Accuracy and reliability

Overall accuracy

For the private sector: The overall accuracy of the statistics is high, as the delimitation of the population is based on Statistics Denmark’s Statistical Business Register (SBR) and the industry codes stated therein. Furthermore, the statistics are mandatory and the reminder procedure is intensive, which is expected to result in a response rate of more than 98 percent.

For workplaces with more than 100 employees that have not reported, data are imputed. The imputation rate is normally below 2.5 percent.

The number of job vacancies typically declines by between 0.5 and 2.0 percent, while the job vacancy rate is most often stable. The job vacancy rate has never declined by more than 0.1 percentage points before the data obtain final status in the same quarter one year after the first publication.

For the public sector: Since public workplaces are legally obliged to post job vacancies on https://jobnet.dk, the administrative data source is assumed to be close to full coverage for the public labour market. However, there will be workplaces that do not post vacancies on https://www.jobnet.dk, even though this is legally required. Methodological decisions have been made based on assumptions in the models for handling intended job vacancies and handling presumed misreporting, and calibrated on the basis of the test survey in May 2024 financed by EUROSTAT. The number of occupied posts is measured as the number of occupied posts at the end of the quarter, obtained from SBR data, and therefore not on the counting date.

Sampling error

For the private sector: As in other EU Member States, the coefficient of variation (CV), which is the standard deviation in relation to the estimate, is used in calculating the uncertainty.

For the total number of occupied posts, the coefficient of variation (CV) was 0.7 percent, while for the total number of job vacancies it was 2.7 percent. The coefficient of variation for the different industry and size groups is relatively high. The fact that the latter is relatively high is not due to the grossing-up or the design and size of the sample. It is instead due to the large variations between the reported number of job vacancies and the many reports of zero job vacancies in each stratum.

For job vacancies, the coefficient of variation (CV) for the 1st quarter of 2026 was:

• Manufacturing, mining and quarrying and utility services (B+C+D+E): 4.1 percent.
• Construction (F): 10.8 percent.
• Trade and transport etc. (G+H+I): 4.9 percent.
• IT and media (JK): 8.7 percent.
• Financial and insurance (L): 9.6 percent.
• Trade, rental and administration of real estate (M):: 15.5 percent.
• Consulting and business services (N+O): 6.9 percent.
• Public administration, education and health (P+Q+R): 9.6 percent.
• Culture, leisure, auto repair and other service activities (S+TUV): 12.2 percent.

For the public sector: Not relevant.

Non-sampling error

For the private sector: In the Job Vacancy Statistics, the sampling frame is the Statistical Business Register, which covers all registered businesses in Denmark. The register is continuously updated on a daily basis and is expected to have full coverage.

Every quarter, a new sample is drawn from the Statistical Business Register, and at that moment there are very few differences between the register and the sample. As time passes, the register is updated; some workplaces are no longer statistically active and new units emerge. Before grossing-up, the sample is updated with information on units that are no longer included in the sample. Unweighted, over-coverage is below 1 percent.

Uncertainty caused by non-response, incorrectly reported data and misunderstandings is sought to be minimised through repeated reminders in the case of non-reporting and through thorough checks of the submitted figures.

Statistics Denmark assumes that the number of job vacancies reported by the workplaces is, to some degree, a variable with non-negligible measurement error. The main source of these non-negligible measurement errors is uncertainty among respondents about how to measure a job vacancy and which unit data must be compiled for (workplace or the entire enterprise). In addition, we have "fictitious workplaces", i.e. units to which employees without a physical workplace are assigned, such as travelling salesmen and sailors. Reporting for these fictitious workplaces can sometimes be difficult.

Statistics Denmark is aware of the problem with these measurement errors. Most of the errors are captured in micro-validation, where reported data with errors are corrected.

It is mandatory to report data electronically via the public platform Virk.dk. It is not possible to submit a questionnaire with blank fields. Item non-response is therefore zero.

The response rate is high because enterprises are legally obliged to report data to the Job Vacancy Statistics. Unweighted, the unit non-response rate is below 5 percent.

The statistics are published quarterly, but data are collected monthly. The sample is divided into three groups, where the 1st group reports data concerning the 1st month of the quarter, the 2nd group reports data concerning the 2nd month of the quarter, and the 3rd group reports data for the 3rd month of the quarter. The variation among the workplaces in each individual stratum is equally distributed, and each individual workplace is only selected once in the quarter. The aim is to cover job vacancies throughout the quarter and avoid fluctuations due to, for example, holiday periods.

The Job Vacancy Statistics only imputes data for missing reports for workplaces with more than 100 employees, and the imputation is carried out immediately before grossing-up.

The method used is ratio imputation from the program Banff. As a first choice, the imputation is done in relation to workplaces within the same industry and, as a second choice, workplaces within the same stratum. We estimate the number of job vacancies (and occupied posts) by using the number of registered (CVR) employees at the workplace as auxiliary information. This is done using a generalised regression method.

The grossing-up itself is carried out using the program CLAN with model-assisted estimation.

For the public sector: Since public workplaces are legally obliged to post job vacancies on Jobnet.dk, the administrative data source is assumed to be very close to job vacancies for the full population in the public sector. However, there will be workplaces that do not post vacancies on Jobnet.dk, even though this is legally required. Methodological decisions have been made based on assumptions in the models for handling intended job vacancies and handling presumed misreporting, and calibrated on the basis of the test survey in May 2024 financed by EUROSTAT. The number of occupied posts is measured as the number of occupied posts at the end of the quarter, obtained from SBR data, and therefore not on the counting date.

Quality management

Statistics Denmark follows the recommendations on organisation and management of quality given in the Code of Practice for European Statistics (CoP) and the implementation guidelines given in the Quality Assurance Framework of the European Statistical System (QAF). A Working Group on Quality and a central quality assurance function have been established to continuously carry through control of products and processes.

Quality assurance

Statistics Denmark follows the principles in the Code of Practice for European Statistics (CoP) and uses the Quality Assurance Framework of the European Statistical System (QAF) for the implementation of the principles. This involves continuous decentralized and central control of products and processes based on documentation following international standards. The central quality assurance function reports to the Working Group on Quality. Reports include suggestions for improvement that are assessed, decided and subsequently implemented.

Quality assessment

For the private sector:

As with all other sample-based statistics, there is some uncertainty associated with the estimates. Information on sampling error is used, among other things, to assess the extent to which changes in, for example, the number of job vacancies reflect uncertainty, or whether they represent real decreases or increases. Sampling error is affected by the size of the sample. As a rule, the uncertainty is halved when the sample size is quadrupled.

As in other EU Member States, the coefficient of variation (CV), which is the standard deviation in relation to the estimate, is used in calculating the uncertainty.

Every quarter, a sample of around 9,000 workplaces (before 2026 the sample was approximately 7,000) with at least one employee at the time of selection is drawn. All workplaces with more than 100 employees are selected for the sample. The remaining part of the sample is based on random selection. The sample is stratified by 19 industries and 6 size groups. When the sample is drawn, the workplaces are placed in the 6 size groups on the basis of the number of employees registered for the workplace in Statistics Denmark’s Statistical Business Register (SBR).

In connection with publication, the workplaces are reclassified so that they are placed in the size groups that correspond to their reported number of employees. This reclassification is due to the fact that the register information is almost one year old. Based on new employment figures in Statistics Denmark’s Statistical Business Register (SBR), the sample is optimised every quarter with regard to movements, so that smaller workplaces that have gained more than 100 employees are compulsorily selected, and large workplaces that have fallen below 100 employees are randomly selected.

Workplaces with more than 100 employees are compulsorily selected and therefore always participate in the sample in accordance with Statistics Denmark's Data Supplier Policy. For workplaces with fewer than 100 employees, 1/12 of the sample is replaced every quarter. This ensures that the sample size is maintained at the same level and that the sample provides an up-to-date picture of the number of occupied posts and job vacancies. This means that a workplace with fewer than 100 employees participates for at most 12 consecutive quarters and is then exempted for a longer period.

The selected businesses are obliged to report data to the Job Vacancy Statistics. Failure to report data will result in a police notification, cf. the Act on Statistics Denmark. The response rate is normally above 95 percent. Incoming reports are continuously checked for errors. To ensure that data are reported for the individual workplace, the reported number of employees is validated against register information on the number of employees from Statistics Denmark’s Statistical Business Register (SBR). If the reported data deviate substantially from SBR, the workplace is contacted to have the report verified.

The number of job vacancies is also checked in relation to the reported number of employees. If the number of job vacancies deviates substantially from what is expected for a workplace of the stated size, the workplace is contacted. Overall, approximately 10 percent of the workplaces are contacted in connection with data editing. It must be assumed that not all reporting errors are detected during data editing. For this reason, the statistics may be subject to uncertainty.

For the compulsorily selected workplaces that have not reported or have been exempted from reporting, imputation is carried out, where possible, on the basis of the report from the same CVR number in the same period. We make efforts to ensure that all compulsorily selected workplaces have reported data, so that imputation is only relevant for workplaces that have been exempted from reporting for one reason or another. The grossing-up itself is carried out using the program CLAN.

For the total number of occupied posts, the coefficient of variation (CV) is below 1 percent, while for the total number of job vacancies it is 2-5 percent. The coefficient of variation for the different industry groups and size groups is relatively high. This is not due to the grossing-up or the design and size of the sample, but is primarily due to the large variations between the reported number of job vacancies and the many reports of zero job vacancies in each stratum.

When a new quarter is published, data are revised one rolling year back in time. The reason for this is delays in reporting data from workplaces as well as continued error-checking of reported data.

The number of job vacancies typically declines by between 0.5 and 2.0 percent, while the job vacancy rate is most often stable and has never declined by more than 0.1 percentage points before the data obtain final status.

For the public sector:

With the use of register data, it is possible to provide data material for approximately one third of the entire labour market and the vast majority of the new industry codes. This means that it is possible to significantly limit the increase in resources for both Statistics Denmark and respondents, even though the population is expanded.

Register data cover the entire population and therefore do not involve the sampling errors associated with survey data, such as statistical uncertainty and possible selective non-response. In addition, the questions in the survey may be understood differently by different respondents, whereas counts of advertised vacancies are linked to workplaces' actual behaviour. However, the measure of job vacancies also requires the inclusion of positions that the employer is prepared to take steps to fill. This is a measure based on the respondent's estimate. The estimate is not captured by this method, but the uniform treatment in the model means that there is greater consistency across workplaces and over time.

In the data from https://www.jobnet.dk, we have far more information; among other things, it is possible to follow developments for specific workplaces and CVR units. It is therefore easier to check and correct any outliers or issues in these data than in survey data, where Statistics Denmark cannot, in the same way, filter out all advertisements from a workplace over a period.

There may be low under-coverage in the advertised vacancies, as there are probably public workplaces that do not post vacancies on Jobnet.dk even though they are legally obliged to do so. However, this is assumed to be marginal, based on research of advertised vacancies on various websites. In addition, the measure of occupied posts is based on the most recent count in SBR and therefore not on the counting date, as in the survey for the private sector.

The problems with presumed misreporting have been calibrated using the analysis of a comprehensive test survey financed by EUROSTAT and conducted in May 2024. The model established makes it possible to use the full data material instead of grossing-up on the basis of the presumed misreporting. This is a major advantage, as we have great confidence in the number of advertised job vacancies at CVR level, since the respondent has to post vacancies and must log in using a CVR number.

It has become possible to receive data directly from Copenhagen Municipality, which improves both the general data quality and the quality of the distribution key used to remedy other CVR units with similar problems.

In the long term, two possible improvements in quality could be: 1) establishing similar agreements with other public CVR units, and 2) succeeding in improving data quality by helping to make public workplaces more aware that they must post vacancies on https://www.jobnet.dk.

Data revision - policy

Statistics Denmark revises published figures in accordance with the Revision Policy for Statistics Denmark. The common procedures and principles of the Revision Policy are for some statistics supplemented by a specific revision practice.

Data revision practice

Every time we publish a new quarter we also publish revised data from the previous four quarters. The main reason for this is that we receive data from business units which do not report data punctually. As a consequence, on-going error correction of reported data takes place.

The number of job vacancies tends to decline between 0.5-2,0 per cent while the job vacancy rates often are stable and never have declined more than 0.1 percentage points before the data obtain the status of final.