Estimates for 2019 and 2020 are based on provisional versions of SILC-19 and SILC-20 data that have not yet been released.
As the profile of the respondents is not the same as that of non-respondents, weighting plays an important part in allowing an extrapolation to the reference population.
The weighting used for SILC-19 is the final standard weighting, whereas that used for SILC-20 (before or during semi-lockdown) is an experimental weighting.
The standard SILC cross-sectional weighting is usually produced roughly 10 months after the end of the field survey. This is because integrity checks, consolidation and the inclusion of register data (available from March 2021 for the 2020 survey) are needed to produce this weighting. However, the tasks of quality control and inclusion of register data are hardly relevant to the subjective variables examined here.
A provisional and experimental weighting was calculated in order to release information as soon as possible on the impact of the Covid-19 crisis on well-being in Switzerland during the first half of 2020. This experimental weighting includes a uniform correction for total non-response, without taking into account the specificities of the profiles. This uniform correction for total non-response was due to a lack of certain auxiliary variables (of profiles) used in the standard weighting to correct for non-response.
In order to make the best possible correction for the potential bias due to total non-response, a final calibration was applied for this provisional and experimental weighting. The final calibration was applied to the same margins as the final calibration of the standard weighting, the only difference being that we used auxiliary information from December 2018 instead of December 2019 (not available before March 2021). For this reason, the experimental weighting used for Covid-19 purposes represents the population at the end of 2018 instead of at the end of 2019, as will be the case for the standard weighting for SILC-20.
The variables used for the final calibration are all taken from registers and are as follows:
- Civil status (source SRPH)
- Sex (source SRPH)
- Nationality group- 4 groups (source SRPH)
- Type of family (source SRPH)
- Household size ( SRPH)
- Major region (source SRPH)
- At risk of poverty status at 60% of median total equivalent household income (source CCO)
- Indicative of total equivalent household income < P10 (source CCO)
- Indicative of total equivalent household income < P50 (source CCO)
- Indicative of total equivalent household income < P20 (source CCO)
- Indicative of total equivalent household income > P80 (source CCO)
The latter partly overlap with the auxiliary variables used for correction for non-response in the standard weighting. The main aspects not covered are as follows:
Variables taken from registers:
- Household composition by nationality
- Type of family and number of children
- Household composition by sex
- Commune typology - 9 groups
- Nationality group - 2 categories
- Size of household’s commune
- Presence of supplementary benefits in household
- Number of unemployment allowances in household
- Number of disability pensions in household
- Number of old-age pensions in household
- Number of incomes from employment in household
- Moved house (change of building) in past 2 years
- Living space per household member
Variables drawn from SILC survey in wave 1 and same values used from wave 2 onwards:
- Employment status in 4 categories
- At risk of poverty status at 60% of median total equivalent household income
- Material deprivation 3 out of 9 items
- Maximum household level of education
- Interest in politics
- Rent and accommodation costs
In order to correct comparability issues arising from the fact that the profiles of respondents at the start of the survey (more willing to cooperate with this statistic) are different from those of respondents at the end of the survey, we processed separately the net sample of respondents before 16 March and that of respondents between 16 March and 20 June (semi-lockdown in Switzerland). This means that the two sub-samples each correspond to the population of December 2018.
The distribution of weights in the experimental weighting of SILC-20-Covid-19 is close to that of the standard SILC-19 weighting.