After adjusting for conceptual and scope differences between data sources, a statistical discrepancy remains between the number of filled jobs as reported by businesses and the number of filled jobs as reported by households.
These discrepancies represent the cumulative impact of data source error, including survey error, and modelling error. Survey error includes both sampling error and non-sampling error. Sampling error is the predictable variability arising from the use of samples, rather than a complete enumeration of the populations of enterprises and households. Non sampling error is all other error in the estimate, and includes:
- Error arising from the reliability of the survey population and related benchmark data, e.g. the accuracy, completeness and timeliness of the Business Register from which business survey samples are drawn, or the reliability of Estimated Resident Population data used in benchmarking the Labour Force Survey;
- Error arising from data used in the estimation and imputation procedures applied in both business and household surveys;
- Error embedded in the estimation and imputation models used in surveys, for example incorrect assumption that missing firm data is similar to that of reporting firms of comparable size in the same industry; and
- Error made by respondents in reporting data, for example the Labour Force Survey relies on one responsible adult in each household to accurately report on the employment status of all other adults in the household, including industry of employment and hours worked in the survey reference week. Industry can be misreported where people are employed by labour hire firms, but actually work in other industries such as Mining, Construction or Manufacturing.
Error can occur in non-survey data sources, such as missing data or misclassification in government administrative records used directly in the Australian Labour Account. Error could occur in the industry classification of sponsored visa holders, or in the reported number of persons in the permanent defence forces.
Modelling error reflects errors embedded in the modelling assumptions used in the Australian Labour Account, for example in assuming that the proportion of children aged under 15 years who work has remained constant since 2006, or in assuming that Quarterly Business Indicators Survey
(ABS cat. no. 5676.0) employment movements accurately reflect quarterly change in the latest available annual data.
The balanced Australian Labour Account estimates apply knowledge of the known strengths and weaknesses of data sources and methodologies, to derive a single estimate of the number of filled jobs.
The balanced estimate of numbers of filled jobs impacts on other data in the Australian Labour Account that incorporate that estimate in their calculation. This includes balanced estimates of numbers of persons employed and hours worked.
Two general observations about data source quality are relevant in deriving a balanced estimate of numbers of filled jobs:
- Household estimates of numbers of filled jobs are considered more reliable at a total economy level. Household data are mainly sourced from the Labour Force Survey, which applies a consistent methodology and asks a consistent set of questions of a statistically robust sample of persons about the number of jobs held by employed persons in their household. By contrast, no single business survey covers the whole economy. Estimates of the total number of filled jobs from the business side are derived from three separate surveys (EAS, SEE and QBIS), supplemented by data obtained from the Australian Business Register. Each source has a different methodology, a different sample, and asks different questions. Adjustments are required to counter overlap. Growth in household side filled jobs is more consistent over time with growth in related economic data (GDP and Compensation of Employees) at a total economy level than growth in business side data.
- Business sources are considered more reliable in estimating the distribution of jobs across industries. The numbers of filled jobs reported by each business survey respondent are automatically coded to the industry classification of that business. This implies that labour input is correctly linked to related production, employment related costs and compensation.
Whilst additional considerations are taken into account at the industry level, the balanced estimate of filled jobs incorporates the advantage of the industry distribution derived from business side data, within a total economy estimate sourced from household side data.