WEIGHTS

BENCHMARKS AND WEIGHTING METHOD

Weighting is the process of adjusting results from a sample survey to infer results for the total in scope population whether that be persons, income units or households. To do this, a 'weight' is allocated to each sample unit e.g. a person or a household. The weight is a value which indicates how many population units are represented by the sample unit. The first step in calculating weights for each unit is to assign an initial weight, which is the inverse of the probability of being selected in the survey. For example, if the probability of a household being selected in the survey was one in 600, then the household would have an initial weight of 600 (that is, it represents 600 households).

The initial weights are then calibrated to align with independent estimates of the population of interest, referred to as benchmarks. Weights calibrated against population benchmarks ensure that the survey estimates conform to the independently estimated distribution of the population rather than to the distribution within the sample itself. The calibration also includes an adjustment to account for differences in the sample enumerated in each of the four quarters over the year. This is a change from previous cycles of SIH where the quarter adjustment was made to initial weights rather than at the calibration step.

Most of the independent person and household benchmarks are based on demography estimates of numbers of persons and households in Australia. The benchmarks are adjusted to include persons and households residing in private dwellings only and to exclude persons living in very remote areas, and therefore do not, and are not intended to, match estimates of the Australian resident population published in other ABS publications. The demography estimates of persons (estimated resident population - ERP) and households used in SIH 2015–16 are built up from the 2011 Census.

SIH weighting

In the 2015–16 SIH, as in 2007–08, 2009–10 and 2011–12 and 2013–14, all persons in each household were assigned a weight. This differs from the 2005–06 SIH where children aged 0–14 years were not given separate weights, but household counts of the number of children were benchmarked to population totals.

As in 2009-10, an additional sample was used to provide detailed income and expenditure information for households where the main source of income is government pensions, benefits and allowances. The additional households completed the full SIH questionnaire and contribute to the published SIH estimates. The pensioner sample is incorporated using a pooled estimation approach. This differs from the method used in 2009-10, where the base HES and additional pensioner samples were combined using composite estimation after weighting each component separately.

The benchmarks used in the calibration of the final weights for the 2015–16 SIH were:

• number of persons -
  • by state or territory by labour force status ('Employed', 'Unemployed' and 'Not in the labour force') (except NT which does not use labour force status); and
  • by state or territory by age by sex;
    • in five year age groups up to 80+ years for all states and territories (excluding NT);
    • in five year age groups up to 70+ years for the NT;

• numbers of households -
  • by state, by indexation quarter by capital city/balance of state (except NT and ACT which only use state);
  • by state, by household composition (number of adults (1, 2 or 3+) and whether or not the household contains children) (except NT which only uses whether or not the household contains children).

• value of government benefit cash transfers -
  • consisting of age pension, family tax benefit, disability support pension, newstart, parenting payments, carer payment, Commonwealth rent assistance, youth allowance, carer allowance, school kids bonus, Austudy/Abstudy, carer supplement, widow allowance, wife pension, sickness allowance, partner allowance, special benefit, utilities allowance, pension supplement, service pension, war widow(er)'s pension.

HES weighting

In the 2009-10 and 2015–16 HES, all persons in each household were assigned a weight. The final SIH weights were used as the initial weights for the HES, with a sub-sampling adjustment being applied to incorporate the additional pensioner sample. This differs from the method used in 2009-10, where the base HES and additional pensioner samples were combined using composite estimation after weighting each component separately.

The HES survey was benchmarked to the in scope estimated resident population (ERP) and the estimated number of households in the population, and to a number of estimates produced from the SIH and the Monthly Population Survey. ERP benchmarks were preliminary, based on the 2011 Census. An additional benchmark using administrative payment data was applied for households that received government payments.

The population benchmarks used in the calibration of the final weights for the 2015-16 HES were:

• number of persons -
  • by state or the ACT by labour force status ('Employed', 'Unemployed' and 'Not in the labour force'); and
  • by state or territory by age by sex;
    • five year age groups up to 80+ years for NSW, Vic., Qld., SA and WA;
    • ten year age groups up to 70+ years for the Tas., NT and ACT;

• number of households -
  • by state, by pension indexation quarter, by greater capital city area/balance of state (excluding NT and the ACT which only use state);
  • by state, by household composition (number of adults (1,2 or 3+) and whether or not the household contains children; excluding the NT which uses only number of adults of 1+).

In addition to the population benchmarks presented above, the following SIH estimates were used as benchmarks at the state level in weighting the HES sample:

• total weekly household income from all sources;
• total weekly household income from own unincorporated business;
• total weekly household income from wages and salaries;
• total weekly household income from government pensions and allowances; and
• household tenure type.

The independent person and household benchmarks are based on demography estimates of numbers of persons and households in Australia, except for the labour force benchmarks which are based on the Monthly Population Survey. The benchmarks are adjusted to include persons and households residing in private dwellings only and to exclude persons living in very remote areas, and therefore do not, and are not intended to, match estimates of the Australian resident population published in other ABS publications.

Composite estimation

Composite estimation was used to obtain the optimal proportions for combining the targeted sample and main HES sample of pensioner households at a state level.

The Pensioner sample was assigned independent initial probability weights as it was selected separately from the SIH and HES samples. The initial weights were calibrated to the same population benchmarks as used in the SIH calibration with broader grouping of the benchmarks to accommodate the smaller sample size. The calibration also included an adjustment to ensure that the observed proportion of identified pensioner households from the subsample selected to complete the survey regardless of screening result was maintained.

In order to derive the most accurate combined estimates for the SIH, HES and additional pensioner samples, composite estimation was used. This method calculates the optimal proportion for combining estimates from the main SIH and the additional pensioner samples at the State by part of state (the additional pensioner sample includes only Metropolitan) by pensioner/non-pensioner household level.

The composite estimates are given by

where Y^_composite is the final composite estimate of some population parameter, Y^_HIES is the SIH sample estimate and Y^_PHES is the additional pensioner sample estimate of that parameter. The composite estimation coefficients (a , 1-a ) of the optimal proportions are the values that minimise the variance of the composite estimate. The values of a were calculated using the total income variable. Weights are then multiplied by a for SIH sample and (1-a) for the additional pensioner sample to achieve optimised composite weights.

After the main SIH and PHES samples were combined a final benchmarking calibration was applied to ensure that the final weights align with population benchmarks.


Consistency of SIH and HES estimates

Although the HES and the SIH are integrated, the estimates for common items published in both the SIH publication Household Income and Wealth, Australia, 2015-16 (cat. no. 6523.0) and HES publication Household Expenditure Survey 2015–16, Summary of Results (cat. no. 6530.0) are unlikely to have exactly the same values, unless calibrated between the surveys. The estimates taken from only the HES subsample are subject to greater sampling variability than the full SIH estimates.

Estimation

Estimates produced from the SIH and HES are usually in the form of averages (e.g. average weekly income of couple households with dependent children), or counts (e.g. total number of households that own their dwelling or total number of persons living in households that own their own dwelling). For counts of households, the estimate was obtained by summing the weights for the responding households in the required group (e.g. those owning their own dwelling). For counts of persons, the household weights were multiplied by the number of persons in the household before summing. The SIH and HES collect data on the number of people, including children, in each household but separate records with income and most detailed data were only collected for people 15 years and older.

Average income values are obtained in two different ways, depending on whether mean gross household income or mean equivalised disposable household income is being derived. Estimates of mean gross household income are calculated on a household weighted basis. They are obtained by multiplying the gross income of each household by the weight of the household, summing across all households and then dividing by the estimated number of households. For example, the mean gross household income of couple households with dependent children is the weighted sum of the gross income of each such household divided by the estimated number of those households.

Estimates of mean equivalised disposable household income are calculated on a person weighted basis. They are obtained by multiplying the equivalised disposable income of each household by the number of people in the household (including children) and by the weight of the household, summing across all households and then dividing by the estimated number of people in the population group.

Averages are obtained by adding the weighted household values, and then dividing by the estimated number of households. For example, average weekly expenditure on clothing and footwear by Victorian households is the weighted sum of the average weekly expenditure of each selected household in Victoria who reported such expenditure, divided by the estimated number of households in Victoria. Note that the denominator is the total number of households and not just the number of households which reported expenditure on a particular item.