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.
North Gate Boulevard du Roi Albert II 16, 1000 Bruxelles
1.6. Contact email address
Confidential because of GDPR
1.7. Contact phone number
Confidential because of GDPR
1.8. Contact fax number
Confidential because of GDPR
2.1. Metadata last certified
6 June 2025
2.2. Metadata last posted
6 June 2025
2.3. Metadata last update
6 June 2025
3.1. Data description
Production index (productie-index/indice de la production).
This index is designed to be an indicator of the economic cycle and it is calculated based on hours worked.
3.2. Classification system
Nace Rev.2 classification (Statistical Classification of Economic Activities) is used for this production index.
3.3. Coverage - sector
The production indicator covers economic activities listed in sections F of Nace rev.2.
Size classes covered
from 5 to 9 persons : sampling rate=20%
from 10 to 19 persons : sampling rate=40%
20 persons and more = sampling rate=100%
3.4. Statistical concepts and definitions
The objective of the production index is to measure changes in the volume of output at close and monthly intervals. It provides a measure of the volume trend in value added over a given reference period. The production index is a theorical measure that must be approximated by practical measures.
For construction sectors, the production index is based on worked hours.
Planned changes in information collected : none
3.5. Statistical unit
Reporting unit : legal unit
Observation unit : legal unit
3.6. Statistical population
Legal units actives in section F of Nace rev.2 and having 5 employees or more in the Belgian business register (DBRIS).
There are 9000 legal units in the population in 2024.
3.7. Reference area
The variable covers hours worked in Belgium.
Only indices are published nationally.
3.8. Coverage - Time
Production in construction indices are available from 2000 for Eurostat. The series are published also from this period in Belgium.
3.9. Base period
Before March 2018, 2010 was the base and the reference year.
From March 2018 onwards, 2015 is the base and the reference year. Indices from year 2000 were re-referenced compared to 2015.
From March 2021 onwards, 2021 is the base and the reference year. Indices from 2021 to 2023 were reweighted. Indices from year 2000 were re-referenced compared to 2021.
These indicators are based on hours worked.
Absolute figures are not published nationally.
The calendar month is the reference period.
6.1. Institutional Mandate - legal acts and other agreements
The legal basis for this indicator is the European Business Statistics (EBS) Regulation:
Regulation (EU) 2019/2152 of the European Parliament and of the Council on European business statistics, repealing 10 legal acts in the field of business statistics (EBS-Regulation
Commission Implementing Regulation 2020/1197 laying down technical specifications and arrangements pursuant to Regulation (EU) 2019/2152 of the European Parliament and of the Council on European business statistics repealing 10 legal acts in the field of business statistic (General Implementing Act)
6.2. Institutional Mandate - data sharing
Data are sent to Eurostat, OECD, UN and IMF.
7.1. Confidentiality - policy
There is a national law concerning the confidentiality of data (statistical law of July 4, 1962): data from reporting units which are individual or which can result in individual data may not be disseminated without permission of the reporting unit. In practice this means indices are not disseminated for sectors where:
There are only one or two reporting units;
There are more than two reporting units but one of them represents 85% or more.
Confidential indices are sent to Eurostat flagged as confidential data.
7.2. Confidentiality - data treatment
Confidential data are treated by suppression.
8.1. Release calendar
Release dates are fixed in advance for a calendar year.
Simultaneous release to all interested parties : Data are transmitted to Eurostat when they are published in Belgium.
For certain reference months, the National Bank of Belgium has a prior access to the data for internal use only.
Identification of internal government access to data before release: The news are sent to the president of our ministry for information and to our general director 1 hour before publication.
Transmission to Eurostat and further use of the statistics:Data are transmitted to Eurostat when they are published in Belgium. The transmission is done by use of EDAMIS using SDMX format (since March 2018). The level of detail corresponds to the 2-digit level of NACE Rev. 2 and the totals.
The data are transmitted monthly to Eurostat and disseminated monthly to national users.
10.1. Dissemination format - News release
There are press releases on this indicator available in Dutch, French, English and German.
In these press releases, the seasonally adjusted production index for the reference month and for the total construction is compared to the same month of year before and the calendar adjusted index is compared to the month before.
News and web pages contain only the highest aggregates total construction.
10.2. Dissemination format - Publications
There are no other publications.
10.3. Dissemination format - online database
The press releases are a web document and are accompanied by an excel table containing :
Gross indexes
Working day adjusted indexes
Trends
Seasonally adjusted indexes
For Belgium, for Nace 2 digits, section F (2021=100).
Data (2010=100 and 2015=100) are not available online. These data are available on demand.
10.4. Dissemination format - microdata access
Microdata are not disseminated.
10.5. Dissemination format - other
Indices are transmitted to Eurostat to be used in calculating European aggregates and to be released as national data.
10.6. Documentation on methodology
There is no available methodology document.
10.7. Quality management - documentation
There is no available quality document.
11.1. Quality assurance
A great effort was made to optimise data collection and calculation methods. For each step of the production chain, the data are controlled. The stability of revisions are also tracked.
Once the figures produced, they are validated by our internal validation service before publication.
11.2. Quality management - assessment
Checks carried out along the production chain of the indices and these are finally validated by our validation service before diffusion.
12.1. Relevance - User Needs
Belgian National Bank, Eurostat, Federations, Professional associations, Universities, Regional public services, other users.
12.2. Relevance - User Satisfaction
No measure of user satisfaction but we receive feedback of users by mail, phone,....
12.3. Completeness
All of the required series are available.
13.1. Accuracy - overall
The sampling error is small.
The non-sampling error is small. This error is more important for less detailed aggregates.
The first estimates are biased.
The principal sources of errors are the non-response error.
13.2. Sampling error
Only one probability sample survey is used to obtain worked hours necessary for the calculation of the production index in construction.
The Belgian index is not adjusted with a productivity coefficient.
An estimate of the accuracy was made via the software POULPE for the production index (base=year-1) for the section F.
The program Poulpe (Programme optimal et universel pour la livraison de la precision des enquetes), based on SAS, has been developed by INSEE and it can incorporate sampling features such as stratification, clustering or multistage-sampling, while it can also approximate variances in case of ppS sampling.
13.3. Non-sampling error
The coverage errors are small.
The data collection and the access errors are small.
The main source of non-sampling error is the unit non-response.
The indexes are revised 3 times.
The survey used for the production index is compulsory but for the first round of calculation, the response rate is about 60-70% and estimations are made at the individual level based on the history of the responding unit or based on the sector evolution.
For the second and the third rounds of calculation, the response rate is about 75-85%. Indices are considered as stable 3 months and 15 days after the reference period.
The non-response is more important for small enterprises (5-9 persons) than for the larger enterprises (10 persons and more).
To increase the response rate, postal reminders are used and enterprises are also contacted by phone.
Response rate: between 90%-100%
Unit non-response: not available
14.1. Timeliness
The production index is released within 1 month and 15 days after the reference period. Then, he is revised twice : two months and 15 days after the reference period and 3 months and 15 days after the reference period (in principle, the third release is final).
Questionnaires are sent to the reporting unit around the 15th of the reference month asking for a reply by the 15th of the following month. Data collection is normally closes 4 days before publication.
14.2. Punctuality
No problem of punctuality.
15.1. Comparability - geographical
The definitions are equally applied to the 3 regions of Belgium.
15.2. Comparability - over time
In 2012, the construction survey was simplified.
Before 2012, the construction survey was based on a cut-off sample : the enterprises of 10 persons or more were surveyed.
After the first January 2012 : we use a rotating stratified design.
from 5 to 9 persons : sampling rate=20%
from 10 to 19 persons : sampling rate=40%
20 persons and more : sampling rate=100%
One third of the sample is replaced every year.
We did not observed a break in time series with the change of the year of reference.
15.3. Coherence - cross domain
No regular confrontation is made, but comparisons are made on an ad-hoc basis, especially when results seem contrary to expectations.
15.4. Coherence - internal
No problem of internal coherence.
Burden on respondents
The survey was simplified in 2012 and in 2014. Before 2012, the survey was exhaustive from 10 persons employed. After 2012, a sample was used. From 2014, only the variable ‘hours worked’ is asked in the survey.
Burden on respondents
year
time to complete the survey
hourly rate
number of respondent
periodicity
Total (€)
2015
0,11
41
2250
12
121770€
Cost associated with the collection and production of the construction production index
~6 persons for ~7600 hours/year
17.1. Data revision - policy
The same revision policy is applied to the data released nationally and transmitted to Eurostat.
Routine revisions : For the production in construction indexes, we recalculate the indices of the last 3 months. The indices are considered as stable 3 months and 15 days after the reference period.
Each month, revisions rates are calculated for the different aggregates. If this rate exceeds a determined threshold, (generally 5%) differences in the microdata are investigated.
Major revisions : If important changes are made in the production chain, we inform the users in the news tab on the page concerning the production indices in construction (Production in construction | Statbel)
Non-scheduled revisions may occur due to errors discovered in the input data after results are considered definitive. In this case, explanations are also given in the published news and to Eurostat.
17.2. Data revision - practice
For the production in construction indices, there is no revision calendar.
The same revision policy applied to data released nationally and transmitted to Eurostat to assure coherence.
The regular source of revision is the replacement of estimates by the actual values reported. For the production in construction indices, the routine revision rates are weak.
The revision rates are not published but are available on demand. In case of major revisions, users are notified and explanations are provided on our website.
Rates of difference between 2 versions of the identical reference month are calculated at Nace 2 digit. If these rates are greater than 5%, investigations are made to find the reasons of the difference.
All of the versions and data are stored in a database with a transaction date. The last version of the indices is stored with '31DEC9999' as transaction date.
Data used to calculate the indices are also stored with a transaction date. We be able to find the data used for the different calculations.
Benchmarking is not carried out with other statistics.
The change in the reference year was done on aggregated bruto indices. There was no significant revision in the raw series.
For the seasonal adjusted data, we have reused the same ARIMA models and the same specifications as for old series (2010=100).
Indicators for the original data series (YoY):
MR = -0.000442
MAR = 0.001872
MaxAR = 0.009642
RMAR = 0.040448
Indicators for the working day adjusted data series (YoY):
MR = 0.000006
MAR = 0.001310
MaxAR = 0.008361
RMAR = 0.029742
18.1. Source data
Type of source : survey
Frame on which the source is based : DBRIS (Belgian business register)
Sample or census : stratified sample survey by Nace 3 level and size class with a rotating design. One third of the sample is replaced every year.
Criteria for stratification : Nace 3 level + size class
Threshold values and percentages :
From 5 to 9 persons : sampling rate=20%
From 10 to 19 persons : sampling rate=40%
20 persons and more = sampling rate=100%
Frequency of updating the sample : annually
18.2. Frequency of data collection
Data are collected monthly.
18.3. Data collection
Questionnaires used in the survey :
Construction survey on monthly hours worked.
For new surveyed enterprises, total of hours of the year before the reference year is also asked.
Planned changes in national questionnaires : none
Data collection media : The data of these questionnaires are mostly collected by a web application and few of these are collected by postal questionnaires. The management of the questionnaires is carried out in our operating system : when a questionnaire is validated by the responding unit, itis granted atemporary status.An agent in our collect centermust then validatethe questionnairebefore having astatusok. The validation is often made based on the past values of the responding unit.
Planned changes in data collection methodology : none
18.4. Data validation
Checks are carried out at the time of data collection and throughout the production chain indices. These are also validated by another service before distribution.
Collected data are daily transferred from the operating system in a datawarehouse. An error reporting with several information (included months for the different survey, included estimations, ...) is sent to all users everyday.
Then, a SAS program allows to check the variation of the different variables for a statistical unit and allows to identify problems. If a problem is detected, the collect center is contacted. In function of the response of our collect center, the declaration is maintained, re-estimated or removed for the index calculation.
The gross indexes based on year-1 are calculated for a reference month. A SAS program allows then to calculate the number of standard errors between the reference month and the 12 months before by Nace 4 digits. If a problem is detected, the collect center is contacted. In function of the response and the remaining time before the publication, the declaration is corrected or removed for the calculation.
Final tables and press releases are validated by our validation service : format, content, missing information, layout of press release,...are verified.
If results appear abnormal, the validation service recontacts the responsible for verification.
18.5. Data compilation
Estimates for non-response : non-response is treated using imputation based on the past of the enterprise or based on the sector
Estimates for grossing-up to population levels : a weight based on the report (number of answering enterprises/number of enterprises in the universe) is attributed to every enterprise having answered in the stratim
Type of index : the production index in construction is a Laspeyres chain index
Method of weighting and chaining : the weights for aggregation are based on value added at factor costs of the year of reference 2021
Planned changes in production methods : none
18.6. Adjustment
Adjustments made for :
quality and productivity : hours worked should be corrected for changes in productivity but for the moment, it is not the case.
Seasonal and calendar adjustments
Agg
Trading day effect
Leap effect
Easter effect
Outliers
Seasonal adjustment decomposition
F
yes
no
yes
yes (5)
additive
F41
yes
no
yes
yes (5)
additive
F42
yes
no
yes
yes (5)
additive
F43
yes
no
yes
yes (6)
additive
More details about seasonal and calendar adjustments are in annex.
Production index (productie-index/indice de la production).
This index is designed to be an indicator of the economic cycle and it is calculated based on hours worked.
6 June 2025
The objective of the production index is to measure changes in the volume of output at close and monthly intervals. It provides a measure of the volume trend in value added over a given reference period. The production index is a theorical measure that must be approximated by practical measures.
For construction sectors, the production index is based on worked hours.
Planned changes in information collected : none
Reporting unit : legal unit
Observation unit : legal unit
Legal units actives in section F of Nace rev.2 and having 5 employees or more in the Belgian business register (DBRIS).
There are 9000 legal units in the population in 2024.
The variable covers hours worked in Belgium.
Only indices are published nationally.
The calendar month is the reference period.
The sampling error is small.
The non-sampling error is small. This error is more important for less detailed aggregates.
The first estimates are biased.
The principal sources of errors are the non-response error.
These indicators are based on hours worked.
Absolute figures are not published nationally.
Estimates for non-response : non-response is treated using imputation based on the past of the enterprise or based on the sector
Estimates for grossing-up to population levels : a weight based on the report (number of answering enterprises/number of enterprises in the universe) is attributed to every enterprise having answered in the stratim
Type of index : the production index in construction is a Laspeyres chain index
Method of weighting and chaining : the weights for aggregation are based on value added at factor costs of the year of reference 2021
Planned changes in production methods : none
Type of source : survey
Frame on which the source is based : DBRIS (Belgian business register)
Sample or census : stratified sample survey by Nace 3 level and size class with a rotating design. One third of the sample is replaced every year.
Criteria for stratification : Nace 3 level + size class
Threshold values and percentages :
From 5 to 9 persons : sampling rate=20%
From 10 to 19 persons : sampling rate=40%
20 persons and more = sampling rate=100%
Frequency of updating the sample : annually
The data are transmitted monthly to Eurostat and disseminated monthly to national users.
The production index is released within 1 month and 15 days after the reference period. Then, he is revised twice : two months and 15 days after the reference period and 3 months and 15 days after the reference period (in principle, the third release is final).
Questionnaires are sent to the reporting unit around the 15th of the reference month asking for a reply by the 15th of the following month. Data collection is normally closes 4 days before publication.
The definitions are equally applied to the 3 regions of Belgium.
In 2012, the construction survey was simplified.
Before 2012, the construction survey was based on a cut-off sample : the enterprises of 10 persons or more were surveyed.
After the first January 2012 : we use a rotating stratified design.
from 5 to 9 persons : sampling rate=20%
from 10 to 19 persons : sampling rate=40%
20 persons and more : sampling rate=100%
One third of the sample is replaced every year.
We did not observed a break in time series with the change of the year of reference.
