1. LAMAS Working Group 13 - 14 June 2018
Agenda Item 3.3 Variance estimation in the LFS
Eurostat

2. Background
Variance of annual net changes, values of the indicators and differences requested by DG EMPL
List extended to 23 indicators this year, including two Yearly indicators
Eurostat in charge of variance estimation for 21 countries
Dissemination only on statistical significance (only Yes/No)
Eurostat

3. Variance of annual net changes
Comparison of NSI and Eurostat results for the 23 indicators :
Same conclusion for all indicators: ES, DE
Different conclusion: IT (3 indicators), SE and AT (4 indicators), LV and HU (1 indicator), FR (2 indicators)
NO comparison not possible
Eurostat compared also NSI/Eurobase indicators
BE and IE break in series, no results this year
The most problematics variables for comparison are Youth indicators
Youth employment rate (different conclusions for Italy) Youth unemployment ratio (different for Sweden and France)
Newly employed different for Italy and Austria
NO because Eurostat has not yet received their new COEFF -> new estimation method with a new coeff.
Eurostat compared also NSI/Eurobase indicators: it shows some issues -> usefull check
Eurostat

4. Net change significance of the NEET rate, age group 18-24, in %
2016
2017
Difference
Significance
Method
Estimated by BE
13.08
12.10 -
(1) BG
22.34
18.61
-3.73
Yes
Taylor
Eurostat CZ
9.14
8.32
-0.82 No DK
7.67
9.21
1.54 DE
9.23
8.58
-0.65
Bootstrap
NSI EE
11.40
12.35
0.95 IE
14.79
13.22 EL
21.89
21.43
-0.46 ES
19.09
17.09
-2.00
Jackknife FR
15.60
14.97
-0.63 HR
21.65
20.16
-1.49 IT
26.04
25.74
-0.30 CY
22.35
22.73
0.39 LV
14.93
14.16
-0.77 LT
12.64
12.11
-0.53 LU
7.99
8.20
0.21
NEET Part 1
FR data corresponds to France metropolitan
Largest difference Bulgaria pp Significant follow by Romania pp Significant
Smallest evolution HU pp logically not significant
HR not significant even if it id pp (small group large variance)

5. Net change significance of the NEET rate, age group 18-24, in %
2016
2017
Difference
Significance
Method
Estimated by HU
14.16
14.11
-0.05 No
Bootstrap
NSI MT
8.28
8.53 -
(1)
Taylor
Eurostat NL
6.07
5.26
-0.82
Yes AT
9.36
8.08
-1.29 PL
14.15
12.83
-1.32 PT
14.93
13.03
-1.90 RO
21.88
19.31
-2.58 SI
10.47
8.04
-2.43 SK
15.49
15.28
-0.22 FI
13.74
12.88
-0.86 SE
8.70
8.30
-0.40
Greg Estimation - CLAN UK
13.99
13.16
-0.84 IS
4.47
4.12
-0.35 NO
7.37
6.29
-1.08
Linearization CH
8.34
7.76
-0.58
Variance cannot be estimated because of a break.
Due to resampling indicators obtained using Bootstrap can be different from Eurobase
Most of the evolutions are significant except for 10 countries (where differences are always lower than 1pp).

6. Net change significance of the Employment rate 20-64, total, in %
2016
2017
Difference
Significance
Method
Estimated by BE
67.68
68.51 -
(1) BG
67.67
71.34
3.67
Yes
Taylor
Eurostat CZ
76.73
78.55
1.82 DK
77.40
76.94
-0.46 No DE
78.62
79.21
0.59
Bootstrap
NSI EE
76.64
78.73
2.09 IE
71.40
72.99 EL
56.19
57.76
1.57 ES
63.87
65.55
1.68
Jackknife FR
70.42
71.05
0.63 HR
61.43
63.56
2.13 IT
61.56
62.34
0.78 CY
68.74
70.70
1.96 LV
73.23
74.76
1.54 LT
75.18
75.98
0.80 LU
70.67
71.53
0.86
(1) Variance cannot be estimated because of a break.
Significant except for 5 countries (above 31)
For these 5 countries differences lower than 1pp

7. Net change significance of the Employment rate 20-64, total, in %
2016
2017
Difference
Significance
Method
Estimated by HU
71.53
73.26
1.74
Yes
Bootstrap
NSI MT
69.64
71.40
1.76
Taylor
Eurostat NL
77.12
78.04
0.92 AT
74.81
75.40
0.60 PL
69.29
70.89
1.59 PT
70.63
73.38
2.75 RO
66.31
68.83
2.52 SI
70.05
73.41
3.36 SK
69.81
71.11
1.30 FI
73.37
74.20
0.83 SE
81.20
81.80
Greg Estimation - CLAN UK
77.54
78.19
0.65 IS
87.77
87.56
-0.21 No NO
78.59
78.27
-0.32
Linearization CH
82.02
82.13
0.11

8. Net change significance of the Employment rate of low skilled persons, age group 20-64, in %
2016
2017
Difference
Significance
Method
Estimated by BE
45.59
45.85 -
(1) BG
39.15
44.39
5.24
Yes
Taylor
Eurostat CZ
43.67
49.20
5.54 DK
61.80
60.23
-1.57 No DE
58.91
59.59
0.67
Bootstrap
NSI EE
65.73
3.92 IE
48.88
50.08 EL
48.07
49.18
1.10 ES
52.86
54.65
1.79
Jackknife FR
51.25
52.59
1.34 HR
37.91
34.71
-3.19 IT
50.32
51.01
0.69 CY
56.76
56.92
0.16 LV
54.80
56.71
-0.47 LT
42.33
44.06
1.73 LU
56.41
56.42
0.01
BG, CZ large evolution > 5 pp logically significant
LT 1.73 NOT significant even if >1 pp

9. Net change significance of the Employment rate of low skilled persons, age group 20-64, in %
2016
2017
Difference
Significance
Method
Estimated by HU
50.71
53.86
3.15
Yes
Bootstrap
NSI MT
57.15
57.89
0.74 No
Taylor
Eurostat NL
60.70
61.19
0.49 AT
53.92
53.91
-0.01 PL
39.53
40.79
1.26 PT
64.72
67.69
2.97 RO
52.26
54.68
2.42 SI
45.50
49.33
3.83 SK
35.95
37.28
1.33 FI
52.05
51.28
-0.77 SE
61.00
61.80
0.80
Greg Estimation - CLAN UK
62.09
63.58
1.50 IS
79.52
77.89
-1.63 NO
60.35
59.94
-0.41
Linearization CH
68.10
67.73
-0.37
Not significant for the majority of the countries (16)
PL, SK, IS not significant even if absolute difference >1 pp

10. LAMAS is invited to take note of
Quality checks done at Eurostat on the sampling design variables (continue every quarters)
Final results to be transmitted to DG EMPL
Results will be also shared with LAMAS delegates
No big changes this year on the variables
Usual checks longitudinal, nb. of PSU STRATA, missing value
FSU cannot be empty etc
Eurostat