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.
Structural business statistics (SBS) describes the structure, conduct and performance of economic activities, down to the most detailed activity level (several hundred economic sectors).
SBS covers all activities of the non-financial business economy with the exception of agricultural activities and personal services. Limited information is available on banking, insurance and pension funds.
Main characteristics (variables) of the SBS data category:
Business demographic variables (e.g. Number of enterprises)
"Output related" variables (e.g. Turnover, Value added)
"Input related" variables: labour input (e.g. Employment, Hours worked); goods and services input (e.g. Total of purchases); capital input (e.g. Material investments)
3.2. Classification system
Statistical Classification of Economic Activities in the European Community (NACE): NACE Rev.1 was used until 2001, NACE Rev. 1.1 since 2002, and NACE Rev 2 is used from 2008 onwards. Key data were double reported in NACE Rev.1.1 and NACE Rev.2 for 2008. From 2009 onwards, only NACE Rev.2 data are available.
The regional breakdown of the EU Member States is based on the Nomenclature of Territorial Units for Statistics (NUTS). Detailed information about the consecutive NUTS Regulations can be found at Eurostat's website
The SBS coverage was limited to Sections C to K of NACE Rev.1.1 until 2007. Starting from the reference year 2008 data is available for Sections B to N and Division S95 of NACE Rev.2. With 2013 as the first reference year information is published on NACE codes K6411, K6419 and K65 and its breakdown.
Tax data is used to calculate core variables such as turnover and total costs. The definitions are coherent, but there is a need to both edit/verify the data and recalculate some of the variables. There are also gaps in the data, and certain variables such as employment are received from other administrative sources or other internal sources. For variables that are more of the conceptual side, such as value added, we include information from a large sample survey. The Enterprise was implemented in the business register and has been used for calculations since 2003. However, the extent of the implementation is still limited and only a few large enterprises are treated correctly. The infrastructure for SBS is in place and when the new initiatives on profiling are finished and the new units are implemented in the register, the SBS deliverables will be compliant. It is not possible to correctly calculate production value and value added without more detailed information. This information is received via a sample survey. The effect is that although we have almost complete data on a micro level we cannot simply aggregate data and have implemented a top-down approach to our calculation. This means that we calculate on the actual level we are disseminating in order to get the best quality of our estimates on the highest levels. This is usually fine, but since the SBS deliverables include data on several NACE levels and other dimensions, size and region, this approach causes minor inconsistencies.
The definition of variable v12170 differs from the definition in the Regulation. The issue is how the variable v12170 has been translated into Swedish. The translation corresponds to a variable in standard Swedish accounting and therefore has a slight different content. Variable v12170 also includes personnel costs. For the data of 2018 and onwards SBS Sweden are changing the calculation of v12170 and will correspond to the definition in the Regulation.
3.5. Statistical unit
The population of interest consists of all Swedish units that carried out market activities in the relevant reference year. The Structural Business Statistics uses three different observation units: enterprise unit (ENT), kind of activity unit (KAU) and local kind of activity unit (LKAU).
SBS is limited to the units available in the Statistical Business Register. The complex Enterprise Units that are available in the business register are in conformity and are used by SBS. However, the number of complex units is not enough.
3.6. Statistical population
There are three target populations identified by the business register, namely:
Institutional: The amount of all non-financial Swedish enterprises that, in the reference year, conducted their main activities as indicated in sections A-S, excluding K and O, according to the Swedish Standard for Industrial Classification, SNI 2007.
Functional: The amount of all Swedish activities that, in the reference year, conducted activities as indicated in sections A-S, excluding K and O, according to the Swedish Standard for Industrial Classification, SNI 2007.
Regional (and functional): The amount of all Swedish regional activities that, in the reference year, conducted activities as indicated in section A-S, excluding K and O, according to the Swedish Standard for Industrial Classification, SNI 2007.
3.7. Reference area
Sweden. Branches of foreign enterprises are included.
3.8. Coverage - Time
1996 to 2020
With the transmission of data from reference year 2018, Sweden have fully implemented the statistical unit ‘enterprise’ in the SBS domain in accordance with Regulation 696/93. The transition from the ‘legal unit’ to the ‘enterprise’ has caused breaks in the series of the SBS characteristics, for example an enterprises consisting of several legal units will only be counted once and intraflows between the legal units within the enterprise will be eliminated (consolidated) in SBS statistics.
3.9. Base period
Not applicable.
Number of enterprises and number of local units are expressed in units.
Monetary data are expressed in thousands of SEK.
Employment variables are expressed in units.
Per head values are expressed in thousands of SEK. per head.
Ratios are expressed in percentages.
2020
The statistical target characteristics are attributed mainly to the calendar year. This applies to both target populations and to most target variables. Target variables that measure a state of things, such as balance sheet variables at institutional level, are exempt from this. These refer to the end of the reference year, that is, 31 December.
6.1. Institutional Mandate - legal acts and other agreements
With regard to confidentiality concerning the agency's specific task in the production of statistics, Chapter 24, Section 8 of the Public Access to Information and Secrecy Act (2009:400) applies.
To safeguard information subject to confidentiality belonging to natural persons or enterprises, it is ensured that the information cannot be disclosed directly or indirectly in the published statistics.
With regard to personal data, that is, information that directly or indirectly relates to a living person, the Official Statistics Act (2001:99), the Official Statistics Ordinance (2001:100) and the Personal Data Act (1998:204) apply.
7.2. Confidentiality - data treatment
Primary confidentiality: We calculate primary confidentiality with the P-Percent rule from 2014 and onwards. We have consent from some of the more important enterprises to publish aggregates that would otherwise be confidential.
Secondary confidentialiy: Generated by a version of TauArgus developed by Statistics Sweden
Futhermore we ask some of the largest enterprises if it is okay to publish the data in their activity. This operation is quite formal and so far only a few enterprises have agreed but this will be expanded upon. The goal is to reach around 30 enterprises that are complicated from a confidentiality perspective, which would allow for no confidentiality on a 3digit NACE level.
8.1. Release calendar
Dissemination follows the publishing calendar for the Official Statistics of Sweden. The dates of planned publications are found in the publishing calendar in advance on Statistics Sweden website.
8.2. Release calendar access
Dissemination follows the publishing calendar for the Official Statistics of Sweden.
8.3. Release policy - user access
We can provide further information which is not available in the dissemination. Everyone with a specific request, given no confidentiality issues, can access the data. Depending on the complexity of the request there may be a cost involved.
Annual
10.1. Dissemination format - News release
Regular press release to the data. The news release is published online at each dissemination.
10.2. Dissemination format - Publications
Regular press release to the data. The electronic publication can be found on SBS Sweden. The publication is available in English and Swedish.
10.3. Dissemination format - online database
The statistics are made accessible in the Statistical Database. All this is available via Statistics Sweden's website www.scb.se/NV0109, on which other tables and figures are also presented.
10.4. Dissemination format - microdata access
MONA (Microdata Online Access) is the standard tool for delivering microdata at Statistics Sweden. Users that get deliveries via MONA can process data over an internet connection without microdata leaving Statistics Sweden. The Public Access to Information and Secrecy Act protects data for individual respondents (microdata), but it is possible to apply for access to data for research or statistical purposes. When a request for data is granted, access to MONA may be given in order to receive the delivery.
10.5. Dissemination format - other
Data are sent to Eurostat to be used in European aggregates and is also released as national data.
10.6. Documentation on methodology
Documentation on methodology is available in Swedish.
The Quality policy includes quality of the statistics, processes and operations. It describes the overall guidelines for Statistics Sweden's quality work.
More information about Statistics Sweden's quality management can be found on the website.
11.2. Quality management - assessment
Sweden uses administrative data for the majority of the enterprises. We supplement the administrative data with surveys. Statistical uncertainty must be taken into account and work is carried out to reduce bias associated with sample surveys (outlier detection/editing/imputation etc). The quality of the statistics is generally considered to be high.
12.1. Relevance - User Needs
The Structural Business Statistics is primarily used by the National Accounts Department, which uses Structural Business Statistics as the basis for calculating the gross national product (GDP) on the production approach, and financial accounts.
Other users of the Structural Business Statistic, who would use the statistics as a basis for analyses of cost levels, productivity, industry development, and the calculation of key figures etc. are as follows:
Internal users:
- Other departments at Statistics Sweden
- The Ministry of Finance
- The National Institute of Economic Research
- Trade organisations, banks, accounting firms, media
- Enterprises/businesses
- Researchers and students
External users:
- Eurostat
- The Organisation for Economic Co-operation and Development (OECD)
- The United Nations
- Other international organisations
12.2. Relevance - User Satisfaction
Not available
12.3. Completeness
v15120 and v15130 are not available in Sweden. NACE H563 does not exist in Sweden by law.
13.1. Accuracy - overall
In no specific order, the most significant sources of errors are modelling errors, coverage errors and measurement errors.
13.2. Sampling error
Most of the estimates are based on administration data (standardised accounting statements SRU, collected from the Swedish Tax Agency), therefore there are no sampling error.
13.3. Non-sampling error
The non-sampling error is regarded to be small. Non-response in the SRU are compensated by mean value imputation with respect to industry and size. The measures taken to minimise the unit non-response is with the help of several postal reminders and telephone reminders for the larger enterprises. An evaluation of the questionnaire is carried out after each survey period. Also, under the Swedish Official Statistics Act (2001:99), information collected directly from enterprises is subject to obligation to provide information.
For the estimation methods, the bias is small with a limited effect in the overall accuracy of the estimate.
14.1. Timeliness
The production time, that is, the time between the endpoint of the reference time and point in time when the statistics are disseminated, is 16 months. The delivery to Eurostat is no more than 18 months.
14.2. Punctuality
No delay of transmission of data to Eurostat.
15.1. Comparability - geographical
There is a high degree of comparability across the different Swedish geographical domains.
15.2. Comparability - over time
Lenght of comparable time series: 1996 to 2020
NACE changes are making it difficult to compare different years. Furthermore the sampling procedure was improved between 2002 and 2003. For the main variables comparability is adequate.
15.3. Coherence - cross domain
Much of the differences between the SBS chacteristics and the same characteristics of other business related statistics are a consequence of different definitions in the respected regulations. The sources for the variables in the Business register differs to SBS.
There are two main differences between SBS and BD for ref year 2020.
Historically SBS and BD have always used slightly different frames, where SBS misses some units that have been active only parts of the reference year
These units will still be part of BD and is the main reason why SBS and BD have never been coherent.
The second and bigger issue for ref year 2020 is however the statistical units used, where SBS use the Enterprise (since ref year 2018) and BD still uses legal units.
Statistics Sweden have 50 000 Enterprises consisting of more than one legal unit which drastically affects the number of enterprises in the population.
We are working towards implementing common and coherent frames for all business statistics.
15.4. Coherence - internal
In the statistical sense yes, in numerical sense not always stictly consistent.
Due to rounding errors aggregates may not be fully consistent with sub-aggregates.
Regional aggregates may not be fully consistent with national aggregates due to missing NUTS-Code on a small number of enterprises.
The burden on respondents are calculated in hours. For year 2020 the burden amounted to 20 000 hrs. The calculation of respondents burden covers both the statistics for national requirements and SBS requirements.
17.1. Data revision - policy
If large errors are detected, revisions are made. Revisions are in this case made annually. For national purposes we implemented a revision policy from reference year 2014, revising data for t-1 when publishing year t.
17.2. Data revision - practice
We are using tax data delivered to Statistics Sweden in August t+1. Another delivery of tax data is done at January t+2 causing some differences between the preliminary and the definitive data. Also some editing is done after the preliminary data delivery.
18.1. Source data
The data source is statistical survey combined with administrative source. The administrative source is the administrative material standardised accounting statements (SRU), collected from the Swedish Tax Agency, which is the information submitted by enterprises in an appendix to the income tax declaration. The administration source contains micro data on 1 100 000 enterprises. SRU is obtained in two batches, in which the first delivery is obtained in August the year after the reference year, and the second delivery is in January the year after, that is, 13 months after the end of the reference year. The administrative material from the Swedish Tax Agency is considered to have high accuracy and should therefore cause only minor measurement errors. Accuracy is due to many observation variables being directly linked to specific accounts according to the current accounting manual BAS account plan. It is used as a main data source, basic data for some characteristics and for imputation in case of non-response. However, SRU is not used for the approximately 400 largest enterprises. Instead, there is a direct collection of data, the "complete form". The SRU data is subjected to several revisions with (increasing) degree of completeness.
The sample design is stratified and carried out by a sample. The stratification criteria is by activity and turnover size class. The selection schemes (sampling rates) are approximately 19 000 enterprises out of approximately 1 100 000 enterprises in the population. The relation between the reporting unit for the survey/administrative data and the enterprise is in 99.9% of the cases 1 to 1. No treshold values are being used.
The frame population is established in November of the reference year. It is delineated using a snapshot of Statistics Sweden's Business Register. To belong to the frame population, the enterprise must be active, according to the Business Register, which it is if it is registered as an employer, is included in the VAT register, or is registered for corporate tax. In addition to those enterprises that are classified as active in the Business Register, the following is added:
- Inactive enterprises holding properties with at least SEK 1 million in assessed value are included as property companies.
- Inactive enterprises that have at least one active subsidiary are included as holding enterprise. (A holding company or investment company is a company whose primary business consists of controlling other companies.)
The frame population and target population are largely consistent, but there are some situations that lead to shortcomings in the frame coverage, such as
- The enterprises registered during December are not included in the frame population. However, major enterprises are manually classified as active. The remaining part constitutes undercoverage.
- Enterprises that operate during the year, but for various reasons terminate operations before November are also not included in the frame population. However, major enterprises are also manually classified here as active. The remaining part constitutes undercoverage.
- Enterprises with out of date information in the Business Register regarding their activity status, due to a backlog in administrative work. Major enterprises are classified manually. Other enterprises constitute overcoverage and undercoverage, depending on whether they are active according to the Business Register, but are actually inactive, or vice versa.
18.2. Frequency of data collection
Annual data collection
18.3. Data collection
The methods used for the basic data collection from the reporting units is by electronic questionnaire. The electronic questionnaires has built-in quality and consistency checks. Direct collection is done via the complete form to about 400 enterprises, instead of using the standardised accounting statements material for these enterprises. This comprises, in principle, the 400 largest enterprises with regard to the previous year's production value. The purpose is to achieve sufficient accuracy when measuring the most important enterprises, which together comprise about one third of the total turnover and value added in the business sector, with regard to both basic and specification variables. Specifications of SRU are collected from the enterprises in the sample surveys. The objective is to give information on primarily detailed income and cost structures, on which the standardised accounting statements does not provide sufficient information. For the specification surveys, and in particular when detailed income and cost data are requested, the risk of measurement error is greater than in the standardised accounting statements material. This is the case even though the instructions and definitions, as far as possible, are adapted to applicable accounting standards. However, company reporting at such a detailed level may vary and sometimes specifications are required that could not be found in the accounting standard. This also applies for specification variables in the complete form. The measures taken to minimise the unit non-response is with the help of several postal reminders and telephone reminders for the larger enterprises. Under the Official Statistics Act (2001:99), information collected directly from enterprises is subject to obligation to provide information. Companies that do not submit information can be submitted with possible sanctions.
Administrative data: The method used to access administrative data (SRU) is by an extraction from a data base at the Swedish Tax Agency and delivered to Statistic Sweden.
18.4. Data validation
In the survey questionnaires there are built-in consistency checks. There are completeness checks (data integrity rules), validity checks (internal consistency) and plausibility checks, we also include time series checks and revisions. The checks are extensive. This is done before dissemination at NSI. Before dissemination and transmitting data validation of format and file structure checks are done. Inter-dataset checks are done only with the help of Eurostat Inputhall validating programme.
18.5. Data compilation
The estimation procedure contains adjustments for non-response, through the use of imputation. A brief description follows below of the use of imputation in basic and specification variables respectively:
Basic variables:
Automatic imputation using mean value imputation is used in most cases. Imputation groups are created by classifying the respondents by legal form, size and industry, where the level of detail of industry depends on the number of available responses.
Specification variables:
A combination of automatic and manual imputation is used for the specification variables. In automatic imputation, the structure of the enterprise's submitted information for the previous year, or mean value imputation, is used. Imputation groups are created by classifying the respondents by industry, where the level of detail of industry depends on the number of available responses. Expert imputation means that data is collected from different sources, such as enterprises' official annual reports.
18.6. Adjustment
We use the accounting year which cover the most of the reference period. If shorter or longer than 12 month we recalculate the figures to be equivalent to 12 month
No further comments
Structural business statistics (SBS) describes the structure, conduct and performance of economic activities, down to the most detailed activity level (several hundred economic sectors).
SBS covers all activities of the non-financial business economy with the exception of agricultural activities and personal services. Limited information is available on banking, insurance and pension funds.
Main characteristics (variables) of the SBS data category:
Business demographic variables (e.g. Number of enterprises)
"Output related" variables (e.g. Turnover, Value added)
"Input related" variables: labour input (e.g. Employment, Hours worked); goods and services input (e.g. Total of purchases); capital input (e.g. Material investments)
Tax data is used to calculate core variables such as turnover and total costs. The definitions are coherent, but there is a need to both edit/verify the data and recalculate some of the variables. There are also gaps in the data, and certain variables such as employment are received from other administrative sources or other internal sources. For variables that are more of the conceptual side, such as value added, we include information from a large sample survey. The Enterprise was implemented in the business register and has been used for calculations since 2003. However, the extent of the implementation is still limited and only a few large enterprises are treated correctly. The infrastructure for SBS is in place and when the new initiatives on profiling are finished and the new units are implemented in the register, the SBS deliverables will be compliant. It is not possible to correctly calculate production value and value added without more detailed information. This information is received via a sample survey. The effect is that although we have almost complete data on a micro level we cannot simply aggregate data and have implemented a top-down approach to our calculation. This means that we calculate on the actual level we are disseminating in order to get the best quality of our estimates on the highest levels. This is usually fine, but since the SBS deliverables include data on several NACE levels and other dimensions, size and region, this approach causes minor inconsistencies.
The definition of variable v12170 differs from the definition in the Regulation. The issue is how the variable v12170 has been translated into Swedish. The translation corresponds to a variable in standard Swedish accounting and therefore has a slight different content. Variable v12170 also includes personnel costs. For the data of 2018 and onwards SBS Sweden are changing the calculation of v12170 and will correspond to the definition in the Regulation.
The population of interest consists of all Swedish units that carried out market activities in the relevant reference year. The Structural Business Statistics uses three different observation units: enterprise unit (ENT), kind of activity unit (KAU) and local kind of activity unit (LKAU).
SBS is limited to the units available in the Statistical Business Register. The complex Enterprise Units that are available in the business register are in conformity and are used by SBS. However, the number of complex units is not enough.
There are three target populations identified by the business register, namely:
Institutional: The amount of all non-financial Swedish enterprises that, in the reference year, conducted their main activities as indicated in sections A-S, excluding K and O, according to the Swedish Standard for Industrial Classification, SNI 2007.
Functional: The amount of all Swedish activities that, in the reference year, conducted activities as indicated in sections A-S, excluding K and O, according to the Swedish Standard for Industrial Classification, SNI 2007.
Regional (and functional): The amount of all Swedish regional activities that, in the reference year, conducted activities as indicated in section A-S, excluding K and O, according to the Swedish Standard for Industrial Classification, SNI 2007.
Sweden. Branches of foreign enterprises are included.
2020
The statistical target characteristics are attributed mainly to the calendar year. This applies to both target populations and to most target variables. Target variables that measure a state of things, such as balance sheet variables at institutional level, are exempt from this. These refer to the end of the reference year, that is, 31 December.
In no specific order, the most significant sources of errors are modelling errors, coverage errors and measurement errors.
Number of enterprises and number of local units are expressed in units.
Monetary data are expressed in thousands of SEK.
Employment variables are expressed in units.
Per head values are expressed in thousands of SEK. per head.
Ratios are expressed in percentages.
The estimation procedure contains adjustments for non-response, through the use of imputation. A brief description follows below of the use of imputation in basic and specification variables respectively:
Basic variables:
Automatic imputation using mean value imputation is used in most cases. Imputation groups are created by classifying the respondents by legal form, size and industry, where the level of detail of industry depends on the number of available responses.
Specification variables:
A combination of automatic and manual imputation is used for the specification variables. In automatic imputation, the structure of the enterprise's submitted information for the previous year, or mean value imputation, is used. Imputation groups are created by classifying the respondents by industry, where the level of detail of industry depends on the number of available responses. Expert imputation means that data is collected from different sources, such as enterprises' official annual reports.
The data source is statistical survey combined with administrative source. The administrative source is the administrative material standardised accounting statements (SRU), collected from the Swedish Tax Agency, which is the information submitted by enterprises in an appendix to the income tax declaration. The administration source contains micro data on 1 100 000 enterprises. SRU is obtained in two batches, in which the first delivery is obtained in August the year after the reference year, and the second delivery is in January the year after, that is, 13 months after the end of the reference year. The administrative material from the Swedish Tax Agency is considered to have high accuracy and should therefore cause only minor measurement errors. Accuracy is due to many observation variables being directly linked to specific accounts according to the current accounting manual BAS account plan. It is used as a main data source, basic data for some characteristics and for imputation in case of non-response. However, SRU is not used for the approximately 400 largest enterprises. Instead, there is a direct collection of data, the "complete form". The SRU data is subjected to several revisions with (increasing) degree of completeness.
The sample design is stratified and carried out by a sample. The stratification criteria is by activity and turnover size class. The selection schemes (sampling rates) are approximately 19 000 enterprises out of approximately 1 100 000 enterprises in the population. The relation between the reporting unit for the survey/administrative data and the enterprise is in 99.9% of the cases 1 to 1. No treshold values are being used.
The frame population is established in November of the reference year. It is delineated using a snapshot of Statistics Sweden's Business Register. To belong to the frame population, the enterprise must be active, according to the Business Register, which it is if it is registered as an employer, is included in the VAT register, or is registered for corporate tax. In addition to those enterprises that are classified as active in the Business Register, the following is added:
- Inactive enterprises holding properties with at least SEK 1 million in assessed value are included as property companies.
- Inactive enterprises that have at least one active subsidiary are included as holding enterprise. (A holding company or investment company is a company whose primary business consists of controlling other companies.)
The frame population and target population are largely consistent, but there are some situations that lead to shortcomings in the frame coverage, such as
- The enterprises registered during December are not included in the frame population. However, major enterprises are manually classified as active. The remaining part constitutes undercoverage.
- Enterprises that operate during the year, but for various reasons terminate operations before November are also not included in the frame population. However, major enterprises are also manually classified here as active. The remaining part constitutes undercoverage.
- Enterprises with out of date information in the Business Register regarding their activity status, due to a backlog in administrative work. Major enterprises are classified manually. Other enterprises constitute overcoverage and undercoverage, depending on whether they are active according to the Business Register, but are actually inactive, or vice versa.
Annual
The production time, that is, the time between the endpoint of the reference time and point in time when the statistics are disseminated, is 16 months. The delivery to Eurostat is no more than 18 months.
There is a high degree of comparability across the different Swedish geographical domains.
Lenght of comparable time series: 1996 to 2020
NACE changes are making it difficult to compare different years. Furthermore the sampling procedure was improved between 2002 and 2003. For the main variables comparability is adequate.
