Australian Institute of Criminology

Skip to content

Analysis of existing data

Stage one of the present study involved the analysis of existing datasets that could assist in determining the nature and extent of international student victimisation in Australia. The only data available was derived from the Australian component of the ICVS and the NHMP.

International Crime Victims Survey

The recent incidents of crime against international students have raised the profile of migrants as key groups who have contact with the criminal justice system but for whom there is little information and data available. To date, crime victimisation surveys have focused on deriving national population estimates and as a result, have invariably under-sampled migrant populations, including international student populations. While limited in scope, a recent exception is the 2004 Australian component of the ICVS.

In 2004, the Australian component of the ICVS surveyed a total of 7,000 respondents; a considerably larger sample than that collected in previous years. Designed to ensure the capture of a nationally representative sample, an over-sample of migrant populations was also collected. Initial analysis of the survey showed that overall, 52 percent of respondents reported at least one criminal victimisation in the five years preceding their interview; while 17 percent had been victims of crime in the 12 months prior. Nine percent of respondents in the 12 months prior to interview had experienced a personal crime (which included robbery, theft of personal property, assault and threats), with rates of personal crime victimisation highest for assault/threats and personal theft, and lowest for robbery.

The AIC undertook a secondary analysis of ICVS data for this study in order to examine the relationship between ethnicity and victimisation; and to examine previous year’s ICVS data to compare changes in victimisation over time.

Students

Those respondents who indicated their current main activity was studying, comprised approximately 10 percent of the 2004 Australian ICVS sample (n=681); of this group, 196 (29%) were born outside of Australia. The number of student respondents who reported having experienced personal victimisation (assault and threats, robbery or personal theft) in the past 12 months and past five years is presented in Table 2; the results are provided separately for both overseas and Australian-born students, and include the country of birth for those born overseas. In order to reduce the sampling bias that resulted from the random population selection process, weighted victimisation rates for each of the four groups have been created (see Figure 1). Overall:

  • Forty-seven percent of Australian-born students reported being a victim of a personal crime in the five years prior to interview. The comparable five year rate of victimisation was 38 percent for overseas-born students, 24 percent for overseas-born non-students and 28 percent for Australian-born non-students.
  • Overseas-born students had the highest rate of recent victimisation (past 12 months), estimated at 15 percent. However, this was not statistically higher than for Australia-born students (13%).
Table 2: Victimisation by country of birth (n)
Country of birthOne yearFive years
Pacific Islands 1 1
United Kingdom/Ireland 1 4
New Zealand 0 2
North America 1 2
South and Central America 1 1
Greece 0 0
Turkey 0 1
Other Europe 4 9
Lebanon 0 2
Other Middle East 1 11
North Africa 0 1
Horn of Africa 0 1
Other Africa 3 5
Central Asia 0 1
South Asia 0 2
China 1 3
Vietnam 2 4
Other East or southeast Asia 4 15
Total born overseas 19 65
Total born in Australia 62 225
Total students 81 290

Source: AIC ICVS 2004 [computer file]

There are a number of caveats that should be considered when interpreting these data. First, despite instructions to the respondents, it is not possible to confirm that a person’s reported experience of crime actually occurred in Australia. This is particularly pertinent for overseas-born students and particularly for five year victimisation rates, given the students may spend a substantial portion of time in their home or other countries before, during or after completion of their studies. Second, specific countries of interest (eg India) have been automatically aggregated into broader geographical regions, preventing further analysis for specific countries. Finally, since this survey was a general population survey, the sample size of overseas-born students is quite small for statistical analysis. Consequently, caution must be exercised when interpreting the results.

Despite these caveats, it appears that overseas-born students were no more or less likely in 2004 to report recent personal victimisation for the past 12 month period. It is unknown, however, whether these differences represent true victimisation differentials associated with a person’s student or migrant status, or whether these differences are the result of other confounding factors such as where a person lived, how much they earned and the types of activities they undertook (eg those who regularly engaged in night-time activities such as work, attending entertainment venues etc would have a higher risk of being a victim of crime).

Figure 1: Past year personal victimisation by migrant and student status (weighted %)

Figure 1: Past year personal victimisation by migrant and student status (weighted %)

Source: AIC ICVS 2004 [computer file]

Isolating the independent effect of a person’s migrant or student status from other factors can be done using a multivariate analysis such as logistic regression. These analyses are designed to identify each factor’s independent contribution to increasing (or decreasing) a person’s probability of being a victim of crime. There were three multivariate questions of particular interest that were able to be tested:

  • Are students generally more likely to be victims of personal crime once other factors are controlled for (alternatively, are other factors more important in explaining victimisation than a person’s student status)?
  • Are those born overseas (irrespective of student status) more or less likely to be victims of personal crime once other factors are controlled for?
  • Are overseas born students in particular, more or less likely than Australian born students to be victims of personal crime?

Table 3 provides the results of two multivariate models. Both include information about a respondent’s migrant and student status, as well as an interaction between the two factors. The second model is an extension of the first, with additional controls for a selected range of personal and situational characteristics.

Table 3: Risk factors for personal victimisation for Australian born (logistic regression)
bOdds ratiop
Model 1
Overseas born (vs Australian born) -0.21 0.81 0.14
Student (vs non-student) 0.48 1.62 0.00
Student+Overseas born (interaction) 0.41 1.51 0.24
Constant -2.38 0.00
Model 2
Male (vs female) -0.10 0.90 0.34
Aged 35 years or older (vs less than 35 years) 0.27 1.31 0.05
Unmarried (vs married) 0.38 1.46 0.00
Income less than $400 per week (vs >$400 per week) -0.34 0.71 0.07
Language other than English (vs English only) -0.18 0.83 0.28
Evening activities almost daily (vs evening activities less than once a week) 0.54 1.72 0.00
Seeing drug-related activity in local area (vs don’t see drugs) 0.91 2.49 0.00
Overseas born (vs Australian born) -0.04 0.96 0.82
Student (vs non-student) 0.00 1.00 0.98
Student+Overseas born (interaction) 0.51 1.67 0.16
Constant -2.72 0.00

Source: AIC ICVS 2004 [computer file]

Model 1 confirmed the findings presented earlier that:

  • being born overseas was not linked to an increase in the risk of personal victimisation;
  • students were statistically more likely to have experienced personal victimisation than non-students;
  • while overseas-born students were at a slightly higher risk of personal victimisation, the modest difference was not statistically significant. There is, therefore, insufficient evidence to suggest that overseas born students are at greater risk of personal victimisation.

Adding additional controls in Model 2 re-adjusted the statistical relationship between a person’s status as a student and their risk of victimisation. This model illustrates that after the effect of other personal and contextual factors are taken into account, there was no difference in the risk of victimisation between students and non-students or, indeed, between overseas-born students and Australian-born students.

The fact that a respondent’s student status in Model 2 was no longer statistically significant suggests that a person’s risk of victimisation isn’t tied to their identification as a student, but rather the other personal and contextual influences that are more prevalent among student populations. In Model 2, there were three other factors found to be statistically significant predictors of personal victimisation. These were:

  • being unmarried (single or divorced);
  • regularly undertaking evening activities, as opposed to staying at home most evenings (or=1.71, p=0.00); and
  • seeing or witnessing drug-related activities in the local area (or=2.48, p=0.00). This may reflect either a geographical relationship (ie some areas are more prone to violence and other criminal activities and therefore, victimisation is higher among residents of those areas), or alternatively, it may reflect the types of activities the respondent is engaged in. For example, respondents that have knowledge of local drug-related activities may engage within their local community in ways that increases their risk of victimisation.

Finally, it should be recognised that these models are of very limited utility in that the pseudo R2 value for Model 2 is estimated at 0.04. In essence, 96 percent of the differences between respondents that might account for differing level of experiencing victimisation remain unexplained by these analyses. Further, while a conventional test—the Hosmer and Lemeshow goodness of fit test—found no significant mis-specification in the model, estimates of the Area Under the Curve (AUC=0.66) suggest that the model has only a modest ability to differentiate victims from non-victims. A model with an AUC value of 0.50 is only as reliable as random chance and the closer the AUC value is to 1.00 the more accurate the model is in minimising false positive and false negative predictions. With an AUC=0.66, Model 2 has what is referred to as only modest predictive validity (Hosmer & Lemeshow 2000).

Overall, these regression diagnostics suggest that to a large extent, the factors associated with victimisation are not well captured in the ICVS dataset and that there remains a significant amount of unexplained variability in victimisation risks. Development of a more appropriately targeted survey, with more fine-grained personal and contextual factors, would be required to improve the understanding of who becomes a victim and the causes, or risk factors, for increased (or decreased) risk of victimisation.

Homicide data

The AIC’s NHMP has collected detailed information on each homicide occurring in Australia since 1990. An analysis of NHMP records was undertaken to break down homicides by race from 1996–97i to 2007–08 (latest available dataset) in order to identify:

  • the number of Indian students killed in Australia; and
  • the number of Indian student homicides that were motivated by racial vilification/discrimination.

The search process yielded eight Indian student homicide victims from seven homicide incidents—one incident involved two Indian student victims. During this same period, a total of 3,530 homicides were recorded across Australia.

Of the eight victims killed:

  • none were indicated to have involved elements of racial vilification or discrimination;
  • three were killed by strangers during the course of the commission of another offence (robbery and police pursuit);
  • four were killed during a domestic or family violence altercation; and
  • one was killed by a friend after an altercation over money.

When examining these deaths against the number of student visas granted to Indian nationals, the rate of homicide per 100,000 offshore student visas granted is low (see Table 4).

Table 4: Indian student victims of homicide, rate per 100,000 relevant student population
YearHomicide victims (n)Visas granted offshore where source country was India (n)Rate per 100,000 student visas granted
1996–97 3
1998–99 1
1999–2000 1
2005–06 1 15,396 6.5
2006–07 1 28,949 3.5
2007–08 1 39,015 2.6

Source: AIC NHMP 2007–08 [computer file]

A further 56 victims of homicide who may have been Indian but were not listed as students were identified. Of these, 21 were identified as Indian nationals from the hardcopy homicide records. Relevant victim characteristics and the recorded motive for the offence are presented in Table 5. It is important to note that none of the incidents were deemed by police officers and/or the AIC to have been racially motivated after detailed assessment of the case files.

While analysis of these sources provides some insight into risk factors and circumstances surrounding crimes against international students, the under-sampling of migrant populations, in particular international students, in crime victimisation surveys and the small number of homicide cases available for analysis highlighted the need for developing a method by which new data could be collected and analysed to answer the research questions. The methodology and findings from the second stage of the AIC’s research, derived from the matching of police and immigration records, are the focus of the remainder of this report.

Table 5: Indian homicide victims—characteristics and motive for offence
YearGenderAge in yrsStudent (Y/N)OccupationMotive
1996–97 Female 48 N Unknown Desertion/termination
1996–97 Female 23 Y Unknown Desertion/termination
1996–97 Male 21 Y Unknown Money
1996–97 Female 22 Y Unknown Money
1997–98 Male 57 Unknown Unknown Unknown
1997–98 Female 50 Unknown Unknown Money
1998–99 Male 25 Y Service station attendant Other argument
1999–2000 Male 31 Y Service station attendant Money
2002–03 Male 32 N Unknown Money
2003–04 Female 39 N Unknown Argument of domestic nature
2004–05 Male 24 N Courier Other argument
2004–05 Male 30 Unemployed Drugs
2005–06 Female 18 Y Unknown Argument of domestic nature
2005–06 Male 25 N Restaurant owner Other argument
2006–07 Male 27 Y Taxi driver Money
2007–08 Male 24 Y Unknown Money
2007–08 Female 27 N Administration Manager Desertion/termination
2007–08 Female 44 N Supreme Court registrar
2007–08 Female 74 N Aged pension/retired No apparent motive
2007–08 Male 51 N Business owner Alcohol-related argument
2007–08 Female 38 n/a Unemployed Incident was domestic related
2007–08 Female 9 n/a Not applicable Incident was domestic related

Source: AIC NHMP 2007–08 [computer file]


End notes

i Data are for homicides from 1996–97 to 2007–08. The ability to identify an Indian student required detailed information about the racial appearance, country of birth and current employment status of the victim. It is possible that some students may not have been identified because information was missing or not provided by the police. In addition, since only one employment status is recorded, some students may have been recorded as ‘part-time employed’ instead of ‘student’ if they were killed during the course of their employment. While every effort was made to crosscheck the files for secondary employment, it is possible that some Indian students were not included because there was insufficient information to identify their status as a student. Moreover, it was not possible to identify the number of Indian students who were living in Australia on a current or expired student visa at the time of their death, nor was it possible to identify how long the deceased had been living in Australia.