Reference metadata describe statistical concepts and methodologies used for the collection and generation of data. They provide information on data quality and, since they are strongly content-oriented, assist users in interpreting the data. Reference metadata, unlike structural metadata, can be decoupled from the data.
The aim of this Survey is to provide data on the amount and structure of labour costs borne by employers on the basis of labour force engagement. The results of the survey provide employers and economic policy makers with an insight into the state and structure of labour costs in the domestic labour market and their comparison with the international labour market, and also show the impact of legal regulations on the amount and structure of labour costs.
The survey collects data on the:
number of full-time and part-time employees;
number of hours actually worked;
number of paid hours;
labour costs.
2.2. Classification system
The following classifications have been used in this survey:
Classification of Activities (2010) - CA (2010), based on the Law on the Classification of Activities (“Official Gazette of the RS”, number 104/09), prescribed by the Government by the regulation of 29 July 2009 (“Official Gazette of the RS”, number 54/10). This classification is comparable with the EU classification of activities (NACE, rev. 2);
Nomenclature of Statistical Territorial Units - NSTJ, based on the Regulation of the Nomenclature of Statistical Territorial Units (“Official Gazette of the RS”, number 109/09 and 46/10).This classification is comparable with the Nomenclature of Territorial Units for Statistics (NUTS).
The data are representative and are processed on the level of CA (2010) sections and divisions; for the territory of the Republic of Serbia and statistical territorial units of Srbija – sever and Srbija – jug (NSTJ 1); and size classes of enterprises by the number of employees (10-49, 50-249, 500-999 and 1000 and more employees).
2.3. Coverage - sector
The Labour Cost Survey in the Republic of Serbia covers all sections of activities of the Classification of Activities (CA 2010) in which enterprises can be registered, (А-S).
Eurostat regulations stipulate that this survey does not cover section A – Agriculture, forestry and fishing.
In order to have data comparable with those of other European countries, separately are presented aggregated data that cover all the sections of activities and aggregated data without section A – Agriculture, forestry and fishing.
Five size categories are distinguished: 10-49, 50-249, 250-499, 500-999 and 1000 and more employees.
2.4. Statistical concepts and definitions
Labour costs refer to the total expenditure borne by employers for the purpose of employing staff. They include employee compensation, which is mainly comprised of gross wages and salaries in cash and in kind and employers' social security contributions, vocational training costs, other expenditure, such as recruitment costs and spending on working clothes, and employment taxes regarded as labour costs minus subsidies received.
Note:Methodological solutions that are fully harmonised with European recommendations (European Statistical Office) are used in this survey, thus making data of the Republic of Serbia comparable with other countries. Used are the definitions and recommendations published in the following regulations: Council Regulation (EC) No 530/1999 of 9 March 1999; Commission Regulation (EC) No 1726/1999 of 27 July 1999. Commission Regulation (EC) No 1737/2005 of 21 October 2005).
2.5. Statistical unit
The reporting units are all active enterprises (legal entities and entrepreneurs) on the territory of the Republic of Serbia that have on average 10 or more employees in the reference year, and are selected by random sampling. The reporting unit transmitted data for the whole enterprise, i.e. in aggregate form for all the organisational parts (local units).
Observation unit is reduced to the level of the local kind of activity, using the appropriate statistical methods.
Based on the data on the number of employees from the SBR, the structures (coefficients) were calculated, which were used to allocate the total labor costs, the number of employees and working hours according to the place of work and field of activity.
Data processing provided data for the following territorial levels: Srbija – sever and Srbija – jug, as well as for CA (2010) divisions. The collected data for enterprises having parts on both mentioned territories or in more activity divisions are broken-down.
2.6. Statistical population
All active business entities (legal entities and entrepreneurs) in the territory of the Republic of Serbia, which have on average 10 or more employees in the reference year, and are registered in the sections of activity A to S (CA 2010).
2.7. Reference area
Republic of Serbia.
Since 1999, the Statistical Office of the Republic of Serbia has no available data for AP Kosovo and Metohia, therefore they are not included in the data for the Republic of Serbia (total).
2.8. Coverage - Time
Year 2012, 2016 and 2020.
2.9. Base period
Not applicable.
3.1. Source data
The Labour Cost Survey 2020 is sample based survey.
The sampling frame is constructed using Statistical Business Register (state March 2021). It contains 23 720 units (active enterprises) with 10 or more employees, which main activity is from the section A-S of NACE Rev. 2 (2010) classification.
A stratified simple random sample of a size of 3807 units has been selected. The stratification was according to the size (five classes according to the number of employees), NACE Rev. 2 divisions and according to the territorial levels NSTJ 1.
The Horvitz-Thompson estimator of the totals and standard errors has been calculated by standard procedure for a stratified simple random sample. the final weights were adjusted for non-responses.
3.2. Frequency of data collection
Four-year frequency.
3.3. Data collection
The survey is based on reporting method by means of the questionnaire LCS, which is to be filled in by every enterprise in the sample. The questionnaire is to be filled in for the enterprise as a whole, i.e. as total for all organisational parts, whether located on the same or another address.
Reporting units are offered to choose whether to submit the information by post on a printed questionnaire or electronically via a web-questionnaire.
3.4. Data validation
Integrated Survey Tool (IST) is used for logical control and corrections of errors. IST is a software package that enables the process of data entry, logical control and production of output tables.
In order to ensure the quality of statistical data, the control is performed during the data entry itself. In applications (web applications and IST applications) for data entry, a logical control was incorporate, which would draw attention to possible errors and inconsistencies when entering data. The following steps in data validation include contingent control, and data analysis to identify potential errors, understand the logic and validity of data, as well as to identify extreme values. Checks are performed by automated procedures and expert data analysis Data are compared with data from the previous period and with available data from other sources.
3.5. Data compilation
The processing steps can be summarized as follows.
During and after data entry, errors and inconsistencies are detected and corrected. Item non-responses are aanalyzed and processed, imputation is performed. The resulting set of clean and complete micro data is weighted. During this phase, some other errors can be noticed, both in the material, and the sampling frame. After the correction, the data are again weighted as needed. After that, a re-analysis is performed, the aggregate data are then compared with related data from other sources.
Before publishing the results, extensive data quality control is carried out. Compliance with other statistical domains is satisfactory. Comparability over time is good. The release of the results is in accordance with the publication calendar and the data is readily available to all users.
5.1. Relevance - User Needs
The main beneficiaries of data obtained by the Labour Cost Survey can be recognized are the Ministry of Labour and Social Affairs, the Ministry of Finance, the Ministry of Economy, the Chamber of Commerce and Industry, Trade unions and Unions of Employers, then the Chamber of of Commerce and foreign employers (investors), also Eurostat to which the data are regularly transmitted, the European Central Bank, the International Monetary Fund and the International Labour Organization.
Methodological solutions that are fully harmonised with European recommendations (European Statistical Office) are used in this survey, thus making data of the Republic of Serbia comparable with other countries. Used are the definitions and recommendations published in the following regulations: Council Regulation (EC) No. 530/1999 of 9 March 1999; Commission Regulation (EC) No 1726/1999 of 27 July 1999, Commission Regulation (EC) No. 1737/2005 of 21 October 2005).The names and definitions of Labour Cost Survey variables are determined by Commission Regulation (EC) No.1737 / 2005of 21 October 2005.
SORS collected and provided all mandatory variables except the variables related to the apprentices, variable D1112.
Missing mandatory variables from LCS 2012:
А13 – Apprentices,
A131 – Apprentices converted into full-time units,
B13 – Hours actually worked by apprentices,
C13 – Paid hours for apprentices,
D112 – Wages and salaries of apprentices,
D123 – Employers’ social contribution for apprentices,
D1112 – Payments to employees’ saving schemes.
The Labour Cost Survey in the Republic of Serbia covers all sections of activities of the Classification of Activities (CA 2010) in which enterprises can be registered. Eurostat regulations stipulate that this survey does not cover section A – Agriculture, forestry and fishing. In order to have data comparable with those of other European countries, presented are aggregated data that cover all the sections of activities and aggregated data without section A – Agriculture, forestry and fishing.
5.3.1. Data completeness - rate
Not available.
6.1. Accuracy - overall
Sampling frame for 2020 was constructed using Statistical Business Register (state March 2021). The sampling frame consisted of enterprises from SBR with the following characteristics:
Their main activity belonged to the sections A - S of NACE Rev. 2 classification of economic activities;
They had 10 and more employees;
Corresponding legal entities were not closed or in liquidation process according to data from SBR, the source for this data is Business Register Agency (BRA).
The frame consisted of 19848 enterprises (sampling units), both from public sector and from non-public sector.
Stratification of frame: units were grouped into size classes according to the number of employees, NACE Rev. 2 divisions and NUTS level 1.
According to the number of employees enterprises were classified into five size classes:
10-49 employees,
50-249 employees,
250-499 employees,
500-999 employees,
1000 and more employees.
According to NACE Rev. 2 divisions enterprises were stratified into 84 classes (01–03, 05–33, 35–38, 41–43, 45–47, 49–53, 55–56, 58-66, 68–75, 77–82, 84-88, 90–96).
According to NUTS level 1, enterprises were stratified into Serbia - North (RS1) and Serbia - South (RS2).
Final stratification is defined by cross classifying size classes, divisions and two territory classes. Strata with 250 and more enterprises were census strata, they were enumerated completely. Census strata were also strata with only one unit.
In all, there were 537 strata, 246 census and 291 non-census strata.
6.2. Sampling error
Sample allocation was obtained as a solution of a non-linear optimization problem under multiple constraints according to the adopted multivariate and multi-domain procedure (Bethel, 1989, ISTAT) [1]. Sample was allocated taking into account that expected sampling errors in terms of coefficient of variations (CV) for estimates of totals of auxiliary variables (in this case it was only average number of employees for the previous 12 months according the CROSO database) in every domain of interest should be less or equal than already fixed planned errors.
Allocated sample consisted of 3807 enterprises. There were 879 census units (enterprises with 250 and more employees and enterprises from strata with only one unit).
Sequential simple random sample was selected using permanent random numbers with Uniform distribution on the interval (0, 1). The starting point for sample selection was 0.75.
The estimates of totals and standard errors (Horvitz-Thompson estimates) were calculated in the standard way for stratified simple random sample. The sampling weights were adjusted for non-response. There were no strata without responding units. Strata with one responding unit, but with more than one unit in the planned sample, were not collapsed with similar strata, because they were from different territory classes (Serbia-North and Serbia-South).
Outliers were detected according to annual costs on the level of NACE Rev. 2 divisions. They were classified into one census stratum and the weights of the remaining units in the strata from which they were excluded were corrected.
Prior to weighting, the values of labour cost variables of enterprises with parts, with respect to NACE Rev. 2 divisions or to two NSTJ 1 territory classes, were broken down on parts (as described in section Statistical units).
In the annex "Planned domains for the final estimates of the results" are shown the maximum planned errors, i.e. coefficients of variations (CV) for sample allocation by domains.
In annex "Coefficient of variation", estimates of the key variables Annual labour costs and Hourly labour costs are presented, as well as their standard errors and coefficients of variations. These estimates are presented on the total level of Serbia, on the level of territory; NACE Rev. 2 sections and on the level of size classes.
________________________
[1] Bathel, J. Sample allocation in multivariate surveys. Survey methodology 15 (1989): 47-57.
Coverage errors most often occur as a consequence of the sampling frame imperfections. If the frame misses some of the scope units there is an under-coverage error. On the other hand, if the frame includes some out of scope units there is an over-coverage error. Over-coverage measures were calculated on the level of an enterprise.
Over-coverage rates, LCS 2020 (%)
Total
Closed down
Not operating (dormant)
In liquidation
In bankruptcy
Without employees
0.53
0.11
0.21
0.08
0.05
0.08
Non-response errors
For in-scope units, the response rate is the ratio of the number of respondents to the total number of in-scope units. In scope units include the respondents and non-respondents.
The overall response rate is 92.3%. The highest response rate is 100% in section D- Electricity, gas, steam and air conditioning supply and the lowest response rate is 81.6% in the section N - Administrative and support service activities.
In the case of item non-response, imputation of missing values, from other appropriate available sources or using deductive principles - estimation of the missing value based on the logical relations of other known variables, was made.
6.3.1. Coverage error
Sampling frame for the target population was constructed on the basis of the SBR that did not contain enterprises in scope of LCS in division ’39’ (NACE Rev. 2).
Table in annex ("Number of enterprises") presents the number of enterprises in the frame, in the sample and the sampling fraction for the total number of enterprises in the sample and for those with parts (local units, according to the two NSTJ 1 territories or NACE Rev. 2 activity divisions (see part - Statistical unit, annex "Procedures for data breakdown according to NSTJ 1 territory levels and NACE Rev. 2 divisions".
There is no reason to believe that under-coverage is the major source of error. Further evaluation has not been carried on.
Over-coverage measures were calculated on the level of an enterprise because auxiliary information from other surveys considering parts of the enterprises (see section Statistical units, Procedures for data breakdown according to NSTJ 1 territory levels and NACE Rev. 2 divisions) was already filtered from potential over-coverage. The number of over-coverage units in the sample was determined on the basis of the information on response:
As in the previous years for which the Research on labor costs was carried out, the most common errors refer to completed or unexecuted hours of work and payments for days not worked.
Table. Percentage of imputation and correction for variables: Hours worked and Payments for days not worked, LCS 2020.
2020
Code and name of variable
% imputation
% corection
B1 Hours actually worked
5,6
3,5
D1113 Payments for days not worked
3,6
6,6
When eliminating these and other errors that appeared, the following methods were combined: re-calling reporting units, comparisons and checks with other available sources (Earnings Survey, final accounts) or using deductive principle ̶ estimating the missing/wrong value based on the logical relationships of other known variables.
It is assumed that the methods used for correction do not adversely affect the accuracy.
6.3.3. Non response error
See below 6.3.3.1 and 6.3.3.2
6.3.3.1. Unit non-response - rate
The unit response rate is the ratio, expressed as a percentage, of the number of responses (respondents) to the total number of in-scope units.
The overall response rate is 92.3%. The highest response rate is 100% in section D- Electricity, gas, steam and air conditioning supply and the lowest response rate is 81.6% in the section N - Administrative and support service activities.
The entry of LCS 2020 data was organized at the regional statistical offices for the corresponding territory, and in the Statistics Department of the City Administration of the City of Belgrade for Belgrade region. Program/application for entering data and logical control was the same, regardless of where the data entry took place. Questionnaires were double-checked, first during the recording – one group of staff, and second time by the team responsible for the results of the survey.
The number of cases needed to be corrected was not so significant to have an impact on the accuracy of the final results.
Table in annex ("Percentage of cases corrected by LCS variables") includes percentages of cases that were corrected at some point in the head office. For data entered in regional offices, statistics on the first level of checking (regional level) is not available. Table in annex refers to cases corrected due to: heavy errors (errors that have to be corrected or questionnaire cannot be considered as valid) and light errors (errors that do not necessarily mean that questionnaire is not valid).
One error may mean that several variables need to be corrected and vice versa – for one inaccurate variable logical control can display several different errors since it is designed to check the relations between the variables. For that reason, the table below shows estimated percentages of cases corrected per each variable.
In case of non-response to a certain question, imputation of missing values from other sources is used. Data are retrieved or estimated from the Earnings Survey, Final Account or using deductive principles by estimating the missing value based on logical relationships of other known variables.
Procedures for data breakdown according to NSTJ 1 territory levels and CA (2010) divisions
To reduce the burden on respondents and survey costs, data were collected for the enterprise as a whole. In data processing, for enterprises with parts, related aggregated data were broken down according to the NSTJ 1 territory levels and CA (2010) divisions.
Based on the data on the number of employees from the SBR, the structures (coefficients) were calculated, which were used to allocate the total labor costs, the number of employees and working hours according to the place of work and field of activity.
The main principles and standards of the revisions of released statistical data are defined by the General Revision Policy of the Statistical Office of the Republic of Serbia.
6.6. Data revision - practice
Not applicable.
6.6.1. Data revision - average size
Not applicable.
The data are published in accordance with the release calendar.
Тhe data are submitted to Eurostat within the prescribed period, 18 months after the reference period.
7.1. Timeliness
The data are published in accordance with the release calendar.
Тhe data are submitted to Eurostat within the prescribed period, 18 months after the reference period.
7.1.1. Time lag - first result
Not applicable.
7.1.2. Time lag - final result
18 months after the reference period.
7.2. Punctuality
-
7.2.1. Punctuality - delivery and publication
Not applicable.
8.1. Comparability - geographical
There are no differences between national and European concepts regarding the population, reference time and classification. Although, the unit of sample selection was the enterprise, usage of appropriate statistical method of estimation had resulted in that observation unit was actually the L- KAU. Concerning the definitions of variables, there are no deviations from the European concepts.
8.1.1. Asymmetry for mirror flow statistics - coefficient
Not applicable.
8.2. Comparability - over time
There were no significant changes in methodology and definitions in 2020 compared to 2016 and 2012, except that data on other taxes (D4) - Taxes, employer costs related to fulfillment or non-fulfillment of the obligation of employing people with disabilities, were not collected in 2012.
8.2.1. Length of comparable time series
2.
8.3. Coherence - cross domain
The degree of coherence is satisfactory. The data are compared with:
- Structural Busines Btatistics (SBS)
- Labour Force Survey (LFS)
- The survey on earnings, as well as
- Labour Cost Index (LCI)
Differences in the data that occur are expected and are the result of different methodologies, coverage and definitions.
Data on compensation of employees (D1) from the national Accounts for 2020 have been estimated both at the level of the total economy and at the level of individual sectors of KD 2010. For now, the SBS publishes only estimates at the level of total economy. they are further treated as confidential and as such have not yet been disclosed.
Data on the number of employees in national accounts for 2020 have also been estimated both at the level of the overall economy and at the level of individual sectors of KD 2010. At the moment, the SBS treats all data on employment in national accounts as confidential and they are still not publish.
8.6. Coherence - internal
Data for variable D4 - Taxes, employer costs related to fulfillment оrnon-fulfillment of the obligation of employing people with disabilities, were not collected in 2012.
The aim of this Survey is to provide data on the amount and structure of labour costs borne by employers on the basis of labour force engagement. The results of the survey provide employers and economic policy makers with an insight into the state and structure of labour costs in the domestic labour market and their comparison with the international labour market, and also show the impact of legal regulations on the amount and structure of labour costs.
The survey collects data on the:
number of full-time and part-time employees;
number of hours actually worked;
number of paid hours;
labour costs.
Not Applicable
Labour costs refer to the total expenditure borne by employers for the purpose of employing staff. They include employee compensation, which is mainly comprised of gross wages and salaries in cash and in kind and employers' social security contributions, vocational training costs, other expenditure, such as recruitment costs and spending on working clothes, and employment taxes regarded as labour costs minus subsidies received.
Note:Methodological solutions that are fully harmonised with European recommendations (European Statistical Office) are used in this survey, thus making data of the Republic of Serbia comparable with other countries. Used are the definitions and recommendations published in the following regulations: Council Regulation (EC) No 530/1999 of 9 March 1999; Commission Regulation (EC) No 1726/1999 of 27 July 1999. Commission Regulation (EC) No 1737/2005 of 21 October 2005).
The reporting units are all active enterprises (legal entities and entrepreneurs) on the territory of the Republic of Serbia that have on average 10 or more employees in the reference year, and are selected by random sampling. The reporting unit transmitted data for the whole enterprise, i.e. in aggregate form for all the organisational parts (local units).
Observation unit is reduced to the level of the local kind of activity, using the appropriate statistical methods.
Based on the data on the number of employees from the SBR, the structures (coefficients) were calculated, which were used to allocate the total labor costs, the number of employees and working hours according to the place of work and field of activity.
Data processing provided data for the following territorial levels: Srbija – sever and Srbija – jug, as well as for CA (2010) divisions. The collected data for enterprises having parts on both mentioned territories or in more activity divisions are broken-down.
All active business entities (legal entities and entrepreneurs) in the territory of the Republic of Serbia, which have on average 10 or more employees in the reference year, and are registered in the sections of activity A to S (CA 2010).
Republic of Serbia.
Since 1999, the Statistical Office of the Republic of Serbia has no available data for AP Kosovo and Metohia, therefore they are not included in the data for the Republic of Serbia (total).
Not Applicable
Sampling frame for 2020 was constructed using Statistical Business Register (state March 2021). The sampling frame consisted of enterprises from SBR with the following characteristics:
Their main activity belonged to the sections A - S of NACE Rev. 2 classification of economic activities;
They had 10 and more employees;
Corresponding legal entities were not closed or in liquidation process according to data from SBR, the source for this data is Business Register Agency (BRA).
The frame consisted of 19848 enterprises (sampling units), both from public sector and from non-public sector.
Stratification of frame: units were grouped into size classes according to the number of employees, NACE Rev. 2 divisions and NUTS level 1.
According to the number of employees enterprises were classified into five size classes:
10-49 employees,
50-249 employees,
250-499 employees,
500-999 employees,
1000 and more employees.
According to NACE Rev. 2 divisions enterprises were stratified into 84 classes (01–03, 05–33, 35–38, 41–43, 45–47, 49–53, 55–56, 58-66, 68–75, 77–82, 84-88, 90–96).
According to NUTS level 1, enterprises were stratified into Serbia - North (RS1) and Serbia - South (RS2).
Final stratification is defined by cross classifying size classes, divisions and two territory classes. Strata with 250 and more enterprises were census strata, they were enumerated completely. Census strata were also strata with only one unit.
In all, there were 537 strata, 246 census and 291 non-census strata.
Not Applicable
The processing steps can be summarized as follows.
During and after data entry, errors and inconsistencies are detected and corrected. Item non-responses are aanalyzed and processed, imputation is performed. The resulting set of clean and complete micro data is weighted. During this phase, some other errors can be noticed, both in the material, and the sampling frame. After the correction, the data are again weighted as needed. After that, a re-analysis is performed, the aggregate data are then compared with related data from other sources.
The Labour Cost Survey 2020 is sample based survey.
The sampling frame is constructed using Statistical Business Register (state March 2021). It contains 23 720 units (active enterprises) with 10 or more employees, which main activity is from the section A-S of NACE Rev. 2 (2010) classification.
A stratified simple random sample of a size of 3807 units has been selected. The stratification was according to the size (five classes according to the number of employees), NACE Rev. 2 divisions and according to the territorial levels NSTJ 1.
The Horvitz-Thompson estimator of the totals and standard errors has been calculated by standard procedure for a stratified simple random sample. the final weights were adjusted for non-responses.
Not Applicable
The data are published in accordance with the release calendar.
Тhe data are submitted to Eurostat within the prescribed period, 18 months after the reference period.
There are no differences between national and European concepts regarding the population, reference time and classification. Although, the unit of sample selection was the enterprise, usage of appropriate statistical method of estimation had resulted in that observation unit was actually the L- KAU. Concerning the definitions of variables, there are no deviations from the European concepts.
There were no significant changes in methodology and definitions in 2020 compared to 2016 and 2012, except that data on other taxes (D4) - Taxes, employer costs related to fulfillment or non-fulfillment of the obligation of employing people with disabilities, were not collected in 2012.
