Information
last updated
15.08.2012.
|
||||||||||||||||||||||||||||||||||||||
|
|
Methodology
|
To
the top of the document
|
||||||||||||||||||||||||||||||||||||
|
1995 – 2001 |
Labour Force Survey (LFS) was launched for the first time in November 1995 [1]. LFS was organised twice a year – in May and in November after that. A sample of dwellings was used for LFS. One household from the sampled dwelling and all corresponding household members aged 15 and more during reference period took part in survey in case of response. The sample was formed as rotating panel sample, where sampling units (dwellings) took part in the survey several times (three times in a row). The dwellings were replaced by other dwellings after the third time of interview, thus ensuring the rotation of dwellings in a panel. The target population of the LFS was divided in two subpopulations – the population of urban and the population of rural areas. Different sampling frames and sampling designs were used in each of these subpopulations. |
|||||||||||||||||||||||||||||||||||||
|
2002 – 2006 |
Work was started on the development of the new LFS methodology in
2000 [2].
The main aims for the new LFS methodology were:
A new sample design was developed. The list of territories covering all private households of Latvia was formed using the census counting areas of the Population and Housing Census 2000. These territories were used as primary sampling units by the new sampling design. The territories were stratified into four strata – Riga, cities under state jurisdiction, towns and rural territories. Two-stage sampling was used for LFS. Territories were selected as primary sampling units at the first stage. Sampling of the territories was done with stratified systematic πps sampling [4]. The second stage sampling units were dwellings. Dwellings were sampled with simple random sampling sample in each of the sampled territories. One household from the sampled dwelling and all corresponding household members aged 15-74 during reference period took part in survey in case of response. The main aim for using two-stage sampling was to reduce the survey costs. Design effect (which increased as the two-stage sample was used) was minimized with high number of primary sampling units in the sample. LFS was organised as rotating panel survey. The rotation scheme of the households was similar as before 2002. Dwellings took part in the survey three times with 26 week interval. Annual sample size of the LFS was 10 296 dwellings. |
|||||||||||||||||||||||||||||||||||||
|
Weighting |
Design weights were adjusted using response homogeneity group method [4]. Each primary sampling unit (territory) was used as response homogeneity group for dwellings. Post-stratification was used to achieve the final weights. Post-stratification was used once for quarterly estimates. Statistic about permanent residents of Latvia at the beginning of the reference year in breakdown by stratum, age group and gender was used as auxiliary information. Post-stratification was used in three steps for the annual estimates:
|
|||||||||||||||||||||||||||||||||||||
|
Estimation of variance |
Jackknife [4] was used for the estimation of variance. The method was implemented using the software SUDAAN. |
|||||||||||||||||||||||||||||||||||||
|
2007 – 2009 |
Sample size of the LFS was increased 2.4 times starting from 2007. This was done in order to achieve more precise estimates of number of employed and unemployed persons in breakdown by region, gender and age group. The rotation scheme was changed to ensure the overlap of the samples of successive quarters. Dwellings participated in the LFS four times with breaks 13 weeks, 39 weeks and 13 weeks. Such rotation scheme ensured the overlap of the samples of successive quarters. Annual sample size of the LFS was 24 128 dwellings. |
|||||||||||||||||||||||||||||||||||||
|
Weighting |
Design weights were adjusted using response homogeneity group method [4]. Each primary sampling unit (territory) was used as response homogeneity group for dwellings. Calibration was used to achieve the final weights. Statistic about permanent residents of Latvia at the beginning of the reference year in breakdown by stratum, age group and gender, statistic on resident population of Latvia at the beginning of the reference year in breakdown by region and statistic on the number of registered job seekers in breakdown by gender and age group was used as auxiliary information. Calibration was implemented using statistical computing environment R and package "sampling". |
|||||||||||||||||||||||||||||||||||||
|
2010 – 2011 |
The frame of territories used for the LFS sample was not updated during the time period 2002 – 2009. A study was made at the end of 2009. It was concluded that existing frame of territories was outdated. It did not characterise the target population of LFS well enough due to the population migration. A new frame of territories for sampling was developed at the end of 2009. The list was developed using the previous frame of territories. The decision was made to perform stepwise update of the LFS sample in 2010. The new sample of territories was sampled for the time period 2010 – 2014 [3]. Dwellings participating in the LFS for the first time in 2010 were selected from the new sample of territories. Dwellings participating in the LFS for the first time before 2010 were selected from the old sample of territories. Two samples of territories were used for LFS in 2010 as the result of sample update. |
|||||||||||||||||||||||||||||||||||||
|
Weighting |
Weights of LFS were calculated at household level (prior at person level) in 2010. All persons aged 15-74 within one household had the same weights. This approach ensured a higher compliance between the estimates of persons and households. Design weights for dwellings were calculated separately for each wave as dwellings were sampled using two samples. Design weights were adjusted using response homogeneity group method [4]. Strata and wave were used as response homogeneity groups. Raking-ratio was used as calibration method for weights (prior bounded regression). Statistic on permanent residents of Latvia at the beginning of the reference year in breakdown by stratum, age group and gender, statistic on permanent residents of Latvia at the beginning of the reference year in breakdown by region and age group and statistic on the number of registered job seekers in breakdown by gender and age group was used as auxiliary information. |
|||||||||||||||||||||||||||||||||||||
|
Additional information |
Total number of employees was estimated with the help of other methodology and other data sources in some sectors. Sectors where the number of employees was not estimated using LFS data are listed (classification according to NACE Rev 2):
|
|||||||||||||||||||||||||||||||||||||
|
2012 |
LFS sample was representative to residents of Latvia living in
private households in age group 15-74. The changes were made to the sampling
frame starting from the fourth quarter of 2010 with aim to make the LFS
sample representative to all population of residents of Latvia living in
private households. The period of five quarters (2010 quarter four – 2011 quarter four) was necessary to make the LFS sample representative to all population of residents of Latvia living in private households because LFS is a rotating panel survey. The LFS sample is representative to all population of residents of Latvia living in private households starting from the first quarter of 2012. |
|||||||||||||||||||||||||||||||||||||
|
Weighting |
It is possible to achieve full correspondence between the estimates of persons and households because the LFS sample is representative to all population of residents of Latvia living in private households. The weights for all persons from one household are equal. The weights of persons are equal to the corresponding household weight. |
|||||||||||||||||||||||||||||||||||||
|
Additional information |
In first quarter
of 2012 total number of employees in the sector of Public administration (O)
is estimated by other methodology and data sources then LFS. |
|||||||||||||||||||||||||||||||||||||
|
Data accuracy |
Standard error Estimates from the Labour Force Survey (LFS) are based on information collected from people in a sample of dwellings, rather than the entire population. Hence the estimates produced may differ from those that would have been produced if the entire population had been included in the survey. The most common measure of the likely difference (or 'sampling error') is the standard error (SE). SE serves also as the output value for the calculation of other quality indicators, for example, for the relative standard error or the coefficient of variation and confidence interval. Calculating the coefficient of variation it can be established whether the obtained result is sufficiently reliable. Example: if the estimation is 90.0 and the standard error is 6.14, the relative standard error or the coefficient of variation (CV) can be obtained as follows: CV=6.14/90*100=6.82. To calculate the confidence interval – the standard error must be multiplied by 1.961. Conf. Int. (+/-)=6.14*1.961=12.03 1 Coefficient used for the 95% confidence interval. Confidence interval When comparing estimates, it is important to use confidence intervals to determine if differences between values are statistically significant. Confidence intervals describe sampling variability and give an indication of the precision of a given estimate. Confidence intervals are calculated by 95% confidence, assuming that the true value lies within the set limits. Example: if the estimate is 90.0 and confidence interval (+/-) 12.03, the confidence limits are (77.97;102.03), with 95% confidence that the true value lies there. At the moment data accuracy indicators (standard error and confidence interval) are available for table NB.061 Employed persons by main kind of economic activity and quarter (NACE rev. 2). |
|||||||||||||||||||||||||||||||||||||
|
Reliability limits |
Quarterly data (thousands)
Annual data (thousands)
|
|||||||||||||||||||||||||||||||||||||
[1] Lapins, J. (1997). Sampling Surveys in Latvia: Current Situation, Problems and Future Development. Statistics in Transition, Vol. 3, No. 2, 281-292
[2] Lapins, J., Vaskis, E., Priede, Z. and Balina, S. (2002). Household Surveys in Latvia. Statistics in Transition, Vol. 5, No. 4, 617-641
[3] Liberts, M. (2010). The Redesign of Latvian Labour Force Survey, in Carlson, Nyquist and Villani (eds), Official Statistics – Methodology and Applications in Honour of Daniel Thorburn, pp. 193-203. Available at officialstatistics.wordpress.com
[4] Särndal, C.-E., Swensson, B. and Wretman, J. (1992). Model Assisted Survey Sampling. New-York: Springer-Verlag