Statistical Office of the Republic of Serbia

Labour Market Statistics Department, Registered Employment, Earnings and Labour Costs Statistics Unit

5 Milana Rakića, Belgrade

The objective of the Structure of Earnings Survey is to provide insight into the structure of earnings, i.e. to obtain data about different monthly and annual wage components (earnings, annual bonuses, earnings related to overtime, special payments for shift work, paid hours, paid overtime hours, annual days of holiday leave, etc.) and on the average hourly earnings, according to the characteristics of the enterprise in which employees work (size of the enterprise, economic activity, type of ownership, etc.), as well as according to the individual characteristics of employees (sex, occupation, education, age, length of service in the enterprise, type of employment contract, etc.).

List of abbreviations:

SES - Structure of Earnings Survey

SBR - Statistical Business Register

CRCSI - Central Registry of Compulsory Social Insurance

SORS - Statistical Office of the Republic of Serbia

CA (2010) - Classification of Activities (2010)

NACE Rev. 2 – Statistical Classification of Economic Activities in the European Community

ISCO-08 – International Standard Classification of Occupations 2008

Classification of Activities (2010) which was prescribed on the basis of the Law on the Classification of Activities (”Official Gazette of the Republic of Serbia”, No 104/09) by the Government on 29 July 2009 (”Official Gazette of the Republic of Serbia”, No 54/10). This classification is comparable with the EU classification of economic activities (NACE, Rev. 2).

Classification of occupations - translation of the International Standard Classification of Occupations 2008 (ISCO-08), extended with certain professions from the Nomenclature of occupations (Serbian: JNZ), which is prescribed as a standard for managing data on occupations in the labour records ("Official Gazette", No 9/98).

Target population covered all enterprises (legal entities and entrepreneurs) in Serbia employing at least 10 employees, operating in any economic activity defined in CA (2010) sections B to S, were active in October 2018 and have submitted the financial statements (balance sheets) for the year 2017 or they were public funds users in 2018.

The survey refers to employees who were under an employment contract in October 2018, whether it was for a fixed or indefinite period of time, as well as employees who worked on the basis of the contract on performing temporary and occasional jobs (regardless of whether they worked full-time or part-time) and who received earnings for October 2018. The employees who were not included in the survey were the ones in enterprises employing fewer than 10 employees, as well as the employees in section A - Agriculture, forestry and fishing.

Data are available by size classes of enterprises (corresponding to the number of employees): 10 to 49 employees, 50 to 249, 250 to 499, 500 to 999, and 1 000 or more employees.

Employees are all persons who had a direct employment contract with the enterprise in October 2018 and received remuneration for the same reference month, whether the contract was for a fixed or indefinite period of time, as well as persons who worked under the contract on performing temporary and occasional jobs, regardless of whether they worked full-time or part-time.

The gross annual earnings include all payments to employees to which tax and contributions are to be paid. Annual earnings include payments for work with full, shorter than full or longer than full working hours (overtime), back-dated arrears or differences in earnings, allowances for on-call, night work and shift work, work on Sundays and public holidays, bonuses which are not paid each month, regular monthly bonuses, compensation for transport costs for going to and from work, meal allowances during work, as well as allowances for hours not worked (annual holiday leave, paid leave, public holidays, sick leave up to 30 days, leave for professional training, and operational delay that is not the fault of workers).

The earnings do not include remunerations: for sick leave longer than 30 days, for time spent on business trips in the country or abroad, for accommodation and meals during work and stay on the field, for costs of funeral services, damage due to injury at work or professional diseases, nor other benefits to which tax and contributions are not to be paid. Also, the earnings do not include wages earned for work abroad.

Bonuses (bonuses which are not paid each month) are the ones paid on the basis of periodical and final accounts, 13th-month payments, allowance for annual holiday leave, and other periodical payments which have the character of salary, as well as jubilee awards and severance payments to employees at the time of retirement or to employees who have been made redundant.

Annual days of holiday leave is the total number of annual holidays, expressed in days, to which an employee is entitled in the reference year, regardless of whether the days were used or not.

Gross monthly earnings are defined as total earnings paid for October 2018, regardless of the month in which they were paid. They cover remuneration in cash paid before any tax deductions and social security contributions payable by wage earners and retained by the employer. The earnings include bonuses paid in the reference month, earnings related to overtime, as well as special payments for shift work (including payments for night work, work on Sundays, or on public holidays).

Earnings related to overtime and special payments for shift work represent part of the total amount of earnings paid for October. Earnings related to overtime refer to the amount of overtime earnings paid for overtime hours. Special payments for shift work are premium payments during the reference month for shift work, night work, or weekend work where these are not treated as overtime. In accordance with Regulation (EC) 1738/2005, the earnings related to overtime and special payments for shift work are presented separately.

The number of paid hours is the total number of hours for which the employee was paid in the reference month (October 2018). Paid hours include hours actually worked, hours not worked, and paid overtime hours. Hours not worked but nevertheless paid are paid hours which the employee did not work due to annual holiday leave, public holidays, paid sick leave up to 30 days, professional training, etc.

The number of paid overtime hours is the number of paid overtime hours which the employee performed outside the contracted working hours, at the request of the employer.

Average gross hourly earnings are defined as the ratio of the monthly earnings paid to the employee for October 2018 and paid hours for the same month.

Note: All statistical concepts and definitions are harmonized with the requirements of Council Regulation 530/1999, the Commission Regulations 1916/2000 and 1738/2005, and the SES 2018: Implementing Arrangements.

Reporting units are active enterprises (legal entities and entrepreneurs) with 10 or more employees.

Units of observation (statistical units) are the employees in enterprises who received earnings for October 2018, regardless of the type of employment contract that they had with the employer: fixed or indefinite period of time or persons who worked under employment contract on performing temporary and occasional jobs, nor the contractual working time (full-time or part-time).

The target population consists of all economically active business entities (companies, enterprises, institutions, cooperatives, and other organizations as well as private entrepreneurs) who have submitted the financial statements for the year 2017 or they were public funds users in 2018, with 10 or more employees. The sampling frame was a list of active enterprises from the Statistical Business Register (SBR), state as in February 2019. The extracted enterprises operate in any economic activity in CA (2010) sections B to S and employed at least 10 employees in October 2018 according to the CRCSI database.

The Republic of Serbia

In Serbian SES data are available for 2014 and 2018.

For the SES 2018, the reference year is 2018 and the reference month is October.

Not applicable.

Data are collected on a random sample of enterprises and employees in these enterprises.

Sample design

The SES 2018 sample is a two-stage stratified random sample, with enterprises as the first stage units (PSUs) and employees as the second stage units (SSUs).

Sampling frame

The target population is comprised of all active enterprises in Serbia employing at least 10 employees operating in any economic activity defined in CA (2010) sections B to S that have submitted the financial statements for the year 2017 or they were public funds users in 2018. The sampling frame was a list of active enterprises from the SBR, state as in February 2019. The extracted enterprise's primary activity was in CA (2010) sections B to S and had 10 or more employees according to CRCSI. In the final sampling frame, 19.525 units were included.

The sampling frame with the first stage units (PSUs) was stratified according to five size classes and CA (2010) divisions.

Size classes were determined on the basis of the data on the number of employees (source: CRCSI). Enterprises were divided into five size classes: E10_E49 (from 10 to 49 employees), E50_249 (from 50 to 249 employees), E250_499 (from 250 to 499 employees), E500_999 (from 500 to 999 employees) and E1000 (with 1000 or more employees). Census (large) enterprises were those with 250 or more employees, in total there are 838 such units in the SES 2018 sample. In all, there are 310 strata; out which 152 is census.

Sample

Allocation of the sample was via optimal multivariate and multi-domain procedure Bethel with auxiliary variables: the total number of emplоyees (from CRCSI database: employees contracted on a fixed or indefinite period of time, as well as persons contracted performing temporary and occasional jobs, state of 31^st of October 2018) and average monthly wages (according to Tax Administration data from the Individual Tax Return Form on calculated taxes and contributions - PPP-PD form, state of 31^st of October 2018). The final sample size for the SES 2018 was 2.185 enterprises (PSUs).

Sample of enterprises was selected using SORS system of coordination that is based on the unique permanent random number (as described in the document Sample coordination of statistical business surveys (http://publikacije.stat.gov.rs/G2014/PdfE/G201410089.pdf). The sequential simple random sample was selected using permanent random numbers, the starting point for selection was 0.75.

Following instructions (on the questionnaire) that describe the procedure for selection of a systematic sample, responsible persons from enterprises were obliged to select the sample of employees (SSUs). The number of employees to be selected depending upon the size of the enterprise. Small enterprises (with 10 to 50 employees) needed to report data for all paid employees. Enterprises with 51 to 99 employees needed to report data for 50% of paid employees; enterprises with 100 to 249 employees reported data for 25% of paid employees; enterprises with 250 to 1499 employees reported data for 10% of paid employees; enterprises with at least 1500 employees reported data for 150 employees. If the enterprises reported data for the insufficient number of employees, the person reporting the data was warned.

Four-yearly.

The reporting method for data collection was applied in the survey, through the SES questionnaire. The questionnaires were submitted by reporting units (i.e. enterprises selected in the sample) through post or in electronic form (by e-mail or using a web questionnaire, which is available on the SORS website).

Data collected with the questionnaire were validated through a set of logical controls.

Data validation is conducted during the chain of activities:

-    control of collected questionnaires by statistical staff;

-    data with „hard“ mistakes could not be transferred via web application;

-    batch editing that is based on a report generated by an application that implements logical rules;

-    data editing during the analysis of data and during the estimation procedure;

-    data validation during the analysis of the aggregates.

The basic characteristics of the estimation procedure used in SES are:

         - Imputation method was used to treat non-response. This was done in case of non-response of large enterprises or enterprises specific by certain characteristics. Missing data are imputed using: data on employee records from the CRCSI database and from the Tax Administration database - data from the Individual Tax Return Form on calculated taxes and contributions (PPP-PD form).

         - Information on the reasons for non-response is used to correct the sampling weight for unit non-response. The realized sample consists of units that: filled out the questionnaire, submitted an incomplete questionnaire (did not submit data for the sufficient number of employees), units for which data were imputed, closed/not operating, in process of bankruptcy, have less than 10 employees and enterprises that have not paid wages for October 2018. Other units are classified in the unrealized sample. The unrealized sample also includes part of over-coverage units (extinguished, in the process of liquidation) in order to compensate, by number, newly created units that a sampling frame does not contain.

         - Units with outlier values for mean gross monthly earnings and average gross hourly earnings (eight primary sampling units) are put into special census strata. The weights of the remaining units in the strata from which they originated were corrected.

         - Horvitz-Thompson estimates, as for the two-stage stratified random sample, are calculated for totals and their standard errors. This means that estimates of totals were calculated by summing the weighted values from the sample. Final weights are the multiplication of the weight from the first stage and weight from the second stage. First stage weight is the weight for enterprises. The second stage weight corresponds to the employees.

Not applicable.

The SORS quality management system is relied on the Serbian official statistics mission and vision, as well as on the European Statistics Code of Practice – CoP and the Total Quality Management – TQM principles, which together make the common quality framework of the European Statistical System (ESS).

For more information, please see the documents on the SORS website.

Through all phases of data production, main quality aspects are controlled by statisticians, sample and IT experts.

At the national level:

In Serbia, the main beneficiaries are the Government of the Republic of Serbia, ministries, Chamber of Commerce and Industry, employer‘s associations, trade unions, media, scientific-research organizations, students, domestic and foreign companies, as well as other SORS divisions (for example statistics of national accounts).

At the international level:

Besides Eurostat, to which the data are regularly transmitted, the European Central Bank, the International Monetary Fund, and the International Labor Organization are the main and the most important users of Structure of Earnings Survey data.

User satisfaction is measured according to the volume of data downloaded from the official SORS website, as well as by the User Satisfaction Survey.

The results of the User Satisfaction Survey are available on SORS website.

All statistical concepts and definitions are fully harmonized with the requirements of Council Regulation (EC) No 530/1999 of 9 March 1999, Commission Regulation (EC) No 1738/2005 of 21 October 2005 amending Regulation (EC) No 1916/2000, and the SES 2018: Implementing Arrangements.

The Structure of Earnings Survey in the Republic of Serbia covers all enterprises with 10 or more employees, operating in any economic activity of CA (2010) sections B to S, except for section A – Agriculture, forestry and fishing.

Data for all mandatory variables and some optional variables were collected and transmitted to Eurostat via Edamis.

Data for the following optional variables were not collected in the SES 2018:

• A17 – Affiliation of the local unit to a group of enterprises,
• B29 – Citizenship and residence of the employee,
• B34 – Other annual days of paid absence,
• B412 – Annual payments in kind,
• B423 – Compulsory social contributions and taxes paid by the employer on behalf of the employee,
• B4231 – Compulsory social-security contributions,
• B4232 – Taxes.

Not available.

The survey is based on a random sample so that sampling error is present in the results. The main non-sampling errors are coverage, non-response, measurement, and processing errors.

The design and organization of the survey, as well as the data collection and processing methods, ensure sufficient overall accuracy. The sample design is such that the estimates for the main domains are of satisfactory precision.

The following non-sampling errors exist in this survey:

- coverage errors that are due to existing time lag in registering the changes of enterprise characteristics (economic activity of an enterprise, number of employees);

- errors that are a consequence of including in the sampling frame only enterprises with 10 or more employees;

- measurement and non-response errors are inevitable, but the process of data collection and data entry are controlled to reduce these errors.

The sampling design ensures results of satisfactory precision for the estimates of parameters by main estimation domains (sex, CA (2010) sections and divisions, size classes of the enterprises, age groups, major occupation groups, etc.).

The sampling weight is corrected for non-response and the presence of outliers. Imputation error, as a component of the estimated variance, is not calculated.

Coefficient of variation of main target variables Gross earnings for the reference month and Average gross hourly earnings in the reference month are calculated for different individual breakdowns. On the level of the Republic of Serbia estimates of the coefficient of variation for variables 4.2 (Gross earnings for the reference month) and 4.3 (Average gross hourly earnings in the reference month) are 0.71% and 0.69%, respectively.

For CA (2010) sections the coefficients of variation (CV) vary from 1% to 8%, most being less than 4%. For occupation level, most coefficients of variation are below 2%, for age groups the coefficients of variation vary from 1% to 3.55%. By size of the enterprise, the CV is the lowest for the E500_999 size class (0.4%) and the biggest for the E50_249 size class (2%).

For more detailed information on coefficients of variation for different individual breakdowns see section: 6.2.1. Sampling error – indicators. 

The response rate was high for both enterprises and employees, 91.8% and 95.2% respectively.

Please see the attached document Coefficients of variation in the Annex with tables on the estimated coefficients of variation for variables 4.2 (Gross earnings for the reference month) and 4.3 (Average gross hourly earnings in the reference month) by: full-time and part-time employees and sex; CA (2010) sections of activities; occupation; age groups and by size class of the enterprise.

Non-sampling errors can be defined as errors arising during all survey activities. The main non-sampling errors in the results of SES 2018 on earnings and employees are coverage, non-response, measurement, and processing errors.

Frame imperfection is the major cause of coverage errors. Auxiliary variables: economic activity, registration status, financial activity, and the number of employees (this variable includes employees in accordance with SES EU regulations) are of crucial importance for the definition of the survey sample design and sample selection of SES 2018. If these variable values are not updated, or with an error, then coverage errors may occur.

Under-coverage errors additionally can arise because the frame lists contain only the data on enterprises with 10 or more employees. Furthermore, only enterprises that have submitted the financial statements for the year 2017 or they were public funds users in 2018 are covered. It does not include enterprises that ceased working in the period November 2018 to January 2019. Although the magnitude of the under-coverage error is not computed it is not considered to be significant.

Over-coverage and non-response errors (or response rates) are calculated using information about the response of the unit.

The units of over-coverage, determined not to belong to the target universe for the survey, are those for which the response is: closed, extinguished; blocked, not working; outside the scope of the survey (does not perform the selected activity); bankruptcy; liquidation; has not paid wages for October 2018 and had less than 10 employees.

The over-coverage rate for the Republic of Serbia for the level of enterprise is 2.6% and for employees is 1.2%. The over-coverage rate is the highest among the enterprises with 10 to 49 employees, 5.3% and 3.5% respectively for enterprises and employees. The main reason for over-coverage is that the business had less than 10 employees.

Non-response units are in-scope units for which data were not collected. Non-response units are those for which the response is: runs a business, refuses to submit a report; not found or moved (undeliverable questionnaire) and other reasons (e.g. unknown reason of non-response).

The average non-response rate for the level of enterprise is 8.2% and for employees is 4.8%. The highest non-response rate is among enterprises with 10 to 49 employees, 16.1% and 14.7% respectively for enterprises and employees.

In order to reduce non-sampling errors the following measures are undertaken:

- collected data are entered in the PC application that represents the so-called "controlled entry", which checks whether the questionnaires are filled out completely and does not allow the entry data with „hard“ mistakes (i.e. mathematically and logically incorrect data),

- the next step is the release of the program for computation-logical control,

- written and e-mail reminders and telephone contacts with the reporting units are used to keep the non-responses to a minimum, as well as for additional clarifications in case of deviations,

- further steps in data editing are based on the analysis of statistical experts and result with correction of discovered mistakes,

- in order to treat and reduce non-response imputations for large (census) non-respondents are done based on administrative data, i.e. Central Registry of Compulsory Social Insurance data and Tax Administration data from the Individual Tax Return Form on calculated taxes and contributions,

- possible deviations in the data observed during the estimation by the sample methodologist are also checked subsequently.

The units of over-coverage, determined not to belong to the target universe for the survey, are those for which the response is: closed, extinguished; blocked, not working; outside the scope of the survey (does not perform the selected activity); bankruptcy; liquidation; has not paid wages for October 2018 and had less than 10 employees.

The over-coverage rate for the Republic of Serbia for the level of enterprise is 2.6% and for employees is 1.2%. The over-coverage rate is the highest among the enterprises in the CA (2010) section S – Other service activities and equals 12.7% for enterprises and 8.7% for employees.

The under-coverage error was not computed.

In total there was 2.6% of over-coverage among enterprises and 1.2% among employees. Higher over-coverage (5.3% and 3.5% respectively for enterprises and employees) in the size class of enterprises with 10 to 49 employees is because of the usual fluctuation of employees among the smallest enterprises with less than 15 employees.

Please see the attached document Over-coverage rate in the Annex with tables on over-coverage rate by size classes of enterprises and by CA (2010) sections.

Not applicable.

The reporting units were enabled by SORS to enter necessary data via a web questionnaire. Respondents were not allowed to send the questionnaire until they filled in all required fields and corrected all "hard" mistakes - mathematically and logically incorrect data, thus the measurement error is reduced to a minimum. Also, employees in the SORS regional departments entered data from the received printed questionnaires into a computer application, which has built-in logical control. In order to correct the observed mistakes, colleagues from the regional departments contacted the reporting units, and later the statistical experts in the SORS carried out additional controls.

The most common measurement errors in SES 2018 were recorded for the variables listed below.

Non-response units are in-scope units for which data were not collected. Non-response units are those for which the response is: runs a business, refuses to submit a report; not found or moved (undeliverable questionnaire) and other reasons (unknown reason of non-response).

The average non-response rate for the level of enterprise is 8.2% and for employees is 4.8%. The non-response rate is the highest among the enterprises in the CA (2010) section I – Accommodation and food service activities (22.2% for enterprises and 15.6% for employees) and section L – Real estate activities (21.6% for enterprises and 17.2% for employees).

The overall unit response rate was 91.8%. The response rate concerning employees was 95.2%.

In case of unit non-response of census units (enterprises with 250 or more employees) data are imputed. Imputation was done using data from administrative sources, i.e. CRCSI database and Tax Administration data from the Individual Tax Return Form on calculated taxes and contributions.

To treat non-response in the estimation process re-weighting was used, i.e. weights were corrected for non-response.

In order to reduce the non-response, the following measures are undertaken: the reporting units could transmit data via web questionnaire, written and e-mail reminders were sent and the reporting units were contacted by telephone.

In order to reduce non-sampling errors caused by human factors, the following measures are undertaken:

-    Collected data are entered into the PC application that represents the so-called "controlled entry". During this phase, the program does not allow the entry of data with „hard“ mistakes.

-    The next step is the release of the program for computation-logical control.

-    Further steps in data editing are based on the analysis of statistical experts and result with correction of discovered errors.

-    If necessary, reporting unit is contacted in order to get correct data.

-    Finally, data validation is realized while comparing data (monthly gross earnings) with administrative sources.

Please see the attached document Non-response rate in the Annex with tables on unit non-response rate by size classes of enterprises and by CA (2010) sections.

See Imputation - rate 6.3.4.1.

Processing errors were corrected during the editing process and process of data validation.

Non-response of census units, i.e. enterprises with 250 or more employees, was mostly due to refusals, which means that these units are still running business, but refused to submit a report. In order to reduce non-response to a minimum and because these units can have a big impact on the result of this survey, data were imputed. Imputation was done using data from administrative sources, i.e. CRCSI database and Tax Administration data from the Individual Tax Return Form on calculated taxes and contributions.

The overall unit imputation rate was 2.8%. Unit imputation rate concerning employees was 4.3%.

No models were used.

Not applicable.

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. http://www.stat.gov.rs/media/2332/general-revision-policy.docx

Not applicable.

Not applicable.

Microdata of the Structure of Earnings Survey, for the reference year 2018, were sent to Eurostat in October 2020, four months after the deadline defined by regulation, i.e. 22 months after the reference period. This delay occurred for various reasons. Main reasons are restrictions caused by coronavirus pandemic and lack of time and human resources.

SORS published a bulletin with the final SES results in December 2020. The bulletin “Structure of Earnings Survey, 2018” can be found (for now only in Serbian) on the SORS website via the following link:

https://publikacije.stat.gov.rs/G2020/Pdf/G20205664.pdf.

Not applicable.

Microdata of the SES 2018 were sent to Eurostat 22 months after the reference period.

The final results were made available to users in December 2020 (24 months from the end of the reference year; only in Serbian). https://publikacije.stat.gov.rs/G2020/Pdf/G20205664.pdf.

The data were published in accordance with the plan, in December 2020. The bulletin "Structure of Earnings Survey, 2018" can be found on the SORS website.

Not applicable.

The SES data were collected according to the European regulations regarding definitions of statistical units, definitions of variables, reference period and classifications used. Enterprises operating in CA (2010) section O – Public administration and defence; compulsory social security, are also covered by the survey, with an exception of CA (2010) class 84.22 - Defence activities, which is not included in the SES 2018 scope. Therefore, data from this survey are comparable with the EU countries.

Not applicable.

For the second time, the SORS conducted the Structure of Earnings Survey, SES 2018. It was first conducted in 2015 as a pilot survey for the reference year 2014. Compared to the SES 2014, there were no significant changes in the definitions, classifications and sample design. The main difference is in the scope of the survey, where, in contrast to SES 2014, SES 2018 also includes CA (2010) section O - Public administration and defence; compulsory social security.

Two time periods, SES 2014 and SES 2018.

not available

not available

Not available.

SES data for enterprises operating in CA (2010) section O - Public administration and defence; compulsory social security, were not collected for the reference year 2014.

Not available.

The bulletin "Structure of Earnings Survey, 2018", for now only in Serbian, was published on the SORS website.

Data from Pilot Survey on the Structure of Earnings for 2014 was published within special publication on the SORS website.

SES data can be found in the Eurostat on-line database via the following link: https://ec.europa.eu/eurostat/web/labour-market/earnings/database

Not available.

Anonymized microdata are accessible at user’s demand (scientific–research organizations, researchers and PhD students, as well as institutions conducting separate project financed by national or international research programs) in accordance with principles defined in Official Statistics Law and rules defined by “Rulebook on the method of using and providing the data produced by the Statistical Office of the Republic of Serbia”. The request can be submitted at the e-mail address: stat@stat.gov.rs, or sent to the postal address of SORS.

not available

Methodological explanation can be found as an integral part of the publication.

Complete methodological documentation can be found on the SORS website, in reference metadata.  https://www.stat.gov.rs/en-US/istrazivanja/referentni-metapodaci

not available

not available

not available

not available

The confidentiality policy is defined in the following documents:

- Official Statistics Law;

- Rulebook on statistical data protection in the Statistical Office of the Republic of Serbia;

- Guidelines on measures of data and information protection in the Statistical Office of the Republic of Serbia and

- Regulation (EC) No 223/2009 of the European Parliament and of the Council, of 11 March 2009, on the transmission of data subject to statistical confidentiality to the Statistical Office of the European Communities.

Confidential data are treated according to the confidentiality related policy rules defined in Official Statistics Law.

none