Skip to start of content

Home → Publications → Reports → Technical and background paper series → 16 → Survey weights (in: The Australian component of the 2004 International Crime Victimisation Survey)

The Australian component of the 2004 International Crime Victimisation Survey

Graham Challice and Holly Johnson
ISBN 0 642 53889 1 ; ISSN 1445-7261
Australian Institute of Criminology, 2005
(Technical and background paper series, no. 16)

Download

Survey weights

The Social Research Centre provided the AIC with a clean SPSS datafile combining the main community sample and the migrant sample. Survey weights serve two main purposes:

  • they enable the survey estimates to be corrected for non-response and coverage problems; and
  • they enable the sample to be adjusted to more accurately represent the population of interest, defined in terms of age group, gender and geographic location.

5.1 Main community sample

A three-stage approach to person weighting was adopted for the 2004 ICVS, which adjusted for:

  • respondent chance of selection within the household, including an adjustment for the disproportionate chance of selection;
  • household chance of selection, based on the number of landlines for private use; and
  • ABS 2001 census population benchmarks for age (16-19, 20-24, 25-29, five-year breaks to 65-69, and 70 plus), and gender within state capital/rest of state location.

Household weights were calculated using ABS 2001 census population benchmarks for number of households with one, two, three or four plus persons aged 16 or older within state capital/rest of state location.

The final merged data file contains several sets of weights including:

  • a person weight and a household weight specific to the main community component (n=6000);
  • a person weight and a household weight specific to the immigrant component (n=1001):
  • a person weight and a household weight specific to all persons of Vietnamese and Middle Eastern background who were interviewed (n=1119), whether interviewed as part of the migrant sample (n=1001), or interviewed as part of the main community component (n=118); and
  • a person weight and a household weight for the total combined sample (n=7001).

The process for deriving merged file person population targets was as follows:

  1. identify target national population from 2001 census data, by gender, age group (16-24, 25-34, 35-54, 55 plus) and location (state capital/rest of state);
  2. subtract persons born in Vietnam and Middle East by age, gender and location, from information provided by ABS special data run;
  3. subtract persons with at least one parent born Vietnam or the Middle East, from information provided by ABS special data run; and
  4. the residual is the target population for the non-Vietnamese, non-Middle East sample.

There are two important points to note in relation to final merged data file person weights.

  • The adoption of a different age group structure within the person weighting matrix for the final merged data file (n=7001), relative to the original main community component data file (n=6000). A similar structure to the original main community file, with finer breaks in age group, would have resulted in too many empty cells or cells with very small bases within the immigrant sample in the final merged file;
  • The target populations for the two components of the migrant sample person weights were slightly different in the merged file (n=7001), compared with the original immigrant file (n=1001) due to differences in the source population data. The special run of ABS data used for Vietnamese-and Middle East-born persons in the final merged data file was based on precise age break information, whereas the target population for the 16-24-year-old group in the original immigrant data file (n=1000) was derived from information provided by DIMIA, specifying the population of interest in terms of 15-24-year-olds.

The process for deriving merged file household population targets was similar to that used for person population targets, with a special ABS run being used to identify the number of households with at least one person born in Vietnam/Middle East, or whose parents were born in Vietnam/Middle East.

5.2 Migrant sample

As for the main community component, a three-stage approach to person weighting was adopted for the original migrant component data file (n=1001). The third stage included identifying an appropriate target population for both first generation (Vietnamese-or Middle East-born population) and second generation (Australian-born persons whose parents were born in Vietnam or the Middle East) migrants. The final approach to person weighting adjusted for:

  • respondent chance of selection within the household from S6b (see questionnaire, Appendix A), including an adjustment for the disproportionate chance of selection;
  • household chance of selection, based on the number of landlines for private use at Q400b (see questionnaire, Appendix A);
  • ABS 2001 population benchmarks (as supplied by DIMIA) for age of persons born in Vietnam or the Middle East (15-24, 25-34, 35-54, 55 plus. The 16-24-year-old population was estimated from this information for the purpose of weighting), and gender at the national level (eight cells per language group); and
  • ABS 2001 population benchmarks (special run by ABS), for age of persons born in Australia (16-24, 25-34, 35-54, 55 plus) whose parents were born in Vietnam or the Middle East, by gender at the national level (eight cells per language group).

Household population targets for the original immigrant file (n=1001) were based on ABS population benchmarks from the 2001 census (special run), using number of residents aged 16 or older in the household (one, two, three or four plus), where at least one person was born in Vietnam/Middle East, or had a parent born in Vietnam/Middle East, at the national level.

For the final merged data file (n=7001), population targets for the migrant component were based on a special ABS run with a final level of geographic precision (state capital/rest of state compared with national for the original migrant component data file) and the same age ranges, but with accurate counts for the 16-24 age group. Due to the number of empty cells, the final migrant sample geographic stratification within the merged file was Sydney/Melbourne/other, giving 24 cells in total.