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 quarterly data are broken down by economic activity (at section level) in accordance with the current NACE Rev. 2 - Statistical classification of economic activities in the European Community.
2.3. Coverage - sector
Sections A-S, according to NACE Rev. 2, enterprises with one or more employees for private and public sector.
2.4. Statistical concepts and definitions
A 'job vacancy' is defined as a paid post that is newly created, unoccupied, or about to become vacant:
for which the employer is taking active steps and is prepared to take further steps to find a suitable candidate from outside the enterprise concerned; and
which the employer intends to fill either immediately or within a specific period of time.
‘Active steps to find a suitable candidate’ include:
notifying the job vacancy to the public employment services,
contacting a private employment agency/head hunters,
advertising the vacancy in the media (for example internet, newspapers, magazines),
advertising the vacancy on a public notice board,
approaching, interviewing or selecting possible candidates/potential recruits directly,
approaching employees and/or personal contacts,
using internships.
'Specific period of time’ refers to the maximum time the vacancy is open and intended to be filled. That period shall be unlimited; all vacancies for which active steps are continuing on the reference date shall be reported.
An 'occupied post’ means a paid post within the organisation to which an employee has been assigned.
2.5. Statistical unit
Q1: Private sector: local unit; public sector (including non-profit institutions serving households): enterprise (legal unit).
Q2-Q4: The job vacancy statistics are based on data from Job openings and recruitment needs, a probability sample survey, and the statistics on occupied post are based on data from Employments, a survey based on administrative data on PAYE tax returns (see also Labour input, number of employees and self-employed persons (europa.eu)). Both surveys are conducted by Statistics Sweden within the national system for official statistics and thus regulated according to the Official Statistics Act (2001:99) and the Official Statistics Ordinance (2001:100).
Coverage
Geographical
Sweden
NACE
All economic activities in sections A-S, according to NACE Rev. 2, and contributing to Swedish production are covered, except for the activities of households as employers and the activities of extraterritorial organisations and bodies.
Enterprise size
At least 1 employee
Definition of the statistical unit
During 2024, two different types of statistical units are used; local unit and legal unit (see 2.5 for more information):
- Local unit: An enterprise or part thereof (e.g. a workshop, factory, warehouse, office, mine of depot) situated in a geographically identified place. At or from this place economic activity is carried out for which – save for certain exceptions – one or more persons work (even if only part-time) for one and the same enterprise.
- Legal unit: A legal person whose existence is recognized by law independently of the individuals or institutions which may own it or are members of it, or a natural person who is engaged in an economic activity in his/her own right.
Remarks
Sampling design
Base used for the sample
Statistics Sweden's business register
Sampling design
Stratified simple random sampling is used, with different stratification principles for the samples used for Q1 and Q2-Q4; for more information, see the section Stratification below. When delimiting the frame, only units that are active according to the information in the business register are included.
For Q1, larger units in terms of number of employees according to the frame (100 employees or more for local units; varying threshold for legal units, depending on stratum) are surveyed every month during the quarter. Remaining units are randomly split, within strata, into three subsamples, where units in subsample i are subject to data collection for reference month i, i=1,2,3, during the quarter.
For Q2-Q4, all units in the sample are randomly split, within strata, into three subsamples, where units in subsample i are subject to data collection for reference month i, i=1,2,3, during the quarter.
Retention/renewal of sampling units
Positive sample coordination over time is implemented using SAMU, a system developed by Statistics Sweden (for more information, see SAMU - The system for co-ordination of frame populations and samples). The sample is updated once a year, before collection of data for the second quarter of the year.
Sample size
Q1: Private sector: 14 500 local units, 4 300 of which are surveyed monthly
Public sector: 1 500 legal units, 700 of which are surveyed monthly
Q2-Q4: 23 450 local units per quarter, all of which are surveyed once per quarter.
Stratification
Q1: The frame is first stratified according to institutional sector. Local units in the private sector are further stratified using information on branch of industry (NACE, 63 categories) and size (number of employees, 5 categories), resulting in 313 strata. Legal units in the public sector are further stratified using information on sub-sector (institutional, 4 categories) and size (number of employees, 5 categories), resulting in 15 strata.
Q2-Q4: The frame is stratified, by means of NACE (sections or groups of sections, 19 categories) and size (number of employees, 6 categories), resulting in 114 strata.
Other sources
Maintenance agency
Q1: The microdata underlying estimates on occupied posts are based on data on number of employees, as derived in the statistical product Short-term employment, a sample survey conducted by Statistics Sweden up until the first quarter of 2024.
Q2-Q4: The microdata underlying estimates on occupied posts are based on data on number of employees, as derived in the statistical product Employments, a survey conducted by Statistics Sweden based on administrative data on PAYE tax returns from the Swedish Tax Agency. (For more information, see Labour input, number of employees and self-employed persons (europa.eu)).
Updating frequency
Q1: Quarterly compilation of collected data.
Q2-Q4: Data on PAYE tax returns are transmitted monthly to Statistics Sweden by the Swedish Tax Agency.
Rules for clearance (of outdated information)
Not applicable
Voluntary/compulsory reporting and sanctions
The data on PAYE tax returns is transmitted following an agreement between the agencies concerned, in line with section 6 of the Official Statistics Ordinance (SFS 2001:100).
Remarks
3.2. Frequency of data collection
Reference dates
Data on job vacancies are collected monthly, with a prespecified Wednesday in the middle of the month as the reference date.
From Q2 and onwards, data on occupied posts are obtained once per quarter, for each of the months in the quarter in question. For occupied posts the data reflect the situation during the month in question, and hence the reference period is month.
3.3. Data collection
Brief description of the data collection method(s)
Remarks
All sample survey data are collected electronically, using a web-based questionnaire. However, as the data on job vacancies come from different surveys for Q1 and Q2-Q4 (see Identification of the source of the data under 3.1), different questionnaires were used for Q1 and Q2-Q4.
From Q2 onwards, the microdata underlying estimates on occupied posts are based on data on number of employees, as derived in the statistical product Employments, and hence obtained electronically in-house.
Units included in the sample survey, are obliged to provide information for this survey according to the Official Statistics Ordinance (SFS 2001:100) and Statistics Sweden's regulations (SCB-FS 2021:22 for Q1 and SCB-FS 2024:10 for Q2-Q4).
3.4. Data validation
To reduce the risk of measurement errors and to facilitate reporting for respondents, a number of logical checks are carried out when the data are submitted. In cases of obvious errors, responses are flagged and must be corrected before the questionnaire can be submitted. In cases of suspected errors, responses are flagged for review and, in some cases, must be commented on before the questionnaire can be submitted.
3.5. Data compilation
Brief description of the weighting method
Weighting dimensions
For a given quarter, target parameters are defined in terms of averages of the corresponding target parameters defined at the monthly level.
For Q1, all statistics are based on a generalized linear regression estimator. Information on number of employees, geographical location and type of economic activity according to Statistics Sweden’s business register are used as auxiliary variables. Moreover, for units included in the part of the sample surveyed every month, imputation is used in case of unit nonresponse (i.e., in case the unit doesn’t provide data for any of the three months). Adjustments for the remaining unit nonresponse are included in the weighting method.
For Q2-Q4, all statistics are based on a nonresponse adjusted version of the Horvitz-Thompson estimator under the sampling design in question.
For more information on nonresponse adjustments, see 6.3.5 below.
See column Brief description of the weighting method.
3.6. Adjustment
Seasonal adjustments are carried out for several time series. For both job vacancy and occupied posts the following NACE sections (or groups of sections) are seasonal adjusted: A, B-E, F, G-I, J, K, L, M-N, O-Q, R-S, and A-S.
For adjustments related to unit nonresponse, see 6.3.5.
4.1. Quality assurance
Not applicable.
4.2. Quality management - assessment
Not applicable.
5.1. Relevance - User Needs
Description of the national users and their main needs
Remarks
Our primary users are the Swedish Public Employment Service, the National Mediation Office, the Ministry of Enterprise, Energy and Communications, the Ministry of Finance and Sweden’s central bank. JVS is also used by universities and independent researchers. We do not have a complete understanding of our users preferences. We provide estimates and annual change on job openings, regional data, size of establishments, and proportions to total number of employees. We also estimate the share of vacancies for which staffing is wanted immediately.
5.2. Relevance - User Satisfaction
Extent to which the needs of national users are satisfied (voluntary)
Remarks
Not available
5.3. Completeness
Description of missing variables and missing breakdowns of the variables
Report progress on the implementation measures regarding quarterly job vacancies statistics of Regulation (EC) No 453/2008, including :
a detailed plan and timetable for completing implementation
a summary of the remaining deviations from EU concepts
None
The JVS produced by Statistics Sweden complies with all compulsory requirement stated by Regulation (EC) No 453/2008. We do not break-down the vacancies by occupation or by permanent position. Statistics Sweden has no plans in the near future to fulfil these optional requirements.
5.3.1. Data completeness - rate
Not applicable.
6.1. Accuracy - overall
The estimation procedure used is based on several assumptions regarding how sources of uncertainty affect accuracy. If the assumptions underlying the chosen estimation method are correct, the estimator is approximately unbiased and an interval of the form statistic ± margin of uncertainty, calculated as 1.96 times the estimated standard error, constitutes an approximate 95 percent confidence interval.
It should be noted that from Q2, although microdata on occupied posts are available at the population level, the statistics are based on data only for the units in response set that provide data on job vacancies. Thus, also the estimates for occupied posts are affected by both sampling error and nonresponse error. However, estimates on number of occupied posts are used in the denominator when estimating the job vacancy rate, and the number of job vacancies and number of occupied posts are likely be positively correlated variables at the unit level. Therefore, the accuracy of the estimator of the rate is likely to benefit from the estimators in the nominator and the denominator being based on the same set of units.
The sources of uncertainty considered to have an impact on accuracy are sampling, frame coverage, measurement, non-response, and model assumptions, with sampling being the source that is considered to have the largest impact on overall uncertainty.
6.2. Sampling error
Standard errors are estimated based on a simplified procedure, under which all monthly subsamples are treated as statistically independent. Moreover, the standard errors are estimated using a procedure intended to also capture most of the random variation caused by other sources of uncertainty, i.e. non-sampling errors. Provided that these sources contribute to the uncertainty as assumed, the method used for estimating the standard errors is likely to be conservative, i.e. on average provide estimates larger than the actual variance.
6.2.1. Sampling error - indicators
Coefficient of variation (taking into account the sampling design) or estimated sampling error for the number of job vacancies (see guidelines).
See annex (coefficient of variation for 2024)
6.3. Non-sampling error
Information on variables with non-negligible measurement and processing errors
Information on main sources of (non-negligible) measurement and processing errors and, if available, on methods applied for correction
Estimation bias: An assessment of the non-sampling errors, in terms of the absolute number of vacant posts, for the total number of job vacancies and, where possible, for aggregation level of NACE Rev. 2 specified in Annex 1 to this Regulation and size classes (1-9, 10 + employees).
Remarks
Overall, the statistics are not deemed to be affected severely by either measurement or processing errors.
However, job vacancy is the variable most likely to be affected by measurement error.
Especially for larger units it might be a challenge to have complete knowledge of all ongoing recruitment processes. This was an issue explicitly taken into consideration when developing the questionnaire that is in use from Q2 and onwards. In developing the questionnaire, both cognitive interviews and pilot testing were used.
Moreover, as indicated in 3.5, logical checks are carried out when the data are submitted, in order to reduce the risk of measurement errors and to facilitate reporting for respondents.
The size of non-sampling error in terms of bias has not been assessed numerically. Nonresponse is the source deemed most likely to impact non-sampling error in terms of bias.
However, there is a strong correlation between the statistics on job vacancies and the statistics on vacant positions released monthly by Arbetsförmedlingen, the official Swedish public employment service, indicating that non-sampling error in terms of bias is not a large problem.
6.3.1. Coverage error
Description of any difference between the reference population and the study population
Description of classification errors
Description of any difference between the reference dates and the reference quarter
Any other relevant information
All local units with at least one employee is surveyed. Local units are very stable and if there is a change of ownership (new corporate registration number) they are still surveyed. If the activity is moved to another local unit, outside the frame, the number of vacancies is still surveyed. These factors make it difficult to measure the impact of coverage-error. The level that Business register at Statistics Sweden operates on, combined with the techniques described above and that we renew our frame and sample once a year, should ensure that coverage error is not a very big issue.
Is very seldom detected.
No difference
6.3.1.1. Over-coverage - rate
Size-weighted over-coverage rates for Q2-Q4 are given below, using number of employees as the size measure:
Q1: 1.9%
Q2: 0.7%,
Q3: 1.3%,
Q4: 2.0%
6.3.1.2. Common units - proportion
The proportion of units covered by both the survey and data from the administrative source in relation to the total number of units in the survey is 1.
6.3.2. Measurement error
See 6.3. non-sampling error.
6.3.3. Non response error
See 6.2, 6.3, 6.3.3.1, and 6.3.5.
6.3.3.1. Unit non-response - rate
Unit response rate
2024 Private sector: Q1: 85.9% Public sector: Q1: 94.0%
Size-weighted unit response rates for Q2-Q4 are given below, using number of employees as the size measure:
Q2: 81.0%, Q3: 81.5%, Q4: 83.5%
6.3.3.2. Item non-response - rate
Due to the construction of the web questionnaire and the logical checks used during the actual data collection, item nonresponse cannot occur.
6.3.4. Processing error
See 6.3. non-sampling error.
6.3.4.1. Imputation - rate
Item imputation rate and methods and, where possible, the effect of imputation on the estimates for the variables transmitted
Non-response is treated by weighting. No imputation is carried out.
6.3.5. Model assumption error
If modelling is used, include a description of the models used. Particular emphasis should be given to models for imputation or grossing-up to correct for unit non-response.
For Q1, for units included in the part of the sample surveyed every month, imputation is used in case of unit nonresponse (i.e., in case the unit doesn’t provide data for any of the three months). The value imputed is the most recent value submitted by the unit in question. For remaining units, nonresponse adjustment is carried out through weighting, assuming that the responding units, per stratum and monthly subsample, constitute a simple random sample from the sampled units.
For Q2-Q4, all unit nonresponse is adjusted through weighting, assuming that the responding units, per stratum and monthly subsample, constitute a simple random sample from the sampled units.
6.4. Seasonal adjustment
Brief description of seasonal adjustment procedures, in particular with regard to the European Statistical System guidelines on seasonal adjustment which have been endorsed and supported by the SPC.
Seasonal adjustment is carried out using the X-13ARIMA-SEATS method, which uses linear filters in combination with time series analysis as the basis for estimating trend-cycle and seasonal components.
As no preliminary statistics are published, no revision policies exist for the surveys underlying the European statistics. For revisions due to unforeseen circumstances, Statistics Sweden’s general policy for revisions applies.
6.6. Data revision - practice
Provide a revision history, including the revisions in the published number of job vacancies and a summary of the reasons for the revisions.
None is applied
6.6.1. Data revision - average size
Not applicable.
7.1. Timeliness
See 7.1.1.
7.1.1. Time lag - first result
Information on the time span between the release of data at national level and the reference period of the data.
The results are published about 55 days after the referenced quarter.
7.1.2. Time lag - final result
Not applicable.
7.2. Punctuality
See 7.2.1.
7.2.1. Punctuality - delivery and publication
Deadlines for the respondents to reply, also covering recalls and follow-ups
Period of the fieldwork
Period of data processing
Dates of publication of first results
Remarks
4 days before our release date.
The fieldwork is conducted 20 days after the start of the quarter until 4 days before our release date.
The process of compiling data ends the day before the release of data.
No release of preliminary data.
Data is released at 08:00 a.m. on the dates announced in our official release calendar. Delays in our releases are rare at Statistics Sweden. JVS have no delays in the last couple of years.
8.1. Comparability - geographical
Information on differences between national and European concepts, and — to the extent possible — their effects on the estimation.
Job openings that does not constitute of newly created positions (such as substitutes being reqcruited for a limited period of time to cover for ordinary staff) is not covered by the european definition of vacancies. The national publication of job vacancy data consists of estimates both with and without these job openings.
8.1.1. Asymmetry for mirror flow statistics - coefficient
Not applicable.
8.2. Comparability - over time
Information on changes in definitions, coverage and methods in any two consecutive quarters, and their effects on the estimation.
Remarks
As indicated in 3.1, from Q2 and onwards the statistics on job vacancies and occupied posts are produced within the scope of a new survey, Job openings and recruitment needs (LOR). The survey design in LOR differs from the design used in its predecessor, Job openings and unmet labour demand (KV), in several ways. Below, the major differences between the two surveys are briefly described:
Changes in sampling and observational unit
In LOR, local unit is used as both sampling and observational unit throughout the economy.
In KV, local unit was used as both sampling and observational unit only in the private sector. In the remainder of the economy, legal unit was used as both sampling and observational unit.
Using local unit throughout allows for more valid estimates and improves the content of the statistics, as it allows for production of more valid statistics at the regional or economic activity level.
Changes in sampling design
In LOR, every sampled unit is in the sample once per quarter, for a specific, randomly, selected month.
In KV, many local units (appr. 4,300 units) and legal units (appr. 700 units) were in the sample three times per quarter, once per month.
Consequently, the sampling design in LOR, in terms of stratification, sample size and sample size allocation, is different from the sampling design in KV.
LOR was designed to substitute not only KV but also two additional surveys used for production of national statistics. In total, taking the changes in design above into account, the respondent burden will decrease by appr. 30 percent.
Changes in estimation methods
In LOR, estimates are based on a nonresponse-adjusted version of the Horvitz-Thompson estimator (HT), whereas in KV, estimates were based on a nonresponse-adjusted version of the generalized regression estimator (GREG).
In LOR, unit nonresponse and extreme, but valid, values are taken into account in a different way than in KV.
Studies based on data from KV indicate that the transition from GREG to HT, which is somewhat easier both to use and communicate, should have a very small, if any, effect on estimates on precision. The changes referred to in the previous bullet point are likely to result in a decrease in bias and an increase in sampling error. Considering the combined effect, an estimator in LOR is expected to have a smaller mean squared error than the corresponding estimator in KV. Thus, LOR is expected to result in more accurate statistics than KV.
Changes in measurement instrument
In LOR, a completely new questionnaire is used for data collection directly from respondents. In comparison, the new questionnaire allows for better separation of job openings corresponding to open positions and job openings to be held by substitutes, resulting in more accurate statistics on job vacancies as defined in Regulation (EC) No 453/2008 of the European Parliament and of the Council.
Changes in sources
In LOR, the microdata underlying estimates on occupied posts are based on data on number of employees, as derived in the statistical product Employments. In Employments, which from 2024Q1 replaced the product Short-term Employment (KS) as the Swedish source on short term business statistics on employment, the variable number of employments is derived at the kind-of-activity unit level, using administrative data on PAYE tax returns (see 3.4 in Labour input, number of employees and self-employed persons (europa.eu)). In LOR, the value for a kind-of-activity unit is allocated to the corresponding local units, using data on number of employees from the business register. In KV, microdata on occupied posts were obtained from KS, which at the sample level was coordinated with KV.
Estimates on number of occupied posts are used in the denominator when estimating the job vacancy rate. The number of job vacancies and number of occupied posts are likely be positively correlated variables at the unit level. Consequently, the accuracy of the estimator of the rate would most likely benefit from the estimators in the nominator and the denominator being based on the same set of units. Therefore, estimates of number of occupied posts are based on data for the response set only, using the same estimation method that is used for job vacancies.
As indicated above, the new statistics are expected to be less biased but more imprecise, with an expected decrease in mean squared error, than before. It is not possible to quantify the combined effect of the changes in the survey design without resorting to methods that rely heavily, or completely, on assumptions. Therefore, no adjustments have been made to previous statistics and caution must be used when making comparisons over time. Thus, it is likely that all series are affected, at least to some extent, by the changes. Moreover, changes to development patterns or seasonal patterns cannot be ruled out.
8.2.1. Length of comparable time series
See 8.2.
8.3. Coherence - cross domain
Comparisons of data on the number of vacant jobs from other relevant sources when available, in total and broken down by NACE at section level when relevant, and reasons if the values differ considerably.
Comments to the attached analysis:
Beveridge curve
The rates initially indicate a positive move along the curve, followed by an unclear trend in 2012-2013. From 2014 we can see a new positive move along the curve. The job vacancy rate started to decrease between 2018-2019, followed by a large drop due to the pandemic in 2020. In 2021-2022 the labour market made a strong recovery, before the job vacancy rate started to decrease again in 2023 and fall even further in 2024 (see annex).
Regarding the comparison of employees from LFS and occupied posts from JVS:
The levels are reasonably well aligned in Sweden, even if especially section O shows a large discrepancy. The reason behind section O's large discrepancy is that we do not present statistics for section O in our national data. Data for section O is only collected to comply with the Eurostat regulation. OBS: Discrepancies between the number of employees and occupied posts should also be affected by the short-term paid layoffs of staff during the pandemic year of 2020 (see annex).
Disclosure rules: Brief description of when data have to be deleted for reasons of confidentiality
Statistics Sweden’s work is regulated by provisions on free access to public records and secrecy in the Public Access to Information and Secrecy Act (SFS 2009:400). Rules of secrecy in the field of statistics is described in chapter 24, §8. The processing of personal data is regulated by the Personal Data Act (1998:204), the Official Statistics Act (2001:99) and Official Statistics Ordinance (2001:100).
A 'job vacancy' is defined as a paid post that is newly created, unoccupied, or about to become vacant:
for which the employer is taking active steps and is prepared to take further steps to find a suitable candidate from outside the enterprise concerned; and
which the employer intends to fill either immediately or within a specific period of time.
‘Active steps to find a suitable candidate’ include:
notifying the job vacancy to the public employment services,
contacting a private employment agency/head hunters,
advertising the vacancy in the media (for example internet, newspapers, magazines),
advertising the vacancy on a public notice board,
approaching, interviewing or selecting possible candidates/potential recruits directly,
approaching employees and/or personal contacts,
using internships.
'Specific period of time’ refers to the maximum time the vacancy is open and intended to be filled. That period shall be unlimited; all vacancies for which active steps are continuing on the reference date shall be reported.
An 'occupied post’ means a paid post within the organisation to which an employee has been assigned.
Q1: Private sector: local unit; public sector (including non-profit institutions serving households): enterprise (legal unit).
Q2-Q4: local unit.
Enterprises with one or more employees.
Sweden.
Not Applicable
The estimation procedure used is based on several assumptions regarding how sources of uncertainty affect accuracy. If the assumptions underlying the chosen estimation method are correct, the estimator is approximately unbiased and an interval of the form statistic ± margin of uncertainty, calculated as 1.96 times the estimated standard error, constitutes an approximate 95 percent confidence interval.
It should be noted that from Q2, although microdata on occupied posts are available at the population level, the statistics are based on data only for the units in response set that provide data on job vacancies. Thus, also the estimates for occupied posts are affected by both sampling error and nonresponse error. However, estimates on number of occupied posts are used in the denominator when estimating the job vacancy rate, and the number of job vacancies and number of occupied posts are likely be positively correlated variables at the unit level. Therefore, the accuracy of the estimator of the rate is likely to benefit from the estimators in the nominator and the denominator being based on the same set of units.
The sources of uncertainty considered to have an impact on accuracy are sampling, frame coverage, measurement, non-response, and model assumptions, with sampling being the source that is considered to have the largest impact on overall uncertainty.
Not Applicable
Brief description of the weighting method
Weighting dimensions
For a given quarter, target parameters are defined in terms of averages of the corresponding target parameters defined at the monthly level.
For Q1, all statistics are based on a generalized linear regression estimator. Information on number of employees, geographical location and type of economic activity according to Statistics Sweden’s business register are used as auxiliary variables. Moreover, for units included in the part of the sample surveyed every month, imputation is used in case of unit nonresponse (i.e., in case the unit doesn’t provide data for any of the three months). Adjustments for the remaining unit nonresponse are included in the weighting method.
For Q2-Q4, all statistics are based on a nonresponse adjusted version of the Horvitz-Thompson estimator under the sampling design in question.
For more information on nonresponse adjustments, see 6.3.5 below.
See column Brief description of the weighting method.
Q2-Q4: The job vacancy statistics are based on data from Job openings and recruitment needs, a probability sample survey, and the statistics on occupied post are based on data from Employments, a survey based on administrative data on PAYE tax returns (see also Labour input, number of employees and self-employed persons (europa.eu)). Both surveys are conducted by Statistics Sweden within the national system for official statistics and thus regulated according to the Official Statistics Act (2001:99) and the Official Statistics Ordinance (2001:100).
Coverage
Geographical
Sweden
NACE
All economic activities in sections A-S, according to NACE Rev. 2, and contributing to Swedish production are covered, except for the activities of households as employers and the activities of extraterritorial organisations and bodies.
Enterprise size
At least 1 employee
Definition of the statistical unit
During 2024, two different types of statistical units are used; local unit and legal unit (see 2.5 for more information):
- Local unit: An enterprise or part thereof (e.g. a workshop, factory, warehouse, office, mine of depot) situated in a geographically identified place. At or from this place economic activity is carried out for which – save for certain exceptions – one or more persons work (even if only part-time) for one and the same enterprise.
- Legal unit: A legal person whose existence is recognized by law independently of the individuals or institutions which may own it or are members of it, or a natural person who is engaged in an economic activity in his/her own right.
Remarks
Sampling design
Base used for the sample
Statistics Sweden's business register
Sampling design
Stratified simple random sampling is used, with different stratification principles for the samples used for Q1 and Q2-Q4; for more information, see the section Stratification below. When delimiting the frame, only units that are active according to the information in the business register are included.
For Q1, larger units in terms of number of employees according to the frame (100 employees or more for local units; varying threshold for legal units, depending on stratum) are surveyed every month during the quarter. Remaining units are randomly split, within strata, into three subsamples, where units in subsample i are subject to data collection for reference month i, i=1,2,3, during the quarter.
For Q2-Q4, all units in the sample are randomly split, within strata, into three subsamples, where units in subsample i are subject to data collection for reference month i, i=1,2,3, during the quarter.
Retention/renewal of sampling units
Positive sample coordination over time is implemented using SAMU, a system developed by Statistics Sweden (for more information, see SAMU - The system for co-ordination of frame populations and samples). The sample is updated once a year, before collection of data for the second quarter of the year.
Sample size
Q1: Private sector: 14 500 local units, 4 300 of which are surveyed monthly
Public sector: 1 500 legal units, 700 of which are surveyed monthly
Q2-Q4: 23 450 local units per quarter, all of which are surveyed once per quarter.
Stratification
Q1: The frame is first stratified according to institutional sector. Local units in the private sector are further stratified using information on branch of industry (NACE, 63 categories) and size (number of employees, 5 categories), resulting in 313 strata. Legal units in the public sector are further stratified using information on sub-sector (institutional, 4 categories) and size (number of employees, 5 categories), resulting in 15 strata.
Q2-Q4: The frame is stratified, by means of NACE (sections or groups of sections, 19 categories) and size (number of employees, 6 categories), resulting in 114 strata.
Other sources
Maintenance agency
Q1: The microdata underlying estimates on occupied posts are based on data on number of employees, as derived in the statistical product Short-term employment, a sample survey conducted by Statistics Sweden up until the first quarter of 2024.
Q2-Q4: The microdata underlying estimates on occupied posts are based on data on number of employees, as derived in the statistical product Employments, a survey conducted by Statistics Sweden based on administrative data on PAYE tax returns from the Swedish Tax Agency. (For more information, see Labour input, number of employees and self-employed persons (europa.eu)).
Updating frequency
Q1: Quarterly compilation of collected data.
Q2-Q4: Data on PAYE tax returns are transmitted monthly to Statistics Sweden by the Swedish Tax Agency.
Rules for clearance (of outdated information)
Not applicable
Voluntary/compulsory reporting and sanctions
The data on PAYE tax returns is transmitted following an agreement between the agencies concerned, in line with section 6 of the Official Statistics Ordinance (SFS 2001:100).
Remarks
Not Applicable
See 7.1.1.
Information on differences between national and European concepts, and — to the extent possible — their effects on the estimation.
Job openings that does not constitute of newly created positions (such as substitutes being reqcruited for a limited period of time to cover for ordinary staff) is not covered by the european definition of vacancies. The national publication of job vacancy data consists of estimates both with and without these job openings.
Information on changes in definitions, coverage and methods in any two consecutive quarters, and their effects on the estimation.
Remarks
As indicated in 3.1, from Q2 and onwards the statistics on job vacancies and occupied posts are produced within the scope of a new survey, Job openings and recruitment needs (LOR). The survey design in LOR differs from the design used in its predecessor, Job openings and unmet labour demand (KV), in several ways. Below, the major differences between the two surveys are briefly described:
Changes in sampling and observational unit
In LOR, local unit is used as both sampling and observational unit throughout the economy.
In KV, local unit was used as both sampling and observational unit only in the private sector. In the remainder of the economy, legal unit was used as both sampling and observational unit.
Using local unit throughout allows for more valid estimates and improves the content of the statistics, as it allows for production of more valid statistics at the regional or economic activity level.
Changes in sampling design
In LOR, every sampled unit is in the sample once per quarter, for a specific, randomly, selected month.
In KV, many local units (appr. 4,300 units) and legal units (appr. 700 units) were in the sample three times per quarter, once per month.
Consequently, the sampling design in LOR, in terms of stratification, sample size and sample size allocation, is different from the sampling design in KV.
LOR was designed to substitute not only KV but also two additional surveys used for production of national statistics. In total, taking the changes in design above into account, the respondent burden will decrease by appr. 30 percent.
Changes in estimation methods
In LOR, estimates are based on a nonresponse-adjusted version of the Horvitz-Thompson estimator (HT), whereas in KV, estimates were based on a nonresponse-adjusted version of the generalized regression estimator (GREG).
In LOR, unit nonresponse and extreme, but valid, values are taken into account in a different way than in KV.
Studies based on data from KV indicate that the transition from GREG to HT, which is somewhat easier both to use and communicate, should have a very small, if any, effect on estimates on precision. The changes referred to in the previous bullet point are likely to result in a decrease in bias and an increase in sampling error. Considering the combined effect, an estimator in LOR is expected to have a smaller mean squared error than the corresponding estimator in KV. Thus, LOR is expected to result in more accurate statistics than KV.
Changes in measurement instrument
In LOR, a completely new questionnaire is used for data collection directly from respondents. In comparison, the new questionnaire allows for better separation of job openings corresponding to open positions and job openings to be held by substitutes, resulting in more accurate statistics on job vacancies as defined in Regulation (EC) No 453/2008 of the European Parliament and of the Council.
Changes in sources
In LOR, the microdata underlying estimates on occupied posts are based on data on number of employees, as derived in the statistical product Employments. In Employments, which from 2024Q1 replaced the product Short-term Employment (KS) as the Swedish source on short term business statistics on employment, the variable number of employments is derived at the kind-of-activity unit level, using administrative data on PAYE tax returns (see 3.4 in Labour input, number of employees and self-employed persons (europa.eu)). In LOR, the value for a kind-of-activity unit is allocated to the corresponding local units, using data on number of employees from the business register. In KV, microdata on occupied posts were obtained from KS, which at the sample level was coordinated with KV.
Estimates on number of occupied posts are used in the denominator when estimating the job vacancy rate. The number of job vacancies and number of occupied posts are likely be positively correlated variables at the unit level. Consequently, the accuracy of the estimator of the rate would most likely benefit from the estimators in the nominator and the denominator being based on the same set of units. Therefore, estimates of number of occupied posts are based on data for the response set only, using the same estimation method that is used for job vacancies.
As indicated above, the new statistics are expected to be less biased but more imprecise, with an expected decrease in mean squared error, than before. It is not possible to quantify the combined effect of the changes in the survey design without resorting to methods that rely heavily, or completely, on assumptions. Therefore, no adjustments have been made to previous statistics and caution must be used when making comparisons over time. Thus, it is likely that all series are affected, at least to some extent, by the changes. Moreover, changes to development patterns or seasonal patterns cannot be ruled out.
