EWCTS 2021 – Weighting

Weighting is applied to survey data in order to improve the accuracy of survey estimates. To ensure that the results of the EWCTS 2021 could be considered representative of workers in the European countries covered, weighting was carried out on the data. It was necessary to apply weighting for two main reasons: first, due to the voluntary nature of the EWCTS 2021, participation may be affected by unequal response rates for different groups of people with the sampling design of Random Digit Dialling (RDD); second, workers with several mobile phones can have a greater likelihood of participating.

The weighting was conducted in two main steps – through calculating design weights and non-response adjustments.

Design weights

In the countries using RDD, because the generated sample is a list of mobile phone numbers rather than people, workers with more than one mobile phone are overrepresented. It is not possible to tell which numbers are affected until the interview, when the respondent is asked how many mobile phone numbers they have.

To account for this factor an adjustment is made, giving a smaller weight to respondents with more phones. Apart from this, within each country all mobile phones have the same likelihood of selection.

Non-response adjustments

A well-known phenomenon in surveys is that certain types of people are easier to contact or more willing to respond. To account for this, non-response adjustments are applied which gives a larger weight to groups which are underrepresented while assigning a smaller weight to overrepresented groups.

Whether a group is overrepresented or underrepresented is determined by comparing the distributions of the EWCTS respondents with the distributions found in Eurostat’s Labour Force Survey employment statistics on age, sex, occupation, sector and region of residence of the respondent. The method used is known as calibration weighting and is applied for each country separately. Numerous checks are made to ensure the quality of the weights.

Final weights

The design weights and non-response adjustments are combined to obtain the final weights. The final weights are constructed as frequency weights and indicate the estimated number of workers represented by each respondent. The weights of all the respondents in a given country add up to the total number of people in employment. They can be used both for country-level analyses or at the European level.

Detailed information on the sampling procedures can be found in the Sampling and weighting report.