Sample Rotation Patterns for the EU Labour Force Sample Survey

1
Sample Rotation Patterns for the EU Labour Force
Sample Survey

• Marinela Kustura Katja Rutar

2
Introduction

• The EU-LFS needs to provide estimates of changes
  in the characteristics of the labour force over
  time.
• These estimates are produced using rotating
  sampling designs.
• The EU Member States include different rotation
  patterns in their sampling designs for the LFS.
• We examine consequences that different rotation
  patterns have on the reliability of estimates and
  costs of the survey.

3
Types of Estimates in the EU-LFS

• The EU-LFS needs to provide different types of
  estimates in terms of time dimension, such as
• Estimates of levels at points in time,
• Average estimates of levels,
• Estimates of net changes in levels between points
  in time, and
• Estimates of gross changes over time at an
  individual level.

4
Estimates of Changes Rotating Sampling Designs

• Estimates of net changes in levels are more
  reliable if the data are collected from the same
  individuals.
• Estimates of gross changes can be obtained only
  if the data are collected from the same
  individuals.
• Rotating sampling designs permit the estimation
  of gross changes.
• Rotating sampling designs enable reliable
  estimates of net changes in levels at different
  point in time.

5
Sample Rotation

• Sample rotation means that proportion of the
  sample is replaced each survey round.
• Partial rotation of a sample is a compromise
  between total rotation of a sample, i.e.
  replacement of a whole sample, and no rotation of
  a sample, i.e. a panel survey.
• In the Labour Force Surveys, a part of the sample
  is replaced each survey round according to a
  pre-defined rotation pattern.

6
Sample Rotation

• The sample consists of number of sub-samples or
  rotation groups.
• Each survey round, a new rotation group is
  introduced into the sample to replace the survey
  round that leaving the sample.
• The part of the sample that is common for two
  survey rounds permits the estimation of gross
  changes and enables reliable estimates of net
  changes.

7
Rotation Patterns

• Rotation pattern is denoted by R-(R)-R, where R
  is the number of successive survey rounds in
  which each rotation group is participating in the
  survey, and (R) is the number of survey rounds
  skipped.
• The rotation pattern 2-(2)-2 means that in 6
  successive quarters each rotation group is
  participating in the survey in the first two
  quarters and in the fifth and sixth quarter, but
  not in the third and fourth quarter.

8
Rotation Patterns

• The rotation pattern 6-(0) means that each
  rotation group is participating in the survey in
  6 successive quarters with none skipped.
• The rotation pattern determines
• The reliability of estimates of net changes in
  levels between points in time,
• The reliability of average estimates of levels.

9
Choosing a Rotation Pattern

• When choosing a rotation pattern, we need to take
  into account
• The relative priorities given to estimates of
  gross changes, net changes, and average estimates
  of levels.
• Over which survey rounds changes are to be
  measured,
• Cost and operational factors,
• The extent to which respondents can be repeatedly
  interviewed.

10
Disadvantages of retaining a part of the sample

• Low response rates
• High attrition rates
• Rotation group bias, i.e. bias resulting from
  continued panel participation
• Rotation group bias is difference in estimates
  produced by different rotation groups.
• Estimates depend upon how many times a rotation
  group has been interviewed previously.

11
The EU-LFS

• The EU-LFS is a rotating random sample survey of
  persons in private households.
• It is a continuous quarterly survey, providing
  quarterly and annual results.
• Sampling designs and rotation patterns are not
  harmonised among the EU Member States.

12
The EU-LFS

• In 2003, 17 EU Member States conducted rotating
  quarterly LFS, and have reported on the details
  of rotation patterns included in their sampling
  designs for the LFS.
• Eight different rotation patterns of the sample
  are applied.
• The impact of these differences on the
  comparability of the results is not assessed.

13
(No Transcript)
14
(No Transcript)
15
Rotation Patterns in the EU-LFS
16
Rotation Patterns in the EU-LFS
17
(No Transcript)
18
(No Transcript)
19
Conclusions

• Rotating sampling designs permit the estimation
  of gross changes and enable reliable estimates of
  net changes in levels at different point in time.
• 17 EU Member States apply eight different sample
  rotation patterns for the LFS.
• The impact of these differences on the
  comparability of the results is not assessed.
• There is more emphasis on the measurement of
  change over successive quarters and less emphasis
  on the measurement of change between quarters in
  two successive years.