1504.0 - Methodological News, Sep 2001  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 14/11/2001   
   Page tools: Print Print Page Print all pages in this productPrint All

2001 MPS REDESIGN

Introduction

The Monthly Population Survey (MPS), incorporating the Labour Force Survey (LFS) and Supplementary questionnaires, is a survey including some 60,000 persons, Australia wide each month. An area based multistage design is used, selecting Collection Districts (CDs) at the first stage, blocks at the second stage and dwellings at the third stage. The survey employs a workforce of some 620 interviewers and costs almost $40 million to run over a five year period. It is one of the largest surveys undertaken by the ABS and is also of key importance to users as it provides the labour force estimates that are input to the National Accounts. Every five years, shortly after the Population Census is conducted, the MPS is redesigned to improve the efficiency of the survey using the latest available cost and variance information.

The purpose of the sample redesign is to choose the design that gives the best trade-off between costs and variance. In order to do this, costs and variances are each expressed as a mathematical function of the sample design parameters: the number of CDs to select; and the number of dwellings selected per CD (referred to as the cluster size). This predicts what the costs and variances will be for a given choice of design parameters, enabling us to minimise variance for a prescribed cost.

New Features

The 2001 redesign is due to be completed in March 2002 and will be phased in during October 2002 to June 2003. The redesign includes two new features:

  • A more coordinated approach towards collecting and producing Indigenous statistics from household surveys. This will be achieved via a separate Indigenous Communities Framework (ICF) in addition to the Private Dwelling and Special Dwelling frameworks.
  • The incorporation of the concept of remoteness into the stratification of the less densely populated rural areas. This will be achieved using the Access/ Remoteness Index for Australia (ARIA) which has been included in the 2001 Australian Standard Geographical Classification (ASGC).

Stratification

The introduction of ARIA has replaced sampled and sparse area types in the stratification, which had been based on SLA level population density. This will have the advantage of keeping more remote towns and their surrounding district together in the same stratum. In addition, the use of ARIA also helps ensure that small islands off the north coast of Australia are treated as very remote rather than like larger towns because of their high density.

Variance Model

A substantial increase in computing power has made it possible to produce a variance model based on fifty different designs, considering all possible samples under each design and emulating the block formation and selection practices used in the MPS. This has resulted in variance models with a high level of fit. In addition variance models were obtained for the first time for smaller area types such as the sampled, sparse and indigenous area types.

The resulting variance model was further adjusted to reflect actual LFS accuracy levels for the middle of the design period. These adjustments allowed for the following differences between Census and MPS:
  • post-stratification estimation used in the MPS;
  • definitional and questionnaire differences between Census and LFS labour force data items;
  • relative changes in population and sample size;
  • sample loss/non-response;
  • changes in variance structure, such as a change in the ratio between within block and between block variance, during the life of the 1996 design; and
  • use of more final census counts as CD measures of size for the 2001 selections.

The Cost Model

The cost model was fitted to detailed cost data obtained in May 1999 which gave a more precise split by cost components and was the first available data to comprehensively represent the cost dynamics under the telephone interviewing approach. The cost model was subsequently adjusted for monthly seasonal changes and to reflect those Payment To Agents (PTA) costs not captured electronically, such as training, Commonwealth vehicles, superannuation and long service leave.

Optimisation Method

An optimisation method for 2001 has been developed to minimise Australian level variances in order to meet a specified cost value and specified relativities in state accuracy levels. This is a departure from previous allocations in which state accuracies were controlled more indirectly by adjusting state sampling fractions. The new approach is transparent in that the desired relativities in state accuracy requirements are specified as input to the optimisation. The method also ensures equal probability sampling within state.

A preliminary sample optimisation has been produced based on projected PTA expenditure for the 2001 design period which delivers slightly lower relative standard errors for employment and unemployment than those that were achieved during 1996. However, due to the increased emphasis on the PTA budget for this design, the optimisation will not be finalised until after the October Management Meeting where a decision will be made on the final budgetary allocation for the 2001 design MPS PTA expenses.

For more details, please contact Daniel Elazar in CO on (02) 6252 6962.

Email: daniel.elazar@abs.gov.au