Income data in the Census
Collecting personal income in the CensusThe 2011 Census collected personal income for all persons aged 15 years and over. People were asked to report the total of all their wages and salaries, government benefits, pensions, allowances and any other income they usually receive, before deductions for tax, superannuation contributions, health insurance, amounts salary sacrificed, or any other automatic deductions. People were asked to report their usual income by selecting one of the following ranges (not in actual dollars).
The ranges presented on the 2011 Census form (in table 1) were selected after analysing data from the 2007-08 Survey of Income and Housing (SIH), in which personal income was collected in actual dollar amounts.
Household and family incomeHousehold and family incomes were not collected in the Census but were derived from personal income data. As it is not possible to aggregate personal income ranges, a specific dollar amount was imputed for each personal income range selected by each family or household member. For the 2011 Census processing, the weighted median estimates of gross weekly personal income from the 2009-10 Survey of Income and Housing (SIH), adjusted for inflation to mid 2011, were calculated for each of the reported ranges in the Census. These medians were then allocated to each person who reported an income range in the Census.
The imputed values for each personal income range are shown below (table 2).
The imputed medians for personal income (in table 2) were then aggregated to create imputed household and family level incomes. The output ranges for family and household variables splits the top personal income range into six separate ranges to accommodate the higher aggregates created when summing imputed personal incomes to imputed family and household incomes (table 3).
Points to consider when using Census income dataThe reported personal income in ranges can be cross-classified by various other personal and demographic characteristics for various types of analysis. The imputed median personal incomes are generally also satisfactory approximations for various populations at aggregated levels (such as Australia, state or territory level, or metropolitan/ ex-metropolitan regions). However, these imputed values may overstate or understate the personal income when conducting in-depth analysis at finer levels. The extent of the over- or under-estimation will depend on personal characteristics such as the person's labour force status, hours worked, occupation, industry, age, educational qualifications, and family composition. The bias in such cases can result from the use of population-wide medians to represent the incomes of people with particular characteristics. Similarly, analyses of imputed personal incomes for smaller regions (below state / territory level) should take account of the imputation biases associated with the differing socio-economic compositions of the regions.
The imputed household and family income provide a practical approximation that would result in the majority of households and families being included in the Census imputed household and family income range that would have been derived had individuals reported their incomes in dollar amounts (rather than in ranges). This approximation is able to support analyses of household and family income ranges at broad areas of geography, but for smaller regions (below state/ territory level) the imputation biases associated with the differing socio-economic compositions of households and families and of the regions should be considered. For this reason, caution should be exercised in using Census imputed household and family income ranges when conducting analysis across geographic areas where socio-economic characteristics may vary, due to the potential for families and households to be allocated to an incorrect income range.
It should also be noted that individuals may tend to understate their incomes on the Census, compared with the amounts that would be reported in surveys designed specifically to measure incomes. The understatement will vary across ranges and by socio-economic characteristics and may be magnified by the imputation process. For example, the only way a person can understate their income is by selecting a Census range that is below the range appropriate to their actual income. Therefore, if a person with actual gross personal incomes near the bottom of one range understates their reported income by selecting the next lower range on the Census form (perhaps by forgetting about a small component of their income), their imputed income will be further dropped to the median income of those people in the SIH reporting actual incomes falling in the lower range. This understatement is magnified when summing imputed personal incomes to family or household levels.
For the above reasons, care should be exercised in the use of Census income information, and particularly when comparing imputed income across geography where the socio-economic characteristics vary across geography. In particular, the reporting of Census income medians or the use of the ratio of Census reported housing costs to Census imputed incomes is likely to significantly overstate such ratios for many subpopulations. Finally, it should be noted that the median income values from the Survey of Income and Housing (SIH) used as the imputed median personal incomes in the Census are themselves subject to sampling error, since SIH is a sample survey.