Measuring victims of crime: a guide to using administrative and survey data

The levels of crime victimisation in the community, as well as people’s perceptions of their safety, are issues that impact directly or indirectly on the quality of people’s lives. Those experiencing direct victimisation in particular, may suffer financially, physically or emotionally. Fear of crime can affect people by restricting community engagement, reducing levels of trust and impacting on social cohesion.

There are a number of ways in which individuals, the community and governments gather information about crime victimisation in Australia. Statistical data is a key component of this information about crime and is critical to decision making, research and policy development.

The suite of statistics produced by the Australian Bureau of Statistics (ABS) has expanded in recent years with the introduction of the redesigned Crime Victimisation Survey in 2008-09 (for further information please refer to Crime Victimisation, Australia cat. no. 4530.0). Faced with a variety of statistical information about crime, it is often difficult for data users to ascertain which data source is best suited for their particular needs. 

Statistics need to be well understood in order for them to be useful in making informed decisions. In some instances results from different data sources may provide a different picture of crime in the community, with administrative data indicating a trend in one direction and survey responses indicating the opposite. The reality is that the different data sources collected via different methods can produce equally valid but quite varied sets of crime victimisation indicators. The challenge for data users is to understand the origins of these different datasets and be able to identify statistics that are most relevant for their needs.

The aim of this paper is to increase community understanding of how the experiences of victims of crime in Australia are measured and to explore why the findings from different data sources may differ. A working example of comparisons between two different data sources is used to demonstrate the impact of varying data collection methodologies on data for a range of offence types. Readers are also offered guidance about which data source is best suited to particular research questions. 

•  Chapter 2 outlines the available ABS data sources for crime statistics and highlights the main differences between the sources;
•  Chapter 3 shows a working example of the differences between administrative and survey data using statistics from Crime Victimisation, Australia and Recorded Crime – Victims, Australia;
•  Chapter 4 provides guidance on the suitability of each data source to answer specific research questions; and
• A technical note steps through the methodology employed for the analysis provided in chapter 3
   

Concepts relating to crime and justice, including victimisation, can be challenging to measure, and the full quantum of crime occurring within the community may never be discovered. One of the key difficulties in determining the true prevalence of crime victimisation within a community is that many victims of crime do not report their experiences to police, and so are not counted in official police recorded data. The reasons for not reporting to the police vary widely, but may be due to the sensitive and personal nature of the criminal offence (for instance, sexual assault or abuse), a belief that the police could not act in response to a victim’s report, fear of reprisal from the offender, simply being too upset or injured following an incident, or wanting to deal with the issue themselves.

There is also a wide variety of information required to effectively describe people’s experiences as victims and answer research and policy questions about crime victimisation. For example, data may be required to:

• estimate the overall risk of becoming a victim,
• monitor trends in crime victimisation over time and across different geographic locations,
• identify risk factors in becoming a victim, to assist in establishing priorities for service providers, such as police and victim support groups, and
• contribute to understanding long-term outcomes for victims of crime.

There is no single data source that is able to provide all the information required to build a full picture of crime victimisation in Australia. Rather, there are multiple sources of information which provide data on different aspects of victimisation. 

The Australian Bureau of Statistics (ABS) maintains national collections on crime victimisation from two different sources: administrative records, such as data from state and territory police agencies; and household surveys. While the collections contain data about closely related concepts, in practice, survey data and administrative data are quite different, and there are advantages and disadvantages to both methods.

At times, the results from different administrative and survey data sources may create divergent pictures of crime in the community. The purpose of this chapter is to explore the key differences between crime victimisation data sources and discuss the implications of these on data comparability.

Administrative data are produced in the course of providing services to client groups or otherwise undertaking the core business of an agency, and as such are shaped by the business practices and the environment within which the agency is operating. There are key advantages associated with collating statistics from administrative data. Generally, administrative data essentially takes the form of a census, in that all the data recorded through the administrative process is potentially available for statistical analysis. Hence the data are not subject to sampling error which is associated with survey data. Administrative data also has the capacity for ongoing, efficient data collection without incurring much of the additional enumeration cost associated with conducting surveys.

There are also disadvantages associated with using administrative data. Data quality may be variable as there can be conflicting demands on record keepers, and statistics are rarely the primary purpose for which the data are collected. For example, crime data is collected as a by-product of police recording systems, the purpose of which is to assist the police agency in managing its business, rather than collect statistics. Furthermore, definitions of data items may not exactly match the definitions of the data items of interest to others outside of the agency and can make comparability challenging.

Administrative data on crime victimisation can be derived from a number of sources including police, hospitals and community service agencies. The ABS publishes data from two administrative data sources about victims of crime:

Data extracted from the Causes of Death and Recorded Crime - Victims collections will produce different results for similar data items due to a wide variety of factors, such as:

• when the information is recorded (e.g. recording of an incident as it comes to the attention of the agency or at some later stage after investigation or other formal process);
• the counting unit used by the agency (e.g. person count or incident count);
• recording rules (e.g. mandatory reporting of some variables); and
• the definition of concepts (a legal interpretation of assault may be different to a hospital treatment interpretation of assault).

A comparison between Causes of Death data and Recorded Crime – Victims homicide data found that Recorded Crime – Victims data consistently yielded higher homicide counts when compared to Causes of Death data (Mouzos, 2003). The analysis highlighted possible reasons for this difference including the difference in when the information is recorded. For the Cause of Death data, deaths are classified as homicides based on the coroner’s determination of the final intent of the incident causing death, whereas Recorded Crime – Victims data is based on the initial recording of criminal incidents by state and territory police agencies. In addition, homicide counts derived from the Cause of Death data are based on year of death registration whereas Recorded Crime – Victims homicide data are based on the year that the homicide becomes known to police.

When collating national data from various state and territory agencies (e.g. state/territory police agencies), differences may also occur between states and territories due to factors such as:

• differences in legislation between states and territories;
• differences in recording practices between relevant state/territory agencies (e.g. police);
• differences in local standards and classifications that may impact on the ability to map to official statistical standards;
• differences in the IT systems used to capture information; and
• differences in data extraction methods.

While the application of national standards and classifications help to minimise the differences between state and territory data, differences in business rules, procedures, systems, policies and recording practices can result in discrepancies in data between states and territories. This is explored in further detail in chapter 3 using Crime Victimisation Survey data and Recorded Crime - Victims administrative data as an example.

Surveys of individuals in the community provide a way of asking people directly about their experiences of crime. Such surveys are designed specifically for statistical purposes. In the field of crime and justice, often one of the primary reasons for conducting a survey is that many victims of crime do not report their experiences to police, and so are not counted in administrative data. Victimisation surveys are a way to tap the ‘dark figure’ of crime (Coleman & Moynihan, 1996). The ‘dark figure’ refers to the volume of incidents that are not officially recorded. Data obtained from surveys may provide a more accurate representation of the ‘true’ figure of crime. 

Surveys use a sample of the population to represent the experiences and perceptions of the entire population. The disadvantage of using this methodology is the sampling error that can be associated with the estimates produced, which refers to the error caused by using a sample of the population rather than surveying the whole population (for further discussion of sampling error, please refer to the technical note). Surveys are also expensive and time-consuming to design and undertake. Crime Victimisation topics are included in four surveys produced by the ABS.

Each of the surveys described collects data about different aspects of crime victimisation, however there are some concepts and data items that are common across all of the surveys. The data produced from these items differ across the surveys due to differences in the survey methodology employed to collect information. Each survey is unique in the following aspects of survey design and operations:

• Sample design and selection – Decisions on the appropriate sample size, distribution and method of selection are dependent on a number of considerations. These include the aims and content of the survey, the level of disaggregation and accuracy at which the survey estimates are required, and the costs and operational constraints of conducting the survey.
• Scope and Coverage – Depending on the survey requirements and constraints, the scope and coverage of the surveys differ. For example, the Crime Victimisation Survey excludes people living in very remote areas, whereas very remote areas are included in the scope of the NATSISS.
• Questionnaire format – Different results can be expected if the questionnaire format differs across the survey instruments. The GSS for example, covers a range of topics within which the crime and justice questions are asked around the middle of the survey. The PSS on the other hand is entirely devoted to questions about the safety of the individual. This may produce a context effect for respondents, which means that the order of the questions could impact on the responses given. This then may impact on the results of the survey.
• Survey procedures – Survey procedures relate to the mode of the data collection and any special procedures required. The Crime Victimisation Survey is generally conducted using a short telephone interview whereas the GSS and the PSS use longer face-to-face interviews. Both methods have advantages and disadvantages that can impact on the survey results. For a full discussion of the impact of different survey procedures, please refer to Information Paper: Measuring Crime Victimisation, Australia: The impact of different collection methodologies, 2002 (cat. no. 4522.0.55.001).
• Response rate – Response rates give an indication of the influence of non-response bias on the data collected in the survey. These vary across the surveys; for example, the 2009-10 Crime Victimisation Survey had a response rate of 87% whereas the response rate for the PSS 2005 was lower at 77%. As such, the influence of non-response bias on the survey data differs between the surveys.

Due to the factors described in relation to survey design and operations, each survey data set should be considered a unique stand-alone dataset and data items, even when appearing to measure similar concepts, should not be compared across the different survey data sources.

As discussed at the beginning of this chapter, there are key differences between administrative data and survey data. There are also specific additional factors unique to crime data that further differentiate administrative and survey data.

Measuring crime victimisation accurately within the community is a challenging task. Throughout this chapter the issues associated with data collection were briefly discussed and it was acknowledged that no single data source is able to comprehensively cover all facets of crime victimisation in the community. Instead, data from both administrative and survey data sources are useful to provide information about particular aspects of victimisation. The ABS maintains two administrative data sources in relation to crime victimisation and four surveys containing crime victimisation information. Each data source produces different results due to the different areas of focus and methodologies employed in collecting the data.

The next chapter explores the differences between the data from different sources in greater detail by using a working example that compares the Crime Victimisation Survey data with police recorded statistics as collected for the Recorded Crime – Victims collection.

The previous chapter examined the broad differences between administrative data and survey data with particular reference to the specific issues in measuring crime victimisation. This chapter continues to examine the differences between data sources using a working example of offence data from Recorded Crime – Victims, Australia(cat. no. 4510.0) and Crime Victimisation, Australia (cat. no. 4530.0). 

This chapter of the Guide aims to answer the questions: What kinds of differences exist between and within different crime victimisation data sources that users need to be aware of when using the data? How can the quality of one source be examined using another source? What do users of crime victimisation data need to keep in mind when looking to bring multiple sources together for analysis or comparison?

In order to assess the degree of comparability between different sources of data it is important to have a clear understanding of the scope, coverage and methodology used for data collection. Where it is not possible to make direct, meaningful comparisons between sources, the trends that arise in the data can still be useful in identifying quality issues and patterns over time. 

To illustrate the process of assessing multiple sources and to highlight some of the different considerations to be taken into account when determining comparability, this chapter will look at an example comparing two different forms of ABS crime victimisation data. The first dataset is created from a sample survey – the Crime Victimisation Survey – and the second dataset is the Recorded Crime – Victims, Australia statistical collection, which comprises recorded crime data sourced directly from policing agencies in each state and territory in Australia. These datasets contain similar concepts: the survey asks respondents about crimes that police have been told about, while the data collected from police represents crimes they have recorded on their administrative systems. These sources can produce quite different statistics, however, and are not easily reconciled. Understanding the different scope and coverage of the sources is imperative to understanding why the resultant statistics can differ.

This illustration explores the differences between the Crime Victimisation Survey and the Recorded Crime – Victims police administrative data collection. There are a number of fundamental differences between the two sources: the way the data is collected, whom it is collected from, the offences that are in scope, the population that is in scope, the geographical regions that are covered and the reference period within which the data is collected.

The collections also cover different offence types. The Recorded Crime – Victims collection only includes selected offences from the range under the remit of policing agencies. The crime victimisation survey similarly focuses on personal and household crimes that are of policy interest and community impact, and where respondents are likely to be able to report their victimisation to an interviewer. For the purposes of this illustration, three personal offence types that are included in both collections are included in this analysis: assault, sexual assault and robbery; as well as two household offences: break-in and motor vehicle theft. Other offence types collected in the Crime Victimisation Survey, such as malicious property damage and other theft do not align with the crime types included in the Recorded Crime - Victims collection and therefore are not included in this analysis.

The survey data and data collected from police differ in population scope and coverage. The police data covers people and organisations who are recorded as victims of crime in almost all areas in Australia (with the exclusion of some spaces under the jurisdiction of the Australian Federal Police, such as major airports and the Cocos (Keeling) Islands) whereas the ABS survey sample does not reach all Australians in all areas. The Crime Victimisation Survey is only asked of Australians selected for the survey who are over 15 years of age (with proxy interviews by parents or guardians possible for respondents under the age of 18) and over 18 years of age for the sexual assault questions. The survey also excludes people who live in areas of Australia that are considered very remote. The exclusion of some areas of Australia from the sample survey is due to the degree of difficulty in accessing some very remote areas and the prohibitive costs of sending interviewers into difficult terrain and areas that are a long way from major centres. For a full listing of the scope and coverage of each collection please refer to the Methodology for Recorded Crime - Victims, Australia (cat. no. 4510.0) and Crime Victimisation, Australia (cat. no. 4530.0).

As a result, if we are to look at comparing the survey and administrative datasets, we need to try and align the coverage of the administrative data with the more restricted survey data. In the example discussed throughout this chapter, we have made the following amendments to the datasets:

• Removal of Recorded Crime – Victims data for persons under the age of 15 years for all general offences, and persons under the age of 18 years for sexual assault.
• Removal of Recorded Crime – Victims data for persons living in very remote areas of the Northern Territory. The Crime Victimisation Survey does not include very remote areas within scope of the survey. For most jurisdictions the exclusion of very remote areas will not affect comparability, given they account for only approximately 1% of the population. For NT, however, this equates to approximately 23% of the population and therefore adjustments have been made to the administrative data to align with the scope of the crime victimisation survey data.
• Removal of Recorded Crime – Victims data about organisations; data is presented for households and persons only.
• Removal of unreported crime incidents from the Crime Victimisation Survey - only victims who say the most recent crime incident was reported to police are included to align more accurately with the police recorded victims data.
   

The Crime Victimisation Survey data is collected on a financial year basis (for example July 2008 to June 2009), while the Recorded Crime – Victims data is collected on a calendar year basis (for example January 2008 to December 2008). Questions within the Crime Victimisation Survey ask respondents to refer to the 12 month period prior to participating in the survey. For example, for survey interviews conducted in September 2009, the respondent is asked to report any incidences of physical force or violence they have experienced since September 2008. This means that the potential span of data from the survey ranges from July 2007 (the earliest offence date for respondents who were interviewed in July 2008) to June 2009. The centre point to the survey recall period is June 2008. June 2008 is also the centre point of the Recorded Crime – Victims data. As such, the 2008-09 data from the survey and the 2008 Recorded Crime – Victims data are considered a suitable approximation for comparison. Similarly the 2009-10 Crime Victimisation Survey data and the 2009 Recorded Crime – Victims data are considered a suitable approximation for comparison.

For a complete outline of the methodology used for this analysis, please refer to the technical note.

Data recorded by police is essentially a census where all counts should be supplied as part of the data collection. The survey data is created from responses given by a sample of the Australian population that is then weighted. Weighting is a process that is undertaken in order to produce estimates of the prevalence we would find for these offences if the entire Australian population had been included in the survey. When using survey data for any analysis, the reliability of the estimates must be considered. Since the estimates in the Crime Victimisation Survey are based on information obtained from a sample of persons, they are subject to sampling variability. Sampling variability means that the estimates may differ from those that would have been produced if all persons were included in the survey.

The standard error (SE) is a measure which indicates the extent to which an estimate might have varied because only a sample was used. Using this measure, the likely range within which the value that would have been produced if all persons had been included in the survey can be calculated. There are about 19 chances in 20 (or a 95% chance) that the difference between the survey estimate and the true value is less than two standard errors based on the Crime Victimisation Survey. We will be referring to this range as the 95% confidence interval. Standard errors for values presented in graphs are displayed below each graph in a table.

When comparing the survey estimates and the administrative data, it is this range above and below the survey estimate that will be considered in the analysis rather than the estimate itself. Where the recorded crime data falls within the range of the estimate, there is a degree of consistency between the data sources. That is, there is insufficient evidence to conclude that there is any difference between the recorded crime value and the population value underpinning the survey estimate (i.e. the values are not statistically significantly different). For example, there are 19 chances in 20 (or a 95% chance) that the true value for the number of persons reporting physical assault in NSW was between 52,800 and 97,200 in 2008-09. The recorded crime figure for total assaults is 72,400, which falls within the confidence interval for the survey estimate, therefore in this case, the administrative data value and the survey estimate are not statistically significantly different.

So far, this illustration has focused on the broad differences that can be seen in the characteristics of datasets across scope, coverage and source. These differences can affect certain offences in different ways due to the specific nature of the offence, the way that victims respond to the event, and sensitivities around reporting to police and other authorities (such as official surveys). By examining each of the selected offences in turn, it is possible to gain a better understanding of the visible differences that can result in statistics as a result of these issues.

Assault statistics from the RCVS collection and estimates from the survey data demonstrate the issues encountered when there is variation in the definitions between data sources and recording practices within data sources. For the survey, the data that is presented relate only to physical assaults that respondents state have been reported to police. The recorded crime data from police records relate to all assaults (attempted and completed). For the detail around these different scopes, please refer to the technical note. This offence demonstrates the impact of different recording and business practices across jurisdictions.

The graphs below compare the estimated number of victims of physical assaults who reported to police from the 2008-09 and 2009-10 Crime Victimisation Survey with the number of victims of total assault recorded by police in the Recorded Crime – Victims collection for 2008 and 2009 respectively. 

The survey estimates have been plotted to include bars to show the range of the 95% confidence interval and the police data has been plotted as direct counts. Despite differences in methodology and content between the data sources, for NSW, SA, WA, NT and the ACT the number of victims of assault as collected in Recorded Crime – Victims falls within less than two standard errors of the survey estimate for the number of victims of physical assault. In this sense, the assault statistics sourced from these two collections can be said to be consistent. 

For Victoria, Queensland, and to a lesser extent, Tasmania, the recorded crime data for total assaults is statistically significantly lower than the survey reported estimates for physical assault.

For the 2009 reference period, the Recorded Crime data for Victoria, Queensland and Tasmania still fall below the range of the survey estimate, with the addition of Western Australia.

As previously discussed, the data that is collected by a police agency for operational purposes is a product of local business rules, context and legislation. In order to try and standardise statistical data across the states and territories to enable the compilation of national statistics, each state and territory police agency worked with the ABS to develop a National Crime Recording Standard (NCRS). Agreed NCRS manuals were distributed to jurisdictions in 2007. 

The NCRS was developed to address the lack of uniformity in initial police recording processes. The NCRS guides the recording and counting of criminal incidents for statistical purposes and enables consistency in recording. While the application of the rules and requirements of the NCRS generally enable the recording of crime in a comparable manner across the jurisdictions, there is some variability in the interpretation and implementation of the rules, in particular the rule which guides what is recorded on the police systems when an incident is reported to police.

Here is the relevant rule in full:

An incident will be recorded as one or more offences if prima facie (on the face of it) and on the balance of probability (more probable than not):

• The circumstances as reported or detected amount to a crime defined by law and fall within the jurisdiction of the police agency; and
• There isn't credible evidence to the contrary.
   

This rule provides a common basis for recording an incident with one or more offences according to the judgement of the police officer (as distinct from evidentiary or prosecutorial reasons).

The difference in the interpretation of this rule is evident in the results of the comparisons between the Crime Victimisation Survey data and the Recorded Crime – Victims data for assaults. For the jurisdictions that interpret the rule as requiring recording based on the report of the victim, the recorded crime data falls within the 95% confidence interval of the survey estimate. For the remaining jurisdictions that require a determination prior to being recorded, the recorded crime data falls outside the 95% confidence interval of the survey estimate. 

There are also more specific issues that can affect a jurisdiction’s data, such as specific legislative differences and administrative practices. Results for the four states where there were differences between the survey results and the administrative data (Victoria, Queensland, Western Australia (in 2009-10 only) and Tasmania) are explored in more detail in the following.

There are two major policy documents that provide direction to Victoria police in the recording of information: one relates to general crime, and the other relates specifically to assault. Policy changes in accordance with the NCRS have been made to the general crime documents, but the recording of assaults continues to be evidentiary based. The assault incidents that are reported to police are initially entered into a patrol duty return, however investigation will be conducted before these are entered into the crime recording system.

For family and domestic violence incidents that involve assault, police will complete a Family Violence Report indicating that assault has occurred but these are not recorded on the crime recording system in all instances. 

One reason for the differences between the recorded crime data sourced from Queensland police and the survey estimates is the more evidentiary approach taken to recording assaults in Queensland, ie the initial report is not taken on face value, in comparison to other jurisdictions, such as New South Wales, South Australia, Western Australia, Northern Territory and Australian Capital Territory. However, there is also a secondary contributor to the difference between the data sources, and this is the application of Queensland family and domestic violence legislation. 

All first instances of domestic violence must be investigated, and action can be taken under the Domestic and Family Violence Protection Act 1989 regarding a domestic incident. If there is evidence of a criminal offence, these matters will also be investigated and appropriate action taken where there is sufficient evidence. If the criminal offence involves an assault, proceedings will only be commenced with the consent of the victim.

The Criminal Code Act 1899 deals with all criminal assault matters. Where a domestic violence incident has occurred which involves an assault, but the victim does not consent to proceeding with the assault charge, then the assault matter will not be recorded on the Queensland Police Records and Information Management Exchange (QPRIME).

In consideration of the legislation described, if a Queensland respondent in the crime victimisation survey stated they were physically assaulted and reported the most recent incident to police, but in reality did not consent for police to proceed with the assault charge, the incident would have been counted in the survey, but would not be included in the police recorded crime data. In other jurisdictions where family and domestic violence incidents are recorded even if the victim does not want to proceed, the incident would be included in the police recorded crime data.

Western Australian police will accept the report of the victim on face value in accordance with their interpretation of the NCRS. Like other jurisdictions that adopt a similar approach, Western Australian recorded crime data for assaults in 2008-09 was lower than the survey estimate, but within the 95% confidence interval of the survey estimate. In 2009-10, however, the number of victims of assault from the recorded crime data is statistically significantly lower than the survey estimate.

Tasmania police operate two separate systems for crime victim recording: the Computer-Aided Dispatch System (CAD) and the Offence Reporting System. All initial reports are recorded through the police communications area in the CAD. The Offence Reporting System records crimes or offences where on face value an offence appears to have been committed. All assaults are reported on the offence reporting system unless there is evidence to suggest that no offence was committed.

Crime statistics as disseminated in Recorded Crime – Victims are sourced from the Offence Reporting System. 

The results of the 2009-10 analysis produced a closer match between the Tasmanian survey results and recorded crime data than 2008-09, however the recorded crime figure was again statistically significantly lower than the survey estimate. The counts for Tasmania are small and as such are subject to relatively more variation than the larger states. 

This example shows the difficulties in obtaining consistent and comparable assault data across jurisdictions (based on police data) due to their different legislative frameworks, business practices, crime recording systems and interpretation of the national crime recording standard. It is therefore recommended that police recorded crime data for assault not be compared across jurisdictions for these reasons. In order to make comparisons between jurisdictions, it is recommended that Crime Victimisation Survey data be used, as this data is not subject to differences in practices or definitions across jurisdictions, and is hence comparable across Australia.

Due to the sensitive nature of sexual assault, it is often a crime that goes unreported to police authorities, services, friends and family. According to the Crime Victimisation Survey, in 2009-10, 36.6% of the most recent incidents of sexual assault were reported to police. The reporting rate for sexual assault is relatively low when compared with other personal offences: the estimated national reporting rate for physical assault in 2009-10 was 50.5% and for robbery the estimated national reporting rate was 60.7%. It follows then that survey respondents may also be reluctant to answer questions about sexual assault via a telephone survey. As such, the questions about sexual assault are asked only of persons aged 18 years and over. Corresponding data from the Recorded Crime – Victims collection has been limited to 18 years and above for the purposes of this analysis.

Sexual assault offences are often reported to the police at a time well after the incident had occurred. That is, the sexual assault victims recorded in the police recorded data may have been reporting in relation to an incident which occurred many years ago, whereas the survey questions limit responses to victimisation events that occurred within 12 months prior to the survey interview. 

For both 2008 and 2009, all recorded crime sexual assault figures were within the 95% confidence interval for the survey estimate, although the confidence intervals are so large as to render the survey estimates too imprecise for comparative purposes.

The low prevalence of sexual assault in the community in comparison with some other surveyed offences should be considered when interpreting the results of this comparison. Low prevalence offences are associated with higher levels of survey sampling error. This higher error makes it more difficult to assess the consistency between the sources to a high degree of precision. It also emphasises the fact that the figures produced from the survey are sample-based estimates only and that the reliability of the data needs to be taken into account when using the data. The Personal Safety Survey delivers more detailed data for this offence type as a result of a more in-depth and private interview methodology and should be referred to if more data about this offence type is required.

Robbery is both a personal and acquisitive offence, which can cause some challenges in definitions and understanding differences in scope across statistical collections. Generally in 2008, for most jurisdictions, the recorded crime data for victims of robbery fell within the 95% confidence interval of the survey estimate of victims who reported a robbery to police. Western Australian and Tasmanian data for robbery was slightly below the survey estimate. 

Reporting rates from the Crime Victimisation Survey for the most recent incident of robbery increased significantly at the national level between 2008-09 and 2009-10, from 39% to 60%. This was driven by particular increases in reporting in Western Australia and Queensland. 

In the crime victimisation survey, break-in covers instances where there has been a break-in to a person’s home (including garages, sheds and caravans). As police recorded crime data is collected on a wider scope (including organisations, retail premises, etc.), the following data has been adjusted so that only residential properties are included. 

For both 2008 and 2009, almost all recorded crime data for break-in falls below the 95% confidence interval of the survey estimate for break-in, with the exception of New South Wales and Tasmania in 2008 and New South Wales and the Australian Capital Territory in 2009.

Motor vehicle theft is relatively simple to define in comparison to some of the other crime types included in the Crime Victimisation Survey. In addition, motor vehicle theft has a high reporting rate; 90% of victims reported their most recent incident of motor vehicle theft to police in 2009-10 nationally. 

With the exception of Queensland in 2008 and Tasmania and Queensland in 2009, jurisdictions’ recorded crime data was within the 95% confidence interval of the survey estimate for motor vehicle theft. The difference between the survey data and the recorded crime data for Tasmania and Queensland is very small. It is worth noting that all respondents in Queensland who said they were a victim of motor vehicle theft also said they reported the theft to police, which means that for motor vehicle theft, Queensland had an estimated reporting rate of 100%. As with all the results from the survey, the reporting rate is subject to sampling error. This may have contributed to the discrepancy between the survey estimate and the recorded crime figure. The true value of the reporting rate if all persons responded may be less than 100%, which would have the effect of reducing the number of victims who reported to police in the survey results.

The differences between the police recorded crime administrative data and the Crime Victimisation Survey data vary depending on the offence type. This is in part due to the nature of the offences themselves. Sexual assault for example has a low prevalence rate within the community and is a sensitive and very personal offence which means it often goes unreported to police or other authorities, whereas motor vehicle theft occurs much more frequently and is almost always reported to police in order to fulfil insurance requirements. Other effects differ between offence types due to the degrees of sampling error associated with the survey estimates, differences in recording practices, and difficulties in defining the offence itself.

The demonstration of the differences between the data sources shows the importance of understanding the purpose, coverage, scope and definitions of a particular source before making a decision to use the data.

Both the victimisation survey data and the administrative data contribute to informing users about the nature and extent of crime victimisation. Neither administrative statistics nor crime victimisation surveys alone can provide comprehensive information about crime, rather the two types of data sources can be used to test alternative hypotheses related to criminal activity (Statistics Canada, 1997). For example, data from victimisation surveys can be used to examine the extent to which particular offence types go unreported to police, which is not possible using police recorded crime data. However, the police recorded crime data is useful in understanding policing activities within each of the states and territories.

The purpose of this chapter is to provide a guide as to which data source is best suited for particular research purposes. All data sources have strengths and limitations, however, and these also need to be considered to determine whether a source is fit for purpose.

Generally, the administrative data sources are most useful for providing a view of crime victimisation that is coming to the attention of the Criminal Justice System and can subsequently become the focus of investigation and other police actions. It is this recorded crime to which operational managers and policy makers are able to respond most effectively. The survey data sources vary in the crime data information available, but generally provide more coverage and detail than the administrative data sources, and are better suited to detailed analysis, extensive use of demographic variables or national level analysis. For example, when looking for data to support research into incidents of current partner violence and the impact that this has on victims' lives, the Personal Safety Survey would be the best source. 

Other measures of data quality can be used to assess different sources of statistical information. The ABS assesses the fitness-for-purpose of data sources using the ABS Quality Framework. The framework contains six dimensions of quality that should be considered when choosing a data source: relevance, accuracy, timeliness, accessibility, interpretability and coherence. These dimensions are discussed in the Data Quality Declarations for each data source. Users should also refer to Methodology in publications in order to assess the usefulness of the data for their purposes.

Ultimately, the choice of which data source to use should be an informed decision made by the user based on the specific research questions to be answered, an understanding of the purpose of the data source and associated underlying methodology.

The key to using crime victimisation data accurately is to thoroughly understand the data source and make informed decisions about which data source to use to meet particular data requirements.

The aim of this paper has been to increase community understanding of how the experiences of victims of crime in Australia are measured and to explore why the findings from different data sources can produce different results. A discussion of the advantages and disadvantages of survey and administrative data collection highlighted the key issues to be aware of when using either data source. A working example of comparisons between different data sources was used to demonstrate the impact of varying data collection methodologies on data for a range of offence types. Guidance about which data source is best suited to particular research questions was also provided.

10 The offence categories used for the 2008 reference period for Recorded Crime – Victims are based on the Australian Standard Offence Classification, 1997 (cat. No. 1234.0) (ASOC97). The offence categories used for the 2009 reference period are based on the Australian Standard Offence Classification, 2008 (second edition) (cat. No. 1234.0) (ASOC08). ASOC provides a uniform national statistical framework for classifying offences.

11 The introduction of ASOC08 had limited impact on the Recorded Crime – Victims data due to minimal change to the activities included or excluded for the selected offences published for the collection. The primary impact was on the offence category of other theft, with data relating to fare evasion coming into scope of the collection. Other theft was not included in the analysis in this paper and the impact on the remaining offence types was limited.

12 During the course of the revised classification being implemented, jurisdictions also rectified a number of local offence codes that were miscoded to ASOC in previous years, thus impacting on data comparability between 2009 and prior years for certain offences. This has primarily affected the offence categories of other theft and assault for a number of jurisdictions resulting in increased victim counts.

13 Within Recorded Crime – Victims, the stage that a police investigation has reached after a period of 30 days has elapsed since the recording of the incident by police is referred to as the outcome of investigation.

14 Data for the 2008 reference period includes counts for outcomes of investigations that were determined as 'no crime', where these can be identified, with the exception of Queensland. The impact of this on the analysis presented in this paper is minimal due to the small proportion of 'no crimes' for the selected offences presented for the remaining jurisdictions that included no crimes in the recorded crime data.

15 Data for the 2009 reference period excludes counts for outcomes of investigations that were determined as ‘no crime’, where these can be identified. The removal of this has resulted in a reduction of victim counts in comparison to 2008, in those jurisdictions where where these counts had been included. Queensland data are not impacted as these outcomes were not included in prior years. Due to systems issues, data for the Northern Territory continues to include these outcomes. The impact of excluding 'no crime' counts from the data for 2009 has led to decreases in victim counts ranging from -0.2% to -9.3%. Most changes are below 5%.

16 For further information about the data quality of the police recorded crime data, please refer to Recorded Crime – Victims, Australia

17 When undertaking any sort of comparisons between data sets, it is imperative to understand the differences between the sources such as the scope of the data, how the data is collected, counting rules applied and definitions employed. The table below outlines the key metadata for the Recorded Crime – Victims and Crime Victimisation, Australia datasets.

18 For both data sources, the counting unit used for the analysis was the victim. For the Recorded Crime – Victims data, the victim count was restricted to persons, households and motor vehicles (for motor vehicle theft). Organisation data was excluded from the analysis as organisations were not in scope for the Crime Victimisation Survey. For the Crime Victimisation Survey data, only victims who stated the incident was reported to police was included in this analysis.

19 Recorded Crime – Victims data is collected in reference to a calendar year while the Crime Victimisation Survey is collected on a financial year basis. As such, comparisons between the data sources have been approximated by using the calendar year reference period for the Recorded Crime – Victims data, with the financial year data for the Crime Victimisation Survey data. This approach was used by the UK Home Office (Smith and Hoare, 2009) in comparing the British Crime Survey with police recorded crime.

20 The diagram below is a graphical representation of this methodology using the 2008 Recorded Crime – Victims reference period and the 2008-09 Crime Victimisation Survey data reference period as an example.

21 The 2008 Recorded Crime – Victims data refers to reports made to police, or an offence being detected by police, between January and December 2008. The 2008-09 Crime Victimisation Survey data refer to the 12 month period prior to the respondent participating in the survey. This means that the potential span of data from the survey ranges from July 2007 (the earliest offence date for respondents who were interviewed in July 2008, ie 12 months prior to interview) to June 2009 (the latest offence date for respondents who were interviewed in June 2009). The centre point to the survey recall period is June 2008, as this is a month which overlaps for all respondents. June 2008 is also the centre point of the Recorded Crime – Victims data. As such, the 2008-09 data for the Crime Victimisation Survey data and the 2008 Recorded Crime – Victims data are considered a suitable approximation for comparison.

22 Recorded Crime – Victims data has been restricted to victims aged 15 years and over for general offence types and 18 years and over for sexual assault, to align with the scope of the Crime Victimisation Survey. For instances where the age of the victim was unknown, these records were also excluded as it could not be determined whether they should be in scope for the analysis.

23 Recorded Crime – Victims data was adjusted for the Northern Territory to exclude victims living in very remote areas, to align with the scope of the Crime Victimisation Survey.

24 Detailed information about how each offence type is defined in the data sources and any adjustments that were made for the data confrontation exercise are provided in the table below.

25 Note that homicide, kidnapping / abduction and blackmail / extortion are not included in the survey so were not part of the analysis in this paper. Likewise, attempted break in and malicious property damage were included in the survey but not included in the Recorded Crime – Victims collection. Other theft is included in both data sources, however due to significant definitional differences it was also excluded from analysis.

This release previously used catalogue number 4500.0.55.001