K. Rutar - An example of longitudinal LFS weights

  • 22,495 views
Uploaded on

9th Workshop on Labour Force Survey Methodology 15-16 maggio

9th Workshop on Labour Force Survey Methodology 15-16 maggio

  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Be the first to comment
    Be the first to like this
No Downloads

Views

Total Views
22,495
On Slideshare
0
From Embeds
0
Number of Embeds
0

Actions

Shares
Downloads
0
Comments
0
Likes
0

Embeds 0

No embeds

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
    No notes for slide

Transcript

  • 1. An example of longitudinal LFS weights Katja.Rutar@gov.si Division for general methodology and standards
  • 2. LFS panel sample sizes (number of households) for the years 2012 and 2013
  • 3. Data collection • First wave – CAPI • Repeated waves – (predominantly) CATI • Nonresponding households from previous quarters are not included in the panel sample • Response rate: in first wave – 66%; in repeated waves – 83% • Average households size for responding households ~ 2,75 members
  • 4. Number of responding individuals
  • 5. Number of responding individuals – 15 years +
  • 6. Non-response distribution of the longitudinal sample – 3rd to 4th quarter 2013 2012Q3 / 2012Q4 Households % 2nd wave 3rd wave 5th wave Sample 3769 1641 1252 876 Ineligible 3 0,1% 0,1% 0,1% 0,0% Response 3135 83,2% 77,2% 85,0% 91,8% Nonresponse 631 16,8% 22,7% 14,9% 8,2% - Refusals 425 67,4% 65,6% 71,7% 65,3% - Noncontacts 84 13,3% 14,8% 11,2% 11,1% - Other 122 19,3% 19,6% 17,1% 23,6%
  • 7. Attrition by selected relevant characteristics of the household (completion rate)
  • 8. Weighting steps • Cross-sectional design weights (on strata and wave level) • Cross-sectional non-response weghts (on strata and wave level) • Longitudinal non-response wave (on strata level, to response statuses at the final quarter) • Grossing up to population total (wave level) • Calibration to demographic data (all data together, to first quarter situation) • Calibration to main employment statuses (all data together, to first quarter situation) + Calibration to demographic data (all data together, to first quarter situation) + Calibration to main employment statuses (all data together, to first quarter situation)
  • 9. Longitudinal population Q – Q (July 1st 2013 vs. Oct 1st 2013) 3th quarter 2013 3th and 4th quarter 2013 4th quarter 2013 Male Female Male Female Male Female 0 - 14 years 153.993 145.295 153.678 144.998 154.432 145.840 15 - 64 years 721.773 682.194 718.959 680.540 720.853 681.283 65 years + 143.892 211.967 142.177 209.948 145.170 213.085 Total 1.019.658 1.039.456 1.014.814 1.035.486 1.020.455 1.040.208 2.059.114 2.050.300 2.060.663
  • 10. Longitudinal population Y – Y (Oct 1st 2012 vs. Oct 1st 2013) 4th quarter 2012 4h quarter 2012 and 2013 4th quarter 2013 Male Female Male Female Male Female 0 - 14 years 152.786 144.173 151.971 143.441 154.432 145.840 15 - 64 years 724.926 685.520 715.330 679.943 720.853 681.283 65 years + 140.765 209.953 133.570 200.967 145.170 213.085 Total 1.018.477 1.039.646 1.000.871 1.024.351 1.020.455 1.040.208 2.058.123 2.025.222 2.060.663
  • 11. Basic descriptive statistics for the Q – Q longitudinal weight, compared to standard quarterly weight, 3rd quarter 2013 n mean stddev min q1 q3 max population W_LONG13q3 8.534 241,28 165,47 14,28 139,87 290,16 1.675,99 2.059.114 W_CROS13q3 15.486 132,97 90,25 10,79 76,44 159,86 1.100,41 2.059.114
  • 12. Changes in main employment statuses for the cohort, interviewed in 3rd and 4th quarter 2013 – using longitudinal weight 2013Q3 / 2013Q4 Unemloyed Employed Inactive Population under 15 in 1000 % in 1000 % in 1000 % in 1000 % Unemployed 48 50 22 23 26 27 Employed 13 1 865 94 45 5 Inactive 21 3 45 6 677 91 Population under 15 4 1 294 99
  • 13. Changes in main employment statuses for the cohort, interviewed in 3rd and 4th quarter 2013 – using longitudinal weight Employed 922 Inactive 741 Uneemployed 96 45 22 4513 26 21 82 vs. 97 752 vs. 753 932 vs. 910
  • 14. Relative standard errors for estimates of changes in main employment statuses for the cohort, interviewed in 3rd and 4th quarter 2013 – using longitudilan weight 2013Q3 / 2013Q4 Unemployed Employed Inactive Population under 15 CV % % % % Unemployed 6 12 10 Employed 15 1 9 Inactive 13 8 1 Population under 15 23 2
  • 15. Changes in main employment statuses for the cohort, interviewed in 3rd and 4th quarter 2013 – unweighted estimates 2013Q3 / 2013Q4 Unemloyed Employed Inactive Population under 15 % % % % Unemployed 51 2 1 28 Employed 1 9 3 5 Inactive 3 7 91 Population under 15 2 98
  • 16. Y - Y
  • 17. Basic descriptive statistics for the Y – Y longitudinal weight, compared to standard quarterly weight, 4rd quarter 2012 n mean stddev min q1 q3 max population W_LONG12q4 4.492 457,76 323,66 41,07 251,14 561,56 3.798,98 2.056.262 W_CROS12q4 14.407 142,73 98,09 15,65 81,96 170,59 1.225,36 2.056.262
  • 18. Changes in main employment statuses for the cohort, interviewed in 4th quarter 2012 and 2013 – using longitudinal weight 2012Q4 / 2013Q4 Unemloyed Employed Inactive Population under 15 in 1000 % in 1000 % in 1000 % in 1000 % Unemployed 38 39 35 36 24 25 Employed 25 3 841 91 56 6 Inactive 27 4 68 9 645 87 Population under 15 2 1 19 7 274 93
  • 19. Relative standard errors for estimates of changes in main employment statuses for the cohort, interviewed in 4th quarter 2012 and 2013 – using longitudilan weight 2012Q4 / 2013Q4 Unemployed Employed Inactive Population under 15 CV % % % % Unemployed 9 13 11 Employed 15 2 11 Inactive 14 9 2 Population under 15 88 16 2
  • 20. Changes in main employment statuses for the cohort, interviewed in 4th quarter 2012 and 2013 – unweighted estimates 2012Q4 / 2013Q4 Unemloyed Employed Inactive Population under 15 % % % % Unemployed 42 3 3 25 Employed 3 9 1 6 Inactive 3 9 88 Population under 15 7 93
  • 21. Conclusions & Open questions • There are many possibilities to calculate longitudinal weights. • At each separate stage of weighting process some assumptions have to be taken. • Should we give priority to consistency of employment statuses, nonresponse adjustment, demographic distribution, at which point of time?