Statistics Belgium

Business Statistics

North Gate
Koning Albert II-laan 16 

1000 Brussels

The Structure of Earnings Survey (SES) is a 4-yearly survey which provides EU-wide harmonised structural data on gross earnings and hours paid which are collected under Council Regulation (EC) No 530/1999 concerning structural statistics on earnings and on labour costs, and Commission Regulation (EC) No 1738/2005 amending Regulation (EC) No 1916/2000 as regards the definition and transmission of information on the structure of earnings.

The following classifications are used:

- NACE Rev. 2 for the economic activity;

- ISCO-08 for the occupation of the worker;

- ISCED for the highest successfully completed level of education and training;

- NUTS for the regional breakdown.

The Belgian SES refers to local units belonging to enterprises with at least 10 employees in the areas of economic activities defined by NACE Rev. 2 sections B to S excluding O.

Employees are all persons who have a direct employment contract with the enterprise or local unit and receive remuneration, irrespective of the type of work performed, the number of hours worked (full or part-time) and the duration of the contract (fixed or indefinite).

Mean annual gross earnings also cover all 'non-standard payments', i.e. payments not occurring in each pay period, such as: 13th or 14th month payments, holiday bonuses, quarterly or annual company bonuses and annual payments in kind.

Mean monthly gross earnings in the reference month cover remuneration in cash paid before any tax deductions and social security contributions payable by wage earners and retained by the employer, and are restricted to gross earnings which are paid in each pay period during the reference month.

Mean hourly gross earnings are defined as gross earnings in the reference month divided by the number of hours paid during the same period.

Number of hours paid includes all normal and overtime hours worked and remunerated by the employer during the reference month. Hours not worked but nevertheless paid are counted as 'paid hours' (e.g. for annual leave, public holidays, paid sick leave, paid vocational training, paid special leave, etc.).

The SES covers all activities defined in NACE Rev. 2 sections B to S (excluding O) for local units that belong to enterprises with at least 10 employees.

The SES refers to local units belonging to enterprises with at least 10 employees in the areas of economic activities defined by NACE Rev. 2 sections B to S excluding O.

The SES refers to local units belonging to enterprises with at least 10 employees in the areas of economic activities defined by NACE Rev. 2 sections B to S excluding O.

Reference year = 2018;

Reference month = October.

Not applicable.

The Belgian SES makes use of three different administrative sources:

• The national statistical register of enterprises (DBRIS)
• The earnings and working hours database of the National Office for Social Security (ONSS)
• The national register of individuals (RN)

A tailor-made questionnaire (NSI) is still necessary for obtaining the information that isn't available in existing registers. The table below gives an overview of the different Eurostat variables and the way Statistics Belgium obtained them:

The Belgian SES is organised on an annual basis.

The SES 2018 was conducted based on a two-stage sampling approach. In the first stage, a stratified random sample of local units is drawn. The stratification criteria are the economic activity, the number of employees and the region of the local unit. The sample fraction equalled to 11.9%.

As indicated at paragraph 3.1, the NSSO register contains individual information for every worker by local unit and enterprise. Worker, local unit and enterprise are identified by an unique key, making it possible to link the NSSO data with other sources. A linked dataset between the NSSO register and the RN data is used as basis for the second sample stage. Given the individual information available in this linked dataset, the total population is divided into strata. For this purpose Statbel makes use of the following three criteria:

• the monthly earnings of the employee. For employees working part-time, the earnings are extrapolated to a full-time position. The salary quartiles are then calculated per division of the economic activity; 

• the age of the employee. In order to limit the number of classes, Statistics Belgium opted for three aggregated age groups. Every group represents around a third of the total population. This results in the three following age groups:
• employees aged 37 or younger;
• employees between 38 and 49;
• employees aged 50 or older.

• the sex is the last criterion. There is still a pay gap between women and men, which means that the monthly earnings do not reflect the population correctly. 

The application of the three stratification criteria results per local unit in a theoretical maximum of 24 groups: 4 salary groups X 3 age groups X 2 codes for the sex. The strata obtained for each local unit are then arranged according to their frequency. The first employee is drawn in the most populated stratum, the second comes from the second most populated stratum, etc. In a large majority of local units, the sample size exceeds the number of available strata. Furthermore, the method proposed aims at an exhaustive coverage of all strata. The selection probability of an employee is considerably higher in sparsely populated strata.

Given that Statbel itself proceeds to the selection of the employees, it is very important that the questionnaire clearly states for which employees the enterprises have to provide the missing information. Within the current legal framework, Statbel can fill in the questionnaire in advance with the national register number, the surname, the name and the birth date of the employee selected. Thanks to this prefill, errors, for example when the employer provides by mistake the data for the wrong employee, can be limited to a minimum.

Data validation consists of global checks and plausibility checks. Global checks are necessary to ensure that complete data are available. Plausibility checks on all variables were done to ensure that the data are reasonable and consistent with other variables.

Not applicable

Not applicable

Before transmission to Eurostat, the results of the SES are checked internally by the validation team. Also Eurostat validates the results by using several global checks and plausibility checks. Finally the Quality Report contains all relevant information to enable the quality of the statistic to be evaluated.

SES microdata are checked for completeness and consistency.

The results and data of the Structure of Earnings Survey (SES) are often used by the Belgian general public. Students, research centres, universities, trade unions, the media, private companies and public administration institutions can all be considered as important users of the SES.

One reason for this interest from a wide range of users comes from the large survey sample. Because of these large numbers, the SES can give a correct image of the gross earnings by several independent variables. Together with the Labour Force Survey, the SES is for example the only national source where earnings can be linked with important personal characteristics such as the level of education or the occupation of the worker. The rather unique combination between individual features on the one hand and enterprise characteristics on the other hand can also explain the broad use of the SES.

Although almost all demands can be met by the existing information and variables, there is still much interest in more personal characteristics of the workers. Aspects such as nationality or household situation of the worker often influence the earnings and labour time. Also in the framework of the gender pay gap this information would be very useful.

Not applicable.

Not applicable.

The SES 2018 was conducted based on a two-stage sampling approach. In the first stage, a stratified random sample of local units is drawn. The stratification criteria are the economic activity, the number of employees and the region of the local unit. The sample fraction equalled to 11.9%.

As indicated at paragraph 3.1, the NSSO register contains individual information for every worker by local unit and enterprise. Worker, local unit and enterprise are identified by an unique key, making it possible to link the NSSO data with other sources. A linked dataset between the NSSO register and the RN data is used as basis for the second sample stage. Given the individual information available in this linked dataset, the total population is divided into strata. For this purpose Statbel makes use of the following three criteria:

• the monthly earnings of the employee. For employees working part-time, the earnings are extrapolated to a full-time position. The salary quartiles are then calculated per division of the economic activity; 

• the age of the employee. In order to limit the number of classes, Statistics Belgium opted for three aggregated age groups. Every group represents around a third of the total population. This results in the three following age groups:
  • employees aged 37 or younger;
  • employees between 38 and 49;
  • employees aged 50 or older.

• the sex is the last criterion. There is still a pay gap between women and men, which means that the monthly earnings do not reflect the population correctly. 

The application of the three stratification criteria results per local unit in a theoretical maximum of 24 groups: 4 salary groups X 3 age groups X 2 codes for the sex. The strata obtained for each local unit are then arranged according to their frequency. The first employee is drawn in the most populated stratum, the second comes from the second most populated stratum, etc. In a large majority of local units, the sample size exceeds the number of available strata. Furthermore, the method proposed aims at an exhaustive coverage of all strata. The selection probability of an employee is considerably higher in sparsely populated strata.

Given that Statbel itself proceeds to the selection of the employees, it is very important that the questionnaire clearly states for which employees the enterprises have to provide the missing information. Within the current legal framework, Statbel can fill in the questionnaire in advance with the national register number, the surname, the name and the birth date of the employee selected. Thanks to this prefill, errors, for example when the employer provides by mistake the data for the wrong employee, can be limited to a minimum.

See 6.2.1. for more information concerning the sampling errors.

Please see the attached document for detailed figures on the Coefficients of variation. For the monthly earnings the CV equals to 0.27 % and for the hourly earnings we obtain a figure of 0.22 %. 

In general the coefficients of variation are low, although for some smaller groups, like the nace section B and L or workers younger than 20 years, figures are above 1 %.  

See further for more details on the non-sampling errors.

For both the DBRIS (first stage) as the NSSO register (second stage), the files for the fourth quarter weren't available yet at the moment of the sampling. Therefore over- or undercoverage are possible. 

With regard to the size of the enterprise Belgium chose to exclude the small enterprises with fewer than ten employees. We also excluded the NACE section O in our sample. This means that the Belgian datasets only include the local units covered by the Regulation.

Not applicable

Not applicable

In order to detect outliers and other quality problems, several aggregated checks are integrated in the procedure. More complicated inconsistency problems were solved internally or by contacting the local unit or the administration responsable for the register on a bilateral basis.

The unit response rate for the SES 2018 amounts to 87 %. The table below gives an overview of the response rates by economic activity: 

See 6.3.3 for more information

See 6.3.3. for more information

See further for more details.

Data completeness is one of the advantages of an administrative register. Consequently, for those variables of which the information is derived from existing datasets, the item imputation rates equalled generally to 0 %.

For the variables included in the questionnaire, the item imputation rate is highest for the level of education (variable B25). For 4.6 % of the workers, this information was imputed. Enterprises often inform us about the difficulties to obtain this specific piece of information. In general, the level of education is less relevant for enterprises, since their workers can compensate a lower degree of education by experience or other skills. This information therefore often remains uncentralised. 

For the variables B26 (length of service in the enterprise) the impuation rate equalled to 1.2 % and for the occupation (B23) we had to impute in 0.9 % of the cases. 

No assumptions are made.

No seasonal adjustment takes place.

SES data aren't revised.

SES data aren't revised.

SES data aren't revised.

The two records with the thoroughly checked and internally approved data were sent to Eurostat on 25 June 2020. The data delivery took place before the deadline stipulated by the Regulation.

The two records with the thoroughly checked and internally approved data were sent to Eurostat on 25 June 2020. The data delivery took place before the deadline stipulated by the Regulation.

The two records with the thoroughly checked and internally approved data were sent to Eurostat on 25 June 2020. The data delivery took place before the deadline stipulated by the Regulation.

The field work started by sending the questionnaires on May 2019. Intensive contact attempts took place by the end of the summer for those respondents wherefore no answer was received. These contactes aimed at convincing respondents to participate and at increasing the unit response rate. For the most important local units phone contacts were replaced with a visit. An employee of Statistics Belgium could then help respondents fill in the questionnaire when this was necessary. The field work was closed in December 2019.

Data processing started as soon as the questionnaires were sent to the respondents. For the overall quality of the data, we aim to check the received data as soon as possible. This allows us to contact the respondent within an acceptable time span when errors or missing items are detected.

Besides the survey data, administrative sources were used to calculate the Eurostat variables. Even if these sources have several advantages, such as the correctness and the completeness of the data, they also have an important drawbacks: some of them were not received before spring 2020.

The final results of the SES 2018 were published in September 2020.

See general point 7.2.

Belgium has strictly followed the content of the Regulation. Geographical comparability should therefore be ensured.

A remark for education professionals is nevertheless in place. In Belgium the contractual labour time of these professionals differs substantially from the effectively performed labour time. This group of employees works several non-paid hours per week preparing classes and correcting student exams. This rather limited number of hours paid influences the amount of the hourly earnings, so variable B32 is used as the denominator there. The hourly earnings for this specific group of employees are therefore rather high and could differ substantially from the earnings available for other countries.

Not applicable

Two major methodological changes took place between SES 2014 and SES 2018: 

• More variables are based on administrative files, including the core variables on monthly and hourly earnings and the number of hours worked;
• From the SES 2018 on, Statistics Belgium also executes the second stage of the sampling, meaning that Statistics Belgium selects the workers wherefore the survey must be completed. 

Both changes result in a limited break in the series. 

See point 8.2.

The following table compares the survey’s variable “Gross annual earnings in the reference year” with the variable “wages and salaries” from the national accounts. Comparisons between these two sources should nevertheless be done carefully, because several differences concerning the target population and the followed methodology exist. The figures of the national accounts include for example businesses with fewer than ten employees, while these enterprises are excluded in the SES.

Given the above remarks, the figures are nevertheless comparable for the majority of economic activities:

Not applicable

See 8.3. for a comparison between National Accounts and the SES.

Not applicable

Every year, Statistics Belgium publishes a press release to explain and illustrate the main results of the SES. This communiqué is generally broadly covered by Belgian newspapers and news agencies.

In Belgium, the SES is organised on a yearly basis and is therefore well-known by the general public. Nevertheless, each year we use several channels to disseminate the SES results.

The results of the SES are also integrated in several publications. Some key results are included in the annual key figures published by Statistics Belgium. Also the annual Gender Pay Gap Report is worth mentioning. This elaborate publication describes the differences between men and women in the Belgian labour market. The SES is by far the most important source of this report.

The main channel to communicate the results is the dedicated section on the website of Statistics Belgium. This webpage contains the most popular aggregated tables, but visitors can also compose their own tables with our dynamic application. The earnings information is one of the most popular sections of the website.

See 9.3. for more information on this point.

Researchers can obtain microdata for their research project. On the website of Statistics Belgium, researchers can find more information on the procedure.  

The Belgian SES anonymised microdata are also accessible via CD-ROM by following the Eurostat procedure.

For users needing more detailed data, Statistics Belgium can produce tailor-made tables. 

The metadata are published on the website of Statistics Belgium.

The metadata are published on the website of Statistics Belgium.

100%

The metadata are published on the website of Statistics Belgium.

Statistics Belgium tries to reduce the burden for enterprises by using intensively administrative datasets. Yet a questionnaire is necessary to complete missing information. The cost for business to complete this questionnaire is measured yearly.

The Belgian confidentiality policy is followed. More information can be found on the website of Statistics Belgium.

The Belgian confidentiality policy is followed. More information can be found on the website of Statistics Belgium.

Not applicable