Title Dynamic Data Editing in online data collection for the Vacant Positions Survey
Author Stax, H.
Workshop Workshop 2011
Year 2011

Slides (ppt)


At the workshop I will present objectives, procedures and results from a pilot-study.


Conventionally, data editing is initiated subsequent to data collection. The preliminary object of post data collection data editing, is to identify possibly defective data. The object proper is to correct errors and improve data quality. Correction of possible defective data may involve costly manual record checks, burdening re-contact with individual businesses/respondents, and possible bias from automatic imputation and statistical editing. In order to minimise costs and burden, the cut off level for possibly defective input data, which is actually edited, may be set low, resulting in low quality output data.

Pilot survey

Data collection for statistics on Vacant  Positions for ESS is fairly non-complex and straight forward - compared to most business statistics at Statistics Denmark. In the questionnaire two questions are asked: Number of vacant positions, and – as a reference variable for data editing - number of persons currently employed. Previously, data editing for this survey was initiated post data collection: For each work unit in the sample, the reported number of vacant positions was compared to the reported number of persons employed. The reported number of employees was also compared to the number of employees recorded in the survey 12 months earlier, and to the number of employees currently recorded for the work unit in the Danish Central Business Register. This was done in order to check, whether the responding businesses had reported vacant positions for the intended work unit (and e.g. not for the whole enterprise). Correction of possible defective data involved manual record checks, re-contact to respondents by phone and statistical imputation and manipulation.

Advanced dynamic online data editing

At Statistics Denmark, Survey and Methods the vacant positions survey was selected as a pilot for “advanced dynamic online data editing”. An online questionnaire was developed, where some of the “old” post data collection edit checks for cross reference were incorporated and run dynamically - as the respondent enters data in the form. The questionnaire for each individual work unit was pre-loaded with: Number of employees recorded in the survey 12 months earlier and number of employees currently recorded in Danish Central Business Register. These values were not actually visible to the respondent in the form, but only used as reference data. If data is entered in the form, which does not comply with the “old” post data collection edit checks, then a warning is generated, and the respondent is asked to re-consider and either modify the response or enter a comment/explanation.