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.
National Statistical Institute of the Netherlands (Statistics Netherlands – CBS: Centraal Bureau voor de Statistiek)
1.2. Contact organisation unit
EBH Business Statistics (Heerlen)
1.3. Contact name
Confidential because of GDPR
1.4. Contact person function
Confidential because of GDPR
1.5. Contact mail address
Centraal Bureau voor de Statistiek Postbus 4481 6401 CZ HEERLEN The Netherlands
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
3 December 2025
2.2. Metadata last posted
3 December 2025
2.3. Metadata last update
3 December 2025
3.1. Data description
The Global Value Chains (GVC) statistics aim to monitor the economic globalization of enterprises by developing harmonised indicators and statistics.
Enterprises use GVCs to organize production worldwide. Complex processes are divided into smaller tasks carried out across multiple countries. By leveraging local advantages such as lower costs or specialized expertise, enterprises remain competitive and deliver goods and services more efficiently. A key feature of this process is the relocation of business functions abroad, referred to as International Sourcing (IS).
The GVC dataset contains ten variables defined in Implementing Regulation (EU) 2022/918. These cover:
(1) employment by business function
(2-5) enterprises purchasing or supplying goods and services abroad
(6) the number of enterprises carrying out international sourcing
(7-8) jobs created and lost resulting from international sourcing
(9) motivations and barriers for sourcing, and
(10) events affecting GVC arrangements.
For variables (2)-(5), the dataset includes only enterprises that report at least EUR 100 000 in goods or services purchased from or supplied abroad in the final year of the reference period.
NACE Rev.2. is used to classify economic activities for all variables.
Specific definitions are established by the Global Value Chains Task Force, composed of participating countries and chaired by Eurostat.
Business functions are created following the Manual on the Classification of Business Function and follow the harmonised list defined in the GVC implementing act, distinguishing core and support functions as well as specific functional categories such as production of goods, ICT services, marketing, engineering, and R&D.
Geographical classifications follow the breakdowns prescribed in Regulation (EU) 2022/918, including EU Member States, Extra-EU, and specific partner regions (e.g. China, India, USA and Canada, Central and South America, Africa).
Classifications for job skills follow the high-skilled / not-high-skilled distinction set out in the GVC Compilers’ Manual.
For the national implementation of the GVC statistics, Statistics Netherlands (CBS) translates these classifications or, where necessary, slightly adapts them to remain conceptually as close as possible to the original intention.
3.3. Coverage - sector
The statistics cover market producers classified under NACE Rev.2, Sections B to N. This sector coverage is defined in Implementing Regulation (EU) 2022/918 and applies uniformly to all variables. Only enterprises with 50 or more employees and self-employed persons in the final year of the reference period are included.
In the Dutch Standard Industrial Classification (SIC), each enterprise is assigned an SIC code consisting of five digits that indicate its main activity. The first four digits of the SIC code correspond almost entirely to the NACE classification, ensuring consistency between the European and national frameworks.
3.4. Statistical concepts and definitions
All concepts and definitions follow Implementing Regulation (EU) 2022/918 and the GVC Compilers’ Manual. International sourcing refers to the partial or full movement of business functions abroad within or outside the enterprise group, during 2021-2023. Variables (2)-(5) apply only to enterprises exceeding the EUR 100 000 threshold for goods/services purchased or supplied abroad. Business functions are divided into core and support categories and specific functional areas (e.g. ICT, management and administration). Motivations and barriers follow the fixed lists provided in the regulation, and importance factors use the standard four-level scale (very important, moderately important, not important, not applicable/do not know).
3.5. Statistical unit
The statistical unit used for the GVC statistics is the enterprise, as defined in Regulation (EU) 2019/2152 on European business statistics and the Business Register Regulation.
The definition of an enterprise is provided by the Statistical Unit Regulation (Council Regulation (EEC) No 696 / 93) as follows: "The enterprise is the smallest combination of legal units that is an organisational unit producing goods or services, which benefits from a certain degree of autonomy in decision-making, especially for the allocation of its current resources. An enterprise carries out one or more activities at one or more locations. An enterprise may be a sole legal unit."
The statistical units (enterprises) used for the sample are drawn directly from the General Business Register of Statistics Netherlands (CBS). Approximately 29% of the enterprises in the GVC sample for the reference period 2021-2023 correspond to a single legal unit, while 30% consist of six or more legal units. For the construction of the business register, information from the Chamber of Commerce (in cases of full ownership) and from the tax authorities (in other cases) is used to determine the ‘top’ legal unit within an enterprise. This unit also serves as the reporting unit for the responding enterprise.
3.6. Statistical population
For all variables, the enterprise population refers to market producers in the Netherlands that fall under Nomenclature of Economic Activities (NACE Rev.2) Sectors B to N (business economy). Enterprises are included if they had an average of at least 50 employed persons (employees and self‑employed) during the final year of the reference period. The average number of employed persons is calculated based on the months in which the unit was registered in the General Business Register of Statistics Netherlands (CBS) during that year.
Additionally, the statistical units of these market producers must have qualified at some point during the reference period as "enterprises" according to the definition in Council Regulation (EEC) No 696/93. These units must also have been active during the reference period. A statistical unit is considered active if it generated positive net turnover, produced output, employed staff, or made investments during that period. Furthermore, enterprises must still be active in the Netherlands at the end of the reference period.
Statistical units that are no longer active in the Netherlands at the end of the reference period, such as those that fully relocated their activities abroad during the period, are not observed in the survey. In total, the target population of the GVC statistics consists of approximately 11,500 enterprises per reference period.
3.7. Reference area
The reference area is the national economic territory of The Netherlands, as defined in Regulation (EU) 2019/2152 for European business statistics. Units located within the national territory and covered by national statistical legislation are included.
3.8. Coverage - Time
The GVC statistics refer to the triennial reference period 2021–2023, as defined in Implementing Regulation (EU) 2022/918. All data transmitted relate exclusively to this three-year reference period.
3.9. Base period
Not applicable. No base year or index is used, as the data are compiled as absolute values within the triennial reference period defined by Implementing Regulation (EU) 2022/918.
All variables in the GVC dataset are compiled and transmitted in absolute values, in accordance with the measurement unit specified in Implementing Regulation (EU) 2022/918. No percentages, indexes, ratios or other derived units are transmitted to Eurostat. This unit applies uniformly to all variables and all breakdowns required by the regulation.
The reference period for the GVC statistics is 2021-2023, in accordance with Implementing Regulation (EU) 2022/918. Variables cover activities that may have occurred during this three-year period. Variables (1)-(5) refer specifically to the final year of the period (2023). Variables (6)-(10) cover actions that took place at any time between 1 January 2021 and 31 December 2023.
6.1. Institutional Mandate - legal acts and other agreements
GVC statistics are produced under the legal framework of European business statistics. The main legal bases are:
Regulation (EU) 2019/2152 on European business statistics (EBS Regulation), the topic Global Value Chains is listed in Annex I.
Implementing Regulation (EU) 2022/918 which defines the variables, population, breakdowns and transmission deadlines for the 2021-2023 reference period.
Statistics Netherlands (CBS) is governed by the Statistics Netherlands Act. This Act establishes CBS’s legal mandate to produce official statistics in the Netherlands. It regulates its authority to collect data, ensures access to administrative sources, and sets out the obligation of enterprises to respond. It also safeguards CBS’s professional independence and defines its role as the national statistical authority.
Statistics Netherlands (CBS) guarantees the privacy of respondents (persons, enterprises and other), the confidential nature of the information provided and the sole use for statistical purposes.
This is captured in national legislation: the Statistics Netherlands Act of 20 November 2003. Section 37 of this Act stipulates that:
The data received by the director general in connection with the performance of his duties to implement this act shall be used solely for statistical purposes.
The data referred to in the first subsection shall not be provided to any persons other than those charged with carrying out the duties of the CBS.
The data referred to in the first subsection shall only be published in such a way that no recognisable data can be derived from them about an individual person, household, company or institution, unless, in the case of data relating to a company or institution, there are good reasons to assume that the company or institution concerned will not have any objections to the publication.
7.2. Confidentiality - data treatment
Statistics Netherlands (CBS) applies the confidentiality guidelines as described in its new Statistical Disclosure Manual: Handboek Statistische Beveiliging 2024(Only available in Dutch). The core principle is that the output of Statistics Netherlands must never lead to the identification of an individual respondent, either directly or indirectly. In statistical disclosure control, this is referred to as 'disclosure': the ability to trace a published value back to an identifiable unit.
Under the strict interpretation of the Statistics Netherlands Act and the General Data Protection Regulation (GDPR), it is irrelevant whether the information was already known to the “discloser” (e.g. via an annual report) or is newly revealed; the mere possibility of re-identification is decisive.
Confidentiality parameters are selected through a careful balance between:
The risk of disclosure (likelihood and degree of identifiability).
The impact on data usability and level of detail.
The specific parameters used to ensure confidentiality are not disclosed. Depending on the type of output table, the following data treatment guidelines apply.
Frequency tables:
For frequency tables (e.g. number of active enterprises), where cell values represent the count of units matching a combination of characteristics, primary confidentiality is ensured through the following checks:
Small cell counts: cells with fewer than a defined weighted number of units are flagged.
Group disclosure: arises when information about a small, identifiable group can be derived from published tables, even without direct identification of individuals. For example, if nearly all enterprises in a cell share the same characteristic, the group itself becomes recognizable.
These checks are performed manually and implemented through scripts. Before publication, frequency tables are reviewed for group disclosure risks according to three rules:
F1: Groups with fewer than three units must not be published.
F2: If a group has fewer than k units and at least 90% of them (minimum row total minus one) are concentrated in one category, the group must not be published. The value of k depends on the potential impact: k = 5 for negligible impact, and between 30 and 100 for sensitive information.
F3: Rule F2 also applies to all meaningful aggregates of other variable categories.
These rules are applied to unprotected tables, before any confidentiality methods such as rounding. Secondary protection is added manually when required and implemented via scripts.
Quantitative tables:
Primary confidentiality is ensured according to P% rule. This approach takes into account that a contributor to a publication group can use own data to make an estimation of competitors values. Furthermore, the weight of secondary confidentiality cells are minimised. The confidentiality treatment is performed using the Tau-Argus package available within Statistics Netherlands. Measures to reduce the number of confidential cells:
Use the P% rule.
Apply the same confidentiality pattern to strongly correlated indicators.
8.1. Release calendar
Data will be made available no later than 24 months after the end of the final year of the reference period (T+24). For the GVC statistics covering the reference period 2021–2023, this implies that the results will be published in the online database (StatLine) of Statistics Netherlands (CBS) by 31 December 2025 at the latest.
8.2. Release calendar access
Expected future updates of the data tables published in the online database (StatLine) of Statistics Netherlands (CBS) are specified in the explanatory notes accompanying each table.
Where applicable, planned publication dates for news articles related to the output of the GVC statistics will be announced in the Statistics Netherlands Publication Calendar.
According to this policy, data provided to third parties (e.g. Eurostat) must also be made available in the online database (StatLine).
Output is published when it concerns mandatory indicators and when it is expected that comparable data will continue to be collected or produced in the future by CBS, thereby enabling the creation of a consistent output series.
GVC statistics are collected, compiled and transmitted every three years in accordance with Implementing Regulation (EU) 2022/918. National dissemination, where applicable, follows the same triennial cycle and usually takes place after the transmission of data to Eurostat.
10.1. Dissemination format - News release
A short news release summarising the key results of the GVC statistics is published at the time of, or at a later stage following, the national dissemination.
10.2. Dissemination format - Publications
The results of the GVC statistics may also be used in more extensive national publications by Statistics Netherlands (CBS) such as long‑form statistical analyses or analytical reports, for example CBS longreads or research reports. These publications are typically longer, methodologically substantiated, and aimed at policymakers, researchers, and specialist audiences.
10.3. Dissemination format - online database
A selection of aggregated GVC indicators is published in the national online statistical database (StatLine) of Statistics Netherlands, under the theme "Enterprises > Globalization".
Results of the GVC statistics may also be presented in thematic reports, statistical yearbooks or public presentations organised by Statistics Netherlands (CBS).
10.6. Documentation on methodology
Methodological information is provided in the national metadata accompanying the GVC data transmission and in the GVC Compilers' Manual published by Eurostat.
In addition, documentation on the methodology of the Dutch GVC statistics is available in the Brief Survey Descriptions section of the Statistics Netherlands (CBS) website, under Global Value Chains.
Statistics Netherlands (CBS) follows the Code of Practice for European Statistics. More information on quality assurance is available on the website: Quality | CBS.
11.2. Quality management - assessment
Overall data quality is assessed through evaluations of relevance, accuracy, coherence and completeness. The General Business Register and the profiling of enterprise units by Statistics Netherlands (CBS) are assessed to be of high quality. Response rates are maximized through reminder procedures and mandatory completion of web questionnaire items. Automated validation and consistency checks are applied. Completed submissions are assigned plausibility scores, but interventions are limited due to the lack of reliable reference sources.
Planned improvements include simplification of language and explanatory text, expanded use of microdata linking methods where quality permits, and intensified cooperation with the Large Cases Unit (LCU) team to further validate enterprise responses, particularly for the sourcing-related questions. Additionally, earlier initiation and improvement of follow‑up calls with enterprises to validate responses will be looked into. This process is currently time‑consuming and conceptually demanding, as it often requires clarifying definitions and requesting resubmission of corrected questionnaires in the absence of reliable reference data.
12.1. Relevance - User Needs
The GVC statistics provide policymakers and researchers with valuable insights into how globalization influences the Dutch economy. It supports evidence-based decision-making in areas such as international trade, labour market developments, and economic growth. The main stakeholder groups therefore include national ministries (e.g. economic affairs), research institutes, academic communities, media organisations, Eurostat, and enterprises themselves.
Stakeholders are particularly interested in the sourcing of business functions: which functions are relocated abroad, the motivations behind these decisions, and the barriers that enterprises face. GVC statistics therefore offer an essential perspective on sourcing structures, the international fragmentation of production, and sectoral exposure to global shocks, while also informing policies on supply chain resilience and strategic dependencies.
12.2. Relevance - User Satisfaction
No formal user feedback on stakeholder satisfaction has been collected by Statistics Netherlands (CBS). However, informal interactions (e.g. via email) provide some indication of user needs. An emerging area of interest concerns the complete relocation or full exit of enterprises from the Netherlands, with ministries and journalists inquiring whether this topic is also covered by the GVC statistics. At present, it is not, as enterprises must still be active at the end of the reference period to be included. In addition, stakeholders such as ministries have ussuallty shown an interest in backsourcing/ the general movement of business functions back to the Netherlands from abroad.
12.3. Completeness
All variables and breakdowns mandated under Implementing Regulation (EU) 2022/918 were delivered to Eurostat for the 2021–2023 reference period. The 1 percent rule was not applied. The dataset includes all required combinations of NACE sections, enterprise size classes, business functions, and geographical areas as outlined in the regulation.
13.1. Accuracy - overall
Overall accuracy is assessed through checks on coverage, coherence, and consistency. Most enterprises in the sample seem to provide usable data, and validation procedures reduce the impact of inconsistent responses. Nevertheless, some uncertainty remains for detailed variables, particularly those affected by respondents’ interpretation of complex breakdowns or requiring subjective assessments such as sourcing motives and barriers.
The smaller the subgroup of enterprises to which a variable refers (e.g. the international sourcing variables), the greater the inaccuracy of the estimates. As the statistics are based on a sample survey, they inevitably differ from true population values due to variability and potential bias. Procedures are in place to minimise these errors, and the survey design aims to ensure that results are as representative as possible of the enterprise population, while acknowledging the limitations inherent in survey‑based statistics.
13.2. Sampling error
The outcomes are primarily based on sample‑based research, and sampling error therefore affects the accuracy of the estimates. The stratified sampling designs ensure adequate representation of size classes and key industries. The stratification and weighting procedures are designed to support representativeness across NACE sections B–N and enterprises with 50 or more persons employed.
The accuracy of the results is highest when they refer to the total population of enterprises. However, a large share of the outcomes concerns the subgroup of enterprises that have sourced business functions abroad or have considered doing so. The results for these subgroups are less accurate, as only a relatively small number of enterprises in the Netherlands belonging to the GVC target population are involved. For indicators related to international sourcing, the standard error (SE) is therefore calculated. When the relative standard error (RSE) exceeds 30%, the corresponding cell is flagged as “low reliability”.
13.3. Non-sampling error
Non-sampling errors constitute the main potential source of inaccuracy in the GVC statistics:
Unit non‑response: Negligible. Because the GVC statistics now fall under the EBS regulation, participation is mandatory for enterprises in the Netherlands. As a result, the response rate is exceptionally high (92%), compared with 70% in the last voluntary survey conducted in 2021
Item non‑response: Does not occur. Respondents must answer all items in the web questionnaire to proceed. A potential risk is that enterprises may quickly select “no” to continue, which should be concidered a misreporting rather than item non‑response.
Misreporting: The respondent burden is high, as the administration of enterprises are not always aligned with the survey concepts and questions can be complex. Mitigation measures include clear questionnaires, guidance notes, built‑in consistency and logic checks, within the web survey. Completed submissions are assigned plausibility scores, but interventions are currently limited. Around 30 enterprises (<1% of responses) were contacted for follow-up validation, but this process is currently time-consuming and challenging due to the lack of reliable reference sources. In addition, some information in the GVC statistics may be perceived as sensitive (e.g. impact of sanctions on Russia). This can increase the risk of selective non‑response or cautious reporting, potentially leading to an underestimation of certain phenomena.
Coverage errors: The completeness of the General Business Register is considered very good, so coverage errors are negligible. Minor issues may arise from timing differences between the reference year and the timing of when the survey is implemented. For the indicator on employment by business function, results depend on the quality of secondary sources such as the dutch Labour Force Survey (LFS), administrative data from the Employment Insurance Agency (UWV), and Eurostat correspondence tables.
Processing errors: Mistakes may occur during data entry, coding, or editing. Automated validation rules, scripts, version control, and double‑checking procedures are applied to minimise such errors.
Overall, non-sampling errors have a moderate but manageable effect on the final results, and established validation and imputation procedures help to minimise their impact.
14.1. Timeliness
Data and metadata were transmitted to Eurostat within the legal deadline of T+21 months.
14.2. Punctuality
The data is published on time.
15.1. Comparability - geographical
The GVC statistics cover enterprises registered in the Netherlands. No regional breakdown within the country is required. Consequently, the same concepts and definitions apply uniformly across all Dutch regions, and there is no evidence that regional differences influence or complicate the interpretation of the results.
Furthermore, the data are comparable across EU Member States. Geographical comparability is ensured through harmonised definitions, classifications, and breakdowns as specified in the GVC Implementing Act. Methodological guidance from Eurostat provides a common framework across the EU. While variations may exist at national level, these do not materially affect overall comparability.
Statistics Netherlands (CBS) participated in all four voluntary pilot GVC surveys carried out in 2007, 2011, 2018, and 2021 prior to the introduction of the EBS framework. Broad conceptual continuity exists between these earlier exercises and the current official collection, as key definitions, business function categories, and geographical breakdowns are based on the same methodological foundations. Nevertheless, differences in survey design, sampling methods, questionnaire content, and follow‑up procedures may limit full comparability with the earlier voluntary results For example, the first survey covered the period 2001–2006. In addition, the samples for the first and second surveys included only enterprises with more than 100 employees, whereas the third survey targeted enterprises with more than 50 employees, with the enterprise group serving as the statistical and reporting unit in the Dutch national implementation of the third pilot survey. The variables on the sourcing of business functions abroad in the fourth survey are broadly comparable to those in the 2021–2023 reference period.
15.3. Coherence - cross domain
TheGeneral Business Register of Statistics Netherlands (CBS) serves as the basis for defining the target population of the GVC statistics. From this register, the enterprise unit is derived, together with other "coordinated enterprise characteristics" such as the Dutch Standard Industrial Classification (SIC), and the size class for persons employed. These coordinated characteristics are also used across by other business statistics for sampling, data weighting, and the calculation of population totals. Consequently, the business register provides the base for the total number of enterprises. For GVC statistics, however, it should be noted that the size class for persons employed is determined using the average number of employed persons recorded in the business register. The target population therefore consists of enterprises for the covered sectors that employed, on average, at least 50 persons in the final year of the reference period.
The output variable “number of employed persons” is produced by combining microdata from the Dutch Labour Force Survey (LFS), administrative data from the Employment Insurance Agency (UWV), and official Eurostat correspondence tables. The weights for the yearly figures derived from the first wave of the Dutch LFS are applied to gross up the data to population totals. These weights are aligned with the employment figures available in administrative registers, ensuring consistency across sources.
Furthermore, related business statistics include SBS, FATS, TEC, and STEC. Differences between these domains may occur due to varying target populations, reference periods, reporting thresholds, or domain-specific definitions. Nevertheless, such discrepancies do not affect the interpretation of GVC results.
15.4. Coherence - internal
Internal coherence is ensured through validation checks and survey design routing, which safeguard consistency across activity, size class, business function, and geographical breakdowns. Logical rules are applied to verify relationships between variables, and follow‑up procedures address any inconsistencies. For instance, enterprises reporting the sourcing of business functions abroad are included in the corresponding business function and geographical breakdowns. Finally, due to the rounding of weights when computing population totals, reported figures may not always correspond exactly to the sum of their components.
The exact costs associated with the production of the GVC statistics are not available.
The overall response burden is assessed as high but manageable. The survey is limited to enterprises with 50 or more persons employed, and several variables can be supported by business register information or pre‑filled where available. However, many enterprises indicate difficulties in relating to the topics covered. This leads to a high perceived burden and occasional frustration among respondents. Manufacturing enterprises generally find the questionnaire more relevant, whereas for other sectors broad applicability remains challenging. Based on feedback, the average completion time was approximately 48 minutes (this is quite high). Burden‑reduction measures include improvements in questionnaire design, the use of clearer language, enhanced guidance materials, and, where possible, the reuse of administrative and secondary data sources through microdata‑linking (MDL) methods. These efforts aim to lessen respondent burden while simultaneously enhancing data quality.
17.1. Data revision - policy
There is no special revision policy in effect at the moment for the GVC statistics.
17.2. Data revision - practice
All data variables are transmitted to Eurostat with definitive status within the legal deadline of T+21 months, except for dynamic variable 10, which was voluntarily transmitted earlier at T+15, at Eurostat’s request, under provisional status and subsequently updated to definitive status when all other variables were delivered.
For larger method shifts or detection of errors, the figures may be revised. Major errors will be corrected at national level right after detection and sent to Eurostat. All revisions wil lbe documented internally and communicated to Eurostat.
At the end of November (T+23), the standard error (SE) for the national publicationwas calculated for the variables related to international sourcing. Based on relative standard error (RSE), unreliability flags were updated, and the datasets were re‑transmitted to Eurostat. In addition, for the employment per business function indicator, four cells were suppressed to ensure compliance with the Dutch Labour Force Survey (LFS) publication guidelines, and this dataset was likewise re‑transmitted.
18.1. Source data
The main sources for the GVC statistics are:
General BusinessRegister: Used to define the target population for the sampling frame and to retrieve relevant enterprise characteristics. It provides core information on economic activity and size class. Register information may also support validation or help reduce respondent burden where appropriate.
GVC Survey: A dedicated survey is conducted among enterprises in the GVC target population to collect detailed information on global value chain activities and international sourcing. From this population, a stratified sample is drawn using the NACE (aggregated 2‑digit division), enterprise size class (having fewer or more than 250 employed persons), and multinational status as stratification criteria. Enterprises classified as "large enterprises" by the Large Cases Unit (LCU) team are fully covered in the survey sample. The survey sample includes up to 4,200 enterprises per reference period.
Dutch Labour Force Survey (LFS): The LFS is a rotating panel survey on labour participation in the Netherlands. It contains information on the occupations of respondents (persons), classified according to the International Standard Classification of Occupations (ISCO). These data allow employment figures to be estimated per business function. Only the data from the first wave of the LFS are used for the GVC statistics.
Administrative data from the Employment Insurance Agency (UWV): The UWV administrative data provides information on employees and enterprises. From these data, the primary employer of each worker in the LFS can be derived. When enterprises in the UWV administrative registers are linked to the General Business Register, the main and secondary NACE codes of enterprises make it possible to determine the core business function of each enterprise in the target population.
Eurostat’s correspondence tables: These tables support microdata‑linking (MDL) methods. They enable the estimation of employment per business function by linking the NACE codes of enterprises in the UWV administrative registers to business functions, and by connecting ISCO occupational codes from employed persons in the LFS to business functions.
Furthermore, data from other related domain statistics may be used as reference or auxiliary variables:
Multinational List (MNE table): Contains information on which enterprise units are part of a multinational group. To enrich the GVC target population and enables a more accurate and representative sample design.
Ultimate Controlling Institute (UCI) list from FATS statistics: Provides information on the control structures of enterprises, enriching the GVC target population for statistical analysis of foreign ownership.
18.2. Frequency of data collection
The data for the GVC statistics are collected every three years in accordance with Implementing Regulation (EU) 2022/918. The most recent data collection covered the reference period 2021-2023 and was carried out once during 2024.
18.3. Data collection
Data for the GVC statistics are primarily collected through a secure electronic webquestionnaire (CAWI: Computer Assisted Web Interview) developed with Blaise and sent to enterprises. In the Netherlands, respondents know this as the "Your Enterprise in International Value Chains" survey. Participation in the survey is mandatory. Invitation letters are sent in June, with an official return deadline set for mid‑August. Enterprises receive two reminder letters, and after the reminder period expires, enforcement measures are applied: a small share of enterprises in the sample are then flagged for non‑response, with the possibility of sanctions if they fail to respond.The CAWI remains open for one year (to allow enterprises under enforcement to submit their response), however, the majority of responses are typically received between June and September in the year in which the survey is conducted.
In addition, secondary data sources are used to produce indicators through microdata linking (MDL), which helps to reduce respondent burden and improve quality. For example, figures on employed persons by business function are derived by combining microdata from the Dutch Labour Force Survey (LFS), administrative data from the Employment Insurance Agency (UWV), and Eurostat correspondence tables.
18.4. Data validation
Validation procedures include automated checks for completeness, internal consistency, logical coherence, and compliance with the definitions and breakdowns specified in the GVC Implementing Act. Validation is performed at both the microdata and aggregated levels, and incorporates feedback from Eurostat’s validation process.
Data validation is primarily based on built‑in procedures within the web survey (CAWI: Computer Assisted Web Interview). These include automatic consistency checks, which significantly reduce the need for manual review and correction of most variables.
Manual review is carried out for flagged cases and detected inconsistencies. Completed submissions are assigned plausibility scores based on logical validation rules, the weight of each rule, and their impact on the output.This ensures that the most important enterprises are reviewed first. However, interventions remain limited due to the lack of reliable reference sources and because follow‑up procedures are time‑consuming and challenging within the context of this survey.
At the aggregated level (macro level), data are examined by NACE section and size class to identify and document major shifts compared to the previous cycle. Since the prior survey was an experimental pilot, larger deviations were expected in this round. Where shifts occur, the underlying microdata are reviewed to explain and, if necessary, correct them. The output of mandatory core indicators and breakdowns is also visualized and compared with previous cycles to assess whether developments align with expectations.
The output series themselves are checked for internal consistency. Validation rules embedded in the processing scripts ensure that all expected breakdown combinations are present. When a combination is missing, the case is flagged and reviewed to determine whether the issue stems from a script error or whether the value is genuinely absent (not observed in the survey).
18.5. Data compilation
GVC survey:
For the GVC sample survey, data compilation follows standard procedures. Each responding enterprise receives a grossing-up weight. Initial inclusion weights are calculated as the inverse of the selection probability and scaled to the target population. To correct for non‑response and improve accuracy, reweighting is applied using auxiliary information reflecting stratification by NACE (output level: sections), enterprise size class (fewer or more than 250 employed persons), and multinational status. Final weights ensure reliable population totals per publication cell. Aggregates are produced according to the required breakdowns, after which confidentiality rules are applied and outputs are checked for errors, correct disclosure control, and internal consistency. Data are rounded to whole numbers only at the final stage of processing.
Microdata linking method - employment per business function indicator:
Employment per (GVC) business function is estimated by combining microdata from the Dutch Labour Force Survey (LFS), administrative data from the Employment Insurance Agency (UWV), and official Eurostat correspondence tables.
From the UWV administrative data, resident persons are linked to their primary employer, restricted to enterprises in the GVC population, and classified using the NACE–CBF correspondence table for core business functions (CBF). From the LFS, the employed population (15–75 years, ILO definition) is filtered and linked to business functions via the ISCO–CBF table, excluding non‑relevant occupations for the GVC statistics. The two sources are merged, with corrections for temporary agency workers and for distinguishing core versus support business functions. Core functions are identified when the occupation‑based business function of a person matches the enterprise’s main business function; otherwise, the occupation is classified as support.
Final aggregated employment figures per business function are produced using LFS annual weights of the first wave, adjusted for agency worker redistribution and reclassification of core/support functions. Reclassification is applied when ISCO occupations predominantly occur in a sector but are mapped to a support function in the correspondence table, ensuring they are nevertheless counted as core based on their prevalence.
Since the number of employed persons is grossed up to the total population using LFS annual weights of the first wave, the publication guidelines of the Dutch LFS for the first wave apply. Variable breakdowns for NACE sections with fewer than 30,000 employed persons are therefore flagged as "low reliability” and set to “confidential”.
18.6. Adjustment
Not applicable. No seasonal or trend adjustments are applied, as the GVC statistics reflect structural information collected every three years.
No further comments.
The Global Value Chains (GVC) statistics aim to monitor the economic globalization of enterprises by developing harmonised indicators and statistics.
Enterprises use GVCs to organize production worldwide. Complex processes are divided into smaller tasks carried out across multiple countries. By leveraging local advantages such as lower costs or specialized expertise, enterprises remain competitive and deliver goods and services more efficiently. A key feature of this process is the relocation of business functions abroad, referred to as International Sourcing (IS).
The GVC dataset contains ten variables defined in Implementing Regulation (EU) 2022/918. These cover:
(1) employment by business function
(2-5) enterprises purchasing or supplying goods and services abroad
(6) the number of enterprises carrying out international sourcing
(7-8) jobs created and lost resulting from international sourcing
(9) motivations and barriers for sourcing, and
(10) events affecting GVC arrangements.
For variables (2)-(5), the dataset includes only enterprises that report at least EUR 100 000 in goods or services purchased from or supplied abroad in the final year of the reference period.
All concepts and definitions follow Implementing Regulation (EU) 2022/918 and the GVC Compilers’ Manual. International sourcing refers to the partial or full movement of business functions abroad within or outside the enterprise group, during 2021-2023. Variables (2)-(5) apply only to enterprises exceeding the EUR 100 000 threshold for goods/services purchased or supplied abroad. Business functions are divided into core and support categories and specific functional areas (e.g. ICT, management and administration). Motivations and barriers follow the fixed lists provided in the regulation, and importance factors use the standard four-level scale (very important, moderately important, not important, not applicable/do not know).
The statistical unit used for the GVC statistics is the enterprise, as defined in Regulation (EU) 2019/2152 on European business statistics and the Business Register Regulation.
The definition of an enterprise is provided by the Statistical Unit Regulation (Council Regulation (EEC) No 696 / 93) as follows: "The enterprise is the smallest combination of legal units that is an organisational unit producing goods or services, which benefits from a certain degree of autonomy in decision-making, especially for the allocation of its current resources. An enterprise carries out one or more activities at one or more locations. An enterprise may be a sole legal unit."
The statistical units (enterprises) used for the sample are drawn directly from the General Business Register of Statistics Netherlands (CBS). Approximately 29% of the enterprises in the GVC sample for the reference period 2021-2023 correspond to a single legal unit, while 30% consist of six or more legal units. For the construction of the business register, information from the Chamber of Commerce (in cases of full ownership) and from the tax authorities (in other cases) is used to determine the ‘top’ legal unit within an enterprise. This unit also serves as the reporting unit for the responding enterprise.
For all variables, the enterprise population refers to market producers in the Netherlands that fall under Nomenclature of Economic Activities (NACE Rev.2) Sectors B to N (business economy). Enterprises are included if they had an average of at least 50 employed persons (employees and self‑employed) during the final year of the reference period. The average number of employed persons is calculated based on the months in which the unit was registered in the General Business Register of Statistics Netherlands (CBS) during that year.
Additionally, the statistical units of these market producers must have qualified at some point during the reference period as "enterprises" according to the definition in Council Regulation (EEC) No 696/93. These units must also have been active during the reference period. A statistical unit is considered active if it generated positive net turnover, produced output, employed staff, or made investments during that period. Furthermore, enterprises must still be active in the Netherlands at the end of the reference period.
Statistical units that are no longer active in the Netherlands at the end of the reference period, such as those that fully relocated their activities abroad during the period, are not observed in the survey. In total, the target population of the GVC statistics consists of approximately 11,500 enterprises per reference period.
The reference area is the national economic territory of The Netherlands, as defined in Regulation (EU) 2019/2152 for European business statistics. Units located within the national territory and covered by national statistical legislation are included.
The reference period for the GVC statistics is 2021-2023, in accordance with Implementing Regulation (EU) 2022/918. Variables cover activities that may have occurred during this three-year period. Variables (1)-(5) refer specifically to the final year of the period (2023). Variables (6)-(10) cover actions that took place at any time between 1 January 2021 and 31 December 2023.
Overall accuracy is assessed through checks on coverage, coherence, and consistency. Most enterprises in the sample seem to provide usable data, and validation procedures reduce the impact of inconsistent responses. Nevertheless, some uncertainty remains for detailed variables, particularly those affected by respondents’ interpretation of complex breakdowns or requiring subjective assessments such as sourcing motives and barriers.
The smaller the subgroup of enterprises to which a variable refers (e.g. the international sourcing variables), the greater the inaccuracy of the estimates. As the statistics are based on a sample survey, they inevitably differ from true population values due to variability and potential bias. Procedures are in place to minimise these errors, and the survey design aims to ensure that results are as representative as possible of the enterprise population, while acknowledging the limitations inherent in survey‑based statistics.
All variables in the GVC dataset are compiled and transmitted in absolute values, in accordance with the measurement unit specified in Implementing Regulation (EU) 2022/918. No percentages, indexes, ratios or other derived units are transmitted to Eurostat. This unit applies uniformly to all variables and all breakdowns required by the regulation.
GVC survey:
For the GVC sample survey, data compilation follows standard procedures. Each responding enterprise receives a grossing-up weight. Initial inclusion weights are calculated as the inverse of the selection probability and scaled to the target population. To correct for non‑response and improve accuracy, reweighting is applied using auxiliary information reflecting stratification by NACE (output level: sections), enterprise size class (fewer or more than 250 employed persons), and multinational status. Final weights ensure reliable population totals per publication cell. Aggregates are produced according to the required breakdowns, after which confidentiality rules are applied and outputs are checked for errors, correct disclosure control, and internal consistency. Data are rounded to whole numbers only at the final stage of processing.
Microdata linking method - employment per business function indicator:
Employment per (GVC) business function is estimated by combining microdata from the Dutch Labour Force Survey (LFS), administrative data from the Employment Insurance Agency (UWV), and official Eurostat correspondence tables.
From the UWV administrative data, resident persons are linked to their primary employer, restricted to enterprises in the GVC population, and classified using the NACE–CBF correspondence table for core business functions (CBF). From the LFS, the employed population (15–75 years, ILO definition) is filtered and linked to business functions via the ISCO–CBF table, excluding non‑relevant occupations for the GVC statistics. The two sources are merged, with corrections for temporary agency workers and for distinguishing core versus support business functions. Core functions are identified when the occupation‑based business function of a person matches the enterprise’s main business function; otherwise, the occupation is classified as support.
Final aggregated employment figures per business function are produced using LFS annual weights of the first wave, adjusted for agency worker redistribution and reclassification of core/support functions. Reclassification is applied when ISCO occupations predominantly occur in a sector but are mapped to a support function in the correspondence table, ensuring they are nevertheless counted as core based on their prevalence.
Since the number of employed persons is grossed up to the total population using LFS annual weights of the first wave, the publication guidelines of the Dutch LFS for the first wave apply. Variable breakdowns for NACE sections with fewer than 30,000 employed persons are therefore flagged as "low reliability” and set to “confidential”.
The main sources for the GVC statistics are:
General BusinessRegister: Used to define the target population for the sampling frame and to retrieve relevant enterprise characteristics. It provides core information on economic activity and size class. Register information may also support validation or help reduce respondent burden where appropriate.
GVC Survey: A dedicated survey is conducted among enterprises in the GVC target population to collect detailed information on global value chain activities and international sourcing. From this population, a stratified sample is drawn using the NACE (aggregated 2‑digit division), enterprise size class (having fewer or more than 250 employed persons), and multinational status as stratification criteria. Enterprises classified as "large enterprises" by the Large Cases Unit (LCU) team are fully covered in the survey sample. The survey sample includes up to 4,200 enterprises per reference period.
Dutch Labour Force Survey (LFS): The LFS is a rotating panel survey on labour participation in the Netherlands. It contains information on the occupations of respondents (persons), classified according to the International Standard Classification of Occupations (ISCO). These data allow employment figures to be estimated per business function. Only the data from the first wave of the LFS are used for the GVC statistics.
Administrative data from the Employment Insurance Agency (UWV): The UWV administrative data provides information on employees and enterprises. From these data, the primary employer of each worker in the LFS can be derived. When enterprises in the UWV administrative registers are linked to the General Business Register, the main and secondary NACE codes of enterprises make it possible to determine the core business function of each enterprise in the target population.
Eurostat’s correspondence tables: These tables support microdata‑linking (MDL) methods. They enable the estimation of employment per business function by linking the NACE codes of enterprises in the UWV administrative registers to business functions, and by connecting ISCO occupational codes from employed persons in the LFS to business functions.
Furthermore, data from other related domain statistics may be used as reference or auxiliary variables:
Multinational List (MNE table): Contains information on which enterprise units are part of a multinational group. To enrich the GVC target population and enables a more accurate and representative sample design.
Ultimate Controlling Institute (UCI) list from FATS statistics: Provides information on the control structures of enterprises, enriching the GVC target population for statistical analysis of foreign ownership.
GVC statistics are collected, compiled and transmitted every three years in accordance with Implementing Regulation (EU) 2022/918. National dissemination, where applicable, follows the same triennial cycle and usually takes place after the transmission of data to Eurostat.
Data and metadata were transmitted to Eurostat within the legal deadline of T+21 months.
The GVC statistics cover enterprises registered in the Netherlands. No regional breakdown within the country is required. Consequently, the same concepts and definitions apply uniformly across all Dutch regions, and there is no evidence that regional differences influence or complicate the interpretation of the results.
Furthermore, the data are comparable across EU Member States. Geographical comparability is ensured through harmonised definitions, classifications, and breakdowns as specified in the GVC Implementing Act. Methodological guidance from Eurostat provides a common framework across the EU. While variations may exist at national level, these do not materially affect overall comparability.
Statistics Netherlands (CBS) participated in all four voluntary pilot GVC surveys carried out in 2007, 2011, 2018, and 2021 prior to the introduction of the EBS framework. Broad conceptual continuity exists between these earlier exercises and the current official collection, as key definitions, business function categories, and geographical breakdowns are based on the same methodological foundations. Nevertheless, differences in survey design, sampling methods, questionnaire content, and follow‑up procedures may limit full comparability with the earlier voluntary results For example, the first survey covered the period 2001–2006. In addition, the samples for the first and second surveys included only enterprises with more than 100 employees, whereas the third survey targeted enterprises with more than 50 employees, with the enterprise group serving as the statistical and reporting unit in the Dutch national implementation of the third pilot survey. The variables on the sourcing of business functions abroad in the fourth survey are broadly comparable to those in the 2021–2023 reference period.
