Skip to content

Statistical processing

Contact info

Science, Technology and Culture, Business Statistics
Anne-Sofie Dam Bjørkman
+45 3917 3396

asd@dst.dk

Get as PDF

Business Enterprise Research and Development (BERD)

The statistics is based on a survey sample of approx. 3.600 units weighted to a frame of approx 20.000 enterprises. The statistics is compiled in one joined questionnaire which covers both the R&D domain and the innovation statistics. An extensive validation process of the data is carried out. One part of the validations is integrated in the data collection in the dynamic web-questionnaire, another part is carried out after the data collection using micro- and macro validation techniques.

Source data

The statistics are an annual, sample-based questionnaire survey based on information collected from approx. 3600 companies from most size classes and all NACE-industries in the Danish enterprise sector. For 2022, information will be collected from 3,545 companies, sampled as a sample from a framework population of approx. 20.000 companies.

The data collection for R&D and innovation in business is carried out as one complete data collection in one questionnaire. For equal reference years, overall data on R&D expenditure and staff as well as the purchase of R&D services are collected. For the odd reference years, a full R&D survey is collected with detailed information on, for example, sources of funding.

The companies for the sample are selected depending on the number of employees and industry. The committee probability is 100 per cent. among the largest companies (companies with 100+ employees or turnover greater than DKK 1 billion), and the committee probability is decreasing as the number of employees decreases. To ensure the overall quality of the statistics, R & D-intensive industries have higher coverage than smaller R & D-intensive industries. The companies in the sample are randomly selected.

Based on a gross population retrieved from the Business Statistics Register, ESR, a survey population (framework population) is formed, from which companies in specific industries or with very few employees (depending on industry) are sorted. The survey population is the population that is later counted up to and thus the population that the survey describes. In 2021, the survey population consists of approx. 20.000 companies.

Frequency of data collection

Yearly. The number of variables varies between even years (overall variables only) and odd years (overall variables supplemented with detailed information).early.

Data collection

It is mandatory to reply to the statistics using the web-based questionnaire from Virk

Data validation

An extensive validation process of the data is carried out. One part of the validations is integrated in the data collection in the dynamic web-questionnaire, another part is carried out after the data collection using micro- and macro validation techniques. The individual reports from the enterprises are compared to former years reports and the registered information on number of employees and turnover. Outlier detection is also used as a validation process.

Data compilation

A stratified random sampling is used on the basis of the activity of the enterprise and the number of employees. By grossing up a weighting and calibration using regression techniques is applied to the weight of the individual enterprise.

Non-response from enterprises of a certain size are imputed with "last known data" (cold deck imputation). If data from former surveys is not available for the specific record then nearest neighbor (hot deck, donor imputation) is used. Other enterprises that have not answered the questionnaire (unit non-response) are handled during the enumeration.

Adjustment

Not relevant for these statistics.