Vacancies, online observation (occupational classification)
What does the survey comprise?
Purpose
The publication of job vacancy figures in the Netherlands, broken down by occupational class and occupational segment, in accordance with the Dutch Standard Classification of Occupations ROA-CBS 2014 (BRC 2014).
Target population
All newly created online vacancies of companies and institutions with personnel in the Netherlands.
Statistical unit
New vacancy.
Start of the survey
Data on vacancies by occupational classification are available from 2018.
Frequency
Annual.
Publication strategy
The results are published approximately five months after the end of the reporting year. The figures are provisional and may be adjusted as a result of an improved weighting method.
How is the survey conducted?
Type of survey
The survey combines online job vacancy data with data from sample surveys and registers. The online vacancies are weighted according to the total number of vacancies in the Quarterly Survey on Vacancies.
Observation method
Online vacancies are collected via web scraping. This involves the systematic search for new vacancies on websites. Subsequently, the full texts of the vacancies are saved and duplicate publications of identical vacancies are removed. The vacancies are classified by occupation and company, amongst others.
Respondents
The online vacancies are collected by Textkernel.
Sample size
All online vacancies in the Netherlands.
Error detection and correction methods
CBS adjusts the SIC (Standard Industrial Classification) code of the online vacancies on the basis of the General Business Register (ABR). The Chamber of Commerce number of the company identified by Textkernel will be used as the statistical linkage key. Links will be checked for consistency by way of a top-down approach.
Weighting
The number of online vacancies does not fully correspond to the total population of new vacancies. This can be explained by the fact that not all vacancies are published or observed; it is also possible that a vacancy is published on multiple websites.
Therefore, the number of online vacancies is weighted so that the number per industry matches the number from the Quarterly Survey on Vacancies (not seasonally adjusted). The weighting is based on a method that was developed by the Employee Insurance Agency (UWV). Each online vacancy is assigned a weight to enable the best possible adjustments for:
- Vacancies at different occupational levels. For basic occupational levels, there is less online recruitment and more use of other channels. Lower occupational levels are therefore given a higher weight than higher occupational levels; this is deduced from the share of employed persons by occupational level according to the Labour Market Survey (LFS);
- Indirect vacancies through intermediaries. Intermediaries may be the only source for a vacancy, but may also cause more duplicate vacancies. If both direct and indirect online recruitment is used for a vacancy, it is not always clear that it concerns the same vacancy and, consequently, deduplication is not always applied. Therefore, the indirect vacancies are given a smaller weight (25%) than the direct vacancies;
- Direct vacancies in an unfamiliar industry. In this way, they are counted towards the correct occupational group in the BRC 2014;
- Matching of the new vacancies by industry with the numbers from the Quarterly Survey on Vacancies for each quarter.
What is the quality of the results?
Accuracy
The results of the survey are subject to a margin of error. Values are therefore rounded to avoid the suggestion of great accuracy.
Three causes of inaccuracy have been identified, from which three conditions for publication have been derived:
- The outcome should not depend too greatly on variation due to incompleteness of online vacancies. To this end, the online vacancy data have been compiled randomly for 50 thousand times using a bootstrap analysis. Condition (1): 90% of the bootstrap results of any value must lie within an interval that deviates less than 10% from the calculated value.
- The outcome should not deviate too much due to variation in the occupational classifications. For this purpose, Textkernel used a correction table from previous research by UWV into the occupational classification. Condition (2): Reclassification results in no more than 10% deviation from the calculated value.
- The outcome should not depend too much on the chosen weight of 25% of the vacancies through intermediaries. Condition (3): The difference between the outcomes of the weighting of vacancies through intermediaries with both a 10% and a 70% weight does not exceed 20% of the calculated value.
Only if the value of a cell meets all three conditions in the years 2018, 2019 and 2020 are the corresponding figures published.
Excluded from condition (2) are the occupations stated in occupational class 06 of the BRC 2014 (Public administration, security and legal occupations). For this occupational class, the condition derived from the UWV survey is considered too strict. See also the publication Classifying occupations with data in online vacancies (Dutch only).
Sequential comparability
The figures are sequentially comparable.
Description of quality strategy
The numbers of new vacancies per company or institution will be checked for plausibility. Incorrect links of vacancies to companies or institutions are corrected if they were to lead to classification in a different industry by mistake.