C.SURVEY METHODS
1.The Sampling Method
The sampling, estimation, and investigation methods used in the Labour Force Surveys have changed since the survey first began in 1954. The methods described in this chapter are those that have been in place since 2012.
At of the beginning of 2012, the Labour Force Survey was also changed from a quarterly to a monthly format. Over the course of 2015, data were collected every month from about 20,400 persons aged 15 and over on an ongoing basis.
For explanations about the methods used in the Labour Force Surveys until 2011 (inclusive) see the publication Labour Force Surveys 2011.[1]
The changes in the survey methods, definitions and questionnaire which were made in 1954–2003 were gathered into a special technical publication.[2]
The sample drawn each year is allocated to 12 mutually exclusive and complementary groups, known as panels. The panels are introduced into the survey investigation in 12 consecutive months, where one panel a month is introduced in the calendar year following the year in which the sample is drawn.
Each panel is investigated eight times according to a pattern known as 4-8-4. In the rotational structure of this pattern, eight investigations are performed for each panel in the following way: the investigation in the introduction month and in the three following months (investigations 1–4), a break of eight months, and four more consecutive investigations in the following months (investigations 5–8). This pattern was chosen based on various uses of the survey estimates, because it reduces sampling errors of the estimates of changes between two consecutive months and two consecutive years.
Each year contains 27 exclusive panels from three different sampling years.
Each following two year period includes 15 shared panels and 12 exclusiveones.
Most of the final sampling units are dwellings, which are sampled in two stages. In the first stage, localities are sampled, with probability proportional to size. In the second stage, dwellings are sampled in the sampled localities, so that the final sampling probability of the dwelling is equivalent to the sampling fraction of that year.
The sample of localities in the survey is taken from a list of localities (the locality sampling frame), and the localities are placed in four main groups, by the size of the locality, as determined by the estimates of number of persons aged 15 and over.
In the first three groups, all localities are included with certainty every year:
a.The 40 largest localities in the country that are included in all 12 new panels of that year’s sample. This group includes and covers about 64% of the persons aged 15 and over in the population.
b.Other large localities which are smaller than those in Group a above are included in three, four, six, and nine panels in the sample for each year. Every locality in this group is represented in at least one panel (but not in every panel) every month of the year. In that way, every month there is at least one interviewing portion of dwellings in each locality. Each year, about 58 localities of this type are sampled, covering about 18% of the persons aged 15 and over in the population.
c.Smaller localities included in only one or two panels a year. Thus, each locality is represented only in some months of the year. Each year about 30 localities of this type are sampled, covering about 4% of the persons aged 15 and over in the population.
d."Probability" localities – small urban localities and all rural ones – are sampled randomly from the list of localities, so that the probability of a locality being sampled is proportional to its size. Each year a sample of localities is drawn, which is dependent on the previous year’s sample. This ensures that many different localities are included in the samples of two adjacent years. Altogether, each year about 190 “probability” localities are sampled from about 1,000 localities in the sampling frame, covering about 14% of the persons aged 15 and over in the country. The sample in each locality is usually placed in one panel.
In every locality that is sampled, a sample of residential dwellings is taken, although sometimes (mostly in non-urban localities) households or other units are sampled.
A small portion of the current sample is derived through other methods, due to their suitability to the sample frame and other constraints, and is not derived through the main method described above. Examples are: samples of student dormitories and immigrant absorption centres, which are not covered in the sample frame of localities, and the sample of new building additions that represent the population entering dwellings whose construction was completed after the sampling process.[3]
Every month, over 9,000 households participate in the survey, so that about 105,000 interviews of households are conducted during the survey year. Notably, the vast majority of households are interviewed several times (up to four interviews) in a given year. In 2015, about 35,200 different households were interviewed, of which about 17,000 households were interviewed in four different months.
2.The Estimation Method
The data collected in the survey are subject to quality control and cleaning procedures (in the Data Editing and Processing System). The process includes about 250 different logical checks to identify logical errors in the collected data.
The data collected in the survey relate to the sample. To transform the data into estimates for the entire population, a “weighting coefficient” is allocated to each individual in the sample, and all of the data for the sample are multiplied by the weighting coefficient. Using the weighting coefficients, estimates are calculated for the monthly sample. The annual estimates presented in this publication are arithmetical averages of the 12 monthly estimates.
The estimation method is intended to reduce both sampling errors and bias that may occur because the characteristics of the households that participated in the survey may differ from those that did not participate.
In the monthly Labour Force Survey, a new estimation method was introduced. The new method, known as Regression Composite Estimation, was proposed by Fuller and Rao (2001). In the composite estimates, data on individuals and households surveyed in a given month are calibrated in terms of the total known population of cross-sections by geographic area, sex and age group, in the same way that calibration was done in the quarterly Labour Force Survey. However, in contrast to the quarterly Labour Force Survey, the composite method also calibrates data on labour force characteristics based on estimates from the previous month.
The method also ensures synchronization of the characteristics of individuals and the characteristics of the households they belong to through the use of an identical weighting coefficient for each person in the same household. The weighting coefficient reflects both the number of households and the number of persons in the population represented by the given household.
The set of weighting coefficients is determined by an iterative process that yields full compatibility between the weighted distributions of persons and the current demographic estimates of the Central Bureau of Statistics, based on the last Census and on the changes made after the Census, according to weighting groups.[4]
The division into weighting groups is independent of religion, by geographic groupings and by population characteristics (age*sex), as seen in Table 3/e, in the CBS publication LabourForce Survey 2014.[5]
In 2012, a number of changes were made in the definitions of the weighting groups, following changes in the distribution of the population.
( 1 )
[1]Central Bureau of Statistics (2012).Labour Force Surveys 2011. Publication No. 1504.Jerusalem: Author, Introduction to Chapter B.
[2]Central Bureau of Statistics (2006).Labour Force Survey: Changes in the Methodology, Definitions and Questionnaire 1954–2003. Technical Publication No. 78.Jerusalem: Author.
[3]For a detailed explanation on the sampling and on special samplings, seeCentral Bureau of Statistics (2017).Labour Force Survey 2014. Publication No. 1663.Jerusalem: Author, Introduction, "methods".
[4]For details, see Central Bureau of Statistics (2017). Labour Force Surveys 2014. Publication No. 1663.Jerusalem: Author, pp. (50)–(54).
[5]Central Bureau of Statistics (2017).Labour Force Survey 2014. Publication No. 1663.Jerusalem: Author, p. (54).