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 Community Innovation Survey (CIS) is a survey about innovation activities in enterprises. The survey is designed to collect the information on types of innovation, processes of development of innovation like cooperation patterns, financing and expenditure, objectives of innovation activities or barriers for initiating or implementing innovation.
The CIS provides statistics by type of innovators, economic activity and size class of enterprises. The survey is currently carried out every two years across the EU Member States, EFTA countries and EU candidate countries.
In order to ensure comparability across countries, Eurostat together with the countries develops a Harmonised Data Collection (HDC) questionnaire and drafts the methodological recommendations for implementation of each survey round.
The CIS 2022 implements the concepts and methodology of the Oslo Manual 4th Edition revised in 2018. The changes that the CIS has undergone due to the revision of the manual and their impact on the indicators collected are described in the Statistics Explained article: Community Innovation Survey – new features.
The legal framework for CIS 2022 is the Commission Implementing Regulation (EU) 2022/1092, which sets out the quality conditions and identifies the obligatory cross-coverage of economic sectors, size class of enterprises and innovation indicators. The target population is enterprises with at least 10 employed persons (sum of employees and self-employed persons) classified in the core NACE economic sectors (see 3.3). Further activities may be covered on a voluntary basis in national datasets. Most statistics are based on the 3-year reference period (t, t-1, t-2), but some use only one calendar year (t or t-2).
3.2. Classification system
Indicators related to the enterprises are classified by country, economic activity (NACE Rev. 2), size class of enterprises and type of innovation.
The main typology of classification of enterprises in reference to innovation is the distinction between innovation-active enterprises (INN) and not innovation-active enterprises (NINN).
The enterprise is considered as innovative (INN) if during the reference period it successfully introduced a a) product or a) business process innovation, c) completed but not yet implemented the innovation, d) had ongoing innovation activities, e) abandoned innovation activities or was f) engaged in in-house R&D or R&D contracted out. Non-innovative (NINN) enterprises had no innovation activity mentioned above whatsoever during the reference period.
3.3. Coverage - sector
3.3.1. Main economic sectors covered - NACE Rev.2
In accordance with the Commission Implementing Regulation (EU) 2022/1092 on innovation statistics, the following sectors of the economic activity are included in the core target population: NACE Sections B, C, D, E, H, J, K, and Divisions 46, 71, 72 and 73.
3.3.1.1. Main economic sectors covered - NACE Rev.2 - national particularities
There were no deviations.
3.3.2. Sector coverage - size class
In accordance with Commission Implementing Regulation (EU) 2022/1092 on innovation statistics, only the enterprises with 10 or more employed persons (sum of employees and self-employed persons) are included in the core target population.
3.3.2.1. Sector coverage - size class - national particularities
We used three size classes (10-49, 50-249, 250 or more), based on the number of employed persons (sum of employees and self-employed persons)
3.4. Statistical concepts and definitions
The description of concepts, definitions and main statistical variables will be available in CIS 2022 European metadata file (ESMS) Results of the community innovation survey 2022 (CIS2022) (inn_cis13) in Eurostat database.
3.5. Statistical unit
In CIS 2022 for the first time the statistical unit for which official statistics were reported was the enterprise; in all previous waves of CIS the statistical unit had been the legal unit.
CIS 2022 was the first innovation survey for which we had to report results at the enterprise level rather than at the level of legal units. As the innovation survey in Belgium is not conducted by the Belgian national statistical office (Statbel) we had to set up a legal construction that would allow Statbel to share enterprise information with our team. This legal construction was not yet in place when the sampling for CIS 2022 was done, so our sampling was still done at the level of legal units. All reporting units that were sent a survey form were legal units. Once the legal construction for cooperation with Statbel was in place, our sample was linked to the official business register maintained by Statbel. Sampled units that were deemed non-market oriented according to the official business register were removed from further processing. Statbel calculated weights at the enterprise level and calibrated those weights for number of enterprises, turnover and number of persons employed, whenever feasible.
The net sample in Belgium contained the following enterprises:
74% were enterprises consisting of 1 legal unit only;
15% were enterprises consisting of 2 legal units;
6% were enterprises consisting of 3 legal units;
5% were enterprises consisting of 4 or more legal units.
For 94% of these enterprises only 1 legal unit was sampled, for 4% of these enterprises 2 legal units were sampled, and for the remaining 2% of enterprises 3 or more legal units were sampled.
The set of responding enterprises contained the following:
For 95% of the responding enterprises a response was given by 1 legal unit only;
For 4% of the responding enterprises 2 legal units within the enterprise had responded;
For the remaining 1% of responding enterprises 3 or more legal units within each enterprise had responded.
For each enterprise Statbel identifies a privileged legal unit, usually this is the most important legal unit, that gives the enterprise its NACE code. For 97% of responding enterprises this privileged legal unit was among the responding legal units. Only for 3% of the responding enterprises only other legal units responded, not the privileged legal unit.
For a total of 216 enterprises responses given by 2 or more legal units needed to be aggregated to obtain one overall response for each enterprise. Not so surprisingly, enterprises for which such aggregation needed to be done predominantly were large and medium size enterprises, and only to a much smaller extent small enteprises: for only 1% of the responding small enterprises aggregation over multiple legal units needed to be done. For medium size enterprises the corresponding percentage was 7%, and for large responding enterprises aggregation over multiple legal units needed to be done for 22% of this set.
For aggregating responses over legal units within enterprises we used a pragmatic approach. We used all responses that were given and assumed additivity overall. When at least one legal unit within an enterprise indicated innovation, the overall response for the enterprise was innovation, etc. Expenditures were summed over legal units within an enterprise. Percentages, such as e.g., for share of turnover due to products new for the market, were first converted into absolute numbers (by multiplying the percentages with turnover) and then summed, and then again written as percentages for the enterprise as a whole.
3.6. Statistical population
Core target population is all enterprises in CORE NACE activities (see 3.3.1) with 10 or more employed persons (sum of employees and self-employed persons).
3.7. Reference area
Belgium is composed of three regions (at NUTS 1 level): Brussels, Flanders, and Wallonia. Each region is endowed with its own statistical office. By law, CIS is conducted in a decentralized way in Belgium, within each of these three regions. A legal agreement is in place to make sure there is ample coordination between the regions to produce the required results for EUROSTAT reporting. By default, results for Belgium are available at the NUTS 1 level.
NUTS1 was used as a stratification dimension for the sampling.
The CIS2022 regional data are calibrated at the Statistical Unit Enterprise level.
3.8. Coverage - Time
Several rounds of Community Innovation Survey have been conducted so far at two-year interval since the end of the 90’s.
3.8.1. Participation in the CIS waves
CIS wave
Reference period
Participation (Yes/No)
Comment (deviation from reference period)
CIS2
1994-1996
Yes
No
CIS3
1998-2000
Yes
No
CIS light
2002-2003*
No
CIS4
2002-2004
Yes
No
CIS2006
2004-2006
Yes
No
CIS2008
2006-2008
Yes
No
CIS2010
2008-2010
Yes
No
CIS2012
2010-2012
Yes
No
CIS2014
2012-2014
Yes
No
CIS2016
2014-2016
Yes
No
CIS2018
2016-2018
Yes
No
CIS2020
2018-2020
Yes
No
CIS2022
2020-2022
Yes
No
*two reference periods can be distinguished for CIS light: 2000-2002 and 2001-2003
3.9. Base period
Not relevant.
CIS indicators are available according to 3 units of measure:
NR: Number for number of enterprises and number of persons employed.
THS_EUR: Thousands of euros. All financial variables are provided in thousands of euros, i.e. Turnover or Innovation expenditure.
PC: Percentage. The percentage is the ratio between the selected combinations of indicators.
For CIS 2022, the time covered by the survey is the 3-year period from the beginning of 2020 to the end of 2022.
Some questions and indicators refer to one year — 2022.
The list of indicators specifying whether they cover the 3-year period or refer to one year according to the HDC will be available in the Annex section of the European metadata (ESMS).
6.1. Institutional Mandate - legal acts and other agreements
This Regulation establishes innovation statistics on a statutory basis and makes the delivery of certain variables compulsory e.g. innovation activities, cooperation, development, expenditures and turnover (see the Regulation). Each survey wave may additionally include further variables.
In addition, the Regulation defines the obligatory cross-coverage of economic sectors and size class of enterprises.
6.1.1. National legislation
Legal agreement on how the three regions and the national government organize the production of official STI statistics: Ejustice website.
6.2. Institutional Mandate - data sharing
A legal construction was set up to allow for data sharing between Statbel, Belgium’s national statistical office, and the three entities that conduct CIS for each of the three regions in Belgium. This data sharing was necessary to implement the enterprise approach in CIS, as it is Statbel that maintains the official business register of Belgium; Statbel also does the business profiling for Belgium, laying out the relationships between legal units and enterprises.
7.1. Confidentiality - policy
The Belgian Interfederal Institute of Statistics (IIS) coordinates the statistics production at the regional and national level in Belgium. It abides by the European Statistics Code of Practice, including its principle 5, on statistical confidentiality and data protection (https://www.iis-statistics.be/doc/CoC_fr.pdf ).
The Belgian legal agreement on how the three regions and the national government organize the production of official STI statistics (Ejustice website) explicitly refers to an earlier regulation of the European Parliament and of the Council on European STI business statistics. This regulation has now been replaced by the regulations mentioned in section 6.1 above, which explicitly state confidentiality is required within the European Statistical System.
7.2. Confidentiality - data treatment
Data cells compiled with data of fewer than 7 units are flagged as confidential. When an observed response represents 80% or more of a certain cell, this cell is also flagged as confidential.
8.1. Release calendar
In the Flemish region, for core R&D statistics derived using the CIS 2022 R&D data, the June 30 publication date is announced by the Flemish statistical office on its website (Vlaanderen - Publicatieagenda). The publication date and the date when the next update will be made available is published next to each individual statistic.
Main innovation statistics are in the Flemish region published by the end of September 2024.
8.2. Release calendar access
In the Flemish region, the publication dates of official Flemish statistics that are published on the website of Statistics Flanders are listed and publicly available. For innovation statistics that are not published on the website of Statistics Flanders, the publication dates are shared well in advance with the most important stakeholders.
8.3. Release policy - user access
In Flanders, two core R&D statistics based on CIS are published online on June 30. More general innovation statistics derived from CIS are published online on September 30.
In Belgium, academic researchers may obtain access to regional microdata for academic purposes by signing confidentiality agreements. They may obtain access to the national microdata by submitting a project proposal to a committee that represents both the national and regional levels, and by signing a confidentiality agreement.
The Belgian Science Policy Office publishes results for CIS on its website (meri.belspo.be) by the end of the year data were transmitted to Eurostat. An interactive tool is available, as well as a short document providing some background for the most salient results.
CIS is conducted and disseminated at two-year interval in pair years.
Accessibility and clarity refer to the simplicity and ease for users to access statistics using simple and user-friendly procedure, obtaining them in an expected form and within an acceptable time period, with the appropriate user information and assistance: a global context which finally enables them to make optimum use of the statistics.
Tabulated statistics for the Flemish region can be downloaded from its STI indicator book website: Vlaamsindicatorenboek website (in Dutch).
10.3.1. Data tables - consultations
Not requested.
10.4. Dissemination format - microdata access
Microdata is available at the Eurostat Safe center. Upon request, microdata can be made available if all three regions have agreed and the user has signed a standard confidentiality agreement statement (adherence to the laws governing statistical practices).
10.4.1. Dissemination of microdata
Mean of dissemination
Availability of microdata
Comments, links, ...
Eurostat SAFE centre
Yes
National SAFE centre
Yes
Upon request, academic researchers can obtain microdata either from a region for regional data, or from Belspo for national microdata, as long as confidentiality agreements are signed. At the national level, data are usually delivered without business enterprise ID's, but are generally not altered nor recoded.
Eurostat: partially anonymised data (SUF)
No
National: partially anonymised data
No
10.5. Dissemination format - other
None.
10.5.1. Metadata - consultations
Not requested.
10.6. Documentation on methodology
Meri belspo website provides some meta-data. Quality Reports can be delivered upon request, but are not generally sought after. Starting with CIS2020, countries' integrated metadata and quality reports are made available in the EUROSTAT database.
10.6.1. Metadata completeness - rate
Not requested.
10.7. Quality management - documentation
In our publications, we refer to the international guidelines (EUROSTAT and OECD (e.g., Oslo Manual)) which we follow. If quality reports were requested, they would be made available to users, but this hasn't happened yet.
11.1. Quality assurance
Belgium follows Eurostat's recommendations.
11.2. Quality management - assessment
The fact that the CIS is not mandatory may reduce international comparability, as in most EU member states it is mandatory, and a randomized experiment by Norway (Wilhelmsen, 2012) has shown that the mere fact of the CIS being voluntary or mandatory yielded clearly different results. They found that innovation rates were higher in the voluntary condition of the randomized experiment (all else being equal).
We pre-fill as much as possible, R&D expenditure and personnel figures from the previous year’s R&D survey are shown as a reference. This helps respondents when making estimates for the reference year, but may also perpetuate a mistake made in earlier years.
The fact that countries differ in the practical implementation of CIS may negatively impact comparability between countries. Recommendations of best practices might help here.
In CIS 2022 there was more integration with other official business statistics in Belgium than in earlier waves of CIS: for the first time results were related to the official business register of Belgium; values for NACE, turnover and number of persons employed were taken from SBS and the official business register; weights were also calibrated on number of enterprises, turnover and number of persons employed whenever feasible.
12.1. Relevance - User Needs
Besides the needs Eurostat, OECD, and the European Innovation Scoreboard have, there are also regional stakeholders and academics that use results obtained with CIS. Academic researchers regularly ask for additional questions to be included in the questionnaire; to the extent possible we accommodate their requests. New questions are always tested first in cognitive interviews.
12.1.1. Needs at national level
User group
Short description of user group
Main needs for CIS data of the user group Users’ needs
Institutions - European Level
Eurostat
Institutions - International organizations
OECD
Institutions - National level
National government
Institutions - Regional level
Regional governments at NUTS 1 level
Various innovation indicators, such as % of innovators by size and industry, amount of R&D expenditure, ...
Researchers and students
Master thesis students, PhD students, professors, researchers within research institutes
Research projects
12.2. Relevance - User Satisfaction
We do not conduct a user satisfaction survey. Occasionally, we do receive requests for more detailed breakdowns, e.g. for certain sector associations (e.g. results at NUTS 2 level (provinces) or national level). We handle each of these requests on a case-by-case basis. Generally, we do not provide results at Nuts 2 level (Provinces).
12.3. Completeness
We covered all the compulsory core NACE sectors. However, even within the compulsory sectors, some cells are missing. This is because no observations were available for these cells (either because there is no firm in the population, or because none of the surveyed firms answered the questionnaire). Belgium is a small country, therefore we also run into confidentiality issues for certain cells.
We did not ask the following (voluntary) questions: the questions of section 2 on business characteristics (2.1, 2.2, 2.3 and 2.4), the voluntary details on innovation expenditure (3.8), the question on tax incentives (4.3), the questions of section 5 on specific factors and actions (5.1, 5.2, 5.3 and 5.4), the question on factors driving decisions on eco-innovations (6.2), the question on the tertiary degree of employees (7.2), the question on general expenditure (7.6), the question on inflows and outflows (7.8) and the question on intra-group loans (7.9).
12.3.1. Data completeness - rate
Not requested.
13.1. Accuracy - overall
Accuracy in the statistical sense denotes the closeness of computations or estimates to the exact or true values. Statistics are not equal with the true values because of variability (the statistics change from implementation to implementation of the survey due to random effects) and bias (the average of the possible values of the statistics from implementation to implementation is not equal to the true value due to systematic effects).
13.2. Sampling error
Restricted from publication
13.2.1. Sampling error - indicators
The main indicator used to measure sampling errors for CIS data is the coefficient of variation (CV).
CV= Coefficient of variation (%) = 100 * (Square root of the estimate of the sampling variance) / (Estimated value)
Formula:
where
and
13.2.1.1. Coefficient of variations for key variables
Coefficient of variation (%) for key variables by NACE categories and for enterprises with 10 or more employed persons
NACE
Size class
(1)
(2)
(3)
Core NACE (B-C-D-E-46-H-J-K-71-72-73)
Total
1.49%
3.88%
2.75%
Core industry (B_C_D_E - excluding construction)
Total
1.98%
5.08%
3.53%
Core Services (46-H-J-K-71-72-73)
Total
2.09%
6.95%
3.92%
(1) = Coefficient of variation for the percentage of innovative enterprises (INN) in the total population of enterprises (ENT) (2) = Coefficient of variation for the turnover of product innovative enterprises with new or improved products (TUR_PRD_NEW_MKT), as a percentage of total turnover of product innovative enterprises [TOVT,INNO_PRD]. (3) = Coefficient of variation for percentage of product and/or process innovative enterprises (incl. enterprises with abandoned and or on-going activities) involved in any innovation co-operation arrangement [COOP_ALL,INN], as a percentage of innovative enterprises (INN).
13.2.1.2. Variance estimation method
Variances and coefficients of variation were estimated using the default of proc surveymeans in SAS. Our variance estimates took into account our sampling design. However, the fact that imputations were made for missing values was NOT taken into account, nor the fact that nonresponse was a major source of uncertainty in our estimates. Hence, the coefficients of variation reported above can be considered to be lower bounds.
13.3. Non-sampling error
Non-sampling errors occur in all phases of a survey. They add to the sampling errors (if present) and contribute to decreasing overall accuracy. It is important to assess their relative weight in the total error and devote appropriate resources for their control and assessment.
13.3.1. Coverage error
Coverage errors (or frame errors) are due to divergences between the target population and the frame population. The frame population is the set of target population members that have a chance to be selected into the survey sample. It is a listing of all items in the population from which the sample is drawn that contains contact details as well as sufficient information to perform stratification and sampling.
13.3.1.1. Over-coverage - rate
Not requested.
13.3.1.2. Common units - proportion
Not requested.
13.3.1.3. Under covered groups of the target population
There may be some undercoverage of recently founded firms due to the fact that the National Social Security Office Employer database we use as frame population is based on information from the previous year. This is unavoidable, however, given the delay in information available from these firms.
13.3.1.4. Coverage errors in coefficient variation
Variances and coefficients of variation were estimated using the default proc surveymeans in SAS. Our variance estimates took into account our sampling design. However, the fact that imputations were made for missing values was NOT taken into account, nor the fact that nonresponse was a major source of uncertainties in our estimates. Hence, the coefficients of variation reported above can be considered to be lower bounds.
13.3.2. Measurement error
Measurement errors occur during data collection and generate bias by recording values different than the true ones. The survey questionnaire used for data collection may have led to the recording of wrong values, or there may be respondent or interviewer bias.
13.3.2.1. Measures for reducing measurement errors
We have not collected any systematic evidence regarding measurement error in CIS2022.
To reduce the risk of measurement errors, we review our questionnaire form every time to improve its clarity and user friendliness. Efforts are made to reduce response burden as much as possible, e.g. by prefilling as many fields as possible. Any comments left on the questionnaire itself or suggestions given by companies are taken into consideration when redesigning the next CIS. We conduct cognitive interviews when (re)designing our survey forms.
13.3.3. Non response error
We did not conduct a non-respondent survey as we found in a randomized experiment conducted in the context of CIS 2014 that responses given to short innovation forms are not directly comparable to responses given to the regular, long innovation form (Hoskens & Debackere, 2024). Innovation rates obtained with the short forms were, all else being equal, higher than those obtained with the regular long forms. We used calibration to adjust the survey weights for non-response. For a small number of large non-responding enterprises we made imputations using nearest neighbor hot decking and ratio means.
13.3.3.1. Unit non-response - rate
See below.
13.3.3.1.1. Un-weighted and weighted unit non-response rate by NACE categories and for enterprises with 10 or more employed persons
Un-weighted and weighted unit non-response rate by NACE categories and for enterprises with 10 or more employed persons
NACE
Number of eligible units with no response
Total number of eligible units in the sample
Un-weighted unit non-response rate (%)
Weighted unit non-response rate (%)
Core NACE (B-C-D-E-46-H-J-K-71-72-73)
3159
7313
43 %
45%
Core industry (B_C_D_E - excluding construction)
1291
3089
42%
43%
Core Services (46-H-J-K-71-72-73)
1868
4224
44%
46%
The number of eligible units is the number of sample units, that ultimately indeed belong to the target population.
13.3.3.1.2. Maximum number of recalls/reminders before coding
Two written reminders are sent out by post, phone call and e-mail reminders are sent to a limited number of enterprises.
Enterprises in Flanders for whom e-mail addresses were available were sent three e-mail reminders and one paper reminder. Enterprises for whom no e-mail address was available (a small minority) were sent two paper reminders. Phone calls were made during three months to encourage enterprises to respond.
13.3.3.2. Item non-response - rate
See below.
13.3.3.2.1. Item non-response rate for Turnover (in Core NACE: B-C-D-E-46-H-J-K-71-72-73 enterprises with 10 or more employed persons)
Item non-response rate for Turnover (in Core NACE: B-C-D-E-46-H-J-K-71-72-73 enterprises with 10 or more employed persons).
Item non-response rate (un-weighted)
(%)
Imputation
(Yes/No)
If imputed, describe method used, mentioning which auxiliary information or stratification is used
Turnover
0 %
No
Turnover of legal units was extracted from a range of administrative sources (VAT, National Bank, Balance sheet, …). Only 42 legal units had no source for turnover, in which case it was left at 0. Enterprise turnover was calculated by summing turnover over the legal units, and subtracting intra-enterprise flows (the amount of VAT sales made between a VAT supplier and a VAT customer belonging to the same enterprise).
13.3.3.2.2. Item non response rate for new questions
Item non-response rate for new questions in CIS t (in Core NACE: B-C-D-E-46-H-J-K-71-72-73 enterprises with 10 or more employed persons)
NEW QUESTIONS IN CIS 2022
Inclusion in national questionnaire (Yes/No)
Item non response rate (un-weighted) (%)
Comments
3.9 -- Reasons for not having more innovation activities
Yes
9.2%
Item non-response rate of 9.2% is taken over unit respondents only; when unit non-respondents are included as well, total item non-response rate considerably higher.
3.10 -- Reasons for having no innovation activities
Yes
6.7 %
Item non-response rate of 6.7% is taken over unit respondents only; when unit non-respondents are included as well, total item non-response rate is considerably higher.
13.3.4. Processing error
We are not aware of any processing errors
13.3.5. Model assumption error
Not requested.
Timeliness and punctuality refer to time and dates, but in a different manner.
14.1. Timeliness
The timeliness of statistics reflects the length of time between data availability and the event or phenomenon they describe.
14.1.1. Time lag - first result
Timeliness of national data – date of first release of national level: July 10, 2024 (delivery to Eurostat); hence 18.3 months after the end of the reference period.
14.1.2. Time lag - final result
Not requested.
14.2. Punctuality
Punctuality refers to the time lag between the release date of data and the target date on which they were scheduled for release as announced officially.
14.2.1. Punctuality - delivery and publication
The date of transmission of complete and validated data to Eurostat was July 10, 2024, which is 10 days after June 30, 2024. The delay was due to the extra processing that needed to happen to convert results to the enterprise level. For this conversion we needed to cooperate with Statbel, Belgium’s national statistical institute. It was the first time this cooperation was in place, usually it took several iterations to streamline the necessary processing steps.
15.1. Comparability - geographical
Comparability aims at measuring the impact of differences in applied statistical concepts and definitions on the comparison of statistics between geographical areas, non-geographical domains, or over time.
The coherence of statistical outputs refers to the degree to which the statistical processes by which they were generated used the same concepts (classifications, definitions, and target populations) and harmonised methods. Coherent statistical outputs have the potential to be validly combined and used jointly.
We use the same statistical concepts and definitions in all regions, and as far as we're aware, those are the same concepts and definitions used in the rest of the EU.
15.1.1. Asymmetry for mirror flow statistics - coefficient
Not requested.
15.1.2. National questionnaire – compliance with Eurostat model questionnaire
Methodological deviations from the CIS Harmonised Data Collection (HDC)
Questions not included in national questionnaire compared to HDC
Comment
We did not ask the following (voluntary) questions: the questions of section 2 on business characteristics (2.1, 2.2, 2.3 and 2.4), the voluntary details on innovation expenditure (3.8), the question on tax incentives (4.3), the questions of section 5 on specific factors and actions (5.1, 5.2, 5.3 and 5.4), the question on factors driving decisions on eco-innovations (6.2), the question on the tertiary degree of employees (7.2), the question on general expenditure (7.6), the question on inflows and outflows (7.8) and the question on intra-group loans (7.9).
Changes in the filtering compared to HDC
Comment
Q 3.8: innovation expenditure
Non-innovators cannot skip this question but have to respond to this question (by ticking “none” for each option).
R&D expenditures were in our forms separated from other innovation expenditures. In Brussels and Wallonia respondents that indicated “no” to the binary questions on internal and external R&D could skip the questions on R&D expenditure. In Flanders respondents indicating “no” to the binary questions on internal and external R&D had to confirm their response in follow-up questions on expenditure and personnel, they were not allowed to skip these questions.
Q 3.9 and Q 3.10: reasons not to innovate (more)
Both questions were combined and were formulated as reasons not to innovate (more), so both innovators and non-innovators had to respond to this question.
Q.3.11: cooperation and Q 3.12: cooperation details
Our filter is different, as we do not make enterprises skip the next question if they only ticked "yes" for option c. This is more in line with the idea to not teach respondents that if they tick "no", they can skip more questions. Only respondents indicating no cooperation at all were allowed to skip the question on more cooperation details. Respondents indicating only cooperation for other business activities had to indicate more details in follow-up question 3.12. When processing the data, however, those details were removed, so that details were only left for firms cooperating on R&D or innovation.
Q 6.1: eco innovation
All respondents were allowed to respond to the question on eco innovation. In our data processing, however, responses to this question were only kept for respondents indicating either product or process innovation (hence, finished innovations). All other responses were removed.
15.1.3. National questionnaire – additional questions
Methodological deviations from the CIS Harmonised Data Collection (HDC)
Additional questions in national questionnaire (not included in HDC)
Comment
We asked for R&D personnel in 2022: both head counts and FTE, and we asked this separately for internal R&D personnel and for external R&D personnel, as well as a more detailed breakdown of costs incurred for R&d in 2022 (the latter only in Brussels and Flanders).
In Flanders, CIS is used to calculate official R&D statistics for even years.
We asked for new-to-world product innovations.
In Brussels and Wallonia only: Did your enterprise conduct R&D in biotechnology,biochemistry, nanotechnology, AI?
Only in Brussels and Flanders we asked the following: Did your enterprise notice an influence on its operations in the 2020-2022 period from the following occurrences? (economic crisis, inflation, shortages, natural disasters, health crisis, war, ...)
15.2. Comparability - over time
Following the publication of the 2018 Oslo Manual, from CIS 2018 onwards business process innovation was more broadly defined and now not only included technological process innovation, but also what before was called organizational and marketing innovation. The conceptual changes of the 2018 Oslo Manual implied big changes for the survey forms for CIS. Surprisingly, however, overall innovation rates remained to be fairly comparable over time: the overall innovation rates for Belgium were 68% and 68% in 2016 and 2018, respectively.
A big change for CIS 2022 was the required implementation of enterprises as statistical unit. In all previous waves of CIS, Belgium had used legal units as statistical unit. Again, surprisingly, overall innovation rates remained fairly stable despite the methodological change: when the CIS 2022 data were analyzed using the old methodology (with legal units as statistical unit) the overall innovation rate for Belgium was 70%. When the CIS 2022 data were analyzed using the new methodology (with enterprises as statistical units) the overall innovation rate for Belgium was also 70%. The fact that overall innovation rates are predominantly determined by small firms, given they are fairly numerous in the economy, and most of those small firms are enterprises consisting of one legal unit, might have something to do with the limited impact of the implemented methodological change. One change that did have some impact on more fine-grained results, however, was the fact that now for the first time CIS results were reported using the official NACE codes of Belgium’s official business register. Those NACE codes are predominantly based on value added. In earlier waves of CIS we did not have access to the official Belgian business register. We had to work with the register of active employers maintained by the Belgian National Social Security Office. The NACE codes in that register are predominantly employment based. So, some shifts between NACE codes may occur.
15.2.1. Length of comparable time series
Not requested.
15.3. Coherence - cross domain
See the comparison between SBS and CIS data in the section 15.3.3 below.
15.3.1. Coherence - sub annual and annual statistics
Not requested.
15.3.2. Coherence - National Accounts
Not requested.
15.3.3. Coherence – Structural Business Statistics (SBS)
This part compares key variables for aggregated CIS data with SBS data Definition of relative difference between CIS and SBS data: DIFF = (SBS/CIS)*100
Comparison between SBS and CIS data (relative difference) by NACE categories and for enterprises with 10 or more employed persons
NACE
Size class
Number of enterprises (SBS/CIS)*
Number of employed persons (SBS/CIS)*
Total Turnover (SBS/CIS)*
Core NACE (B-C-D-E-46-H-J-K-71-72-73)
Total
105%
106%
120%
Core industry (B_C_D_E - excluding construction)
Total
103%
101%
107%
Core Services (46-H-J-K-71-72-73)
Total
106%
111%
132%
* Numbers are to be provided for the last year of the reference period (t)
15.4. Coherence - internal
Not requested.
Restricted from publication
16.1. Cost
Restricted from publication
17.1. Data revision - policy
Not requested.
17.2. Data revision - practice
Not requested.
17.2.1. Data revision - average size
Not requested.
18.1. Source data
Raw data are the responses given by enterprises to CIS 2022.
18.1.1. Sampling frame (or census frame)
At the time of sampling we did not yet have access to Belgium’s official business register. As a proxy, the register of active employers maintained by the Belgian National Social Security Office was used as sampling frame.
18.1.2. Sampling design
Sampling was done before we obtained access to the official Belgian business register. Legal units were sampled, using as population frame the register of active employers maintained by the Belgian National Social Security Office. We used their size categories and NACE codes. As CIS is conducted in a decentralized way in Belgium, within each of its three NUTS 1 regions, sampling was also done within each of these three regions.
1. For the Brussels Region:
A census was done for all large and medium-sized firms, as well as small firms belonging to NACE 20-22, 33, 26-30, 59, 61-63, and 71-72.
2. For the Walloon Region:
A census was done for all large firms, for medium-sized firms belonging to NACE 8-46 and 58-73, as well as small firms belonging to NACE 20-21, 26-27, and 72.
3. For the Flemish region:
Besides firm size and sector, a third stratification variable that was taken into account for sampling in the Flemish region reflected whether a firm was known to have continuous R&D spending, was active in biotechnology, recently received public funding for R&D, was involved in R&D or innovation projects with government R&D centers or was involved in specific government projects promoting R&D. The inventory of firms with continuous R&D spending as obtained from the 2022 R&D survey was used as a base for the first of these sources.
Census sampling was done for all large size firms (250 or more employees), for all medium size firms (50-249 employees) and for small size firms (10-49 employees) of NACE 19-22, 26-30, 59-63, and 71-72. Census sampling was also done of the small size firms in the other NACE sectors when they were selected based on the third stratification variable (continuous R&D, public support for R&D, etc.).
For the remaining small size firms an overall sample size of 1000 firms to be randomly selected was set. Neymann allocation was then used to determine the sample sizes for NACE aggregates that would be used for reporting, taking into account the five precision criteria EUROSTAT has specified for CIS, as well as the expected percentage of enterprises with internal R&D. Taking into account expected levels of non-response, a minimum sample size of 50 firms was set for each NACE aggregate. The NACE aggregates considered were: NACE 05-09,12,32,35-39; NACE 10-11; NACE 13-15; NACE 16-18,31; NACE 19,22-23; NACE 24-25,33; NACE 46; and NACE 49-53,58,64-66,73.
18.1.3. Target population and sample size
Sample/census indicator
Number of enterprises
Target population (A) (*)
14314
Sample (B = C+D)
7313
In case of combination sample/census:
Sampled units (C)
2220
Enumerated units/census (D)
5093
Overall sample rate (E = 100*B/A)
51.09%
(*) CIS core population, i.e. NACE Rev.2 B-C-D-E-46-H-J-K-71-72-73 enterprises with 10 or more employed persons.
18.1.4. Data source for pre-filled variables
Variables and indicators filled or prefilled from other sources.
Variables/Indicators
Source
Reference year
Turnover and number of employees
Statistical Business Register
2022
Member of an enterprise group
Balance sheet data, company websites, Bel-first (a commercial database owned by Moody’s, containing balance sheet data and ownership data).
EGR 2022 (+Statistical Business Register mainly for all resident groups)
2022-2024
Country of head office for enterprise groups
Balance sheet data, company websites, Bel-first (a commercial database owned by Moody’s, containing balance sheet data and ownership data)
2022-2024
18.1.5. Data source and variables used for derivation and weighting
Item
Response
Data source used for deriving population totals
Turnover and number of persons employed: obtained from the Enterprise Business Register.
Variables used for weighting
Non response correction by post-stratification was formed by NUTS 1 region, NACE, size, and in one region also R&D status. Calibration was done on number of enterprises, turnover and number of persons employed, whenever feasible.
18.2. Frequency of data collection
According to the Commission Implementing Regulation (EU) 2022/1092, the innovation statistics shall be provided to Eurostat every two years in each even year. The data collection takes place every second year in year t-2 preceding the data provision.
18.3. Data collection
In Brussels and Wallonia sampled reporting units (legal units) are sent a paper form and given the option to return their response on paper in an enclosed envelope, or enter their response in a web survey form. They are sent a first reminder consisting of a postcard, a second reminder is sent in a third wave, consisting of a letter with link and login information.
In Flanders, the majority of sampled reporting units (legal units) are sent a survey invitation by e-mail. Respondents are invited to respond to the survey in a secure web survey form. Only a small minority of sampled reporting units for whom no valid e-mail address could be found are sent a paper form. The paper form allows respondents to return their response on paper free of charge, or to enter their response in a secure web form. A first reminder is sent by e-mail, followed by a second reminder on paper, enclosing the paper survey form. A third reminder is sent again by e-mail. Sampled reporting units for whom no valid e-mail address could be found are sent a second paper form by postal mail. A fourth reminder is again sent by e-mail. Following the first paper reminder, phone calls are made to encourage firms to respond.
The web forms conduct error checks on the responses respondents enter. The error checks yield warning messages, but respondents are free to ignore them. Follow-up questions that may be skipped following a “no” response are greyed out.
Paper responses are entered twice by our staff in the web survey interface. Both inputs are compared to detect potential data entry errors.
18.3.1. Survey participation
The survey is voluntary.
18.3.2. Survey type
We used a combination of sample and census, depending on the size of the population in the various strata in the population.
18.3.3. Combination of sample survey and census data
See 18.3.4. for more details on the sampling procedure.
18.3.4. Census criteria
Only in the more populated cells in our sampling design, random sampling was applied. Generally, these were the smaller, and more low-tech firms.
As mentioned before, sampling was done before we had access to the official Belgian business register. Legal units were sampled, using as population frame the register of active employers maintained by the Belgian National Social Security Office. We used their size categories and NACE codes.
For the Brussels Region: A census was done for all large and medium-sized firms, as well as small firms belonging to NACE 20-22, 23-30, 30, and 71-72.
For the Walloon Region: A census was done for all large firms, for medium-sized firms belonging to NACE 8-46 and 58-73, as well as small firms belonging to NACE 16-18, 20-21, 26-27, and 72.
For the Flemish Region: Besides firm size and sector, a third stratification variable that was taken into account for sampling in the Flemish region reflected whether a firm was known to have continuous R&D spending, was active in biotechnology, recently received public funding for R&D, was involved in R&D or innovation projects with government R&D centers or was involved in specific government projects promoting R&D. The inventory of firms with continuous R&D spending as obtained from the 2022 R&D survey was used as a base for the first of these sources.
Census sampling was done for all large size firms (250 or more employees), for all medium size firms
(50-249 employees) and for small size firms (10-49 employees) of NACE 19-22, 26-30, 59-63, and 71-72. Census sampling was also done of the small size firms in the other NACE sectors when they were selected based on the third stratification variable (continuous R&D, public support for R&D, etc.).
For the remaining small size firms an overall sample size of 1000 firms to be randomly selected was set. Neymann allocation was then used to determine the sample sizes for NACE aggregates that would be used for reporting, taking into account the five precision criteria EUROSTAT has specified for CIS, as well as the expected percentage of enterprises with internal R&D. Taking into account expected levels of non-response, a minimum sample size of 50 firms was set for each NACE aggregate. The NACE aggregates considered were: NACE 05-09,12,32,35-39; NACE 10-11; NACE 13-15; NACE 16-18,31; NACE 19,22-23; NACE 24-25,33; NACE 46; and NACE 49-53,58,64-66,73.
18.3.5. Data collection method
Data collection method
Survey method
Yes/No
Comment
Face-to-face interview
No
Physical meetings were set up with some large enterprises to make sure their responses were in line with the guidelines of the Oslo Manual and the Frascati Manual. The visited enterprises then usually entered their final responses online, in the web survey form.
Telephone interview
Yes
A limited number of responses resulted from telephone interviews: some respondents asked to be interviewed over the phone, others who initially refused to respond when contacted over the phone to remind them of the survey, were converted to respondents.
Postal questionnaire
Yes
This was the second most used response mode
Electronic questionnaire (format Word or PDF to send back by email)
Yes
This response mode was rarely used.
Web survey (online survey available on the platform via URL)
Yes
This was the most used response mode.
Other
No
18.4. Data validation
Not requested.
18.5. Data compilation
We use the SAS routine that was originally made available by Eurostat for CIS 4 for data editing and data imputation. We updated this routine for use for CIS 2022. In this routine continuous variables are estimated using weighted ratio means, and nominal and ordinal variables are estimated using nearest neighbor hot deck imputation. More details are given in the document describing EUROSTAT’s methodological recommendations for CIS 2022, in section 5.5 and Annex 7.
In our data processing for CIS 2022 there were several changes compared to previous waves of CIS:
Results are now reported for enterprises. In all previous waves of CIS results were reported for legal units, a lower level of aggregation.
Employment numbers are now numbers for persons employed taken from the official business register, which include besides employees on payroll also owners, family workers and outworkers whose income is a function of the value of the outputs of the enterprise. In previous waves of CIS, employment numbers reflected number of employees only and were taken from balance sheets or from survey responses.
Turnover numbers are now those of the official business register and are at the enterprise level. In earlier waves of CIS turnover was reported at the level of legal units and was taken from balance sheets or from survey responses.
Survey weights are now calibrated for number of enterprises, turnover and number of persons employed, when feasible. In earlier waves of CIS survey weights were inverse sampling probabilities, corrected for nonresponse.
NACE codes used are now those of Belgium’s official business register, and are predominantly based on value added. In previous waves of CIS we did not have access to Belgium’s official business register. The NACE codes of the National Social Security Office were then used instead. These values were pre-filled in the survey forms, and respondents were able to indicate corrections if needed. The NACE codes of the National Social Security Office are predominantly employment-based.
18.5.1. Imputation - rate
Imputation is the method of creating plausible (but artificial) substitute values for all those missing.
Definition of imputation rate:Imputation rate (for the variable x) (%) = 100 * (Number of replaced values) / (Total number of values for a given variable)
Definition of weighted imputation rate: Weighted imputation rate= 100 * (Number of total weighted replaced values) / (Total number of weighted values for a given variable)
18.5.1.1. Imputation rate for metric variables
Imputation rate (%) for metric variables by NACE categories and for enterprises with 10 or more employed persons:
NACE
Size class
Total Turnover (1)
Turnover from products new to the market (2)
R&D expenditure in-house (3)
Unweighted
Weighted
Unweighted
Weighted
Unweighted
Weighted
Core NACE (B-C-D-E-46-H-J-K-71-72-73)
Total
0%
0%
9%
24%
9%
9%
Core industry (B_C_D_E - excluding construction)
Total
0%
0%
8%
34%
9 %
9%
Core Services (46-H-J-K-71-72-73)
Total
0%
0%
9%
9%
9%
9%
(1) = Imputation rate (%) for the total turnover in the last year of the reference period (t) (TUR)
(2) = Imputation rate (%) for the share of the turnover in the last year of the reference period (t) due to new or improved product new to the market in the total turnover for product innovative enterprises TUR_PRD_NEW_MKT/TOVT(INNO_PRD)
(3) = Imputation rate (%) for the R&D expenditure performed in-house (EXP_INNO_RND_IH)
18.5.2. Weights calculation
Weights calculation method for sample surveys
Method
Selected applied method
Comments
Inverse sampling fraction
This was used as starting point in step 1.
Non-respondent adjustments
Post-stratification to account for non-response was done in a second step.
Other
In a third step, calibration was done using the Calmar macro available for SAS. The truncated linear method was used. Calibration was done on number of enterprises, turnover, and number of persons employed, whenever feasible.
Calibration was not feasible in all cells of our stratification design. Calibration on turnover and employment was feasible for 1/3rd, 2/3rd and 90% of the cells of the stratification design for the regions Brussels, Wallonia and Flanders, respectively.
A limited number of specific observations are considered separately by giving
them a weight=1. These are observations for which the R&D expenditures are
so high compared to the rest of the sample that multiplying these observations
by more than one would bias the final result for R&D and innovation
expenditures variables.
18.6. Adjustment
The survey was carried out at the level of legal units, but results are reported at the level of enterprises. In a small number of cases responses were obtained from multiple legal units that belonged to the same enterprise, hence, their responses had to be aggregated to obtain one response per enterprise. More details on this adjustment are given in section 3.5, on statistical unit, above.
Calibration of weights was done whenever feasible, in order to aim for overall totals for number of enterprises, turnover and number of persons employed that are in line with the overall totals obtained for these variables in the structural business survey and in the official business register of Belgium. More details on this calibration are given in sections 18.1.5 (data source and variables used for derivation and weighting) and 18.5.2 (weights calculation).
The Community Innovation Survey (CIS) is a survey about innovation activities in enterprises. The survey is designed to collect the information on types of innovation, processes of development of innovation like cooperation patterns, financing and expenditure, objectives of innovation activities or barriers for initiating or implementing innovation.
The CIS provides statistics by type of innovators, economic activity and size class of enterprises. The survey is currently carried out every two years across the EU Member States, EFTA countries and EU candidate countries.
In order to ensure comparability across countries, Eurostat together with the countries develops a Harmonised Data Collection (HDC) questionnaire and drafts the methodological recommendations for implementation of each survey round.
The CIS 2022 implements the concepts and methodology of the Oslo Manual 4th Edition revised in 2018. The changes that the CIS has undergone due to the revision of the manual and their impact on the indicators collected are described in the Statistics Explained article: Community Innovation Survey – new features.
The legal framework for CIS 2022 is the Commission Implementing Regulation (EU) 2022/1092, which sets out the quality conditions and identifies the obligatory cross-coverage of economic sectors, size class of enterprises and innovation indicators. The target population is enterprises with at least 10 employed persons (sum of employees and self-employed persons) classified in the core NACE economic sectors (see 3.3). Further activities may be covered on a voluntary basis in national datasets. Most statistics are based on the 3-year reference period (t, t-1, t-2), but some use only one calendar year (t or t-2).
10 July 2024
The description of concepts, definitions and main statistical variables will be available in CIS 2022 European metadata file (ESMS) Results of the community innovation survey 2022 (CIS2022) (inn_cis13) in Eurostat database.
In CIS 2022 for the first time the statistical unit for which official statistics were reported was the enterprise; in all previous waves of CIS the statistical unit had been the legal unit.
CIS 2022 was the first innovation survey for which we had to report results at the enterprise level rather than at the level of legal units. As the innovation survey in Belgium is not conducted by the Belgian national statistical office (Statbel) we had to set up a legal construction that would allow Statbel to share enterprise information with our team. This legal construction was not yet in place when the sampling for CIS 2022 was done, so our sampling was still done at the level of legal units. All reporting units that were sent a survey form were legal units. Once the legal construction for cooperation with Statbel was in place, our sample was linked to the official business register maintained by Statbel. Sampled units that were deemed non-market oriented according to the official business register were removed from further processing. Statbel calculated weights at the enterprise level and calibrated those weights for number of enterprises, turnover and number of persons employed, whenever feasible.
The net sample in Belgium contained the following enterprises:
74% were enterprises consisting of 1 legal unit only;
15% were enterprises consisting of 2 legal units;
6% were enterprises consisting of 3 legal units;
5% were enterprises consisting of 4 or more legal units.
For 94% of these enterprises only 1 legal unit was sampled, for 4% of these enterprises 2 legal units were sampled, and for the remaining 2% of enterprises 3 or more legal units were sampled.
The set of responding enterprises contained the following:
For 95% of the responding enterprises a response was given by 1 legal unit only;
For 4% of the responding enterprises 2 legal units within the enterprise had responded;
For the remaining 1% of responding enterprises 3 or more legal units within each enterprise had responded.
For each enterprise Statbel identifies a privileged legal unit, usually this is the most important legal unit, that gives the enterprise its NACE code. For 97% of responding enterprises this privileged legal unit was among the responding legal units. Only for 3% of the responding enterprises only other legal units responded, not the privileged legal unit.
For a total of 216 enterprises responses given by 2 or more legal units needed to be aggregated to obtain one overall response for each enterprise. Not so surprisingly, enterprises for which such aggregation needed to be done predominantly were large and medium size enterprises, and only to a much smaller extent small enteprises: for only 1% of the responding small enterprises aggregation over multiple legal units needed to be done. For medium size enterprises the corresponding percentage was 7%, and for large responding enterprises aggregation over multiple legal units needed to be done for 22% of this set.
For aggregating responses over legal units within enterprises we used a pragmatic approach. We used all responses that were given and assumed additivity overall. When at least one legal unit within an enterprise indicated innovation, the overall response for the enterprise was innovation, etc. Expenditures were summed over legal units within an enterprise. Percentages, such as e.g., for share of turnover due to products new for the market, were first converted into absolute numbers (by multiplying the percentages with turnover) and then summed, and then again written as percentages for the enterprise as a whole.
Core target population is all enterprises in CORE NACE activities (see 3.3.1) with 10 or more employed persons (sum of employees and self-employed persons).
Belgium is composed of three regions (at NUTS 1 level): Brussels, Flanders, and Wallonia. Each region is endowed with its own statistical office. By law, CIS is conducted in a decentralized way in Belgium, within each of these three regions. A legal agreement is in place to make sure there is ample coordination between the regions to produce the required results for EUROSTAT reporting. By default, results for Belgium are available at the NUTS 1 level.
NUTS1 was used as a stratification dimension for the sampling.
The CIS2022 regional data are calibrated at the Statistical Unit Enterprise level.
For CIS 2022, the time covered by the survey is the 3-year period from the beginning of 2020 to the end of 2022.
Some questions and indicators refer to one year — 2022.
The list of indicators specifying whether they cover the 3-year period or refer to one year according to the HDC will be available in the Annex section of the European metadata (ESMS).
Accuracy in the statistical sense denotes the closeness of computations or estimates to the exact or true values. Statistics are not equal with the true values because of variability (the statistics change from implementation to implementation of the survey due to random effects) and bias (the average of the possible values of the statistics from implementation to implementation is not equal to the true value due to systematic effects).
CIS indicators are available according to 3 units of measure:
NR: Number for number of enterprises and number of persons employed.
THS_EUR: Thousands of euros. All financial variables are provided in thousands of euros, i.e. Turnover or Innovation expenditure.
PC: Percentage. The percentage is the ratio between the selected combinations of indicators.
We use the SAS routine that was originally made available by Eurostat for CIS 4 for data editing and data imputation. We updated this routine for use for CIS 2022. In this routine continuous variables are estimated using weighted ratio means, and nominal and ordinal variables are estimated using nearest neighbor hot deck imputation. More details are given in the document describing EUROSTAT’s methodological recommendations for CIS 2022, in section 5.5 and Annex 7.
In our data processing for CIS 2022 there were several changes compared to previous waves of CIS:
Results are now reported for enterprises. In all previous waves of CIS results were reported for legal units, a lower level of aggregation.
Employment numbers are now numbers for persons employed taken from the official business register, which include besides employees on payroll also owners, family workers and outworkers whose income is a function of the value of the outputs of the enterprise. In previous waves of CIS, employment numbers reflected number of employees only and were taken from balance sheets or from survey responses.
Turnover numbers are now those of the official business register and are at the enterprise level. In earlier waves of CIS turnover was reported at the level of legal units and was taken from balance sheets or from survey responses.
Survey weights are now calibrated for number of enterprises, turnover and number of persons employed, when feasible. In earlier waves of CIS survey weights were inverse sampling probabilities, corrected for nonresponse.
NACE codes used are now those of Belgium’s official business register, and are predominantly based on value added. In previous waves of CIS we did not have access to Belgium’s official business register. The NACE codes of the National Social Security Office were then used instead. These values were pre-filled in the survey forms, and respondents were able to indicate corrections if needed. The NACE codes of the National Social Security Office are predominantly employment-based.
Raw data are the responses given by enterprises to CIS 2022.
CIS is conducted and disseminated at two-year interval in pair years.
The timeliness of statistics reflects the length of time between data availability and the event or phenomenon they describe.
Comparability aims at measuring the impact of differences in applied statistical concepts and definitions on the comparison of statistics between geographical areas, non-geographical domains, or over time.
The coherence of statistical outputs refers to the degree to which the statistical processes by which they were generated used the same concepts (classifications, definitions, and target populations) and harmonised methods. Coherent statistical outputs have the potential to be validly combined and used jointly.
We use the same statistical concepts and definitions in all regions, and as far as we're aware, those are the same concepts and definitions used in the rest of the EU.
Following the publication of the 2018 Oslo Manual, from CIS 2018 onwards business process innovation was more broadly defined and now not only included technological process innovation, but also what before was called organizational and marketing innovation. The conceptual changes of the 2018 Oslo Manual implied big changes for the survey forms for CIS. Surprisingly, however, overall innovation rates remained to be fairly comparable over time: the overall innovation rates for Belgium were 68% and 68% in 2016 and 2018, respectively.
A big change for CIS 2022 was the required implementation of enterprises as statistical unit. In all previous waves of CIS, Belgium had used legal units as statistical unit. Again, surprisingly, overall innovation rates remained fairly stable despite the methodological change: when the CIS 2022 data were analyzed using the old methodology (with legal units as statistical unit) the overall innovation rate for Belgium was 70%. When the CIS 2022 data were analyzed using the new methodology (with enterprises as statistical units) the overall innovation rate for Belgium was also 70%. The fact that overall innovation rates are predominantly determined by small firms, given they are fairly numerous in the economy, and most of those small firms are enterprises consisting of one legal unit, might have something to do with the limited impact of the implemented methodological change. One change that did have some impact on more fine-grained results, however, was the fact that now for the first time CIS results were reported using the official NACE codes of Belgium’s official business register. Those NACE codes are predominantly based on value added. In earlier waves of CIS we did not have access to the official Belgian business register. We had to work with the register of active employers maintained by the Belgian National Social Security Office. The NACE codes in that register are predominantly employment based. So, some shifts between NACE codes may occur.
