Sampling method

Household Labor Force Survey is designed to produce estimations on annually, quarterly (3 months) and monthly basis over 3 months moving average by carrying out the survey at each month in the country.

Sample Size

Sample size of the survey is calculated in order to have annual estimations on Nuts1 x urban-rural and Nuts2 level. 

For the determination of the sample size, two studies were carried out:

In the first study, the initial selection probabilities, f0, were calculated in parallel with the year of 2004. The number of households was allocated to the Nuts2xurban-rural groups (52) proportionally. Then, in order to achieve the sufficient sample size in each group, the number of households in the urban groups were weighted by 1.5*f0 and in the rural groups by f0. By this weighting, some groups had still under or over sample sizes. These groups were reweighted by f0 or 2*f0. Hence the final sample sizes from the first study were obtained.    

In the second study, the requirement of Eurostat 577/98 regulation was taken into account. The instructions in this regulation were applied on the 2007 data set and the sample sizes in each strata were calculated independently. Following the regulation, firstly, %5 of the working age population was calculated and the corresponding groups belonging to this %5 of the working age population were determined.  The groups were chosen from age, gender and education level groups. Then the sample sizes for each strata (52) were calculated depending on both the %8 coefficient of variation criteria and the values of unemployment rate, design effect, overlapping factors between quarters and correlation coefficient values in each of the selected age, gender and education level groups.

 

The achieved sample sizes from the two studies were examined and the maximum ones in each stratum were chosen as the final sample size of the survey. Annual sample size of 2009 LFS was determined as approximately 168000 households. Accordingly, the quarterly sample size consists of approximately 42000 households. 

Annual sample design of LFS allows;

  • Producing quarterly estimations
  • Measuring variation between consecutive quarters
  • Cumulating quarterly estimations for annual estimations
  • Measuring variation between same quarters of the consecutive years
  • Monthly estimations over 3 month moving average approach.

The rotation pattern is applied by the use of 8 subsamples in each quarter. Each subsample constitutes 350 clusters. The addresses to be surveyed from each selected cluster are divided into two sets namely A and B. In each quarter, only one of these sets is included in the survey.  Hence the %50 overlapping ratio between consecutive quarters and same quarters between consecutive years are guaranteed. Number of addresses in each cluster is 15. This value was determined by taken into account the rate of homogeneity value.

Household Labour Force Survey Rotation Pattern

According to the scheme above, in the first quarter of 2009, 1 of the 8 subsamples comes from the new design (2009), while the others are from the previous design.  In this way, the transition from the old design to the new one is spread over time. In the year 2010, all subsamples come from the new design.

The address frame of 2009 survey is National Address Data Base (UAVT) which is updated regularly and linked with Address Based Population Register System (ADNKS). Each new subsample to be included in the survey is selected from the updated UAVT. Therefore listing study in the field is not needed which was the case of the survey before.

Sampling Methodology

First stage sampling unit: Blocks, which constitutes approximately 100 household addresses. While forming the sampling frame of the blocks, updated UAVT is used. Each of villages that don’t have municipalities is defined as one block. The blocks are selected by proportional to size.
Second stage sampling unit: Addresses. Number of 30 addresses is selected at once. The selection is done systematically then the selected addresses are divided into two sets (A and B). In each quarter, only one of these sets from the same block is included in the survey.

Stratification: Nuts2, urban-rural

Sampling Error Estimation

Sampling errors related to proportion and total estimates of the survey are calculated based on Taylor Series approximation using SAS module.

Weighting

Results of the household labour force survey have been weighted and published by the most recent population projections. Until year 2009 aforesaid population projections was calculated based on the general population censuses. In 2007 Address Based Population Registration System (ABPRS) was established and there exists some differences in the distribution of population by age, sex and regions in ABPRS compared to the censuses. In this direction, the new population projections were produced using the most recent population data obtained from this new system. In this direction, national and regional population projections were renewed by the results of 2008 ABPRS and also using the results of Turkey Population and Health Survey realized by the Hacettepe University Institute of Population Studies.
 
New population projections have started to be used in household labour force survey results starting from January 2009 onwards and in order to obtain comparability, monthly results for 2005-2008 terms and 2004 annual results were also revised by the new population projections.

2013 results given in this CD were also weighted by the revised population projections.

Reference and application period

The first week of each month starting with Monday and ending with Sunday has been used as the reference period.

The field application starts after the reference week and is completed within 15 days.

The method of collecting data

All the information has been collected by interviewers using Computer Assisted Personal Interviewing (CAPI) Method.

Top of page