meetings. The purpose of these meetings is to look at crime trends using data and intelligence for individual PSAs, and to discuss local priorities and resource allocations for responding to emerging crime trends.

We reviewed tasking and coordination meeting documents from three PSAs and found that they include:

  • regional intelligence updates, such as for drug crime
  • the number of offender debriefs completed—informal reports of conversations with offenders
  • the number of outstanding warrants
  • family violence activities and trends
  • trends in other crime categories
  • scorecards—monthly crime comparisons with percentage changes
  • information on known drivers of crime—for example, the link between alcohol consumption and assault
  • updates on how data supports claims that crime reduction efforts are successful
  • tasking opportunities—activities to help address issues identified
  • discussion of the causes behind spikes in trends and links to other offending and proposed actions.

Victoria Police's Intelligence and Response Unit provides guidance on how to put tasks and priorities arising from tasking and coordination meetings into effect. Decisions made at divisional and regional meetings also influence PSA responses. These mechanisms can work together to influence PSAs and how they arrange their resources to tackle different categories of crime.

The ready availability of data on the intranet and through regular communications makes it easy for PSAs, regions and divisions to compare their performance to others.

Back to Top