Agency for Statistics of Bosnia and Herzegovina

Labour Market Department

Zelenih beretki 26

71 000, Sarajevo

Bosnia and Herzegovina

The Bosnia and Herzegovina has been conducting Labour Cost Survey since 2012 (every four year), according to the Eurostat’s methodological recommendations and regulations.

For the survey is used KDBiH 2010 (BiH national classification) which corresponds to NACE Rev.2 Classification of Economic Activities, excluding: A Agriculture, forestry and fishing.

Data are collected for activities B-S.

In line with Commission Regulation (EC) No 1737/2005 of 21 October 2005 amending Regulation (EC) No 1726/1999

The units of observation in this survey are legal entities. Data were collected for legal entities as a whole, and not according to the corresponding local units.

All enterprises with 10 or more persons employed in the covered sectors.

Bosnia nad Herzegovina

The data refer to whole year of 2020.

No based period.

The data source for LCS are business entities (companies) with 10 or more employees.

Four-year frequency.

The enterprises received a paper version of the questionnaire and guidelines.

The questionnarie and instructions can be found in the attached document.

[Not requested]

[Not requested]

[Not requested]

Agency for statistics of Bosnia and Herzegovina follows the principles in The European Statistics Code of Practice (ES CoP) and The Quality Assurance Framework of the European Statistical System (ESS QAF) and other standards and quality management.

[Not requested]

Key users of the Labour Market statistics data are: Council of Ministers of B&H, Directorate for Economic Planning, Directorate for European Integration; Labour and Employment Agency of Bosnia and Herzegovina, IMF, researchers, media, employer’s associations, trade unions, etc.

User satisfaction was not measured for this survey.

[Not requested]

[Not requested]

The overall accuracy of the results is assessed in the following sections.

B&H Statistical Business Register (SBR) was used as a basis for construction of frame population. The survey covered all active enterprises from SBR which had  information about the number of persons employed from annual financial statements for the year 2020. For active units that do not have this data from the annual financial statements, the number of persons employed is taken from  the Unique system of registration, control and collection of contributions which is used as secondary administrative  source.  Dormant (temporarily inactive) and inactive units are excluded.

Enterprises with less than 10 employed persons, as well as enterprises from A (by NACE Rev 2 ) were excluded from the frame population.

There were some enterprises with local units existing in different entities, but data were collected according to the headquarter location of the corresponding enterprise.

Total frame  was stratified by entity, NACE Rev 2 division (2 digits) and employment classes (10 - 19, 20 - 49, 50 – 249 and 250+ persons employed). Employment classes were created on the basis of the original SBR classes.

Enterprises were allocated to strata using "the Neyman rule" on the variable number of employees (from balance sheets) with some modifications of practical nature. The Neyman rule states that sampling units should be allocated proportionally to the number of elements from the frame population by strata multiplied with standard deviation of number of employed persons from SBR by strata.

The Neyman rule was modified in following manner:

–        enterprises with 50+ persons employed were completely enumerated

–        if strata contained less than 5 enterprises, all of them were sampled

–        if the Neyman rule imposed less than 5 enterprises, then 5 enterprises were sampled.

On the basis of described allocation, stratified random sample was drawn from the total frame population. Total sample size was 4.292 enterprises for B&H.

The coefficients of variation (CV) for the key variables of the survey, Annual labour cost and Hourly labour cost[1] are given in annex. The coefficients of variation are produced by R software using survey package and are calculated at population level and breakdowns by NACE Rev 2 sections and the size band respectively.

[1]Annual labour cost = D1+D2+D3-D5,Hourly labour cost = (D1+D2+D3-D5)/B1

D1(compensation of employees), D2(vocational training costs), D3(other expenditure paid by the employer), D5(subsidies received by the employer), B1(total hours actually worked).

There is no item D4(tax) in the labour cost in Bosnia and Herzegovina.

Coverage errors (or frame errors) arise due to differences between the population that is covered by the framework and target populations. We distinguish three types of errors: insufficient coverage, excessive coverage and multiple listing (duplication).

The rate of over-coverage is the proportion of units (irrelevant) available within the framework that do not belong to the target population. Since we do not usually observe all the frame units, but only those selected in the sample, we need to estimate that proportion by the (un) relevant units in the sample.

Given that the sample selection framework is SBR (Statistical Business Register), with the status of 31.12. of the reference year, it happens that in the sample are selected business entities that on average have no more than 10 employees annually, which is an overcoverage, since only the business entities with 10 or more employees are covered by the Regulation.

Also, SBR data include the calculated employer's costs in the reference year, while the LCS cover business entities that have paid salaries and other labor costs in 2020.

All cases were treated as non-response although they should be treated as overcoverage, which directly affects the results of assessments.

Total frame population for B&H contained of 9 533 legal entities.

[Not requested]

The part of job referring to measurements and processing errors was performed in entities, but we don’t have application that would follow record on type of corrections made and on which variables it was made. So, it is only possible to asses measurements and processing errors.

Reasons for the occurrence of measurement errors may be:
- design of implementation instruments (questionnaires, manuals)
- non-response of data providers
- lack of records for certain statistics

The response rate can be defined in different ways depending for example on how over-coverage is treated. The table below contains information about the number of units that has responded, not responded and are over-coverage.

In annex, the response rates have been calculated in three different ways and are depending on how the over-coverage is handled in the sample.

No imputation has been done for LCS 2020.

[Not requested]

[Not requested]

Data referring to Federation of Bosnia and Herzegovina are:

Prior to entering data from completed questionnaires into the Blaise computer program, persons in charge of input and control of input data visually inspect all questionnaires. If some irregularities are detected in the data, which can not be corrected based on the other data from the form, a telephone contact with the business entity is established and thus the correct information on the basis of which the data is corrected in the questionnaire is established.

During the data entry itself to the electronic database (blase product), data is controlled for the second time. In the program for entering data from the form, controls are installed that prevent the calculation of logical and logical data in the database. The computer program warns us of wrong collections, missing values, irregular connections, disharmony, etc. These are hard and soft controls that either stop the entry, do not allow the entry of invalid data, or just send a message warning that the data is beyond the specified set limits. If necessary, and at this stage, the business entity is again contacted by phone, in order to correct the data.
This type of input ensures that after completion of the entry we have a database with calculating data.
The correction monitoring mechanism at this stage has not been established, so we can not provide information on the number of corrections during field work.
Additional controls were performed in the central statistical office (all controls and corrections were made in cooperation with data providers).

Measurement errors are possible at the data collection stage due to the lack of understanding of some items by the person who completes the form, as well as when the data itself is entered into the program itself.
Also frequent occurrence in the practice of this research is that people who fill up the ITR questionnaire do not read the instructions on how to fill in. It is mostly used practice that in some segments does not coincide with the methodology of this research.
The most common mistakes in the ITR 2020 occurred when filling out items about executed and non-executed hours of work, as well as fees for the same, since most companies did not keep a separate record of these items but were subsequently reckoned. In addition to this, it often happened that the item contributed to salaries, which were presented separately, doubled in the total gross salary, and the same amount was shown twice, which needed to be corrected.

No imputation has been done.

Small enterprises with less than 10 employees and sections A (Nace Rev.2) are optional, and have not been included and not accounted for.

No data is given for apprentices. The reason is that they are very rare in B&H. So rare, that it was not considered worthwhile to specifically ask about apprentices.

Variables D11 and D111 have the same values, due to the lack of inclusion of apprentices in LCS 2020.

The variable full-time equivalence for part-time employees (A121) was not included in the questionnaire for LCS 2020 in B&H. For calculating this variable this model has been used:

Model:  A121= C12 / (C11/A11)

A11 = Number of full-time employees

A121 = Number of full-time equivalence for part-time employees

C11 = Paid hours for full-time employees

C12 = Paid hours for part-time employees

C11/A11= Average paid hours for full-time employees

If model like this is used an assumption is made that part-time workers have the same working pattern as full-time workers. This is not true, but may be good enough.

The tax (variable D4) that Eurostat asks for is tax for employer for having employees. In B&H we do not have that kind of taxes, so we can not collect it.

For the same reason we also did not have in LCS 2020 next four variables:

1. D.1112            Payments to employees’ savings schemes
2. D.11144          Stock options and share purchase schemes
3. D.1222            Employers' imputed social contributions for pensions and health care
4. D.1224            Other imputed social contributions of the employer

and variables in connection with apprentices: A13, A131, B13, C13, D112, D123

[Not requested]

[Not requested]

[Not requested]

[Not requested]

First release of LCS 2020 for B&H were published in paper form and on our website in May 2022.

Tables A an B were sent to Eurostat in June 2022.

[Not requested]

[Not requested]

Data were published as planned.

[Not requested]

Comparison with results in the European Statistical System is possible, because the survey was carried out in accordance with the Regulations of the European Commission No. 530/199, 1726/1999 and 1737/2005.

[Not requested]

 The 2016 and 2020 Labour Cost Surveys were based on the same methodology.

[Not requested]

1. Labour Cost Survey vs. Labour Force Survey

The graph in annex ("Hours actually worked, expressed per employee, 2020 - NACE sections excl. A,T,U") shows a comparison of Hours actually worked[1]expressed per employee during 2020 in the LCS and the average actual hours worked in the main job per year of the LFS2020, also expressed per employee.

Differences between LFS and LCS are reflected in the following:

• LFS covers the whole labour market and LCS only enterprises with 10 and more employees.
• LFS collected data for sections A, T and U, and LCS did not cover them.
• LFS includes all (formal and informal) employment while LCS includes only formal, registered employment

The LFS contains information on hours worked for the reference week. Hours worked in the reference week multiplied by 52 weeks and thus get the average hours actually worked on the main job per year.

2. Labour Cost Survey vs. Structural Business Statistics

When comparing the LCS and the SBS one must know that there are few significant differences between the two statistics:

• Firstly, enterprises with less than 10 employees were excluded in the LCS while they were included in the SBS.
• Secondly, the public sector is not included in SBS, which can be good to keep in mind when analysing the data.
• In SBS sections K, P, Q and R are excluded. Also, section S does not cover division 94.
• SBS collect data for market activities in the non- financial sector.

The graph in annex ("Wages and salaries per employee, 2020") shows the Wages and salaries[2], expressed per employee from the LCS compared to SBS (Structural Business Statistics).

¹⁾ Code D11, divided by the value of code A1, in appendix 1 to Regulation(EC)No1726/1999

D11(wages and salaries), A1(number of employees)

²⁾ Code B1, divided by the value of code A1, in appendix 1 to Regulation(EC)No1726/1999.

B1(number of hours actually worked), A1(number of employees)

Labour Cost Index is still not produced in B&H.

National Accounts department of Agency for Statistics of Bosnia and Herzegovina don’t have data on compensations of employees, so it is impossible to compare Labour Cost Survey with National Accounts.

In LCS 2020 there are no major changes in the definitions and methods in comparison with LCS 2016.

Press release-website

Users of statistical data can easily and quickly access the data, which are published at the BHAS website: https://bhas.gov.ba/data/Publikacije/Saopstenja/2022/LAB_08_2020_Y1_1_BS.pdf

Not available

[Not requested]

[Not requested]

[Not requested]

Beside this quality report, the survey questionnaire and guidelines explain in detail survey methodology. As a part of the data, which are published on our web-site, we also have methodological explanations related to the survey.

[Not requested]

[Not requested]

[Not requested]

[Not requested]

[Not requested]

[Not requested]

-